Process Monitoring Method Study Based On Information Transfer Technology

Process monitoring is one of important means in industrial production to ensure operate stably and securely. Process monitoring based on data-driven methods need much information to monitor process variables; however, the process monitoring model will emit wrong alarms with insufficient information and influence manufacture. This thesis studies information transfer technologies to establish the process monitoring model to solve the problem of insufficient process information. The main contents of this thesis are as follows:Firstly, the principle of PCA and PLS and the ways to establish process monitoring model are elaborated. The methods of EPCR and JYPLS process information transfer technologies on the basis of PCA and PLS are studied.Secondly, the method of establishing process monitoring with EPCA is researched. With regard to change of process conditions and the problem about inaccurate monitoring model, the thesis expounds the method of recursive principal components analysis (RPCA) and the way to update monitoring model with local model based just-in-time learning.Thirdly, the method of establishing process monitoring with JYPLS is researched. The way to update monitoring model with local JYPLS model based just-in-time learning is researched to solve the problem of process change, and at the same time some methods of seleceting similarity data.Fourthly, the process monitoring model with EPCA and JYPLS in the grinding process is established respectively, and do a series of simulations, which get the conclusion that the methods are very effective and practical.