To address the high-performance sustainable manufacturing requirements of ceramic composite, surface roughness is regarded as an indispensable technical indicator for improving the machined product quality. Whereas, the complicated material removal mechanism of heterogeneous composite materials in the micro-milling operation would limit the in-process prediction performance of surface roughness. In order to improve prediction performance and overcome solving difficulties, this paper develops a PSO-GPR-based stochastic-aware recognition system of in-process surface roughness for micro-milling operation of ceramic composite. Considering the close association with surface generation of machined products, the instantaneous specific machining energy consumption model of ceramic composite in micro-milling operation is proposed as the foundation for recognizing in-process surface roughness values. More importantly, compared with the traditional cutting mechanism analysis, the real-time volumetric fraction of carbon fiber in ceramic composite is introduced into the heterogeneous material removal mechanism, in which the influence of stochastic fiber distribution has been considered. Accordingly, the in-process surface roughness recognition model derived from the stochastic fiber distribution of ceramic composite is established by integrating the particle swarm optimization (PSO) methodology and the Gaussian process regression (GPR) algorithm, which is independent of extensive experiment data and training data. Furthermore, the prediction performance of the proposed PSO-GPR-based stochastic-aware recognition methodology for in-process surface roughness values has been validated by conducting a series of micro-milling experiments with ceramic composite. The average predictive error is 2.79% and the maximum predictive error is 13.27%. Correspondingly, the values of MSE, MAE and MAPE are 0.00039, 0.0150 and 2.77%, respectively.
扫码关注我们
求助内容:
应助结果提醒方式:
