Enhanced FMEA for supply chain risk identification

Lu, Lu; Zhou, Rong; de Souza, Robert Conference Paper Enhanced FMEA for supply chain risk identification Provided in Cooperation with: Hamburg University of Technology (TUHH), Institute of Business Logistics and General Management Suggested Citation: Lu, Lu; Zhou, Rong; de Souza, Robert (2018) : Enhanced FMEA for supply chain risk identification, In: Kersten, Wolfgang Blecker, Thorsten Ringle, Christian M. (Ed.): The Road to a Digitalized Supply Chain Management: Smart and Digital Solutions for Supply Chain Management. Proceedings of the Hamburg International Conference of Logistics (HICL), Vol. 25, ISBN 978-3-7467-6535-8, epubli GmbH, Berlin, pp. 311-330, https://doi.org/10.15480/882.1783 This Version is available at: https://hdl.handle.net/10419/209355 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-sa/4.0/ Enhanced FMEA for Supply Chain Risk Identification Lu Lu1, Zhou Rong1, Robert de Souza1 1 – National University of Singapore Supply chain risk identification is fundamental for supply chain risk management. Its main purpose is to find critical risk factors for further attention. The failure mode effect analysis (FMEA) is well adopted in supply chain risk identification for its simplicity. It relies on domain experts’ opinions in giving rankings to risk factors regarding three decision factors, e.g. occurrence frequency, detectability, and severity equally. However, it may suffer from subjective bias of domain experts and inaccuracy caused by treating three decision factors as equal. In this study, we propose a methodology to improve the traditional FMEA using fuzzy theory and grey system theory. Through fuzzy theory, we design semantic items, which can cover a range of numerical ranking scores assessed by experts. Thus, different scores may actually represent the same semantic item in different degrees determined by membership functions. In this way, the bias of expert judgement can be reduced. Furthermore, in order to build an appropriate membership function, experts are required to think thoroughly to provide three parameters. As the results, they are enabled to give more reliable judgement. Finally, we improve the ranking accuracy by differentiating the relative importance of decision factors. Grey system theory is proposed to find the appropriate weights for those decision factors through identifying the internal relationship among them represented by grey correlation coefficients. Theresultsofthe casestudyshowthe improvedFMEA does produce different rankings from the traditional FMEA. This is meaningful for identifying really critical risk factors for further management. Keywords: supply chain risk identification; FMEA; grey system theory; fuzzy set theory First recieved: 01.Jul.2018 Revised: 10.Jul.2018 Accepted: 15.Jul.2018 311 Enhanced FMEA for Supply Chain Risk Identification 1 Introduction Riskidentificationinvolvingbothriskclassificationandrisk rankingcanbeseen as a fundamental work for risks assessment. It identifies critical risks that need further assessment and treatment (Berman and Putu, 2012). Generally, researchers categorize risks into several groups for systematically risk identification. In our study, risks are classified into three levels—the macro level, the company level and the industry level. Risks on the macro level may influence the whole supply chain’s operation; risks on the company level are from operation activities of a company; risks on the industry level are from the development of industry (Zhou et al., 2012). Subsequently, we still need to identify most relevant risk factorsso that only those relevant and important ones are studied further. As a large number of risk factors that may be involved, the easiest way is through risk ranking so that a company can effectively mitigate them (Chopra and Sodhi, 2004). A comparison of nine risk ranking techniques is shown in Table 1, where techniques are compared in five attributes—complexity of application, risk consequence analysis, risk probability analysis, quantified output, and objectivity. 