Scholarlink

T91 steel is widely used in electric power, petrochemical and nuclear industries due to its excellent high temperature strength and creep resistance. However, traditional creep life estimation methods have the disadvantages of complex calculation and low accuracy. In this paper, a deep learning model based on artificial intelligence is constructed, which uses deep learning to extract characteristic data, screen variables and optimize hyperparameters to improve the accuracy and adaptability of prediction results. Compared with traditional algorithms, this algorithm can effectively improve the safety performance of the device, reduce the probability of failure, and also estimate the life of superalloy materials. Provide an intelligent technical means. Compared with traditional methods, the model reduces prediction error by 18.2% (RMSE = 12.3 MPa), improves accuracy to 94.7% (R2 = 0.963), and shortens computational time by 42%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Study on Deep Learning Model for Creep Life Evaluation of T91 Steel at High Temperature

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dongcun Luan,
  • Jiarui Zhu,
  • Decai Zhang,
  • Weidong Li,
  • Rongrong Wei

摘要

T91 steel is widely used in electric power, petrochemical and nuclear industries due to its excellent high temperature strength and creep resistance. However, traditional creep life estimation methods have the disadvantages of complex calculation and low accuracy. In this paper, a deep learning model based on artificial intelligence is constructed, which uses deep learning to extract characteristic data, screen variables and optimize hyperparameters to improve the accuracy and adaptability of prediction results. Compared with traditional algorithms, this algorithm can effectively improve the safety performance of the device, reduce the probability of failure, and also estimate the life of superalloy materials. Provide an intelligent technical means. Compared with traditional methods, the model reduces prediction error by 18.2% (RMSE = 12.3 MPa), improves accuracy to 94.7% (R2 = 0.963), and shortens computational time by 42%.