Study on derivating qualitative sensitivities from experimental results and qualitative optimization

In the practical design process, we make up design models which can express the behaviors of the real design objects and we innovate the designs based on those models. However, making up accurate design models adequately sometimes becomes a hard burden to the designer and consequently such modeling process may consume the time for designers to have creative designs. In such circumstances, there exist some approaches which are making design models based on experimental data. Advantages of these approaches can reduce the burden of the designers. However, existed those approaches are not so robust to the errors made in experiments, and also they need some kinds of approximation method with some kinds of errors. To overcome these disadvantages, we need some methods which are robust to the experimental and approximation errors. We have been studying the optimum oriented method based on qualitative sensitivities. In this study, we will propose a method to derive qualitative sensitivities only by using model experimental data. And using those sensitivity, we will carry out qualitative optimization. We applied the proposed system to the design of shaft of the rotor as one of examples and show the effectiveness of the method.