International Journal for Numerical Methods in Fluids

doi: 10.1002/fld.1650150901pmid: N/A

Yu, Yun‐Sheng; Guangte, Wang

doi: 10.1002/fld.1650150902pmid: N/A

Existing discrete, linear rainfall‐run‐off models generally require the effective rainfall of a given storm as the input for computing the run‐off hydrograph. This paper proposes a rainfall‐run‐off model which uses the rainfall hyetograph as input and directly accounts for rainfall losses. The model combines an ARMA model and a modified Philip equation for rainfall losses due to infiltration. For a given watershed with measured rainfall hyetograph and the corresponding run‐off hydrograph, optimal values of model parameters are estimated by using a non‐linear iterative technique. Applications of the model to two different watersheds show that the computed run‐off hydrographs agree well with the measurements. The proposed model is a viable alternative to the widely used unit‐hydrograph method.

Daniels, H.; Rouvé, G.

doi: 10.1002/fld.1650150903pmid: N/A

Computer simulations may contribute significantly to the optimal design of air‐conditioning systems. To capture the effects of partially permeable walls such as bookshelves on the movement of air and heat, it is necessary to extend the density‐dependent Navier‐Stokes equations by an additional friction term. The finite element technique is convenient to approximate the extended equations in spatial co‐ordinates. For the time co‐ordinate a recently proposed semi‐implicit finite difference method is very efficient in terms of accuracy and computational complexity. A pressure correction approach is most appropriate to decouple the primitive variables in the extended Navier‐Stokes equations. The resulting algorithm has the interesting feature that small symmetric positive definite systems of equations can be solved sequentially for each of the primitive variables. Iterative solution of the systems of equations with preconditioned conjugate gradients combined with a compressed storage scheme allows fine grid computations at affordable costs. As an example a two‐dimensional version of the code was applied to an enclosure which was heated from the side and contained a porous wall. The time‐dependent computational results are compared with measurement data.

Ermakov, M. K.; Gryaznov, V. L.; Nikitin, S. A.; Pavlovsky, D. S.; Polezhaev, V. I.

doi: 10.1002/fld.1650150904pmid: N/A

A PC‐based system for modelling of convection in enclosures on the basis of the Navier‐Stokes equations is described and a number of test results are given. New examples of mixed convection in a square chamber and thermal convection in ordinary and porous (isotropic and anisotropic) vertical layers are presented which may be of interest in civil and building engineering.

Taniguchi, Takeo; Holz, Klaus‐Peter; Ohta, Chikashi

doi: 10.1002/fld.1650150905pmid: N/A

A grid generation method is proposed for an arbitrary two‐dimensional domain. The method, based on the Delaunay triangulation, is modified so that it can be used as a grid generator for an arbitrary two‐dimensional area with complex boundary geometry. Input data for the method are the co‐ordinates of all nodes and the ordering of nodes on each boundary. Its efficiency is examined through a number of actual problems, and a numerical experiment clarifies that the grid generation requires a CPU time which is proportional to the number of nodes.

Okajima, Atsushi; Ueno, Hisanori; Sakai, Haruhisa

doi: 10.1002/fld.1650150906pmid: N/A

By a finite volume method, laminar flows around bluff bodies with a rectangular cross‐section of various width‐to‐height ratios from 0·2 to 10 and with a cross‐section of a round leading edge and a square trailing edge have been computed on body‐fitted curvilinear co‐ordinates at Reynolds numbers of (1, 4, 7) × 103. Turbulent flows have also been computed by a standard k‐ϵ turbulence model. Computed results are compared with experimental data at a Reynolds number of 103 and clearly show the effects of the shape of the bluff body on the aerodynamic characteristics. We can successfully simulate some interesting phenomena whereby the flow pattern changes critically when the side ratio B/H is about 2·1 and 6; that is, a fully separated flow, an alternately reattached one and a stationarily reattached one. The results also reveal interactions between the wake and separation bubbles. There are, however, significant discrepancies between the results from the k‐ϵ turbulence model and experiments.

