Design optimization using fast annealing evolution algorithms.

摘要

In this thesis, Fast Annealing Evolution Algorithm (FAEA), a stochastic global optimum search method is implemented using MatLab programming. The original FAEA is for solving unconstrained optimization problems (FAEA was developed to optimize the Lennard-Jones clusters problem in computational chemistry). In this thesis, FAEA has been extended to solve constrained optimization problems. Additionally, FAEA is also extended to find multiple solutions in the design domain. This is called Multiple Solution Fast Annealing Evolution Algorithm (MS-FAEA) in this thesis. The effects of controlling parameters on the result and number of function evaluations are found by numerical experiments on standard test problems. To apply Fast Annealing Evolution Algorithm to solve design optimization problems, a new hybrid constraint handling method is designed and implement into FAEA and MS-FAEA. The algorithm has also been modified to handle mix variables. Three engineering design problems are tested on the single solution and multi-solution methods. These problems are the truss design problem, the spring design problem and the pressure design problem.