Fault Diagnosis & Field Measurement Prediction Techniques for a Gas Metering System

Nizamuddin, Siti Asfarina (2014) Fault Diagnosis & Field Measurement Prediction Techniques for a Gas Metering System. [Final Year Project] (Unpublished)

[thumbnail of 13808_Siti Asfarina_Final Report.pdf]
13808_Siti Asfarina_Final Report.pdf

Download (1MB) | Preview


This report discusses on research regarding fault diagnosis system for a process plant. In this project, the process studied is Petronas gas metering system to Kapar Power Plant. There are two parts to this project. The first part is focused on proposing a backup fault diagnosis method for this gas metering system. The second part of the project is to propose suitable field measurement prediction techniques, which could be used in the event of a fault or intermediate condition.
In order to achieve the first objective, this report first discusses the potential fault diagnosis methods which can be applied to the metering system. The advantages and disadvantages of each method were evaluated. From evaluation, it was chosen to propose fault diagnosis system using Adaptive Neuro Fuzzy Inference System (ANFIS). In order to carry out fault diagnosis, data is first filtered into fault data and healthy data. The faults filtered in this report include transmitter fault and hang fault for parameters of Temperature, Pressure and Gross Volume. Once healthy data was identified, it was further classified into normal and intermediate categories. This process was done through three different methods, which are the hyperbox model, linear model and ANFIS model. Once these models were analysed, the writer has chosen to proceed with ANFIS model for data classification. Classified data was then grouped into clusters.
The second part of the project is focused on proposing suitable field measurement prediction technique using ANFIS that can be used in the event of fault or intermediate conditions. Six different ANFIS models were developed to estimate parameters Temperature, Pressure and Gross Volume during transmitter and hang fault. Five variables such as ANFIS input, data division, number of epoch for training, type of membership function and randomisation of data were varied in order to develop the best model. ANFIS prediction model for Temperature produced satisfactory results of less than 1% error. ANFIS prediction model for Pressure and Gross Volume on the other hand need to be further developed to meet industrial requirements.

Item Type: Final Year Project
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments / MOR / COE: Engineering > Electrical and Electronic
Depositing User: Users 2053 not found.
Date Deposited: 01 Oct 2014 16:50
Last Modified: 25 Jan 2017 09:37
URI: http://utpedia.utp.edu.my/id/eprint/14198

Actions (login required)

View Item
View Item