Scholarlink

<p>The degradation process was described by combining stress concentration behavior, thermal response variation, and time-dependent experimental observations, which were then reorganized as sequential inputs for RUL estimation. Tooth-root stress identifies fatigue-sensitive regions, while temperature evolution reflects frictional heat, contact instability, and crack initiation. Integrating these features into the long short-term memory (LSTM) model improves prediction reliability, interpretability, and engineering applicability. However, gear RUL prediction still faces challenges due to limited degradation datasets and the difficulty of reproducing realistic operating conditions. To address these limitations, this study develops a physics-informed artificial intelligence framework for predicting the degradation and remaining useful life of FDM-printed PLA spur gears under dynamic meshing conditions. Finite element simulations, thermal experiments, and LSTM modeling were integrated to evaluate load-dependent stress, contact behavior, temperature evolution, and prediction stability. Grid independence analysis identified 1.0&#xa0;mm as the optimal mesh size. Static analysis showed a maximum principal stress of 166.25&#xa0;MPa at the tooth root, while transient stresses reached 9.778, 13.032, and 14.667&#xa0;MPa under 10, 12.5, and 15&#xa0;kg loads, respectively. Temperature-validation errors remained below 1%, confirming reliable thermo-mechanical prediction.</p>

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

Physics-informed data-driven remaining useful life prediction of additively manufactured gears using finite element stress features

  • 期刊论文

  • 发表日期：

    2026年7月1日

查看全文
  • Chil-Chyuan Kuo,
  • Yu-Heng Cai,
  • Armaan Farooqui,
  • Song-Hua Huang

摘要

The degradation process was described by combining stress concentration behavior, thermal response variation, and time-dependent experimental observations, which were then reorganized as sequential inputs for RUL estimation. Tooth-root stress identifies fatigue-sensitive regions, while temperature evolution reflects frictional heat, contact instability, and crack initiation. Integrating these features into the long short-term memory (LSTM) model improves prediction reliability, interpretability, and engineering applicability. However, gear RUL prediction still faces challenges due to limited degradation datasets and the difficulty of reproducing realistic operating conditions. To address these limitations, this study develops a physics-informed artificial intelligence framework for predicting the degradation and remaining useful life of FDM-printed PLA spur gears under dynamic meshing conditions. Finite element simulations, thermal experiments, and LSTM modeling were integrated to evaluate load-dependent stress, contact behavior, temperature evolution, and prediction stability. Grid independence analysis identified 1.0 mm as the optimal mesh size. Static analysis showed a maximum principal stress of 166.25 MPa at the tooth root, while transient stresses reached 9.778, 13.032, and 14.667 MPa under 10, 12.5, and 15 kg loads, respectively. Temperature-validation errors remained below 1%, confirming reliable thermo-mechanical prediction.