A global optimization method is proposed to improve the conventional method of simulated annealing. By introducing the probability distribution function for the objective function and the concept of stable energy for detecting thermal equilibrium during annealing, the selection of initial temperature and equilibrium criterion becomes easy and effective. Furthermore, the efficiency and robustness of the proposed method is retained by employing the technique of region reduction and an adaptive neighborhood structure. In the case where multiple (global) optima may exist, a technique based on the method of simulated evolution is developed to circumvent the difficulty of convergence of population. Numerical studies of some standard test functions and an optimum structural design problem show that the proposed method is effective in solving global optimization problems.