Presentations of the SuPreAM & NewAIMS Workshop "Optimization of Microstructure and Surface Integrity of Steels to Tackle New Challenges in Additive Manufacturing"

TECHNICAL WORKSHOP SuPreAM project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112346 Predictive simulation of finishing operations in steel Additive Manufacturing for optimal Surface integrity SuPreAM and the Tribological Study on Burnished Surface of Additive Manufactured Metal Adrián Travieso Scientific Researcher, Eurecat Technology Centre [email protected] 3 4 SuPreAM objectives WP5 - Surface integrity characterization Project Overview 01 02 03 Vibration Assisted Ball Burnishing (VABB) 04 05 Results 06 SuPreAM activities 4 Project overview 5 PREdictive simulation of finishing operations in steel Additive Manufacturing for optimal SUrface integrity •Coordinator: Eurecat Technology Centre •Call: RFCS-2022 •Start date: 1st July 2023 •End date: 31st December 2026 •Total costs: 2 608 967.35 € (budget), 1 565 380.41 € (grant) 6 •Challenge 1: Limited surface quality and dimensional accuracy of AMed steel components. •Solution 1: SuPreAM develops a predictive model to quantify surface roughness and residual stresses, enabling targeted finishing strategies that enhance surface integrity and meet tight tolerance requirements. •Challenge 2: Poor machinability of AMed steels due to inhomogeneities such as porosity and anisotropic microstructures. •Solution 2: The project investigates tool wear mechanisms and cutting dynamics specific to AMed steels, optimizing machining parameters and tool selection to improve process reliability and part quality. Challenge identification 7 •Challenge 3: High scrap rates and repeated reprocessing during finishing operations, increasing manufacturing costs. •Solution 3:By identifying critical parameters affecting surface integrity and linking them to functional performance, SuPreAM enables first-time-right manufacturing, reducing scrap by up to 35%and defective parts below 2%. •Challenge 4: Limited industrial knowledge on suitable finishing strategies (machining and polishing) for AMed components. •Solution 4: SuPreAM will generate in-depth knowledge on how printing parameters, machining strategies, and polishing processes affect surface integrity—providing industry with clear guidelines for post-processing AMed steels. Challenge identification 8 SuPreAM objectives 9 To optimize the surface integrity of Additive Manufactured and Machined steel components and to reduce manufacturing expenditures at the steel industrial sector through the minimization of the material scrap up to 35% and reducing the number of re-processing loops during finishing operations keeping defective parts below 2% of additive manufactured components. The main objectives to be tackled to this purpose can be summarised as follows: •#OB1: To develop and improve the performance and capabilities of a predictive simulation model of finishing operations in steel Additive Manufacturing (AM). •#OB2: To optimize the surface integrity of steel additive manufactured (AMed) components. Our objectives Predictive simulation of finishing operations in steel Additive Manufacturing for optimal Surface integrity 3D Numerical Modelling of the Cutting Process: Recent Developments PRESENTER NAME: Josep Maria Carbonell COMPANY: CIMNE EMAIL: [email protected]c.edu 2 4 The PFEM applied to the modelling of Machining Problems, solutions and examples Conclusions and Future research lines Key features of the PFEM 01 02 03 04 3 KEY FEATURES OF THE PFEM 4 PFEM SOLUTION ALGORITHM GOOD MESH GOOD MESH Δ𝑡 4 5 Mesh at 𝑡𝑛 1 5 PFEM SOLUTION ALGORITHM 6 Mesh at 𝑡𝑛+1 after FEM computation 2 6 BAD MESH PFEM SOLUTION ALGORITHM 7 1𝑠𝑡 remeshing step: Erase all elements, maintain the nodes 3 7 PFEM SOLUTION ALGORITHM 8 2𝑠𝑡 remeshing step: Delaunay triangulation 4 8 PFEM SOLUTION ALGORITHM 9 3𝑠𝑡 remeshing step: Alpha Shape Method Erase element if: 𝑟 ≥ 𝛼 ∙ ℎ𝑚𝑒𝑎𝑛 𝑟 ℎ 5 9 GOOD MESH PFEM SOLUTION ALGORITHM 10 PFEM SOLVER ❑Advantages: ✓Allows to model severe changes of topology ✓Uses standard FEM theory which means robustness and reliability in the solution ✓Easy coupling with another techniques (e.g. FEM and DEM) 10 17 Meshing the outside domain •Detection of contact and active set •An ancillary interface mesh is created by the Delaunay tessellation using alpha-shapes 17 CONTACT MECHANICS 18 •Based on the interface mesh generated between the shrunken contacting domains. 