Probabilistic Fuzzy System for Evaluation and Classification in Failure Mode and Effect Analysis

Abstract

Failure Mode and Effect Analysis (FMEA) is an essential risk analysis tool that is widely applicable in various industrial sectors. This structured technique allows us to identify and assign priority levels to potential failures that violate the reliability of a system or process. Failure evaluation occurs in a decision-making environment with uncertainty. This study proposes a probabilistic fuzzy system that integrates linguistic and stochastic uncertainty based on a Mamdani-type model to strengthen the FMEA technique. The system is based on analyzing the frequency of failures and obtaining the parameters to determine the probability of occurrence through the Poisson distribution. In addition, the severity and detection criteria were evaluated by the experts and modeled using the Binomial distribution. The evaluation result is a discrete value analogous to the process of obtaining the success or failure of the expert generating the evaluation of 10 Bernoulli experiments. Three fuzzy inference expert systems were developed to combine multiple experts’ opinions and reduce linguistic subjectivity. The case study was implemented in the knitting area of a textile company in the south of Guanajuato to validate the proposed approach. The potential failure of the knitting machinery, which compromises the top tension subsystem’s performance and the product’s quality, was analyzed. The proposed system, which is based on a robust mathematical model, allows for reliable fault evaluation with a simple scale. The classification performed by the system and the one performed by the experts has similar behavior. The results show that the proposed approach supports decision-making by prioritizing failure modes.