312 1 Introduction Table 1: Comparison of different techniques Techniques Complexity of application Risk consequence analysis Risk probability analysis Quantitative output Objectivity Structured ”Whatif”(SWIFT) Low A A No Low Fault tree analysis Medium NA A Yes High Cause and consequence analysis High A A Yes High Cause and effect analysis Medium A NA No Medium Decision tree High A A Yes Medium FMEA Medium A A Yes Medium Hazard analysis and critical control points Medium A NA No Medium Analysis hierarchy process Medium NA NA Yes High Bayesian statistics and Bayes nets High A NA Yes High 313 Enhanced FMEA for Supply Chain Risk Identification According to the above table, risk ranking techniques can be classified into four big categories. The first category is a supporting method, which can only give a general analysis about the risks. For example, SWIFT uses a series of “What if” questions to identify the deviations from normal conditions with the help of a predefinedchecklist. Thesecondcategory usesscenarioanalysis, whichisgoodat analyzing the causes of risks. Fault tree analysis, cause and consequence analysis, cause and effect analysis, and decision tree belong to this category (Dakas et al., 2009, Hauptmanns, 2010, Hichemand Pepijn, 2007). The third category is function analysis methodincluding FMEA, hazardanalysis, and criticalcontrol points. They focus on analyzing the effects of risks. The final category is the statistical method, which applies the statistical knowledge into the analysis process. AHP and BBN belong to this category. FMEA has been adopted widely as it can produce quantitative output, which is desirable for risk ranking. However, it may be biased as the opinions of domain experts can be subjective. The target of the current study is to improve FMEA for its objectivity using fuzzy set theory and grey system theory. The structure of the paper is as follows. The next section describes the traditional FMEA; section 3 presents the enhanced FMEA. A case study is provided in section 4. Finally, conclusions are made in section 5. 2 The Traditional FMEA FMEA identifies failure modes and mechanisms as well as their effects. There are several types of FMEA, e.g. design FMEA, system FMEA, process FMEA, service FMEA, software FMEA, etc. The current study adopts the FMEA methodology. In the study, the system means the supply chain while the failure mode refers to the potential supply chain risk. For each risk, experts give three scores between one and ten regarding the risk’s occurrence frequency (OF), detectability and severity. Then the risk priority number (RPN) can be calculated through multiplying these three score and represents the risk impact of the risk factor. The higher is the RPN, the more critical is the risk. 314 2 The Traditional FMEA Table 2: Scores marked by experts Risk Experts OF Detectability Severity Risk 1 B1 4 5 9 B2 5 1 5 B3 2 2 8 Average 3.7 2.7 7.3 Risk 2 B1 9 3 4 B2 6 2 2 B3 8 3 5 Average 7.7 2.7 3.7 2.1 An Example We assume that there are three domain experts—B1, B2 and B3, who give ranks regarding the risk impacts of two risk factors: risk 1 and risk 2 in a scale of 1 to 10. The greatest rank, 10, refers to the greatest risk impact. The summary of scores marked by experts is shown in Table 2. Then, using the average numbers of three experts’ rankings, RPNs for risks 1 and 2 can be calculated as RPN 1 =3.7*2.7*7.3=7.3 and RPN 2 =7.7*2.7*3.7=7.7, respectively. Since RPN 2 is greater than RPN 1 , risk 2 is more risky than risk 1 according to those three experts. 2.2 Limitations of Traditional FMEA Through the above example, three limitations of the traditional FMEA can be recognized. Firstly, there may be ranking differences among experts, which can lead to the inaccuracy of outcomes. For example, for the same degree of risk impact, expert B1 may score 9 while expert B2 scores 7. Secondly, the approach depends on the experience and knowledge of experts to a large degree and the outcome can be very subjective. Finally, the RPN formula above does not consider therelativeimportanceof threedecisionfactors. The severityof arisk factorcould be more important than OF or detectability while in the current approach, three decision factors are treated as equal. As a result, the above RPN may not be able to give accurate risk rankings. 