Premium
Parameter identification of river current and diffusion by reduced Kalman filter finite element method
Author(s) -
Kanai Osamu,
Kawahara Mutsuto
Publication year - 2011
Publication title -
international journal for numerical methods in fluids
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.938
H-Index - 112
eISSN - 1097-0363
pISSN - 0271-2091
DOI - 10.1002/fld.2694
Subject(s) - diffusion , kalman filter , finite element method , mathematics , computation , time domain , filter (signal processing) , constant (computer programming) , mechanics , mathematical analysis , physics , statistics , thermodynamics , algorithm , computer science , computer vision , programming language
SUMMARY The objective of this study is to propose a parameter identification of a river current and diffusion coefficients by using the reduced Kalman filter finite element method, which has been previously presented and now extended by the authors. For comparison, the estimation computations of velocity, water elevation, and chemical oxygen demand (COD) concentration are carried out on the basis of nonlinear shallow water flow and compared with the observations carried out at the Teganuma river in Japan. A marked discrepancy in COD concentration is found between the computed and observed results. The correlation factor between the computed and observed results is 0.51. To reduce the discrepancy, the authors believe that the diffusion coefficients should be identified. Assuming that the diffusion coefficient is constant in the entire domain and over the entire total duration, satisfactory results were not obtained. Thus, the computational domain is divided into four subdomains according to the water depth. Assuming that the diffusion coefficients are constant in each subdomain, the identification is carried out. Relatively good, albeit unsatisfactory, results are obtained. The discrepancy between the computed and observed COD concentration has special features. In some time zones, the computed results are larger whereas in other time zones, they are smaller than the observed results. To compensate this discrepancy, we assumed that the diffusion coefficient is a function of COD concentration. The correlation factor is improved to be 0.73. The identified diffusion coefficients are time functions that change cyclically with a period of 24 h . This fact suggests that biological phenomenas occurred in the river. Copyright © 2011 John Wiley & Sons, Ltd.