which the singular element appears in the optimum process and the result may be a partial optimum solution are two disadvantages when using the ESO method to optimize 2- D model. In this paper,two algorithms are proposed for solving these disadvantages. One is filter and rejection algorithm based on the trait of singular element. Using this algorithm,only one parameter is needed to control filtering and deleting,and the parameter is determined by the number of elements linked around singular element. The other is interval approximation algorithm. Minimum initial rejection rate R,0min,which fulfill removed conditions was ascertained by using this algorithm and it was used as starting value,thus other initial rejection rate R,0, were obtained by setting certain increment. Two algorithms mentioned above were implanted in the existing ESO method,then optimizing respectively with different initial rejection rate which were obtained before and the available structures from the results was filtered out in terms of performance parameters. The results of optimization for numerical 2-D structures indicate that the improved ESO method has better suitability and stability than the existing one,and it avoids shortcoming that the result may be a partial optimum solution when using the existing ESO method.