DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)

Control valve stiction can be considered as one the primary causes that can affect a control system performance. Since control valve act as the final part of control element, it can cause disturbances towards an operation. In 1989, an initiative has been created where a valve stiction detection meth...

Full description

Main Author: ROSLAN, MOHAMED YASSIN
Format: Final Year Project
Language: English
Institution: Universiti Teknologi Petronas
Record Id / ISBN-0: utp-utpedia.18047 /
Published: IRC 2017
Subjects:
TP Chemical technology
Online Access: http://utpedia.utp.edu.my/18047/1/Mohamed%20Yassin%20bin%20Roslan_17123_Dissertation%20Report.pdf
http://utpedia.utp.edu.my/18047/
Tags: Add Tag
No Tags, Be the first to tag this record!
id utp-utpedia.18047
recordtype eprints
spelling utp-utpedia.180472018-08-01T09:32:16Z http://utpedia.utp.edu.my/18047/ DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN) ROSLAN, MOHAMED YASSIN TP Chemical technology Control valve stiction can be considered as one the primary causes that can affect a control system performance. Since control valve act as the final part of control element, it can cause disturbances towards an operation. In 1989, an initiative has been created where a valve stiction detection method is technologically advanced to detect stiction in a control valve. Afterwards, numerous methods are made and redeveloped to improvise the technique of detection the stiction in a control valve. However, none of the methods so far are using the neural network methods. Therefore, this project will cover on several neural network methods on valve stiction detection and will be tested for the effectiveness in detecting the fault. This project will be conducted by using MATLAB© Simulink to compute the simulation model. The initial objective of the design is to determine the number of neurons and the type of transfer function used in the hidden layer for the feedforward neural network model in this project. The outcome of the feed forward model use 11 units in the hidden layer, as the result produced is favorable by using 11 neurons compared to different number of neurons. As for the transfer function, the feedforward model uses tansig as the transfer function in the hidden layer and purelin as the transfer function in the output layer. As for the neural network strategy, three approaches are tested and analyzed. First approach and third approach produced unfavorable result. However, second approach produced effective and favorable result but still unpractical to be applied due to high amount of error generated. IRC 2017-01 Final Year Project NonPeerReviewed application/pdf en http://utpedia.utp.edu.my/18047/1/Mohamed%20Yassin%20bin%20Roslan_17123_Dissertation%20Report.pdf ROSLAN, MOHAMED YASSIN (2017) DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN). IRC, Universiti Teknologi PETRONAS.
institution Universiti Teknologi Petronas
collection UTPedia
language English
topic TP Chemical technology
spellingShingle TP Chemical technology
ROSLAN, MOHAMED YASSIN
DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
description Control valve stiction can be considered as one the primary causes that can affect a control system performance. Since control valve act as the final part of control element, it can cause disturbances towards an operation. In 1989, an initiative has been created where a valve stiction detection method is technologically advanced to detect stiction in a control valve. Afterwards, numerous methods are made and redeveloped to improvise the technique of detection the stiction in a control valve. However, none of the methods so far are using the neural network methods. Therefore, this project will cover on several neural network methods on valve stiction detection and will be tested for the effectiveness in detecting the fault. This project will be conducted by using MATLAB© Simulink to compute the simulation model. The initial objective of the design is to determine the number of neurons and the type of transfer function used in the hidden layer for the feedforward neural network model in this project. The outcome of the feed forward model use 11 units in the hidden layer, as the result produced is favorable by using 11 neurons compared to different number of neurons. As for the transfer function, the feedforward model uses tansig as the transfer function in the hidden layer and purelin as the transfer function in the output layer. As for the neural network strategy, three approaches are tested and analyzed. First approach and third approach produced unfavorable result. However, second approach produced effective and favorable result but still unpractical to be applied due to high amount of error generated.
format Final Year Project
author ROSLAN, MOHAMED YASSIN
author_sort ROSLAN, MOHAMED YASSIN
title DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
title_short DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
title_full DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
title_fullStr DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
title_full_unstemmed DEVELOPMENT OF CONTROL VALVE STICTION DETECTION METHOD USING NEURAL NETWORK (NN)
title_sort development of control valve stiction detection method using neural network (nn)
publisher IRC
publishDate 2017
url http://utpedia.utp.edu.my/18047/1/Mohamed%20Yassin%20bin%20Roslan_17123_Dissertation%20Report.pdf
http://utpedia.utp.edu.my/18047/
_version_ 1741195286659727360
score 11.62408

Similar Items