AI - driven predictive modeling for micro plates

Abstract

In the realm of micro electromechanical systems, comprehending their behaviour is paramount importance for practical applications. Analysis of these structures often involves grappling with size effects, rendering traditional methods like finite element analysis complex. To confront this challenge head-on, this study endeavors to develop a micro-plate model by integrating the modified couple stress theory, thereby capturing size effects inherent in electromechanical systems. The primary focus lies in probing critical loads and resonant frequencies of micro-plates through a meticulous parametric and comparative inquiry. This investigation encompasses an examination of the influence of length scale parameters, material gradient in- dices, and aspect ratios as prescribed by the modified couple stress theory. To surmount the constraints of conventional techniques, we establish a comprehensive database encompassing diverse geometric and material properties, alongside an array of boundary conditions. This rich dataset amalgamates results derived from the analysis of microplates utilizing methodologies like the Differential Quadrature or Finite Element Method (DRM or FEM), in addition to findings from other researchers. Moreover, this paper pioneers the integration of Deep Learning Neural Networks (DLNN) and Support Vector Machines (SVM) to predict critical loads and natural frequencies for any microplate within the specified parameters. The SVM lever- ages hyperplanes and Kernel functions for data classification and regression, while the DLNN, with its intricate network architecture, minimizes errors through forward and backward propagation. The trained models exhibit remarkable adaptability across various problem types, representing a significant stride in applying Artificial Intelligence (AI) methodologies to delve into the stability and dynamics of microplates.