14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions, Proceedings

1 Braunschw eig, Germany 14 th International Symposium on Hazards, Pre v ention and Mitigation of Industrial Explosions Braunschweig, Germany | July 11 – 15, 2022 Proceedings Physikalisch -Technische Bundesan stal t Otto von Gueric ke Universität Magdeburg Proceedings of the 14 th International Symposiu m on Hazards, Prevention and Mitigation of Industrial Expl osions DOI: 10.7795/81 0.20221124 Citation Templat e: Author 1, Author 2, ... : "Title of articl e", pp. n1 - n2 . I n: "Proceedings of th e 14 th International Symposium on Hazards, Prevention and Mitigati on of Ind ustrial Explosions (ISHP MIE 2022)", Braunschweig, G ermany, 202 2. DOI: 10.7795/81 0.20221154 Herausgeber: Physikalisch-Techn ische Bundesanstalt Bundesallee 100 38116 Braunsch weig, Germany Dr. Michael Bey er Dr. Arnas Lucassen 3.7 | Fundamentals of Explosion Prot ection phone: +49 531 592- 3700 e-mail: michael.be [email protected] ishpmie2022.ptb.de [email protected] Published under CC-B Y-ND 4.0 DOI: 10.7795/81 0.20221124 Contents Message from the L ocal Organizing Committee (Michael Be yer) 9 Message from the Pro gram Committee (Holger Großhans) 10 List of Symposium Com mittees 11 Plenary and Review ( Holger G roßhans ) Hybrid mix ture explosions – A brief review Peng Zhao, Dejian Wu, Stefan H. Spitzer, Arne Krietsch, Pau l Amyotte, Ulrich Kraus e (#8 ) 12 Session 01: Advances in explosi on protection: Strategies, measures, and protective equipment I (Dieter Gabel) Flamep roof en closu re and p ressure relief: Status quo Sabrina Herbst, Johanna Gerl ach; Leo Siegle, Fran k Eng el mann, Tho rsten Arnhold (#8) 55 Influence of vent distribution on the violence of a gas explosion Jérôme Daubech, Emmanuel Lepret te, Christophe Prou st (#66) 65 Gas phase explosions in oxygen enriche d atmospheres Simon Egan (#69 ) 79 Session 02: Explosion modelling and simulation I (Josué Melguizo-Gavilanes) Investigation on flame propa gation and pa rticle decompo sition behavior of dust explosion in MIKE 3 apparatus Yangyue Pan, Christoph Spijker, Harald Raupenstrauch (#19) 89 TGA-FTIR for kinetic and evolved gas analysis of the co al particles in dust deflagrati on Yangyue Pan, Christoph Spijker, Harald Raupenstrauch (#20) 98 A Numerical Model for the Minimu m Ignition Temperature of Dust Clouds Tengfei Chen, Jo van Caneghem, Jan Degrève, Jan Berghmans, Filip Verplaetsen, Maarten Vanierschot (#53) 115 Session 03: Explosion testing I (Regis Bauwens) Acceleration sensitivity of pi ezoe lectric pressure sensor s and the influence on the measurement of explosion pressures Tim Krause, Jens Brunzendorf, Detlev Mark us, Harun Kanbur, Niels Springer, Otto Walch, Christian Herr (#29) 128 A compara tive study between two ignition sources: electric igniter versus pyrotechnic igniter Stephane Bernard, Chayma El Gadha, Mame William Louis (#30) 148 Measuremen t of dust flame propagati on and temperatur e wi th lo w wall influence Christoph Spijker, Stefan Puttinger, Simon Schneiderbauer, Stefan Pirker, Georg Meyer, Christoph Buchner, Tino Lindner-Silwes ter (#37) 158 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 4 Session 04: Hydrogen safety I (Chunkan Yu) Large-scale tests to investigate the consequences of exposing cryogenic storage vessels containing liquid hydrogen to a fire lo ad Martin Kluge, Abdel Karim Habib, Kees van Wingerden (#49 ) 168 Experimen tal investigation i nto the consequences o f rel ease of liquified hydrogen onto and under water Martin Kluge, Abdel Karim Habib, Kees van Wingerden (#50 ) 182 Minimum diameters for CH4/H2-air mix tures: implications for natural gas cooktop burners Paola Cristian Mejía-Botero, Josué Melguizo -Gavilanes, Fer nando Veiga-López (#57 ) 197 Session 05: Explosion p revent ion I (Ritsu Dobashi) Experimen tal study of humidity influence on triboelectri c cha rging of particle-laden duct flows Holger Grosshans, Wenchao Xu (#1 5) 210 Assessment o f explosion ris ks ca used by cone discharg es when filling silos and containers with pellets Martin Glor, Ute Hesener (#46) 220 Review on CFD modeling of elec trostatic powder charging during pneumatic conveyi ng Holger Grosshans, Si mon Jant ac (#84) 231 Session 06: Dust explosions I (Maria Po rtarapillo) Classification of Dispersibili ty for Combustible Dust based on Hausner Ratio Yajie Bu, Albert Addo, Paul Am yotte , Yuan Chunmiao (#11) 244 Moderation of Fe dust explosion by nano-sized Fe2O3 and Fe3 O4 powders Yongzheng Guo, Kaiyue Ren, Peng Zhao, Weixing Huang, Aizhu Wei, Dejian Wu (#60) 256 Dust cloud behaviou r in the modified Hartmann tube Enrico Danzi, Olivier Dufaud, Faus to Franchini, Luca Marmo, Matteo Pietraccini (#88) 268 Session 07: Explosion modelling and simulation II (Holger Großhans) A CFD Based Methodology to Design an Explosion Prevention System For Li-Ion Based Batte ry Energy Storage Sys tem Anil Kapahi Kapahi, Sunil Lakshmipathy, Stefan Kraft, Jens Conzen, Alberto Alvarez-Rodriguez (#71) 281 Modelling Methodology for Deflagration Vent Design of Battery Energy Storage Systems Sunil Lakshmipathy, Anil Kapahi Kapahi , Jerome Ta vea u, St efan Kraft, Jens Conzen (#73) 299 Probabilis tic Risk Assessme nt for domino effect due to Vapor Cloud Explosions usin g an i ntegrated Petri-Bayesian netwo rk appr oach Julio Ariel Dueñas San tana, Orelvis González Gómez, Jesus Luis Orozco (#93) 311 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 5 Session 08: Explosion p revent ion I I (Hannes Kern) Inert gas influence on limiting expe rimental safe gap of fuel-air mixtures at various in itial p ressures Maria Mitu, Sabine Zakel, Thomas Stolz (#9) 322 The capabilities of flame a rr esters on ex plosive mixtures with increased inert gas content Frank Stolpe, Samah Bendad a (#4 3) 332 Session 09: Explosion-p rote cted devices I (Jérôme Taveau) Thermal evaluation of junction and connection boxes in explo sion protection Florian Koch, Detlev Markus, Ulric h Krause, Peter Thurnherr (#7) 340 Analysis of flanged joints for flameproo f enclosed devic es of gas group IIC according to the standard IEC 60079-1 Victor H ugo Padron Herrera (# 24) 348 An experimental and a numerical study of a vented dust explosion i n an 11.5 m3 vessel Chen Huang Huang (#38) 360 Session 10: Explosion modelling and simulation III (Lo ren z Boeck) Dust Particle Sedimenta tion in the 20 L Standard Vessel f or Dust Explosion Tests Maria Porta rapillo, Almerinda Di Bene detto, Marco Trofa, Ro berto Sanchirico (#25) 371 Improved modelling of hydrogen explo sions – representing effects of varying concentration and reduced oxygen atmospheres Melodía Lucas, Helene Hi sken, Trygve Skjold, Bjørn Arntzen (#64) 381 Characterizing the Rea ctivity of Large-Scale Explosions Using a Dimensionless Two- Parame ter Combustion Model Regis Bauwens, Lorenz Boec k, Sergey Dorofeev (#70) 393 Session 11: Hybrid mixtu re ex plosions I (Sabine Zakel) The role of vapor fraction in hydroca rbon mist explosion Stéphanie El-Zahlanieh, Amelie Jean, Alexis Vignes, Olivier D ufau d (#68) 404 Where one plus one equals three: the MIT of hybrid mixt ures Paul Geoerg, Stefan Spitzer, Dieter Gabel, Ulrich Krause (#17) 417 Session 13: Dust Explosions II (Chris Cloney) Dust Explosions in Vessel -Pipe Systems at Large Scale Lorenz Boeck, Regis Bauwens, Sergey Dorofeev (#16) 429 Thermal Structure of Aluminum Dust Explosion with Additional Carbon Dioxide Pojul Chang, To shio Mogi, Rit su Dobashi (#26) 440 Ageing Effect on Ignition Sensitivity of Lignocellulosic Dusts Maria Porta rapillo, Almerinda Di Bene detto, Roberto Sanchir ico, Enri co Da nzi, Luca Marmo (#59) 452 Recycling of plastics: dust explos ion risk evaluation Enrico Danzi, Luca Marmo, Maria Portarapillo, Roberto Sanchirico, Almerinda Di Benedetto (#61) 465 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 6 Session 14: Hydrogen Safety III (Stefan Essmann) Schlieren imaging investigations of hot gas ke rnel expan sion caus ed by slo w c ontact -break discharges Carsten Uber, Steffen Franke, Michael Hilbert, Di rk Uhrland, Niklas Schüler (#3 ) 477 Investigating the influence of additi onal inductivity in electrical circuits during ignition caused by contact break discharges Bogdan Barbu, Carsten Uber, Frank Berger, Michael Hilbert ( #4) 488 Investigation of the safe ty re lated ignition processes of laminar strained premixed NH3-H2-air flames Chunkan Yu, Detlev Markus, Robert Schießl, Ulrich Maas, Bo Shu, Sven Eckart, Harmut Krause, Stefan Essmann, Agustin Valera-Medina (#67) 498 Experimen tal and numerical modelling of igniting hyd rog en/air mixtures in inter-connected vessels Bisha m McCarthy-Singh, Tim Krause, Detlev Markus , Holger Grosshans, Manideep Manc hi katla, Alberto Gambaruto (#74) 508 Session 15: Flame propagati on and acceleration I (Arnas Lucassen) Research on the flame stability of b iodiesel/diesel blends in small-s cale combustion based on data fusion Gongping Mao, Huan Mao, Zh ijian Xu , Chunkan Yu (#21) 525 Influence of the flamma ble cloud geometry on the gas explosion effects Jérôme Daubech, Jérô me Héb rard, Emmanuel Lep rette (#65) 537 Session 16: Explosion p revent ion I II (Paul Amyotte) Explosive properties of se le cted aerosols dete rmined in the spherica l 5 l test chamber Adrian Toman, Wojciech Adamus (#45) 549 Holi powder potential dust explosions Blanca Castells, Isabel Ame z, Javier Garcia -Torrent, David León (#33) 560 Session 17: Dust Explosions III (Dejian Wu ) Investigation of Marginally Explosi ble Dusts Albert Addo, Mar ia Portarapillo, Almerinda Di Benedetto, Yajie Bu, Paul Amyotte, Yuan Chu nmiao, Ashok Dastidar, Faisal Khan (#10) 575 Turbulence generated by dust dispersion in the standard 1 m3 vessel Zdzisła w D yduch (#72) 592 Session 18: Case studies I (Oswald Losert) Qualification in explosion protection - use of an adaptive learning platform Patrick Dyrba (#28) 600 Learning From the Past: The Importa nce of Risk Assessment in Aluminium Dust Processes Marco Barozzi, Marco Deru di, M artina Silvia Scotton, Sabrina Copelli (#41) 607 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 7 Session 19: Hybrid mixtu re ex plosions II (Enrico Danzi) The maximum rate of pressure rise of hybrid mixtures Stefan Spitzer, Enis Askar, Paul Geoerg, Dieter Gabel, Arne Krietsch, Ulrich Krause (#5) 626 1st Inte rnation al Round Robin Test on Safety Characteristics of Hybrid Mixtu res Stefan Spitzer, Arne Krietsch (#18) 639 Making hyb rid mi xtures explosion a common case Matteo Pietraccini, Pierre-Alexand re Glaude, Anthony Dufour , Oliv ier Dufaud (#77) 650 Session 20: Ignition phenomena I (Michael Hilbert) Experimen tal study on inert products, moisture, and particle size effect on the minimum ignition energy o f combustible dusts Isabel Amez , Blanca Castells, David León, Javier Garcia-Tor rent, Ljiljana Medic (# 32) 662 Thermal Ignition: Effec ts of Fuel, Ambient Pressure and Nitrogen Dilution Conor Martin, Joseph Sheph er d (#48) 678 Ignition energy and flame propagation in ethylene oxide- air mix tures Christophe Proust (#85) 693 Session 23: Explosion modelling and experiments I (Bo Shu) Experimen tal study of Unco nfined Lean Hydrogen-Oxyg en Explosions Tomoyuki Johzaki, Taku ma Endo, Wookyung Kim, Keita Tanaka, Akihiro Ueda, Yangkyun Kim (#54) 701 CFD modelling of vented explosio ns for chambers of tw o differe nt scales Guillaume Lecocq, Jérôme Daubech, Emmanuel Leprette (#51) 713 Experimen tal and numerical study of the fuel effect o n fla me prop agation in long ope n tubes Guillaume Lecocq, Jérôme Daubech, Emmanuel Leprette (#52) 724 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 8 14 th International Symposium on Hazards, Prevention, and Mitigation of Industrial Explosions Braun sc hweig, GE RMANY – July 11-15, 2022 Greetings from the Local Organizing Committee Chair On behalf of the Local Organizing Committ ee, I w elcome you to the 14th international Sympo sium on Haz ards, Preve ntion and Mitigation of Industrial Explosions in the week from Jul y 11 to 15, 2022. The ho st is the Phy sikalisch-Techni sche Bundesanstalt in Braunschweig together with the Otto v on G uericke Un iversity Magdeburg. The se ries of I nte rna tional Symp osia on Hazards, Prevention and Mitigation of Industrial Explosion s h as a long tra diti on. After 12 undisturbed symposia, the past ISHPMIE 2020 had to take place as an online forum due to the corona pandemic. Despite the short-term organization of the ISHPMIE 2020 Forum, we had 575 registered participants. That could be regarded as a very good success for these circumstances. Now we are very pleased to ho st the ISHPMIE 20 22 as well. I would like to thank the International Organizin g Committee fo r giving us the opportunity to meet at the 14th Symposium in Braunschweig. Unfortunately, we are not yet f ree from the obstructi ve cir cumstances of the Corona Pandemic. As a r esult, many scie nti s ts especially f rom Asia cannot come and present their contribution s in person. The political situation related to the Russian war in Ukraine led to further restrictions. However, we did not let this discourage u s and we hope to ex perie nce a motivating face-to-face conference again. We have used the opportun ity for stimulating additions. We will try out the "Indu s try meets science" forum for the first time. Here we invited representatives from industry and associations to discuss thei r view on the needs and future challenges in research with the participants. Another new element is the "Young researcher's night" where the young scie ntists are brought together to expand their network. As w ith previous symposia, Best P aper Aw ards will b e pre sen ted , and selected arti cles wil l be proposed for publication in a special volume of the Jou rnal of Loss Prevention in the Pro ces s Indu stries. Finally, I wou ld like to th ank all colleagues for the ir suppo rt. Without them, the preparation and implementation of suc h a sympo sium would not be possib le: First and foremost the ISHPMIE 2022 organization team and the many helping hands of the PTB and the Otto von Guericke University Magdeburg, my co-chairs Ulrich Kraus and Detlev Markus, the program chair Holger Großhans together with the authors, reviewers and modera tors, the supportin g orga niz ations of the symposium mentioned on the website as well as Trygve Skjold and the Internat ional Organizing Committee. Thanks very much! I wish all part ici pants an inte resti ng conference week with many contacts and fruitful discussion s. Michael Beyer Local Organiz ing Commit tee C hair Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 9 c MEC + y LEL =1 (1) Bartknecht’s model: c MEC = ( y LEL -1 ) 2 (2) where c and y mean ref er to the dust concentration in the flammable hybrid mixture (g/m 3 ) and the volume concentration of the flammable gas in the hybrid mixture (vol %) respectively. MEC is the minimum dust explosion concentration (g/m 3 ), L EL is the lower explosion limit of the flammable ga s (vol %). Fig. 1 Explosion r egimes in th e plane methane content/nicotinic acid con centration (Gar cia-Agr eda et al., 201 1) . Based on th is eval uating explosion regime , a lot of researchers conduct ed more studies on hybrid mixtures. Th e explos ion characteristics of different types of hybrid mi xtures were further investigated (Addai et al., 2015b; Zhao et al., 2020 & 2021) . For further developing this evaluating explosion regime diagram, the stoichiometric line was introduced into the diagram by Sanchirico et al. (2011) to study the severity of hybrid mixture explosions and comparison to pure dust /ai r and vapour /air explosions, the area of dust driven explos ion, dual- fuel explosion and gas driven explosion were studied ( Fig. 2) . But according to the intensive experimental data associated with hybrid mixtures, the pre dictive ability of those two empirical models are quite condition -dependence (Khalili et al., 2012) . The results show ed that the c lassical models developed by Bartknec ht or Le Chatelier are not always conservative from a safety point of view, and th e s imilar experimental phenomenon was also found in other literatures (Addai et al., 2015b; Zhao et al., 2020) , indicating that these c lassical models have a strong conditional dependence. Therefore, focusing on the exponent of the power function, more comprehensive and syst ematically predictive model s were developed and proposed. Jiang et al. (2014 & 2015) developed a new formula based on the classical models developed by Bartknecht or Le Chatelier by introducing two indexes: K St and K G , and the experimental data show ed that the proposed formula accurately predicts the explosion and non-explosion boundary. Jiang’s model (2014 & 2015) : c MEC = ( 1- y LEL ) ( 1.12±0.03 ) K St K G ( 3 ) Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 16 where the K St is the dust explosion index and K G is the gas explosion index. By comparing those predictive models, it was found that they all predict the L EL of a hybri d mixture by constructing a func tional relationship between the flammable gas concentration (y), and the dust concentration (c). These models are all c onsistent in the structural form of the function, which is a power function relationship between the flammable gas concentration, and the dust concentration, so a more comprehensive model w as proposed (Ji et al., 2022) . Ji’s model (2022) : c MEC = (1 − y LEL ) λ (4) λ=5.12×10 -7 ×e ( P max G +P max St ) 2 ∙ lg ( K G +K St ) 0.34 +1.1 (5) where 𝑃 𝑚𝑎𝑥 𝐺 is the maximum explosion pressure of the combustible gas, 𝑃 𝑚𝑎𝑥 𝑆 𝑡 is the maximum explosion pressure of dust. Fig. 2 Explosion r egimes at t v ¼ 60 ms (Cir cles dimension s ar e pr oportional to K st ) (Sanchirico et al., 201 1) . Besides, a simple mathematical model has been derived from the enthal py balance of the whole system assuming that the combustion kinetics of pure species are independent and unchanged by the presence of other combustible species, complete conversion of the reacta nts and no heat losses was proposed (Abbas et al., 2019) . 𝐿𝐸𝐿 𝐻 = 𝐶 𝑝,𝐴 ∆𝑇 𝑥 𝑑 ( ( −∆ℎ 𝑅,𝑑 ) −𝐶 𝑝,𝐴 ∆𝑇 )+𝑥 𝑔 ( ( −∆ℎ 𝑅,𝑑 ) −𝐶 𝑝,𝑔 ∆𝑇 )+𝐶 𝑝,𝐴 ∆𝑇 . 100 (6) where 𝐿𝐸𝐿 𝐻 is lower explosion limit of hybrid mixture (kmol/kmol), 𝑥 𝑑 and 𝑥 𝑔 are the mole fraction of dust and gas (kmol/kmol) respectively, 𝐶 𝑝,𝐴 and 𝐶 𝑝,𝑔 are the molar specific heat of air and gas at constant pressure (kJ/kmol.K) respectively, −∆ℎ 𝑅 ,𝑑 is the reaction heat of combustible dust (kJ/kmol) and ∆T = (𝑇 𝑓 − 𝑇 𝑖 ) refers to the rise in temperature from ambient or initial to the flame temperature. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 17 By coll ecting, comparing and summarizing the experimental data associated with hybrid mixtures available in literatures (Denkevits, 200 7; Denkevits and Hoess, 2015) , an appr oach for evaluating the explosion regime was proposed based on the different ignition energy (Cloney et al., 2013 & 2017) . The d iagram can be re-divided according to the di fferent of amount o f igni tion energy (high ignition pyrotechnic igniter) and low ignition energy (electric spark ignitor) and combusti on mechanism, gas-driven, dual-fuel regimes and dust dr iven regimes were ref ined. As shown in Fig. 3 . Fig. 3. Hybrid explo sion regime diagram for hi gh-reactivity (left), low-reactivit y (right) dus t reproduced from the experimenta l data of for spark ignition and 60 ms dela y. The largest values of P m (bar) and K m (bar-m/s) are indicated fo r each recorded gas concentration as text (Cloney et a l., 2017) . The deviations of the mathematical mod els may ari se from the fact, that in non e of the previous works the gas amount was verified and small deviations of it have an impact on the characteristics. 2.2 Chemical kinetics and combustion mechanism of hybrid mixtures More than one cent ury ago, in 1885, the experimental phenomenon that mixing coal dusts with methane at a concentration lower than the methane lower flammability li mit (i.e., 4.1% in air) would allow the explosion of the dust/gas mixture was observed (Engler, 1885) . From then on, the studies on the explosion characteristics and combustion mechanism of hybrid mixtures have been ongoin g. In general, the combustion behaviour thr ough hybrid mixtures depends on a number of steps of the oxidation process of the combustible fraction. 1. The devolatilization of the volatile fraction of the soli d fuel (not for dust with non -volatile), 2. Mixing of th e volat iles with oxygen in the gas phase, 3. Combustion of the flammable gas in the gas phase, 4. Combustion of the remaining solid fraction (mostly char) (Krause and Kasch, 2000) . For combustible dust with or without volatiles, the chemical kinetics and combustion mechanisms of them during combustion can be very different. For orga nic dusts, a two-phase combustion model was developed by (Slezak et al., 1985) to investigate the flame propagation in rich mi xtures of coal dust in air environment, which including heterogeneous combustion, pyrolysis of the coal, and homogeneous combustion of volatile ma tter and the optically thick limit for radiative heat transfer . After that, many researchers conducted experiments to investigate the combustion mechanisms of dust explosion s, their research mainly started with the particle size, flame propagation and temperature profile (Ju et al., 1998; Han et al., 2000; Dobashi and Senda, 2002 & 20 06) . Moreover, the inerting effect on the combusti on of hybrid mixtures by introducing some inert substances like carbon dioxide, nitrogen and other comp onents Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 18 was also studied to investi gate the influence on the combustion mechanism (Wang et al., 2020b) . Based on the experimental data available in the li terature, the combustion mechanisms of hybrid mixtures including the effect of inert gas was concluded in Fig. 4 . For most of the metal dust, li ke iron powder, copper powder, silicon powder etc . the combustion behaviour will occur in the surface of the particle with the help of high enough ignition ener gy (Sun et al., 1998a, 2001a & 2003) . In par ticular, iron is a suitable candidate as it burns purely heterogeneously by surface reactions. The iron – air adiabatic flame temperature is similar with the methane flame temperature, which allows the observation of both flames sim ultaneously. It has also been shown that iron ca n burn in either a diffusion-controlled or a kinetically-controlled regime depending on, among other things, the oxidizing environment (Julien et al., 2015a) . But, with the addition of methane, the ignition energy will decrease and the combustion will occu r more easily (Julien et al., 2014 & 2015a) , and correspondingly, the flame structure will change. Fig 4. Schematics of combustion mechanism of hybrid mixtures including inerting effect (Wu et al., 2 021b & Wu et al., 2022 ). As the promising candidate of clean combustion ener gy, aluminium was also studied intensivel y either using the pure dust or with add ition of flammable gases, like methane or hydrogen. Unlike the combustion of iron, the combustion of aluminium suspension has two characteristics: 1. The devolatilization will occur. 2. The formation of Alania coat which cover the aluminium core, which makes the combustion mechanism more complex than that o f iron powder . Therefo re, understanding the combustion mechanism of aluminium particles in details is the basement for understanding th e flame propagation of alu minium dust suspension and aluminium-based hybrid mixtures in dep th. For decades, lots of res earchers dedicated their contribution to research the combus tion of not only the isolated individual aluminium parti cles but the aluminium suspension (Bucher et al., 1996; Goroshin et al., 1996 b; Dufaud et al., 2010) . With the addition o f flammable gases like hydrogen, the combustion mec hanism of aluminium powder will change and the combustion mechanism was concluded ( Fig. 5 (Yu et al., 2020)) . The combustion mechanism of a luminium particles was revealed Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 19 by (Bucher et al., 1996; Jing et al., 2021 & 2022) , which is shown in Fig. 6 . Fig. 5. Reaction mecha nism of aluminium du st explosion in hydrogen/air environment (Yu et al., 2020 ) . ( A ) ( B ) (C) Fig. 6. (A) Melting-oxidation mechanism, (B) Growth mechanism of oxide layer, (C) transient reactio n mechanism of flake al uminium dust (Jing et al., 202 1 & 2022) . In particular, heat transfe r plays an important role during the combustion of dust suspension and fl ame propagation including conductive heat transfer and thermal radiation (Badiola and Dreizin, 2013; Thimothé e et al., 2016) . The role of thermal rad iation was investigated by Christophe et al. (2017) who studied the thermal radiation in dust flame propagation and proposed a new experimental measurement of thermal radiation in dust flames together with a physical interpretation, shown in Fig. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 20 7 . By generati ng and analysing the pyrolysis gases of some car bohydrates such as starch, Dufau d et al. (2011) demonstrated that the pyrolysis phase could not be systematically neglected when considering the dust explosion kinetics. Fig. 7 . Cassel's problem an d definitions (Christoph e et al., 2017) . 3 Experimental measurements for different explosion safety characteristics 3.1 Experimental apparatus For now, there is still no sta ndard apparatus and procedure available espe cially for the study of explosion behaviour of hybri d mi xtures. So, to better experimental study on the hybrid mixture explosion, a variety of standardized and modified experimental apparatus were adopted for experiment tests, including ISO 1 m 3 explosion chamber, 20L- sphere , Godbert- Greenwald (G-G) furnace or BAM furnace , Hartman tube or MIKE 3, which originally used in the experiment of combustible dust explosion. At the very beginni ng, Hartmann tube was widely used in the res earch of dust cloud explos ion due to it s characteristics of low cost, simple structure and easy maintenance (Sweis, 2006; Nifuku et al., 2007) . However, the accuracy of the res ults measured by Hartm ann tubes is poor because of the small volume, the use of ele ctrostatic ignition and the poor flame propagation path. Since the experimental results measured by ISO 1 m 3 device were close to practical industries (Going et al., 2000; Dastidar et al., 2001) , therefore, it has long been used as the only standard device to test the explosion charac teristics of dust (Zhen and Leuckel, 1997; Garcia-Torrent et al., 1998; La´ zaro and Torrent, 2000) . Numerous researchers have attempted to reduce the volume of the 1 m 3 device and have found that to achieve a balance between ease of testing and realistic data, the volume of the vessel must not be less than 20 L (Eckhoff, 2 003) . On this basis, the 20L-sphere was developed (Siwek, 1996) , shown in Fig. 8, and the experimental results were conf irmed to be in good agreement with the ISO 1 m 3 devic e (Cashdollar and Chatrathi, 1993; Proust et al., 2007) . A USBM 20 L laboratory explosibility cham ber was used to conduc t experiment s and the results indicated, that the Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 21 results from experiments show relatively good agreement with those from full-scale experimental mine tests (Cashdollar, 1996) . The 20L-sphere has the advantages of small size, easy operation, low experimental costs, high accur acy and gradually replaced ISO 1 m3 as the main apparatus of laboratory research dust explosion equipment. In addition, G-G fur nace has also been w idely used for determining the minimum ignition temperatur e (MIT) (Addai et al., 2016b) , and the modified Hartman tube was also often used for the measurement of the minim um ignition energy (MIE) (Norman et al., 2015) . Fig. 8 & Fig. 9 give the p ictures of those apparatus mentioned above. Table 1 lists the apparatus used popularly for measuring the flame propagation behaviour. a. ISO 1 m 3 explosion chamber b. 20L-sphere Fig. 8 . Apparatus used for the experimental r esear ch of dust cloud a. MIKE III apparatus b. G-G furnace Fig. 9 . Apparatus used for the measur ement of MIE and MIT T ab le 1 : Apparatus used for the r esear c h of flame pr opagation. Source Main apparatus Supplementary appa ratus Remark V ertical tube Liu et al. (2007) Y in et al. (2009) Rectangular tube, 8 × 8 × 50 cm. A h igh-speed v ideo camera. A thermocouple. A Sch lieren opt ical part. A time controller . The upper end of the tube is open Gao et al. (2014a, 2014b) Cyl. vertical tube, d = 68 mm. A h igh-speed v ideo camera. A thermal infrared imaging device. The tube has di ffe rent lengths including 30 0 mm, 600 mm, 900 mm. The upper end of the tube is open. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 22 Chen et al. (1996) Ju et al. (1998a, 1998b ) Dobashi et al. (2006) Sun et al. (2000,2001a, 2001b, 2005) Gao et al. (2015, 2017) Zhang et al. (2016, 2017, 2020) Y u et al. (2016, 2020) Cyl. ver tical tube, d upper = 64 mm, d lower = 44 mm d upper = 60 mm, d central = 96 mm, d lower = 84 mm d upper = 94 mm V tube = 1 L, d = 76 mm d upper = 95 mm V com. chamber = 1 L, d central = 96 mm A h igh-speed v ideo camera. A b and-pass filt er . A light attenuator . A laser-scattering opti cal instrument. A two-dimensional particle images. V elocimetry system. At the end of spray , the middle part of the duct was moved down to its bottom position suc h that the combustible pa rticle cloud was provi ded in an open space. Y uzuriha et al. (2017) Rectangular tube, 7 × 7 × 30 cm A h igh-speed v ideo camera. A mesh was attached to the top of the duct, which prevented the particles out of the duct. Zhang et al. (2017) Rectangular tube, 8 × 8 × 50 cm A h igh-speed v ideo camera. A f ine thermoc ouple and an ion current probe. The top of the duct is open and the bottom is closed. Chang et al. (2020) A modified Hartmann tube, 7 × 7 × 29 cm A 25 μm R-type thermocouple. - Y u et al. (2021) Rectangular tube, 10 × 10 × 50 cm A high-speed video camera. A light attenuator . A f ast respo nse pressure sensor with. an acquisi tion frequency of 100 kHz. A closed chamber . T rans par ent latex balloons Skjold et al. (2013) A transparent latex balloon. A b alloon hold er . A dispersion nozzle. A downw ard-f acing spark gap. A h igh-speed v ideo camera. Using chemical i gniter . Cheng et al. (2018) A transparent latex balloon. A b alloon hold er . A d ispersion nozzl e. A h igh-speed v ideo camera. Using chemical i gniter . Julien et al. (2015b , 2015c) V ickery et al. (2017) A h igh-speed v ideo camera. A n eutral densi ty filters. A pho todi ode and a microphone. Spark ignition. “Bunsen burne r” Goroshin et al. (199 6a, 1996b) Burner An Ocean Optics USB 4000 spectrometer c oupled with a - Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 23 Soo et al. (2013) Julien et al. (2015a) 100-micro optical fiber . A h igh-resolution di gital camera A h igh-speed v ideo camera A v ariable neut ral-density filter Other apparatus Chen et al. (2005) Cyl. horizontal tube, d = 14 cm, L = 12 m 20 sets of specially desi gned dispersion systems. Six piezoelectric trans ducers. One end of the combustion tube is joined to a 10 m 3 da mp- tank. The other end is sealed with a fl ange mounted evenly with six ignitors. Krause et al. (2006) Cyl. vertical tube, d = 30 cm, L = 1.45 m A h igh-speed v ideo camera - Kern et al. (.2015) Cyl. vertical tube, d = 14 cm Five silicon photodiodes An optical system with photodiodes A h igh-speed v ideo camera The dust feeding de vice was located at the to p of the tube. W ang et al. (2016) Rectangular tube, 8 × 8 × 50 cm. A g as supply unit. T wo thermocouples. An ignition system. A h igh-speed v ideo camera. A synchronization controller The bottom end of the chamber was closed and the top end was ope n. Xia et al. (202 1) A constant volume spherical chamber: d = 200 mm, H = 280 mm. V = 6.19 × 10 -3 m 3 Three types of photog raphy were used to c apture the flame propagation: direct imaging, OH radical photography , and schlieren photography . Capacitor dischar ge ignition. Arne et al. (2021) Cyl. vertical tube, d = 7 cm, L = 1 m A h igh-speed v ideo camera A dust concentration measurement system - 3.2 Determination of explosion severities parameters The determination of explosion severities parameters of hybrid mixtures is significant ly important for better understanding of the e xplosion behaviour of hybrid mixtures and the rationalization safety desi gn of in dustrial process involving the combustible dust materials (Cashdollar , 2000) , especially for the hybrid mixtures. Therefore, many researchers paid their attention to the research of severit y parameters, incl uding the maximum explosion pressure ( P max ), the maximum rate of explosion pressure rise ((dp/d t ) max ) and the explosion index ( K st ) (Denkevits, 2007; Li et al., 2012; Ji et al., 2018) . It is important to know , that the addition of flammable gases into the dust suspension system will increase the explosion severity compared with pure dust suspension (Dufaud et al. 2008; Addai et al., 2015a; Kundu et al., 2018; Song et al., 2019) . The ef fect of the Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 24 addition of flammable gas on the explosion severity was shown in Fig. 10 . Fig. 10 . The effect of smal l amount of CH 4 on the explosio n pr ocess of 20 g/m 3 bituminous coal dust (Zhao et al., 2020) . But the explosion severity cannot be promoted all the time with addition of fl ammable within the confined explosion chamber due to the oxygen consumption caused deeply by flammable gas presence and increased the oxygen dif fusion resistance in the hy brid mixtures, which reduced the reaction intensity of oxygen in dust particle surfaces. The decrease of dust particle burning ratio reduced hybrid mixture's explosive property (Li et al., 2012) , and thi s phenomenon can be seen in Fig. 1 1 . Fig. 1 1 . Effect of methane fraction on coal dust explosion parameters (Li et al., 2012) . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 25 Chemic al ignitor: 10 kJ coal dust with lower volatile content Song et al., (2019) Coal- dust/metha ne D 50 : 28 μ m for coal du st #1 and 76 μm for coal dust #2, 133 μm for coal dust #3. C dust : 50 - 400 g/m 3 , C gas : 5, 8, 10 vol%. The maximum explosion pressure and maximum rate of pressure rise keep rising with the increasing initial pressure. Addai et al., (2015) Starch/met hane/aceto ne C dust : 5- 10 00 g/m 3 , C gas : 1-4 vol% 1. MEC. 2. P max . 3. (d P /dt) max . Spark ignitior : 10 J A hybrid mixture explosion is possible even when dust, gas and vapor concentrations are respectivel y lower than their minimum explosion concentration (MEC) of dust and lower explosion limit (LEL) of gas and vapor . Kosinski et al., (2013) Carbon black/prop ane C dust : 0-500 g/m 3 , C gas : 0-5 vol% 1. P max . 2. (d P /dt) max . Chemic al ignitor: 1 kJ. Addition of some quantiti es of combustible gases (here: propane) may sustain combustion processes. The addition of Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 32 Y u et al., (2021) Aluminum dust / hydrogen D 50 : 56.1 8 μm C dust : 1000, 1500 and 2000 g/m 3 , C gas : 0-30 vol% 1. Flame morphology . 2 . Flame propagation velocity . Spark ignitior hydrogen can significantly af fect the ignition delay time and the flame propagation velocity of the hybrid mixtures Y u et al., (2020) D 50 : 56.1 8 μm C dust : 1000, 1500 and 2000 g/m 3 , C gas : 0, 5, 10 vol% 1 . Flame morphology and microstructures. 2. Flame propagation velocities Attributable to a variety of intermediate products competing f or oxygen and absorbing he at, the hybrid explosion residues cooled faster , porous oxide layers and incompletely oxidized aluminum spheres with small particle sizes were formed In summary , dust explosions are time-dependent flame propagation process, therefore, safety characteristics are the critical parameters, including ignition temperature, ignition ener gy , lower explosion limitation and oxygen concentration, etc. to maintain the sustainable flame propagation. Knowing these characteristic parameters in depth will have s ignificant ly beneficial to understand the flame propagation behaviour and the prevention and mitigation of potential explos ion risk in industrial scenario. 