基于深度学习技术的金属构件残余应力场反演

金属构件中残余应力会对材料结构的力学性质及稳定性产生重要的影响,因此准确高效地确定构件残余应力分布至关重要。传统用于确定残余应力的方法主要有实验法,有限元仿真和反演方法,然而这些方法成本高昂,计算效率低无法满足实际工程的需要。为了解决上述问题,本文发展了一种基于深度学习技术的高效残余应力反演方法。该方法使用人工神经网络技术代替了有限元模型修正的过程,仅通过试件表面有限测点即可获得整个表面的残余应力分布。与传统有限元模型修正法相比,该方法显著地提高了反演效率,并通过一个方形开孔弹塑性仿真模拟对其效率和准确性进行了验证。

Inverse Identification of Residual Stress Field in Metal Component Based on Deep Learning

Residual stress within a metal component can significantly affect the mechanical performance and stability of a structure. Therefore, accurate measurement of residual stress distribution within the component is of primal importance. Conventional methods for residual stress determination primarily include experimental testing, finite element simulation and inverse identification. However, these methods suffer from disadvantages of high testing costs and low inverse efficiency, which cannot be easily adopted in practical engineering. To solve these shortcomings, this study developed a high-performance method based on a deep learning technique. In this method, an artificial neural network was used to replace the finite element calculation in the finite element model updating (FEMU) technique and the residual stress distribution of the component was inversely obtained based on the measured residual stresses on a finite number of measuring points. Compared with the conventional FEMU technique, the inverse identification efficiency of the proposed method was considerably improved. Additionally, the accuracy and efficiency of the method were verified by a simulated square open-hole experiment considering an elastic-plastic metal material.