Evolving Interpretable Fuzzy Rule-Based Systems with Genetic Programming for Predictive Maintenance

Abstract

The project addresses the issue of predictive maintenance by proposing a new method that crosses the boundaries of both genetic programming and fuzzy logic to evolve interpretable fuzzy rule-based systems. The targeted method attempts to achieve a highly accurate predictability of the remaining useful life of industrial equipment with the interpretable evolved rules at the same time. The National Aeronautics and Space Administration turbofan engine degradation dataset [1] is used in the assessment of the validity of the method. The main goals of research are to achieve high accuracy of predictions and to generate easily decipherable linguistic rules that may give an overall look on the degradation process. This is the first study found in the literature that combines genetic programming and fuzzy logic in predictive maintenance tasks. This research’s importance lies in the possibility of plugging the gap between the accuracy of data-driven techniques and the interpretability of rule – based systems. Hence, it results in better predictive maintenance systems.

Keywords: Adaptive Systems, Machine Learning, Predictive Maintenance, Genetic Programming, Fuzzy Rule-based Systems, Remaining Useful Life Prediction