Prediction of stability and buckling strength plays a vital role in determining the safety of the structures at design loads. The analysis of composite plated structures being anisotropic in nature is more involved. Artificial Neural Network (ANN) has been found to be a tool, which can accommodate the non-linear behavior of the systems. This has led to the increasing application of ANN for the solution of highly involved structural engineering problems. Prediction of the stability of composite plates with geometry singularities using artificial neural networks (ANN) has been investigated in the present paper. Using the available experimental data, an ANN model with input parameters consists of aspect ratio and geometric singularities, tested under uni-axial loading was found to be acceptable in predicting the buckling behavior of composite plates. The network is developed using the available experimental results and the efficacy of the so developed network is tested with a new set of available experimental data.
Keywords: Stability; composite plate; feed forward neural networks; training; testing.
[...] RESULTS AND DISCUSSION In the present investigation a neural network model is developed for GRP plate having a modular ratio of 40 with eccentric cutouts of different shapes. The cut out shapes considered are square, circle, a combination of circle and ellipse, square and rectangle. The different cases are analysed using feedforward network with the two different architectures 3-18- 6-1 and 3-12-1. The input variables are taken as plate aspect ratio eccentricity ratio in both the directions as and (ey/b). [...]
[...] Formulation of the Problem: The present study is focused on the behaviour of composite plates with eccentric holes. In order to emphasize the importance of the presence of a hole, its eccentricity and the shape are varied. The present investigation is carried out by considering holes of square, circular, rectangular and elliptical shape eccentrically about the centroid of the plate. Since the plate with an eccentric hole is a structural component, the buckling behavior of these plates under in-plane loading is studied thoroughly by setting the plate aspect ratio, eccentricity ratio in both the directions and as input variables. [...]
[...] The preparation of data is a matter of considerable importance in training the neural network. In the present study the number of training data is arrived at based on the hypercube rule by . The hypercube is an imaginary cube on which all the combinations of input are located. The corners of the cube and the value combination on the mid-point of the cube face represent the boundary values of the given input. Further, the point inside the cube should be selected. [...]
[...] It is found that the stability of composite plates is primarily a function of thickness, plate aspect ratio, hole size ratio and the orientation of the layers. Here the modeling is carried out for a plate of 5mm thickness with symmetric fibre orientation. Hence these parameters make up the input vector for the neural network while the output is the buckling coefficient. Inorder to obtain acceptable and balanced neural network performance, normalization of the input and output data was conducted using the relation given by where C is a constant between - 0.25 and 0.25 to ensure that the values are in the range of 0.2 and n is the number of digits in the integer part of the variable V. [...]
[...] In the present study a neural network model is developed to predict the stability of composite plates with cutouts under uni-axial loading. The network is developed using the available experimental results. The parameters that affect the stability of the composite plate are taken as the plate aspect ratio and the eccentricity ratio and the stability is defined in terms of the buckling coefficient. After a detailed literature survey to get the optimum architecture, two networks, one with one hidden layer and the other with two hidden layers were arrived at. [...]
using our reader.