Digital twin-driven centering process optimization for high-precision glass lens

Abstract

Glass lenses are key photoelectric components of laser systems in semiconductor equipment. The circularity, surface quality, and optical axis precision of the glass lens directly affect the efficiency of a laser system. In the processing of conventional glass lenses, centering is often used to meet specification requirements. However, high-precision processing of materials with high hardness and low ductility is difficult. The process of centering a glass lens is performed according to processing parameters determined by experience, the degree of wear of the grinding wheel, and specifications of the glass lens. Circularity errors, edge cracks, and optical axis errors are common defects in lens processing; such defects must be identified through manual inspection and thus result in high scrap rates and production costs. To address this problem, a digital twin-driven (DTD) centering process optimization system for high-precision glass lenses was developed in this study. The proposed system combines an omnidirectional information model of production line manufacturing data, a virtual process model based on sensor signals, and a design of experiment–based genetic algorithm (DOE-GA). The models of the system are connected through data flow and operate simultaneously to create a high-fidelity virtual simulation of the centering process. To facilitate evaluation of the performance of the proposed DTD centering process, the process was implemented in an industrial setting. The test in the real-world production line shows that the proposed DTD system reduces the time of process development from 4 h to 1 h, decreases the inspection from full to 10% sampling inspection and improves the yield rate by 20%.