Second, the general steps of DOE
Experimental design DOE(Design of Experiments) experimental design aims at the improvement of product flow, product research and development, and product production process, analyzes important factors, optimizes results, and improves design robustness. By selecting the relevant parameters that have great influence on the product characteristics, we can determine the indicators in the factors and choose the indicators that are more in line with the product characteristics. Simply put, we can examine various parameters that affect product characteristics and determine which parameters are combined and which product characteristics are the best. DOE experimental design has been widely used in all aspects of improving product quality and reducing production cost. Experimental design can also be used in enterprise management, product structure adjustment, production efficiency and higher production plan.
The process of human understanding of nature is a gradual process. Generally, we first put forward an idea based on the existing knowledge, design an experiment to verify or deny it, and the data obtained from the experiment help people to verify or correct the original idea, then put forward a newer and deeper idea, and then design a new experiment. This iterative process will continue until a relatively complete theoretical result is formed. Conducting experiments is also a learning process, which cannot be accomplished overnight. Generally speaking, the test should be carried out in several batches, and the following steps are generally adopted:
1, the screening factors were designed by partial factors.
At first, the situation was not very clear. Considering that there may be many factors (greater than or equal to 5) affecting the response variables, the factors should be screened out in a larger test range, and the partially realized factor test design method is usually adopted. The result obtained in this way may be rough, but it can greatly reduce the number of tests and achieve the purpose of screening. If considering some factors, the cost of the test is still too expensive, we can use the "Plaekett-Burman design" method with fewer tests to screen the factors.
2. Comprehensive analysis of factor effect and interaction is made by the method of total factor experimental design.
When the number of screened factors is less than or equal to 5, we can further carry out all-factor experimental design in a slightly smaller range to obtain accurate information on the effects and interactions of each factor, and further screen the factors until the number of factors does not exceed 3.
3. Determine the regression relationship and find the optimal setting by response surface method.
When the number of factors is less than three, it is conditional to adopt more detailed response surface design analysis method to fit a quadratic equation to the response variable in a small area containing the best point, so as to get the best point in the experimental area.
The above steps are typical. In practical work, you may skip a link or repeat a step several times. In short, we should constantly screen factors, constantly adjust the scope of testing and the selection of factor levels, and finally achieve the overall goal of testing after several rounds of testing.