Design and Intelligent Optimization of Z-Shaped MEMS Accelerometer Structure for Heart Sound Detection

Digital heart sound auscultation technology is crucial for the early diagnosis of coronary heart disease, and the optimal structural design and parameter optimization of high-performance heart sound sensor devices can significantly enhance the quality of heart sound signal acquisition. The determination of conventional sensor structural parameters primarily relies on empirical selection based on simulation analysis results and repeated experimental validation, which is time-consuming and inefficient, making it difficult to reduce research and development cycles and costs. Based on this, this study proposes a novel Z-shaped cantilever beam accelerometer-based heart sound sensor structure and designs a PSO-back-propagation neural network (BPNN) optimization model that combines particle swarm optimization (PSO) and back propagation (BP) neural networks to achieve intelligent analysis and optimization selection of the parameters for the Z-shaped cantilever beam acceleration heart sound sensing structure. The results indicate that the PSO-BPNN demonstrates extremely high accuracy in predicting the maximum stress (S) and natural frequency (F) of the sensor, with relative errors as low as 1.3% and 0.1%, respectively. Furthermore, transient simulation tests using COMSOL with heart sound interpolation signals show that the optimized Z-shaped sensor performs excellently, with an output voltage amplitude reaching 0.025 V, significantly improving its response sensitivity compared to the empirical parameter structure. Therefore, this study not only enhances the efficiency of design and development while reducing empirical errors but also theoretically validates the effectiveness of intelligent analysis and design methods, laying a solid foundation for technological advancements in sensor design and the widespread application of intelligent analysis and design methods.