Optimization of material composition of nonhomogeneous hollow sphere for thermal stress relaxation making use of neural network

In this study, a neural network is applied to optimization problems of mate rial compositions for a nonhomogeneous hollow sphere with arbitrarily distr ibuted and continuously varied material properties such as functionally gra ded material (FGM). Using the analytical procedure of a laminated hollow sp here model, the analytical temperature solution for the nonhomogeneous holl ow sphere is derived approximately. Furthermore, the thermal stress compone nts are formulated under the mechanical condition of being traction free. A s a numerical example, the nonhomogeneous hollow sphere composed of zirconi um oxide and titanium alloy is considered. Also, as the optimization proble m of minimizing the thermal stress distribution, the numerical calculations are carried out making use of neural network, and the optimum material com position is determined taking into account the effect of temperature-depend ency of material properties. Furthermore, the results obtained by neural ne twork and ordinary nonlinear programming method are compared. (C) 1999 Else vier Science S.A. All rights reserved.