首页|基于潜高斯过程引入理论先验的钛合金结构件残余应力场推断方法

基于潜高斯过程引入理论先验的钛合金结构件残余应力场推断方法

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
残余应力(Residual stress，RS)是导致钛合金结构件加工变形的主要原因。钛合金结构件的残余应力包括初始残余应力(Initial residual stress，IRS)和加工表层残余应力(Machined surface residual stress，MSRS)，其中，MSRS是加工表层区域的IRS与高水平的加工残余应力(Machining-induced residual stress，MIRS)耦合作用的结果。结构件的加工变形控制是航空航天工业亟需解决的重要问题，准确获取结构件残余应力场的是加工变形精确预测和控制的基础。然而，现有的残余应力预测方法难以考虑零件制造过程中的各种不确定性，导致预测精度有限。现有的测量方法仅能在样件中测量局部的残余应力，对于大型结构件残余应力场测量，测量效率低。针对以上挑战，本文提出了一种贝叶斯框架下同时推断钛合金结构件IRS和MSRS的方法。该方法将不可观测的IRS和MSRS建模为潜高斯过程，将不同区域的MSRS场之间存在相关性这一先验知识通过具有共享协方差的核函数融入潜高斯过程，并利用可观测的变形力对残余应力场进行推断。本方法提供了一种从概率角度利用变形力数据推断零件残余应力场的有效手段，为后续变形控制策略优化提供了可靠依据。
Inference Method for Residual Stress Field of Titanium Alloy Parts Based on Latent Gaussian Process Introducing Theoretical Prior
Residual stress(RS)within titanium alloy structural components is the primary factor contributing to machining deformation.It comprises initial residual stress(IRS)and machined surface residual stress(MSRS),resulting from the interplay between IRS and high-level machining-induced residual stress(MIRS).Machining deformation of components poses a significant challenge in the aerospace industry,and accurately assessing RS is crucial for precise prediction and control.However,current RS prediction methods struggle to account for various uncertainties in the component manufacturing process,leading to limited prediction accuracy.Furthermore,existing measurement methods can only gauge local RS in samples,which proves inefficient and unreliable for measuring RS fields in large components.Addressing these challenges,this paper introduces a method for simultaneously estimating IRS and MSRS within titanium alloy aircraft components using a Bayesian framework.This approach treats IRS and MSRS as unobservable fields modeled by Gaussian processes.It leverages observable deformation force data to estimate IRS and MSRS while incorporating prior correlations between MSRS fields.In this context,the prior correlation between MSRS fields is represented as a latent Gaussian process with a shared covariance function.The proposed method offers an effective means of estimating the RS field using deformation force data from a probabilistic perspective.It serves as a dependable foundation for optimizing subsequent deformation control strategies.

titanium alloyresidual stress field inferencelatent Gaussian processmachining deformation

陈俊松、刘长青、赵智伟、王伟、向兵飞、危震坤、李迎光

展开 >

南京航空航天大学机电学院,南京 210016,中国

舍夫德大学工程科学学院,舍夫德,瑞典

江西洪都航空工业集团有限责任公司,南昌 330096,中国

钛合金 残余应力场推断 潜高斯过程 加工变形

National Key R&D Program of ChinaNational Science Fund for Distinguished Young ScholarsGeneral Program of the National Natural Science Foundation of China

2022YFB34026005192550552175467

2024

10.16356/j.1005?1120.2024.02.001

南京航空航天大学学报(英文版)
南京航空航天大学

南京航空航天大学学报(英文版)

CSTPCD
影响因子：0.279
ISSN：1005-1120
年,卷(期)：2024.41(2)
  • 22