4 Flame characteristics and propagation 4.1 Flame pr opagation behaviours during the dust explosion pr ocess As early as 1968, the flame propagation of combustible dust and flammable gases mixtures was already studied (Singer and Liebman, 1968) , and a comparison was made between the combustion flames of aluminium powder and coal powder in a methane/air environment. T o better understand the flame propagation of dust cloud combustion, the combustion mec hanism of isolated particle is Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 33 essential. A theoretical study on the combustion of dust clouds from the scale of dust particles was performed (Nomura and T anaka, 1992) , they established the combustion model of particles based on some assumption. The picture of combustion model was shown in Fig. 12 . Fig. 12 . Du st cloud model (Nomura and T anaka, 1992) . Dahoe et al. (1996) stud ied the role of flame thickness in the validit y of the “cube -root” law about the dust explosion in the spherical vessels and developed a model (the three -zone model) for the pressu re evolution of confined dust explosions in spherical vessels which tak es the flame thickness into account, the pressure-time curves that are generated with this model show a good resemblance with those measured in practice. The model was shown in T able 4 . T ab le 4 . An overview of the differ ent phases that the flame fr ont goes thr ough during a n explosion (Dah oe et al., 1996) . Case 1 δ  R vessel Case 2 δ  R vessel Phase 1a: r rear = 0.0 r front  δ Phase 2a: r rear = 0.0 r front  R vessel Phase 1b: r rear = r front – δ δ  r front  R vessel Phase 2b: r rear = 0.0 r front = R vessel Phase 1c: δ  r rear  R vessel r front = R vessel Phase 2c: 0.0  r rear  R vessel r front = R vessel As we all know , for better understand the flame propagation behaviour of hybrid mixtures, the research on the pure dust is esse ntial and basement. The structure of flames propagating through metal particle clouds and the behaviour of metal particles near th e flames have b een examined experimentally (Sun et al., 1998 & 2006a) , and the result show that the combustion zone consi sts of luminous particles without gas- phase flame and the velocity of particles at the leading edge of combustion zone is nea rly proportional to the flame ve locity . After that, the temperature profile across the combustion zone propagating through an iron particle cloud and the concentrati on profile of particles across a flame propa gating through an iron particle cloud were also investigated (Sun et al., 2001 & 2003) . Julien et al. (2015a ) studied the flame structures and particle combustion regimes are studied in hybrid fue l mixtures of methane and iron using a modified Bunsen burne r with two different oxidizing environments: stoichiometric methane–air mi xture and lea n methane–oxygen–nitrogen mixture, making the content of oxygen different in those two cases, and the result showed that existing a critical concentration of iron powder that a coupled flame front in the combustion products of the methane flame can be formed. Furtherm ore, after the metal -powder fl ame formation, a double front structure separated by a dark zone is observed in the kinetic regime, whereas the two flames over lap Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 34 and form a single Bunsen cone in the case of the dif fusion regime , as shown in Fig. 13 & 14 . Fig. 13 . Ir on–methane hybrid fla mes (Julien et al., 2015a ) . Fig. 14. Differ ent combustion r egimes observed in the work of Julien et al. (2015 a) . Besides, the motion behaviour of particles ahead of the flame front were investigation by combining th e particle image velocimetry (PIV) techniques, the combustion mechanism was rev ealed by Gao et al. (2015a) and Haghiri and Bidabadi (201 1) and shown in Fig. 15 . Fig. 15 . Flame pr opag ation mechanism (Gao et al., 2015a) . There are lot of influential factors that can affect the flame propagati on behaviours, including the particle size distribution, particles concentration (Gan et al. 2018a; Zhang et al., 2018) , radiation (Cao et al., 2014a; Bidabadi and Azad, 2015; Chr istophe et al., 2017) , turbulence and material thermal characteristics (Gao et al., 2012 & 2013) , etc. Besides, the volatile conte nt invol ved in the dust also have effect on the combustion of dust and flame p ropagation, i.e., coal dusts with higher volatile matter contents ( V daf ), lower vitrinite reflectance ( R o,max ) and less ash contents ( A ad ) show stronger Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 35 explosion severity (Li et al., 2018a) . Based on the analysis on the post -explosion residues, experimental results from Li et al. (2018) indicated that there is a linear relationship between explosion flame propagation speed and dust concentration and quantitatively analysed by standard method the emi ssion properties of gaseous p roducts during coal dust explosion, explained the structure evolution during coal particle explosion process. 4.2 Effect of particle size and its distribution on the flame pr opagation According to the research results obtained from the literatures (Dufa ud et al., 2010; Kuai et al., 201 1; Zhan g et al., 2017b; Zhang et al., 2017a; Liu et al., 2018; Ichinose et al., 2020) . The flame propagation behaviour varied with the particle size distributions , even if Sauter mean diameter ( D 32 ) was same, and flame can propagate very fast in small and monodispersed particles which didn't contain larg e particles (Y uzuriha et al., 2017) . Gan et al. (2018 b) studied the ef fects of polyethylen e particle size distributions on flame propagations of hybri d mi xtures of ethylene/polyethylene, the results showed that flame propa gation velocities and maximum flame temper atures increased with the decrease of particle size distribution as a whole. On the other hand, p articles with nano-size and micro-size will display quite dif ferent flame propagation characteristics during the combusti on of its suspension. In general, compared with mi cro-size combustible particles, par ticles with nano-size hav e some unique properties (Y ett er et al., 2009; Krietsch et al., 2015) : 1. The increased specific surface area may lead to an increase in ignition sensitivity and rea ction severity . 2. Some powders may show pyrophoric behaviour wh en sized down to nano- scale. 3. Oxygen adsorption at reactive surfaces of individual particles may result in a passivation o f the powder . 4. Powders may tend to form agglomerates which are of microscale. Experimental results have shown that, in the one hand, for nano-particles, flame was characterized by a regular spherical shape and spatially continuous combustion structure combined with a number of luminous spot flames. The flame propagation mechanism was similar to that of a premixed gas flame coupled with solid surface combustion of the agglomerates. On the other hand, smaller particles maintained the leading part of the propagating fl ame and gover ned the combustion process of PMMA dust clouds (Zhang et al., 2016) . Similar experimental phenomenon was also observed from other relevant literatures (Bouillard et al., 2010; Escot Bocanegra et al., 201 1; Li et al., 2016; Y u et al., 2016; Gao et al., 2017; Chang et al., 2020; Danzi et a l., 2021;) . And the flame propagation mechanism of nano- si ze and micro-size particle suspension was concluded as shown in Fig. 16 . ( A ) ( B ) Fig. 16 . Flame pr opag ation and structur e thr ough 100 nm ( A)/30 mm (B) PM MA dust cloud (Zhang et al., 2016) . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 36 4.3 Flame quenching distance One of the important fundamental parameters in laminar flame theory is the flame quenching distance, as it is closely related to the characteristic laminar flame thi ckness and structure. T ogether with burning velocity , the quenching distance is a fundamental flame characteristic that reflects both the flame propagation mode and its structure (Palecka et al., 2015) . The quenching distance and fl ame speed under aluminium-oxygen-nitrogen and alum inium-oxygen-helium environment was investigated by Goroshin et al. (1996a) and found that the quenching distance and fl ame speed are very weak functions of dust concentration in rich mixtures and the substitution of nitrogen for helium in air increases the minimum quenching distance from 5 to 7 mm, as shown in Fig. 17 . Under the experimental conditions of red uced-gravity environment, T ang et al. (2009) studied flame quenching distance using iron particles, the results showed that the flame quenching distance increases linearly with particle size from less than 2 mm quenching distance for the 3 μm-sized dust to 10 mm quenching distance for the 27 μ m -sized dust, which is agreement with the numerical results from Bidabadi et al. (2016) and Bozor g et al. (2019) , who found that flame propagation through iron powders composed of smaller particles is faster , and the quenching distance is lower . And the reason can be explained that increasing particle s ize increases the flame thickness and decreases the burning velocity (i.e., increases the quench ing distance (Jarosinski et al., 1988) . Palecka et al. (2015) investigated quenching distance of flames in hybrid meth ane–aluminium mixtures, and found that Coupled aluminium–methane fl ame fronts only appear above a thr eshold aluminium concentration around 300 g/m3. Below this concentration, the appearance and quenching behavi ours of the methane flame seeded with reactive aluminium and inert SiC dusts are similar , as shown in Fig. 18 . Fig. 17. Quenching dista nce as a function of dust concentration (Gor oshin et al., 1996 a) . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 37 Fig. 18. Results of flame quenching experiments at differ ent concentratio ns of aluminium dust in 16.3% O 2 /8.1% CH 4 / 75.6% N 2 /Al mixtur es (Pa lecka et al., 2015) . 4.4 Burning velocity Burning velocity of fuel-air mixtures are always used to characterize the reactivity of fuel-air mixtures. In distinction to other explosion parameters, e.g. , K G or K st , burning velocity show the advantage to emphasize the influence of flow velocities on the fl ame propagation (Krause and Kasch, 2000) . As mentioned ab ove, there is a lack of the standa rd protocol associated with the measurement of burning for dust-air suspension, the methods available in the literatures for me asuring the burning velocity of dust-air suspension are derived from that of f lammable gas, including tube met hod and burner method, soap bubble, explosion vessel. In particular , the tube method is best for burning velocity up to 80 cm/s, and that the constant volume bomb method is best for the higher burnin g velocity (Andrews and Bradley , 1972) . Some empirical models were introduced to calculate the value of burning velocity with some assumptions. 𝑆 𝑢 = 𝐴 𝐴 𝑓 𝑆 𝑠 (1 5) where 𝑆 𝑢 is the burning velocity , A is the front flame surface and A f is the surface of the cross section of the tube. This model will valid under the assumptions of 𝑆 𝑢 is constant value over the entire cross section of the tube, the density of unburnt fuel ahead of the flame front 𝜌 𝑢 is constant and the flame speed is uniform over the tube cross section. In another case, because the flame aerodynamics changes as the flame propagates along the tube and invalidate the assumption of a constant 𝑆 𝑢 over the cross section, the flame area A f is also variable, therefore, the following model was proposed: 𝑆 𝑢 = 𝐴 𝐴 𝑓 (𝑆 𝑠 − 𝑆 𝑔 ) (16) where 𝑆 𝑔 is the mean unburnt gas velocity averaged over the tube cross-section area A f . Based on those models, the burning velocity of metal particles with nano -sized were investigated experimentally (Krietsch et al., 2021) . At the same, considering the stretched of the flame due to the turbulence during the propagation and the partial confinement of the tube method, the thermal expansion factor α and flame’ s stretching factor K cal led Karlovitz’ s factor were introduced into the Equetion-1, and the un-stretc hed flame burning velocit y was inves tigated (Cuervo et al., 2017; Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 38 T orrado et al., 2017) , and the results show that he str etching of a gas flame is strongly influenced by the addition of dusts. Nevertheless, for lower gas concentrations and larger dust concentrations called ‘dust-driven regime’, the presence of powders tends to limit the flame velocity to that of the less reactive compound. K = 1 𝐴 𝑓 𝑑 𝐴 𝑓 𝑑𝑡 (17) 𝑆 𝑢 = − 𝐿𝐾 + 𝑆 𝑢 0 (18) Burner method were also used for investigati ng the burning velocity of hybrid mixtures by many researchers (Goroshin et al., 1996b; Soo et al., 2013; Julien et al., 2014) . Goroshin et al. (996b) studied the burning velocity in fuel-rich aluminium dusts cloud, and found that burning velocity was a strong function of the molecular tr ansport properties of th e carrier gas, and the weak dependence of the flame speed on dust concentration is a direct result of the weak dependence of the particle b urning rate in the diffusive regime on the flame temperature. In addition, a sim plified, ti me-dependent numerical model that considers the influence of both diffusional and kinetic rates on the par ticle combustion process was proposed to study the thermal structure and burning velocity of flames in non -volatile fuel suspensions (Soo et al., 2016) . Over the years, McGill University has been developing alternative experimental techniques based on direct observation of dust flames, yielding reliable fundamental parameters such as flame burning velocity , temperature and structure, a new apparatus for in vestigating flame propagation in turbulent dust clouds at near constant pressure conditions was designed by Skjol d et al. (2013) . After that, an extensive series of balloon experiments were performed (Julien et al., 2015b & 2015c) , and found that the flame speed of stable flames is found to be a str ong function of the heat conductivity of the gas mixture, and pulsating and spiral -like flames are discovered in fuel -lean mi xtures, and flames with cellular patterns occur in very -fuel-rich suspensions. By using the same balloon setup, V ickery et al. (2017) studied the propagation of isobaric spherical flames in hybrid aluminium- methane fuel mixtures and indicated that the dif ference in behaviour at low concentrations in mi xtures with and without excess oxygen is explained by the abilit y of aluminium particles reacting with free oxygen to ignite and burn in the dif fusion-limited combustion mode. The maximum ef fective burning velocity was proposed to be used as th e substitute of the K St but shows less apparatus dependent than the correspo nding K St values (Pu et al., 2007) . For metal hydride suspension, the flame propagation behaviours and influential factors of T iH 2 dust explos ions at a constant pressure were also studied (Cheng et al., 2018) , and the results indicated that the burning mec hanism of T iH 2 dust is thought to be mainly controlled by dif fusion regime, the appearance of hydrogen gas accelerates the combustion rate of T iH 2 particles and also makes the T iH 2 dust changed from a di screte media to a continuum, which may account for the phenomen on that the flame speed in dust cloud of T iH 2 is lar ger than that of T i at the same concentration s no matter in air or oxygen atmosphere. The burni ng velocity of this kind of apparatus, i.e., isobaric condition created by the tran sparent balloon, was calculated from the following equation: 𝑆 𝐿 = 𝑆 𝑓 [1 − ( 𝑟 𝑏 3 −𝑟 𝑏0 3 ) 𝑟 𝑓 3 ] (1 9) where 𝑟 𝑏 , 𝑟 𝑏0 , 𝑟 𝑓 are radii of the initial prefilled balloon , the followi ng inflated b alloon and the dust flame respectively , which can be obtained from measuring th e still frames of flame front, and the 𝑆 𝑓 represents the flame speed propagated in dust clo ud (Skjold et al., 2013) . 4.5 Flame pr opagation mechanism Understanding the flame propagation mechanism in depth is essential for designing the Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 39 protective measurements in industrial scenario. According to the dif ferent combustion steps between particles with low-volatile and particle with high -volatile, the research on the flame propagation mechanism should be start with two categories roughly: heterogeneous combustion (the combustion occurs in the sur face of the particles) and homogeneous combustion (the combustion occurs in the gas phase), in fact, in some cases, the combustion process can be combined with heterogeneous combustion and homogeneous combustion (T an et al., 2020) . In spite of int ensive contribution associated with hybrid mixtures explosion have been done (Chen and Fan, 2005; Jinhua et al., 2006; W ang et al., 2006; Liu et al., 2007; Kern et al., 2015; Gao et al ., 2015; Ajrash et al., 2017) , the f lame propagation mechanism of hybrid mixtures still cannot be clarified, especially on the interaction between the gas phase and soli d phase, including the heat transfer (conductive and thermal radiation), mass transfer and flow state. Similar with the gas flames, flames in particulate suspensions at the labor atory scale are primarily driven by molecular heat diffusion and have comparable burning velocities (Bergthorson et al., 2015) . However , they exhibit several significant differences in their structure and behaviour from homogeneous flames due to their multiphase nature. The flame propagation behaviours of aluminium suspension were studied intensively (Marmo et al., 2004; Chen and Fan, 2005; Dufaud et al., 2010; Julien et al., 2015b; Julien et al., 2015 c; Li et al., 2016; Sun et al., 2006b; Y u et al., 2016 ) . Furthermore, the main dist inctive fea ture of a flame in a solid suspension is the ability of particles to ignite. Namely , to transition from a combustion regime limited by reaction kinetics to a regime limited by diffusion of the oxidizing gas towards the particle surface, or in the case of evaporating particles, towards the micro-flame enveloping each individual particle. After ignition, the temperature of the particle or micro-flame can excee d the gas temperature by sever al hundre d degrees, often surpassing the adiabatic flame temperature for fuel-lea n mixtures. The particle combustion rate in th e diffusion combustion regime is a weak, non -Arrhenius, function of gas temperature (Soo et al., 2016) . Unlike gas flames, the width of the flame reaction zone in particle su spensions can span a larg e temperature range and can be comparable to, or even exc eed, that of the pre heat zone (Goroshin et al., 1996a) . The existence of diffusi on micro-flames within a global fl ame-front (in effect, flames within the flame), which are insensitive to the bulk gas temper ature, make s dust flames resistant to heat loss (F rank-Kamenetskii, 1969; T ang et al., 201; Ber gthorson et al., 2015;) and also serves to maintain a constant burning velocity with increasing fuel concentration in fuel -rich mi xtures (Goroshin et al., 1996) . The addition of flammable gas into the particulat e suspension can have significant ef fect o n the flame propagation c haracteristics, including the flame velocity (Y u et al., 2020) , flame structure (Soo et al., 2013) , burning velocity (Soo et al., 2013; Julien et al., 2015c) and the flame propagation stability (Y u et al., 2020) , etc. The most popular hybrid mixtures used in the laboratory is the combination of coal dust and flammable gases, like methane, hydrogen and carbon monoxide, etc. According to the intensive experimental data available in the literatures (Liu et al., 2007; Li et al., 2012; Addai et al., 2015a, 2015b & 2017; Song et al., 2019; Zhao et al., 202 0 & 2021) , the syner getic ef fect between in the hybrid mixtures play a significant role in promoting the combustion and decreasing th e sensitivity of combustion of hybrid mixtures. For investigating the synerg etic effect of flammable gas and or ganic dust at the condition of LE L concent ration, lycopodium and methane were selected (Abbas et al., 2022) , and the result indicated that for hybrid mixtures of carbonaceous dusts (like lycopodium) at their LE L, ignition occurs in the gas phase, howe ver , flame propagation is only possible through a two-way interaction of dust and gas during the cou rse of combustion. Gao et al. (2015) studied the flame propagation mechanism in du st explosions, there were two dif ferent combustion regimes : kinetics-controlled regime and devolatilization controlled regime, which observed during the Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 40 experiment, seen in Fig. 19 . (a) Kinetics controlled regime (b) Devolatilization -controlled regime. Fig. 19 . Schemati cs of flame pr opagation mechanisms in dust explo sions. (a). Kinetics contr olled r egime, (b). Devolatilization -contr olled r egime (Gao et a l., 2015) . Based on the simulation results and the res earch results from Garcia-Agreda et al (201 1 ) , Cloney (2018) fur ther investigated the explosion evaluating regime by taking the equivalence ratio analysis and timescale analysis into account and explained the flame propagation mechanism in details, as shown in Figs. 20 & 21 . The results demonstrated that maximum burning velocity for hybrid m ixtures may occur along li nes of constant Φ ℎ 𝑣 (volatile component equivalence ratio) for small particle sizes and that a kinetic-limited combustion regime may be pre sent for Φ ℎ 𝑡 (total equivalence ratio) and an impeded gas flame regime where t he dust could not react fast enough to contribute to energy release in the flame front. (a) (b) Fig. 20 . Burning velocity and combustion r egime diagram for hybrid mixtur es of (a) 10 μm, (b ) 33 μm coal dust particles and methane g as (Cloney , 2018) . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 41 Dufaud, O., Pettin, L., T raoré, M., 2008. Dust/vapour explosion: Hybrid beh aviours? J. Loss Prev . Process. Ind. 21(4), 481-484. Dufaud, O., P errin, L ., T raoré, M., Cha zelet, S., Thomas, D., 2009. Expl osions of vapour/dust hybrid mixtures: A particular class. Powder T echnol. 190, 269-273. Dufaud, O ., Poupeau, M., Khalili, I., Cuervo, N., Christo doulou, M., Olcese, R., Dufour , A., Perrin, L., 201 1. C omparing Pyrolysis Gases and Dusts Explosi vities: A Clue to Understa nding Hybrid Mixtures Expl osions? Ind. Eng. Chem. Res. 51, 7656-7662. Dufaud, O., T raoré, M., Perrin, L., Chazelet, S., Thomas, D., 2010. Experimental investigation and modelling of aluminum dust s explosions in the 20 L sphere. J. Loss Prev . Process. Ind. 23, 226-236. Eckhof f, R.K., 1975. T owards Absolute Minimum Ignition Ener gies For Dust Clouds? Combust. and Fla me 24, 53-64. Eckhof f, R.K., 2003. D ust explosions in the process in dustries, in: Publishi ng, A.G.P . (Ed.). Eckhof f, R.K., 2005a. Current sta tus and expected fu ture tre nds in dust e xplosi on research. J. Loss Prev . Process. Ind. 18, 225-237. Eckhof f, R.K., 2005b. Gas and dust explosions caused by smolderin g combustion in powder layers and deposits. Explosion Hazards in t he Process In dustries. Eckhof f, R.K., 2016. Gas and Dust Explosions Cau sed by Smouldering Combustion in Powder Layers and Deposits. Explosion Hazards in t he Process In dustries, 2nd Edition. Eckhof f, R.K., 2019. Meas urement of minimum ignition ener gies (MIEs) of dust clouds – History , present, future. J. Loss Prev . Process. In d. 61, 14 7-159. Eckhof f, R.K., Enstad , G., 1976. Why Are "Long " Electric Sp arks More Ef fective Dust Explosion Initiators Than "Short" Ones? 27, 129-131. Engler , C., 1885. Beitra ge zur Kenntniss der Staubexplosionen Chemische Industrie, 17 1 -173. Escot Bocanegra, P ., Sarou-Kanian, V ., Davidenko, D., Chauvea u, C., Gökalp, I., 201 1. Studies on the burning of micro - and nanoaluminum parti cle clouds in air . 47-62. Foniok, R., 1985. Hybrid dispersi ve mixtures and inertized mixtures of coal dust. explosivenes s and ignitability . Staub Reinhalt. Luft 45, 151 -154. Frank-Kamenetskii, D.A., 19 69. Dif fusion and Heat Exchange in Chemic al Kinetics. Plenum pres s, New Y ork. Franke, H., 1978. Bestimmung der Minderstzude nergie von Kohlenstaub/Methan/Luft Gemisches (hybrid e Gemische). VDI-Berichte 304, 69 -72. Gan B., Li B., Jiang H., Zhang D., Bi M., Gao W . 2018a. Ethylene/polyethylene hybrid explosions: Part 1. Effects of ethylene concentrations on flame propagatio ns. J. Loss Prev . Process. Ind. 54, 93-102. Gan B., Li B., Jiang H., Zhang D., Bi M., Gao W . 2018b. Ethylene/polyethylene hybrid explosions: Part 2. Effects of polyethylene particl e size distributions on fla me propagations. J. Loss Prev . Process. Ind. 55, 134-143. Gao, W ., Dobashi, R., Mogi, T ., Sun, J., Shen, X., 2012. Ef fects of particle characteristics on flame propagatio n behavior during or ganic dust explosi ons in a half-closed chamber . J. Loss Prev . Process. Ind. 25, 993- 999. Gao, W ., Mogi, T ., Rong, J., Y u, J., Y an, X., Dobashi, R., 2015a. Motion behaviors of the unburned particles ahead of flame front in hexa decanol dust explosio n. Powder T echnol. 271, 125-133. Gao, W ., Mogi, T ., Y u, J., Y an, X., Sun, J., Dobashi, R., 2015b. Flame propagation mechani sms in dust e xplosions. J. Loss Prev . Process. Ind. 36 , 1 86 -194. Gao, W ., Zhang, X., Zhang, D., Peng, Q., Zhang, Q., Dobashi, R., 2017. Flame propagation beh aviours in nano-metal dust explosions. Powde r T echnol. 321, 154-162. Gao, W ., Z hong, S., Mogi, T ., Li u, H., Rong, J., Dobashi, R., 2013. Study on the influence of material thermal characteristics on dust explosion parameters of three l ong-chain monobasi c alcohols. J. Loss Prev . Process. Ind. 26, 186-196. Garcia-Agreda, A., Di Benedetto, A., Russo, P ., Salzano, E., Sanchirico, R., 201 1. Dust/gas mixtures explosion regimes. Powder T echnol. 205, 81-86. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 48 Garcia-T orrent, J., Conde-Lazaro, E., W iien, C., Rautalin, A., 1998. Biomass dust explosibili ty at elevated init ial pressure s. Fuel 77, 1093-1097. Glassman, I., 1996. Combustion, 3rd ed, Academic Press, San Diego, USA . Going, J.E., Chatrathi, K., Cash dollar , K.L., 2000. Flammability limit measurements for dusts in 20 -L and 1-m3 vessels. J. Loss Prev . Process. In d. 13, 209-219. Goroshin, S ., Bidabadi, M., Lee, J.H.S., 1996a. Quenching Distance of Laminar Flame in Aluminum Dust Clouds. Combust. and Flame 10 5, 147-160. Goroshin, S., Fomenko, I., Lee, J.H.S., 1996b. Burning velocities in fuel-rich aluminum dust clouds. T wenty-Sixth Symposium (Inter national) on Combustion 1966, 1961-1967. Haghiri, A., Bidabadi, M., 201 1. Dynamic behavior of particles across flame propagation through micro -iron dust cloud with thermal radiati on ef fect. Fuel 90, 2413-2421. Han, O.-S. , Y ashima, M., Matsuda, T ., Mats ui, H., Miyake, A., Ogawa, T ., 2000. Behavior of fla mes pro pagating th rough lycopodium dust clouds i n a vertical duct. J. Loss Pre v . Process. Ind. 13, 449–457. Hossain, M.N., Amyotte, P ., Abuswer , M., Dastidar , A., Khan, F ., Eckhoff, R., Chunmiao, Y ., 2014. Influence of liquid and vapourized sol vents on explosibility of pharmaceutical excipient dusts. Process Saf. Prog. 33, 374-379. Hosseinzadeh, S., Norman, F ., V erplaetsen, F ., Be rghma ns, J., V an den Bulck, E., 2015. Minimum i gnition energy of mixtures of combustible dusts. J. Loss Prev . Process. Ind. 36, 9 2-97. Hosseinzadeh, S., Norman, F ., V erplaetsen, F ., Ber ghmans, J., V an den Bulck, E., 2016. A study on the effects of using diffe rent ignition sources on explosion severity characteristics of coals in oxy-fuel atmospheres. J. Loss Prev . Process. Ind. 43, 5 3-60. Ichinose, K., Mogi, T ., Dobashi, R., 2020. Effects of the Particle Size and Agglome ration on the Minimum Explosible Concentration and Flame Propagation V elocity in Dust C louds. Combust. Explo. S hock+. 56, 421-426. Ji, W ., W ang, Y ., Y ang, J., He, J., W en, X., W ang, Y ., 2022. Me thods to predict variations of lower explosion limit associated with hybrid mixt ures of flammable gas and dust. Fuel 310, 122138. Ji, W ., Y u, J., Y u, X., Y an, X., 2018. Experimental investigation into the ve nted hybrid mixture explosions of lyc opodium dust and methane. J. L oss Prev . Process. Ind. 51, 102-1 1 1. Jiang, J., Liu, Y ., Mannan, M.S., 2014. A correlation of the lower flammability limit for hybrid mixtures. J. Loss Prev . Process. Ind. 32, 1 20-126. Jiang, J ., Liu, Y ., Mas huga, C.V ., Mannan, M.S., 20 15. V alidation of a new formula for pr edicti ng the lower flammability limit of hybrid mixtures. J. Loss Prev . Process. Ind. 35, 52-58. Jing, Q., W ang, D., Liu, Q., R en, L., W ang, Y ., Liu, C., Shen, Y ., W ang, Z., 2021. Study on transient reaction mecha nism and explosion intensity parameters of micron-sized fla ke aluminum dust in ai r . Chem. Eng. Sci. 246, 1 16884. Jing, Q., W ang, D., Liu, Q., Shi, C., Liu , J., 2022. Ignition sensitivity a nd expl osion be haviors of micron-sized aluminum powder: Compariso n between flake alumi num powder and spherical alu minum powder . Chem. Engi. Sci. 252. Jarosinski, J., Lee, J. H., Knystautas, R., Crowley , J.D. , 1988. Quenching of dust-air flames. Symposium (International) on Combusiton. 21 ( 1), 1917-1924. Ju, W . Dobashi, R., Hirano, T ., 1998. Reaction zone structures and propagation mechanisms of flames in stearic acid particle clouds. J. L oss Prev . Process. Ind. 1 1, 423-430. Julien, P ., S oo, M., Goroshin, S., Frost, D.L., Ber gthorson, J.M., Glumac, N., Zhang, F ., 2014. Combustion of Aluminum Suspensions in H ydrocarbon Flame Products. J. Propul. Poweer . 30, 1047-1054. Julien, P ., Whiteley , S., Goro shin, S., So o, M.J., Frost, D.L., Ber gthorson, J.M., 2015a. Flame structure and particle - combustion regimes i n premixed methane–iron –air suspensions. Proc. Comb ust. Inst. 35, 2431-2438. Julien, P ., V ickery , J., Goroshin, S., Frost, D.L., Bergt horson, J.M., 2015 b. Free ly -propagating flames in aluminum dust clouds. Combust. an d Flame 162, 4241-4253. Julien, P ., V ickery , J., Whitel ey , S., W right, A., Goroshin, S., Bergthorso n, J.M., Frost, D.L., 2015 c. Ef fect of scale on freely propagating flames in aluminum dust clouds. J. Loss Prev . Process. Ind. 36, 230- 236. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 49 Kern, H., W ieser , G.J., Raupen strauch, H., 2015. Flame propagation in lycopodium/air mixtures below atmospheric pressure. J. Loss Prev . Process. I nd. 36, 281-286. Khalil, Y .F ., 2013. Experimental investigation of the complex deflagration phenomena of hybrid mixtures of activated carbon dust/hydro gen/air . J. L oss Prev . Process. Ind. 26, 1027-1038. Khalili, I., Dufaud, O., P oupeau, M., Cuervo-Rodriguez, N., Perrin, L ., 2012. Ignition sensitivit y of gas–vapor/dust hybrid mixtures. Powder T echnol. 217, 199-206. Kim, W ., Soga, T ., Johzaki, T ., Endo, T ., Kato, T ., Choi, K., 2019. Minimum ignition ener gy and minimum explosible concentration of L-isoleuci ne and glycine powder . Powde r T echnol. 347, 207-214. Krause, U., Kasch, T ., 2000. The influence of flow an d turbulence on flame propagation through dust-air mixtures. J . Loss Prev . Process. Ind. 13, 291-298. Krietsch, A., Reyes Rodriguez, M., Kristen, A., Kadoke, D., Abbas, Z., Krause, U., 2021. Ignition temperatures and flame velocities of met allic nanomaterials. J. Loss Prev . Process. Ind. 71, 104482. Krietsch, A., Scheid, M., Schmidt, M., Krause, U., 2015 . Explosion behaviour of metallic nano powders. J. Loss Prev . Process. Ind. 36, 2 37-243. Kuai, N., Huang, W ., Du, B., Y uan, J., Li, Z., Gan, Y ., T an, J., 2013. Experiment-based investigations on the effect of ignition ener gy on dust explosion behaviors. J. Loss Pre v . Process. Ind. 26, 869-877. Kuai, N., Li, J., Chen, Z., Huang, W ., Y uan, J., Xu, W ., 201 1. Experi ment -based investigations of magnesiu m dust explosion characte ristics. J. Loss Prev . Process. Ind. 24, 302-313. Kundu, S.K., Zanganeh, J., Eschebach, D., Moghtaderi, B., 2018. Explosio n seve rity of methane –coal dust hybri d mixtures in a ducted sp herical vessel. Powder T echnol. 323, 95-102. La´zaro, E .C., T orrent, J. G.a., 20 00. Experi mental research on explosibility at high initial pressures of combustible dusts. J. Loss Prev . Process. In d. 13, 221-228. Li, G., Y uan, C., Fu, Y ., Zhong, Y ., Chen, B., 2009. Inerting of magnesium dust cloud with Ar , N 2 and CO 2 . J Hazard Mater 170, 180-183. Li, H., Chen, X., Deng, J., Shu, C., Kuo, C., Y u, Y ., Hu, X., 2020a. CFD analysis a nd experimental study o n the ef fect of oxygen level, particle size, and dust concentration on the flame evolution characteristics and explosion severity of cornstarch dust clou d deflagration in a spherical c hamber . Powder T echnol. 372, 585-599. Li, H., Deng, J., Chen, X., Shu, C., Kuo, C. , Hu, X., 2020b. Influence of ign ition delay on explosion severities of the methane–coal particl e hybrid mixture at elev ated injection pressures. Pow der T echnol. 367, 860-876. Li, H., Deng, J., Chen, X., Shu, C., Kuo, C. , Zhai, X. , W ang, Q., Hu, X., 2020c. Qualitative and quantitative characterisation for explosion severity and gaseous –solid residues during methane–coal particle hybrid explosions: An approach to estimating the safety degree for undergro und coal mines. Process Saf. Environ. Protect. 141, 150-166. Li, H., Deng, J., Chen, X., Shu, C ., Kuo, C. , Zhai, X., W ang, Q., Hu, X., 2020d. T ransient temperature evolution of pulverized coal cloud deflagration in a m etha ne –oxygen atmosphe re. Powder T echnol. 366, 294-304. Li, H., Deng, J., Shu, C., Kuo, C. , Y u, Y ., Hu, X. , 2020e. Flame behaviours and deflagration severities of aluminium powder–air mixture i n a 20-L sphere: Computational fluid dynamics modelli ng and experimental vali dation. Fuel 276, 1 18028. Li, Q., Lin, B., Dai, H., Zhao, S., 2012. Explosion character istics of H 2 /CH 4 /air and CH4/coal dust/air mixtures. Powder T echnol. 229, 222-22 8. Li, Q., T ao, Q., Y uan, C., Zheng, Y ., Zhang, G., Liu, J., 2018a. Investig ati on on the structure evolution of pre and post explosion of c oal dust using X-ray dif fraction. Int. J. Heat Mass T ran. 120, 1 162-1 172. Li, Q., W ang, K., Zheng, Y ., Mei, X., Lin, B., 2016. Explosion se verity of micro-sized aluminum dust and its flame propagation properties in 2 0 L spherical vessel. Powder T echnol. 301, 1299-1308. Li, Q., Y uan, C., T ao, Q., Zheng, Y ., Zhao, Y ., 2018b. Experimental analysis on post -explosion residues for evaluating coal dust explosion seve rity and flame propagation be haviors. Fuel 215, 417- 428. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 50 Liu, S., Cheng, Y ., Meng, X., Ma, H., Song, S., Liu, W ., Shen, Z. , 2018. Influence of particle size polydispersity on coal dust explosibility . J. Loss Prev . P rocess. Ind. 56, 444-450. Liu, Y ., Sun, J., Chen, D., 20 07. Flame propagation in hybrid mixt ure of coal dust and methane. J. Loss Prev . Process. Ind. 20, 691-697. Man, C.K., Gibbins, J.R., 201 1. Factors af fecting coal particle ignition under oxyfuel combustion atmospheres. Fuel 90, 294-304. Marmo, L., Cavallero, D., Debernardi, M.L., 2004. Aluminium dust explosion risk analysis in metal workings. J. Loss Prev . Process. Ind. 17, 449-465. Moroń, W ., Ferens, W ., Czajka, K.M., 2016. Explosion of dif ferent ranks coal dust in oxy -fuel atmosphere. Fuel Processing T echnol. 148, 388 -394. Murillo, C., Dufaud, O. , Bardin-Monnier , N., López, O., Munoz, F ., Perrin, L., 2013. Dust explosions: CFD m odeling as a tool to characte rize the relevant paramete rs of the dust dispersion. Chem. E ng. Sci. 10 4, 10 3-1 16. NFP A-69, 2014. Standard on Explosion Preventii on Systems. National Fir e Protection Agency , Quincy , MA. Nifuku, M., Koyanaka, S., Ohya, H., Barre, C., Hatori, M., Fujiwara, S., Horiguchi, S., Sochet, I., 2007. Ignitability characteristics of aluminium and magnesium dusts that are generated during the shredding of post -consumer wastes. J. Loss Prev . Process. Ind. 20, 322-329. Nomura, S., T anaka, T ., 1992. Theoretical analysis of d ust explosions Powder T echnol. 71, 189-196. Norman, F ., Hosseinzadeh, S., Bulck, E.V ., Ber ghmans, J., V erplaetsen, F ., 2015. Dust explosion severity characteristics of Indonesian Sebuku c oal in oxy-fuel atmosphe res. 7th European Combu stion Meeting (ECM 2015). Palecka, J., Julien, P ., Goroshin, S., Ber gthorson, J.M., Frost, D.L., Higgins, A.J., 2015. Quenching dist ance of flames in hybrid methane–alumi num mixtures. Pr oc. Combust. Inst. 35, 2463-24 70. Pang L., Cao J., Zhao Y ., Y uan C. Y ang K., Zhang Z., 2021. Minimum ign ition ener gy of LDPE dust/ethylene hybrid mixture. J. Loss Prev . Process. In d. 72, 104546. Pellmont, G., 1979. Explosions-und Zündve rhalten von hybriden Gemisc hen aus brennbaren Stäuben und Brenngase n. ETH Zurich. Pellmont, G., 1980. Minimum ignition energy of combus tible dusts and explosion behaviour of hybri d mixtures, 3rd International Symposi um on Loss Prevention a nd Safety Promotion in t he Process Industries, pp. 851-862. Pico, P ., Ratkovich, N., Muñoz, F ., Dufaud, O., 2020a. Analysis of t he explosion behaviour of wheat starch/pyrolysis gases hybrid mixtures through experimentation and C FD-DPM simulations. Powde r T echnol. 374, 330-347. Pico, P ., Ratkovich, N., Muñoz, F ., Dufaud, O., 2020b. CFD-DPM and exp erime ntal study of the dynamics of wh eat starch powder/pyrolysis gases hybrid mixtures in t he 20-L Sphere. Powder T echnol. 372, 638-658. Pilao, R., Ramalho, E. , Pinho, C., 2006. Overall characterization of cork dust e xplosion. J Hazard Mat er 133, 183-195. Proust, C., Accorsi, A., Dupont, L., 2007. Measuring the violence of dust explosions with the “20l sphere” and with the standard “ISO 1m3 vessel”. J. Loss Prev . Process. Ind. 20, 599-606. Pu, Y .K., Jia, F ., W ang, S., Skjold, T ., 2007. Determination of the maximum effective burning velocity of dust–air mixt ures in constant volume c ombustion. J. Loss Prev . Process. Ind. 20, 462-469. Radford, J., 2015. "What happened tr agic day of Mt Kembla Mi ne disaater " (Ⅰ llawarra Mercury website,). http://www .illawarramecury .com.au/sto ry/2441786/what-happened-tragic- day- of - mt -kembla -mine-disaster/) (Accessed: 14 August 2015). Redlinger , R., 2015. Numeric al simulation of hybrid dust/gas expl osion experiments in the standar d 20 -L sphere. Fusion Eng. Des. 100, 419- 424. Russo, P ., Bene dettob, A.D., Sanchirico, R ., 2012. T heoretical Evaluation of the E xplosion Regimes of Hybri d Mixtures. Chemical Engineeri ng T ransa ctions 26, 51-56. Salamonowicz, Z., Kotowski, M., Półka , M., Barnat, W ., 2015. Numerical simulation of dust explosion in t he s pherical 20l vessel. Bulleti n of the Polish Academy of Sciences T echnical Scie nces 63, 289-293. Sanchirico, R., Di Benedetto, A., Garcia -Agreda, A., Russo, P ., 201 1. Study of the severity of hybr id mixture explosions Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 51 and comparison to pure dust–air and vapour–air explosio ns. J. Loss Prev . Process. Ind. 24, 6 48-655. Sanchirico, R., Russo, P ., Saliva A., Doussot, A., Di Sarli D, V ., Di Benedetto, A., 2015a. Explosion of lycopodium - nicotinic acid methane complex hybrid mixt ures. J. Loss Prev . Process. Ind. 36, 505- 508. Sanchirico, R., Russo, P ., Di Sarli, V ., Di Benedetto, A., 2015b. On the explosion and flammability behavior of mixtures of combustible dusts. Process Saf. Environ. Protect . 