Structure Design Optimization of a Differential Capacitive MEMS Accelerometer Based on a Multi-Objective Elitist Genetic Algorithm

This article describes a global structure optimization methodology for microelectromechanical system devices based on a multi-objective elitist genetic algorithm. By integrating a parameterized model with a multi-objective evolutionary framework, the approach can efficiently explore design space and concurrently optimize multiple metrics. A differential capacitive MEMS accelerometer is presented to demonstrate the method. Four key objectives, including resonant frequency, static capacitance, dynamic capacitance, and feedback force, are simultaneously optimized to enhance sensitivity, bandwidth, and closed-loop driving capability. After 25 generations, the algorithm converged to a uniformly distributed Pareto front. The experimental results indicate that, compared with the initial design, the sensitivity-oriented design achieves a 56.1% reduction in static capacitance and an 85.5% improvement in sensitivity. The global multi-objective optimization achieves a normalized hypervolume of 35.8%, notably higher than the local structure optimization, demonstrating its superior design space coverage and trade-off capability. Compared to single-objective optimization, the multi-objective approach offers a superior strategy by avoiding the limitation of overemphasizing resonant frequency at the expense of other metrics, thereby enabling a comprehensive exploration of the design space.