Case Studies of Survival Analysis for Predictive Maintenance in Manufacturing

Abstract

The Predictive Maintenance (PdM) as a tool for detecting future failures in manufacturing was recognized as an innovative and effective method. Different approaches for PdM have been developed to compromise the availability of data and the demanding needs for probability estimation. The Survival Analysis (SA) method was used in this paper for the probability estimation of machine failure. The paper presents the use of the two most popular SA models: Kaplan-Meier non-parametric and Cox proportional hazard models on two different datasets to present the methodology and the possibilities for applications in manufacturing. By using the first SA model, the results show the probability of a machine or component part to survive a certain amount of time. The Cox proportional model was used to find out the most significant covariates in the observed dataset which have an influence on survival time. The analysis showed that the use of SA in the PdM is a challenging task and can be used as an additional tool for failure analysis and maintenance planning.

Article history: Received (July 18, 2024); Revised (August 11, 2024); Accepted (September 14, 2024); Published online (November 6, 2024)