Fatigue Life Prediction Technology of Metal Structural Materials Based on Deep Learning

Fatigue life is a key indicator for evaluating the reliability of metal materials, while traditional prediction methods have limitations. This study proposes a hybrid model that integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM) to improve the accuracy of predictions. The model extracts spatial features from the metal fatigue process using CNN, captures dynamic changes over time with LSTM, and outputs prediction results through fully connected layers. Experimental results show that the model has high precision and reliability in predicting metal fatigue life, with prediction values very close to actual values. Furthermore, compared with LSTM and ARIMA models, the CNN+LSTM model demonstrates superior predictive performance. The Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE) of the CNN+LSTM model are the lowest, at 22.3607, 3.0787, and 16.6667, respectively, indicating that it outperforms LSTM and ARIMA models in prediction accuracy, with the smallest prediction error and highest prediction accuracy. The deep learning model of this study provides a scientific basis and technical support for the design, maintenance, and management of engineering structures, showing great application potential in the field of metal fatigue life prediction.