Robustness estimation of simple lens systems by machine learning

Abstract

Tolerance analysis and tolerance sensitivity optimization (desensitization) are important and necessary for manufacturability. However, compared to the optimization of optical performance, tolerance analysis is still time-consuming. A machine learning approach for the fast robustness estimation of lens systems is proposed. The results of the machine learning estimation and the other four different methods are compared with the results of the Monte Carlo analysis. The proposed model is added to the merit function in commercial software for optimization to reduce the sensitivity.