Citation

Antomarioni S, Ciarapica FE, Bevilacqua M. Safety Sci. 2022; 150: e105711.

Copyright

(Copyright © 2022, Elsevier Publishing)

DOI

10.1016/j.ssci.2022.105711

PMID

unavailable

Abstract

The ongoing digital transformation enables the collection of a vast amount of data from the maintenance and production processes. Hence, the techniques traditionally applied for monitoring the operations and defining the maintenance approaches for improving the asset reliability can be accompanied by new methodologies more oriented to data analytics. In this context, this work proposes a data-driven framework for supporting the analysis of the production processes in terms of failures and related effects through the well-known and widely applied Failure Mode, Effect, and Criticality Analysis (FMECA). Indeed, after developing the Failure Mode Effects and Criticality Analysis, the results obtained for the plant under investigation are elaborated through Machine Learning techniques such as the Association Rule Mining to define the cause-effect relationships that led to a failure. The association rules extracted are then processed through the Social Network Analysis to represent such relationships, facilitate their comprehension, and identify the existence of communities of nodes to detect critical patterns and locate the most influential nodes. The knowledge of such details can provide helpful support in terms of awareness of the plant and the development of intelligent maintenance procedures. The proposed approach is applied to an off-shore and on-shore platform to assess the impact of the theoretical analysis on the practical implementation by highlighting unknown relations among the analyzed variables and showing new cause-effect relationships.

Language: en

Keywords

Association Rule Mining; Data-driven approach; Failure Mode Effects and Criticality Analysis; Machine Learning; Risk-Based Maintenance; Social Network Analysis