Kondo, Norio; Tosaka, Nobuyoshi; Nishimura, Toshio

doi: 10.1002/fld.1650150907pmid: N/A

A third‐order upwind finite element scheme is presented for numerical solutions of incompressible viscous flow problems. In order to achieve the third‐order upwind approximation for only the convection term in the Navier‐Stokes equations, a simplified Petrov‐Galerkin formulation in which a modified weighting function is expressed by the sum of a standard weighting function and its second and third spatial derivatives is employed. The mixed method is also employed in the formulation so that a discretization with high‐order accuracy in space is carried out by the use of linear elements. Because a truncation error caused by the third‐order upwind approximation is smaller than that of a first‐order upwind scheme, it is expected that the third‐order upwind scheme will greatly improve the numerical solutions of the Navier‐Stokes equations. Numerical results in one and two dimensions are presented to illustrate the effectiveness of the proposed scheme.

Ozono, Shigehira; Ohya, Yuji; Nakamura, Yasuharu; Nakayama, Ryuzo

doi: 10.1002/fld.1650150908pmid: N/A

The unsteady viscous flow around flat plates with square leading and trailing edges is investigated by using a finite difference computation of the incompressible two‐dimensional Navier‐Stokes equations. The chord‐to‐thickness ratio of a plate ranges from d/h=3 to 9, with a Reynolds number based on the plate's thickness equal to 103. The numerical analysis confirms the finding obtained in our previous experiment that vortex shedding from flat plates with square leading and trailing edges is caused by the impinging shear layer instability. The Strouhal number based on the plate's chord increases stepwise with increasing d/h in accordance with the experiment. The numerical analysis also gives some crucial information on the complicated vortical flow occurring near the trailing edge. Finally, the mechanism of the impinging shear layer instability is discussed in the light of the experimental and numerical findings.

Kashiyama, Kazuo; Okada, Tsuyoshi

doi: 10.1002/fld.1650150909pmid: N/A

This paper presents a new automatic mesh generation method for the finite element analysis of shallow water flow. The key feature of this method is that the finite element mesh can be generated so that the element Courant number is nearly constant in the whole domain. It follows that the numerical stability and accuracy improve automatically. Moreover, the finite element mesh data, including the data of water depth, can be prepared automatically. The three‐node triangular element is used for the finite element. In order to show the efficiency of the method presented, the mesh obtained by this method is applied to some shallow water flow analyses. This method is shown to be a useful and powerful tool for the preparation of optimal finite element mesh data for shallow water flow analysis.

Onishi, Yasuo; Trent, Donald S.

doi: 10.1002/fld.1650150910pmid: N/A

In this study the adequacy of the k‐ϵ turbulence model and the feasibility of the three‐dimensional hydrodynamic‐transport models TEMPEST and FLESCOT for deep ocean radionuclide disposal assessment were evaluated qualitatively. TEMPEST and FLESCOT were applied to a hypothetical, two‐dimensional, deep ocean case with and without stratifications. TEMPEST with the k‐ϵ model was applied to obtain quasi‐steady state eddy viscosity distributions. With calculated eddy viscosity distributions as part of the input, FLESCOT then calculated distributions of velocity, water temperature, sediment and the dissolved and sediment‐sorbed radionuclide, assuming that the radionuclide was disposed on the ocean bottom. Results revealed that the computed eddy viscosity increased almost linearly with vertical distance near the ocean bottom, then rapidly decreased towards a molecular viscosity value when the vertical gradient of the velocity distribution became very small. The results also demonstrate the importance of the density gradient to suppress the turbulent kinetic energy production, resulting in reduced eddy viscosity, producing the maximum computed eddy viscosity of 0·2 Pa s, which compares well with the reported value of 0·07 Pa s in the deep ocean. Thus the k‐ϵ turbulence model appears to be qualitatively applicable to the deep ocean environment.