Information-Driven Optical Monitoring of Microforming Process by Training Sensed Data Related to Mechanical Motion, Deformation, and Variables

High-quality microforming depends on various process states, but the multiple sensors lead to instability in dynamic processes. In microforming, information-driven optical monitoring is proposed to diagnose the multistate variables by training the in-process sensed data. The objective is to online trace mechanical motion, deformation, and variables. First, the microforming process was informatized by optical sensing dynamic changes. In order to predict the fluctuation data by prior probability information fitting of the screened nearest sensed data, K-Gaussian process correction (KGPC) was constructed to correct input process variables to the actual process variables. Finally, the errors of monitoring information were analyzed to determine the decision thresholds adapted to the model. It is shown that the data variation of optic sensing is sensitive to multiple process states. The KGPC monitors process variables and mechanical motion with an accuracy of 96%–98% and predicts mechanical deformation only by 65% due to equipment fluctuation. Moreover, it may control information accuracy in process diagnosis by decision thresholds.