TOWARDS DEVELOPING A NOVEL FRAMEWORK FOR PRACTICAL PHM: A SEQUENTIAL DECISION PROBLEM SOLVED BY REINFORCEMENT LEARNING AND ARTIFICIAL NEURAL NETWORKS

Abstract

The heart of prognostics and health management (PHM) is to predict the equipment degradation evolution and, thus, its Remaining Useful Life (RUL). These predictions drive the decisions on the equipment Operation and Maintenance (O&M), and these in turn influence the equipment degradation evolution itself. In this paper, we propose a novel PHM framework based on Sequential Decision Problem (SDP), Artificial Neural Networks (ANNs) and Reinforcement Learning (RL), which allows properly considering this feedback loop for optimal sequential O&M decision making. The framework is applied to a scaled-down case study concerning a real mechanical equipment equipped with PHM capabilities. A comparison of the proposed framework with traditional PHM is performed.

Keywords

• phm
• maintenance planning
• artificial neural network
• sequential decision problem
• reinforcement learning