基于贝叶斯高斯过程的激光喷丸固有应变响应建模研究  

Research on inherent strain response modeling of laser peen forming based on Bayesian Gaussian process

作　　者：林杰 童一飞[1] Lin Jie;Tong Yifei(School of Mechanical Engineering,Nanjing University of Technology,Jiangsu Nanjing,210094,China)

机构地区：[1]南京理工大学机械工程学院,江苏南京210094

出　　处：《机械设计与制造工程》2024年第12期1-10,共10页Machine Design and Manufacturing Engineering

基　　金：国家自然科学基金(51375305,51075271)。

摘　　要：基于数值仿真的激光喷丸成形固有应变反求是一个过程复杂、计算成本高且结果不精确的物理模型,为解决该问题,聚焦于激光喷丸成形过程中的一个重要中间物理量——固有应变,将其作为核心研究对象进行深入探讨。首先,基于数值仿真预测模型与贝叶斯高斯过程方法,建立了激光喷丸成形固有应变响应模型。然后,通过马尔科夫蒙特卡洛采样方法实现了对基于数值仿真反求固有应变模型的参数标定,并预测了带有不确定性量化分析的固有应变后验分布结果。最后,利用固有应变的标定结果作为训练数据集,通过贝叶斯高斯过程的数据分析技术,成功构建了针对固有应变的函数型响应模型,进而运用此模型实现了对工件形变的精准预测,有效提高了激光喷丸成形工艺预测的计算效率与可靠性。The inverse determination of laser peen forming inherent strain based on numerical simulation is a physical model,which is a complex process with low-accuracy outputs and highly expensive computational cost.To solve these problems,this paper develops a response model of laser peen forming inherent strain based on numerical simulation prediction model and Bayesian Gaussian process method.Finally,using MCMC sampling method to deal with the parameter calibration of the model,it inversely determines the inherent strain based on numerical simulation,and prediction of the inherent strain posterior distribution with quantitative analysis of uncertainty.This effectively ameliorates the calculation efficiency and reliability of the prediction to the laser peen forming process.

关 键 词：激光喷丸成形工艺 固有应变 贝叶斯高斯过程 马尔科夫蒙特卡洛采样方法 参数标定 

分 类 号：TN249[电子电信—物理电子学] V261[航空宇航科学与技术—航空宇航制造工程]