94, 410-419. Singer , J.M., Liebman, I., 1968. Spherical Flames of Spark-ignited Dust Clouds. Central Electricity Generating Board 12. Siwek, R., 1996. Determination of technical safety indices and factors influencing hazard evaluation of dusts. J . Loss Prev . Process. Ind. 9, 21-31. Skjold, T ., Olsen, K.L., Castellanos, D., 2013. A constant pressure dust explosi on experiment. J. Loss Prev . Process. Ind. 26, 562-570. Slezak, S.E., Buckius, R.O., Krier , H., 1985. A Model of Flame Propagation in Rich Mixtures of Coal Dust in Air . Combust. and Flame 59, 25 1-265. Song, S., Cheng, Y ., Meng, X., Ma, H., Dai, H., Kan, J. , Shen, Z., 2019. Hybrid CH4/coal dust explosions in a 20- L spherical vesse l. Process Saf. Environ. Prote ct. 122, 281-287. Soo, M., Julien, P ., Goroshin, S., Bergtho rson, J.M., Frost, D.L., 2013. Stabilized flame s in hybrid aluminum-methane- air mixtures. Proc. Comb ust. Inst. 34, 2213-2220. Soo, M., Kumashiro, K., Goroshin, S., Frost, D.L., Bergthorson, J.M., 2016. Thermal structure and burning velocity of flames in non-volatile fuel s uspensions. Proc. Combust. Inst. 36, 23 51-2358. Sun, J., Dobashi, R., Hirano, T ., 1998a. Structure of flames propagating through metal particle clouds and behavior of particles. T wenty-Seventh Symp osium (International) o n Combustion/The Combustion Institute, 2405-241 1. Sun, J., Dobashi, R., Hirano, T ., 2001a. T emperature profile across the combustion zone propagatin g through an iron particle cloud. J. L oss Prev . Process. Ind. 14, 463-467. Sun, J ., Dobashi, R., Hirano, T ., 2006a. V elocit y and number density profiles of particles across upward and downward flame propagating t hrough iron particle clo uds. J. Loss Prev . Process. Ind. 19, 135-141. Sun, J. , Dobashi, R., Hirano, T ., 2006b. Structure of flames propagating through aluminum parti cles clou d and combustion process of partic les. J. Loss Prev . Process. Ind. 19, 769-773. Sun, J., Dobashi, R., Hirano, T ., 1 998b. Structure of flames propagating through metal particle clouds and behavior of particles. T wenty-Seventh Symp osium (International) o n Combustion/The Combustion Institute, 2405–241 1. Sun, J., Dobashi, R., Hirano, T ., 2001b. T emperature profile across the combustion z one propagatin g an iron particle cloud. J. Loss Prev . Process. In d. 14, 463–467. Sun, J., Dobashi, R., Hirano, T ., 2003. Concentrati on profile of particles across a flam e propagating through an iron particle cloud. Com bust. and Flame 134, 381-387. Sweis, F .K., 2006. T he effect of admixed material on the minimum explosible concentration of oil shale. J. Loss Prev . Process. Ind. 19, 7 01-704. T an, X., Schmidt, M., Zhao, P ., W ei, A., H uang, W ., Qian, X., W u, D., 2020. Mi nimum ignition temperature of carbonaceous dust clo uds in air with CH 4 /H 2 /CO below the gas lower explosion limit . Fuel 264, 1 1681 1. T ang, F .-D., Goroshin, S ., Higgins, A., Lee, J. , 20 09. Fla me propa gation and quenching in iron dust cl ouds. Proc. Combust . Inst. 32, 1905-1912. T ang, F .-D., Goroshin, S., Higgins, A.J., 201 1. Modes of particle combustion in iron dust flames. Proc. Combust. Inst. 33, 1975-1982. Thimothée, R., Chauveau, C ., Halter , F ., Gökalp, I., 2016. Experimental Investigation of the Mechanisms of Cellular Instabilities Develo ping on Spherical T wo-Phase Flames. Combust. Sci. T echnol. 188, 2026-2043. T orrado, D., Cuervo, N., Paca ult, S., Glaude, P. -A., Dufaud, O. , 2017. Influence of carbon black nanopart icles on the front flame velocity of metha ne/air explosions. J. Loss Pre v . Process. Ind. 49, 919-928. T sai, Y ., Ho, S., Huang, A., Shu, C. , 2018. Potential explosion hazard of polyester resin dust formed fr om a granula tion process: Limiting ox ygen concentration with dif fer ent pressures . A ppl. Therm. Eng. 135, 74- 82. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 52 V an Laar , G. , 1983. Infl uence of Moisture Content on the Minimum Ignition Ener gy (MIE) of Dust/Air Mixtures. Report on the W orking Group Minimum Igniti on Energy . TNO Prins Mauti ts Laboratorium. V ickery , J., Julien, P ., Goroshin, S., Ber gthorson, J.M., Frost, D.L., 2017. Propagation of isobaric spherical flames in hybrid aluminum-met hane fuel mixtures. J. Loss Pre v . Process. Ind. 49, 472-480. W ang, D., Qian, X. , W u, D., Ji, T ., Z hang, Q., Huang, P ., 2020a. Numerical study on hydrodynamics and expl osion haza rds of corn starch at hig h-temperature environm ents. Powder T echnol. 360, 1067 -1078. W ang, Q., Sun, Y ., Jiang, J., Deng, J., Shu, C., Luo, Z., W ang, Q., 2020b. Inhibitin g ef fects of gas–particle mixtures containing CO2, Mg(OH)2 particles, and NH4H2PO4 part icles on methane explosi on i n a 20-L closed vessel. J. Loss Prev . Process. Ind. 64. W ang, S., Pu, Y ., Jia , F ., Gutkows ki, A., Jarosinski, J ., 2006. An Experimental Study on Flame Propa gation in Cornstarch Dust Clouds. Combust . Sci. T echnol. 178, 1957-1975. W ang, Y ., Qi, Y ., Gan, X., Pei, B., W en, X., Ji, W ., 2020c. Influences of coal du st components on the explosibility of hybrid mixtures of met hane and coal dust. J. Loss Prev . Process. Ind. 67, 104222. W u, D., Huang, X., Norman, F ., V erplaetsen, F ., Ber ghma ns, J., V an den Bulck, E., 2015. Experimental investigation on the self-ignition behaviou r of coal dust accumulat ions in oxy-fuel combustion system . Fuel 160, 245-254. W u, D., Kriets ch, A., Schmidt, M., Krause, U., 2022. Ef fect of oxygen concentration, ine rt gas and CH 4 /H 2 addition on minimum ignit ion ener gy of coal dusts. J. Loss Prev . Proces s. Ind. 77 , 104772. W u, D., Norman, F ., V anierschot, M., V erplaetsen, F ., Berghma ns, J., 2018. Self -heating characteristics of coal dust deposits by a hot gas fl ow in oxy-fuel atmosphe res. Appl. Therm. Eng. 131, 947-95 7. W u, D., Norman, F ., V erplaetsen, F ., V an den Bulck, E., 2016a. Experimental study on the minimum ig nition te mperature of coal dust clouds in o xy-fuel combustion at mospheres. J Hazard Mater 307, 274-280. W u, D., V anierschot, M., V erplaetsen , F ., Berghmans, J., V an den Bulck, E., 2016b. Numerical study on the ignition behavior of coal dust layer s in air and O2/CO2 at mospheres. Appl. Them. Eng. 109, 709-717. W u, D., Norman, F ., Schmidt, M., V anierschot, M., V erplaetsen, F ., Ber ghmans, J., V an den Bulck, E., 2017a. Numerical investigation on the self-ignition behaviour of coal dust accumulations: The roles of oxygen, diluent gas and dust volume. Fuel 188, 500-5 10. W u, D., Schmidt, M., Hua ng, X., V erplaetse n, F ., 2017b. Self-ignition and smoldering characteristics of coal dust accumulations in O2/N 2 and O2/CO2 atmos pheres. Proc. Combust. Inst. 36, 3195-3202. W u, D., Son g, Z., Schmidt, M., Zhang, Q., Q ian, X., 2019. Theoretical and numerical stu dy on ignition behaviour of c oal dust layers on a hot s urface with correcte d kinetic parameters. J Hazar d Mater 368, 156-162. W u, D., T an, X., W ei, A., Dua n, Q., Huang, W ., Schmidt, M., 2021a. Ignit ion temperatu re and mechanism of carbonaceous dust clouds: The roles of volatile matter , CH4 addition, O2 mo le fraction and diluent gas. J Hazard Mater 405, 124189. W u, W ., W ei, A., Huang, W ., Zhao, P ., Schmidt, M., Krause, U., W u, D., 2021b. Experimental and theoretical study on the inh ibiti on ef fect of CO2/N2 blends on the ignitio n behavior of carbonaceous dust clouds. Process Saf. Environ. Protect. 153, 1-10. Y etter , R.A., Risha, G.A., Son, S.F ., 2009. Metal particle combustion and nanotechnology . Proc. Combust. Inst. 32, 1819-1838. Y u, J., Zhang, X., Z hang, Q., W ang, L., Ji, K., Peng, L., Ga o, W ., 2016. Combust ion beha viors and flame microstruct ures of micro- and nano-ti tanium dust explosions. Fuel 181, 785-792. Y u, X., Y u, J., W ang, C., Lv , X., W ang, Y ., Hou, Y ., Y an, X., 2021. Experimental study on the overpressure and flame propagation of hybri d hydrogen/aluminum dust explosions in a square clos ed vessel. Fuel 285, 1 19222. Y u, X., Y u, J., Zhang, X., Ji, W ., L v , X., Hou, Y ., Li, Z., Y an, X., 2020. C ombustion be haviors and residues characteri stics in hydrogen/alumin um dust hybrid explosi ons. Process Saf. Environ. Prot ect. 134, 343-352. Y u, X., Y u, J., Zhang, X., Ji, W ., Y an, X., 2019. Combustion behaviors and residues characteristics in hydrogen/aluminum dust hybrid eplosio ns. Process Saf. Environ. Protect . 134, 343-352. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 53 Y uan, J., Hu ang, W ., Du, B., Kuai, N., Li, Z., T an, J., 2012a. An Extensive Discussion on Experimental T est of Dust Minimum Explosible C oncentration. Procedi a Eng. 43, 343-347. Y uan, J., Hua ng, W ., Ji, H., Kuai, N., W u, Y ., 2012b. Experimental investigation of dust MEC measurement. Powder T echnol. 217, 245-25 1. Y uzuriha, Y ., Gao, W ., Mogi, T ., Dobashi, R., 2017. Ef fects of particle size distributions on flame propagation behavior through dust clouds of PMM A. J. Loss Prev . Process. Ind. 49, 852-858. Zhang, H., Chen, X., Zhang, Y ., Niu, Y ., Y uan, B., Dai, H., He, S ., 2017a. Effects of particle size on flame structures through corn starch d ust explosions. J. Loss P rev . Process. Ind. 50, 7-14. Zhang, Q., Li u, L., Shen, S., 2018. Ef fect of turbulence on explosion of al uminum dust at vari ous c oncentrations in air . Powder T echnol. 325, 467-47 5. Zhang, X. , Y u, J., Gao, W ., Zhang, D. , S un, J., Guo, S., Doba shi, R., 2017b. Ef fects of particle size distributions on PMMA dust flame propagatio n behaviors. Powder T echnol. 317, 197-208. Zhang, X., Y u, J., Y an, X., Xie, Q. , Gao, W ., 2016. Flame propagation behaviors of na no- and mic ro-scale PMMA dust explosions. J. Loss Pre v . Process. Ind. 40, 101-1 1 1. Zhang, Y ., Zha o, J., Ma, Z. , Y ang, F ., Cheng, F ., 2019. Ef fect of oxygen c on centrati on on oxy-fuel combustion characteristic and interacti ons of coal gangue and pine sawdust. W aste Manag e. 87, 288-294. Zhao, F ., Rogers, W .J., Sam Mannan, M., 2009. Experimental measurement and numerical analysis of binary hydrocarbon mixture flammabili ty limits. Process Saf. En viron. Protect. 87, 94- 104. Zhao, P ., Schmidt, M., Krause, U., Duan, Q., Krietsch, A., W u, D., 2021. Ex perimental study on the minimum explosion concentration of anthracite dust: The roles of O2 mole fraction, inert gas and CH 4 addition. J. Loss Prev . Process. Ind. 71, 104490. Zhao, P ., T an, X. , Schmidt, M., W ei, A., Huang, W ., Qian, X., W u, D., 2020. Minimum explosion concentration of coal dusts in air with s mall amount of CH 4 /H 2 /CO unde r 10-kJ ignition ener gy conditions. Fuel 260, 1 16 401. Zhen, G., Leuckel, W ., 1997. Effects of ignitors and turbulence on dust. J. Loss Prev . Process. Ind. 10, 317- 324. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 54 14th I nt ernat ion al Sym posi um o n H azards , Pre venti on, and Mit iga tion of I ndus trial Expl osio ns Brauns c hwei g, GE RMANY - July 11 - 15, 2022 Flameproof enclosure and pressure relief: Status quo S. Herbst a , J . Gerlach a ; L. Sieg le a ; F. Engelmann a & T. A rnhold b a Ernst Ab be Universi t y Jena (I NNOK Researc h Group, Jena, Germ any) b R. STAHL Ak t iengesellsc ha ft (VP Technolog y, Waldenburg, G ermany ) E- m ail: sabr ina.herbst@ea h- jena.de Abs tract The purpose of the ignition protection t y pe flameproof enclosure is to contain th e permissible explosion ins ide and prevent its transfer to the surrounding explosive atmosphere. The exothermic reactio n r equir es th at chem ical en er g y can b e convert ed i nto t hermal an d mech anica l en erg y. Consequently, the flameproof enclosure must be designed for the possibl e increase in temperature and pressure th at may r esult from this sudden oxidation/degradation reaction. This is realised by designing the housing in ac cordance with the stress es that may a rise, taking into a ccount a max. 4 - fold safet y facto r and flameproof joints. This principle of ac tion is questioned by the E rnst -Abbe- University of Applied Sciences Jena and the comp an y R. Stah l Sch al tger äte GmbH. Pressure re lief is intende d to m inimise the therma l and mechan ical ener g y inside the flameproof en closure caused b y th e ex pl osio n. Pr essure r eli ef, th ere fore, has the function of an energy store, among other th ings. Consequently, the load a cting on the housing is reduced a nd the temperature of the ga s flowing out of the pressure r elief is lowered to a pe rmissible value. Accordingly , th e overa ll requirements for flameproof enclosures ch ange and an en clo sure fo r this ignition protection type can be des igned in a more m aterial - s a v i n g w a y. This approach has be en st udi ed ex tensi vel y as p art of sever al res earch p ro jects. After s ucces sful comple tion of the se basic tests, the deter mination of inf luencing pa rameters and their effect on small enclosures f ollow ed with the help of differe nt test series. The knowled ge gained wa s tr ansferre d to larger e nclosures and the effect investigated through further test ser ies, in order to develop a marketable product based on this knowledg e. Ke y words: flameproof en closure, pressure relief, product innovation 1. Intr oduct ion Companies are c onstant ly required to adapt their products and busine ss models to future developments, in order to maintain their competitiveness. The megatrends currentl y i nfluencing tomorrow’s world and the environment include digitalisation, g lobalisation, resource scarcit y an d climate protection, including decarbonisation, EY (2020). In the area of unit goods produc tion, the megatrends of globa lisati on, a lack of r esources a nd climate protec tion have alr ead y b een havi ng a n increasing impac t for y e ars through a w ide variet y of influences. This industr y is characterised by high mater i al, production and transport costs. In the field of ex plosion protection, a sub - ar ea of s afet y t echn olog y, a v er y high p rofi le o f requireme nts must be fulfilled b y t he products. For ex ample, the products are used in a wide r ange of industries, from harsh industrial and mar itime environments on drilling pla tforms, to cleanrooms in the pharmaceutical industry . This leads to a very wide range of environmental conditions – from the Arc tic cold in Siberia to san dstorms in the Middle East. Likewis e, requir ements of restr ictive Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 55 leg islation a nd stand ardisation must be taken int o account, which may h ave na tional and regional differen ces . A globally proven technical solution principle for enabling equipment for use in potentiall y explosive atmo spheres is t he fl ameproof enclosure t y p e of p rotection, Wit tler, M. (2005). This t y pe of i gnition protection has existed in its basic oper ating principle since the beg innin g of the 20th ce ntury until today. I n order to adapt to today ’s me ga trend requirements, this principle can be challenged, in order to develop a r esource-saving flameproof enclosure with r educed material use and a low wei ght. Within t he fr amework of several research projects, t his active principle was questioned b y the Erns t - Abbe-University of Applied Science s J ena a nd the compan y R. Stahl Schaltg erät e GmbH 2. Flam eproof encl osu re The task of the ignition protection t y pe flamepro of enclosure is to c ontain a permissib le explosio n inside and preve nt its transfer to the surrounding explosive atmosphere. This ena bles non-explosion- proof equipment to be qualified for use in a potentiall y ex plosive atmosphere. The non-explosi on- proof equipment is inserted into the flameproof enclosure and installed with i t in the hazardous area. If, for ex am ple, a ho t su rface – caus ed by a malfunction in the non -explosion-pr otec ted equipment – ignites the explosive ga s atmos phere , the explosion is contained by th e flameproof enclosure and transmission to the outside is prevente d. The ex othermic re action of the explosion means that chemi cal en erg y can be c onvert ed in to th ermal an d m echanical en erg y. Con sequen tl y , th e flameproof enclosure must be designed f or the possible increase in te mpera ture and pressur e that may result fr om this sudden oxi dation/de gradation reaction. This is realised b y the following functional elements, the design of which is recommended in accordance with the specifications of the IEC 60079 - 1 standard (DIN 2015): • Appropriate de si gn of t he enc losure according to the stresses that may arise, taking i nto account a m ax. 4 - fold s afet y factor • F lame proof joints • Appro priat el y designed a ccessori es su ch as fas ten ers, con nect ors, et c. In the field of electric al explosion protection, this ignition protec tion t y pe is often used for switchgears, c ontrol and display devices, controls, mot ors, transfor mers, heaters and l uminaires. I n non- electr i cal explosion protection, the use of this ignition protection t y pe enabl es th e use of frict ion clut ches, b rakes , cat al y ti c conv ert ers and cartri dg e heate rs in poten tial l y e x plo sive at mos pheres. The require ments for the ignition pr otect ion t y p e flameproof enclosure necessit ate a ver y materi al - intensive construction of t he enc losures in which the equipment, which has a potential ignition source, is used. The conseque nces are high material a nd m anufac turing costs, which lead to price -intensive products. Currently commerciall y available a luminium flameproof enclosures with a volume of about two litres have a w eight o f about four kilograms, such as the enclosure of the 8265 se ries, size 2 from R. Stahl AG, Stahl (2022) or the enclosure of t ype 07 - 4C, GUB 0 A0A0 from Bartec Top Holding GmbH, Bar t ex (2022). The deve lopment of the flameproof enclosures is also characterised by the elabo rate tes tin g and cert ifi catio n proces ses. C omb ined, these fact ors l ead to ex ten siv e challen ges in the development of new innovative produc ts. 3. Pressu re rel ief 3.1 Pressure relief in technology Pres sure is a forc e actin g on a su rface. T he res ul t i s mechan ical t ensio n in t he bod y. Reli ef occu rs when the re duction in pressure takes place. Ac cordi ng to the phy sical princi ples, this can be r ealised by minimising th e force , as well as increasing the s u rface ar ea. As a resu lt, th e mechani cal t ensi ons crea t ed in the body are minimised. This relationship forms the phy sical basis of the general function of pressure re lief . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 56 From a technical discipli nary perspective, there are a lar ge number of appli cations of the pressure relief function, which are, however, based on different types of function and operating principles. On the one hand, a distinction must be made be tween ac tive and passive function fulfilment. An activ atin g and activ e me chani sm rep res ents a ctive function fulfilment. I f the function carrier can ge n era t e the required function throug h its geome try and the materials used, the function is fulfilled passively. O n the other hand, the active principles used to realise the pre ssure relie f function di ffer. Mos t press ure reli ef s y s t ems achi eve th e functi o n b y inc reasi ng th e avail abl e volu me and s urface area . Another option is the conversion of the act ing e ner gy, as the example of flameless pressure relief demon strat es. The ex i sti ng mechani cal and therm al ener g y is t rans fer red an d abso rbed b y a flame filter . This results in a minimisation o f the pr essure - causin g for ce. Table 1 : Operating principles of the pressure relief function i n technical system s Techni cal syst e m wit h the pr essure relief functi on Active (A)/ Passi ve (P) functi on carrier Principle o f act ion: Volu m e enlarg e m e nt Principle o f action:  En largem ent of the lo aded surface Principle o f action:  Heat transfer/ conver sion of existing energy Unli mited volum e L i mi te d volume Ball valve with pressure relief A ● Safe ty va lve s A ● ● Rupt ur e d iscs / B ur sti ng plu gs A ● ● Pressure relief flaps A ● ● Explosio n flap A ● ● Pressure relief do m e A ● ● Pressure relief louvre A ● ● Fla meless p ress ure relief A/ P ● ● ● ● Table 1 shows that the func tion of pressure relief is primarily realised by integrating a dditional components actively into machines and sy stems. A possi ble integ ration of the pressure relief function as a passive f un ction in existing eleme nts of the machine s and sy stems minimises the o rigin ally requi red eff ort and can m ai ntain the ex is ting in stal lati on sp ace as wel l as re duce i t if neces sar y 3.2 Flameproof enclosure: Integration pressure relief The pr essure relie f fun ction o ffers m an y innovativ e approaches in the field of explosion protection. In particular , the ignition protection t y pe flame proof enclosure har bours ex tensive optimisation potential, which can, f or example, induce a reduction in the necessary use of materia ls. I n order to withstand the pressure c aused b y an explos ion, massive w all thicknesses are r equired, among other thi ngs. If the c ausat ive p aramet ers of an exp losi on and its consequ ences c an be i nflu enced at t he si te of explosion generation by relieving the pressure , i t might b e possible to eff ectively minimise the forces an d he at that occur . As a res ult , a flame proof enclosure would have to meet lowe r requirements and can b e des i gned in a materi al -saving wa y. This approach has already be en looked at in more detail with differ ent ideas and approache s. Among other things, pa tents describing different wa y s of relieving pressure should be highlighted. Hornig (2013) provides the f irst design guide lines for the design of flameproof enclosures with pressure relief openings ca use d b y porous materia ls. In the context of this investig ation, mater ials mad e of ta pe winding, sintere d p article bulk and sintered short fibres were a vailable. The overall objective of the research projects pres ented he re is to gain knowledge on th e desi gn o f pass ive pres sure r elief el em ents for the ignition pr otection t yp e flameproof enclosure . The passive Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 57 Referen ces Bar t ec (2022). Druckfeste Gehäuse GUB Ex d IIC, Alumi nium & Edelstahl. https://www.bartec.de/ProdCatalogue /Ass ets/Datasheets/lng_1/413607_GUB_d.pdf Bund (2022 ). BUN DESREP UBLIK DEU TSCH L AND, VER TR ETEN DURCH D AS BUNDESMIN I S TER IUM FÜR W I RTS CHAFT, DIESES W I EDERUM VERTR ETEN DURCH DEN PRÄS IDENTEN DER PHYS I KA LISC H - TECHNI SC HEN BUNDESA NSTALT BRAUNSCHWEIG UND BERLIN, 38116 BRAUNSCHWE I G, DE. Verfahren zur Führung von im lnnern eines Gehäuses gebildeten Explosionsgasen durch zünddurchschlagsicher ausgebildete Spalte und explosionsgeschütztes Gehäuse. Er finder: U. KLAUSMEYER. Anmeldung: 17. J uni 1998. Deutschland. DE 198 26 911 A1 EY (2020). Are y ou refra ming your future or is the future reframin g you?: Understanding megatrends will help y ou see opportunities wh ere others don’t. EYGM Limited. https://asse ts.ey .com/content/da m/ey - s ites/ey - com /en_gl/to pics/mega trends/ey - me gatrend s - 2020.pdf DIN (2015). D I N Deuts ches Institut für Normung e. V. D IN EN 6007 9 - 1, Explosionsgefährdete Berei che – Te il 1: Geräte schutz durch druckfeste Kapselung „d“ (IEC 60079 - 1:2014); Deutsche Fassung EN 60079- 1:201 4, April 2015. Herbst, S., Enge lmann, F., Grote, K. - H. (2017) : Der Einsatz von D ruckentlastung im Explosi onssc hutz – eine innovative Basis für den Leichtbau. Vortrag und Beitrag beim 15. Gemeinsames Kolloquium Konstruktionstechnik 2017. Universität Duisburg - Es sen vom 05. bis 06. Oktober 2017 in Duisburg Herbst, S. (2018). Druckentlastung im Explosionsschutz, Magdeburg: Otto -von- Gueri cke - Universität Magdeburg, Diss., Hornig, J. (2013). Perme able Werkstoffe zur zün ddurchsc hlagsicheren Ex plosionsdruckentlastung. G estaltung und Anwendungsmöglichkeiten. Zugl.: Magdeburg, Univ., Fak. für Maschinenbau, Diss., 2012. Aachen: Shaker, Fortschritte in der Maschinenkonstruktion . I S BN 978 -3-8440-1835- 6. Stahl (2015) Ex d Gehäuse aus Leichtmetall, „Druckfeste Kapselung“. Reihe 8265, https://r- stahl.com/uploads/tx _ai meos/Files/i_/de/8265_Empt y Enclosures_EK00_ III _de.pdf Stahl (2022). Ex d Gehäuse aus Leichtme tall reihe 8265 Art. Nr. 165781 . https://r- stahl.com/de/global/produkte/kabel-und- leitun gse i nfuehrunge n-komponent en-und- ge h aeuse/ le ergeh aeus e/e x -d- aluminium/e x -d- gehaeuse - aus - leic htmetall - reihe -8265- 46888/165781-46925/ Wittl er, M. (2005) . EXAM Neuer Name, gr oße Tr adition. EXZeitsc hrift, 37/2005: 58-62. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 64 14th International Symposium on Hazards, Prevention, and Mitigat io n of Industrial Explosions Braunschweig, GERMANY – July 11 - 15 , 2022 Influence of vent distribu tion on the viole nce of a gas explosion Daubec h J. 1 ,*, Leprette E. 1 , P roust C. 1,2 1 Institut National de l’ Environnement I ndustriel et des Risques, Parc Technologique ALATA, BP 2, 60550 Verneuil- en -Halatte, France. 2 Sorbonne Universités, UTC-TIMR, 1 rue Dr Schweitzer, 60200 Compiègne, France. *Corresponding author’ s email: [email protected] Abstract The development of n ew energies (battery, fu el cell, elect rolyser, sto rage, et c.) le ads to the implementation of these applications in ISO shipping containers. Depending on the tec hnology, such containment can lead to a risk of acc umulation of flammable gas and ther efore explosion. I nd eed, the explosion of an I SO container can have major effec ts, as in the ca se of th e accident that o ccurred in British Columbia (Canada) in 2013. This accident destroyed completely th e container. The adjacent containers were de formed, and hous es’ and vehicles’ windows were blow n out within 150 m. One door of the container was projected to 100 meters. To avoid such future disa sters, an e fficient strategy of mi tigation should be implemented. The use of vent panels is one of the most popular protection techniques to mitigate the confine d explosion risk. The role of explosion vents is to discharge the excess gas p roduced by the combustion to limit the explosion overpressure to an acceptable value compatible with the mechanical strength of the enclosure to be protected. The c alculation of the ve nt area can be difficult due to the mul tiplicity of parameters that will influen ce the venting process. In most cases, the widely studi ed sit uation is that of an enclosure filled with a flammable mixture with explosion discharge orifices generally concentrat ed in one single area. There is little research into the influence of the distribution of the explosion venting area on the surface of the enclosure to be protected. This paper pr esents the results of an experimental study where 1.2 m 2 of vent area was spread over the surface of a 37 m 3 explosion chamber. Four configurations of vent areas distribution are studied. Two flammable mixtures are used respectively 15.5% and 17.4 % hydrogen-air with two locations of the ignition source. Keywor ds : Vented gas explosion, vent, secondary explosion 1. Introduction The development of n ew energies (battery, fu el cell, elect rolyser, sto rage, et c.) le ads to the implementation of these applications in ISO shipping containers. Depending on the tec hnology, such containment can lead to a risk of acc umulation of flammable gas and ther efore explosion. I nd eed, the explosion of an ISO container can have major effects, as in the case of the accident (WorkSafeBC, 2013) that occurred on April 13 th , 2013 in S aanich in British Columbia (Canada). At around 6:3 0 AM, an explosion occ urred in a shipping container used for various storage, resulting in the destruction of the container and damage to the sur roundings. Adjace nt containers we re deformed, and houses and vehi cles’ win dows were blown out. The container was torn off, the walls we re flattened, and some parts of the container were landed up to 275 meters away. The explosion blew out both ends of the container, hurtled the roof 15 meters, and shot one of the doors about 100 meters across Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 65 the site (Fig 1.). The blast also damaged window s in the surroundings " as far about a half a block away” i.e., glasses were broken around 150 meters. The c ause of the explosion (WorkSafe BC, 2013) is a leak from the 20lb propane tank of a barbec ue, store d inside a lar ge steel shipping container. The leak resulting fr om the open va lves crea ted an exp losive a tmosphere inside the shipping containe r. A soft drink cooling circuit that cycled and spa rked ignited the flammable mixture and trigg ered t he explosion. Damage ana lyses allow to estimate the maximum overpre ssure involved in the explosion. The lev el of overpressure insi de the container is ar ound 2.5 bar. So, to avoi d this kind of disaster , it’s necessary to define some strategy of mitigation. a. Container after the explo sion b. Container’s d oor projected at 100 m c. Container after the explo sion d. Broken win dows on the apartment blo ck Fig. 1. Damage of explosion in ISO container (from WorkSafeB C, 2013) The use of vent panels i s one of the most popular protection techniques to miti gate the confined explosion risk. The role of vents is to discharge t he excess gas p roduced by combustion outsi de a volume to limit the explosion overpressure to an acceptable value compatible with the mechanical strength of the enc losure to be protected. The physics of vented deflagration has b een studied for a long time (Coop er et al, 1986, Bimson et al, 1993, Catlin et a l, 1996, Proust e t al, 2010, B au wens et a l, 2011, Daubech et al, 2011). The widely studied situation is that of an enclosure filled with a flammable mixture with explosion discharge orifices genera lly concentrated in one single area. After the igni tion of th e flammable mixture, the initial flame growth leads to the production of hot combustion products. Due to the thermal expansion of the bu rnt gases, the internal pressure in the enclosure inc rea ses. When the pre ssure in the enclosure reaches the ope ning pressure of the explosion vents, the reactive mixtu re can be e xpelled to the outsi de forming a swirling cloud. Whe n th e fl ame Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 66 reac hes the openings, it c an ignite this reactive cloud, producing a secondary explosion in front o f the vent. Catlin (1991) accurately studied the evolution o f the flame f ront contour from ignition to flame e xit through the vent. He noti ced 3 diffe rent stages during flame propaga tion. Immediately after ignition, the flame front grows spheric ally around the ignition source. It then elongates into an ellipsoid of revolution and evolves into an ellipsoid truncated from it s base as the flame approaches the vent. As a result of the pressu re increase in the enclosure, part of the flammable cloud is expelled from the enclosure and takes the shape of a mushroom (Cooper et al, 1986). Oth er authors agree that it is shaped like a ball with a radius of the same order of magnitude as the hydraulic diameter of the vent (Harrison et al, 1987). The physics of the combustion of the outer cloud i n the secondary explosion is poorly understood. But it is possible to describe the combustion process (Catlin, 1991) of thi s flammable cloud in 3 phases: • Phase I: The fl ame velo city does not vary signi ficantly from the flame velocity inside the enclosure, while th e mushroom shape of the vortex c ontinues to develop. During this pha se, there are no high leve ls o f overpressure associated with the secondary explosion. • Phase II: The flame arrives at the head of the mushroom and suffers from an abrupt increase in the surface area . It is at this po int that the pressure effects of this explosion occur. • Phase III: This last phase is charac terized by a spherical flame propagation through the remaining reactive volume Some authors (Bauwens, 2010, 2012, Daubech, 2013, S ommersel, 2017, Skjold, 2019) show experimentally the influe nce of different parameters such as the initial concentra tion of the flammable mixture, the initial turbu lence, and the presence of obstacles or the position of igniti on source the explosion chamber. But the major problem with vent installation is that vent panels are generally located in the sam e area. This situation ca n crea t e a very huge external explosion (Daube ch et al, 2011) with a high level o f overpressure. The distribution of vents on th e surface of the enclosure to be protected has been little studied. This paper proposes to st udy the influence of v ent distribution on the violen ce o f a gas e xplosion. The vent holes are distributed on the w alls of the 37 m 3 explosion chamber . Th e total su rface of vents is 1.2 m 2 . Several configur ations of vent surface installation are p roposed. Two hydrogen flammabl e mixtures are use d with t wo positions of an ignition sourc e. 2. Experimental set- up The explosion chamb er i s a 37.5 m 3 parallelepipedal volume (internal dimensions: 6 m long, 2.5 m wide, and 2.5 m high) de signed to withstand an explosion overpre ssure of 2 bar (Fig 2.). I t is a metal structure made of H-irons and modular side fr ames which can be fitted with s olid walls to completely block the surface or ca n be customized to accommodate the vent panels or viewing windows. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 67 Fig. 2 . 37.5 m 3 explosion c hamber The injection of flammable gas into the e xplosion chamber is carried out from a 5 m 3 tank. The gas injection pressure is mon itored by a pr essure sens or located upstream of th e inj ection pilot valve. I t is set to 5 bar. The injection is carried out through a 20 mm diameter circular orifice. The leak age rate is a bout 50 g/s. The H2 concentrations are mon itored by 6 oxygen an alysers located at different heights in the explosion chamber. The turbulence generated by the leak is sufficient to completely homogenize the flammable atmosphere. The turbulence generated by the leak is su fficie nt to completely homogenize the fla mmable atmosphere. To ensure that the fl ammable atmosphere is completely quiescent, a ti me of 30 s is allowed between th e end of the hydro gen inj ection and ignition. The instrumentation includes 2 int erna l pressure sensors P 1 and P2 (Kistler 0 -2 b ar pi ezoresistive sensors) and 2 external pressure s ensors settled i n lens supports Lent 1 a nd lent 2 (Kistler 0 -2 bar piezoresistive sensors ). To visualize the explosion and opening of the v ents, a high-speed cam era (Phantom MIRO – 2000 i/s) and an HD camera are used. Th e flammable mixture is ignited with a pyrotechnica l match of 6 0 J. A tot al vent area of 1.2 m 2 divided int o 4 unit areas of 0.3 m 2 (0.6 m x 0.5 m) is used to study the influence of vent distribu tion on the explosion dynamics. Four vent distributi on configurations were studied (Fig. 3). Th e first configuration is the reference for which the 4 unit vents are in the same area . For the following configur ations, the unit ve nts are distributed on two sides of the explosion chamber. The external pressure sensors are always located on the vent axes so that their location changes from one config uration to another. Fig 3. also presents an overview of the instrumentation location for each c onfiguration. Two ignition source locations are studied: • at the centre of the wall opposite the 4 unit vents of configur ation 1, • at the centre of the e xplosion chamber. Two quiescent fla mmable mixtures are used: • 15.5 % vol. hydrogen-air mixture, • 17.4 % vol. hydrogen-air mixture. For each ve nt panel distribution, one experimental configuration is conducted twice to test the reproduc ibility with a go od level of success. T ranspa rent wa ll Vent pa nel 2,5 m 6 m 2,5 m Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 68 Configu ration 1 Configu ration 2 Configu ration 3 Configu ration 4 Fig. 3. Four vent distribution configurations and instrumentation positions P P I gnitio n s ou rce P1 P2 Lent2 Lent1 Norm al ca mera 3 m 3 m Fas t ca m era P I n te rnal p ress ure s ens or Exter nal p res su re s ens or Fr a ngible w al l Cl osed s ur face Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 69 The unit safety vents h ave an a rea of 0.3 m 2 . Vents are made of a PE plastic sheet held in place by a 5 cm square or flat iron frame screwed to the support plate (Fig. 4) depending on the chosen configuration. The op ening overpressure of this plastic sheet is about 80 mbar when held by the square iron frame, and 50 mbar when h eld by the flat iron fram e. Th e val ues of opening overpressure are experimentally determined. The surface density of the plastic sheet is 0.150 kg/m 2 . Fig. 4. Fixing iron frame 16 tests were carr ied out crossing: • the 4 frangible wall distributions (Configuration 1 to 4) • the 2 explosive atmospheres • the 2 ignition positions 3. Results and discussion Typical results Fig 5. presents the characteristic int erna l and exte rnal overpressures for the first configuration with the ignition on the side opposite the 4 uni t vents loca ted on the same flange for the 15.5 % and 17.4 % hydrogen-air quiescent mixtures. The phenomenology observed was alre ady presented by Daubech et al (2013) a nd can b e divided into the following steps: 1. Ignition 2. Flame propagation and pressure increase in the ex plosion chamber, 3. Opening of the vent as soon as the interna l pressur e reache s its opening pre ssure, 4. Discharge of part of the flammable cloud to the outside, 5. Formation of the externa l vortex, 6. Competition betwee n the production of burnt gases by the flame an d the discharge of the gase s through the vent, 7. Ignition of the externa l cloud by the flame rea ching the vent and explosion 8. Internal combustion in progress 9. Discharge of combustion products through the ve nt and pressure drop in the enclosure 10. End of internal combustion For the 15.5 % H 2 -air mixture, the maximum inside overpressure reaches 250 mbar and the external overpre ssu re reac hes 50 mbar. For the 17.4 % H 2 -air mixture, the maximum inside overpressure reaches 400 mbar and the external overpre ssu re reac hes 165 mbar. