A random tunneling algorithm for the structural optimization problem

Abstract

Based on the concept of subenergy transformation and the terminal repeller in the terminal repeller unconstrained subenergy tunneling (TRUST) algorithm, a global optimization algorithm, called the random tunneling algorithm (RTA), is proposed. RTA is a two-phase optimization method in which a global phase is carried out by random tunneling and a local phase by gradient optimization with the BFGS method. In RTA, the population of start points is generated randomly, and the similarity checking is performed during the cycles of the two phases. RTA was compared with other well-known stochastic methods in the optimization of the standard test functions. The potential energy function of Lennard-Jones clusters was also selected as an instance to investigate the performance of the RTA in high-dimensional optimization. All the global minima of LJ clusters containing up to 100 atoms were successfully located. The results indicate that RTA may be a good tool for the structural optimization problem.