Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.576960
Title: Multivariate statistical modelling for fault analysis and quality prediction in batch processes
Author: Hong, Jeong Jin
Awarding Body: University of Newcastle Upon Tyne
Current Institution: University of Newcastle upon Tyne
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Abstract:
Multivariate statistical process control (MSPC) has emerged as an effective technique for monitoring processes with a large number of correlated process variables. MSPC techniques use principal component analysis (PCA) and partial least squares (PLS) to project the high dimensional correlated process variables onto a low dimensional principal component or latent variable space and process monitoring is carried out in this low dimensional space. This study is focused on developing enhanced MSPC techniques for fault diagnosis and quality prediction in batch processes. A progressive modelling method is developed in this study to facilitate fault analysis and fault localisation. A PCA model is developed from normal process operation data and is used for on-line process monitoring. Once a fault is detected by the PCA model, process variables that are related to the fault are identified using contribution analysis. The time information on when abnormalities occurred in these variables is identified using time series plot of the squared prediction errors (SPE) on these variables. These variables are then removed and another PCA model is developed using the remaining variables. If the faulty batch cannot be detected by the new PCA model, then the remaining variables are not related to the fault. If the faulty batch can still be detected by the new PCA model, then further variables associated with the fault are identified from SPE contribution analysis. The procedure is repeated until the faulty batch can no longer be detected using the remaining variables. Multi-block methods are then applied with the progressive modelling scheme to enhance fault analysis and localisation efficiency. The methods are tested on a benchmark simulated penicillin production process and real industrial data. An enhanced multi-block PLS predictive modelling method is developed in this study. It is based on the hypothesis that meaningful variable selection can lead to better prediction performance. A data partitioning method for enhanced predictive process modelling is proposed and it enables data to be separated into blocks by different measuring time. Model parameters can be used to express contributions
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.576960  DOI: Not available
Share: