Residual stresses field estimation based on deformation force data using Gaussian process latent variable model

Residual stresses field inside the bulk material is the main factor of the part deformation, which is a vital issue for large-scaled monolithic components manufacturing in aircraft industry. The estimation of residual stresses is the base of the part deformation control. Non-uniform residual stresses field is distributed among the whole part, and differs from part to part. However, existing residual stress measurement’s methods can only measure the sample’s residual stress and the detection depths are limited by physical principle, which makes it hard to measure the heavy thickness part. In order to address this issue, this paper presents a method of estimating part residual stresses fields and their uncertainty in machining process based on a novel Bayesian statistical model integrated with observation data (deformation force). Deformation forces data can be easily and accurately monitored during machining processing via several fixture devices, and contain information that can be related to the residual stress. In order to solve the problem of how to infer an unobservable residual stresses field by using sparse observed deformation force data, this work introduces a Bayesian framework. The unmeasurable and unobservable residual stresses field is deemed as the latent physics field and the Gaussian process is specified for the latent field as a prior with parameters in kernel function to reduce the solution space and also introduces proper Distribution hypothesis on deformation force data. The model provides an acceptable estimation result with the limited observation data. The proposed method provides an effective way to estimate interior residual stress field in the machining processing of monolithic components part in terms of probability, and will help the selection of deformation control strategy.