Research Of Product Design Based On Process Data

In industry process, producers often obtain a variety of different grades of product by changing the operation conditions or the ratio of ingredients in the feed, so that different consumption needs are met. It’s a common issue that how we can find or adjust the corresponding operation conditions based on the customers’desired quality specifications. The theory and method on product design is an effective tool to solve this problem, especially for those products which are manufactured at the expense of the high cost of raw materials, long production cycle. If an appropriate method is found out to confirm the operation conditions being able to produce a product having a specified set of quality characteristics, it will in general take less time and production costs, improve productivity to some degree. So the study of product design is so much meaningful both in practice and in theory.Product design is involved with modeling. In this paper, several commonly used modeling methods are introduced firstly. Among these methods, regression analysis based on process data has advantages of quick speed of modeling, transparent internal model structure, no need of mastering much more process knowledge and so forth, leading to an extensive attention and research. In the light of serious problem that regression parameter matrix becomes invalid because of co-linear operation variables in standard regression model, a method based on latent variable model is applied to product design. Here, PCR and PLS are mainly studied. Afterwards, the approaches mentioned above are used to solve product design of low density polyethylene. The simulation results show that product design based on latent variable model is effective. In view of the insufficiency of process information in product design, a strategy of information migration on the basis of similar processes is proposed. Specifically, two methods, extended PCR (EPCR) and joint-Y PLS (JYPLS) are discussed deeply and proved to be practical and effective through the simulation.Time-varying is unavoidable. When comes to the issue of product design with time-varying characteristics, a recursive PCR, updating the process model by including the newest samples is proposed. As the data size for modeling increases, recursive PCR may be difficult to implement in practice because it leads to a reduction in the speed of model adaptation. As a result, a recursion strategy with a window sliding along the data sets, i.e. including the newest samples and excluding the oldest ones, is more appropriate for time-varying processes. In the end, a simulation experiment is carried out and the results have shown the effectiveness of the proposed thoughts and approaches.