The Research Of Uniform Experimental Design On Evaluation And Optimization Analysis Method

Objective Study the idea, methods, principles, traits of Uniform Experimental Design. Use uniform design table correctly to be tested. To explore optimal experimental conditions in multi-factors and multi-levels experiment using orthogonal experimental design and Uniform Experimental Design. Compared with orthogonal experimental design and evaluated in multi-factors and multi-levels. It obtained the expected results by using uniform design with few times of tests. However, many practical investigators have raised a query on the reliability of the results. We must test and verify its feasibility, reliability with simulation examples in this study. It obtained regression analysis model by uniform design software. By using this model carry on the significance analysis of factors and the estimation of the results, the forecasting and the optimization, explore methods of optimization analysis and choosing optimal experiment condition.Methods Choosing uniform design table correctly and arranging the head of the table rationally to fit regression model, the model carries out multiple linearity regression analysis and quadratic regression analysis. Make use of the random error to obtain the experiment data. Carry on the inspection to various factors, screen the optimal experimental conditions by using uniform design and orthogonal experimental design. Compared two methods of the comparability and the rationality of optimize results.Result In the evaluation of uniform design optimization, use different designs in the same experiment, and compute simulated data by using SAS. With orthogonal design analysis, the optimal experimental condition, which arises from the variance analysis and calculates, is in line with using regression analysis. With uniform design analysis, the optimal experimental condition is basically consistent between screening variable by the stepwise regression and using orthogonal design analysis. In the sample analysis, choose appropriate method to explore optimal conditions according to establishing regression model. If the main effect of factors was not significant while the interactions were significant, we may use contour chart to seek the optimal experimental conditions. The study used genetic algorithm, the latest global search method to determine the optimal experimental conditions. Conclusion Treatment at many factors, the level of many experiments, the uniform design is a very useful tool. When the commonly used design still did not have the feasibility, uses the homogeneous design. It may use few times test to reveal the influence and regularity of each experimental factor, getting more information, showing the greater advantages of saving a lot of time and funding, but also easy to put into effect and the results are reliable. As long as the strict control of experimental error, selecting the appropriate uniform design table, the results of uniform design are reliable.