18 Condensed Lagrange Multipliers or Penalty approach CONTACT DOMAIN METHOD (Hartmann et al.) 19 19 20 20 21 21 22 PROBLEM OF INTEREST : MACHINING 23 MACHINING PROCESS 23 •MACHININIG / CUTTING METAL Material is cut by means of a cutting tool detaching the material and producing different types of chips 24 24 MACHINING PROCESS 25 TEMPERATURE RIGID TOOL DEFORMABLE TOOL SIMULATION OF MACHINING •Continuous chip: Johnson-Cook plasticity model 25 26 Adaptive insertion of particles •Segmented chip: Modified Johnson-Cook plasticity model (TANH) 26 SIMULATION OF MACHINING 33 33 34 •2D Modeling 34 3hr simulation 35 •3D Modeling 35 20 h simulation 36 •2D PFEM 36 + •3D surface topography model HYBRID MODELLING PROCESS [1] Wang, Jianing, et al. "A high efficiency 3D surface topography model for face milling processes." Journal of Manufacturing Processes 107 (2023): 74-87. 37 Surface topography: A1 MODELLING SURFACE TOPOGRAPHY 38 Surface topography: A1 Modeling Experiment [1] Wang, Jianing, et al. "A high efficiency 3D surface topography model for face milling processes." Journal of Manufacturing Processes 107 (2023): 74-87. MODELLING SURFACE TOPOGRAPHY 39 2D surface roundness contour line: A1 Modeling Experiment [1] Wang, Jianing, et al. "A high efficiency 3D surface topography model for face milling processes." Journal of Manufacturing Processes 107 (2023): 74-87. MODELLING SURFACE TOPOGRAPHY 40 ✓PFEM allows for the study of machining by direct simulation: ✓The modelling material deformations and contact ✓Determining cutting forces and temperatures, including: •The thermo-mechanical material behaviour. •The frictional heat generation, effects of lubrication and refrigeration. •The effects of coating and wear on the cutting tool. ✓Current research is focused in: •Robust and efficient 3D PFEM modelling •Hybrid model : 2D/3D PFEM model + 3D surface topography model CONCLUSIONS 40 41 Thank you for your attention Josep Maria Carbonell Puigbó [email protected] Dr. Hadi Bakhshan - Dr. Fernando Rastellini –Prof. Eugenio Oñate 41 TECHNICAL WORKSHOP Optimising steels microstructure and surface integrity to face new challenges in Additive Manufacturing SuPreAM project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112346 8 •Transient Heat Equation •Nonlinear due to temperature-dependent properties. •Purely thermal model: no fluid or plasma dynamics. •Material removal triggered by boiling point Tb. Thermal model for EDM 9 9 •On boundaries not affected by the spark heat input, convective and radiative heat flux: •On the spark boundary, a Gaussian heat flux: •Once the material reaches the boling point, it is removed from the computational domain EDM Governing equations - Boundary conditions 10 10 •Energy distribution ratio: EDM Governing equations - Boundary conditions 11 4 EDM Governing equations Numerical approximation Numerical examples Introduction 01 02 03 04 Conclusions 05 12 12 •Time discretization using a backward differences scheme: •BDF2: •Galerkin variational form of the problem: Numerical approximation –Embedded finite element method 13 13 •Ghost stabilization terms for small cuts instabilities: Numerical approximation –Embedded finite element method 14 14 Numerical approximation –Final stabilized formulation 15 15 •A level set function is used to define the active domain of the problem: Numerical approximation –Level set function and element subintegration 16 16 •We use our in-house adaptive refinement library Refficientlib •Geometrical refinement criteria: Numerical approximation –Adaptive mesh refinement strategy and refinement criteria 17 17 •MPI based parallelization, domain decomposition with adaptivity, load rebalancing •Electrode is represented by means of a CAD file, and from there a distance function to the electrode for each node is built (OCCT