Structural Optimization Using the Simultaneous Perturbation Stochastic Approximation Algorithm

Abstract

This paper presents an application of the simultaneous perturbation stochastic approximation (SPSA) method to size optimization of structures. This method can predict a gradient approximation that needs only two measurements of the objective function regardless of optimization problem dimension. This characteristic is very promising in reducing the computational cost of optimization process, especially in problems with a large number of variables to be optimized. Furthermore, the stochastic nature of SPSA can enhance the convergence of the method to achieve the global optimum. Some test examples are considered to demonstrate the effectiveness of the method when compared with the other optimization methods found in the literature. Numerical results reveal the computational merits of the SPSA-based method for structural optimization.