Predicting Structural Deterioration Condition of Individual Storm-Water Pipes Using Probabilistic Neural Networks and Multiple Logistic Regression Models

After several decades in service, the deterioration of storm-water pipe assets is inevitable. The deterioration of storm-water pipes is characterized by structural deterioration and hydraulic deterioration. Condition assessment using closed circuit television (CCTV) inspection is often carried out to assess the structural condition of pipes. However, the knowledge on the condition of storm-water pipe assets is still limited for strategic planning of maintenance and rehabilitation, because generally only a small sample is CCTV-inspected and in almost all cases, these pipes are inspected once only due to high costs. The challenge for researchers is to use the sample of CCTV-inspected pipes for developing mathematical models that can predict the structural condition of remaining pipes as well as the future condition of pipes. In this present study, the deterioration pattern of storm-water pipes is constructed on the basis that each pipe has its own deterioration rate due to its pipe factors. Based on this, two mathematical models using multiple logistic regression (MLR) and probabilistic neural networks (PNN) are developed for predicting the structural condition of individual pipes. The MLR model was calibrated using the maximum likelihood method and the PNN model was trained using a genetic algorithm (GA). The predictive performances of both models were compared using CCTV data collected for a local government authority in Melbourne, Australia. The results showed that the PNN model was more suited for modeling the structural deterioration of individual storm-water pipes than the MLR model. Furthermore, the use of GA improved the training results of the PNN model compared to the trial and error method.

[1]  Samuel T. Ariaratnam,et al.  Innovative method for assessment of underground sewer pipe condition , 2006 .

[2]  Jim Albert,et al.  Ordinal Data Modeling , 2000 .

[3]  Maha N. Hajmeer,et al.  Comparison of logistic regression and neural network-based classifiers for bacterial growth , 2003 .

[4]  Paul Davis,et al.  Application of probabilistic neural networks in modelling structural deterioration of stormwater pipes , 2006 .

[5]  J. P. Davies,et al.  The structural condition of rigid sewer pipes : a statistical investigation , 2001 .

[6]  George Morcous,et al.  Case-Based Reasoning System for Modeling Infrastructure Deterioration , 2002 .

[7]  David E. Goldberg,et al.  Genetic Algorithms in Search Optimization and Machine Learning , 1988 .

[8]  PingSun Leung,et al.  Predicting shrimp disease occurrence: artificial neural networks vs. logistic regression , 2000 .

[9]  Xiangming Zhou,et al.  Accelerated Assessment and Fuzzy Evaluation of Concrete Durability , 2005 .

[10]  Tarek Hegazy,et al.  Predicting cost deviation in reconstruction projects: Artificial neural networks versus regression , 2003 .

[11]  Kerry J. McManus,et al.  The effect of Thornthwaite Moisture Index Changes in ground movement predictions in Australian soils , 2004 .

[12]  Dulcy M. Abraham,et al.  Estimating Transition Probabilities in Markov Chain-Based Deterioration Models for Management of Wastewater Systems , 2006 .

[13]  G. Kuczera,et al.  Markov Model for Storm Water Pipe Deterioration , 2002 .

[14]  Gerardo W. Flintsch,et al.  Soft Computing Applications in Infrastructure Management , 2004 .

[15]  Seongkyu Chang,et al.  Application of Probabilistic Neural Networks for Prediction of Concrete Strength , 2005 .

[16]  Anne Ng,et al.  Selection of genetic algorithm operators for river water quality model calibration , 2003 .

[17]  Dulcy M. Abraham,et al.  CHALLENGING ISSUES IN MODELING DETERIORATION OF COMBINED SEWERS , 2001 .

[18]  Samuel T. Ariaratnam,et al.  Assessment of Infrastructure Inspection Needs Using Logistic Models , 2001 .