Risk assessment by failure mode and effects analysis (FMEA) using an interval number based logistic regression model

• Published in: Safety science 2020-12, Vol.132, p.104967, Article 104967
• Main Authors: Bhattacharjee, Pushparenu, Dey, Vidyut, Mandal, U.K.
• Format: Article
• Language: English
• Subjects: Causality; Electric power generation; Failure analysis; Failure modes; FMEA; Interval number; Logistic regression; Logit models; Machine learning; Methodology; Power plants; Probability; Probability of risk of failure; Pumps; Regression analysis; Regression models; Risk assessment; Risk factors; Risk Priority Number (RPN); Risk reduction; Safety research; Statistical analysis
• ISSN: 0925-7535; 1879-1042
• DOI: 10.1016/j.ssci.2020.104967

•This paper addresses the shortcomings of traditional RPN calculation during failure mode and effect analysis, and proposes a systematic approach for identifying and evaluating potential failures using the methodology of interval number based logistic regression approach. A comparative analysis of the existing approaches and proposed methodology is also provided.•Logistic regression allows comparing the effects of variables measured on different scales and it indicates the significant relationships between dependent variable and independent variable. Thus, it helps to associate coefficients with the risk factors of failure. The coefficient in the logistic regression equation helps to know the degree of importance of each risk factor in the failure.•In failure analysis, it is used to predict the likelihood of failure, and investigating the importance of related factors contributing to failure. With the help of data of submersible pumps as a case study, an equation of probability of risk of failure 'P' is found through R software. Thus, a novel attempt has been made for the first time to investigate the limitations of traditional RPN calculation statistically. In order to reduce risks of failure, industries use a methodology called Failure Mode and Effects Analysis (FMEA) in terms of the Risk Priority Number (RPN). The RPN number is a product of ordinal scale variables, severity (S), occurrence (O) and detection (D) and product of such ordinal variables is debatable. The three risk attributes (S, O, and D) are generally given equal weightage, but this assumption may not be suitable for real-world applications. Apart from severity, occurrence, and detection, the presence of other risk attributes may also influence the risk of failure and hence should be considered for achieving a holistic approach towards mitigating failure modes. This paper proposes a systematic approach for developing a standard equation for RPN measure, using the methodology of interval number based logistic regression. Instead of utilizing RPN in product form for each failure, this method is benefited from decisions based on probability of risk of failure, 'P' which is more realistic in practical applications. A case study is presented to illustrate the application of the proposed methodology in finding the risk of failure of high capacity submersible pumps in the power plant. The developed logistic regression model (logit model) using R software helped in generating the probability of ri