Computational Engineering and Physical Modeling (Oct 2022)
Adaptive Neuro-Fuzzy Inference System (ANFIS) Integrated with Genetic Algorithm to Optimize Piezoelectric Cantilever-Oscillator-Spring Energy Harvester: Verification with Closed-Form Solution
Abstract
Piezoelectric vibration-based energy harvesters (PVEHs) are designed to convert mechanical energy into electric energy. Researchers deal with issues like inefficient amount of energy and frequency bandwidth. Optimizing and widening the PVEH can address the issues. As a modification to the dynamic magnification concept of conventional PVEH, a novel integrated oscillatory multisystem of cantilever-oscillator-spring is proposed. In this project maximizing the widened effective frequency bandwidth with respect to the oscillator mass and spring constant is the main goal. The closed-form voltage function obtained numerically-analytically is expensive in terms of computational time and cannot be used in the genetic optimization. In this regard, soft computing techniques is adopted. Utilizing adaptive-neuro-fuzzy-inference-system (ANFIS), a regressor model is designed to estimate voltage function evaluations in the genetic optimization, such fuzzy system is tuned with decent type and number of membership functions according to the root-mean-square-error criteria. Fuzzy inference system (FIS) is implemented using 64 and 49 fuzzy rules derived from Gaussian membership functions (MFs) and passed to the genetic algorithm initiating with 100 iterations and 30 populations. Using roulette wheel, tournament, and random selection methods, optimal values of the mass and stiffness ratios are found to yield the most widened frequency bandwidth. Findings reveal integration of the proposed oscillator-spring subsystem drastically reinforces utmost generated voltage. Furthermore, tuning parameters result the maximum widened frequency bandwidth which improves the harvester performance up to 3 times the conventional values.
Keywords
