Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system

Nowadays, modern process plants are following the trend of highly integrated and complex processes and highly instrumented chemical processes. In the most chemical processes, the need of monitoring various process variables is driven by the large amount of data produced by the instruments. Eventuall...

Full description

Bibliographic Details
Main Author:	Mohd Farid, Mohd Na’aim
Format:	Undergraduates Project Papers
Language:	English
Published:	2014
Subjects:	QA Mathematics
Online Access:	http://umpir.ump.edu.my/id/eprint/8878/ http://umpir.ump.edu.my/id/eprint/8878/ http://umpir.ump.edu.my/id/eprint/8878/1/CD8682%20%40%2055.pdf

id	ump-8878
recordtype	eprints
spelling	ump-88782015-10-26T06:24:59Z http://umpir.ump.edu.my/id/eprint/8878/ Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system Mohd Farid, Mohd Na’aim QA Mathematics Nowadays, modern process plants are following the trend of highly integrated and complex processes and highly instrumented chemical processes. In the most chemical processes, the need of monitoring various process variables is driven by the large amount of data produced by the instruments. Eventually, data will overload and wasted. Therefore, something needs to be done to monitor process data in lesser dimension as well as retain the most variations present. Principal Component Analysis (PCA) based MSPM technique is introduced to help us monitor and control processes. However, we have to retain too much principal component (PC) scores which complicate the fault detection operation. Thus, new MSPM technique, Multiple Linear Regression (MLR) is introduced to be utilized together with PCA to monitor predictor variables from criterion variables by equation that relates both variables. Hence, we only need to monitor criterion variables. The hypothesis for this study is if MLR is implemented, the lesser the number of variables need to be monitored. This proposed method is applied to the on-line monitoring of a simulated continuous stirred tank reactor with recycle (CSTRwR) from case study of Zhang, Martin, & Morris(1995). MATLAB software is utilized in this study. The general framework of fault detection comprises Phase I and Phase II. Phase I starts from normalisation of NOC data, PC scores formulated, monitoring statistics SPE and T2 are calculated and lastly 95% and 99% control limits developed. Phase II starts from standardisation of fault data with respect to NOC data, PC scores developed, monitoring statistics SPE and T2 are developed and lastly fault detection using control limits developed in Phase I. System A is the CSTRwR monitored by PCA-based MSPM system for the original set of variables. System B is the CSTRwR monitored by new MLR-PCA based MSPM system for the criterion variables which are the main product variables. Dynamic model was developed in Phase I. Then, fault data was introduced in the Phase II for fault detection. For System A, using abrupt fault data No.1 (F01a), faults were detected successfully by monitoring five variables out of 13 variables meanwhile System B only monitor two variables out of three variables with almost identical outcomes. Hence, new MSPM technique, MLR was successfully proven to be an efficient monitoring tool with quick detection and isolability while retaining as much as possible variations in lesser dimension 2014-07 Undergraduates Project Papers NonPeerReviewed application/pdf en http://umpir.ump.edu.my/id/eprint/8878/1/CD8682%20%40%2055.pdf Mohd Farid, Mohd Na’aim (2014) Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system. Faculty of Chemical & Natural Resources Engineering, Universiti Malaysia Pahang. http://iportal.ump.edu.my/lib/item?id=chamo:84802&theme=UMP2
repository_type	Digital Repository
institution_category	Local University
institution	Universiti Malaysia Pahang
building	UMP Institutional Repository
collection	Online Access
language	English
topic	QA Mathematics
spellingShingle	QA Mathematics Mohd Farid, Mohd Na’aim Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
description	Nowadays, modern process plants are following the trend of highly integrated and complex processes and highly instrumented chemical processes. In the most chemical processes, the need of monitoring various process variables is driven by the large amount of data produced by the instruments. Eventually, data will overload and wasted. Therefore, something needs to be done to monitor process data in lesser dimension as well as retain the most variations present. Principal Component Analysis (PCA) based MSPM technique is introduced to help us monitor and control processes. However, we have to retain too much principal component (PC) scores which complicate the fault detection operation. Thus, new MSPM technique, Multiple Linear Regression (MLR) is introduced to be utilized together with PCA to monitor predictor variables from criterion variables by equation that relates both variables. Hence, we only need to monitor criterion variables. The hypothesis for this study is if MLR is implemented, the lesser the number of variables need to be monitored. This proposed method is applied to the on-line monitoring of a simulated continuous stirred tank reactor with recycle (CSTRwR) from case study of Zhang, Martin, & Morris(1995). MATLAB software is utilized in this study. The general framework of fault detection comprises Phase I and Phase II. Phase I starts from normalisation of NOC data, PC scores formulated, monitoring statistics SPE and T2 are calculated and lastly 95% and 99% control limits developed. Phase II starts from standardisation of fault data with respect to NOC data, PC scores developed, monitoring statistics SPE and T2 are developed and lastly fault detection using control limits developed in Phase I. System A is the CSTRwR monitored by PCA-based MSPM system for the original set of variables. System B is the CSTRwR monitored by new MLR-PCA based MSPM system for the criterion variables which are the main product variables. Dynamic model was developed in Phase I. Then, fault data was introduced in the Phase II for fault detection. For System A, using abrupt fault data No.1 (F01a), faults were detected successfully by monitoring five variables out of 13 variables meanwhile System B only monitor two variables out of three variables with almost identical outcomes. Hence, new MSPM technique, MLR was successfully proven to be an efficient monitoring tool with quick detection and isolability while retaining as much as possible variations in lesser dimension
format	Undergraduates Project Papers
author	Mohd Farid, Mohd Na’aim
author_facet	Mohd Farid, Mohd Na’aim
author_sort	Mohd Farid, Mohd Na’aim
title	Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
title_short	Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
title_full	Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
title_fullStr	Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
title_full_unstemmed	Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
title_sort	utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system
publishDate	2014
url	http://umpir.ump.edu.my/id/eprint/8878/ http://umpir.ump.edu.my/id/eprint/8878/ http://umpir.ump.edu.my/id/eprint/8878/1/CD8682%20%40%2055.pdf
first_indexed	2023-09-18T22:06:56Z
last_indexed	2023-09-18T22:06:56Z
_version_	1777414770316541952

Utilising multiple linear regression (mlr) technique in multivariate statistical process monitoring (mspm) system

Similar Items