Free Vibration Analysis of Jute-Epoxy Polymer as Skin Layers and Tea Waste Epoxy as a Core Layer-Based Sandwich Plate Through Experimental and AI Approach

Abstract

The current research work is based on the free vibration analysis of the sandwich plate with jute-reinforced composite skins and tea-waste epoxy-based core through numerical and experimental approaches. The fabrication of skin and core layers for the proposed sandwich plate is accomplished through the hand layup technique. The elastic constants of the core layer and skin layers are obtained through the uniaxial tensile test in the universal testing machine (UTM) INSTRON 5967 as per ISO-527-5 standards. The natural frequency of the proposed sandwich plate is obtained by performing a set of experiments using the modal impact hammer test and Fast Fourier Transform (FFT) analyzer working on Pulse Labshop. The numerical simulation of the sandwich plate is performed using the finite element software ABAQUS by utilizing the elastic constants obtained from the uniaxial tensile test. The 3D solid element having three degrees of freedom per node is adopted for the numerical simulation of the proposed sandwich plate. The accuracy of the numerical results has been verified with available literature and found to be in very close agreement. Later, the effect of various parameters such as aspect ratio, core thickness ratio and fibre ply orientation under different edge conditions on the natural frequency of the sandwich plate is investigated. Subsequently, the prediction model for the natural frequency of the sandwich plate is developed through the Adaptive Neuro-Fuzzy Inference System (ANFIS) under different sets of input variables. An extensive data set obtained from numerical simulation is used to train and test the ANFIS models, and several statistical metrics are used to assess each model's accuracy as well. The natural frequency of sandwich plates obtained for different sets of unseen input parameters through ANFIS and numerical simulation in ABAQUS is very close with less than 5% errors.