An Optimization Approach to Leak Detection in Pipe Networks Using Simulated Annealing

Abstract

A numerical approach, named as leak detection simulated annealing (LDSA), is developed based on the combination of transient flow simulations and simulated annealing to detect leaks in a pipe network. In the LDSA framework, an analysis model for pipe networks is first used to determine the steady-state nodal head solution of the network. For detecting the leaks, a valve closing suddenly at downstream side of the network is designed to create a transient event and then the method of characteristic is applied to simulate the transient flow in the network. Results from an experiment on leak detection in a single pipeline presented in the literature are used to verify the capability of LDSA in identifying the leaks. The result shows that the leaks represented by orifices in two cases can be accurately detected. The effect of C (d) A (discharge coefficient times effective area of orifice) on the detection is evaluated. In addition, three scenarios with three cases in each scenario are designed to test the performance of LDSA in numerical experimental studies for a synthetic network. For a very small value of C (d) A, it is found that the use of multiple observation points can increase the detection capability.