Record Details

A genetic algorithm based back propagation network for simulation of stress–strain response of ceramic-matrix-composites

DSpace at IIT Bombay

View Archive Info
 
 
Field Value
 
Title A genetic algorithm based back propagation network for simulation of stress–strain response of ceramic-matrix-composites
 
Creator RAO, HSUDARSANA
GHORPADE, VAISHALI G
MUKHERJEE, ASMITA
 
Subject composite materials
computer simulation
neural networks
stress analysis
 
Description Ceramic-matrix-composites (CMCs) are fast replacing other materials in many applications where the higher production costs can be offset by significant improvement in performance. In applications such as cutting and forming tools, wear parts in machinery, nozzles, valve seals and bearings, improvement in toughness and hardness translate into longer life. However, the recent resurgence in the field of development of CMCs has been due to their potential use for the Space Transport systems, Combustion engines and other energy conversion systems. The CMCs are ideal structural material for these applications. However, due to their lack of toughness, they are prone to brittle fractures. Therefore, the main consideration in the development of CMCs has been to toughen them. To achieve this, the bi-material interface should be weak and must allow debonding, resulting in crack deflection. In the present work, the stress–strain response of Al2O3 (matrix)/SiC (whisker) ceramic composite has been simulated using a back propagation neural network (BPN), which incorporates the effect of interface shear strength (IFS) in the analysis. For efficient and quick training, the weights for the BPN have been obtained by using a genetic algorithm (GA). The GA has been modelled with 150 genes and a chromosome string length of 750. The network simulation is based on the stress–strain response obtained from the finite element analysis. A three noded isoparametric interface element has been employed to model the whisker/matrix interface in finite element analysis. The finite element analysis has been carried out only for a limited number of specimens. However, the simulation model is capable of predicting the stress–strain relationship for a new interface shear strength even with this limited information. Thus, the robustness and the generalisation capability of the neural network model is demonstrated. The development stages of the GA/BPN model such as the preparation of training set, selection of a network configuration, training of the net and a testing scheme, etc., have been addressed at length in this paper.
 
Publisher Elsevier
 
Date 2009-03-23T05:10:38Z
2011-11-25T19:59:54Z
2011-12-26T13:07:56Z
2011-12-27T05:55:57Z
2009-03-23T05:10:38Z
2011-11-25T19:59:54Z
2011-12-26T13:07:56Z
2011-12-27T05:55:57Z
2006
 
Type Article
 
Identifier Computers & Structures 84(5-6), 330-339
0045-7949
http://dx.doi.org/10.1016/j.compstruc.2005.09.022
http://hdl.handle.net/10054/1047
http://dspace.library.iitb.ac.in/xmlui/handle/10054/1047
 
Language en