Energy harvesting enhancement of nonuniform functionally graded piezoelectric beam using artificial neural networks and Lichtenberg algorithm

In this paper, we present a novel optimization approach employing Artificial Neural Networks (ANN) to enhance electric power generation in functionally graded piezoelectric tapered harvesters (FGPEH). The FGPEH consists of a host beam composed of functionally graded material, with piezoelectric layers covering its top and bottom surfaces. We formulate a finite element model using Kirchhoff plate assumptions and a four-node quadrilateral element to derive the electromechanical governing equation for the proposed harvester. To maximize the output voltage of the harvester, we optimize five material and geometrical parameters, including tapering parameters, host beam material properties, and the thickness of the piezoelectric layer. The optimization cost is significantly reduced by employing an ANN predictive model, trained on data obtained from numerous random simulations of the numerical model. The trained model achieved an optimal predicted response that surpassed the best database response by 323%, all while reducing computational costs by 80%, utilizing a novel lightning-inspired algorithm known as the Lichtenberg Algorithm (LA). The coupling of ANN and LA in the optimization process results in two designs. The first design, a single non-uniform FGPEH, increases harvested energy by 63%. The second design, a dual non-uniform FGPEH, further enhances energy harvesting by nearly 72%.

Automated summary

This paper presents a novel optimization approach using Artificial Neural Networks (ANN) to enhance the efficiency of a piezoelectric energy harvester. By optimizing material and geometrical parameters, the output voltage of the harvester is maximized, with significant reductions in computational costs achieved through ANN predictive modeling. The coupling of ANN and a lightning-inspired algorithm resulted in two designs that significantly increased energy harvesting.