library) •Position of the next spark is based on a minimum distance of the active nodes to the electrode (easy to parallelize through gather and scatter operations). •To accelerate simulation time, multiple sparks if they are sufficiently separated and do not affect each other, can be simulated (more involved to parallelize) Numerical approximation –Parallelization and simultaneous sparks simulation 24 4 EDM Governing equations Numerical approximation Numerical examples Introduction 01 02 03 04 Conclusions 05 25 25 An efficient algorithm for EDM process simulation has been presented ✓Thermal simulation ✓Level set based on eliminating boiled volume ✓Embedded finite element method ✓Ghost stabilization ✓Adaptive mesh refinement ✓Coupling with geometrical representation of the electrode by means of CAD ✓Simulataneous simulation of multiple sparks (sufficiently separated) Numerical results have been presented ✓Single spark validation ✓Part-scale simulation The presented strategy is capable of properly representing the EDM process Next steps will include further calibration, residual stress evaluation Conclusions - overview 26 26 •J. Baiges, H. Venghaus, M. Dias. Adaptive Numerical Simulation of Electro Discharge Machining finishing operations using an Embedded Approach. In preparation, 2025 •J. Baiges and C. Bayona. RefficientLib: An efficient load-rebalanced adaptive mesh refinement algorithm for high performance computational physics meshes. SIAM Journal of Scientific Computing 2017 Conclusions - references TECHNICAL WORKSHOP Optimising steels microstructure and surface integrity to face new challenges in Additive Manufacturing SuPreAM project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112346 Predictive simulation of finishing operations in steel Additive Manufacturing for optimal Surface integrity Material characterisation and surface roughness prediction Simon Larsson Associate Professor at the Division of Solid Mechanics, Luleå University of Technology [email protected] 2 Material characterisation 3 Introduction •Characterisation of Additively manufactured materials at elevated strain-rates and temperatures. •Calibration of constitutive model. 4 Constitutive model –Johnson-Cook 𝜎 = 𝐴+𝐵∙ 𝜀𝑛∙ 1 + 𝐶∙ln ሶ𝜀 ሶ𝜀0 ∙ 1 − 𝑇 − 𝑇0 𝑇𝑚− 𝑇0 𝑚 Material constants A Initial yield strength B Flow stress effect n Flow stress effect C Strain -rate effect m Thermal softening effect Variables 𝜺 Equivalent plastic strain ሶ 𝜺 Plastic strain-rate ሶ 𝜺𝟎 Reference plastic strain-rate 𝑻 Temperature 𝑻𝟎 Reference temperature 𝑻m Melt temperature Plastic deformation Strain-rate dependency Thermal softening 5 •Compressive testing of cylindrical samples •Low strain-rate: Gleeble 3800 GTC •High strain-rate: split-Hopkinson pressure bar (SHPB) •3 Temperatures: 20°C, 200°Cand 400°C •4 Strain-rates: quasi-static, 1000 s-1,3000 s-1 and 6000 s-1 Experimental setups 6 Experimental results –quasi-static compression 13 •Machining of 2 sample parts, each tested under: •End milling (smooth & aggressive conditions) •Face milling (smooth & aggressive conditions) •Parameters recorded: •Depth of cut (axial or radial) •Feed rate •Spindle speed (fixed ~995 rpm) •Mean spindle load (Nm) •Surface roughness measured: Sa, Sq (μm) •Dataset: 8 experimental cases total Experiments & Data Collection Face milling End milling Milling case Depth of cut (mm) Feed rate (mm/min) Spindle load (Nm) Sa (µm) Sq (µm) End milling –aggressive 2.0 500 4.06 0.61 0.83 End milling –smooth 1.0 500 0.70 0.69 0.82 Face milling – aggressive 2.5 125 5.26 1.18 1.42 Face milling –smooth 0.5 (×4 passes) 1990 6.47 0.30 0.38 14 •Predictors: •log(depth of cut) •log(feed rate) •Spindle load •Responses: •log(Sa), log(Sq) •Linear regression structure: log 𝑆𝑎 = 𝛽0+ 𝛽1log 𝑑 + 𝛽2log 𝑓 + 𝛽3𝐿𝑠 log 𝑆𝑞 = 𝛾0+ 𝛾1log 𝑑 + 𝛾2log 𝑓 + 𝛾3𝐿𝑠 •Predictions are exponentiated back to μm scale. •Implemented in Python (scikit-learn). •Missing spindle load →replaced by mean spindle load. Model development Inputs Logtransform Regression Exponentiate Predicted Sa, Sq 15 •Performance (training data): •Sa model: R2= 0.994, RMSE = 0.025 μm •Sq model: R2= 0.998, RMSE = 0.015 μm •Model captures main trends well despite very small dataset (8 samples) Results and model performance Predicted vs Actual values for Sa and Sq. Dashed line = perfect agreement. 