SUGGESTION RBFNN MODEL FOR THICK COMPOSITE LAMINATED PLATES STRESS ANALYSIS

Abstract

This work is an attempt to model thick composite laminated plates stress data using radial Basis function neural network architectures. A finite element model with embedded degraded zone in laminated composite plates is developed through training the neural network, because in such case involves many data sets and all of these data are difficult to generate using experiments. Therefore, Training data was performed by using both experimental and finite element program for different dimensions (a, b, h /m) with static and dynamic loads conditions. The neural network model has 6 input nodes representing the load (L) and thick composite size (a=length, b=width and h=thickness), type of lamination, and time. While eighteen nodes at hidden layer and one output node representing the Stress. Dimensions of Plates [a(0.125-0.25)m, b(0.125-0.25)m, h(0.0125-0.025)m], and Static load condition (10) qo(kN/m2), the dynamic load condition (10-100) qo(kN/m2), to(0.05-0.0005sec.) Analytical studies on the performance of the neural network are compared to independent solutions and measured data and found to be in excellent agreement. Verifications are made for both finite element code results and utilization of neural networks without performing too many, long experiments.