Fuzzy-FMECA Integrated with K-Means Clustering for Data-Driven Risk Classification of Industrial Equipment

Conference paper

Bouchaala Mohammed Abdelghani, Mawloud Titah, Guetarni Islam Hadj Mohamed, Tarfaya Anis, Bachir Bouiadjra Bachir
National Conference on Innovations in Industrial Safety and Environment (IISE), Department of Industrial Safety and Environment, Institute of Maintenance and Industrial Safety, University of Oran 2 Mohamed Ben Ahmed, Oran, Algeria, Oct 21

Conference website

Cite

APA Click to copy
Mohammed Abdelghani, B., Mawloud, T., Islam Hadj Mohamed, G., Anis, T., & Bachir, B. B. Fuzzy-FMECA Integrated with K-Means Clustering for Data-Driven Risk Classification of Industrial Equipment. In National Conference on Innovations in Industrial Safety and Environment (IISE). Oran, Algeria: Department of Industrial Safety and Environment, Institute of Maintenance and Industrial Safety, University of Oran 2 Mohamed Ben Ahmed.

Chicago/Turabian Click to copy
Mohammed Abdelghani, Bouchaala, Titah Mawloud, Guetarni Islam Hadj Mohamed, Tarfaya Anis, and Bachir Bouiadjra Bachir. “Fuzzy-FMECA Integrated with K-Means Clustering for Data-Driven Risk Classification of Industrial Equipment.” In National Conference on Innovations in Industrial Safety and Environment (IISE). Oran, Algeria: Department of Industrial Safety and Environment, Institute of Maintenance and Industrial Safety, University of Oran 2 Mohamed Ben Ahmed, n.d.

MLA Click to copy
Mohammed Abdelghani, Bouchaala, et al. “Fuzzy-FMECA Integrated with K-Means Clustering for Data-Driven Risk Classification of Industrial Equipment.” National Conference on Innovations in Industrial Safety and Environment (IISE), Department of Industrial Safety and Environment, Institute of Maintenance and Industrial Safety, University of Oran 2 Mohamed Ben Ahmed.

BibTeX Click to copy

@inproceedings{mohammed-a,
  title = {Fuzzy-FMECA Integrated with K-Means Clustering for Data-Driven Risk Classification of Industrial Equipment},
  month = oct,
  day = {21},
  address = {Oran, Algeria},
  journal = {National Conference on Innovations in Industrial Safety and Environment (IISE)},
  organization = {Department of Industrial Safety and Environment, Institute of Maintenance and Industrial Safety, University of Oran 2 Mohamed Ben Ahmed},
  author = {Mohammed Abdelghani, Bouchaala and Mawloud, Titah and Islam Hadj Mohamed, Guetarni and Anis, Tarfaya and Bachir, Bachir Bouiadjra},
  month_numeric = {10}
}

This work proposes a hybrid Fuzzy-FMECA–K-Means framework to improve risk classification and maintenance prioritisation for industrial equipment. The approach addresses key limitations of conventional Failure Mode, Effects, and Criticality Analysis (FMECA), particularly the imprecision of expert assessments, the subjectivity of predefined risk thresholds, and the limited ability to discriminate between failure modes with similar criticality indices. Triangular fuzzy numbers are used to represent linguistic evaluations of Occurrence, Severity, and Detectability, while centre-of-gravity defuzzification yields a crisp criticality index, (C*). K-Means clustering is then applied to classify failure modes into four risk categories—Critical, High, Moderate, and Low—without relying on predefined thresholds for the criticality index. The framework is applied to 22 failure modes of a sugar centrifuge in an Algerian industrial context and compared with classical FMECA. The results show 31.8% exact agreement between the two classifications and a Spearman rank correlation coefficient of 0.788, indicating substantial consistency in relative prioritisation despite differences in risk categorisation. Both approaches identify critical failure modes associated with the discharge valve, while the proposed framework redistributes several intermediate- and high-risk failure modes across different categories, offering a more differentiated basis for maintenance prioritisation. The number of clusters was predefined to obtain four interpretable risk categories, while a fixed random seed and multiple K-Means replicates support computational reproducibility. This case study demonstrates how the proposed framework combines fuzzy representation of expert uncertainty with unsupervised clustering to provide a practical and transparent tool for risk-based maintenance decision-making. It constitutes a proof of concept that merits further investigation. Future work will focus on applying the framework to larger-scale industrial systems to assess its generalisability and performance, while investigating the automatic identification of the optimal number of clusters to enhance its adaptability across different industrial contexts.