Black-box Multi-objective Optimization System Research And Development

Optimization design is a modern design method which is based on the optimization theory in mathematics, computer as the tool to seek the best solution in a variety of Pareto solution. Because the traditional design method of the design of low quality and satisfactory design scheme is time-consuming, so optimization design technology is widely used in different engineering fields, and the study is receiving more and more attention.According to the optimization of multi-objective parameter optimization based on "black-box", the purpose is on the basis of study, based on Matlab software, to develop a multi-objective optimization system, in order to solve the problems such as complex, dealing with complicated, difficult to application. The development system optimization process includes several aspects:analysis of concrete conditions to select the proper experiment design method, then select the appropriate experiment design table; On the basis of experiment design table to arrange test plan, conduct a test; selection contains all patterns of samples for neural network, set up a network model of the relationship between parameters and target values. Using a variant of NSGA-II to complete multi-objective optimization of parameters, the Pareto solution is obtained, and carries on the screen to meet specific Pareto range constraint conditions and the target value of the solution. Adopt the method of fuzzy degree to evaluate the Pareto, to determine the optimal solution. The system also introduced external interface technology, increasing the system's utility optimization.The author based on Matlab platform, through language programming to achieve a variety of seeking optimal solution algorithm for the optimization process of visualization. Through the example analysis and application, the multi-objective optimization system is developed in this paper can gain better optimization results, it improves the previous optimal algorithm can only attain the Pareto solution and unable to obtain the optimal solution, improved calculating speed is slow and lack of accuracy is low, for the "black-box" of the optimization problem provides an effective solution.