بهینه‌سازی ابعادی برداشت کننده انرژی ارتعاشی پیزوالکتریک با استفاده از شبکه عصبی و الگوریتم ژنتیک

Given the growing demand for clean energy, renewable energy sources have recently attracted significant attention. One such source is vibrational energy, which is inexpensive and environmentally friendly, making it suitable for low-power devices such as sensors, actuators, and wearable smart devices, while eliminating the need for batteries. This study focuses on the analysis and dimensional optimization of a piezoelectric vibration energy harvester consisting of a silicon-based cantilever beam with an attached tip mass. For this purpose, 48,000 data points were generated using the finite element method for various geometric configurations. After filtering, the data were used to train two separate neural networks: one for predicting the natural frequency and the other for estimating the output voltage. A genetic algorithm with 300 generations was then applied to these neural network models to determine the optimal beam dimensions. To validate the optimization results, the Euler–Bernoulli beam vibration equations were employed. The results show that the genetic algorithm increased the figure of merit from 0.187 (the maximum value among the initial data set) to 0.251 for the optimized design. For these optimal dimensions, the predicted natural frequency and output voltage were 87.4 Hz and 0.237 V, respectively. The corresponding values obtained from the beam vibration equations were 84.9 Hz for the natural frequency and 0.248 V for the output voltage. which demonstrates good agreement between the results of FEM and equations.