315 Enhanced FMEA for Supply Chain Risk Identification 3 Improved FMEA To reduce the limitations of the traditional FMEA, we propose to improve it using fuzzysettheoryandgreyrelationanalysis. Theprimaryprocedureoftheimproved FMEA is as follows. — Identify the relevant risk factors (regarding the risk categories) and domain experts. — Reduce the expert bias using Fuzzy Set theory. — Experts reach consensus for each risk. — Improve assessment precision through Grey Correlation Analysis. — Ranking risk factors. Specially, we use Fuzzy Set theory in step 2 to reduce experts’ ranking difference and Grey Correlation Analysis in step 4 to improve assessment precision through applying appropriate weightages to decision factors. 3.1 Fuzzy Set In the classical discrete sets, an element either belongs to a set or it does not. But for fuzzy sets, their elements have degrees of membership according to certain membership functions (Abdelgawad and Fayek, 2011). There are many types of membership functions based on the graphs, e.g. triangular, trapezoidal, Gaussian, generalized bell, sigmoid, and others. We choose the triangular membership function to improve the traditional FMEA. A triangular membership function is specified by three parameters, a, b, and c in formula (1): M(x) =              0,x≤a x−a b−a,a≤x≤b x−a b−a,b≤x≤c 0,c≤x (1) 316 4 Case Study where M refers to the membership of score x. The value of the membership ranges from 0 to 1. The value of 1 represents the full membership. 3.2 Grey Relation Analysis Grey system theory was first developed in 1982 (Deng, 1982). A grey system generally refers to a system lacking certain information, e.g. structure message, operation mechanism, or behavior document. The aims of the grey system theory are to provide theory, techniques, notions and ideas for resolving latent and intricate systems (Deng, 1982). This study adopts the grey relation analysis, which describes the relationships between one main factor and all the other factors in a given system. The degree of correlation among different factors is measured by their grey correlation coefficient. The greater is the value of coefficient, the closer relationship is between the two factors. 4 Case Study In this section, we use a case study to improve the traditional FMEA through the proposed methodology. Assuming there are three risk factor, e.g. raw material shortage, labour availability, and natural disaster as well as three experts, e.g. B1, B2, B3, the ranking process follows the five steps illustrated in sections 4.1 to 4.5. 4.1 Identify the Relevant Risks List and Experts Table 3 summarizes the risk factors and experts, which are identified for the ranking process. Risks identified are raw material shortage (R1), labor availability (R2), and natural disaster (R3) while three experts, B1, B2 and B3 are from supply chain, operation, and R&D departments, respectively. 317 Enhanced FMEA for Supply Chain Risk Identification Table 8: Three experts’ average judgment for risk 2 VH H M L VL Occurrence frequency 0 2(67%) 1(33%) 0 0 Detectability 0 0 2(67%) 1(33%) 0 Severity 0 0 3(100%) 0 0 Table 9: Three experts’ average judgment for risk 3 VH H M L VL Occurrence frequency 0 0 2(67%) 1(33%) 0 Detectability 0 0 1(33%) 2(67%) 0 Severity 1(33%) 2(67%) 0 0 0 324 4 Case Study x1=    x1 o1x1 o2x1 o3x1 o4x1 o5 x1 d1x1 d2x1 d3x1 d4x1 d5 x1 s1x1 s2x1 s3x1 s4x1 s5     =  0 0 1 0 0 0 0 0.33 0.67 0 0.67 0.33 0 0 0 (4) x2=    x2 o1x2 o2x2 o3x2 o4x2 o5 x2 d1x2 d2x2 d3x2 d4x2 d5 x2 s1x2 s2x2 s3x2 s4x2 s5     =  0 0.67 0.33 0 0 0 0 0.67 0.33 0 0 0 0 0 0 (5) x3=    x3 o1x3 o2x3 o3x3 o4x3 o5 x3 d1x3 d2x3 d3x3 d4x3 d5 x3 s1x3 s2x3 s3x3 s4x3 s5     =  0 0 0.67 0.33 0 0 0 0.33 0.67 0 0.33 0.67 0 0 0 (6) Subsequently, the judgment matrices for risks 1, 2 and 3 are listed in matrices of (4) (5) and (6). They represent the rankings from three experts given in terms of five semantic items. 