Structural Optimization Design of MEMS Piezoelectric Hydrophones Based on Intelligent Learning Algorithm

Hydrophones rely on high sensitivity to detect weak hydroacoustic signals, with the diaphragm structural design significantly impacting the sensitivity of MEMS piezoelectric hydrophones. Traditional optimization methods for hydrophone structure involve creating an equivalent circuit mathematical model or a finite-element analysis (FEA) model through simulation, starting with a structural model based on equivalent analogy or empirical knowledge, and proceeding with a parametric analysis of geometric structural parameters. However, these methods are computationally intensive and time-consuming, impeding structural design optimization. This article presents an intelligent learning (IL) algorithm-based method for optimizing MEMS hydrophone structures, which aims to simplify the design process and enhance hydroacoustic reception. Orthogonal tests and FEA models are utilized to collect the simulation data, and neural network models are developed to illustrate the relationship between the structural parameters and hydrophone performance. The diaphragm electrode thickness, piezoelectric layer thickness, diaphragm radius, and top electrode radius are optimized through a backpropagation (BP) neural network enhanced by the particle swarm optimization algorithm. The model’s accuracy is validated through simulation testing, with the optimized hydrophone sensitivity and effective electromechanical coupling coefficient averaging 99.77% and 96.03%, respectively. A MEMS hydrophone is fabricated based on the design model’s optimization results, and experimental testing confirmed the design performance. This method can meet the design requirements of MEMS hydrophones with improved efficiency and precision and can be extended to other MEMS sensors and actuators requiring structural parametric design.