dc.contributor.advisor | Wang, Kewen | |
dc.contributor.advisor | Liew, Wee-Chung | |
dc.contributor.author | Hao, Lin | |
dc.date.accessioned | 2019-08-07T01:43:27Z | |
dc.date.available | 2019-08-07T01:43:27Z | |
dc.date.issued | 2019-07-15 | |
dc.identifier.doi | 10.25904/1912/1787 | |
dc.identifier.uri | http://hdl.handle.net/10072/386541 | |
dc.description.abstract | The introduction of water quality monitoring systems (WQMSs) into the city wastewater
discharging and recycling system has been considered as an effective practice to address
the potential risks to wastewater plants as well as public concerns associated with
pollution source localization. Under such background, a WQMS was developed and
deployed in a large Australia Research Council (ARC) Linkage project entitled “A New
Management System for Effective Wastewater Source Control” to meet the requirements
of several major Australian water utilities.
A WQMS was developed for this ARC Linkage project, which aims to monitor the
wastewater quality in city wastewater pipelines to safeguard purified recycled water
(PRW) system in water treatment companies. The developed WQMS contains both welldesigned
hardware and reliable software and implements primary functions including
data collection, data processing, event detection, alarming, a central server and web-based
user interface (UI).
To explore the potentials of existing event detection methods using in wastewater
application environment, six widely used event detection methods in drinking and natural
water area are evaluated, including Cumulative Sum (CUSUM), Linear Prediction,
Moving Median, Exponentially Weighted Moving Average (EWMA), Autoregressive
Integrated Moving Average (ARIMA) and Artificial Neural Networks (ANNs). The
evaluation is performed on real datasets collected from 10 different pump stations with
WQMS deployed.
Referring to the results of the performance evaluation on some existing event detection
methods, we designed and developed two methods according to the requirements of the
practical application. The first modified Bayesian analysis method achieves good
performance in testing, and it has been applied in deployed WQMSs and trialed for over
four years without any significant issues. The second one is an ANN-based method with
multiple classifiers integrated. It achieves excellent overall performance among ten
testing sites although the high computational cost limits its practical value. | |
dc.language | English | |
dc.language.iso | en | |
dc.publisher | Griffith University | |
dc.publisher.place | Brisbane | |
dc.subject.keywords | Water quality monitoring systems | |
dc.subject.keywords | Wastewater | |
dc.subject.keywords | Event detection methods | |
dc.subject.keywords | Water recycling systems | |
dc.title | Abnormal Event Detection Platform Design for a Wastewater Quality Monitoring System | |
dc.type | Griffith thesis | |
gro.faculty | Science, Environment, Engineering and Technology | |
gro.rights.copyright | The author owns the copyright in this thesis, unless stated otherwise. | |
gro.hasfulltext | Full Text | |
dc.contributor.otheradvisor | Zhao, Huijun | |
gro.thesis.degreelevel | Thesis (PhD Doctorate) | |
gro.thesis.degreeprogram | Doctor of Philosophy (PhD) | |
gro.department | School of Info & Comm Tech | |
gro.griffith.author | Hao, Lin | |