4.4 Improve Precision through Grey System Theory In the current section, we firstly establish an assessment matrix based on step 3 andthenareferencematrixusingthemostrisky semanticitem. Subsequently, the degree of relevancy is measured through the grey correlation coefficient. Finally, we apply weights to decision factors (occurrence frequency, detectability, and severity) to differentiate their relevant importance. 325 Enhanced FMEA for Supply Chain Risk Identification 4.4.1 Establishment of assessment matrix and reference matrix The assessment matrix is built based on the judgment matrices of (4) (5) and (6) and specific numbers of sematic items. The assessment matrix for three risks are formed as follow: R=  r1or1dr1s r2or2dr2s r3or3dr3s (7) Where rio =Ivh ×xi o1+Ih×xi o2+Im×xi o3+Il×xi o4+Ivl ×xi o5 is the score of “occurrence frequency” for risk i. rid =Ivh ×xi d1+Ih×xi d2+Im×xi d3+Il×xi d4+Ivl ×xi d5 is the score of “detectability” for risk i. ris =Ivh ×xi s1+Ih×xi s2+Im×xi s3+Il×xi s4+Ivl ×xi s5 is the score of “severity” for risk i. Withspecificnumbersforfivesematicitems, e.g. Ivl = 0.65 , Il= 3.4 , Im= 5.7 , Ih= 8.2, and Ivh = 9.7, we have: — For risk 1, r1o= 5.7,r1d= 4.2,r1s= 8.9; — For risk 2, r2o= 7.4,r2d= 4.9,r2s= 5.7; — For risk 3, r3o= 4.9,r3d= 4.2,r3s= 8.7; Then, we have the assessment matrix in (8). R=  r1or1dr1s r2or2dr2s r3or3dr3s =  5.7 4.2 8.9 7.4 4.9 5.7 4.9 4.2 8.7 (8) Furthermore, we use the specific number of semantic item “very high”, e.g. 9.7 to establish the reference matrix in (9). Rf=rfo rfd rfs=9.7 9.7 9.7(9) 326 4 Case Study Thisreferencematrixrepresentsa very riskysituationofa risk factorwherethe“occurrence frequency”, “detectability”, and “severity” are all “very high” (Table 4). 4.4.2 Calculating the grey correlation coefficient Nowwith theassessment matrix R in(8) andthe referencematrix Rf in (9), wecan calculate the grey correlation coefficient between them using the grey correlation coefficient function (10) (Du et al., 2011): λ(xf j , xij ) = min i|xfj −xij |+vmax i|xfj −xij | |xfj −xij |+vmax i|xfj −xij | (10) Where irefers to risk i; jrefers to decision factor j; frefers to the reference matrix entry; λ(xf j , xij ) refers to the grey correlation coefficient of entries xfj and xij ; υ is the distinguishing coefficient; its value is within [0,1] and normally v= 0.5. Table 10 presents the results of all |xfj −xij | and the minimal and maximal values are min i|xf−xi|= 0.8and max i|xf−xi|= 5.5. Thus, we have λ1o= min i|xfj −xij |+0.5max i|xfj −xij | |xfj −xij |+0.5max i|xfj −xij | =0.8+0.5×5.5 4+0.5×5.5= 0.52 (11) 327 Enhanced FMEA for Supply Chain Risk Identification Table 10: Results of all |xfj −xij | No. Occurrence frequency Detectability Severity ∆1j=|xfj −x1j| |xfo −x1o|= 4 |xfd −x1d|= 5.5|xfs −x1s|= 0.8 ∆2j=|xfj −x2j| |xfo −x2o|= 2.3|xfd −x2d|= 4.8|xfs −x2s|= 4 ∆3j=|xfj −x3j| |xfo −x3o|= 4.8|xfd −x3d|= 5.5|xfs −x3s|= 1 Similarly, we can get the grey correlation coefficient matrix (12). λ=  λ1oλ1dλ1s λ2oλ2dλ2s λ3oλ3dλ3s =  0.52 0.43 1 0.7 0.47 0.52 0.47 0.43 0.95 (12) Furthermore, assuming the weights of decision factors are given in matrix (13). ω=0.3 0.2 0.5(13) Where ωo is the weight of Occurrence frequency, ωd the weight of detectability, and ωsthe weight of severity. We have G=  ωoλ1o+ωdλ1d+ωsλ1s ωoλ2o+ωdλ2d+ωsλ2s ωoλ3o+ωdλ3d+ωsλ3s =  0.742 0.564 0.702 (14) Matrix G is the final rankings of three risks regarding three decision factors considering three experts’ judgement. The final scores of risks 1, 2, and 3 are 0.742, 0.564, and 0.702, respectively. As the 0.742 is the greatest, R1 is the most risky one while R3 is the second and R2, the third. In summary, from the highest to the lowest in term of risk impacts, the ranking of studied risks is R1>R3>R2. 