16 •A simple linear regression model can predict surface roughness from basic milling parameters. •Good fit on limited dataset. •Handles missing inputs (spindle load). •Limitations: only 8 samples →risk of overfitting. •Possible next steps: •Collect more data (different tools, materials, conditions). •Try ridge regression or leave-one-out validation. •Extend model to include spindle speed, tool wear, and material effects. Conclusions & next steps TECHNICAL WORKSHOP SuPreAM project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112346 TECHNICAL WORKSHOP The NewAIMS project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112371 New approach to Additive Manufacturing of Microstructurally Optimised Steels Conceptualisation, study and demonstration of strategies to obtain cost-effective highperformance steel in metal 3D printing processes Eduard Garcia-Llamas, Dr. EURECAT 3 4 NewAims Project NewAims in Practice Expected Outcomes Introduction 01 02 03 04 4 Introduction 5 Opening Question Image 1 and 2: https://www.shutterstock.com/es/search/3d-printed-metal-parts Image 3: https://3dwithus.com/3d-printing-in-medicine Intrincate structures Jet engine Medical implants 12 Project structure Figure 1: Project Methodology conceptualized in three Research Pillars contributing to an overarching goal 13 NewAIMS in Practice 14 Microstructural Analysis via Dilatometry 10-2 10-1 100101102103104105106 0 100 200 300 400 500 600 700 800 900 start transf end transf M A+B B F+P A+F+P MS=190ºC Temperature (ºC) Time (s) Ac3= 828ºC Ac1= 762ºC A 842 214 277 342 395 385 419 483 562 678 720 395 not finished HV1 15 L-PBF Additive manufactured samples Aconity MINI 16 L-PBF Additive manufactured samples Additive Manufacturing Observations Room-temperature builds: •Caused cracking and delamination due to high thermal gradients Preheating (350 °C): •Optimized laser power and scan speed •Eliminated major defects Material-specific observations: •HTCS® : keyhole-induced porosity Outcome: •Crack-free builds achieved •Highlights importance of thermal control and process tuning 17 E-PBF Additive manufactured samples Modified Arcam S12 EB-PBF system (3 kW electron gun, 60 kV, partial He pressure 2×10⁻³ mbar) ⌀90 mm build tank with gravity-fed hopper and integrated rake Samples built on 60×60×10 mm 304 stainless steel plates Temperature monitored via K-type thermocouple 18 Expected Outcomes 19 Outcomes and Impact •Development of two new high-performance steels specifically designed for Additive Manufacturing (AM). •Production of demonstrator tooling parts to validate performance. •Establishment of a clear link between process, microstructure, and performance, which is essential for industrial adoption. •Expected Impact •Enables stronger, more reliable, and cost-effective steel components for AM. •Boosts Europe’s manufacturing competitiveness and supports sustainability by reducing material waste. •Accelerates the industrial use of AM, particularly in tooling and automotive sectors. TECHNICAL WORKSHOP The NewAIMS project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112371 TECHNICAL WORKSHOP The NewAIMS project has received funding from the European Union’s Research Fund for Coal and Steel (RFCS): project num. 101112371 8 Thermal history experienced by the material during the build process Track 1 XTrack 2 Track 3 Track 4 Track 5 Last printed layer T = 350 °C, t = 0.1 s T2= 1250 °C, t = 0.2 s Temperature [°C] Time [s] T = 300 °C, t = 0.1 s T3= 1059 °C, t = 0.2 s T4= 800 °C, t = 0.2 s T5= 700 °C, t = 0.2 T1= 650 °C t = 0.2 s Track1 Track2 Track3 Track4 Track5 HR = 1500 °C/s CR = 1012 °C/s Scheel et al., Additive Manufacturing Letters 6 (2023) 100150 Ji et al., Journal of Materials Processing Tech. 301 (2022) 117452 Ashby et al., Additive Manufacturing 53 (2022) 102669 9 Thermal history experienced by the material during the build process 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 012345678 Temperature [°C] Time [s] T 1 = 714 °C T 2 = 1276 °C T = 286 °C T 3 = 1120 °C T 4 = 868 °C T 5 = 767 °C T = 380 °C V 304-1275 °C = 1478 °C/s V 1201-381 °C = 1001 °C/s V 25-714 °C = 1583 °C/s V 680-286 °C = 914 °C/s V 409-1120 °C = 1169 °C/s V 701-23 °C = 290 °C/s V 1110-377 °C = 1030 °C/s V 407-865 °C = 1058 °C/s V 810-380 °C = 1032 °C/s V 412-767 °C = 839 °C/s T = 377 °C T = 381 °C BAHR DIL 805A Quenching Dilatomer. Hollow cylindrical samples. Maraging Steel. As-cast material 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 012345678910 11 Temperature [ºC] Time [s] Track 5 Track 3 Track 4 Track 2 Track 1 As-cast Microstructure evolution of as-cast Maraging steel under LPBF thermal history 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 012345678910 11 Temperature [ºC] Time [s] Track 5 Track 3 Track 4 Track 2 Track 1 As-cast Microstructure evolution of as-cast Maraging steel under LPBF thermal history 12 Sample Volume percentage (%) HV5 BCT FCC As-cast 75 25 574 Track 1 99 1476 Track 2 96 4315 Track 3 99 1320 Track 4 99 1310 Track 5 99 1310 As-built 96 4372* *value from the intermediate T section of the sample 0 5 10 15 20 25 30 0 100 200 300 400 500 600 Austenite volume fraction [%] Hardness [HV5] Hardness Volume fraction Track 1 XTrack 2 Track 3 Track 4 Track 5 Last printed layer Microstructure evolution of as-cast Maraging steel under LPBF thermal history 13 Low magnification SE image in SEM BF image in TEMHigh magnification SE image in SEM Transverse section Cellular Structure of Maraging Steel Layer thickness is a key parameter to modify the solidification cell size of the as-built samples. Cellular Structure of Maraging Steel [-210] Transverse section Retained Austenite in As-Built Structure Rwp: weighted summation of residual of the least squares fit, Rexp: statistically expected least squares fit, GoF: goodness of fit (sometimes referred as chi-squared); GoF=Rwp/Rexp; a GoF=1.0 means a perfect fitting. Retained Austenite in As-Built Structure BCC FCC Step size = 0.08 µm EBSD phase map Retained Austenite in As-Built Structure EBSD phase distribution map with retained austenite, BCT and high angle grain boundaries BCC FCC Step size = 0.2 µm FCC phase distribution SE image with the FCC phase overlaid BSE micrograph with FCC phase overlaid Longitudinal section Precipitation Processes during Aging – Tag = 540 ºC Ni, Ti enriched rod-like and Mo enriched spherical phases Ni, Ti enriched rod-like and Mo enriched spherical phases Austenite Reversion during Aging Transverse section Hardness of LPBF Maraging Steel 27 Remarks Refined cellular microstructure is aunique feature in alloys fabricated by laser powder bed fusion. Results evidenced the importance of layer thickness as akey parameter to modify the solidification cell size of the as-built samples. Additive manufacturing of Maraging steels results in complex microstructures formed by a tetragonal martensitic matrix and aheterogeneous distribution of retained austenite. The high dislocation density in combination with solute segregation in the as-fabricated material can promote the precipitation of intermetallic phases that are responsible for the high strength. It is not clear if the austenite growth/reversion and the precipitation of the intermetallic phases during ageing occur as competitive or collaborative phase transformations. 28 Thanks for your attention Retained Austenite in As-Built Structure Journal of Materials Research and Technology, 2023; 25: 6898-6912 Rwp: weighted summation of residual of the least squares fit, Rexp: statistically expected least squares fit, GoF: goodness of fit (sometimes referred as chi-squared); GoF=Rwp/Rexp; a GoF=1.0 means a perfect fitting. Tetragonality of Martensite in As-Built Structure Low magnification High magnification Solute Distribution in As-Built Structure Area: 1000 µm x 1000 µm; Step size: 1 µm Area: 100 µm x100 µm; Step size: 0.1 µm EOS - 40 µm - CW laser mode Longitudinal section Additive Manufacturing 94 (2024) 104494 Texture in As-Built Structure Scientific Reports, 2022; 12: 16168 Multiples of Random Distribution (MRD) Texture in As-Built Structure Multiples of Random Distribution (MRD)