An Explainable Artificial Intelligence Approach for Remaining Useful Life Prediction

Prognosis and health management depend on sufficient prior knowledge of the degradation process critical components to predict remaining useful life. This task is composed two phases: learning prediction. The first phase uses available information learn system’s behavior. second predicts future behavior based system estimates its lifetime. Deep approaches achieve good prognostic performance but usually suffer from a high computational load lack interpretability. Complex feature extraction models do not solve this problem, as they lose in thus have poor prognosis for A new prepossessing approach used with clustering address issue. It allows restructuring data into homogeneous groups strongly related each other using simple architecture LSTM model. advantageous terms time possibility limited capabilities. Then, we focus interpretability deep Explainable AI interpretable RUL proposed offers model improvement enhanced interpretability, enabling better understanding contributions. Experimental results NASA C-MAPSS dataset show compared common methods.