Investigation and hybrid deep belief neural network-based validation of piezoelectric bimorph cantilever composites assisted with tip mass

Abstract

This work fabricated the piezoelectrical material using PMMA (polymethyl methacrylate) and Ce-doped ZnO nano-powders. The study was conducted to validate the piezoelectric performance of the proposed material and its suitability for the cantilever structure. The PMMA/Ce-ZnO makes the cantilever structure more flexible and performs better. The frequency response is applied to the beam using an electrical circuit to analyze the power and voltage output. The acceleration is given to the cantilever beams to analyze their resonant frequencies. The change in resonant frequencies results in a high voltage and power output. The resistive loads are used in the circuit to find the electrical load. The frequency response is analyzed in three different inner (18 mm, 20 mm, 22 mm) and outer (23 mm, 25mm, 27 mm) active layer lengths of beams. As a result, the maximum voltage of 21.94 V with 13.67 mW power and 2.9 mA current is obtained at a resonant frequency of 51.06 Hz and 1 g acceleration amplitude, which are approximately 42%, 45% and 15% higher voltage, power and current obtained from the lowest performer of proposed piezoelectric cantilevers. The experiment results are validated using the hybrid DBN-SSO (Deep Belief Network based Salp Swarm Optimization) machine learning technique. The proposed DBN-SSO achieved 0.9998 and 0.9966 regression coefficients for voltage and power outputs, thus proving the fitness of predicted results with the experiments. As per findings, a 27 mm inner and 18 mm outer active layer based energy harvesting system is suggested as a suitable energy source where ever 10-12 mW power, 2.5-2.9 mA current and 20-21.94 V voltage are applicable.