An efficient algorithm for high accuracy particle tracking in finite elements

被引:15
作者
Pokrajac, D
Lazic, R
机构
[1] Univ Aberdeen, Dept Engn, Aberdeen AB24 3UE, Scotland
[2] Metrol Technol Ltd, Aberdeen AB21 0GU, Scotland
关键词
particle tracking; path line; FEM; spatial interpolation;
D O I
10.1016/S0309-1708(02)00012-X
中图分类号
TV21 [水资源调查与水利规划];
学科分类号
081501 ;
摘要
We propose an algorithm for particle tracking based on Cheng's method [Int. J. Numer. Meth. 39 (1996) 1111-1136]. Velocities in a flow field are known at a series of points and interpolated between them by finite element local functions. Tracking is performed in local coordinates, element by element, using any standard ODE solution method. The exit from an element is found using the polynomials to interpolate between the tracking points. The algorithm was tested and compared to Pollock's and Cheng's method in a series of numerical experiments, in which the Euler, Runge-Kutta 2, Runge-Kutta 5(4) and Runge-Kutta 6(4) ODE solution methods were combined with first-, second-, third- and fifth-order exit polynomials. The known velocities had a random error with standard deviation of 0%, 0.1% and 1% of the velocity. Meaningful results were obtained only when the spatial interpolation error and the error of the tracking method were calculated separately, otherwise some results were misleading. The numerical experiments confirmed that the accuracy of the exit polynomial has to be consistent with the ODE solution method. Quadratic interpolation of velocities on a coarser mesh often gives more accurate path lines and requires less computational time than linear interpolation. Pollock's method for particle tracking is viable only if input data are rather inaccurate and path lines nearly straight. Cheng's method is appropriate for moderately accurate input data, while the proposed algorithm with Runge-Kutta 5(4) or Runge-Kutta 6(4) method and fifth-order exit polynomial has excellent accuracy. Computational time is about 10 times longer than for Cheng's method while the accuracy is increased by several orders of magnitude. (C) 2002 Published by Elsevier Science Ltd.
引用
收藏
页码:353 / 369
页数:17
相关论文
共 39 条
[1]  
Bear J., 1979, HYDRAULICS GROUNDWAT
[2]   An adaptive pathline-based particle tracking algorithm for the Eulerian-Lagrangian method [J].
Bensabat, J ;
Zhou, QL ;
Bear, J .
ADVANCES IN WATER RESOURCES, 2000, 23 (04) :383-397
[3]  
BORELI M, 1982, P IAHR INT S HYDR PR
[4]   A VARIABLE ORDER RUNGE-KUTTA METHOD FOR INITIAL-VALUE PROBLEMS WITH RAPIDLY VARYING RIGHT-HAND SIDES [J].
CASH, JR ;
KARP, AH .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1990, 16 (03) :201-222
[5]   BRKF45 - A FORTRAN SUBROUTINE FOR SOLVING 1ST-ORDER SYSTEMS OF NONSTIFF INITIAL-VALUE PROBLEMS FOR ORDINARY DIFFERENTIAL-EQUATIONS [J].
CASH, JR .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1989, 15 (01) :29-30
[6]   A BLOCK 6(4) RUNGE-KUTTA FORMULA FOR NONSTIFF INITIAL-VALUE PROBLEMS [J].
CASH, JR .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1989, 15 (01) :15-28
[7]   MODELING GROUNDWATER-FLOW FIELDS CONTAINING POINT SINGULARITIES - STREAMLINES, TRAVEL-TIMES, AND BREAKTHROUGH CURVES [J].
CHARBENEAU, RJ ;
STREET, RL .
WATER RESOURCES RESEARCH, 1979, 15 (06) :1445-1450
[8]  
Cheng HP, 1996, INT J NUMER METH ENG, V39, P1115, DOI 10.1002/(SICI)1097-0207(19960415)39:7<1115::AID-NME895>3.0.CO
[9]  
2-4
[10]   EULERIAN-LAGRANGIAN SOLUTION OF THE CONVECTION-DISPERSION EQUATION IN NATURAL COORDINATES [J].
CHENG, RT ;
CASULLI, V ;
MILFORD, SN .
WATER RESOURCES RESEARCH, 1984, 20 (07) :944-952