Scholarlink

<p>As the preferred joining technique for aero-engine rotor components, inertia friction welding (IFW) encounters narrower process windows with the increasing alloy strength, making it imperative to develop simulation models with high accuracy and good generalization capability within a given alloy system under different welding parameters and component sizes. The friction coefficient is the critical parameter to determine the interfacial behavior in the numerical simulation of IFW. Therefore, a thermomechanical simulation system was firstly established in this study. Then, a friction coefficient database covering a series of sliding velocities, temperatures, and axial pressures was obtained through friction experiments. Considering the strong multicollinearity among these variables, the ridge regression machine learning model was employed to effectively decouple them, constructing a unified friction constitutive model that describes both Coulomb and shear friction states. On this basis, a fully coupled thermomechanical finite element model adopting the proposed friction model was established, whose predictive capability and generalization performance with respect to variations in welding parameters and geometric size were systematically validated by comparisons with IFW experiments under different conditions. The simulation accurately reproduced the rotational speed decay, temperature evolution, axial shortening distance, and flash morphology, and the prediction errors were generally &lt; 10%, except for a few outliers. Comparisons with conventional friction models further demonstrated the superior generalization performance of the machine learning–enhanced friction model. The weld forming behavior under different welding parameters was also discussed for welding components.</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

A novel simulation model for inertia friction welding with a machine learning–enhanced friction constitutive relation

  • 期刊论文

  • 发表日期：

    2026年5月21日

查看全文
  • Rui Zhan,
  • Haoran Li,
  • Yuyuan Cai,
  • Qing Liu,
  • Chunbo Zhang,
  • Hang Liang,
  • Lei Cui,
  • Yongchang Liu

摘要

As the preferred joining technique for aero-engine rotor components, inertia friction welding (IFW) encounters narrower process windows with the increasing alloy strength, making it imperative to develop simulation models with high accuracy and good generalization capability within a given alloy system under different welding parameters and component sizes. The friction coefficient is the critical parameter to determine the interfacial behavior in the numerical simulation of IFW. Therefore, a thermomechanical simulation system was firstly established in this study. Then, a friction coefficient database covering a series of sliding velocities, temperatures, and axial pressures was obtained through friction experiments. Considering the strong multicollinearity among these variables, the ridge regression machine learning model was employed to effectively decouple them, constructing a unified friction constitutive model that describes both Coulomb and shear friction states. On this basis, a fully coupled thermomechanical finite element model adopting the proposed friction model was established, whose predictive capability and generalization performance with respect to variations in welding parameters and geometric size were systematically validated by comparisons with IFW experiments under different conditions. The simulation accurately reproduced the rotational speed decay, temperature evolution, axial shortening distance, and flash morphology, and the prediction errors were generally < 10%, except for a few outliers. Comparisons with conventional friction models further demonstrated the superior generalization performance of the machine learning–enhanced friction model. The weld forming behavior under different welding parameters was also discussed for welding components.