Transient-Simulation-Driven Model Order Reduction for Efficient Periodic Steady-State Analysis via Shooting Method

This paper presents an efficient method for analyzing the periodic steady-state of non-linear circuits via shooting method. Conventional Newton Shooting Method (NSM) suffers from expensive computational complexity and intensive memory usage when dealing with large-scale circuits. The proposed technique integrates Model Order Reduction (MOR) into the NSM framework, creating a smaller, surrogate model that leads to substantial reduction in both CPU time and memory consumption. The proposed method also demonstrates the advantages of using a least squares Petrov-Galerkin (LSPG) projection against the standard Galerkin projection (GP). The resulting ROM, once generated, can be reused for different input stimuli, not limited to the nominal input, enabling exploration of stimulus variations. To further alleviate computational overhead proportional to the full-order model size, the approach leverages hyper-reduction via Interpolatory Proper Orthogonal Decomposition (IPOD). The method is validated through several numerical examples, showcasing its accuracy, efficiency, and capability to handle different input stimuli with a single reduced model.