Creep-Fatigue Interaction Life Prediction for P91 Steel Based on Improved Elman Neural Network Model

Abstract

A creep-fatigue life prediction model for P91 steel using Improved Elman Neural Network is founded based on the creep-fatigue experiment datas in this paper. The load keeping time is selected as the model input and the creep life or fatigue life as the output in the model. The prediction ability is validated from training the sample data. The results show the model has a very high prediction accuracy for training samples. But the model generalization ability is relatively weak, so the forecasting accuracy for test sample is lower than the training samples. Through increasing test sample number and decreasing the uncertain factors during the experiment in the follow-on working, the prediction accuracy and the model generalization ability will be increased.