Risk assessment and prediction of gas leakage; Case study in Noord-Holland

Organisator Laboratory of Geo-information Science and Remote Sensing

wo 9 april 2014 12:30 tot 13:00

Locatie Gaia, building number 101
Droevendaalsesteeg 3
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 1

by Yan Zeng (China)


The replacement of old gas pipe materials requires huge investments. Phased renovation is also necessary because during maintenance the network should remain operational. A risk-based maintenance programme that prioritizes pipes with the highest failure risks seems appropriate for this purpose. The cause of pipe failure is very complicated; it necessary to detect the most important factors that have correlation with pipe failure. This study used archived leakage records to compute the prediction failure probability of individual pipes for most of Noord-Holland. It also explores whether environmental variables (e.g. land use type, soil type, water storage capacity and population density) can improve the model compared to only using physical characteristics of pipes (e.g. pipe age, pipe material, pressure, diameter, etc.). Logistic regression and tree method were used to develop prediction models. The validity of the statistical models was assessed based on the Root-mean-squared error (RMSE) and explained variance (R2) using a separate validation dataset. After that, individual risks were assessed by calculating the zones of influence of each type of hazard and the number of affected people should the pipe failure happen. Results showed that the environmental variables used in this research did not improve the prediction model. The results computed by classification tree were better than logistic regression. However, both methods had low predictive power with an extremely low RMSE. Nevertheless, the analysis of potential consequences of network failure demonstrated bottlenecks in the gas network. The study also highlights the need of improving the quality of the pipe database especially the pipe installed date and joint type and suggests environmental variable that may be stronger correlated with pipe breakages.

Keywords: gas distribution network; probability of pipe leakage; individual risk; data analysis