Online Fatigue Life Prediction and Crack Length Estimation Based on LSTM

In this paper, a prediction method based on LSTM model is proposed for the online prediction problem of fatigue life and crack length in crack propagation and fracture stage. The proposed method uses the online collected data to predict the health indicator (HI) of the specimen based on a long short-term memory (LSTM) model; Then a particle filtering algorithm and polynomial fitting are used to convert the health indicator into the remaining fatigue life and crack length. Tests were performed with several experimental data under the same load, and the results showed that the deviation between the predicted life and the real life of the model is small, with an error value of less than 5% relative to the total fatigue life in the middle and later stages of the prediction.