Please use this identifier to cite or link to this item: http://hdl.handle.net/2440/80948
Type: Conference item
Title: Minimization algorithms and experimental inverse transient leak detection
Author: Vitkovsky, J.
Simpson, A.
Lambert, M.
Citation: Conference on Water Resources Planning and Management, Roanoke, Virginia, USA, 2002: 10 p
Publisher: ASCE
Issue Date: 2002
Conference Name: Conference on Water Resources Planning and Management (2002 : Roanoke, Virginia, USA)
Statement of
Responsibility: 
John P. Vítkovský, Angus R. Simpson, and Martin F. Lambert
Abstract: The inverse transient method has been verified for leak detection from a laboratory pipeline using a least-squares minimization. The process determines the size and location of a leak. In previous numerical studies the Levenberg-Marquardt algorithm and genetic algorithms were used to perform this minimization. In the current experiment, the Levenberg-Marquardt algorithm yields incorrect leak locations and magnitudes. The complexity of the objective function causes the Levenberg-Marquardt algorithm to fail. The inverse transient method has been verified for leak detection from a laboratory pipeline using a least-squares minimization. The process determines the size and location of a leak. In previous numerical studies the Levenberg-Marquardt algorithm and genetic algorithms were used to perform this minimization. In the current experiment, the Levenberg-Marquardt algorithm yields incorrect leak locations and magnitudes. The complexity of the objective function surface causes the Levenberg-Marquardt algorithm to fail. The use of the shuffled complex evolution algorithm—a more global minimization algorithm—improves results, however convergence is slow. A systematic approach to the application of the Levenberg-Marquardt algorithm determines the correct leak location and magnitude more efficiently.
Rights: Copyright status unknown
RMID: 0030000289
Appears in Collections:Civil and Environmental Engineering publications

Files in This Item:
There are no files associated with this item.


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.