Optimization of material composition of nonhomogeneous hollow sphere for thermal stress relaxation making use of neural network (Q1968616)

The paper deals with an application of neural networks to the optimization of material composition for a nonhomogeneous hollow sphere with arbitrary distributed and continuously varied material properties (such as functionally graded materials). Using an analytical procedure for a laminated hollow sphere model, the authors obtain an approximate analytical solution for the temperature of the nonhomogeneous hollow sphere. Furthermore, the thermal stress components are determined under the mechanical condition of being traction free. As a numerical example, the authors consider a nonhomogeneous hollow sphere composed of zirconium oxide and titanium alloy. Also, as the optimization problem of minimizing the thermal stress distribution, the numerical calculations are carried out making use of a neural network, and the optimum material composition is determined taking into account the dependence of material properties on temperature. Finally, the authors compare the results obtained by using the neural network and the ordinary nonlinear programming method.