Application Research Of Multivariable Statistical Process Control In Industrial Production

This dissertation investigates the application of MSPC and its shortage, gives settlement and tries to explain the means of principle component, making MSPC more practical. The main research work includes as following:1. We tell the relation between quality management and SPC, summarize the application and development of SPC and sum up its characteristics, introduce the basic concept, main types, techniques and implementation procedure of SPC.2. Several mathematical tools used in MSPC are introduced, including Principal Component Analysis (PCA) and Multi-way Principal Component Analysis (MPCA), Principal Component Recursion (PCR), Partial Least Square (PLS). We introduce the methods of PCA and MPCA in detail, point out the essential of PCA, give out several algorithm of PCA and the application field. The meaning and means to interpret the practice means of PC are brought forward. Basing on the theories above, PCR and PLS are presented. For continuous process, the normal PCA can be used to analyze the process data. For batch process, MPCA is the appropriate technique. We put forward a new technique, named Glide Windows Multi-way SPC, based on MPCA.3. Single-variable SPC and its disadvantages are pointed out,then some statistical variables and techniques of MSPC are emphasized. Multi-dimension data processing is transformed to the analysis of a single statistical variable by defining different statistical variables.4. The application of MSPC in continuous process and batch process is introduced, basing on the processes of synthesizing of Phosphonic Acid Dimethyl and batch cooling crystallizer. For the process of synthesizing of Phosphonic Acid Dimethyl, the significance of application of MSPC on it is discussed, the parameters and original data are deduced and estimated, and process monitoring and diagnostic simulation are given based on simplified model. For the process of batch cooling crystallizer, Glide Windows MSPC is applied based on the original data from simulation programs, and it works well.5. Lastly, some future research areas are highlighted.