【摘要】 振动是金属构件疲劳失效的重要因素,残余应力可以表征金属构件疲劳状态。然而残余应力在构件疲劳过程中演化行为复杂,传统寿命模型通常适用于残余应力缓慢松弛过程,且较少考虑初始残余应力、冷作硬化、材料差异性等影响,评估误差大。该研究融合Kodama物理模型和基于高斯过程的数据驱动方法,建立了物理模型和高斯过程融合驱动的疲劳状态评估模型。通过铝合金疲劳试验,证明该方法可减少试样材料、表面强化工艺等差异性对评估结果的影响,提高残余应力演化模型准确性。更多还原

【Abstract】 Vibration is an important factor in the fatigue failure of metal components, while residual stress can characterize the fatigue state of metal components. So, residual stress is an important factor affecting the fatigue life of metal components. The introduction of residual compressive stress through surface strengthening technique, like shot peening, can improve the life of components. However, the evolution behavior of residual stress in the process of fatigue is complex. The traditional life model is usually suitable for the slow relaxation process of residual stress, and less consideration is given to the effects of initial residual stress, cold work hardening, material difference and so on. Large evaluation error will result by using these models. In the paper, a physical model and the data-driven method based on Gaussian process were fused, and a component cycle prediction model driven by the fusion of Kodama physical model and Gaussian process(K-GP) was established. The fatigue tests of 2024 aluminum alloy shot peening specimens prove that the model proposed can reduce the influence of differences in specimen materials and surface strengthening processes on the evaluation results and improve the accuracy of the residual stress evolution model.更多还原