(598k) Robustness Anlysis in Pipe Prediction Model of Water Distribution Network

Gobong Choi, Shinje Lee, Dongwhi Jeong, Jong Min Lee*, Gibaek Lee, and Jungchul Seo

School of Chemical and Biological Engineering, Seoul National University,

1 Gwanak-ro, Gwanak-gu, Seoul, 151-744, Korea

Department of Chemical and Biological Engineering, Korea National University of Transportation

50 Daehak-ro Chungju-si, Chungbuk 380-702, Republic of Korea

Samchully, Samchully Bldg 35-6, Youido-dong, Yongdungpo-gu, Seoul 150-885 Republic of Korea

 Water distribution network is a system that contains several components for supplying consumer with enough water of good quality. This distribution network can detorirate with not only enviromental conditions (soil type, acidity, etc.), but operational conditions (e.g., operational pressure), resulting in an serious economic burden and social impacts. Nevertheless, it is difficult to assess condition of the buried infrastrucure. Aging of pipe is used as a sole and primary criterion for replacement or rehabiliatation of pipe. It is also common that old water pipes can satisfactorily function while new pipes may show poor performance; this is because there are many other factors leading to pipe failures, including environmental conditions, physical characteristics, and historical factors. Hence, it can be costly to make decisions for only based on the age of pipe. There are two kinds of approaches for prediction of water main condition: first-principle and statistical approaches [1]. The first-principle approach constructs a mechanistic model to describe the physical mechanism leading to pipe breakage. It evalautes pipe failure due to the stesse applied on pipe under various conditions. Failure prediction of water main using this model is limited due the model complexity and expensive data requirement for reliable parameter estimation. On the other hand, statistical methods utilize historical data including recorded historical failure and enviromental factors and relates these to estimate pipe condition. While these models are easy to apply, obtaining such data sets is time-consuming for constructing reliable model. Hence, it is necessary to select a relevant set of variables before collecting data and determining a model structure. Towards this, this study first proposes a scoring scheme to rank the ease of data acquisition for each variable. Furthermore, robustness of existing statistical models for pipe condition assessment is examined using global sensitivity analysis and Monte-Carlo simulations. Incorporating these two schemes will aid in selecting the most sensitive factors in constructing statistical empirical model.

References

[1]    Kleiner, Y. AND Rajani, B., “Comprehensive review of structural deterioration of water mains: statistical models”, Urban Water, 3, pp. 131-150, May 2001