Efficient detection of piezoelectric material defects in smart structures using nonlinear vibration and neural networks validation

This paper presents a new technique for detecting defects in piezoelectric materials located inside smart structures, mainly the troubleshooting of functionally graded piezoelectric (FGP) porous plates excited by an electric field. This research accounts for the Von-Karman nonlinearity to study the influence of mechanical and electrical loadings on the dynamic behavior of world material. Maxwell's equations are used to describe the coupling of electric fields with the piezoelectric properties of the plate, and porosity is closely examined for its impact on the defect detection process. Hamilton's principle is used to obtain the equations of motion, which can replicate the system's non-linear dynamics. The harmonic differential quadrature method (HDQM) is used to achieve numerical results, while equations governing the response of the system under different boundary conditions can be discretized appropriately. Deep learning models known as deep neural networks (DNN) are used to validate the mathematical model and extract more information from this complex, large dataset, which is obtained from these simulations. The DNN model provides a robust framework for detecting defects based on learning the complex relationships between different related features of the system, and its efficient classification aids in the detection and classification of defects. The study also explores the parameter selection and optimization in the DNN algorithm, so as to balance the model accuracy and computational efficiency. The results are important as they provide a cost-effective, accurate technique to detect defects in piezoelectric materials and help in smart structure health monitoring, which is a rapidly growing field. The experimental setup detailed in this research can form the basis for future developments of structural diagnostics and the implementation of smart materials in engineering applications.