The implementation of CLEAR algorithm for incompressible fluid flow and heat transfer on a non-staggered grid
Zhiguo Qu,Wenquan Tao,Yaling
2004-01-01
Abstract:In this paper, a detailed implementation procedure of CLEAR algorithm on a non-staggered grid system are formulated. Discussion is made on the method to predict velocity and pressure in a unique manner that is independent of relaxation factor. In order to verify the performance of CLEAR algorithm, four numerical examples on non-staggered grid of forced convective fluid flow and natural convection with available solutions are provided to compare the convergence performance between the CLEAR and SIMPLER It is shown that on the collocated grid, the CLEAR algorithm can also greatly enhance the convergence rate based on the iteration number and the corresponding consumed CPU time compared with the SIMPLER algorithm with similar robustness. INTRODUCTION The pressure correction method that belongs to segregated algorithm is an approach widely employed for numerically solving the Navier-Stokes equations for fluid flow and heat transfer problems. The SIMPLE algorithm proposed by Patankar and Spalding (1972) is the first such algorithm widely used in literature. Since then several variants are reported to enhance its convergence rate, among whom are SIMPLER (Patankar 1981), SIMPLEC (Van Doormal and Raithby, 1984), SIMPLEX (Van Doormal and Raithby, 1985, Raithby and Schneider, 1988), PISO (Issa, 1985), SIMPLET (Sheng and Shoukri, 1998), and MSIMPLER (Yu eral., 2001).In addition, Yen and Liu (1993) presented the explicit correction step method to accelerate the convergence rate. All the above-mentioned methods are usually named SIMPLE-like or SIMPLE-family algorithm. Moukalled and Darwish (2000) made a comprehensive review and given a unified reorganization expression form for all the pressure correction algorithms. The common characteristic of these algorithms is that a pressure correction term is introduced to the calculation procedure of each segregated solution step to improve the velocity that is solved from momentum-discretized equation as the intermediate value. The pressure correction term in the SIMPLE-family algorithms is used to improve the intermediate velocity solved from the momentum equation such that the modified velocity can satisfy the mass conservation condition for each control volume at each iteration level, that is critical for iteration convergence as indicated by Blosch and Shyy(1993) .In the derivation of pressure correction equation, the effects of the pressure corrections of the neighboring grid points are neglected that will not affect the final solutions when the iterative process converges (Patankar,1980), but does affect the convergence rate (Shyy and Mittal,1998). Because of this neglecting, all the above-mentioned algorithms are of semi-implicit type. Very recently, the present authors (Tao etal, 2004) proposed a fully implicit segregated algorithm CLEAR (Couple and Linked Equations Algorithm Revised) for incompressible fluid flow and heat transfer. The CLEAR algorithm avoids introducing the pressure correction term and velocity correction term to discard the basic assumption of the SIMPLE series algorithm. The new algorithm is tested with six examples on staggered grid system, and it is proved that it can enhance convergence rate to a great extent. In this study, the CLEAR algorithm is extended to collocated grid system to test its feasibility. To fulfill the fundamental requirement for eliminating the decoupling problem between velocity and pressure and making the solution to be independent of under-relaxation factor, detailed investigation is made on the previous related methods. For some mildly irregular computation domain , the domain extension method (Tao, 2001) is applied, and is further formulated in this study to meet the special requirement on the collocated grid to treat the solid region in the computational domain. Based on the above preliminary research, the calculation procedures of SIMPLER and CLEAR on collocated grids are presented for comparison. Six 2-D numerical examples with available solution are computed to investigate the performance of the CLEAR and SIMPLER algorithm on collocated grid system. Discussion on Momentum Interpolation The staggered grid is widely employed in CFD/NHT literatures since it can efficiently guarantee the coupling between velocity and pressure. However it shows its inconvenience of complicity for the code development in unstructured grid and curvilinear bodyfitted grid, especially for 3-D computation. While on the non-staggered grids, such complicity can be greatly alleviated. The crucial issue in using non-staggered grids is how to eliminate the decoupling between pressure and velocity. In 1980’s the momentum interpolation method (MIM) on non-staggered grid is first presented by Rhie and Chow (1983) to prevent the decoupling problem. It is subsequently reformulated by Peric (1985) and Majumdar (1986). Later, Majumdar (1988) and Miller and Schmidt (1988) pointed out that the Rhie and Chow’s method has the weakness that the solution is under-relaxation factor dependent in some extent despite of removing the false pressure field. To make a remedy for this unpleasant deficiency, a few schemes are proposed lately. An easy technique (Kobayashi and Pereira, 1991) is to set the under-relaxation factor 1 α = before momentum interpolation is implemented, but it may decrease the robustness of the algorithm to some extent. Hence, we can see that in order to make a reliable and efficient computation on non-staggered grid, the following three aspects should be guaranteed. (1) The algorithm can avoid the checkerboard pressure distribution; (2) The converged resolution is independent of the under-relaxation factor; (3) The algorithm should possess required robustness. In order to develop a computation scheme on the collocated grid, which possesses the above-mentioned three features, we first make a brief review on the existing implementation procedures on the collocated grids, take advantage of some successful practice and then propose a new implementation scheme for this study. In the following, a brief description of the governing equations and the discretization procedure will be presented. For simplicity of presentation, we take two-dimensional incompressible laminar steady fluid flow in Cartesian coordinates as an example. The system for non-staggered grid is depicted in Figure. 1.