328 5 Conclusion 4.5 Compare the Traditional and the Improved FMEA For the assessment matrix R (8), the RPN applying the traditional FMEA is as follows. RP N =  r1o×r1d×r1s r2o×r2d×r2s r3o×r3d×r3s =  5.7×4.2×8.9 7.4×4.9×5.7 4.9×4.2×8.7 =  213 206 179 (15) Thus, the ranking of risks is R1>R2>R3, which is different from the outcome of the improved FMEA in matrix (14). The reason is that the improved FMEA considers weights (matrix (13)) for three decision factors. Furthermore, if we directly include those weights into the assessment matrix (8) without applying the grey correlation coefficient. The result is as follows. RP N1=  5.7×0.3×4.2×0.2×8.9×0.5 7.4×0.3×4.9×0.2×5.7×0.5 4.9×0.3×4.2×0.2×8.7×0.5 =  6.39 6.18 5.37 (16) The new ranking becomes the same as the one from the traditional FMEA, e.g. R1>R2>R3, but different from the improved FMEA. This emphasizes the importance of including the grey correlation coefficient in allocating appropriate weightages to decision factors. 5 Conclusion Inthisstudy, themethodologytoimproveFMEAforsupplychainriskidentification is proposed in order to reduce the bias from domain experts and improve the ranking accuracy. First of all, the subjective bias in ranking from experts can be reduced through establishing semantic items, which are linked to numerical scores through fuzzy membership functions. In this way, even though experts give difference scores for the same level of risk impact, those scores can still represent the same semantic meaning, perhaps in different degrees. In this way, the bias from experts can be reduced. 329 Enhanced FMEA for Supply Chain Risk Identification Furthermore, in order to build a membership function, three parameters are requested to represent the coverage of a semantic item in terms of numerical scores. This enables experts to think thoroughly and further improves the reliability of their judgement. Finally, in the traditional FMEA, decision factors are treated equally in their roles to determine the impact of a risk. This may not rational. In the improved FMEA, we differentiate the importance of decision factors in ranking risk impacts. The grey correlation coefficient is adopted to extract appropriate weights for decision factors. This further improves the accuracy of the ranking. References Abdelgawad, M. and A. Fayek (2011). “Fuzzy reliable analyzer:quantitative assessment of risk events in the construction industry using fuzzy fault-tree analysis”. In: Journal of Construction Engineering and Management 137.4, pp. 294–302. Berman, K. and D. K. Putu (2011). “SCRIS: a knowledge-based system tool for assisting manufacturing organizations in identifying supply chain risks”. In: Journal of Manufacturing Technology Management 23.7, pp. 834–852. Chen, M. S. (2010). “Evaluating the rate of aggregative risk in software development using fuzzy set theory”. In: Cybernetics and Systems: An International Journal 30.1, pp. 57–75. Chopra, S. and M. S. Sodhi (2004). “Managing risk to avoid supply chain breakdown”. In: MIT Slogan Management Review 46.1, pp. 53–62. Dakas, I. M., D. A. Karras, and D. C. Panagiotakopoulos (2009). “Fault tree analysis and fuzzy expert systems: early warning and emergency response of landfill operations”. In: Environmental Modeling & Software 24.1, pp. 8–25. Deng, J. (1982). “Control problems of grey systems”. In: Systems and Control Letter 5, pp. 288–294. Du, D. L., J. Qiu, and H. Y. Zhao (2011). “Risk assessment study of manufacturing green supply chain based on grey theory”. In: 2011 China located International Conference on Information Systems for Crisis Response and Management. Faisal, A. and L. S. Sarah (2015). “A fuzzy-based integrated framework for supply chain risk assessment”. In: International journal of production economics 161, pp. 54–63. Hauptmanns,U.(2010).“Adecision-making framework forprotectingprocess plants from flooding based on fault tree analysis”. In: Reliability Engineering & Sytem Safety 95.9, pp. 970–980. Hichem,B.andC.Pepijn(2007).“Dynamicfaulttreeanalysis using input/outputinteractivemarkov chains. dependable systems and networks”. In: Internal Conference on 37th Annual IEEE/IFIP, pp. 708–717. Zhou, R., R. De Souza, and M. Goh (2013). “Risk management of complex supply chains part 1: Supply chain risk and complex systems”. In: Vol. 12-Nov-SCI-09. 330