The rapid advancement in technology and engineering leads to increasingly complex structural working conditions. Especially, in the field of aeronautics and astronautics, structures are frequently subjected to high temperatures together with external forces, posing great threat to structural health. Consequently, the identification of both mechanical and thermal loads is crucial for structural health monitoring. However, the presence of thermal fields introduces additional challenges to load identification problems. In this paper, a model-based mechanical-thermal load identification method is firstly proposed with structural mechanical-thermal modal analysis. A data-driven self-adaptive model updating strategy based on artificial neural networks (ANN) is then developed in order to account for the effect of thermal fields on the structure, including changes in material properties arisen from high temperatures and changes in structural properties arisen from thermal stress. Furthermore, the study considers two types of uncertainties — Gaussian white noises and structural uncertainties and interval estimate of load identification based on Taylor expansion and uncertainty, and carries out uncertainty propagation analysis to quantify their effects on load identification. Eventually, simulation and experiment examples are provided to validate the effectiveness of the proposed methodology. The results of numerical examples demonstrate that the mechanical-thermal identification method converges rapidly, achieving an overall accuracy exceeding 95 % (namely, an error <5 %) in the identification of both mechanical and thermal loads, and the interval estimate can successfully envelope the real values of the loads.

中文翻译：

---

一种考虑多源不确定性的基于模型的混合数据驱动的机械-热动力载荷辨识方法


技术和工程的快速发展导致结构工作条件越来越复杂。特别是在航空航天领域，结构经常受到高温和外力的影响，对结构健康构成巨大威胁。因此，识别机械载荷和热载荷对于结构健康监测至关重要。然而，热场的存在给载荷识别问题带来了额外的挑战。本文首先提出了一种基于模型的机械-热载荷辨识方法，并进行了结构--热-模态分析。然后开发一种基于人工神经网络 （ANN） 的数据驱动的自适应模型更新策略，以考虑热场对结构的影响，包括高温引起的材料特性变化和热应力引起的结构特性变化。此外，该研究考虑了两种类型的不确定性——高斯白噪声和结构不确定性以及基于泰勒展开和不确定性的载荷识别区间估计，并进行了不确定性传播分析以量化它们对载荷识别的影响。最终，提供了仿真和实验实例来验证所提出的方法的有效性。数值算例结果表明，机械-热识别方法收敛迅速，在机械载荷和热载荷的识别中实现了超过 95% 的总体精度（即误差 <5 %），并且区间估计可以成功地包络载荷的实际值。