American Society of Civil Engineers


Forecasting the Remaining Useful Life of Cast Iron Water Mains


by Mohamed Fahmy, P.E., (corresponding author), M.ASCE, (Ph.D. Candidate, Dept. of Building, Civil, and Environmental Engineering, Concordia Univ., 1455 Blvd. de Maisonneuve W, Montreal QC, Canada H3G 1M8 E-mail: ma_fahmy@alcor.concordia.ca) and Osama Moselhi, P.E., F.ASCE, (Professor, Dept. of Building, Civil, and Environmental Engineering, Concordia Univ., 1455 Blvd. de Maisonneuve W, Montreal QC, Canada H3G 1M8. E-mail: moselhi@encs.concordia.ca)

Journal of Performance of Constructed Facilities, Vol. 23, No. 4, July/August 2009, pp. 269-275, (doi:  http://dx.doi.org/10.1061/(ASCE)0887-3828(2009)23:4(269))

     Access full text
     Purchase Subscription
     Permissions for Reuse  

Document type: Journal Paper
Abstract: Effective asset management strategy of civil infrastructure systems requires integration of technical and financial plans. This is particularly true in managing water mains, which requires knowledge of their current condition and their forecasted remaining useful life. This paper presents a model designed to forecast the remaining useful life of cast iron water mains. The model is easy to use and its generated results are utilized in determining condition rating of the water mains being considered. The model considers factors related to pipe properties, its operating conditions, and the external environment that surrounds the pipe. In addition, it overcomes limitations associated with existing models. Three different data-driven techniques are considered in the model development; each is used to study the relationship between remaining useful life and a set of deterioration factors, and to forecast remaining useful life of cast iron water mains. These techniques are multiple regression and two types of artificial neural networks: multilayer perceptron; and general regression neural network. The data used in model development were acquired from 16 municipalities in Canada and the United States. The results produced by the developed models correlate well with the actual conditions.


ASCE Subject Headings:
Water pipelines
Neural networks
Failures
Forecasting
Cast iron
Water distribution systems