Optimizing Segmented Bimorph Piezoelectric Harvesters Using Advanced Bidirectional Long Short Term Memory Network

This study aims to optimize the geometric design of a bimorph segmented linearly tapered piezoelectric harvester (BSLTPH) to enhance low-frequency and multimode vibration energy harvesting. The objective is to achieve reduced resonance frequencies and improved power generation through simultaneous optimization of beam length, width, and taper ratio.

A finite element method (FEM)-based dataset was generated using multiple BSLTPH configurations with varying geometric parameters. An Advanced Multihead Cross-Attention Bidirectional Long Short-Term Memory (AMCA-BiLSTM) network was developed to establish the relationship between harvester geometry and performance characteristics. The network hyperparameters were optimized using the Coati Optimization Algorithm (COA), while a Genetic Algorithm (GA) was employed to identify the optimal structural design. The FEM model was validated against established analytical and numerical benchmark models prior to optimization.

The optimized harvester exhibited enhanced multimode vibration energy harvesting performance and reduced operating frequencies compared with conventional rectangular and trapezoidal piezoelectric harvesters. The optimal configuration consisted of a beam length of 121.3 mm, width of 71.56 mm, and taper ratio of 0.7682. The optimized design operated within frequency bands of 38-608 Hz and 57.5-455 Hz and demonstrated substantially higher voltage and power outputs across multiple vibration modes. Comparative evaluation showed that the segmented trapezoidal harvester generated higher power than conventional configurations, particularly in the second and third vibration modes.

The proposed FEM-AMCA-BiLSTM-GA framework provides an effective approach for the design optimization of piezoelectric vibration energy harvesters. The optimized BSLTPH achieves improved multimode energy harvesting capability, lower resonance frequencies, and enhanced power generation, making it suitable for low-frequency vibration-powered sensing and self-powered electronic applications.