Leak detection systems for liquid pipelines are installed to minimize spillage in case of a leak. Therefore reliability, sensitivity and detection time under practical operating conditions are the most important parameters of a leak detection system. Noise factors to be considered among others are unknown fluid property data, friction factor, instrument errors, transient flow, slack-line operation and SCADA update time. The opening characteristics and the size of leaks differ considerably from case to case.

Each software-based leak detection method available today has its particular strength. As long as just one or two of these methods are applied to a pipeline a compromise has to be found for the key parameters of the leak detection system.

The paper proposed illustrates how a combination of several different software-based leak detection methods together with observer-type system identification and a knowledge-based evaluation can improve leak detection. Special focus is given to leak detection and automated leak locating under transient flow conditions. Practical results are shown for a crude oil pipeline and a product pipeline.

This content is only available via PDF.