Enhancing FMEA with Machine Learning for Predictive Safety in Industrial Equipment: Case Study at GLA Gas Lift Compressor

Abstract

This thesis presents an enhanced approach to Failure Modes and Effects Analysis (FMEA) by integrating machine learning and Monte Carlo simulation to improve predictive safety and reliability in industrial equipment. Being realized at SONATRACH’s Guellala Production Center, the study applies the latest AIAG-VDA Action Priority standard for risk evaluation. A machine learning model is employed to predict probability of failure modes within 90 days using three features timestamp, pressure, and temperature, transforming subjective expert assessments into objective, data-driven probabilities. The results demonstrate that despite of the limitations this integrated methodology offers more accurate risk prioritization and supports proactive maintenance planning, contributing to improved safety and operational efficiency in critical gas lift compressor systems.