Parameter quality conditions in open-channel inverse problems
作者:
Rahman H. Khatibi,
Peter R. Wormleaton,
John J. R. Williams,
期刊:
Journal of Hydraulic Research
(Taylor Available online 2000)
卷期:
Volume 38,
issue 6
页码: 447-458
ISSN:0022-1686
年代: 2000
DOI:10.1080/00221680009498298
出版商: Taylor & Francis Group
数据来源: Taylor
摘要:
Some open-channel empirical parameters lack exact values due to (a) not having definitive measurement methods and/or (b) not rendering a "properly-posed" system of equations in the formulation of their inverse problems. The values of these parameters are therefore susceptible to dataerrors, imperfections in governing equations or insufficiency of the gauged data used in inverse problems. Best values of such parameters in a mathematical sense can be identified by the implementation of inverse problems using optimisation methods but there are potential pitfalls. A comprehensive review of inverse problems is presented in this paper outlining some of the pitfalls in their implementation and illustrating the parameter quality conditions of "identifiability, uniqueness and stability". The uniqueness condition was approached by statistical methods in this paper. Other considerations of equal importance were found to include the compliance with the assumptions underlying optimisation methods and a careful selection of the objective function.
点击下载:
PDF (11822KB)
返 回