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 70 a b Fig. 5. Internal and external overpressures for configuration 1 - Ignition on the side opposite the 4 unit vent surface located on the same flange – 15.5 % (a) and 17.4 % (b) hydrogen-air quiescent mixtures. Influence of ignition location Fig 6. pr esents the intern al and external overpressures for the first configu ration with the ba ckwall and the ce ntral ignit ion for the 15.5 % hydr og en-air quiescent mixtures. For an identica l venting configuration, the internal overpressure is greater when the igni tion takes place as far as possible from the vents (250 mbar for the backwall igniti on v s. 160 mbar for the central ignition). With the b ackwall igni tion, the flame t akes longer to r eac h the discharge su rfaces of the explosion than for the c entral ignition. The amount of burnt gas produced by the flame when it runs every 6 m of the explosion chamber is greater th an the amount of burnt g as p roduce d for a central ignition. The external overpressure related to the secondary explosion is greater when the ignition is central compared to the backwa ll igni tion (90 mbar vs. 50 mbar). It can be expl ained by the fact that the flame reaches the ve nt earlier in the central ignition The externa l c loud is still in the for m of a vo rtex at the time of ignition. In the case of backwall igniti on, the externa l vortex can degenerate in a form of a fresh gas jet during t he interna l flame dev elopments alrea dy as observed by Da ubech e t al (2017 ). The fla me propagation in a compact e xternal cloud as a vortex leads to a more powerful secondary explosion. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 71 Fig. 6. Internal and external overpressures for configuration 1 - Backwall and ce ntral ignition – 15.5 % hydrogen-air quiescent mixtures Influence of the distribution of vent are as on th e explosion violence Fig 7. presents the internal and exter nal overpre ssures for configur ations 1 to 4 with the backwall ignition for the 15.5 % hydrogen-air quiescent mixtures. Table 1 sums up the values of overpressures. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 72 Fig. 7. Internal and external overpressures for configurations 1 to 4 - Backwall ignition – 15.5 % hydrogen-air quiescent mixtures Table 1. Internal and ex t ernal overpressure for co nfigurations 1 to 4 - Backwall ignition – 15.5 % hydrogen-air quiescent mixtures n° Configura tion Internal overpre ssu re (mbar) External overpre ssu re – Lent 1 (mbar) b External overpre ssu re – Lent 2 (mbar) Config 1 250 50 25 Config 2 90 16 7 Config 3 110 30 28 Config 4 95 18 16 We notice th at higher int ernal and exter nal ov erpressures ar e reached for configuration 1. The internal overpre ssu res for configurations 2, 3 , and 4 have the same order of magnitude around 100 mbar. Configurations 2 and 4 g ive the same order of ma gnitude f or internal and external overpressures. I n comparison, the overpre s sures of c onfiguration 3 are higher. Fig 8. presents the internal and exter nal overpre ssures for configur ations 1 to 4 with the backwall ignition for the 17.4 % hydrogen-air quiescent mixtures. Ta bl e 2 sums up the values of overpre ssures. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 73 2.2 The facts The installation is com plex with several reactors and other pieces of equipm ent. The vent lines from each equipm ent item are connected to a vent header network. Gases produced in each equipm ent item pass through the vent header network and are tr eated before release to the atmosphere. The treatment takes place eith er in a scrubber to rem ove toxic gases or in a thermal treatm ent unit to remove Volatile Organic Compounds (VOC). On th e day of the explosion, a chem ical operation was carried out in a reactor in three successive phases: 1. A mixture of various chemicals was heated to distill off the dichloromethane that it contained , 2. Hydrogen peroxide 35% was fed in over 10 hours to carry out an oxidation reaction, 3. Potassium hydroxide 45% was added to d ecom pose the excess of hydrogen peroxide. During the first and s econd phases, the gases produ ced were sent to the scr ubber, because the therm al treatment unit was stopped for m ain tenance. During the third phase, the gases produced were sent to the thermal treatm ent unit. The explosi on occurred during this last phase. 2.3 The explanation During the first phase, the dichlorom ethane, which ha d been distilled off, c ooled and condensed in the lowest points of the vent network. During the third phase, the reaction b etween hydrogen peroxide and potassium hydroxide produced pure oxygen. This gas also went through the vent network. Vapors of dichloromethane m ixed with oxygen gas. This mixture was inside explosive limits and it ignited in the thermal treatm ent unit. A flame fr ont blew back along the ve nt line. The overpressure shattered a piece of glass equipment and ruptured a length of polypropylene pipe. 2.4 The lessons 2.1.1 Flammability of dichloromethane Dichloromethane has no flash point and is not cla ssified as a flammable liquid under international regulations. However mixtures of dichloromethane vapour and air are explosive. The following data are given by the GESTIS database (h ttps://gestis-datab ase.dguv.de/search)  Melting point = -97 °C  Boiling point under atmospheric pressure = 40 °C  Saturated vapour pressu re at 20 °C = 470 hPa  Upper Explosive Limit (UEL) = 22% v/v  Lower Explosive Limit (LEL) = 13% v/v  Autoignition temperature (AIT) = 605 °C  IEC/ATEX explosion group = IIA  Maximum Explosion Pressure = 5.9 bar  Minimum Ignition Energy (MIE) = 9300 m J The last piece of data shows why dichloromethane shows no flas h point. Its Minimum Ignition Energy is higher than the energy of sparks used in flash point measurements. According to Britton (1999), the Minimum Ignition Energy of dichlorm ethane is five orders of ma gnitude lower in p ure oxygen than in air and the e xplosive limits are wider:  Upper Explosive Limit (UEL) = 68% v/v  Lower Explosive Limit (LEL) = 11.7% v/v  Minimum Ignition Energy (MIE) = 0.137 m J Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 80 2.1.2 Decomposition of hydrogen peroxide to give oxygen Hydrogen peroxide is sold as a so lution in water with a mass concentr ation of 20 % to 70 %. It is slightly acidic, with a pH of about 3. It deco mposes to give water and gaseous oxygen, as shown below: H 2 O 2 (l) → H 2 O (l) + ½O 2 (g) The decomposition of hydrogen pero xide is slow at am bient tem perature and mild ly acidic pH. It is catalysed by strong acids and bases as well as by very many metal ions and indeed other substances. 2.1.3 Blow-back of flame from thermal oxidizers In order to reduce emissions of Volatile Organic Com pounds, it is usual to treat waste gas streams using a thermal oxidizer. This usually consists of a hot ceramic bed. The waste gas is passed over the bed and heated to temperatures in excess of 7 00 °C. At such temperatures the organic com pounds are burnt, giving water, carbon dioxide etc. However, it is im portant to note that the waste gas fed to the thermal oxidize r must be outside of explosive lim its. If waste gas inside exp losive limits is fed to a thermal oxidizer, ignition will occur and a fl am e front will blow back along the vent line to the source of the waste gas. The flam e front will accelera te as it travels back dow n the vent line. It m ay well reach the speed of sound in the gas mixture a nd so chan ge from a deflagration to a detonation regime. 2.5 Where did we go wrong? A risk analysis had been carried for the proc ess. But the working group was not aware that dichloromethane was flammable. They knew that oxygen was produced in the third phase of this operation but they did not realise that the flammab ility lim its of dichlorom ethane are much wider in oxygen than in air. Nor did they imagine that di chloromethane, which was boiled off in the first phase of the process, would conde nse in the vent network and then mix with the oxygen gas produced in the third phase. 3. Explosion of a gaseous mixt ure of hydrocarbon, polar solvent, oxygen and nitrogen 3.1 Description An explosion occurred inside a bunker used for res earch purposes. A flexible stainless steel hose ruptured in several places and one person suffered a slight ear injury (tinn itus). The building structure was not affected and there was no releas e of material to the environm ent. Fig. 1. The flexible hose after the incident Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 81 3.2 The facts The accident occurred during the la st of approximately 250 experim e nts designed to optimise a new process for manufacture of a commercial product by oxidation of a common hydrocarbon. The experiments were all carried out in a 250 m l pressure vessel (design pressure 200 bar), situated inside a bunker. The reaction invol ved the following steps. 1. Charge the reactor with 75 g of the hydrocarbon, a polar solvent and a catalyst. 2. Close the reactor and pressurize with nitrogen to (10 to 30 bar gauge). 3. Start the stirrer and heat to th e desired temperature (60 to 90 °C). 4. Inject the following simu ltaneously to the liquid phase: a. a mixture of nitrogen and oxygen b. aqueous hydrogen peroxide. We had assessed the safety of this process in our Process Safety Laboratories before the experimental work started. We had shown that, at the desire d temperature range and at pressures up to 10 bar gauge, the concentration of flammable gases (i.e. hydr ocarbon plus solvent) in the gas phase of the reactor would always be between the Lower Explosive Limit (LEL) and the Upper Explosive Lim it (UEL). For this reason, it was recommended to k eep the ratio of oxygen to nitrogen below the Limiting Oxygen Concentration (LO C) of 7 % (see figure 2 below). However, with a ratio [O2]/([O2]+[N2]) below the LOC of 7 %, the reaction yield and process throughput were found to be rather poor. The work ing temperature was raised from 60 °C to 90 °C and the working pressure from 10 to 30 bar gauge. Progressively higher levels of oxygen were then tried, generally in the range of 41 % to 45 %, instead of the value of less than 7 % which had been advised. In the final experiment, the working pressure was 30 bar and the level of oxygen m easured in the gas phase was at a record high level of 52 %. An explosion occurred, causing the rupture of a flexible line, certified for 206 bar. The reactor itself was not damaged, but its rupture disc had blown. A person was present in the bunker at this point and he suffered an ear in jury (tinnitu s). Fig. 2. Illustrative flammability diag ram for a hydrocarbon in air Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 82 3.3 The explanation After the incident we established that the gas phas e of the reactor had been inside exp losive limits for most if not all of the 250 experiments carried out. It seems that an ignition source was only found on the last experiment, where the level of oxygen was highe st. In general, it is well established that the energy required for ignition of a given fuel is mu ch reduced at high levels of oxygen, as we saw in the case of dichlorom e thane. 3.4 The lessons The pattern of damage to the flexible hose, which was ruptured in severa l places, suggests tha t the explosion propagated as a detonation . A gas phase detonation propagates at a speed which is at least equal to the speed of sound in the gas mixture. That m eans that ru pture at one point , releasing gas and reducing the pressure at that point, does not stop the pr opagation of the flame front. 3.5 Where did we go wrong? This incident demonstrates that recipients of proc ess safety information must be able to understand, interpret and apply that informati on. Despite the availability of a thorough report detailing the hazards and safety limits of the chem ical system in question, the group operati ng the equipm ent did not understand how to apply its recommendations. For the same reason they were not in a position to correctly interpret the ef fect of changed operatin g c onditions . In this case, they believed an incre ase in operating pressure would com pensate for a rise in operating temperature and place the system below the LEL; but this was not the case. This type of reaction is carried out in a bunker, because it is recognised that the reactions are experimental, using m aterials whic h are dangerous. There are two se parate reactors in the sa me bunker, each controlled from outside the bunker by a tech nician. According to the Standard Operatin g Procedure, no person shall be present when a reaction is being carried out in one or other r eactor . In this case, a technician was in the bunker whilst th e reaction was being carried out. I f the Standard Operating Procedure had been followed, ther e would have been no human consequences. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 83 4. Explosion of a gaseous mixture of isopropanol and oxygen 4.1 Description A chemical process was scaled up and transferred from one location to another. On th e first batch in the new location, the gaskets on the uppermost part of the reactor were blown out and a fire started. The fire was extinguished. Nobody was injured and there was no damage to the environment. Fig. 3. Top of the reactor after the incident 4.2 The facts We were developing a process for a new produc t grade. The process involved mixing water, isopropanol, guar (a naturally o ccurring polym er of sugar), hydroge n peroxide and aqueous sodium hydroxide solution. It was operated at atmospheric pr essure at tem peratures up to 55 °C. As part of the developm ent, it was tried out at various locations , increasing the scale. The incident occurred on the first batch carried out in a 5 m 3 glass lined mild steel reactor. The process was carried ou t as intended. But, during the operati on, the operators sudde nly heard the sound of a loud explosion and saw that the reactor was on fire. The s ite fire brigade extinguished the fire. Nobody was injured. The damage to the equipm ent included blown gaskets on the lid of the reactor (see figure 4 above) and one broken gl ass seal pot on a breather line. We estim ate that the pressure inside the reactor reached 12 to 13 bar gauge, i.e. at least twice the design pressure of 6 bar gauge. 4.3 The explanation After the incident, we showed that th e reaction generates oxyg en under normal conditions. That is to say, the mixture is alkaline, so the hydrogen peroxide solution which is fed in is converted to water and oxygen. The reactor was initially under nitroge n, but the rate of ge neration of oxygen would have been enough to create an oxygen enriched atmo sphere in the h eadspace of the reactor in four minutes. This would have formed a m ixture of o xygen and isopropanol vapor inside explosive limits. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 84 4.4 The lessons Mixtures of flamm able gas with pure oxygen or air enriched with oxygen ar e very easy to ignite compared to m ixtures of the same gas with air. The MIE is comm only reduced by a factor of 100 or more (see table 2 below). NFPA 53 cites severa l inciden ts involving the spontaneous ignition of deposits of grease etc. in oxygen rich atmospheres at am bient temperature. Also, an explosion of a mixture of flammable gas with oxygen gives a much higher pressure than a m ixture of the same gas with air, so this one deviation increas es bot h the probability and severity of the scenario. 4.5 Where did we go wrong? In the previous trials of the process , some or all of the oxygen form ed had been displaced by a continuous purge with nitrogen. In the new location we did not realize th at this was necessary. We thought that it was sufficient to ensure that the ve ssel was under a nitrogen atmosphere at the start of the batch. 5. Comparison of gas phase explosions in air and in oxygen Table 2 lists the main flammability characteristics of some typical fuels in air and in pure oxygen. Some trends are obvious: 1. Minimum Ignition Energy (MIE) is reduced, in most cases by two orders of magnitude. 2. Autoignition tem perature (AIT) is reduced in some cases. 3. Flammability Limits are wider i.e. th e LEL is s lightly reduced whilst the UEL is much higher. According to NFPA 53, the pressu re ratio, i.e. the final absolute pres sure divided by the initial absolute pressure, is about 7 for typical hydrocarbons in air but 33 in oxygen. In elongated geometries, such as pipes, flam e accelerati on leading to detonation is m ore likely. Table 2: Characteristics of different fuels in air and pure oxygen Fuel gas Air Oxygen MIE mJ AIT °C LEL %v/ v UEL %v/ v MIE mJ AIT °C LEL %v/ v UEL %v/ v Acetone 1 .15 465 2.6 12.8 0.0024 2.5 60 Acetylene 0.017 305 2.5 1 00 0.0002 296 2 .5 100 n-but ane 0.25 288 1. 6 8.4 0.00 9 27 8 1.6 49 Dichloro methane 9300 615 13 2 2 0.137 606 11.7 68 Diethyl ether 0.19 193 1.9 36 0.0012 1 82 2.0 82 Ethane 0.24 515 3.0 12.5 0.0019 506 3 .0 66 Ethyle ne 0.084 490 2. 7 36 0.00094 485 3. 0 80 n-hexane 0.24 225 1. 1 7.5 0.006 218 1.2 52* Hydrog en 0.016 5 20 4 .0 7 5 0.0012 400 4.0 95 Methane 0 .21 630 5.0 15 0.0027 5.1 61 Propan e 0.25 2.1 9.5 0.0021 The values are taken from BRITTON ( 1999), NFPA 53 and the GESTIS database * determined at 93 °C 6. Risk analysis Everybody knows you should do a Process Risk Analys is on a chemical process. But anybody who has ever tried knows that it is di fficult to do so thoroughly. Chemi cal installations are com plex. Chemical processes are difficult to understand. Chemicals often r eact in ways we do not an ticipate or with consequences that we did no t imagine, as was the case in the thre e incidents describ ed above. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 85 The main hazards of the chem ical industry are as follows:  Physical explosi on or implosion  Gas phase explosion  Dust explosion  Fire  Runaway reaction  Condensed phase detonation  Loss of containment leading to acute effects on hum an health and the ecosystem. To assess the risk of a gas phase explosion for a process involving a chem ical reaction, it is vital to ask three questions: 1. Is an oxidant gas used or generated by the primary (wanted) reactio n or by a secondary (unwanted) reaction? Oxidant gases include oxygen (O 2 ), ozone (O 3 ), nitric oxide (NO), nitrogen dioxide (NO 2 ,), nitrous oxide (N 2 O), fluorine (F 2 ,) and chlorine (Cl 2 ). For exam ple, EGAN (2015) describes a gas phase explosion in a diazotis ation reactor. In this case the oxidant gas was nitric oxide. 2. Is a fuel, such as a hydrocarbon present in the form of a gas or vapor? At least 99% of chemical processes I have dealt with have some kind of fuel, e ither as a solvent or as a reagent or as a product. 3. Does the reaction involve m aterials that could act as ignition sources? An example would be activated Raney Nickel. “Raney ni ckel” is an allo y of nickel and aluminium, usually s upplied in the form of granules. Activation means treating it wi th aqueous sodium hydroxide, which dissolves most of the aluminium, leaving granules containing nickel in finely divided form, which can be used as a hydrogenation catalyst. Activated Ra ney nickel must always be ke pt under water or in an inert atmosphere. If it com es into cont act with air, the spontaneous oxidati on of the fine particles of nickel make it glow red hot and it beco m es a powerful ignition source. If a danger of gas phase explosion does exist for a gi ven chemical proces s, the next step i s to ac quire all of the data need to carry out a risk assessment, such as:  Explosion limits of the fuel in the oxid ant gas which are usually expressed by: o LEL (Lower Explosive Limit), o UEL (Upper Explosive Limit) o LOC (Limiting Oxygen Concentration)  Parameters linked to its ignition, such as: o Gas group (A to D under the NFPA system and IIA, IIB or IIC under the IEC/ATEX system) o MIE (Minimum Ignition Energy) o AIT (Autoigniton Temperature)  Parameters linked to explosion violence, such as: o P Max (Maximum Explosion Pressure) o K G (Normalised ra te of pressure rise) o Fundamental burning velocity All of this data must be acquired under conditions com parable with thos e of the process. For example, if a process operates under a pressu re of 10 bar gauge and 200 °C, the explosive limits determined at atmospheric pressure and ambient tem perature will not be representative. Indeed, in general, the explosive range (i.e. the gap between LEL and U EL) increases with in creasing pressure and temperature. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 86 With such data, determ ined under representative c onditions, we can then evaluate possible scenarios of gas phase explosion where all three of the following causes are pres ent in the same place at the sa me t i me: 1. Oxidant gas in sufficient concentration ( ≥ LOC) 2. Fuel gas within explosive limits (LE L to UEL) 3. Effective ignition source (i.e. ener gy > MIE or temperature > AIT). In the case of a gas phase explosi on inside a vessel, a fourth caus e is sometim es added, for example when the maximum explosion pressure is above th e vessel design pressure but below the pressure expected to cause rupture (typical ly 3 tim es the design pressure): 4. Rupture frequency based on a comparison of th e maximum explosion pressure, the vessel design pressure and the estimated rupture pressure of the vessel. Each cause is given a frequency of occurrence which m ay be:  Given (occurring more than ten tim es per year)  Very frequent (occurring between one and ten times per year)  Frequent (occurring between once every ten years and once per year)  Possible (occurring between once every thousan d years and once every ten years)  Improbable (occurring between once every hundred t housand years and once every thousand years) EGAN (2016-A) describes the method used to estimate ignition frequency, which depends on:  The characteristics (MIE, AIT, etc. of the explosive m ixture)  The volume within explosive lim its (for exampl e many more ignition sources are to be found outside a vessel then inside)  The various ignition sources, such as electrical equipment, wh ich m ay be present and their suitability for explosive atmospheres. We assess all ignition sources (ele ctrical equipm ent, static electrical disc harges, friction, electromagnetic radiation, heat, etc.) and for a given scen ario, we base the over all ignition f requency on the ignition source which we cons ider to be the most frequent fo r the gaseous mixture concerned. The following assessments of igni tion frequency are typical:  Given =1/1 Hydrogen released to an area not rate d for that gas,  Very Frequent ≈ 1/10 Methane released to an area with unrated electr ical equ ipment,  Frequent ≈ 1/100 Methane released to an ar ea with equipm ent rated for zone 2,  Possible ≈ 1/1000 Methane released to an area with equipment rated for zone 1. Our estimation of ignition frequency is based on feedback from everyday experience handling hydrocarbons, such as gasoline, under air containi ng 21 % volum e of oxygen. For example, we know from experience tha t mixtures of gasoline vapor and air within explosive limits will of ten be present when handling this material but co ntrol of ignition sources is an effective way to prev ent explosions and fires. The same thing cannot be said for gaseous m ixtures of hydrocarbons and air en riched in oxygen or in other gases such as chlorine. Fo r such mixtures, experience indicat es that ignition nearly always occurs by both th ermal and electrostatic m echanisms . Therefore, the igniti on frequency is “Given”, or, if there is sufficient practical experience and data on the mixture, at best “Very Frequent”. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 87 EGAN (2016-B) describes the method we use to asse ss scenarios in a Process Risk Analysis. From the frequency of the necessary, sufficient and independent causes of each s cenario, we estim ate its Probability of occurrence, over a period of one year, on a scale ranging from 1 to 10 -6 . We assess the Severity level of the scenario in term s of human and environment consequences on a scale from Low (any human or environm ental consequences are clea rly reversible) to Disa strous (exposure of 100 people to a risk of dying or very long term pollution). We then assess the Risk level of th e scenario from its Probability a nd its Severity. We define three levels of Risk:  Risk 1 (Unacceptable) – it is manda tory to reduce the Probability or Severity of such scenarios by suitable means such as adding pr eventive or protective safeguards.  Risk 2 (Interm ediate) – the Probability or Seve rity of such scenarios should be reduced by suitable means such as adding preventive o r protective safeguards.  Risk 3 (Acceptable) – the ex isting safeguards must be maintained in good working order. The higher the level of the Severity of Consequences the lower needs to be the P robability to get an “Acceptable” level of Risk. In the case of a gas ph ase explosion, preventive safeguard s might include:  On line analysis of the level of oxygen in the gas phase of a re acto r, such as one u sed for oxidation of a hydrocarbon, with automatic shutdow n of the system if the level of oxygen is too high,  Explosion venting,  Explosion suppression. Protective safeguards might include placing a reactor in a bunker. 7. Conclusions Three incidents involving gas phase explosions occu rred in the Solvay gro up of chemical companies over a period of seven years. In each case , the oxygen was derived from the decomposition of hydrogen peroxide used in the process. The incide nts show how important it is to identify chem ical reactions which release oxygen and to take th is in to account when assessing the risk. A gas phase explosion of a given fuel gas or vapor in oxygen may be com pared with an explosion of the same fuel in air as follows: 1. Minimum Ignition Energy is reduced, in most cases by two orders of m agnitude. 2. Autoignition temperature is reduced in som e cases. 3. Flammability Limits are wider i.e. th e LEL is s lightly reduced whilst the UEL is much higher. 4. Explosion pressure ratio is higher : ≈ 7 in air and ≈ 33 in oxygen. 5. In elongated geometries, such as pipes, flam e accel eration leading to detonation is more likely. References Britton (1999). Avoiding Stat ic Ignition Hazards in Chem ical Operations (CCPS). Egan, S. M. (2015). Learning less ons from eight gas explosions, Loss Prevention Bulletin, 243: 18- 27. Egan, S. M. (2016-A). Ignition Frequ ency in Process Risk Analysis at Solvay, Chemical Engineering Transactions, 48: 289-295. Egan, S. M. (2016-B). Process Risk Analysis within Solvay, Chemical Engineering Transactions, 48: 823-828. NFPA 53 (2016) Recommended Pr actice on Materials, Equipment, and Systems Used in Oxygen- EnrichedAtmospheres Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 88 In v estigation on flame pr opagation and particle decomposition beha vior of dust explosion in MIKE 3 apparatus Y angyue P an a , Christoph Spijker a & Harald Raupenstrauch a a Montanuni versitaet Leoben, Austria E-mail: yangyue [email protected] Abstract Due to high turb ulence and high particle load in the MIKE 3 apparatus, the flame behavior is hard to be observ ed. The previous study sho wed that particle clusters are the reason for the irre gular flame front (P an et al., 2022b). In this study , we in vestigate the influence of dust suspension conditions of dif ferent particle sizes as well as the particle decomposition beha vior during the explosion in e xperimental and simulation methods. The e xperiment of coal dust e xplosion was conducted in the MIKE 3 apparatus. Fi ve size ranges: 32 - 45 µ m , 45 - 63 µ m , 63 - 71 µ m , 71 - 125 µ m , 125 - 250 µ m were tested and the flame propagation process w as recorded by a high-speed camera. The flame front ov er time was detected using the MA TLAB image post-processing code. In the simulation, the default solv er coalChemistryF oam in OpenFO AM was used. Three mono-sized particles were simulated: 25 µ m , 125 µ m and 250 µ m . The particle positions of each case were extracted from the simulation of dust dispersions in MIKE 3 (P an et al., 2020). The particle v olatile content was modified to match the coal dust characters. The flame propagates to wards the particle clusters and the intense radiation sho ws a luminous flame. Due to the high turbul ence and une venly distrib uted coal clouds, the flame shape is asymmetrical. The simulation result sho ws that the unev en distribution of particles has little influence on the flame propagation in the MIKE 3 apparatus. K eywords: dust e xplosion, MIKE 3, OpenFO AM, Eulerian-La grangian 1 Intr oduction The simulation of dust explosion still f aces challenges due to the complexity of multiphase flo w and chemical reactions. T o in vestig ate the process, proper simulation models should be used accordingly . The simulation methods for multiphase phase flo w are the Eulerian and Lagrangian approaches. For the Eulerian method, particles and air are described as continuum phases, where the solid particle is implemented with a pseudo fluid model. The percentage of each phase in a cell is described with a coef ficient. This method is able to determine the flo w field properties such as flame propagation pro- file out of the embedded chemical reaction model. Ho we ver , the particle trajectories are not predicted due to the assumption of the pseudo-fluid model. The Lagrangian method provides the possibility of detailed in v estigation of particle ef fects. The particles are seen as individuals and data such as compo- nents and reaction rate is a vailable. At the same time, the biggest drawback of this method is the high requirement of computation time. The experimental instruments are able to pro vide repeatable results b ut also many stays unkno wn to us. F or example, in the pre vious study , the dust concentration along the MIKE 3 tube prior to ignition v aries due to different particle sizes. Moreo ver , the rapid explosion makes it hard to observ e the chemical reactions and transport phenomena. In this study , continuous work is carried out to study the e xplosion process of dust explosion in the MIKE 3 apparatus in e x- perimental and simulation methods. The open-source toolkit OpenFO AM version 5x w as used in the simulation work. 2 Experiment setup The e xperimental apparatus used in the study is sho wn in Fig. 1. MIKE 3 is a modified Hartmann tube that is widely used in laboratories to test the minimum ignition ener gy of po wders. The appa- 14 t h International Symposium on Hazar ds, Pr evention and Mitigation of Industrial Explosions Braunsc hweig, GERMANY - J uly 11-15, 2022 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 89 F ig. 6 : Flame fr ont pr opagation with dif fer ent initial particle positions in the simulations. 0 0.05 0.1 0.15 0.2 0.25 0.3 Height (m) 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Particle concentration (-) 250 m particle concentration evolve 0 s 0.03 s 0.06 s 0.12 s 0.15 s 0.18 s 0.21 s 0.24 s Ygas (a) Evenly distrib uted initial positions. 0 0.05 0.1 0.15 0.2 0.25 0.3 Height (m) 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 Particle concentration (-) 250 m particle concentration evolve 0 s 0.03 s 0.06 s 0.09 s 0.12 s 0.15 s 0.18 s 0.21 s Ygas (b) Unevenly distrib uted initial positions. F ig. 7 : 250 µ m particle concentration and volatile content e volve over the axis of the tube . source o ver time. Particles that are lo wer than the ignition point descend to the bottom of the tube. The shape ’U’ becomes wider o ver time as a result of the flame propag ation. In Fig. 7b, the initial po- sition of particles are distrib uted in the domain with dif ferent concentrations. There are more particles between 0.1 m and 0.25 m and less particles near the bottom of the tube.The particle concentration at the bottom of the tube does not increase remarkably and resulting in a higher particle concentration at the top of the tube comparing to the Fig.7a. The particle v olatile content is marked as circles and the shade of the filling represents the remained v olatile, where black means that all of the volatiles are present, and blank means that all of the v olatiles ha ve been released.On the other hand, they can be representati ve of the flame front position.The lo west particle concentration can be found near the center of the flame due to the gas expans ion. The particles near the ignition source start to decompose as the flame forms. Then, more particles are b urnt when the flame propagates. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 96 4 Conclusions The flame propagation beha vior in the MIKE 3 apparatus was studied in this paper with e xperiments and CFD simulations. The dust explosion of 32 - 45 µ m , 45 - 63 µ m , 63 - 71 µ m , 71 - 125 µ m , 125 - 250 µ m particles were carried out with concentration of 750 g / m 3 . The flame in the experiment tends to propagate in the direction of the particle clusters. As a result, the flame shape is asymmetrical ov er the axis. The flame cluster indicates the une v en distrib ution of particles at the instance of ignition. When the particle size increases, the flame speed decrease. It is also found in the simulation result.A comparati ve simulation with e venly distrib uted particle concentration was carried out. Acknowledgements PY ackno wledges the financial support of the China Scholarship Council (CSC, No. 201808420277). The authors are grateful for the computing time granted by the Chair of Thermal Processing T echnol- ogy (TPT), Uni versity of Leoben. Refer ences CEN EN 13821 (2003). P otentially explosive atmospher es - Explosion pr evention and pr otection - Determination of minimum ignition ener gy of dust/air mixtur es. European standard, Brussels. Cesana, C., Siwek, R. (2010). Manual MIKE 3 appar atus . K uehner A G. Clone y , C. T ., Riple y , R. C., Pegg, M. J., Amyotte, P . R. (2018). Laminar burning velocity and structur e of coal dust flames using a unity Le wis number CFD model . Comb ustion and Flame, 190:87–102. ISSN 00102180. doi:10.1016/j.comb ustflame.2017.11.010. Field, M. A. (1969). Rate of comb ustion of size-gr aded fractions of c har fr om a low-r ank coal between 1 200°k and 2 000°k . Comb ustion and Flame, 13:237–252. Hosseinzadeh, S., V anierschot, M., Norman, F ., V erplaetsen, F ., Berghmans, J. (2018). Flame pr opa- gation and flow field measur ements in a Hartmann dust e xplosion tube . Po wder T echnology . ISSN 1873328X. doi:10.1016/j.po wtec.2017.10.001. K ern, H., W ieser , G. J., Raupenstrauch, H. (2015). Flame pr opa gation in lycopodium/air mixtur es below atmospheric pr essur e . Journal of Loss Pre v ention in the Process Industries, 36:281–286. Morsi, S., Ale xander , A. (1972). An in vestigation of particle tr ajectories in two-phase flow systems . Journal of Fluid mechanics, 55(2):193–208. P an, Y ., Spijk er , C., Raupenstrauch, H. (2020). In vestigations on the ef fect of particle size on dust dispersion in mik e 3 appar atus . In 13th International Symposium on Hazar ds, Pr evention, and Mitigation of Industrial Explosions (ISHPMIE) . P an, Y ., Spijk er , C., Raupenstrauch, H. (2022a). CFD modeling of particle dispersion behavior in the MIKE 3 appar atus . Alexandria Engineering Journal, 61(12):9305–9313. ISSN 11100168. doi:10.1016/j.aej.2022.03.039. P an, Y ., Spijk er , C., Raupenstrauch, H. (2022b). Numerical in vestigations on dust e xplosion pr ocess in MIKE 3 appar atus . In Pr oceedings of the T enth International Seminar on F ir e and Explosion Hazar ds , pages 56–64. Oslo,Norway . