The compressive residual stresses and grain refinement induced by laser shock peening (LSP) are two very important surface strengthening mechanisms that are responsible for the significant improvements of the mechanical property and fatigue performance of metallic materials. A computation framework linking the dislocation-mechanism-based constitutive model with the dislocation density evolution model was developed and implemented into ABAQUS/Explicit code for numerical simulations of the residual stresses and grain refinement induced by LSP of TC4 titanium alloy. Two kinds of three-dimensional finite element models involving the plate model and TC4 blade model were respectively created to simulate the multi-LSP processes, and the effect of laser spot overlap ratio was accordingly investigated in detail. The obtained results show that with the increase of laser spot overlap ratio, the depths of both the compressive residual stresses and grain refinement increase, and with respect to the same depth, the compressive residual stresses increase and the refined cell size decreases. An attempt was made to employ the artificial neural networks (ANNs) to estimate the residual stresses and grain refinement induced by LSP of the whole TC4 blade model, and the ANNs-estimation approach presents the excellent efficiency in predicting the multi-LSPed results with inexpensive cost of numerical computation.

激光冲击喷丸（LSP）引起的压缩残余应力和晶粒细化是两个非常重要的表面强化机制，它们可显着改善金属材料的机械性能和疲劳性能。开发了将基于位错机制的本构模型与位错密度演化模型联系起来的计算框架，并将其实施到ABAQUS / Explicit代码中，以对TC4钛合金的LSP引起的残余应力和晶粒细化进行数值模拟。分别创建了包括平板模型和TC4叶片模型的两种三维有限元模型来模拟multi-LSP过程，并据此详细研究了激光光斑重叠率的影响。所得结果表明，随着激光光斑重叠率的增加，残余压应力和晶粒细化的深度均增加，而相对于同一深度，残余压应力增加，细化的晶胞尺寸减小。尝试使用人工神经网络（ANN）估计整个TC4叶片模型的LSP引起的残余应力和晶粒细化，并且ANNs估计方法以低廉的价格提供了预测多LSPed结果的出色效率。数值计算成本。