Metamodeling for robust design of energy harvesting devices using polynomial chaos expansion and artificial neural networks

Abstract

The generation of electrical energy using piezoelectric devices represents a promising alternative due to the high charge density these materials can generate. Cantilever beam devices modeled using finite element methods are commonly used in studies focused on the conversion of mechanical energy into electrical energy. With this, the influence of specific variables and parameters can be analyzed through the Frequency Response Function (FRF) of power output. In the search for optimal solutions, it is important to consider uncertainties in parameters in order to design robust devices. Due to the high computational cost of robustness analysis, particularly when considering the mean and relative dispersion, the generation of computational metamodels becomes interesting for their characteristics. This work aims to develop and evaluate a metamodeling approach for the mentioned FRF using polynomial chaos expansion and artificial neural networks, assessing which method provides better accuracy. After the analysis, a method is chosen to generate the metamodel and multi-objective optimization with algorithm NSGA-II is applied to maximize the mean and minimize the dispersion of FRF. The results demonstrate that metamodels can effectively approximate the outcomes obtained from the original function in scenarios characterized by significant uncertainties, with relatively low computational effort.