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 97 TGA-FTIR f or kinetic and e volv ed gas analysis of the coal particles in dust deflagration Y angyue P an a , Christoph Spijker a & Harald Raupenstrauch a a Montanuni versitaet Leoben, Austria E-mail: yangyue [email protected] Abstract The common approach in the dust deflagration simulations ignores the temperature gradient inside of the particles. Therefore, the reaction rate of the particle at one temperature remains constant. In order to e xplore the mass loss and e volv e gas characters during the coal particle decomposition procedures, a single-particle model was created using OpenFO AM tool kit. In this study , the pyrolysis characteristics and gas properties of the coal sample were determined by TGA-FTIR. The e v olution of gases in real-time was in vestigated and implemented as kinetic models in the dust deflagration. T o solv e the heat and mass transfer of the single-particle, a two-phase solver based on the eulerian method was de v eloped based on reactingFoam.The porosity of the coal particle w as included with respect to the coal mass. The result of the heat and mass transfer of the single-particle model agrees well with the e xperiment. In order to simulate the particle behavior in the dust e xplosion, new boundary conditions e xtracted from dust explosion simulations will be implemented. The final goal of the single-particle model is to implement the ne w particle decomposition beha vior into the full scale of dust explosion simulations. K eywords: dust deflagr ation, coal, single-particle model, TGA-FTIR, kinetics, OpenFO AM 1 Intr oduction A dust e xplosion is fundamentally a phenomenon from solid fuel comb ustion. In the safety science field, dust explosion is unkno wn due to its complexity . The lev el of se verity is highly dependent on the particle species, the humidity and the particle agglomeration. The comb ustion mechanisms of the single-particle is an intrinsic vie w to understand the dust explosion itself and to pro vide reliable disaster control parameters. The dust explosion is a rapid chemical reaction process where useful data from the e xplosion studies are hard to get. The numerical methods and hardware de velopment mak e the computational fluid dynamics (CFD) applicable in more research fields by using reliable sub- models (Skjold, 2007). The details of chemistry and physical processes will be re vealed and adopted by industries in system designs. The chemical and physical sub-models in CFD simulation are critical in dust e xplosion to ha v e a reproduction of the process in dust deflagration. The open-source CFD tool OpenFO AM, which has e xisting functions and is flexible for ne w models, is used in the study . Thermogra vimetric analysis aided with Fourier transform infrared spectroscopy (TGA-FTIR) has been used in coal particle analysis. The temperature-programmed system records the mass loss in coal particles and e volv ed gas from the pyrolysis. In the particle combustion stages, the pyrolysis of particles will ha ve an impact on the temperature and the gas composition of the surrounding fluid phase due to heat and mass transfer . On the other hand, due to the combustion and interaction between the two phases, the TGA-FTIR results are not suf ficient in predicting the particle beha vior for dust deflagration. The complex physical transition and chemical reaction require proper boundary condi- tions in order to model physical phenomenons in the dust e xplosion. A new solver loaded with ne w particle models is created to solv e the mass, heat and momentum transport in the particle. The single particle is resolv ed and Eulerian approach is used in order to present details of the gas production and transport. A porosity variable ξ is introduced in the model. 14 t h International Symposium on Hazar ds, Pr evention and Mitigation of Industrial Explosions Braunsc hweig, GERMANY - J uly 11-15, 2022 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 98 2 Experiment method T o b uild up the single-particle model, a coal sample was used. In order to minimize the ef fect of inner particle reactions, the bulk coal w as processed into a small particle size. A ball mill w as used firstly to grind the b ulk coal and then the coal dust was sie ved into a particle size range of 20-32 µ m. The coal was dried in an o ven at 100 ◦ C for 12 hours. The element analysis result of the coal sample is sho wn in T able. 1. Fig. 1 sho ws the scanning electron microscopy (SEM) images of the coal sample. Fig. 1a sho ws a non-consistent shape and rough surfaces of the particle. In Fig. 1b, a porous structure was found. (a) Overall vie w of particle shape. (b) P or ous structur e of the coal particle . F ig. 1 : Surface structur e of the coal sample . T able 1 : Element analysis of the coal sample, air -dry basis. Element analysis (wt. %) C HONS Coal 73.9 4.87 4.81 1.47 - The thermogra vimetric analysis was carried out in the chair of Process Engineering of Industrial En vironmental Protection, Uni versity of Leoben. In each heating rate, 10 mg of sample was heated up from 30 ◦ C to 1100 ◦ C at heating rates of 30 ◦ C/min. The inert purging g as N 2 was used to displace air in the pyrolysis zone with a flo w rate of 100 ml/min. By this means, the sample was not oxidized by Oxygen. The ev olved g as was measured using the complementary FTIR spectrometer . 3 Kinetic analysis 3.1 TGA experiment r esult The TGA result is sho wn in Fig. 2.The mass of coal samples was normalized by their initial v alue. The first deri v ati ve of mass change on time ( d m / d t , DTG curve) is also plotted into the figure. The highest peak occurs at between 700 - 800 K . It corresponds to the main decomposition stage. In this stage, the coal mass decreases remarkably due to acti ve p yrolysis. In the meantime, volatile g ases, CO, CO 2 , CH 4 and light hydrocarbons are released and measured in the FTIR analyzer . The mass of coal decreased around 30% of the total weight in this stage. The second reaction rate is not able to shape a "step" in the TG curv e and can only be observed in the DTG curv e. The second peak in the DTG occurs between 900 - 1100 K . In this stage, the reaction rate v alue of the peak is distinctly lo wer than the first peak. The mass loss of the coal in this stage is from the second p yrolysis of the gaseous tar . The total mass loss of the coal during the experiment is 34.98% ( β = 30 K / min ). The reaction rate of the first and second peaks are 10 . 66 × 10 − 4 s − 1 at 749.7 K and 1 . 411 × 10 − 4 s − 1 at 1050 K . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 99 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 Temperature (K) 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 -2 0 2 4 6 8 10 12 Reaction rate ( s -1 ) 10 -4 Normalized weight (-) F ig. 2 : TG and DTG data of coal sample at 30 K/min. Due to the comple x reaction of the coal, the kinetics of the coal decomposition can not be described by one reaction. Instead of describing the kinetics of coal decomposition, one approach is to use the production of e volv ed gases. By studying the reaction of each gas species indi vidually , the kinetic parameters are obtained. This method gi ves the opportunity to di vide the total kinetics of the coal into se veral reactions of the p yrolysis gases. The sum of the produced gases is equal to the total mass loss of the coal. 3.2 Evolved gas analysis FTIR e xamined the gaseous products of TGA in order to in v estigate the kinetics further . In the test, mid-infrared light w as used to detect the gas molecules. Molecules and functional groups can be identified at specific bands. Furthermore, a real-time measurement was applied to obtain the spectrogram in a three-dimensional way . Fig. 3 is the spectrum of the ev olved g ases at 30 K / min . The three dimensions are absorbance, temperature and wa venumber , respecti vely . The absorbance v aries from 0 to 1, indicating the intensity of infrared light absorbed by the gases at a specific wa venumber . If the absorbance is 0, the infrared light is not absorbed. According to Beer Lambert’ s la w (Euqation. (1)), the gas concentration is proportional to the absorbance. A = ε c l (1) where A is absorbance. ε is molar absorption coef ficient, L / ( mol · cm ) . c is molar concentration, mol / m 3 . l is the path length, cm . The coef ficient is a characteristic constant when the temperature and wa venumber is fix ed. It describes the ability of the species to absorb light in a specific wa velength (King et al., 2002). Therefore, the absorbance can reflect the concentration of each gaseous products. Gaseous products can be identified with their wa venumber . The pyrolysis product of the coal that are detected are CO (2190 cm − 1 ),CO 2 (2345 cm − 1 ),CH 4 (1295 cm − 1 and 2810 cm − 1 ), and light carbohydrates (1450 cm − 1 ). W e assume the light carbohydrates as C 2 H 4 . Molecules such as CH 4 absorbs both bands at 1295 cm − 1 and 2810 cm − 1 , sho wing two peaks in the spectrum. The absorbance peak of each e volv ed gas is mark ed in the spectrum graphic. In Fig. 3, the gas release at the be ginning of the experiment is v ery lo w . It can be observed that the coal starts to release CO 2 at the be ginning, which is slightly earlier than the other gases. Then the CO 2 gradually rises as the temperature increases. The H 2 O is not observed because the coal w as Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 100 pre-processed in the drying o ven. The intensi ve g as release is between 800 K and 1100 K , which is in correspondence with the acti ve mass loss period of the coal. F or CH 4 and C 2 H 4 , the y ha v e relati vely small ranges. The concentration of CO 2 reduces after the intensi v e gas release period and then increases remarkably as a result of tar pyrolysis. At the end of the measurement, the continuous temperature rise enables char decomposition and gas production. Ho wev er , in the study of dust ex- plosion, the coal dust is not fully comb usted, which means that we only need to consider the pyrolysis stage of the coal. F ig. 3 : FTIR thr ee-dimensional spectrum of e volved gas at 30 K / min (Kittinger, 2019). The molar absorption coef ficients of e v olv ed gases (T able. 2) are used to calculate their concentration. The concentration of each gas is calculated and plotted in Fig. 4. It can be observed that the C 2 H 4 and the CO 2 gases ha ve higher concentrations. There is a high release amount of CO and CO 2 after 1200 K due to the decomposition of the char , which will not be discussed in this study . The production of C 2 H 4 and CH 4 are located at the acti ve decomposition temperature range. In this range, the concentration of CO 2 is on the same le vel as C 2 H 4 , thus leading to the f ast mass loss of the coal. Around 1050 K , the high concentrations of CO 2 and C 2 H 4 are the results of the second reaction due to the tar decomposition. T able 2 : Molar abosrption coefficient of e volved gases (Dong et al., 2017). CO L / mol · cm CO 2 L / mol · cm CH 4 L / mol · cm C 2 H 4 L / mol · cm 1146 945 784 612 By rearranging the concentration, the mass loss due to each ev olv ed gas is plotted in a sequence to compare with the coal mass loss, as sho wn in Fig. 5. For e xample, the final v alue of CO at 1400 K is the start v alue of CO 2 at 400 K . Therefore, the final v alue of the last gas, C 2 H 4 , shows the residual coal mass. CO 2 takes the highest mass fraction among the gas products. The start point of production of CO and CO 2 are earlier , which is around 650 K . Each gas will be analyzed to obtain its kinetic data in the follo wing section. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 101 400 500 600 700 800 900 1000 1100 1200 1300 1400 T e m p e r a t u r e (K) 0 1 2 3 4 5 6 E v o l v e d g a s c o n c e n t r a t i o n ( k g / m 3 ) 10 -4 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 N o r m a l i z e d w e ig h t (-) CO Coal C 2 H 4 CO 2 CH 4 F ig. 4 : Evolved gas concentr ation compar es with normalized coal mass loss. 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0 1 1 0 0 1 2 0 0 1 3 0 0 1 4 0 0 Temperature (K) 0.65 0 . 7 0 . 7 5 0 . 8 0 . 8 5 0 . 9 0 . 9 51 CO Coal C 2 H 4 CO 2 CH 4 Normalized weight (-) F ig. 5 : T otal coal mass loss and the mass loss due to e volved gases. 3.3 Evolved gas kinetics F or a mass-related pyrolysis speed r pyr o j ( k g / m 3 · s ) of a gas component i from a single, independent parallel reaction j applies: r pyr o j = A j e x p ( − E j RT )( V max , j − V j ) n j j = 1 , ..., n pyro (2) where V max , j is the mass fraction determined experimentally . For the calculation of the formation rates r pyr o j of the indi vidual independent, parallel pyrolysis reactions ( n pyr o ) were considered at a certain point in time t . A mass fraction of V j from solid is already volatilized to this point in time. These solid fractions that are released by the decomposition reactions of the solid will be represented by the e volv ed-gas mass fractions. The total formation rate of one gas component i results from the sum of the indi vidual independent parallel reactions: Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 102 r pyr o i = ∑ j = 1 r pyr o j i = 1 , ..., n gas (3) In order to implement the kinetic data into simulation models, the kinetics of each gas e v olution is fitted by using parallel gas reaction kinetic equations. T ake the gas specie CH 4 as an e xample, the reaction rate is calculated out of the gas concentration: r CH 4 = A 1 e x p ( − E 1 RT )( V max , 1 − V 1 ) n 1 (4) T ake log arithmic on both sides of the Equation (4) for CH 4 : l n ( r CH 4 ) = l n A 1 + ( − E 1 RT ) + n 1 ∗ ( V max , 1 − V 1 ) (5) The kinetic parameters can be fitted out of Logarithmic-linear method. Due to two reaction peaks being observ ed in the reaction rate of CO 2 (Fig. 4), two parallel reaction models were used in MA TLAB© to calculate its kinetics. The fitting curves are sho wn in Fig. 6, compared with the e xperiment result. Compared to the first peak, the second one is not distinct but essential for describing the tar decomposing stage. The reaction stages are not separated into primary and secondary pyrolysis stages in this method. This simplification will reduce the computation time when considering high solid loading comb ustions. The total mass ov er the temperature range is plot- ted in Fig. 6b. The dif ference between the model and experiment result is about 2%. The models of other gas species are in Fig. 7-9.The gas release o ver time is in good agreement with the e xperimental data. Therefore, the model can predict the gas e volv e profile. The kinetic data of the coal profiled using e volv ed gases determined experimentally is sho wn in T able. 3, where T b = E j / R . T able 3 : Pyr olysis kinetic data of coal (30 K / min). Gas A(1/s) T b ( K ) V max ( k g / k g ) Order , n (-) CO 1.60E+08 14000 0.9778 2 7.00E+03 10500 0.9426 2 CO 2 9.00E+08 15000 0.86 2.4 8.00E+03 10000 0.8353 2 CH 4 1.00E+04 9000 0.781 2 C 2 H 4 1.00E+06 13000 0.6523 1.6 In the modeling of the formation of a single gas component (e.g., CO, CO 2 ), up to two independent parallel reactions j were used in the calculations for coal pyrolysis. Howe v er , if necessary , more inde- pendent parallel reactions can be used to describe the formation of a single pyrolysis g as component. 4 Model implementation In this part, the single-particle model will be set up using the TGA data for the coal particles. The SEM picture of one coal particle sho ws the porous channel on the surface with a width of around 1 µ m . Therefore, the steps for setting up the single-particle model start from implementing the physical property of a single coal particle and then applying the kinetic of particle decomposition into the simulation. The modification of the solv er is based on reactingFoam. It is a transient solver for the turb ulent flow of compressible fluids, including the chemical reactions used for dust comb ustion. The already e xisting solver rhoPorousSimpleF oam is not used because the diameter of the pores is fix ed in the solv er and does not include chemical properties for further reaction modeling. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 103 400 600 800 1000 1200 1400 Temperature (K) -0.5 0 0.5 1 1.5 2 2.5 3 3.5 Reaction rate ( s -1 ) 10 -3 Test Model Reaction 1 Reaction 2 (a) CO 2 total r eaction rate and sub r eaction r ates over temper atur e. 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.02 0.04 0.06 0.08 0.1 0.12 Concentration (kg/m 3 ) Test Model (b) CO 2 mass incr ease over temper atur e. F ig. 6 : Comparison of e volved gas r eaction model with the experimental r esult: CO 2 . 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.5 1 1.5 2 Reaction rate (1/s) 10 -3 Test Model (a) CO total r eaction r ate over temper atur e . 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.01 0.02 0.03 0.04 0.05 0.06 Concentration (kg/m 3 ) Test Model (b) CO mass incr ease over temper atur e. F ig. 7 : Comparison of e volved gas r eaction model with the experimental r esult: CO. 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.5 1 1.5 2 2.5 3 3.5 4 Reaction rate (1/s) 10 -3 Test Model (a) CH 4 total r eaction r ate over temper atur e . 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.01 0.02 0.03 0.04 0.05 0.06 Concentration (kg/m 3 ) Test Model (b) CH 4 mass incr ease over temper atur e. F ig. 8 : Comparison of e volved gas r eaction model with the experimental r esult: CH 4 . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 104 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.5 1 1.5 2 2.5 3 3.5 Reaction rate (1/s) 10 -3 Test Model (a) C 2 H 4 total r eaction r ate over temper atur e . 300 400 500 600 700 800 900 1000 1100 1200 Temperature (K) 0 0.01 0.02 0.03 0.04 0.05 0.06 Concentration (kg/m 3 ) Test Model (b) C 2 H 4 mass incr ease over temper atur e. F ig. 9 : Comparison of e volved gas r eaction model with the experimental r esult:C 2 H 4 . 4.1 T ransport ef fect The heterogeneous reactions and phase transitions in coal particles will lead to a mass fraction change in species. Furthermore, due to the comb ustion, the enthalpy in the gas phase is increased. Therefore the temperature increases. The temperature increase results in pressure increase or decrease in the flo w , thus creating a con vecti v e flow in the g as phase inside the particle. T o model this flow , an Euler- Euler approach was used for the g as phase. The pressure drop is then deri v ed by considering the pore diameter and the v oid volume. The porous structure of the coal particle is assumed as a parallel channel (see in Fig. 1b). The Navier - Stokes equation for a 2D flo w in the channel can be written in the cartesian coordinates on x direction as, u ∂ U ∂ x + v ∂ U ∂ y − ν ∇ 2 U = − 1 ρ ∂ p ∂ x (6) where, u , v are the v elocities on x, y coordinates, m / s , ν is the kinematic viscosity of the fluid, m 2 / s , ρ is the density of the fluid, k g / m 3 . The flo w is assumed on x direction, so v = 0 (see in Fig. 10). For a fully de v eloped flo w in the steady state, u only changes along the y coordinate,and it keeps constant o ver the x coordinates at an y y position.Therefore, dU / d x = 0. First two terms in Equation. (6) will be eliminated. The third term of Equation. (6) can be written as, F ig. 10 : Sc hematic of the fully developed fluid flow in a 2D c hannel. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 105 300 400 500 600 700 800 900 1000 1100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Temperature (K) Time (s) 25 μ m 1000K Tc fluid 25 μ m 1000K Tc solid 25 μ m 1000K Ts fluid 25 μ m 1000K Ts solid (a) 25 µ m 300 400 500 600 700 800 900 1000 1100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Temperature (K) Time (s) 50 µ m 1000K Tc fluid 50 µ m 1000K Tc solid 50 µ m 1000K Ts fluid 50 µ m 1000K Ts solid (b) 50 µ m 300 400 500 600 700 800 900 1000 1100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Temperature (K) Time (s) 125 µ m 1000K Tc fluid 125 µ m 1000K Tc solid 125 µ m 1000K Ts fluid 125 µ m 1000K Ts solid (c) 125 µ m 300 400 500 600 700 800 900 1000 1100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Temperature (K) Time (s) 250 µ m 1000K Tc fluid 250 µ m 1000K Tc solid 250 µ m 1000K Ts fluid 250 µ m 1000K Ts solid (d) 250 µ m F ig. 14 : The temper atur e incr ease over time of four particle sizes: (a)25 µ m;(b)50 µ m;(c)125 µ m;(d)250 µ m. (T bc = − h mix λ s ( T s − 1000 K ) ). Due to the intensi ve heat e xchange, the temperature of the particle increases to 1000 K within a very short time range. The particle also under goes rapid heterogeneous reactions and produces pyrolysis gases. The dif ference between the particle core and solid is less remarkable than in Fig. 12. The temperature gradient in the particle will be further studied in the next steps by using dif ferent heating temperature. When the particle size increases, the time to heat up the particle increase. In Fig. 14d, the temperature increase rate decreases significantly . This is due to the increase of the coal pore size and a lo wer heat transfer between the boundary and the particle. 6.3 0D comparative model In order to e xplore the inner particle temperature gradient ef fect on the prediction of particle reac- tions. A 0D model using MA TLAB was created to model the reaction rate without the temperature gradient. The surf ace temperature increased to 2000 K in the comparison work, which is a more e xtreme situation than the 1000 K in the pre vious part. The v olume a v eraged temperature from the 1D model was used in the model. The kinetic models of the e volv e gases as well as the particle size, the simulation time keep the same as the 1D OpenFO AM model. The reaction rates of the e v olve gases from the 0D and 1D models are compared in Fig. 15. The dots represent the 0D model results.The 0D results are almost the same as the 1D model.Each gas specie reaction rate from dif ferent particle size is compared. When the particle size becomes larger , the maximum reaction rate of particles becomes lo wer . It takes more time for lar ger particles to heat up Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 112 0 0.2 0.4 0.6 0.8 1 Time (s) 0 2 4 6 8 10 12 14 16 Reaction rate (1/s) The gas evolve of CO 0D 25 m SIM 25 m 0D 50 m SIM 50 m 0D 125 m SIM 125 m 0D 250 m SIM 250 m (a) CO 0 0.2 0.4 0.6 0.8 1 Time (s) 0 2 4 6 8 10 12 Reaction rate (1/s) The gas evolve of CO2 0D 25 m SIM 25 m 0D 50 m SIM 50 m 0D 125 m SIM 125 m 0D 250 m SIM 250 m (b) CO 2 0 0.2 0.4 0.6 0.8 1 Time (s) 0 50 100 150 Reaction rate (1/s) The gas evolve of CH4 0D 25 m SIM 25 m 0D 50 m SIM 50 m 0D 125 m SIM 125 m 0D 250 m SIM 250 m (c) CH 4 0 0.2 0.4 0.6 0.8 1 Time (s) 0 5 10 15 20 25 30 35 40 45 Reaction rate (1/s) The gas evolve of C2H4 0D 25 m SIM 25 m 0D 50 m SIM 50 m 0D 125 m SIM 125 m 0D 250 m SIM 250 m (d) C 2 H 4 F ig. 15 : 2000 K shoc k test r esults:comparison of the evolve gas r eaction r ate between 0-D and 1-D mod- els.Each gas e volve is compar ed between differ ent particle sizes.(a)CO; (b)CO 2 ; (c)CH 4 ; (d)C 2 H 4 . and therefore the particles releases gases slo wer . In the deflagration, the CO 2 has the widest reaction peak. And the CH 4 is the least one but with very high maximum reaction rate v alues. Ho we ver , the dots from 0D cases fall on the simulation results e xactly . They do not mak e a remarkable dif ference. Therefore, the temperature gradient inside of the particles which are not larger than 250 µ m can be ignored in the dust deflagration. 7 Conclusions In this paper , a single-particle model was de veloped for the deflagration of coal dust. The TGA e xperiments were conducted to obtain the kinetic data of the coal sample. T wo acti v e pyrolysis phases were identified in the TG curv e: at 700 - 800 K and 900 - 1000 K . The second reaction rate peak is due to the gases tar reaction when the temperature increase to a higher range. In the dust deflagration simulations, the second pyrolysis should not be ignored because it w ould lead to a lower reaction rate and not promising predictions. Therefore, an alternativ e approach by counting the e volv ed gases was used. The e volv ed gases during the pyrolysis of coal are analyzed by the FTIR instrument. In order to implement the kinetic data and e volv ed gas content into the dust deflagration simulations, sub- models of the gas e v olvement are applied. The kinetic data of the gases were obtained, some of which Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 113 are two-function models. The models were implemented into the CFD simulations, which is a two- phase Eulerian model including the particle porosity , heat transport of conduction and con v ection, the pyrolysis of particle and species transport in the porous structure.Finally , a 0D model was created in order to e xplore the inner particle temperature gradient ef fect. The single-particle model sho ws promising results in the prediction of dust decomposition, including the heterogeneous reaction and mass and heat transport phenomenon. From the 0D and 1D simulation results, the temperature gradient does not ef fect strongly on the e volution of the v olatile content of the particle if the particle size is under 250 µ m . The further steps would be using the single-particle model to predict the dust e xplosion process. By using a simplified model, the computation time would be reduced remarkably and a ne w method in predicting the dust explosion process is proposed. Acknowledgements PY ackno wledges the financial support of the China Scholarship Council (CSC, No. 201808420277). The authors are grateful for the computing time granted by the Chair of Thermal Processing T echnol- ogy (TPT), Uni versity of Leoben. Refer ences Badzioch S, Field MA, G. D. (1964). In vestigation of temper atur e variation of thermal conductivity and thermal dif fusivity of coal . Fuel, 43(4):267–272. Dong, M., Zheng, C., Miao, S., Zhang, Y ., Du, Q., W ang, Y ., T ittel, F . K. (2017). Development and measur ements of a mid-infr ar ed multi-gas sensor system for CO, CO2 and CH4 detection . Sensors (Switzerland), 17(10). ISSN 14248220. doi:10.3390/s17102221. Gschaider , B. F . (2009). gr oovybc . King, P . L., V ennemann, T . W ., Hollo w ay , J. R., Hervig, R. L., Lo wenstern, J. B., Forneris, J. F . (2002). Analytical tec hniques for volatiles: A case study using intermediate (andesitic) glasses . American Mineralogist, 87(8-9):1077–1089. ISSN 0003004X. doi:10.2138/am- 2002- 8- 904. Kittinger , F . (2019). Simultaneous thermal analysis with ftir gas analysis . T echnical report, Monta- nuni veristaet Leoben. P an, Y ., Spijk er , C., Raupenstrauch, H. (2020). In vestigations on the ef fect of particle size on dust dispersion in mik e 3 appar atus . In 13th International Symposium on Hazar ds, Pr evention, and Mitigation of Industrial Explosions (ISHPMIE) . Ranz, W ., Marshall (1952). Evaporation fr om dr ops . Chemical Engineering Progress, 48:141–146. Skjold, T . (2003). Selected aspects of turbulence and comb ustion in 20-Litr e e xplosion vessels: De- velopment of Experimental Appar atus and Experimental In vestigation . (June):324. Skjold, T . (2007). Revie w of the DESC pr oject . Journal of Loss Pre vention in the Process Industries, 20(4-6):291–302. ISSN 09504230. doi:10.1016/j.jlp.2007.04.017. Spijker , C., Raupenstrauch, H. (2017). Numerical in vestigation on inner particle effects in ly- copodium/air dust deflagr ations . Journal of Loss Pre vention in the Process Industries, 49:870–879. T omeczek, J., P alugniok, H. (1996). Specific heat capacity and enthalpy of coal pyr olysis at elevated temper atur es . Fuel, 75(9):1089–1093. ISSN 00162361. doi:10.1016/0016- 2361(96)00067- 1. W ittig, K., Golia, A., Nikrityuk, P . A. (2012). 3D numerical study on the influence of particle por osity on heat and fluid flow . Progress in Computational Fluid Dynamics, An International Journal, 12(2/3):207. ISSN 1468-4349. doi:10.1504/PCFD.2012.047463. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 114 14th International Symposium on Hazards, Prevention, and Mitigat ion of Industrial Explosions Braunschweig, GERMANY – July 11-15, 2022 A Numerical Model for the Minimum I gnition Temperature of Dust Cloud s Tengfei Chen a , Jo Van C aneghem b , Jan Degrève c , J an B erghmans d , Filip Verplaetsen c,e , & Maarten Vanier s chot a a KU Leuven, Depa r tm ent of Mechan ical Engin eering, Group T Leuven Cam pus, A. Vesalius straat 13, B - 3000 Leuven, Be l gium b KU Leuven, De partm ent of Materi als Enginee ring, G roup T Leuven Cam pus, A. Vesalius straat 13, B -3000 Leuven, Belg ium c KU Leuven, Dep artment of Chem ical Engineering , Celes tijnenlaan 200F , B - 3001 Leuven, Belgium d KU Leuven, Depa r tm ent of Mechan ical Engin eering, Celesti jnenlaan 30 0A, B - 3001 Leuven, Belg ium e Adinex NV, Br ouwerijstraat 11, B- 2200 Herentals, Belg ium E-mail: tengfe i.chen@kule uve n.be & chentf19946@1 63.com Abstract This paper p resents a numerical model for the calculation of the minimum ignition temperature (MIT) of dust clouds b ased on the Godbert-Greenwald f urnace . The model considers heat transfer between the air and dust partic les, the dust particle re action kinetics, and the re sidence times of dust clouds in the furnace. For all the 13 dust s amples studied, deviations between the calcula ted and the experimental M I T values are within 20% (6.57% o n average). Th ere is also g ood agreement betwe en the ex perimental and the numerical MIT variation trends against dust concentration and p article size. Two different ignition modes are discovered. The first one consists in ignition near the furnace wall for bigger particles char acterized b y r ather short residence ti mes. In the second mode, the ignition starts from the center of t he furnace b y self -heating of the dust cloud fo r smaller particles with longer residence times. For magnesium, as dust concentration increases, the lowest ig nition tempera ture of the dust cloud IT (c onc ) decreases first, then transits to increase at a certain point. The transition happens at different dust concentra tions fo r dif fere nt p article sizes. Moreover, th e M I T of the magnesium dust cloud genera ll y inc reases as particle size increases, but the increa sin g trend stagnates within a certain me dium particle size rang e. Keywords: minimum ignition temperature, dust clouds, numerical mode l 1. Introduction The minimum ignition temperature (M IT) of a dust c loud is a n important parameter for the assessment of the dust cloud ex plosion risks with hot surface i gnition sources in related process industries . Numerous experimental studies on the MI T of dust clouds have been underta ken in recent y ears (Nifuku et al. (2007), Cao et al. (2012 ), Boilard et al. (2013), Mittal (2014), J anès et al.(2014), Miao et al.(2016), Addai et al. (2016a, 2016b, 2017), Mishra et al. (2018), Cao et al. (2019), Deng et al. (2019), Tan et al. (2020), Wang et al. (2020), Sun et al. (2020 ), Ga bel e t al. (2021 ) and Krietsch et a l. (2021) ). The most popular apparatus to measure the M I T is the Godbert-Greenwald furnace (G - G furnace) (Eckhoff, 2019) . With the G-G fur nace, the MIT is defined as th e “ the lowest temperatur e of a hot surface on whic h the most ignitable mix ture of the dust with air is i gnited under specified t est conditions ” (I SO/ IEC 80079-20-2, 2016). The te sts have to be performed under varia tion of bot h dispersion pressure and dust concentration until the MIT is found. Although there are theoretical models developed for the MIT c alculation (Yuan et al. (2012), Addai et al. ( 2016c) and Arsha d et al. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 115 (2021)), thes e models ignore the temperature gradient and evolution insi de a dust cloud du ring the ignition proce ss. In this stud y, a numeri cal model is established for t he M IT calculation of dust clouds based on the G- G furnace, with consideration of the heat transfer between the air and dust particles, the dust particle reac tion kinetics, and the residence time of dust p articles in the furnace. A s mentioned above, G- G furnace is the most popular M IT testing equipment, thus most of the available experimental M I T data are also generated with the G-G furnace. The refore, numerical data with t he model dev eloped b ased on the G-G fu rnace expected to be more comp arable with the ex perime ntal result s. Temperature gradient and evolution in the dust cloud is studied in detail for different dusts and different furnace temperatures, then the calculated M I T data are compared with the ex perime ntal results. F inall y , the influence of du st concent ration and particle size on the MIT is discussed. 2. Methods and materials 2.1 Physical model The ph ysica l model for the M IT calculation of dust clouds in Fig. 1 is based on the G- G fur nace MIT testing equipment illustrated in F i g. 2. I n such a furnace, the dust particles a re not pre-mixed with the hot gas ini tially present in the furnace chamber , but are blown into the chamber with the cold pressurized gas forming an ini tial cold dust cloud with a mbient temperature. The initiall y cold dust cloud is then gradually h eated up du ring its down wards movement. Therefore, the i gnition of a dust cloud not only depends on the furnace tempe rature, but also on the dust cloud residence time in the furnace chamber before t he dust cloud falls out from bottom of the furnace . I f t he dust particles cannot absorb enou gh heat for their ignition during their residence time, then it is like ly that the ignition fails. Fig. 1. Physical model for the MIT calculation r Heat flow Hot boundary condit ion with furnac e temperature Cold boundary condit ion with the ambient temperature Initi al c old dust cloud layer with the ambient temperat ure Space inside t he furnace Dust cloud layer heated up by the furnac e as it moves downwards Furnace heat ing stops af ter the resi dence time Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 116 Fig. 2. Schematic diagram for G-G furnace set up (Tan e t al., 2020 ) In Fig . 1, an initiall y col d dust cloud lay e r in a cylindrical space with the size of the furnace heating chamber is used to simulate the dust cloud ignition process in the G -G furnace MIT test. I deall y , th e dust particles are assumed to be distributed uniformly in the dust cloud lay er . Since dust clouds in the test are not pre-heated before ente ring the h eating chamber , the initial dust cloud laye r temp erature in Fig . 1 is set as the ambient temperature. Durin g the dust cloud residence ti me in the furnac e , the cy lindrical dust cloud lay er boundar y temp erature is set as the furnace h eat ing surf ace tempe rature. Afterwards, the boundary temperature is set to be the ambient temper ature to simulate the scenario after the dust cloud l a y er drops out of the furn ace heating c hamber . I t must b e noted that the numer ical model for the M I T calculation is a on e-dimensional model, so in Fig. 1 the axial direc tion is assumed to be adiabatic. I n the G-G furnace test s, suc cessful ignition of the dust cloud requires a visual observation of the flame. There fore, in this numerical stud y the criterion for successful ignition of the dust cloud is also the occurre nce of i gnition in the dust cloud a nd the propa ga tion of igni tion through the dust cloud. 2.2 Governing equations The dust cloud residence time in the furnace heating chamber can b e estim ated based on the settling velocity of the dust particles obtained by th e Stokes law (Hosseinzadeh, 2018): 𝑣 𝑠 = (𝜌 𝑠 −𝜌 𝑔 )𝑑 𝑝 2 𝑔 18 𝜇 (1) 𝑡 𝑟𝑒𝑠 = 𝐻 𝑓𝑢𝑟 𝑣 𝑠 (2) where 𝜌 𝑠 — pa rticle density , k g/m 3 ; 𝜌 𝑔 — g as d ensity , k g/m 3 ; 𝑑 𝑝 — particle size, m ; 𝑔 — gravity acceleration, 9.81 m/s 2 ; 𝜇 — dynamic viscosit y , N·s/m 2 , calculated with Sutherland's formula as a function of the furnace inner surface tempera ture ( White, F.M., 2006); Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 117 𝑣 𝑠 — dust particle settling velocit y , m/s; 𝑡 𝑟𝑒𝑠 — dust cloud residence time in the G-G furnace , s ; 𝐻 𝑓𝑢𝑟 — height of the G-G furnace, m. It must b e noted that in a r eal dust cloud in the G- G fur nace, the residence ti me can be di f ferent a mong the dust partic les and bet wee n the particles and the ga s ph ase. The residence time is also subject to the dust blowing pre ssure. Therefore, Eq. (1) and Eq. (2) can be vi ewed as an ideal mathematica l expectation of the dust cloud residence time in the furnace. Moreover , the sudden drop of the boundary t emperature aft er the residence time is also not a re alistic setting. I t must be admitted th at the physical model in Fig. 1 is after all a hi ghly sim plified model, focusing on the ver y basic character istics of the dust cloud ignition process in the G -G furnace. Similar to the governing equations of the dust cloud mi nimum ignition energ y (MI E) calculation (Chen et al., 2020; C hen et al., 2021), the MIT calculation in this study is also based upon the heat transfer in the gas phase and from the gas to the d ust partic les. Considering the cy lindri cal shape o f the fur nace, it is assumed here that the heat transfer is mainly in the ra dial direction. This leads to th e following energy ba l ance equation (Taler and Duba, 2006)): 𝜀 𝑔 𝜌 𝑔 𝐶 𝑝_𝑔 𝜕 𝑇 𝑔 (𝑟 ,𝑡 ) 𝜕𝑡 = 1 𝑟 𝜕 𝜕𝑟 (𝜀 𝑔 𝑘 𝑔 𝑟 𝜕 𝑇 𝑔 (𝑟 ,𝑡 ) 𝜕𝑟 ) − ( 1 − 𝜀 𝑔 ) ℎ 𝑠 𝑎 𝑝 (𝑇 𝑔 ( 𝑟 , 𝑡 ) − 𝑇 𝑠 ( 𝑟, 𝑡 ) ) + (1 − 𝜀 𝑔 )𝜔 𝑜𝑟𝑔 𝑄 𝑜𝑟𝑔 (3) with initial conditions: 𝑇 𝑔 ( 𝑟 , 0 ) = 𝑇 𝑎 and boundary conditions: 𝑇 𝑔 (𝑟 𝑓𝑢𝑟 , 𝑡 ≤ 𝑡 𝑟𝑒𝑠 ) = 𝑇 𝑓𝑢𝑟 , 𝑇 𝑔 (𝑟 𝑓𝑢𝑟 , 𝑡 > 𝑡 𝑟𝑒𝑠 ) = 𝑇 𝑎 and 𝜕 𝑇 𝑔 (0,𝑡) 𝜕𝑟 = 0 where 𝜀 𝑔 — volume fraction of the gas phase (in this study i t is the air) in the dust cloud; 𝐶 𝑝_𝑔 — gas specific heat, J·(kg·K) -1 ; 𝑘 𝑔 — g as thermal conductivit y , W ·(m·K) -1 ; 𝑎 𝑝 — 𝑆 𝑉 𝑠 = 6 𝑑 𝑝 , specific e xternal surface area of particle, 1/m; 𝑉 𝑠 — volume of a single particle, 𝑉 𝑠 = 4 3 𝜋 ( 𝑑 𝑝 2 ) 3 ; ℎ 𝑠 — convective heat transfer coefficient, ℎ 𝑠 = 𝑁 𝑢 𝑘 𝑔 𝑑 𝑝 , W·(m 2 ·K) -1 , 𝑁 𝑢 — Nusse lt number, assumed constant for dust partic les (Nu = 2) (Yarin and Hetsroni, 2004); 𝑆 — surface area of a sin gle particle, 𝑆 = 4π ( 𝑑 𝑝 2 ) 2 , m 2 ; 𝑇 𝑔 (𝑟, 𝑡 ) — air temperature, K; 𝑇 𝑠 ( 𝑟, 𝑡 ) — particle te mperature, K; 𝑇 𝑎 — ambient air temperature, 300 K; 𝑇 𝑓𝑢𝑟 — furnace inner surface tempera ture, K; 𝑟 𝑓𝑢𝑟 — furnace inner radius, m; 𝜔 𝑜𝑟𝑔 — volume reac tion rate o f organic dust particles, kg·(m 3 ·s) -1 ; 𝑄 𝑜𝑟𝑔 — heat of combustion per mass of the o rganic material, J/kg. