Correlated Multi-response Experimental Design And Robust Optimization Using RSM

In the actual production process, the product often has multiple quality characteristics, and has correlation between characteristics. This dissertation mainly studies the correlated multi-response robust optimization problems, and mainly focuses on acquiring robust optimal solutions when considering tolerance factor. To achieve this goal, a model for influence factors and response variables using response surface method is build, and Mahalanobis distance function method is used to consider correlation between the responses. When tolerance factor is existing, how to obtain the robust optimal solution is worth studying. The topics discussed in this paper are outlined as following.Firstly, this dissertation introduces the experimental design and response surface method, and summarized the concepts and methods of the multi-response robust optimization. The Mahalanobis distance function which considered the variance-covariance matrix is selected to transform multi-response optimization problem into the minimization of overall distance function.Secondly, we developed hybrid intelligent algorithm which combining genetic algorithm and the pattern search to find the minimum of overall distance function. The genetic algorithm is used to global searching in the feasible region and then the pattern search algorithm is used to find accurate local optimization solutions. Compared with a single pattern search algorithm, hybrid intelligent algorithm can better deal with high complexity function optimization problem, and can improve the accuracy of the optimal solution than single intelligent algorithm.Finally, analyze the influence of tolerance factors to the Mahalanobis distance function, improved the Mahalanobis distance function which has effected by tolerance factors, and use hybrid intelligent algorithm which combined genetic algorithm and pattern search to search the robust optimal solution. Results show that this method can obtain the robust optimal solution which fall in robust feasible region, and this solution is not sensitive to the fluctuation of tolerance factors.