Uncertainty analysis and measurement of machining distortion based on Bayesian network

During the machining process of large components from rigid blanks into thin-wall and weakly rigid complex structural parts, geometrical structure characteristics, dynamic structure properties and initial residual stress distribution as well as the state of the surface layer are constantly changing, so the machining of thin-wall parts is full of many uncertainties. Based on the stress-strain mechanical relationship, the influence of residual stress uncertainty on machining distortion is systematically analyzed from the view of stress state uncertainty. The uncertainty of initial residual stress and surface residual stress was evaluated by the joint evaluation method of moment estimation and autoregressive model or the joint evaluation method of least square fitting extrapolation and autoregressive model. The Bayesian network model of machining distortion is constructed, the inference process of machining distortion uncertainty is carried out, and the influence of uncertainty of input factors on machining distortion is quantified by a posterior probability. The results show that the uncertainties of initial residual stress and surface residual stress increase the machining distortion by 17.8% and 1.0%-6.4%, respectively.