283731 On-Line Fault Detection and Diagnosis of Pipeline in Water Distribution Network Systems

Monday, October 29, 2012
Hall B (Convention Center )
Shin Je Lee1, Go Bong Choi2, Dae Shik Kim3, Jung Chul Suh4, Gi Baek Lee5 and Jong Min Lee1, (1)School of Chemical and Biological Engineering, Seoul National University, Seoul, South Korea, (2)School of Chemical and Biological Engineering, Seoul National University, Seoul, 151744, South Korea, (3)Seoul National University, (4)Samchully, Seoul, South Korea, (5)Korea National University of Transportation, Chungju-si, South Korea

Water pipe networks are generally installed underground. Once they are equipped, it is difficult to recognize the state of pipes when faults such as leak or burst happen. Accordingly, post management is often delayed, thus making the loss of pipelines increase. Therefore, a systematic fault management system of water pipe network is required to prevent accidents and minimize the loss. The conventional fault detecting methods include listening, noise logging, tracer gas, and GPR. Those methods are easy to implement, but they have a limitation that considerable time, labor, and finances are required to apply those methods. Therefore, a novel approach for detecting and diagnosing faults of water pipelines is required.

In this work, we develop online fault detection and diagnosis system of water pipe networks that detects faults and diagnoses the location and cause of the faults using online data such as pressure or flow rate. The data are measured and gathered into communication devices in lab scales. The proposed approach takes the following steps. First, we decide the locations of sensors for measuring pressure or flow rate and the locations of fault tests for leak or burst. Genetic algorithm (GA) and inverse transient analysis (ITA) are applied to find optimal sensor locations and fault test locations, respectively. A modified scale space filtering (SSF) technique is applied to filter out noise included in the collected data. The proposed SSF approach reduces computational complexity and addresses the end-effect. Fault detection and diagnosis schemes are then applied to the filtered data. Cumulative sum (CUSUM), Fast Fourier transform (FFT), and Discrete wavelet transform (DWT) are applied for effective diagnosis. The proposed approach will be illustrated on a practically-sized pipe network with field test data.


Extended Abstract: File Not Uploaded