- Jayasooriya, Mahinda, Dahlhaus, Peter, Barton, Andrew, Gell, Peter
- Authors: Jayasooriya, Mahinda , Dahlhaus, Peter , Barton, Andrew , Gell, Peter
- Date: 2015
- Type: Text , Conference proceedings
- Relation: 36th Hydrology and Water Resources Symposium: The Art and Science of Water, HWRS 2015; Hobart, Tasmania; 7th-10th December 2016. p. 436-442
- Full Text: false
- Reviewed:
- Description: Inflow and infiltration into separate sewer systems is an ongoing challenge experienced by water utilities in managing sewer networks across the world. An accurate estimation of groundwater infiltration in terms of volume and flow rate is important for making decisions on sewer rehabilitation and for the effective operation of sewer networks. The fast response of surface inflow to sewers occurs during or immediately after a prolonged or intense precipitation event and can often be exacerbated by illegal stormwater connections into the sewer network. The slow response of inflow to sewers can be attributed to deep infiltration or the discharge of groundwater into the sewer network. A common practice for most Australian water utilities in combatting the problem of infiltration and inflow is to undertake short to medium term sewer network flow monitoring, while collecting contemporaneous rainfall data, to assess the various volumes and their origin in their sewer networks. This paper presents a review of the current data collection practices, using the City of Ballarat in south eastern Australia as a case study. Discussion is provided around gaps in data collection practices to properly understand the problem and recommendations are made on what additional monitoring works should be performed so that infiltration, in particular, can be assessed on a sound scientific basis. © 2015, Engineers Australia. All rights reserved.
- Description: The Art and Science of Water - 36th Hydrology and Water Resources Symposium, HWRS 2015
Accurate prediction of sewer network flows and assessment of groundwater infiltration volumes in sewer networks
- Authors: Jayasooriya, Mahinda
- Date: 2022
- Type: Text , Thesis , PhD
- Full Text:
- Description: Excessive flows due to inflow and infiltration (I&I) in sewer networks contribute to hydraulically overloaded sanitary sewer networks. Understanding the cause–effect relationships associated with sewer network flow generation and accurate quantification of excessive flow volumes entering sewer networks is crucial to developing solutions to resolve this issue. A systematic literature review found a gap in understanding the cause–effect relationships of I&I and no universally accepted method for I&I estimation. This research used three years of data from an Australian sewer catchment in Ballarat, Victoria, as a case study to enhance understanding of the network flow generation process. The aims of this research were to accurately predict sewer network flows, reliably separate the groundwater infiltration (GWI) volume of sewer network flow, improve understanding of associated processes and identify the influential parameters that impact sewer network flow. This research used three methods to achieve these aims. First, commercially available software was used to develop and calibrate a sewer network model to predict flows with data from the case study catchment. The best practice guidelines available in Australia were used to calibrate the model, and a new method was introduced to separate the sewer network flow components. Second, two classical hydrology applications, the recursive filter and the flow duration curve (FDC) methods, were used to separate the GWI component of the total sewer network flow and understand the performance of the sewer catchments. Third, an artificial neural network (ANN) model was developed to predict sewer network flow. Parameters known to influence sewer flows were incrementally introduced to the ANN model to improve performance and understand their relative importance. This study finds that commercially available software can predict total sewer network flows with reasonable accuracy using the Wallingford or triangular hydrograph (RTK) hydrological implementation methods. However, the software cannot replicate the dynamic nature of the complex I&I processes that occur over time or reliably separate the sewer network flow components. A key finding from this study is that commercially available software has clear limitations in separating sewer network flow components. Recursive filter and FDC methods can be successfully used as standalone techniques to calculate the GWI volume. These methods have further value in comparing the I&I between catchments to determine flow characteristics and are straightforward for practising engineers to use. The ANN model predicts sewer flow with a high level of accuracy, and soil moisture is the most critical predictor of the total flow regime. The ANN model is also able to predict the dynamic and complex nature of I&I into sewer networks. More research is needed to achieve effective and reliable separation of individual components in an overall sewer flow regime. ANN and other emerging machine learning techniques show much potential to effectively predict sewer flows across a range of sewer network and catchment characteristics.
- Description: Doctor of Philosophy
- Authors: Jayasooriya, Mahinda
- Date: 2022
- Type: Text , Thesis , PhD
- Full Text:
- Description: Excessive flows due to inflow and infiltration (I&I) in sewer networks contribute to hydraulically overloaded sanitary sewer networks. Understanding the cause–effect relationships associated with sewer network flow generation and accurate quantification of excessive flow volumes entering sewer networks is crucial to developing solutions to resolve this issue. A systematic literature review found a gap in understanding the cause–effect relationships of I&I and no universally accepted method for I&I estimation. This research used three years of data from an Australian sewer catchment in Ballarat, Victoria, as a case study to enhance understanding of the network flow generation process. The aims of this research were to accurately predict sewer network flows, reliably separate the groundwater infiltration (GWI) volume of sewer network flow, improve understanding of associated processes and identify the influential parameters that impact sewer network flow. This research used three methods to achieve these aims. First, commercially available software was used to develop and calibrate a sewer network model to predict flows with data from the case study catchment. The best practice guidelines available in Australia were used to calibrate the model, and a new method was introduced to separate the sewer network flow components. Second, two classical hydrology applications, the recursive filter and the flow duration curve (FDC) methods, were used to separate the GWI component of the total sewer network flow and understand the performance of the sewer catchments. Third, an artificial neural network (ANN) model was developed to predict sewer network flow. Parameters known to influence sewer flows were incrementally introduced to the ANN model to improve performance and understand their relative importance. This study finds that commercially available software can predict total sewer network flows with reasonable accuracy using the Wallingford or triangular hydrograph (RTK) hydrological implementation methods. However, the software cannot replicate the dynamic nature of the complex I&I processes that occur over time or reliably separate the sewer network flow components. A key finding from this study is that commercially available software has clear limitations in separating sewer network flow components. Recursive filter and FDC methods can be successfully used as standalone techniques to calculate the GWI volume. These methods have further value in comparing the I&I between catchments to determine flow characteristics and are straightforward for practising engineers to use. The ANN model predicts sewer flow with a high level of accuracy, and soil moisture is the most critical predictor of the total flow regime. The ANN model is also able to predict the dynamic and complex nature of I&I into sewer networks. More research is needed to achieve effective and reliable separation of individual components in an overall sewer flow regime. ANN and other emerging machine learning techniques show much potential to effectively predict sewer flows across a range of sewer network and catchment characteristics.
- Description: Doctor of Philosophy
- «
- ‹
- 1
- ›
- »