Development of a Machine Learning-Based Predictive Maintenance System for Maintenance Enhancement

Abstract

This research utilized machine learning to build a predictive model for the GE MARK VIe industrial control system. The model aimed to estimate the G2.TTXD output based on three input variables: Inlet Pressure, Inlet Temperature, and IGV Position. Data gathered from the Hassi R'mel gas field was preprocessed and used to train and evaluate three regression algorithms: Linear Regression, Random Forest, and Support Vector Regression (SVR). Model performance was assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), and R-squared (R²), along with analyses of residuals and value evolution. Among the models, SVR delivered the most accurate and reliable results. These findings demonstrate the potential of machine learning to enhance industrial control systems through predictive maintenance, real-time monitoring, and increased operational efficiency key objectives of Industry 4.0.