An interpretability research of the Xgboost algorithm in remaining useful life prediction

Traditional industrial health management (PHM) and prediction relies on maintenance experience or work mechanism models to acquire the remaining useful life (RUL) of the equipment is becoming increasingly difficult to obtain. In this paper, the data-driven xgboost ensemble learning method is adapted to predict the RUL of aero-engine, the raw data is input into the ensemble learning model after a simple preprocessing to obtain the prediction results directly. In order to verify the validity of the method, the results obtained are tested on a c-mapss dataset provided by NASA, and the impact of the main influences on the accuracy of the model is analyzed. Finally, the xgboost method is compared with several other machine learning methods on the same data set, and the results show that the xgboost method has higher accuracy, which verified the validity of the method. At last, in order to explain the model, each parameter is analyzed based on the actual meaning, that avoided the simple input-output "black box" phenomenon, and enhanced the interpretability of the model.