OMS-DNN: A Surrogate Model for Forward Design of Mirror Structure and Optical System

In the digital era, the demand for real-time diagnostics and control in mirror structure and optical system design is rapidly increasing, which in turn has led to a growing reliance on resource-intensive numerical computations. To lower the entry barrier for designers and address the time-consuming nature of these computations, we introduce OMS-DNN. This framework employs a fully data-driven deep learning architecture designed to replace computationally expensive finite element analysis and ray tracing while integrating optimization algorithms to generate target design solutions for forward design. OMS-DNN leverages deep neural networks to capture complex nonlinear relationships and incorporates a graph convolutional network (GCN) to uncover the topological relationships between parameters. We conducted experiments on largescale datasets, including 20,000 structural design samples and 100,000 Three-Mirror Anastigmat (TMA) optical design samples. OMS-DNN achieved mean absolute percentage errors (MAPE) of 0.98%, 4.09%, and 1.66% across three evaluation metrics. To meet forward design requirements, we integrated the proposed architecture with an optimization algorithm. Experimental results demonstrate that our method can rapidly generate constraint-compliant parameter design solutions tailored to various needs, thereby lowering the entry barrier for designers. These findings validate the effectiveness of OMS-DNN and its potential to transform traditional design workflows.