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 118 The evolution of the part icle temperature is governed b y the followin g equation (Yarin and Hetsron i (2004)): 𝜌 𝑠 𝐶 𝑝_𝑠 𝑑𝑇 𝑠 (𝑟 , 𝑡 ) 𝑑𝑡 = ℎ 𝑠 𝑎 𝑝 (𝑇 𝑔 ( 𝑟, 𝑡 ) − 𝑇 𝑠 ( 𝑟 , 𝑡 ) ) + 𝜔 𝑚𝑒 𝑡 𝑄 𝑚𝑒𝑡 (4) With initial condition: 𝑇 𝑠 ( 𝑟 , 0 ) = 𝑇 𝑎 where: 𝐶 𝑝_𝑠 — particle specific heat, J·(kg·K ) -1 ; 𝜔 𝑚𝑒𝑡 — volume reaction rate of metal dust particles, kg·(m 3 ·s) -1 ; 𝑄 𝑚𝑒𝑡 — heat of combustion p er mass of the me t al material, J/kg. It must be noted that it is assumed that for organic dust particles, their comb ustion happens mainl y in the gas phase (Wu et al., 2022), and for m etal p articles their combustion happens usuall y on the particle surf ace (Yuan et al., 2012). Therefore, in this study the heat generation from or ga ni c dust particles 𝜔 𝑜𝑟𝑔 𝑄 𝑜𝑟𝑔 is put in Eq. (3) and from metal dust particles 𝜔 𝑚𝑒𝑡 𝑄 𝑚𝑒𝑡 is put in Eq. (4). With the discret e particle distri bution in space , the a bove e quations ca n only be solved numer icall y . I n this study, the same numer ical approa ch for the MIE c alculation in Chen et al. (2020, 2021) is applied for the MI T calcul ation using the finite difference an d energy balance method. 2.3 Materials The dust materials used for the M I T calculation in this study are listed in Table 1 , along with the corre spondin g G-G furnace p arameters of 𝑟 𝑓𝑢𝑟 and 𝐻 𝑓𝑢𝑟 . Basic properties such as particle densit y , heat of combustion and heat capa city ca n all b e found in Chen et a l. (2021). Table 1 also includes the forms of Arrhenius equations for the volumetric reaction rate 𝜔 from liter ature for all the studied dust materials. Table 1 : Material properties and furnace parameters Mater ial 𝒓 𝒇𝒖𝒓 /m 𝑯 𝒇𝒖𝒓 /m 𝝎 / kg·(m 3 ·s) -1 Magnesium (Nifuku et al., 2007) 0.03 0.52 6 𝜔 𝑀𝑔 = 𝐴 𝑀𝑔 𝜌 𝑔 𝑌 𝑂 2 ,𝑆 𝑎 𝑝 𝑒 −𝐸 𝑀𝑔 𝑅𝑇 𝑠 , 𝐴 𝑀𝑔 — 8.6·10 8 m /s, 𝜌 𝑔 — 1.20 5 kg /m 3 , 𝑌 𝑂 2 , 𝑆 — 0.23 2, 𝐸 𝑀𝑔 — 1 88780 J/mol. (Gure vich et al., 1 968) Aluminum (Nif uku et al., 20 07) 0.03 0.52 6 𝜔 𝐴𝑙 = 𝐴 𝐴𝑙 𝑎 𝑝 𝑒 − 𝐸 𝐴𝑙 𝑅𝑇 𝑠 , 𝐴 𝐴𝑙 — 1.6·10 7 kg·(m 2 ·s) -1 , 𝐸 𝐴𝑙 — 57600lgd p +562600 J/m o l. (Zh ang et al., 2020 ) Zirconium (Cao et al., 20 19) 0.01 95 0.23 𝜔 𝑍𝑟 = 𝐴 𝑍𝑟 𝜌 𝑠 6(𝑑 𝑝 −2𝜀 0 ) 2 𝑑 𝑝 2 (𝑑 𝑝 −2𝜀 0 )𝜀 0 𝑒 −𝐸 𝑍𝑟 𝑅𝑇 𝑠 , 𝐴 𝑍𝑟 — 4.36 1·10 -5 m 2 ·s -1 , 𝐸 𝑍𝑟 — 36400lgd p +266200 J/mol, 𝜀 0 — 1.45 nm. ( Zhang et al., 2020; Bakradze, 2011) Lycopod ium (Addai et al., 2016 a) 0.01 75 0.42 𝜔 𝐿𝑦𝑐𝑜 = 𝐴 𝐿𝑦𝑐𝑜 𝜌 𝑠 𝑒 −𝐸 𝐿𝑦𝑐𝑜 𝑅𝑇 𝑎 , 𝐴 𝐿𝑦𝑐𝑜 — 10 7.04 s -1 , 𝐸 𝐿𝑦𝑐𝑜 — 1 21102 J/mol. (Zhou, 201 3) Corn starch (Addai et al., 20 16a) 0.01 75 0.42 𝜔 𝐶𝑜𝑟𝑛 = 𝐴 𝐶𝑜𝑟𝑛 𝜌 𝑠 𝑒 −𝐸 𝐶𝑜𝑟𝑛 𝑅𝑇 𝑎 , 𝐴 𝐶𝑜𝑟𝑛 — 10 12.2 s -1 , 𝐸 𝐶𝑜𝑟𝑛 — 218650 J /mol. (Zhou, 201 3) Mahogany wood (Chen et al., 2 016) 0.01 95 0.23 𝜔 𝑊𝑜𝑜𝑑 = 𝐴 𝑊𝑜𝑜𝑑 𝜌 𝑠 𝑒 −𝐸 𝑊𝑜𝑜 𝑑 𝑅𝑇 𝑎 , 𝐴 𝑊𝑜𝑜𝑑 — 3. 04·10 6 s -1 , 𝐸 𝑊𝑜𝑜𝑑 — 1 04800 J/mol. (W en et al., 2004) 𝐴 — pre -exponential factor ; 𝐸 — activ ation energy; 𝑌 𝑂 2 ,𝑆 — m a ss fractio n o f the oxygen o n metal surface; 𝑅 — universal gas co nstant, 8.31 4 J/(mol·K); 𝜀 0 — initial o xide thickness of metal par ticles. It must be noted that in Cao et al. (2019) and Chen et al. (2016), 𝑟 𝑓𝑢𝑟 and 𝐻 𝑓𝑢𝑟 of the furnace used for their tests are not clearl y st ated. However, th e commonl y us ed G -G f urnac e has 𝑟 𝑓𝑢𝑟 of around Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 119 0.0195 m, and 𝐻 𝑓𝑢𝑟 of appr oximately 0.23 m with a volume of around 0.27 L (Mittal, 2014). Therefore, in thi s stud y, these common values for 𝑟 𝑓𝑢𝑟 and 𝐻 𝑓𝑢𝑟 are assumed for the calculation of the zirconium powder (Cao et al., 2019) and mahogany wood powder (Chen et al., 2016). 3. Results and discussion 3.1 Temperature profiles Fig ure 3 and 4 show gas t empera ture prof iles of a 29 µm and a 90 µm magnesium dust cloud at critical furnace temperatures, re spectivel y. During the i gnition proc ess, the gas phas e acts as heat ca rrier from the furnace to the particles and the heat exchan ger betwe en partic les, so here the gas temper ature profiles are used to help in understanding the ignition process. Cle arly , the 250 g/m 3 29 µm magnesium dust cloud can be successfull y i gnited with 𝑇 𝑓𝑢𝑟 of 762 K which gives rise to a sudd en temperature jum p . While with 𝑇 𝑓𝑢𝑟 of 761 K, su ch su dden temperature jump cannot be obs erved in the dust cloud so the ig nition fails at this temp erature. Therefore, the MIT of 250 g/m 3 29 µm magnesium dust cloud is 762 K. Similarly , based on F ig. 4 , th e M IT of 1000 g/m 3 90 µm magne sium dust cloud is 829 K. Comparison between Fi g. 3(a) a nd Fig. 4( a) also re veals some interesting aspects regarding the ignition process. Figure. 3(a ) refers to smaller p articles of 29 µm characterized by a rather longer residence ti me of 23.2 s (Eq. (1) and (2)) with 𝑇 𝑓 𝑢𝑟 = 762 K. It shows that by the time of 23.2 s, gas temperature profiles across the dust cloud alread y rise to the same level of the furnace tem p era ture. Then even a fter 23.2 s when the dust cloud laye r falls out from the furnace , the self-he ating from the chemical reaction of dust particles sti ll enables a ther mal runaway at the ce nter of the dust cloud. Afterwards, the ignition propag ates from the cent er outwards th roughout the dust cloud. On the oth er hand, Fig. 4(a) refers to the bigger particles of 90 µm with shorter residence time of 2.54 s with 𝑇 𝑓𝑢𝑟 = 829 K. As a result, b y the time of 2.54 s, Fig. 4(a) shows that at radial positions with 𝑟 < 0.02 m the gas temperature profiles sti ll stay near the ambient level. Therefor e, the self-heating of the dust cloud is rather insi gnificant with such a sho rt residence time, and the i g ni tion starts from p articles near the hot furnace bo undary and then propagates inwards to the cent er of the dust cloud. In conclusion, for dust clou ds with shorter residence ti mes in the furnac e, su cce ssful i gnition depends on the thermal ig nition of dust particles ne ar the ho t furnace boundar y withi n the residenc e time. For dust clouds with long er residence times in the furnace, su ccessful ignition r elies more on whe ther the self-hea tin g of the dust cloud during ( and even after) the residence time can tri gger a therm al explosion inside the dust cloud. It must be noted that in this numerica l model there is a sudden drop of the boundary temper ature of the dust cloud f ro m the furnace temperature to the ambient temperature. This is a ver y ideal assumption. I n reality , the drop o f the bound ary te mperature takes time, so the assumption of sudden drop of the bounda ry temperature can result in r ather high MIT predictio ns. However, on the othe r hand, after the dust cloud drops out of the furnace, in reality the dust clou d will potentiall y expand since the physical restriction of the furnace wall is lifted. This ex pansion factor is a disadvantage for the heat accumulation in the dust cloud, but it is also not considered in the numerical model, so thi s will lead to potential lower M I T predictions. It must be admitted that the ph y sical behaviors of the dust cloud during the MIT test is rather complex , with the current simple numerical model some influencing fa ctors ca nnot be revealed. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 120 a. 𝑇 𝑓𝑢𝑟 = 762 K b. 𝑇 𝑓𝑢𝑟 = 761 K Fig. 3. Gas phase temperature profi les of 250 g/m 3 29 µm magnesium dust cloud with 𝑇 𝑓𝑢𝑟 of 762 K and 761 K a. 𝑇 𝑓𝑢𝑟 = 829 K Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 121 14th International Symposium on Hazards, Prevention, and Mitigat ion of Industrial Explosions Braunschweig, GERMANY – July 11 - 15 , 2022 Acceleration s ensitivity of pi ezoelectric pressure sen sors and the influe nce on the me asurement of explosion pressures Tim Krause a , Harun K anbur a , Niels Springer a , Je ns Brunzendorf a , De tlev Markus a , Otto Walch b & Christian Heer c a Physikalisch-Te chnische Bundesanstal t (PTB), Bra unschwe ig, Germany b R. STAH L Schaltgeräte GmbH, W al denbu r g, Germa ny c Kistler Inst rumente AG, Winterthur, Switze rland E-mail: tim.k [email protected] Abstract To measure the explosion pre ssure inside an enclos ure, it is common to install a piezo electric pressure sensor in the enclosure w all. The pre ssure wave of the internal e xplosion inevitably leads to vibrations of the enclosure walls. This unwanted but naturally occurring motion is also transmitted to the pressure sensor mounted in the enclosure wall and results in iner tial forces a ffec ting the piezoelectric element. Dur ing the me asurement o f the explosion pressure, this affects the output signal of the pressure sensor since an undesired signal due to the acceler ation of the pressure sensor is superimposed on the de si red pr essure signal. This behaviour of the s ensor is describe d as acceleration sensitivity. The level of acceleration sensitivity depends on the type and c onstruction design of the pressure sensor. Even though this sensor behaviour is basically not a new phenomenon, the evaluation of an int ernationa l comp arison between Ex testing la boratories in the field of flameproof enclosure s has shown that the consi deration of this issue is a major challenge in daily practice concerning the measurement of explosion pressures and is even often completely neglected. This work evaluates the behaviour o f various piezoelectric pr essure sen sors with respect to the influence of acc eleration and investigates the specific impact on the e xplosi on pressure mea surement in the field of flamep roof enclosures. For this purpose, ex plosions fr om typically used explosive mixtures such as hydrogen, propane and ethyne in air are exa mined. These investigations involve simple model enclosures with various specifications as well as a commercially available equipment for ha zardous areas. By using blind holes and s pecially designed adapters, a pra ctical method is applied to be able to detect the effect of acceleration on the sensor signal separately from the pressure signal. F or this purpose, both the discrete-time pressure curves and the fr equency compon ents are analysed using Fast Four ier Transform. The use of signal filters as a practical and fast approach to address these unwanted s ignal components is discussed and e valuated. This paper p rovides gui delines for typical end -users in the field o f flam eproof enclosures ho w t o handle acceleration of piez oelectric pressure sensors and the influence on the measurement o f explosion pressures c orrectly. Keywords: flam eproof enclosures, explosion pressure, piezoelectric pressure sensor, a cceleration sensitivity Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 128 1. Introduction Equipment intended for u se in potentially explosiv e atmospheres (Ex equip ment) may only be plac ed on the mark et if it is certified to meet th e spec i al technical requirements o f the c orresponding type o f protection as defined by e.g., IEC standards of the IEC 60079 series. Ex testing laboratories conduct various tests a ccording to th ese standards on test samples provided by the manufacturer. For Ex equipment of the type of prote ction flameproof enclosures, the ability of the enclosure to withstan d the pressure of an internal explosion in accordance with I EC 60079-1 (IEC, 2014) is one of the main properties to be tested. Here t he basis of these tests is the determination of the explosion pressure, which acts as a reference pressure fo r further ov erpressure tests. These ove rpressure tests ar e executed with 1.5 to 4 times of the refe rence pressure, dep ending on the certification procedure . The correct measurement of the explosi on pressure as reference pr essure is th erefore importa nt, as any measurement devi ation is mul tiplied subsequently. The analysis of an international comparison between Ex testing laboratories on this subject as well as current discu ssions from the field of certification show that the influence o f ac celeration re spectively the acceleration sensitivity of the pressure sensors used i s an insufficiently consid ered influencing factor when determining the explosion pressure. The explosion pressure inside a flameproof enclosure inevitably leads to a vibration of the enclosure walls (Krause, 2017). Th e deflection of this vibration depends on the material and wall thickness, as well as on the level of th e explosion pressure (Spörhase, 2021). The movement of the enclosure wall is a lso transmitted to th e pressure se nsor mount ed in it and leads to inertial forces acting on the piezoelectric element. This can influe nce the outpu t signal of the pressure sensor during the explosion pressure measurement since an undesired signal due to the acceleration of the pressure sensor is superimposed on the desi red pressure signal (Tichý , 1980). This behaviour of piezoelectric p ressure sensors is called acceleration sensitivity and is usua lly differe ntiated according to the directions of action "axial" and " radial". The respective level of acceleration sensit ivity depends on the construction design of the pressure sensor. The inf luence of acceleration for specific applications and the compensation of a cce leration -related effects are addr essed in several scientific studies (Ren, 2013). In addition to methods for changing the sensor design with re gard to the piezoelectric element (Wang, 2021), t here are also approaches to compensate acceleration -relate d influ ences with digital filters based on suitable mathematica l models (Xu, 2019). Howe ver, these pa rtially promising approac hes are not universally applicable, as they often refer to specific applications, might produce incorrect re sults due to incomplete discrimination betwee n the acceleration-induced and the pr essure- induced signals, may involve very complex mathe matical calculations or are limited e.g., mechanical adjustments to improve a ce rtain property cause another property to deteri orate. The reason fo r the insuf ficient attention to this issue is often the lack of awareness and thus of recognising the problem in the first place. The fact that in practice there are many different pressu re curves in the field of fl ameproof enclosure due to changing test param eters such as enclosure geometry, combustible-air mixture, sensor position, sensor type, etc., intensifies the challenge of detecting an influence of acceleration on the pressure sensor that causes an altered pressure signal. The aim of this work is ther efore to discuss practic al methods to detect acceleration influences on th e explosion pressure measurement in general and to minimise or eliminate these unwanted influen ces accordingly. 2. Experimental setup The experiments performed are based on the IEC 60079-1 standard rega rding the test requirements for deter mining explosi on pressure for flameproof enclosures. The experiments are initially performe d with all the c ombustible-air mixtures and pre ssure sensors described in the following subsections. Later, the consideration of the combusti ble is reduced to hydrogen-air mixtures since the influence of acceleration is most significa nt here. Also, the focus regarding the pressure sensors is laid on those types with the highest and lowest acceleration sensitivity. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 129 2.1 Test setup and test samples The sch ematic r epresentation of the experimental setup can b e se en in Figure 1. Two diff erent test samples are considered. The first one (included in Figure 1) is a simple model sample of a flamepr oof enclosure consisting of a combination of two c ylindrical chambers . The chambers are mad e of stainless steel of the sa me inner diame ter (Ø 1 61.5 mm) with differe nt lengths (short cha mber l = 250 mm; long chamber l = 500 mm) conn ected by an orifi ce plate with a 15 mm hole in the centre . The long chamber is closed with a stainless steel fl ange of 22 mm thickness a nd represents the ignition side, with a centrally loc ated spark plug on the flange. The short chamber represents the pre ssure measurement side, which is also closed by a fla nge. To vary the effect of a cce leration forces both a 22 mm thick stainless steel flange and a 16 mm thick a luminium flange a re examined. As shown in Figure 2 for pressure me asurement a centrally located threaded hole for th e pressure sensor is used on each of the two va rying flanges . I n addition, a blind hole with the sam e depth is a pplied slightly offset for both flanges (without connection to the e xplosion volume). Another pressure sensor of the same type is installed in this blind hole to provide the output signal without the actual pressure signal, but only the influence of the ac celeration on th e pressure sensor. As a se cond test sample, a real Ex equipment consist ing of a M CC B (Molded Case Circuit Br eaker) in the design o f a flameproof enclosure is used (see Figure 3). The enclosure is made of glass fibre reinforced polyamide (PA 66) with external dimensions of 254 mm x 149 mm x 189 mm (length x width x height). The spark plug as igni tion source is located in the lower area of the length side. The pressure sensor for pressure measurement is installed by means of a threaded hole on the front side. Due to the thin wall thickness of the enc losure, a blind hole for the use of a second pressure sensor to measure th e influence of acceleration is not practical. Furthermore, due to the complex geometry without many straight surfaces, it is also not possible to use a second measuring point in a comparable position to the pressure sensor. For this reason, instea d of a bl ind hole, a specially designed adapter with a cl osed front is used for the determination of the acceleration influence directly with the threaded hole of the actual pressure measu rement. The combustible s used for the experiments are hydrog en (31 ± 1 vol. % H 2 ), propane (4.6 ± 0.3 vol. % C 3 H 8 ) and ethyne (14 ± 1 vol. % C 2 H 2 ) in air. The resp ective combust ible-air mixture is purged into the test sample by using mass flow controllers and the concentration ratio is verified by an oxygen ana lyser. After the igni tion of the combustible-air mixture within the test sample by the spark plug, th e pressure c urve of the e xplosion is recorded by piezoelectric pressu re sensors. The pressure sensors g enerate a charge in linear Fig. 1. Schematic represe ntation of the experimental setup including the test sample consisting of cylindrical chambers, orifice plat e and exchangeable flanges (16 mm alu minium flange / 22 mm stainless steel flange) Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 130 Fig. 2. Details of the constructional implementation of pressure measurement and the us e of a pressure sensor to investi gate acc eleration influen ces: a) Section of through hole and blind hole (top view), b) Section of cross -sectional view for throu gh hole and blind hole, c) Pressure sensor install ed in blind hole, d) Pressure sensor installed in through hole. correlation to the pressure which is converted into a voltage by a charge amplifier ( Type: Kistler LabAmp 5167). This voltage signal is re corded with a transient recorder before it is processed further using P ython software. T he pressure sensors used for determining the acceleration influences in the blind hole or when using the closed adapters wo rk with the same measuring cha in. During signal acquisition, the sensor signals are filtered with a low -pass filter with a cut-off frequency of 20 kHz (Butterworth, 2nd order) in the charge amplifier to pre vent aliasing effects (Grünigen, 2004). F urther filters such as the low-pass filter with a 3 dB point of 5 kHz ±0.5 kHz , which is standard a ccording to standa rd I EC 60079-1, or the band-stop filters that are use d in these investigations are realised in the post -processing via P ython. The sampling rate for all me asurements is 10 MS/s. To protect the pressure sensors from thermal shock effects caused by the explosion h eat, the sensor membranes are prepared with a 1 mm layer of RTV silicone (Krause, 2021). To exclude the influe nce of vari ations in temperature on the explosion pressure and the vibrations of the enclos ure walls, the tests were conducted within the ra nge of 25 °C to 28 °C with temperature monitoring. Fig. 3. Second test sample: “ Real Ex e quipment ” consisting of an MCCB (Molded Case Circuit Breaker) designed as a fl ameproof enclosure with gas inle t (green marking), ignition source (orange marking) and pre ssure se nsor (yellow marking). (Gas outlet not visible on the back). All units in mm. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 131 2.2 Piezoelectric pressure sensors The pressure se nsors con sidered in this work are all piez oelectric pressure sensors since piezoelectric pressure measurement t echnology is widespread in the determination of explosion p ressures according to I EC 60097-1 and represents the current state of the technology. Piezoresistive pressure sensors are of minor rele vance for the target grou p considere d in these inv estigations, as they have further practical disadv antages in addition to the lower flexibility regarding measurement ranges (Krause, 2021). This is a lso clea rly reflected in the feedback from the Ex testing labora tories participating in the international comparison regarding the pressure m easurement te chnology used. All partic ipants exclusively use d piezoelectric pre ssure sensors. The types of pressure sensors considered in this work are listed in Table 1 together with re lev ant tech nical specifications. The design and principle of piezoelectric pressure sensors ar e compa rable across manufacturers. T o exclude possible influences due to different man ufacturing quality, sensors from one manufacturer are used. For these pres sure sensors measurements to determine the acceleration sensitivity were conducted prior to the a ctual experimental explo sion tests to measure the explosion pressures. For this purpose, two pressure sensors of eac h type were measured in axial direction three times by a shaker (Type: Tira S522 0) for six frequency ranges (100 Hz, 1 kH z, 2 k Hz, 3 kHz, 4 kHz, 5 kHz). The experimental setup is shown in Figure 4 and the ave raged results in Figure 5. Table 1: Pressure sensors to be investigated with relevant tec hnical specifications (Kistler, 2021) Type of sensor Pressure range in bar Natural frequency in kHz Sensitivity (ty p .) in pC/bar Kistler 601 CAA 0 … 25 0 >215 - 37 Kistler 601 H 0 … 10 00 ≈150 - 16 Kistler 603 CAA 0 … 10 00 >500 -5 Kistler 603 1 0 … 25 0 ≈160 - 14 Fig. 4. Pressure sensors during the determination of acceleration sensitivity by using a shaker Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 132 Fig. 5. Acceleration sensitivity for different pressure sensors and different freque ncies in mbar p er unit of the free fall acceleration g = 9.81 m/s 2 . In the results for the acceleration se nsitivity for the pre ssure sensors considered in F igure 5, the 601CAA sensor with 𝑎  acc. sen, 601CAA = 1.8 mbar /g has the highest averaged a cceleration sensitivity for the fr equencies consider ed followed by sensor 601H with 𝑎  acc, sen, 601H = 0. 8 mbar/g. The senso rs 603CAA with 𝑎  acc., sen, 603CAA = 0.07 mbar /g and 6031 with 𝑎  acc., se n, 6031 = 0.05 mbar/g have an accelera tion compensated design with lower acceleration sensitivity in a comparable range. 3. Acce l eration sensitivity on piezoelec tric pressure sensors The behaviour of the pressure sensor types u nder consideration is investigated with varying accelera tion. Figure 6 sho ws the results of the explosion pressure measurement of the pressure sensors for the two different flange types and thr ee different combustib le -air mixtures. The e xplosion pressures r esult in each c ase from the mea n values of five individual measurements. The results for a hydrogen-air mixture usi ng the 22 mm stainless stee l flange show comparable explosion pressures with p ex , 601CA A = 12.37 bar, p ex , 601H = 13.06 ba r, p ex , 603CAA = 12.51 bar and p ex , 6031 = 11.79 bar within the expected scatter fo r all four pressure sensor types and are all within the expected order of magnitude. The standa rd deviations are also in a comparable r ange with σ 601H = 0.45 bar, σ 603CAA = 0.40 bar, σ 6031 = 0.43 bar. Th e standard deviation of the 601CAA is slightly smaller with σ 601CAA = 0.27 bar. Changing the flange to the 1 6 mm aluminium flange while keeping all other specifica tions when performing the experiments should produce comparable explosion pressures as expected, since none of the para meters influencing the explosion pressure change. While this expectation is fulfilled for the pr essure s ensor types 601H, 603CAA and 6031 (with p ex, 601H = 12.57 bar, p e x, 603CAA = 12.74 bar and p ex, 6031 = 11.88 bar), it can be obs erve d for th e pressure sensor type 601 CAA that the re sult for the explosion pressure is about 32 % higher with p ex , 601CAA = 16.31 bar. I n addition to this, the standa rd deviation has more than doubled, which is a lso not observed for the othe r pre ssu re s ensors in this scale. A similar but less pronounced picture also arises for the results of the prop ane-air mixture and the ethyne-air mixture when considering the two pressure s ensors 601CA A (with the highest acceleration sensitivity) and the 6031 (with the lowest accelera tion s ensitivity). While the explosion pressures for the 6031 sensors are in a comparable range within the scatter for both flange types (for C 3 H 8 : p ex , 6031, 22 mm = 15.81 bar, p ex , 6 031, 16 mm = 15.80 bar; Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 133 Fig. 6. Explosion pressur es of different pressure sensor types for two flange types and three differe nt combustible-air mixtures. for C 2 H 2 : p ex , 6031, 22 mm = 17.22 bar, p ex , 6031, 16 mm = 18.31 bar), a mor e signif icant incr ease occurs fo r the sensors 601CAA when changing to the 16 mm flange (for C 3 H 8 : p ex , 601CAA, 22 mm = 16.67 bar, p ex , 601CAA, 16 mm = 18.14 bar; for C 2 H 2 : p ex , 601CAA, 22 mm = 18.48 bar, p ex , 601CA A,16 mm = 20.89 bar). To investigate the r eason f or this behaviour, it is useful to analyse the pressure curves for the configurations considered. For this purpo se, F igure 7 shows the pressure curves for the pressure sensors of all four pressure sensor typ es for a hydrogen -air mixture for both flanges. Since the avera ging of the pressure curves from the five individual measurements of ea ch pressure sensor type is not meaningful due to time shifts, the pressure curve of an indivi dual expe riment is used as an example for the re spective pressure senso r. Diagram a) shows the output signals of the pressure sensors installed in the centre through hole and in direc t contact with the explosion volume using the 22 mm stainless steel flange. Diagram b) shows the output signals of the pressure sensors a ssigned to the same e xperiment, w hich are installed in the blind hole and where ther e is no contact to the explosion volum e - thus the sensor is exposed to the same mech anical vibrations, but not to the explosion pressure. It can be seen that, corresponding to the comparable explosion pressures, the pressure curves of all four pressure sensors also show almost identical cha racteristics. For the pressure sensors in the bli nd holes , there are als o signal curves that do not reflect the explosion pre ssure curve but result from the effect of acceleration on the pressure s ensors. The relatively low output signals vary depending on the sensor between approx. ± 1 bar for sensor 601CAA and approx. + 0.34 bar/- 0.02 bar for sensor 6031. The se nsor signal of 603CAA is in a sim ilarly low range as the signal of sensor 6031. The sensor signal of 601H is with + 0.6 bar/- 0.32 bar between the values of sensors 6031 and 601CAA. The level of the output signals here refle cts the accelera tion sensit ivities of the respective sensors ( see F igu re 5). The higher the a cceleration sensitivity of the pressure sensor in the blind hole, the hig her the corresponding ou tput signal. However, the direct influen ce on the explosion pressures as well as the pressure curve s is small, which is shown by the compara ble results. The a nalysis of the pressure curves and the output signals of the pressure sensors mounted in the blind holes using the 16 mm aluminium flange comes to a differe nt result. In a ccordance with the pre vious discussion of diagrams a) and b), the pressure cu rves of the pressure sensors in the through hole (diagram c)) and the out put signals of the pr essure sensors in the blind h oles (diagram d)) are als o shown in the diagr ams of Figure 7. As discussed fo r Figure 6, comparable va lues result for the explo - Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 134 Fig. 7. Pressure curves ( diagram a): 22 mm stainless steel fl ange and diagram c): 16 mm aluminium flange) and output signa ls of the pressure sensors mounted in the blind hole (diagram b): 22 mm stainless steel flange and diagr am d): 16 mm aluminium flange) of different pressure sensor types for a hydrogen-air mixture (31 ± 1 vol. % H 2 ) on the test sample consisting of cylinders and orifice. sion pressures o f the pr essure sensors 601 H, 603CAA and 6031. This is also shown in the pressur e curves considered in diagram c). The signal curv es of th e pressure sensor s 603CAA and 6031 show an almost identica l behaviour as with the 22 mm stainless steel fla nge. The signal curve s of the 601H and 601CAA sensors, in contrast, deviate significantly. While this deviation does not have a noticeable e ffect on the measured valu e for the explosion pressure in the case of the 601H pressure sensor, but is rather characterised by an incr eased undershoot, the sensor signal of the 601CAA is superimposed by a strongly oscillating signal, both in the positive and in the negative direction. An increased explosion pressure erroneously follows from this overshooting. The reason for th e oscillations in the pressure sensors 601H and 601CAA superimposed on the pressure signal bec omes obvious when lookin g at the output signals of the pressure senso rs in the bli nd holes, shown in diagram d). Here it can be seen that the a cceleration in the cas e of the pressure sensor 601H and even more in the case of the pressure sensor 601CAA leads to an oscillation behav iour in the output signal, which is superimposed on the corr esponding pres sure signals in diagram c) in an almost identical curve form. The maximum amplitudes of the output signals of the pressure sensors 6031 and 603CAA installed in the blind hole have approximately tripled compared to the output signals of the 22 mm stainless steel flange an d are in a range of approx. ± 1 bar. With the 601H sensor, the maximum amplitude of the output signal has increa sed by a factor of approx. 9 with + 3.82 bar/- 5.35 bar, with a change o f the maximum amplitude into the negative range. With the pr essure sensor 601CAA, th e largest increase o f the out put signal with + 8.57 bar/ - 10.7 bar oc curs in neg ative amplitude by a facto r of approx . 11 and in posi tive amplitude by app rox. 8.6. These results also reflect the ranking of the accelera tion sensitivity of the pressure senso rs. Furthermore , it can be s een for the configurations considered th at accelerations that lead to output s ignals of ≤ ± 1 bar for the pressure sensors installed in the blind hole do not cause any significant cha nges in the output sign als of the pressure s ensors installed in the centre through hole that reflect the pressure curve. In th e following, the pressure curves of th e experimental results for the propa ne - air mi xture (see Figure 8) and the ethyne-air mixture (see Figure 9) are presented in the same way as in the previously Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 135 Fig. 8. Pressure curves ( diagram a): 22 mm stainless steel fl ange and diagram c): 16 m m aluminium flange) and output signa ls of the pressure sensors mounted in the bl ind hole (diagram b): 22 mm stainless steel flange and diagr am d): 16 mm aluminium flange) of different pressure sensor types for a propane-air mixture (4.6 ± 0.3 vol. % C 3 H 8 ) on the test sample consisting of cylinders and orifice. performe d analysis of the hydrogen-air mixtures but reduc ed to the se nsors 601CAA (high accelera tion s ensitivity) and 6031 (low acceleration sensitivity). For the propane-air mixture in Figure 8, it ca n be se en in genera l that the pressure curve differs from a pressure curve after ignition Fig. 9. Pressure curves ( diagram a): 22 mm stainless steel fl ange and diagram c): 16 m m aluminium flange) and output signa ls of the pressure sensors mounted in the bl ind hole (diagram b): 22 mm stainless steel flange and diagr am d): 16 mm aluminium flange) of different pressure sensor types for an ethyne-air mixture (14 ± 1 vol. % C 2 H 2 ) on the test sample c onsisting of cylinders and orifice. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 136 of a hydrogen-air mixture. Both the pressure rise tim e and the deca y of the pressure are significantly slower, which is due to the lower laminar combusti on speed of propane . Whil e the overall explosion pressure is higher, the re flective pr essure wave superimposed on the glob al pressure waveform is visible but significantly r educed. The explosion pre ssures of the two pressure sensors 601CAA and 6031 shown in diagra m a) are of a comparable order of magnitude with p ex , 601CAA, 22 mm = 16.67 bar and p ex , 6031, 22 mm = 15.71 bar within the scatter. The maximum amplitudes o f the output signals of the pressure sensors installed in the blind holes (di agram b)) are small with + 0.46 ba r for the sensor 601CAA and with + 0.14 bar for the sensor 6031 and about half as larg e compared to the hyd rogen - air mixture. Also, here the sit uation changes with the transition to a 16 mm aluminium flange. Diagram c) shows that the explosion pressure increases by approx. 11 % to p ex , 601CAA, 16 mm = 18.51 bar wh en using the pressure senso r 601CAA, whi le a comparable value is achieve d with the pressure sensor 6031 with a c hange of a pprox. - 0.5 % to p ex , 6031, 16 mm = 15.63 bar . With the 601CAA pressure sensor, an osc illating signal of higher frequency is super imposed on the actual pressure signal, which lea ds to the increa sed a mplitude. The signal curve of the 6031 is almost unchanged, only a sli ghtly increased oscillation on the pressure curve can b e seen. I n diagr am d) the output signals of the pressure sensors installed in the blind hole also show the ca use of the signal superimposition. Due to an oscillation of the flange and a re sulting ac celeration of the pressure sensor, an amplitude of + 3.68 bar/- 3.22 bar r esults for the sensor 601CAA and a value range of + 1.50 bar/- 0.96 bar for the sensor 6031. For the sensor 601CAA this results in an increase of th e output signal by a fac tor of approx. 8 for the positive range as well as a deflection int o the nega tive range for the fir st time. T he fac torial increase fo r sensor 6031 is of an ord er of 10. When analysing the data of an e thyne- air mixture in Figure 9, a similar behaviour can be seen as with the two combustible-air mixtures investigated previously. Using the 22 mm stainless steel flange (see diagram a)), the two pressure sensor types 601CAA with p ex , 601CAA, 22 mm = 18.84 ba r a nd 6031 with p ex , 6031, 22 mm = 17.22 ba r give compa rable expl osion pressures and curve characteristics. Th e reflective pressu re w ave superimposed on the glo bal pr essure is cle arly vi sible, an imposed signal due to accelera tion is not im mediately rec ognisable here. The maximum amplitudes of the output signals of the pressure sensors when using the blind hole (se e diagram b)) are in the ra nge of ± 1 bar. Both the pre ssure curve s and the maximum ampli tudes of the output signals of the sensors in the blind hole are similar to the results of the hyd rogen-air mixture. Diagr ams c) a nd d) show the signal curves when using a 16 mm alumi nium flange. Here, p artly clear changes in the signal character istics are recognisable. Ac cording to diagram c) the ex plosion pressures for t he sensor 601CAA with p ex , 601CAA, 16 mm = 20.89 bar have increased by approx. 