Analysis and inference of stream of dimensional errors in multistage machining process based on an improved semiparametric model

In the multistage manufacturing process, the dimensional errors generated in previous stage will affect the dimensional errors in the next stage. Considering that the machining process is a typical time-varying nonlinear system, the determination of dimensional errors and the analysis of stream of errors have great challenges. The analysis and inference approaches based on theoretical model and data-driven model have some shortcomings in accuracy. In order to solve the above problems, an improved semiparametric model integrating engineering knowledge, mechanism model and measurement data is proposed. Transfer error, system error and random error among multistage machining process are decomposed and analyzed. Besides, the uncertainties within transfer coefficient and system error are quantified and calibrated through Bayesian framework. The effectiveness of the model is verified through the multistage machining process of turboshaft. The deviation between inference results and measurement results is 6.9026μm. Compared with the traditional parametric model, nonparametric model and semiparametric model, the proposed improved semiparametric model shows higher inference accuracy.