Application of probabilistic neural networks in modelling structural deterioration of stormwater pipes

Tran, H, NG, A, Perera, B, Burn, S and Davis, P 2006, 'Application of probabilistic neural networks in modelling structural deterioration of stormwater pipes', Urban Water Journal, vol. 3, no. 3, pp. 175-184.


Document type: Journal Article
Collection: Journal Articles

Title Application of probabilistic neural networks in modelling structural deterioration of stormwater pipes
Author(s) Tran, H
NG, A
Perera, B
Burn, S
Davis, P
Year 2006
Journal name Urban Water Journal
Volume number 3
Issue number 3
Start page 175
End page 184
Total pages 10
Publisher Taylor and Francis Ltd
Abstract In Australia, when stormwater systems were first introduced over 100 years ago, they were constructed independently of the sewer systems, and they are normally the responsibility of the third level of government, i.e., local government or city councils. Because of the increasing age of these stormwater systems and their worsening performance, there are serious concerns in a significant number of city councils regarding their deterioration. A study has been conducted on the structural deterioration of concrete pipes that make up the bulk of the stormwater pipe systems in these councils. In an attempt to look for a reliable deterioration model, a probabilistic neural network (PNN) model was developed using the data set supplied from participating councils. The PNN model was validated with snapshot-based sample data, which makes up the data set. The predictive performance of the PNN model was compared with a traditional parametric model using discriminant analysis on the same data set. Structural deterioration was hypothesised to be influenced by a set of explanatory factors, including pipe design and construction factors—such as pipe size, buried depth—and site factors— such as soil type, moisture index, tree root intrusion, etc. The results show that the PNN model has a better predictive power and uses significantly more input variables (i.e., explanatory factors) than the discriminant model. More importantly, the key factors for prediction in the PNN model are difficult to interpret, suggesting that besides prediction accuracy, model interpretation is an important issue for further investigation.
Subject Infrastructure Engineering and Asset Management
Keyword(s) Deterioration model
Probabilistic neural networks
Stormwater pipes
Discriminant analysis
DOI - identifier 10.1080/15730620600961684
Copyright notice © 2006 Taylor & Francis
ISSN 1573-062X
Versions
Version Filter Type
Citation counts: Scopus Citation Count Cited 20 times in Scopus Article | Citations
Altmetric details:
Access Statistics: 176 Abstract Views  -  Detailed Statistics
Created: Tue, 19 Feb 2013, 13:10:00 EST by Catalyst Administrator
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us