Predicting the Fatigue Crack Life of Train Axles Using a Collaborative Network of Yolov5 and LSTM for Acoustic Emission Signals

In the field of axles, life forecasting can be used to predict the lifespan of the axle and to determine more effective maintenance schedules. The life forecasting can have a significant positive impact on the safe operation of train axles, which can be used to predict the lifespan of the axle and to determine more effective maintenance schedules. Traditional methods for predicting the life of the shaft are based on experimental and statistical analysis, but these methods can be expensive and time-consuming. The LSTM-based Lifetime Forecasting method uses machine learning technology and historical data on the axles to more accurately predict remaining lifespan, thereby reducing costs and improving efficiency. This invention is innovative in that it introduces a semi-monitoring technique for estimating an axle RUL’s full life based on YOLOv5 and LSTM training. A large number of unmarked data are trained along with YOLOv5 and LSTM to determine interrupted data to RUL predictions. The use of a laboratory-collected sound crack on the 5 million, 7 million and 10 million signal data collection and RMSE MAE values, for verifying the performance of the model, respectively, for the full cycle of life of 5, 7 million, 10 million vehicles, is predicted with a strong difference between the predictable life of 5M and the remaining value of the forecast model.