Bayesian Inference of Contaminant Source in Water Distribution Systems
by Hui Wang, (North Carolina State University, Campus Box 7908, Raleigh, NC 27695-7908) and Kenneth W. Harrison, (Earth System Science Interdisciplinary Center, College Park, MD 20740-3823; and Hydrological Sciences Branch, NASA Goddard Space Flight Center, Greenbelt, MD, 20771)
Section: 13th Annual Water Distribution Systems Analysis Symposium, pp. 268-275, (doi: http://dx.doi.org/10.1061/41173(414)29)
Access full text
Purchase Subscription
Permissions for Reuse 
| Document type: |
Conference Proceeding Paper |
| Part of: |
World Environmental and Water Resources Congress 2011: Bearing Knowledge for Sustainability |
| Abstract: |
Bayesian approach has the advantage of incorporating new observations in inferring contaminant history in water distribution system. A tailored Markov Chain Monte Carlo(MCMC) procedure is proposed by Harrison and Wang(2009,2010) on EWRI conference. As an extension to previous work, full history of the contaminant event, including injection node, contamination magnitude, starting time, and injection duration, is inferred under this framework. It provides probabilistic inference about the location of contaminant node by incorporating the hydraulic uncertainties in water distribution networks. A comparison between deterministic water demand scenario and uncertain water demand scenario shows the advantage of Bayesian inference. Parallel computing is applied while implementing the MCMC algorithm and several methods to reduce computing time in estimation of the likelihood function are also discussed. |
|
