Leak detection in pipelines by inverse backward transient analysis

Abstract

A new method of leak detection in pipelines is introduced based on the inverse transient analysis. The transient state can be initiated in pipes by closing the downstream end valve. The time history of pressure fluctuations is measured at the valve location as a downstream boundary condition. The duration of valve closure is considered long enough to minimize effects of the friction factor and of waterhammer. The numerical modeling of transient analysis in time domain uses the backward Method Of Characteristics (MOC).An inverse problem is applied for leak detection by defining the least-squares criterion objective function. The leak areas at characteristic nodes are considered as decision variables. Leak parameters are determined using the Sequential Quadratic Programming (SQP) method. To eliminate undesirable noisy and uncertainties due to the valve modeling, a method independent of valve type, closure duration and valve operation was considered. Finally, numerical examples are solved based on this method.