13 % compared to the result with the 22 mm stainless steel flange. Fo r the sensor 6031, with p ex , 6031, 16 mm = 18.31 bar, there is also an inc rease, albeit smaller, of about 6 % in comparison. The p ressure cu rves show clear signal overla ys, especially for the 601CAA sensor, which is the reason for the incr ease in explosion pressure . The cause of these signal overlays also results here from the observation of the output signals of the pressure sensors in the blind hole, shown in diagram d). Due to the accelerati on of the pressure sensors, the amplitudes of the s ensor signal for the 601CAA sensor are in the ra nge of + 6.45 bar/- 6.85 bar, i.e., about 6 to 7 times higher than for the 22 mm stainless steel flange. The amplitudes of the sensor signal of the 6031 ar e in the range of + 1.8 3 bar/- 2.89 bar, i.e., increased by a factor of 2-3. I t c an be clearly seen tha t the signal curves of the output signals of the sensors in the blind hole are also reflec ted in the pressure curves of the sensors. According to the lev el o f th e acceleration sensitiviti es of the pressure senso rs, the signal overlay is stronger for sensor 601CAA than for sensor 6031. It can be stated that all pressure sensors examined in this sec tion a re affected by ac celeration due to the vibration of the enclosure wall. The level of this acceleration influence depends on the one hand on the real acceleration l evel due to th e explosion effect (the higher the explosion pressure inc reas e over ti me ( dp / dt ) ex the higher the acceleration) a nd the vibration behaviour of the e nclosure wall (flange material and thickness). On the other hand, the acceleration sensitivity due to the respective sensor design plays a decisive role . This acce lerati on influe nce can be represented very well by using the pre ssure sensor with a blind hole without connection to the explosion volume. Even though there Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 137 Fig. 15. Frequency spectra and pressure curves for the pressure sensor 601CAA for the 16 mm aluminium flange in comparison to the 22 mm stainless steel flange using the filter adapting method for a hydrogen-air mixture (31 ± 1 vol. % H 2 ) (diagrams a) and b), propan e-air mixture (4.6 ± 0.3 vol. % C 3 H 8 ) (diagrams c) and d)) and an ethyne-air mixture (14 ± 1 vol. % C 2 H 2) (diagrams e) and f) ) with a),c) and e ): frequency spectra of the pressure se nsors from the blind holes after filter adapting ; and b) ,d) an f ): filter ed pressure curve of the 601CAA. detected between the sig nal curves, which is a ch ara cteristic property o f the applied filter func tions . In analogy, improved matche s of the frequency spe ctra and signal character isti cs can also be achieved for the propan e-air mixtures using a band-stop filter for the frequency range 1.8 kHz to 2.5 kHz (Butterworth, 2nd order) in addition to the standard filter (see di agrams c ) and d) ) as well as for the ethyne-air-mixture applying two filter applications consisting of a band-stop filter (fre quency r ange: 1.9 kHz to 3.2 kHz; Butt erworth, 2nd order) and a low -pass filter (cut-off frequency: 4 kHz, Bessel, 2nd order) (see di agrams e) and f)). In summ ary, it can b e stated th at for the configurations considered, when v arying three combustible-air mi xtures, it is possible by adapting the filter functions used to reduce the acceleration components influencing the pressure signal to such an extent that they no longer significantly alter the pressure signal. For this method, i t is also necessary to record th e signal curve of a pressure sensor in parallel, which has no direct connection to the explosion volume (blind hole) and thus only represents the effect of the acceleration of the s ensor in the sensor signal . Due to the large number of possible pressure curves resulting from the almost infinite variations in the boundary conditions of the test with regard to different geometries, c ombustible-air mixtures, pressure sensors, filter options, etc., it is not possible to specify a particula r filter for this procedure Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 144 in advanc e. It is a lways necessary to analyze to wh at extent the acceleration sensitivity of the pressure sensor affec ts the p ressure signal and subsequently apply an appropriate f il ter function. 6. Influence of acceler atio n on the pressure signal using the example of a r eal Ex equipment In this part of the work, the knowle dge gained from the previous investi gations is applied to determine the explosion pressure on a real Ex equipment w ith substantial acce leration of the enclosure walls. For this purpose, the se cond test sample (see F igure 3) and the exp erimental setup specified i n section 2.1 are used. I n this case, as described, instead of using a second pr essure sensor o f the same type for det ermining the influence of acceleration in the blind hole, a second measurement is conducted with the identical sensor, but using an adapter with a closed fr ont. Figure 16 shows the sensor signals for sensor types 601CAA and 6031 for the use of a hydrogen -air mixture, both for the through hole for measuring the pressure and for the closed adapter to d etermine the a cceleration effects. Dia gram a) shows that the explosion of the hydrogen-air mixture inside the test sample leads to differe nt pressure curves and explosion pressures depe nding on the type of pressure sensor used. A c lear oscillation signal is superimposed on t he pre ssure curve of the 601CAA, which with p ex, 601CAA = 7.67 bar leads to an explosion pressure increased by app rox. 38% compared to the 6031 with p ex, 6031 = 5.56 bar. The frequency spe ctra assigned to the pressure curves in diag ram c) als o cause o f this superimposed oscillation becomes evident when looking at t he sensor signals of th e pressure sensors using the closed adapter (see dia gram b)). For the 601CA A sensor, the acceleration influences lead to higher signal oscillations due to the hi gher acceleration sensitivity than for the accelera tion compensated 6031 sensor. Th e co rresponding frequency sp ectrum in diagram d) also shows the frequency com ponent with increased am plitude of the 601CAA in the range around 1 kH z. In order to consider th e identified acceleration influences on the press ure signal of the sensor 601CAA, the second method of adapting the filter function is a pplied. In this ca se, the use of the first method discussed for subtracting the frequency spectra does not lead to a suf fic ient reduction in the influence of acceleration on the sensor signal. Th e reason for this is that the pressure me asurement Fig. 16. Pressure curves and frequen cy spectra f or the pressure sensors 601CAA and 6031 for th e second test sample “real Ex equipment” for a hydrogen -air mixture (31 ± 1 vol. % H 2 ); diagram a): Pressure curves of the pressure sensors; diagram b): Signal c urves of the pressure sensors with closed adapter; diagram c): Frequency spectra of the pressure curves of the pressure sensors; diagram d): Frequency spectra of the signal curves of the pressure sensors with closed adapter. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 145 using the through hole is not mea sured in para ll el with the sensor signal usi ng the closed adapter, du e to the enclosure geometr y. This always leads to slightly differe nt signal curves due to differe nt explosions. The filter fu nction adapted using the signal curve from the sensor in the closed adapter and its freque ncy spectrum is applied to the pressure signal mea sured by the 601CAA. In addition to the standard low-pass filter with a 3 dB point at 5 kHz (Butterworth, 2nd or der), a band -stop filter is used in the frequency range from 0.6 kHz to 3 kHz (Bessel, 2nd o rder). The resulting new signal curves for the pressure sensor (diagram a )) and the sensor for determining the acc eleration (diagram b)) are shown in Figure 17. The influe nce of the acceleration on the sensor signal is significantly reduced by adapting the filter function. The ex plosion pressure with p ex, 601CAA, a dj. = 5.58 bar is now in the comparable range to the explosion pressure of p ex, 6031 = 5.56 bar measured by the sensor 6031. In addition, the consistenc y of the characteristics o f the signal curv es are im proved compared to th e original one s (s ee Figure 16). Also, for re al Ex equipment without idealized housing geometry, it is shown that with the help of a procedure for determining the acce lerati on influences on the sensor type used (here by using the closed adapter) and a corresponding adaptation of the filter functions, the acceleration influences affecting the pressure signal can be re duced. 7. Conclusions When me asuring explosi on pressures, vibrations of the enclosure walls occurring as a result of the explosion c an be trans mitted to the pre ssu re sensors used a nd, depending on the acceleration sensitivity of the sensor, alter the measurement result. This can lead to an unwanted signal being superimposed on the de sired pressure signal. I n this pape r it is demonstrated that this unwanted additional signal can be of the order o f several bar. If thi s parasitic signal component remains undetected, this can lead to incorrect test results and hence to higher r equirements, e.g., for t esting and certifica tion of Ex equipment. Thus, it is first im portant to be able to deter mine that there is a n unwanted influence on the sensor signal due to a cceleration effects. To d etermine this influence, a Fig. 17. Pre s sure c ur ves for the pre ssure sensor 601CAA for the test sampl e “real Ex equipment” in comparison to sensor 6031 using the fi lter adapti ng method for a hydrogen -air mixture (31 ± 1 vol. % H 2 ) in diagram a) and the associated sensor signals of the identical sensors using closed adapters in diagram b) Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 146 simple and practical procedure is to install a pressure sensor of th e same type close to the actual measuring location, but without contact to the explosion volume. This ca n be in the for m of a blind hole or a closed adapter, depending on the test samp le characteristics. With th is information, measures can be taken to reduce the influen ce of the acceleration sensit ivity of pressure sensors on th e measurement of explosion pressure s. The simplest, but rare ly practicabl e option is to use appropriate pressure sensors with whic h the problem doe s not occur (low acce leration sensitivity). How ever, this depends on the leve l of acceleration and is therefore not a general solution. Especially since accelera tion sensitivity is not the only decisive technical specification of relevance for pressure sensors. The othe r more universal and thus more practica l possibility is to eliminate the acceleration contributions via the subtraction of the f requency spectra or to apply individu al filter func tions that specifica lly filter out the acceler ation components. Which method is more appropriate depends on th e test parameters and must be decided individually. The difficulty regarding the determination of explosion pressures in th e context of flameproof enclosures according to IEC 60079 -1 is that only a low-pass filter with a 3 dB point of 5 kHz ±0.5 kHz shall be used to smooth the signal. However, this filter specification does not solve the problem; on the contrary, it severely limits the possi bilities for meeting the challenge w ith the methods presented in thi s work. The authors therefore recommend that the acc eleration sensitivity of pre ssure sensors be considered when de termining explosion pressures by a) using pr essure senso rs that are insensitive to acceleration or b) adapting the strict specifica tion o f the filter in the IEC 60079 -1 standard. Whe n adjusting the f ilter specifications, c are must be taken to e nsure that thi s does not lead to a signi ficant underestimation of the actual explosion pressure. Reference s Internationa l Electrotechnical Comm ission (2014 ), IEC 60079-1, Explosive Atmosphere s - Part 1: Equipment Protection by Flameproof Enclosures “d” , Edition 7.0 Krause, T., Bewersdorff, J., Markus, D. (2017), Investigations of static and dynamic stresses of flameproof enclosures , Journa l of Loss P reve ntion in the Process Industries , Volume 49, Part B, pp. 775-784, Elsevier; doi: https://doi.org/10.1016/j.jlp.2017.04.015 Spörhase, S. , Brombach, F., Eckhardt, F., K rause, T., Markus, D., Küstner, B. , Walch, O. (2021 ), Untersuchungen zur Vergleichbarkeit der statisc hen und dyna mischen Ü berdruckprüfung v on druckfesten Kapselungen , Paper submitted to Forschung im Ingenieur wesen in September 2021, Springer; ISSN (elec troni c): 1434-086 Tichý, J., Gautschi, G. (1980), Piezoelektrische Meßtechnik , Springer-Verlag; ISBN 3-540-09448 Ren, Z., Jia, Z., Zhang, J., Shang, Y., Gao, S. (2013), Research on the Dynamic Error and Acceleration Compe nsation for the Piezoelec tric Sensor, Sensors & Transduce rs, Vol. 153, Issue 6, pp. 118-123; I SSN: 17 26-5479 (Online) Wang, G., Li, Y., Cui, H., Yang, X., Yang, C., Chen, N. (2 021), Acceleration self -compensation mechanism and experimental research on shock wave piezoel ectric pressure sensor , Me chanical Systems and Signal Processing, Volume 150, Elsevier; doi: https://doi.org/10.1016/j.ymssp.2020.107303 Xu, F ., Ma, T. (2019), Modeling and Studying Ac celeration-Induced Effe cts of Piezoelectric Pressure Sensors Using System Identification Theory , Sensors (Basel), 19(5): 1052; doi: 10.3390/s19051052 Grüningen, D. (2004), Digitale Signalverarbeitung , 3. Auflage, Fachbuchverlag Leipzig; ISBN: 3-446-22861-6 Krause, T., Meier, M., Brunzendorf, J. (2021), Influence of thermal shock of piezoelectric pressure sensors on the measurement of explosion pressures , Journal of Loss Prevention in the Process Industrie s, Volume 71, E lsevier; doi: https://doi.org/10.1016/j.jlp.2021.104523 Kistler (2021), Data shee ts piezoelectric presser sensors type 601CAA, 601H, 603CAA, 6031 , Kistler Group, Switzerland Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 147 A comparative study between two ignition sources: electric igniter versus pyrotechnic igniter Chayma EL G ADHA, St éphane BERN ARD and Mame WILLIAM-LOUIS La boratoire PR I SME EA 4229, IUT de Bourges, University of Orleans, INSA -CVL, France E-m ail: stephane .bernard@univ-orlean s.fr chay ma.el-gadha@ univ-orleans.f r Abstract The risk assessment of combustible e x plosive dust is base d on the de termination of the probability of dust dispersion, the identification of potential ig nit ion sources and the e valuation of explosi on severity . It is achieve d in most of c ases with the two main experimental normalized devices such as the Hart mann tube (sp ark ignition) and the 20 L spherical bomb (with the 5 kJ pyrotechnic ignitors) . Ignition energ y of the 5kJ ignitor is we ll c alibrated and g en erates a r eproducible ig nition. But on the other hand this ig nition is not punctual and the over pressure produced is nearl y 2 bars. Moreover, the p y rot echnic ig niter accelerates the combustion with multi ignition points in a large volum e and that disturbs the kinetics. I n thi s way, thi s ignition sourc e doe s not allow to analy se the c ombustion products because the composition of the pyrote chnic igniter w as found in the combustion products. This article deals with the comparison of two ignition sources in the 20 L sph eric al bomb. A large pan el of classical ex plosive dusts is studi ed with electrical and p y r otechnic ignitors, in order to e valuate the possibilit y to establish a correspondence between parameters obtained with these two ignition technics. Severity parameters of Aluminium powder, titanium alloy and nicotinic acid CaRo 11 were measured b y using the t wo t y pes of i gnition s y stem in our 20 L sph erical bomb equipped with the Kühner dihedral injector. The maximum explosion overpressure P max and dus t deflagration index K st were m easured in a large range of conce ntration allowing to prop ose correlations between electrical and p y r otechnic i gnition for e ach parameter and each t y p e o f powder . The relevance of these c orrelations wil l be discussed. Keywords : Dust explosions, Pyrotec hnic ignitor, Electric spark ignitor, Aluminium, Titanium, Nicotinic acid. 1. Introduction Safety features ar e essential for determining the risks of explosion when handling combustible dust and for designing safety me asures. This dust safet y -related prop erties are often inspected without c onsidering the influence of the i gnition source as mentioned by ( Askar & S chröder, 2019). The explosion charac teri stics severity m aximum explosion pressure P max and maximum ra te of explosion pressure rise (dp /d t) max are de termined in closed vessels such as the 20 -L-sphere. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 148 Tests are preformed according to the international standards, for example the EN 14034 series (2004-2006). For ex plosion tests on dusts, chemical igniters are p rimarily used (D I N EN 1 4034; ASTM1226). There are stand ardized chemical igniters with defined energies ran ging from 10 00J to 10000J. The igniters ge nerate a large spheri cal flame with a volume over the 20 L bomb, as described by (Proust, et al., 2007), and the i g nition is not punctual but mul tipoint. This last point contradicts one of the o riginal obj ectives of the sphere which was to h ave a central point ignition. Moreover, the py rotechnic igniter accelerates the combustion and disturbs the kinetics. In this wa y , this ignition source does not allow to ana l y ze the c ombustion products because the composition of the pyrotechnic ig niter was found in the combustion products. The need for alternative ignition sources a rises due to several drawbacks of pyrotechnic i gniters, in addition to their high cost and hi gh en ergy inp ut compared to most ignition sources used in practice . This type of ignition source should be less expensiv e, g enerall y available, and allow the operator to use it without a certificate of competenc y . Another i gnition source that fulfils these requirements is the " spark ignition". A previous paper from ( Scheid, et al., 2013) reports the test results of a comparative stud y between two ignition so urces: p yrotechnica l igniter and exploding wire. P max and (dp/dt) max values from 5 different dusts were determined wit h both ignition sources in the range of 100 J to 1 kJ. The P max values d etermined with exploding wire were less than 10% lower tha n values determined according to t he test standard. The influence of the ignition energy o f the igniter on the maximum explosion pressure seems to be almost negligible. In contrast to that (dp/dt) max values deter mined accord ing to the test standa rd le d to 30% higher values. (Spitzer, et al., 2021) present results of a comparative calorimetric and v isual stud y between four different ty p es of ignition sources (Exploding wire, Chemical ig niter, Induction spark and Surface-gap sp ark). The influence of the e lectro de-orientation, dist ance as well as ig nition energy on the reproducibility of th e exploding wire ignite r was tested. For the stud y max imum explosion pressure and maximum explosion pressure rise values from CaRo11, aluminium and titanium dusts were determined with both igni tion sources. Dust explosions are ge nerall y characterized b y complex reac tion mechanisms, which depe nd on the chemical composition of the dust, dust concentrations and flow conditions. (Van der W el, 1993) distinguishes between different reac tions mechanisms depending on whether the re action takes place in th e gas phase by evaporation or a t solid surface in form of g aseous products or solid or liquid material. This paper concerns the c omparison of two i gnition sources in the 20 L sphe rical: ele ctric i gniter versus p y rotechnic i gniter. A large pane l of classical explosive dus ts is studied with these ignitors, in order to evaluate the possibility to establish a correspondence betwee n parameters obtained with these tw o ignition technics. The dusts were s elected such that different combustion mechanisms were considered: Aluminium powder, titanium alloy and nicotinic acid CaRo 11. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 149 2. Mater ials and E quipement 2.1 Materials Aluminium powder used in the pre sent work is a comm erc ial micron sized aluminium powder (purity>99.7%) supplied by the compan y M&C “Métaux & Chimie” a nd referenced F3915. For titanium, Ti6Al4V powders w ere manufactu red by TL S T echnik spezialpulver. The size distribution of these powders was determined with a laser diffraction technique (Malvern).The result is presented in Table 1 . This measure prov ides a statistical anal y sis via D v (10), D v (50) and D v (90). Table 1: Particles size AlF3915 Ti6 Al4V D 10 , µ m 8 11.8 8 D 50 , µ m D 90 ,µ m 35 80 26.4 2 44.3 Fig. 1. SEM photography of pure aluminium Fig. 2. SEM photography of Ti6Al4V 26 μ m The SEM photography Fig. 1 and Fig. 2 showed that AlF3915 and Ti6Al4V particles are spherical. (Millogo, et al., 2018) and (Millogo, et al., 2020) 2.2 20L sphere The explosions were carried out in a sphe rical 20 L ex plosion vessel designed at the P RISME La borator y ( Fig. 3 ) in accordance with the international standard EN 14034 -3 (British Standards I nsti tution, 2006) in order to character ize the ability of powde rs to explode. The parameters mea su red are the maxi mum explosion overpr essure 𝑃 𝑚𝑎𝑥 and the max imum rate of pressure rise in the sphere, ( 𝑑𝑃 𝑑𝑡 ) 𝑚𝑎𝑥 over the explosive range for a g iv en dust. This experimental setup consists of a holl ow sphe re made of allo y ed steel, a dust storage container connected with the chamber throu gh a dust outlet electro valve (K ü hner A G). The dust was injected at the bottom of the tank with the standard compliant d ihedral inj ector. A rebound noz zle (K ü hn er AG) ( Fig. 4 ) disperses the p articles with the air in the chamber ge neratin g a turbulent flow in order to reach ho mogeneous dust dispersi on. At the end o f the injection period, the atmospheric pr essure is reached. The explosi on signal is measure d b y a Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 150 Kistler 701A sensor and the ac quisition is perf ormed by a Keysight digital oscilloscope and recorded on a computer. The tests with the p y rote chnical igniters (5kJ ) as an i gnition source were perf ormed using the standard control unit of t he 20 L sphere. For t ests with electric spark, a spark generator devic e is made b y the labora tor y itself according to the standard requirements, was used. I ts design was detailed in (B ernard, et al., 2010) Fig. 3 . Design of spheri cal chamber of 20 litters volume Fig. 4 . The rebound nozzle 2.3 Ignition source 2.3.1 Ignition processes Ignition of combustible dust clouds occurs onl y in the presence of a flam e source or a suf ficient heat sourc e. In general, ignition is triggere d in a va riety of ways f rom low -energy to hi gh-energy ignition source s (Am y ott e & Eckhoff, 2010). Howeve r, the ignition sourc e differs in terms of power and energ y . The ignition source ca n significantl y influence the d ynamics and course of the scattered dust explosi on (Yuan, et al., 2015). In the industry, th ere can be several t ypes of significa nt ignition sources that can cause dust to ex plode. These are t y pically a spark, a hot surface, ov erheating , di rect fire, etc. These i gnition sources are characterized b y th e Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 151 fundamental parameter of "ignition energ y "(the energ y of the i gnition source that ignites a cloud of dust) and the d y namic s of the ignition process (spa rk is punctual sour ce, fire/combustion is a "slow" source, in term s of dust bu rning rate) , ( Amyotte, 2014, Eckhoff, 2002 , Ku racina, et a l. , 2021). One of the fundamental differences of these ignitors is the ignition mechanism. While for the ignition spark an electric arc is generated the p y r otechnical igniter emits f lames and bu rning solids. 2.3.2 Electrical ignition system The elec tric arc ignition system consists of a high voltage generator, whose discharge initi ates an arc betwee n the electrodes. A generator p rovides the ignition energ y b y capacitor dischar ge , and whose delivered voltage and current, as well as the ar c holdin g time, are adjustable. Th e system produces spark at nearly constant power a nd controls the spark energ y b y controllin g the dura tion of the spark. The spark current is adjustable between 2 and 8 A and wa s s et to 4 A in this study . The a rc voltage i s 82.5 V. As the volt age a nd current int ensity a re consta nt, energy is inly proportional to the spark duration : E = U arc . I ar c x τ arc , t his time could be changed ove r the range of 1µ s-100ms. The arc en ergy value achi eved with such a n arran gement is in the range from 10 mJ to 500 J, making possi ble to measure the ignition energ y of the less ignitable dusts as detailed by (Bernard, e t al., 2010).The e nergy de posited in the discharg e r eaches 66 J. I n this paper we only recall the electric scheme ( Fig. 5 ). The spark generator has, in addition, a "trigger" output (s y n chronized on the beginning o f the c ycle at cyc le at time t 0 ) which allows to trig ger an external device. P ointed electrodes made of tungsten were used with a 2.4 mm diameter. The electrodes’ extremities are shaped at an angle of 40°. This configuration allows minimal erosion of the elec trodes and provides a conical shape to generate th e spark. The e xperimental ignition c onditions for the electrical ignition system are summarized in Table 2 (Bernard, et al., 2017) Table 2 : Ignition condition in the 20 L spherical bomb Parameter Value Spark current intensity ,A 4 Spark power,W 330 Spark energy ,J 66 Electrodes ga p,mm 4 Electrode sha pe Conical (40°) Fig. 5. Electric scheme of the spark gene rator (Bernard, et al., 2012) Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 152 2.3.3 Pyrotechnic I gniter The ASTM E1226-12a standard specifies the exact composition and quantity of th e mixture in a chemical igniter. The chemical igniters consist of small plastic or aluminium buckets filled with a firing cha rge of 40% zirconium, 30% bariu m nitrate and 30% ba rium peroxide. A sealing cap ke eps the firing charge inside An ele ctrical fuse head is connected to two wires for a precise electrically controlled i gnition from the outside of the apparatus (see D IN EN 14034) . The ignition source was pla ced in the middle of the sphere and c onnected to th e electrodes of the lid of the chamber, such as it is described in the test standards. 3. Results and discussion 3.1 Severity parameter results: K st and P max 3.1.1 CaRo11 The calibration of our 20  L sphere was performed by using Caro 11 po wder and compared to “ round robin results ” . The K st max value obtained with the pyrotechnical igniter is well in the range of re sults given b y “ round robin tests ” . While for the electric sp ark igniter, the ratio between round robin measure m ents and our tests is 1.47 (Bernard, et al., 2017). Fo r e ach experimental test, at least two runs were performe d. Fig. 6 pre sent the e volution of the pressure and the k st a s a function of the dust concentration with both ignition sources . The highest v alue of P max (6.64 bara) is obtained at a concentration of 500g/m 3 for the electric spark igniter and for the pyrotechnica l igniter, the highest value of P max (7.99 ba ra) is obtained a t a concentration of 500g/m 3 . While the Kst max for the pyrotechnical igniter was 242bar.m/s at a c oncentration of 750g/m 3 , it was 134 bar.m/s for the electric spark igniter at a concentration of 750g/m 3 . Fig. 6. Severity parameters of CaRo 11 as function of conc entration determined with electric spark igniter and pyrotechnical igniter The curve s were fitted according to a polynomial of order 2 and the K st is written as follows as a function of the conce ntration C: 𝐾 𝑠𝑡 = 𝑎 0 + 𝑎 1 . 𝐶 + 𝑎 2 . 𝐶 2 Where the fitting coefficients for the electrical spark igniter are respectivel y a 0 = -339.03, a 1 =1.4275, a 2 = -1.063 10 - 3 and for the pyrotechnical igniter ar e a 0 =-628.62, a 1 =2.76989 and a 2 =-2.144 10 - 5 . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 153 14th International Symposium on Hazards, Prevention, and Miti gation of Industrial Explosions Braunschweig, GERMANY – July 11-15, 2022 Moderation of Fe dust expl osion by nano - sized Fe 2 O 3 and Fe 3 O 4 powders Y ongzheng Guo a , Kaiyue Ren a , Peng Zhao b ,c , W eixing Huang a , Aizhu W ei a , Dejian W u b ,c * a School of Chemical Engineering, Sichuan Univ ersity , Chengdu 610065, Chin a b Division 2.1 ‘ ‘Ex plosion Protection Gases and Dusts’ ’, Bundesanstalt fü r Materialforschung und - prüfung (BAM), Unter den Eichen 87, D - 12205 Berlin, Germany c Otto von Guericke Univer sity , Universitätsplatz 2, D - 39106 Magdeburg, Germany E-mail : [email protected] Abstract: Iron powder , as one of the most abundant metal fuels that can be used as recyclable carriers of clean energy , is a prom ising alternative to fos sil fuels in a future low - carbon economy . It may pose a potential explosion hazard during the process of processing, storage, tr ansport and reduction/oxidation (r edox). The explosion characteristics of iron dust in air were unde rtaken via a 20 L spherical explosion chamber with an emphasis on minimum explosion concentration (MEC) of iron dust. The alternative method of combustion duration time ( t c ) was used to determine MEC, and compared with the sta ndardized over pressure method . T wo kinds of nano - sized iron oxides (Fe 2 O 3 and Fe 3 O 4 ) were used as inertants to determine the inhibition ef fect of dif ferent oxidation products. The experimental results showed that adding Fe 2 O 3 or Fe 3 O 4 could reduce the explosion severity and sensitivity of iron dust. The MEC data determined by both methods were comparable. The addition of 5 % oxide has obvious inhibition ef fect under 1500 g/m 3 concentration. W ith the increase of oxide concentration to 1 0 %, the inerting ef fect increases, and the MEC of iron dust increases more than 3 times. The increase of dust conc entration will weaken the inerting ef fect. When the concentration increases from 500 g/m 3 to 3000 g/m 3 , the weakening effect of 10 % Fe 2 O 3 on the maximum explosion pressure decreases from 38.45 % to 2.24 %, and 10 % Fe 3 O 4 decreases from 46.21 % to 10.63 %. These results provide a fundamental basis to mitigate the iron dust explosion via solid inerting method without adding extra elements. Keywords: ir on; ir on oxides; MEC; combustion duration tim e; overpr essur e 1. Introduction As a recyclable carrier of clean ener gy , metal fuels are expected to replace fossil fuels in the zero - carbon or extra low - carbon economy of the future ( Ber gthorson 2018 ) . Among all materials, iron has been studied due to its abundance ( Popok et al. 2015; Lissianski et al. 2001 ). Due to the heterogeneous combustion of iron on the surface, its oxidation products a re easil y captured and collected, and then used for reduction and recovery with clean energy such as solar energy to achieve the purpose of green re cycling ( Ber gthorson 2018; Ber gthorson et al. 2015 ). Metal fuels can be ef fectively kept indefinitely if they are protected from humidity and ambient air in airtight containers ( Bar dsley 2008; Shkolnikov et al. 201 1 ), making them part of energy reserves or strategic reserves. Fully understanding the combustion behavior of iron dust is of significance for industrial safety , combustion science and technology . Previous studies have shown that the dust explosion risk is related to many factors such as dust particle siz e ( T ascón 2018 ), concentration ( Li et al. 2016; Zhang et al. 2018 ), dispersion ( Di S arli et al. 2014 ), environmental oxyg en content ( Li et al. 2009 ) and ignition energy ( Going et al. 2000; W u et al. 2009 ). Among them, Sun et al . ( 2001 & 2003 ) conducted a series of studies on iron dust, such as the flame propaga tion mechanism, flame temperature and dust cloud behavior . They pointed out that there is no gas phase fl ame in the combustion of iron dust, and the number density of iron particles in the dust cloud varies with the distance fr om the leading edge of the co mbustion zone. The maxi mum value of number density is Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 256 about 2.6 times larger than that at the region far ahead of the flame, which may affect the measurement of MEC. Gao et al . ( 2017 ) obser ved the high - resolution spectrum of iron combustion products and pointed out that the spectrum was decomposed into four pairs of spin orbit contributions, indicating that the combustion oxides contained Fe 2 O 3 , Fe 3 O 4 and FeO. Dreiz in ( 2000 ) revealed t hrough the phase structure diagram of burning iron that in high concentration of oxygen, oxygen needs to be d iffused twice from th e phase interface to react with fresh iron, and the burnin g rate of iron at high temperature may be limited by the iron oxidat ion kinetics. Danzi et al. ( 2021 ) found that the laminar burning velocity of non - porous dust can be calculated by the maximum explosion pressure rise rate in a purely dif fusive regime. However , contact with air during processing, storage, transportation and reduction/oxidation (redox) is still unavoidable, resulting in a potential explosion hazard. Krietsch et al. ( 2015 ) even pointed out that spontaneous combustion occurs when iron powder par ticle size decreases to nanometer level. The US Chemical Safe ty Board has reported an accident that from January to May 201 1, three iron dust explosions occurred at the Hoeganaes Corporation plant in Gallatin, T enness ee, resulting in four deaths and four injuries ( CSB 2011 ). Therefore, it is necessary to adopt reasonable technology to prevent and mitigate explos ion risk in industrial processes involving combus tible metal dust. The inerting of dust explosion based on the principle of intrinsic safety substi tution and attenuation is the most commonly used met hod. Nitrogen, carbon dioxide and ar gon are common gas inertants ( Li et al. 2009 ). However , the use of gas inertants is often limited by the envir onmen t, which is dif ficult to achieve in open space s and can also cause asphyxiation risks ( Eckhoff 2005 ). Solid inertants can well make up for this defect. Commonly used solid inertants include KCl, CaCO 3 , NaHCO 3, NH 4 H 2 PO 4 , e tc . ( Amyotte 2006; Kuai et al. 201 1a ). How to choose solid inerta nts depends on material properties. Experiments ( Chen et al. 2017; Dai et al. 2020; Dastidar and Amyotte 2002; Going & Snoeys 2002; T aveau et al. 2015 ) show that kaoli n, talcum, silicone oil, MET - L - X and Furex 770 cannot ef fectively inhibit the explosion of aluminum dust while the addition of carbonate and phosphate ef fectively inhi bits the severity of explosion. In particular , metal oxides as combustion products are also a good choice for solid inerting agents ( Bu et al. 2020; Jin et al. 2021 ). At the same time, the addition of metal oxides will not change the composition of products, which is a feature that other organic dusts do not have. In order to b etter understa nd the role of metal oxides in inhibiting metal dust explosion, the explosion characteri stics of iron dust and its oxide dust mixed by 20 L explosion sphere chamber were studied. These results provide a basis for reducing iron dust explosion by solid inerting method without adding additional elements. 2. Experimental 2.1 Experimental Materials The Fe dust and Fe 3 O 4 dust used in the experiment were purchased from Chengdu Shengshi Technology Co., Ltd., and Fe 2 O 3 dust was purchased fr om Chengdu Kelong Chemical Preparation Plant. The particle size distribution of dust obtained by laser particle size analyzer and Zeta nano particle size analyzer is shown in Fig. 1. The surface structure o f the dust samples observed by scanning electron microscopy (SEM) is shown in Fig. 2. It can be seen from the diagram that Fe has a wide particle size distribution, irregular shape and no agglomeration, while Fe 2 O 3 and Fe 3 O 4 have regular spherical shapes and serious agglomeration. In the experiment, the powder was well mixed by stirring, which was used to simulate the pre-mixing of combustible dust and inerting agent in practical work ( Dufaud et al. 2014 ). Before the experiment, the experimental samples were dried at 50 ° C in a vacuum oven for about 2 h to keep the moisture content below 5 wt.% ( ASTM 2007 ). Then the required dust is weighed proportionally and mixed uniformly for testing. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 257 Fig. 1. Particle size distributions of the dust samples: (a) Fe, (b) Fe 2 O 3 , (c) Fe 3 O 4 Fig. 2. SEM images of the dust samples: (a) Fe, (b) Fe 2 O 3 , (c) Fe 3 O 4 2.2 Experimental apparatus, pr ocedur e and explosion criteria The dust explosion oc curs when the dust concentration e xceeds a critical value, and the explosion below that concentration will not propagate. This concent ration is defined as the lower explosion limit (LEL) or minimum explosion c oncentration (MEC) of dust. The European standard EN14034 ( EN 2006 ) gives the classic MEC test method: namely, explosion overpressure method. When the measured explosion overpressure relative to the initial pressure P i is ≥ 0.03MPa, dust suspension ignition or explosion is considered to occur. More detailed operation proce ss is sh own in Fig. 3. A series of combustion tests were carried out on ir on dust using a standard 20L spher ical explosive device as shown in Fig. 4. During the test, the pre -weighed dusts were filled into the d ust container with a volume of 0.6 L, and then were dispersed into the 20-L spherical chamber that had been evacuated to 0.04 MPa with the help of premixed compressed gas mi xture (2 MPa) and ignited by a 5-kJ chemical ignitor ( Taveau et al. 2017 ). For the chemical ignitor used in the exper iment, the measured average explosion pressure was 0.05MPa (i.e., P i =0.05MPa). In our previous study, a MEC discriminant method based on combusti on duration time was proposed and used ( Yuan et al. 2012; Zhao et al. 2020 & 2021; Tan et al. 2019; Wu et al. 2022 ) . Dust explosions are time-dependent flame propagati on processes. When the dust concentration is lower than MEC, the spacing between particles is too large, and the combustion of single particle cannot ignite its surrounding particles, and the combustion flame no longer propagates, so dust explosion will not occur. When the dust concentration reaches MEC, the spacing between particles decreases, a nd the spot fl ame formed by single par ticle combus tion is sufficient to ig nite the surrounding particles, forming a continuous combustion flame (producing flame propagation velocity), which forms a dust explosion. After that, with the increase of dust concentration, the total amount of combustible materials will gradually increase, an d the flame propagation velocity will gradually increase. Therefore, there is a minimum value of flame propagation velocity at MEC, which indicates that MEC can be determined by finding the minimum value of fl ame propagation velocity of dust at low concentration. Using the inverse relationship between minimum flame propagation veloc ity and combustion duration, MEC can be determined by the maximum combustion duration. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 258 Fig. 3. MEC test pr ocedur e Fig. 4. Schematic diagram of the 20 - L spherical explo sion test system Fig. 5 shows the explosion pressure-time ( P - t ) curve of iron dust at the concentration of 350 g/m 3 and the ignition ener gy of 5 kJ. It can be seen that the evolution process of dust explos ion pressure can be divided into three stages. The fi rst stage refers to the p rocess of dust injection from t 1 to t 2 , where t ig is defined as ignition delay time. The second stage refers to the development of dust explosion from t 2 to t 3 , where t c is defined as combustion duration time. After t 3 the explosion ended, the pressure began to decay. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 259 Fig. 5. Typical explosion process curve of iron dust in the standardized 20 -L spheri cal chamber 3. Results and discussion 3.1 Minimum explosion concentration of ir on dust Fig. 6 shows the MEC of iron dust in air measured by two methods, from which it ca n be seen that the MEC measured by the standardized over pressure method is 300 g/m 3 , and measured by the combustion duration time method is 310 g/m 3 . The relative error of the results obtained by the two test methods is within 5 %, indicating that the proposed combustion duration time method is also applicable to the judgment of MEC, and can be used as an alternative ex plosion criterion. Fig. 6. The MEC of ir on powder measur ed by two methods In fact, the combusti on duration time criterion directly determines the MEC from the explosion dynamics, and the test repeatability is less af fected by equipment. The standardized over pressure criterion determines the dust MEC from the thermodynamic point of view , which is related to the total heat released by the explosion. In the low concentration range, the measured e xplosion pressure fluctuates, and the repeatability of the test results is poor ( T an et al. 2019 ). Therefore, the combustion duration time criterion method for testing MEC has better stability and accuracy . Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 260 3.2 Severity of iron dust explosion Fig. 7 shows the explosion characteristic parameters of iron dust at different concentrations. I t can be seen from the fi gure that the maximum explosion pressure ( P max ) and the maximum explosion pressure rise rate ((d P /d t ) m ax ) first increase and then decr ease, wh ich is very close to the experimental results of Clouthier et al . ( 2019 ). The dif fer ence is that the maximum value of P max obtained by Clouthier et al . is at the concentration of 2000 g/m 3 , while thi s experiment is at 2500 g/m 3 , which may be caused by the over - driving ef fect of 10 - kJ ignitor ( T aveau et al. 2017 ). When the iron dust concentration is low , the oxygen content in the spher ical chamber is suf ficient, and the main influencing facto r of P max is iron dust concentration. W ith the increase of iron d ust concentration, the ener gy rel eased by combustion increases, so the P max increases. When the concentration of iron dust reache s about 2500 g/m 3 , the reaction between iron and oxygen is the most suf ficient, and the released ener gy reaches the peak. When the dust concentration is too high, oxygen is relatively insuf ficient, which makes some iron particles cannot be fully burne d. These iron dust will compete for the heat released by combustion, making the net energy released by explosion decreased. Fe 2 O 3 , Fe 3 O 4 and FeO are three oxides of ir on. The calculation shows that the stoichiometric concentration of iron reacted with 20 L air for formation of Fe 2 O 3 (Fe+3/4O 2 →1/2Fe 2 O 3 ) is about 660 g/m 3 , for formation of Fe 3 O 4 (Fe+2/3O 2 →1/3Fe 3 O 4 ) is about 740 g/m 3 and for formation o f FeO (Fe+1/2O 2 →FeO) is about 990 g/m 3 which is far less than the concentration when the P ma x and (d P /d t ) max are the larg est in the experiment. This is because in the experiment, it is assumed that the added dust can be dispersed well in the 20 - L spherical chamber , and it is often difficult to achieve in reality , especially for dust clouds with lar ge density and concentration ( Cashdollar and Zlochower 2007 ), which makes the act ual explosion concentration less than the nominal concentration set. The same phenomenon was observed by Kuai et al. (201 1b). In addition, incomplete combustion will also lead to this phenomenon. Fig. 7. Effect of concentration on explosion characteristics of iron dust: (a) P max , (b) t c and ( d P/ d t) max Combustion duration time ( t c ) monotonically decreases with the increase of concentration in a higher concentration range. The qualitative reason is ment ioned in Section 2.2. Br oumand et al . ( 2013 ) deduced the combustion of dust under small Biot number through mass and energ y conservation equations, and obtained the flame propagation velocity of iron dust: 𝑣 𝑓 2 = 𝜆 𝜌𝑐𝜏 ( 𝐶 𝑑 ,𝑢 𝑄 𝜌𝑐 ( 𝑇 𝑖 − 𝑇 ∞ ) − 1 ) (1) where 𝑣 𝑓 is the flame propagation velocity , Q is the heat of rea ction, 𝐶 𝑑 ,𝑢 is the dust concentration, 𝑇 𝑖 is the ignition temperature of the micron - sized particles, 𝑇 ∞ is the temperature of unburned mi xture, ρ , c , λ are the mixture density , heat capacity and thermal conductivity respectively and τ is the burning time of single particle dust: Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 261 𝜏 = 𝜌𝑑 2 8𝜌 g 𝐷𝑙 𝑛 (1 + 𝑖 𝑌 ∞ ) (2) where d is particle diameter, ρ g is the gas densi ty, D is the gas mass diffusivity, i is the mass stoichiometric fuel –oxidant ratio and Y ∞ is the mass fraction of oxygen in the preheat zone. Assuming th at the products of each experiment is the same, then the burning time of single particl e dust τ is a constant in the current experiments. Therefore, for equation (1), other parameters rema in unchanged except for dust concentration C d,u . The flame propagation velocity can be simplified as: 𝑣 𝑓 2 = A𝐶 𝑑 ,𝑢 + B (3) where A and B are two parameters related to other physical properties of iron dust. For 20 - L spherical chamber, assuming that the char acteristic combustion time is approximately equal to the combustion duration time, then the combustion duration time can be expressed as ( Zhao et al. 2021 ): 𝑡 𝑐 = R vessel 𝑣 𝑓 (4) where R vessel is radius of the spherical chamber . T aking Eq.(4) into Eq.(3) : ( R vessel 𝑡 𝑐 ) 2 = A𝐶 𝑑 ,𝑢 + B (5) In orde r to simplify the calculation, the values of A and B c an be obtained by subst ituting the experimental data with concentrations of 1000 g/m 3 and 3000 g/m 3 into the Eq.(5). The calculation results are shown in T able 1. Fig. 8 shows the distribution of cal culation results and exper imental results. It can also be seen from the calculation results that t c decreases with the increase of concentration when other conditions remain unchanged. Since the dust is im possible to maintain a consistent size and perfect spherical shape, and the simplified calculation is af fected by the deterministic conditions, these will lead to inevitable e rrors. The relative error of the formula is less than 20 % for the concentration range from 500 g/m 3 to 3000 g/m 3 , which is withi n the acceptable range. Table 1: Deterministic conditions of solution and calculation results Deterministic conditi ons of solution R vessel A B C d,u =1000 kg/m 3 , t c =0.2828 s C d,u =3000 kg/m 3 , t c =0.2014 s 17 cm 1.7556 1857.9355 Fig. 8. The distribution of calculation r esul ts and experimental r esults 3.3 The inerting effect of oxides on ir on powder The inhibition effect of oxide on iron dust is shown in Fig. 9. The addition of Fe 2 O 3 and Fe 3 O 4 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 262 can red uce the P max and the (d P /d t ) m ax of iron dust. This is because, firstly, the nano-powders have strong inter-particle force and physical adsorption ( Bu et al. 2020 ). The addition of nano-oxides promotes the agglomeration of iron dust into larger groups, resulting in poor dispersion. What's more, the aggregation makes the oxidant diffuse inward through the pores inside the aggregates. During the ignition process, the penetration depth of the oxidant inside the aggregates decreases rapidly, and the concentration of the oxidant decreases with the consumption of the rea ction particles (Soo et al. 2018) . There may even be a “dead” core inside the aggr egates, where the particles stop the rea ction due to the lack of oxidant. Secondly, the oxide itself has a certain cooling effect and thermal resistance effect, which can absorb the heat generated by partial combustion and increase the flame propagation resistance. To better quantitatively analyze the inhibition effect of iron oxides, a weakening efficiency (W E) is defined as: WE = 𝑃 max ,1 − 𝑃 max,2 𝑃 max,1 (6) where P max ,1 is the maximum explosion pressure before inerting, and P max ,2 is the maximum explosion pressure after inerting. It is worth noting that for the addition of a small amount of oxides, when the concentration of combustible dust increases, the inerting effect decreases, as shown in table 2. When the concentration is lower than 1500 g/m 3 , the addition of oxides decreases the P max by more than 10 %. However , whe n the concentration is higher than 1500 g/m 3 , th e weakening ef fect of oxides decreases to less than 10 % except for 10 % Fe 3 O 4 . This may be because with the increase of dust concentration, the total heat released by combustion becomes larger, the energy proportion absorbed by oxide decreases, and the lar ger explosion pressure makes the turbulence in the 20-L spherical chamber stronger, so that some aggregates are separated and the reaction contact area of iron dust increases again. Of course, it may also be because the dust concentration incre ases , making oxide and iron dust not fully mixed, so that the inerting effect does not appear the desired effect. Table 2: Weakening efficiency of P max by 5 % and 10 % oxides at different concentrations Solid inerts proportion 500 g/m 3 1000 g/m 3 1500 g/m 3 2000 g/m 3 2500 g/m 3 3000 g/m 3 Fe 2 O 3 5% 31.72 % 29.35 % 10.94 % 3.40 % 3.38 % 1.13 % 10% 37.82 % 38.45 % 15.33 % 5.47 % 4.78 % 2.24 % Fe 3 O 4 5% 17.96 % 21.32 % 12.06 % 6.75 % 3.91 % 6.71 % 10% 41.18 % 46.21 % 27.39 % 11.33 % 10.63 % 11.65 % Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 263 Fig. 9. The inerting effect of oxides: (a) on P max (b) on (dP/dt) max (c) on t c As mentioned above, the particle agglomeration becomes stronger and the particle diameter d becomes larger with the addition of nano -oxide. From Eqs. (1), (2) and (4), it can be seen that the increase of particle size will lead to the increase of particle combus tion time, the decrease of fl ame propagation veloc ity, and the overall combustion duration time becomes longer . It should be noted that t c is much shorter than that without oxide in figure 9 -c where the virtual coil comes out. This is because the addition of oxides causes the original explosion concentration to change into non-explosion concentration, which make s the combustion end earlier and forms a smaller t c . The added oxide covers the surface of iron dust, which increases the resist ance of oxygen dif fusion to iron dust ( Chen and Y uen 2003 ), thus slowing down the combustion p rocess and prolonging the combustion duration time. 3.4 Comparison of inhibition effe ct Fig. 10 shows the influence of the same content of oxides on the P max . When 5 % oxide is added, there is a little difference between Fe 2 O 3 and Fe 3 O 4 for high concentration of inerting. However , it is worth noting that the addition of 5 % Fe 2 O 3 makes the MEC of iron dust exceed 1000 g/m 3 , while the addition of 5 % Fe 3 O 4 still makes the MEC of iron dust remain below 1000 g/m 3 . However, when the oxide content is added to 10 %, it is obvious that the inerting effect of Fe 3 O 4 is better than that of Fe 2 O 3 , and both o f them increase the ME C of iron dust to more than three times. 10 % Fe 3 O 4 reduces th e maximum value of P max by 22 kPa, while 10 % Fe 2 O 3 reduces the maximum value of P max by only 9.9 kPa. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 264 Fig. 10. Comparison of iron oxide and ferric oxide inerting effect: (a) 5 % quality percentage, (b) 10 % quality percentage The rea son for this difference may be related to the properties of iron and its oxides. F irst, the combustion temperature of ir on is lower t han its boiling temperature, which makes ir on present the combustion mode of shrinking core combustion ( Ber gthorson et al. 2015 ). Secondly, Fe 2 O 3 is stable in high partial pressures of oxygen, but is unstable and can be reduced to Fe 3 O 4 under low partial pressures of oxygen ( T akeda et al. 2009 ). Finally, the stru cture of Fe 2 O 3 is loose while Fe 3 O 4 is close, and the diffusion of oxygen in Fe 3 O 4 is less than that in Fe 2 O 3 ( Chen and Y uen 2003 ). Therefore, before the oxidant is added, the iron dust first combusts on the surface to for m an Fe 2 O 3 layer, and th en shrinks to th e internal c ombustion. Due to the decrease of oxygen, Fe 3 O 4 layer and FeO laye r are formed in turn. After adding iron oxide, a layer of Fe 2 O 3 is formed on the edge of iron dust. Due to its resistance to oxyg en diffusion, agglo meration and cooling effect, the P max was reduced. When Fe 3 O 4 is added, an oxide film was also formed on the edge of iron dust. Because of its greater resistance to oxygen diffusion, the Fe 2 O 3 layer formed by surface combustion is under low partial pressures of oxygen. On the one hand, low partial pressures of oxygen make the internal combustion advance into the stage of Fe 3 O 4 . On the other hand, the low oxygen parti al pressure causes the generated Fe 2 O 3 layer to decompose into Fe 3 O 4 , which further increases the oxygen diffusion resistance and reduces the total heat generated by combustion. In addition, the combustion heat of three oxides generated by iron combustion is (Gao et al. 2017): Fe ( s ) + 3 4 ⁄ O 2 (s)→ 1 2 Fe 2 O 3 ⁄ (s) ∆H comb 0 =412.1 kJ/mol (Ⅰ ) Fe ( s ) + 2 3 ⁄ O 2 (s)→ 1 3 Fe 3 O 4 ⁄ (s) ∆H comb 0 =372.8 kJ/mol (Ⅱ) Fe ( s ) + 1 2 ⁄ O 2 (s)→FeO(s) ∆H comb 0 =272 kJ/mol (Ⅲ) Mi et al. ( 2022 ) pointed out that the mai n products of iro n dust combustion are Fe 4 O 3 and FeO, Fe 2 O 3 only generates a small layer on the surface, so there may be reaction competition between Fe 4 O 3 and FeO in the combustion process. Therefore, in terms of chemical reaction, the addition of Fe 4 O 3 will weaken the reaction (II) and prom ote the reaction (III), thus maki ng the reaction (III) dominant, which will greatly reduce the total heat released by the iron combustion and reduce the P max . 4. Conclusions The explosion characteristics of iron dus t and the inhibition effect of i ts two oxidation products were studied by using 20L spherical explosion chamber. The results show that the combustion duration time method is not only suitable for carbon -based dust, but also for metal dust, whi ch provides a new idea for the test of the MEC. The addition of Fe 2 O 3 and Fe 3 O 4 c an reduce the explosion severity of iron powder and weaken the combustion progress. The addition of a small Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 265 Fig. 3. Visualisation of the cloud, different DeltaX, as in the legend, test#12 @50 ms. Fig. 4: Visualizati on of t he dust cloud, different DeltaX, as in the legend, test#12 @70 ms Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 272 Fig. 5: Visualizati on of t he dust cloud, different DeltaX, as in the legend, test#12 @125 ms From t he above images, it could be observed ho w the Delta20 is accurate in the final stage of th e dispersion. At the same time , the Delta1 and 5 ar e likely more helpful to detec t the cloud structure at the early stage of dispersion, when the cloud rise velocity is higher. This difference likely oc curs due to the actual Del taX method functioning in the early phase o f the dispersion, when the DeltaX i s higher (x = 20 ), given the higher momentu m of t he dust cloud rise, the turbule nce eddies w o uld tra vel a great distance along with x fr ames (in the ver tical direc tion) and change shap e dur ing the shooting, so they ar e no longer recognizable w hen X gr ows. Figure 6 shows the effect of the DeltaX choice with respe ct to the actual size of pa rticle a nd their visualisation with time evolution. On the contrary, when w e consider the im ages ta ken after 0.1 s, when flow has r educed his init ial momentum, the optimal choice would be Delta 5. As expected, the turbulence scale has significantly decreased with respect t o early stage. Finally, t he images taken a fter about 0.2 seconds from the dispersion, when the cloud is prac tically in free fa ll , are optimally proce ssed by delta 20: turbulence scale has fu rther decreased, eddies are barely visibl e, likely beca use the dust concentration is almost homogeneous. Figure 6 : Schematic representation of DeltaX effect on particle motion and frame acquisition Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 273 The second video-processing step allows detecting the luminance variation on a singl e row of the video fr aming. Thus, obt aining the luminance’s ma ximum, avera ge and stan dard deviation is possible during dust dispersion. G raphic representation is reported in Fig. 7, where a wav eform graph and an intensity history are plotted. From this post-processing, further information is obtained, such as:  An estimate of dust cloud front rise velocity in the tube.  The distribution of the dust clusters in height and time. Cloud front was detected using a “detector value” (the average, maximum or st andard deviation of luminance). In addition, a threshold value and a background v alue (when no particles are present) were defined. When v alues higher than t he detect or threshold are r egistered in the vid eo, the arrival of the cloud front c ould b e identified. Fig. 7. Intensity history and waveform graph representation as in the LabVIEW application This t hreshold is set differe ntly depe nding on the dispersion tests. Finally, rise velocity is estimated as the derivative of the front rise in time. 3. Results and di scussion In the followings, the resul ts of the different approaches adopted fo r the post-treatment are reported: intensity of the luminance of the pixels is firstly analysed as it is, in terms of pe ak and average of the values on the same pixels row in the video frame. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 274 Fig. 8: Intensity luminance peak vs height, De lta2 0 approach for three samples (300 g/m3, 3.5 bar). The intensity evolution i n time shows the cloud concentration history. A lum inance peak is evident at the early disp ersion time (50 ms), while anoth er peak is obse rved after 100 ms fr om dispersion at a higher location in the tube . After 100 m s the dust cloud seems to be more homogeneous in space and time. Any clearly defined peaks could be obs erved f rom this time on. Fig. 8 (right) r eveals ho w the cloud front position and velocity may be studied with this elabora tion. Two approaches were initially adopte d to i dentify the optimal solution regarding t he amount o f information recovered from the e xperimental tes ts: intensity ave rage and intensity peak in ti me, respectively. From F ig. 9 , some considerations may be drawn. Fig. 9. Intensity luminance peak vs. height, Delta20 approach for thr ee samples (te sts#1, 11, 15). The highest peak is in the lower p art of the tu be, below t he electrodes (the st ar represents the luminance effect due to electrode movement; this peak should be brushed i n the elaboration). Intensity is higher for S tarch and Silica tests than Fe-Starch within the first half of the tube, whil e Fe- Starch intensity is higher a fterw ards. A clear se cond pea k i s observed for this test at about 20 cm from the bottom. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 275 Fig. 10: Intensity peak by height vs. time for three samples (tests#1, 11, 15) . The intensity history is f ound if the m aximum value s a ccording to tube height is calculated, as i n Fig. 10. From these data, the maximum peak in intensity is reached for diffe rent samples at different times after the dispersion: maximum dust c loud “ optical concentration” is re ached at different time intervals depending on t he dust sample nature. This first outcome reveals the variabili ty of dus t cloud dyna mics inside the tube in terms of conce ntrations rang e and timing at the moment of ignition. As observed in Danzi et al. (2021), it is ne cessary to optimise the delta frame (DeltaX, where X is the delta fra m e value) based on the local cloud speed or cloud rise steps. When applying diff erent Delta fr ames, the dif ferent i nformation contained i n the intensity graph could be observed in Fig. 11. Fig. 11: Intensity evolution in height (left) and time (right) for test#1, with different DeltaX approaches. The av erage intensity v alue is higher fo r Delta20 up to a bout 22 cm from the tube bott om; afterwards, its value decay rapidly, w hile Delta5 and Delta1 stay constant. If peak value s are considered, Delta20 presents higher values in all dispersion time with respect to other Delta approaches. The interruptions in Fig. 11: for Delta different than 20 are due t o unexp ecte d spikes at maximum intensity value (255) during dispersion time (likely due to lab light frequency oscillations). Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 276 Fig. 12: Standard deviation values of intensity (peak and average of pixels rows) vs. dispersion time, for test#1. A different elabora tion was performed, focusing on the Standard deviation value of intensit y, intended to i dentify some corre spondences between this value and the turbulent structure ch aracteristics in terms of scale and velocity. 3.1 Dust cloud rise evaluation The cloud front is detected with the LabVI EW routi ne as the rise of the lum ina nce threshold is defined a priori (depending on background lumi nance). Th e purpose of the threshold is to distinguish betwee n actual dust particles and fluctuations due t o background or exte rnal lights. I f adequate detection criteria a re adopted ( average, maximum or stand ard deviation values) and the luminance threshold is well-imposed, the cloud rise is extrapolated from video fra mes. Fig. 13: reports the difference betwe en the rise o f an Iron/Maize starch cloud and a S ilica powder cloud, at the same dispersion conditi ons (nominal concentration and press ure). Sil ica po wder seems more easily lifted by the air blow, while fail to rise to the end of the tub e, reaching a quasi-stil l condition and then starting to settle down in the last interval (fr om almost 0. 16 seconds on). Iron-Starc h cloud rise les s steeply, likely reaching the top of the tube, without any settling. The different rise could be influenced by factor s such as density of dusts, dispersion e fficiency, agglomeration/breakage of clusters in the vertical direction. Hence, as estimated here, the dust cloud front reaches the elec trode position at differe nt times for different dusts, thus implying the maximum concentration will occur at diffe rent times close to the elec trodes. The i gnition delays should be related to these discrepancies. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 277 Fig. 13: Cloud rise comparison, estimated with Initial diff erence detector on light int ensity, tests#3 & 11. Rise velocity is calc ulated from the first derivativ e of the c loud height rise in time, Fig. 14 shows the differe nt velocity estimat ion of three different po wders. The compa rison could be m ade among them. Fig. 14: Cloud front rise v elocity, estimated from elaboration, for three differe nt powders (tests#1, 11, 15) Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 278 Figure 15: Cloud rise velocity, test#1, with focus on the early step of dispersion (right). From Figure 15, rise vel ocity oscillations could be observed in the right-h and i mage: this could be corre l ated to the turbulence i ntensity in this interval, which i s greater du e t o the air pulse contribute . This approac h do es not clearly identify the induced turbulence due to the el ectrode obstacles. The turbulen ce course seems to peak in the early phase. At the same tim e, d eca y is observed, until an asymptotic value, near the end of the dispersi on time. Further investigations c oul d support this evaluation, identify turbulent vortexes during the cloud rise, and help estimate the turbulence scale and velocity. 3.2 Dust cloud rise evaluation: comparisons with previous works As reported above, similar experimentations on the fluid-dynamics of Hartman tube could be found in literature, although only minor investigations are pre sent (with respect to 20L apparatus). Data from Hosseinzadeh et al. (2018) could be partially compar ed with the elaboration results from the present work on the particle velocity. If dust cloud rise velocity is compared to particle velocity vs. time in Hosseinzadeh et al., (2018) similar evolution in time could be observed: an early rise up to about 7 m/s and a rapid decrease to values smaller than 1 m/s after about 100 ms from the dispersion. The compar is on is made betwee n the f ront rise velocity obtained by applying the I nitial Difference appro ach with a detector threshold on the average value (see section 3.1); data from Hosseinzadeh et al. (2018) are referred to the single particle ve lo city in 59 runs. Ensemble averaged velocity values reached a maximum value of about 3.9 m/s, which is about one half the maximum value obtained in this work. 4. Conclusions This work aims to define a novel approac h to studying t he cloud behaviour and dynamics inside the Hartman modified tube, adopted fo r t he flammability screening and the MIE determination in the combustible dust explosion risk assessment. It impl ies high-speed movies post-treatment that reveal som e fluid dynamics aspects of the dust cloud :  The rise in time and space  The dust cloud front rise velocity  The distribution of clusters during the tube rise  Information about the turbulence during the test Once the m ethod is finalised (optimisation of delta frame procedure and tun ing conce rning di fferent powders), it will be a val id alternative to m ore time consuming and comp lex methods, such as P IV tracing of particles. Further works will foresee the realisation of videos with a different ca m era setup, i.e., with two differe nt cameras, set at differe nt shooting angle s, to register a wider sectio n of the t ube and r ealise a Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 279 “quasi” 3D framing, close to the central axe of the tube. The images o f the two-camera will be overlapped for this purpose. The cloud pa rticle size distribution need also to be foc used on: a LabVIEW routine wil l be used with a digital imaging elaboration tool to identify the motion and si ze of the c lusters during dust dispersion. The cor relation between cluster size and an associated turbulence sca le wil l be studied. Reference s Danzi, E., Franchini, F., Dufaud, O., Pietra ccini, M., Mar mo, L. (2021). Investigation of the fluid dynamic of the m odified Hartmann tube equipment by high-speed vi deo processing. Chemical Engineering Transactions , 2021, 86, pp. 367–372 Hosseinzadeh, S., Vanierschot, M., Norman, F ., Verplaetsen, F ., Berghmans, J. (2018). Flame propagation and flow field measurements in a Hart mann dust explosion tube. Powder Technology, 323, 346–356. 10.1016/J.POWTEC.2017.10.001 ISO / IEC 80079-20-2: 2016 Explosive atmospher es - P art 2 0-2: Material characteristi cs - Combustible dusts t est m ethods Marmo, L., Danzi, E. (2018), Metal waste dusts from mechanica l workings - explosibilit y paramete rs investigation, Chemical Engineering Transactions, 67, pp. 205–210. 10.3303/C ET1867035 Marmo, L., Ferri, A., & Danzi, E. (2019). Dust explosi on hazar d in the textile indust ry. Jo urnal of Loss Prevention in the Process Industries, 62. https://doi.org/10.1016/ j.jlp.2 019.103935 Marmo, L., S anchirico, R., Di B enedetto, A., Di Sarli, V., Riccio, D., Danzi, E. (2018), S tudy of the explosible properties of textile dusts Journal of Loss P revention in the Proce ss Industries, 54, pp. 110–122. 10.1016/j.jlp.2018.03.003 Murillo, C ., Dufaud, O., Bardin-Monnie r, N., López, O., Munoz, F., & P errin, L. (2013). Dust explosions: CFD m odeling as a tool to characterise the relevant par ameters of the dust dispersion. Chemical Engineering Science, 104, 103–116. 10.1016/J.CES.2013.07.02 Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 280 14th International Symposium on Hazards, Prevention, and Mitigat ion of Industrial Explosions Braunschweig, GERMANY – July 11-15, 2022 A CFD Based Methodology to Design an Explo sion Preven tion System for Li -Ion Based Battery Energ y Storage Sys tem Anil Kapahi, Alberto Alvarez-Rodriguez, Stefan Kraft, Jens Conzen, and Sunil Lakshmipathy Jensen Hughe s, Baltimore, MD USA E-mail: akapah i@jensenh ughes.com Abstract This work developed a performance-based methodology to design a mechanical exhaust ventilation system for explosion prevention in Li-Ion-based stationary battery energy storage systems (BESS). The design methodology consists of identifying the hazard, developing failure scenarios, and providing mitigation measures to detec t the battery gas and maintain its global concentration lower than 25% of the lower flammability limit (LFL) to meet the prescriptive performance criterion of NFPA 69 – Standard on Explosion Prevention Systems. R epre sentative UL 9540A test data is used to define the battery gas composition, release rate, and release duration to describe the failure scenario involving thermal runawa y propagation. In addition, an exemplar BESS enclosure geometry is define d to m odel the failure scenarios using a computational fluid dynamics (CFD) solver. A grid convergence study is performed to estimate the gr id resolution required to perform the CFD analysis. I n addition, s ensitivity studies for diff erent input parameters are performed to understand the impac t of input s on the detec tion times and ventilation perfo rmance. The approach used in this work provides a systematic procedure for the fire protection engineering community to understand the explosion prevention requirement for a BESS installation. The explosion prevention system functionality presented in this work is limited to removing flammable battery ga s generated due to the non-flaring decomposition of batteries. 1. Introduction Energy storage is playing a pivotal role in empowering the dec arbonization of transportation and enabling power grids to function with more re silience. Lithium-Ion based batteries ha ve come a long way from their usa ge in consumer electronics with tens of Wh (watt-hour) capacity to approximate 100 KW h c apacity ba ttery systems i n modern electric vehicles. Decarbonizing the electricity ge neration process is a big issue and critical to supporting the changing landsca pe in the automotive industry. Addressing this issue ensures we do not deal with greenhouse gases at the electricity generation source. Lithium-Ion based energy storage is one of the leading technologies for sustainable a nd emissi on-free energy. The a dvantage of storing green energy, such as solar or wind, during off-peak hours and using it during peak hours is gaining traction as various governments in the world look toward renewable sources of energy. The growth in the energy capacity is tremendous, with the United States having less than 1 GW of large e nergy storage installations in 2019 to adding a capacity of 6 GW in 2021 and forecasted to achieve an additional 9 GW in 2022 [1]. Like many other energy sources, Lithium-Ion based batteries present some hazards related to fire, explosion, and toxic exposure risk. Although the battery technology is considered safe and is continuously improving, the batte ry cells can undergo thermal runway when they experience a short circuit leading to a sudden release of thermal and electrochemical energy to the surroundings. Cyclical thermal/e lectrical loading and unloading, manufac turing defects, a nd thermal, mechanical, or electrical abuse are ma ny reasons that can c ause an exother mic reaction inside the batteries. Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 281 the detectors would actuate and ac tivate the mecha nical exha ust system, simultaneously deactivating the H VAC c ooling system. FDS can calcula t e the changes in said paramete rs over time. The model divides a given geometry into a series of small control volumes. Each of the volumes is then individually eva luated by the model via a serie s o f conservation equa tions for mass and momentum transfer. Once the computations are c ompleted for each time interval, the results are compiled to form the numerical and visual results of the model. The model can also consider the f low of ai r and ga ses through vents and other openings. The FDS model presents an idealize d representatio n of a real compa rtment or enclosure, but one that has proven to be very useful and accurate f or many engin eering applications. The model has been subjec ted to a wide range of engineering scrutiny and comparisons with experimental data. FDS has been deve lop ed by a recognized governm ent authority and is not biased or influence d by any specific financial int erest or association to a particular industry. Documentation of the model, including validation studies, is rea dily available [1 4]. 2.4 Sequence of Operations A sequence of operations is defined for the C FD model that results in the release of battery gas at the beginning of the simulation. The battery ga s disperses in the container and is detected by one of the two hydrogen detectors. The hydrogen detec tio n results in the activation of the e xplos ion prevention system. A sequence of op erations depicting the event and the c orresponding e vent tim e is shown in Table 3 . Table 3: Sequence of operations for the CFD analysis Event Event Time (s) Start of battery ga s release 0 H 2 detection thre shold is reached at one of the two H 2 detectors t 1 Activation of the exha ust syst em t 1 + 10 seconds The exhaust system reaches its full capac ity after a linear ramp of 20 seconds t 1 + 30 seconds Battery gas re lease stops 600 seconds This event timeline is used for a ll of the simulations shown in this work except for the simulations involving the grid convergence study. The grid convergence analysis is limited to the detec tion of battery gas. 3. Modeling Methodology This section provides an overall modeling methodology and a list of scenarios that were modeled. The 3D CAD geometry of the enclosure shown in Figure 5 was imported into FDS via the software PyroSim developed by Thunderhe ad Engineering. The model was augmented with point devices to monitor the hydrogen concentra tion with time. In addition, the HVAC module of F DS w as used to set up the cooling HVAC supply and return nodes. At this point, t he containe r had all of its original features ca ptured that can be used for the CFD analysis. The explosion preve ntion system require d adding further details of a standalone exhaust and supply louvers to the model. The model adde d two suppl y louvers of size 1 ft wide by 2 ft high and a 1ft 2 opening for the e xhaust f an. The exact locations for these ope nings are provided in Figure 6. The exhaust location wa s sele cted to be at the ceiling leve l as battery gas is expected to accumulate at the ceiling level as it is hot and buoya nt. The supply l ocations are provided in the middle of the side Proceedings of the 14th International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions DOI: 10.7795/810.20221124 288 [Document text truncated for crawler view.]