WO2002003173A2

WO2002003173A2 - Equal order method for fluid flow simulation

Info

Publication number: WO2002003173A2
Application number: PCT/US2001/041228
Authority: WO
Inventors: David Marc Waite; Shaupoh Wang; Jenn-Yeu Nieh
Original assignee: Aemp Corporation
Priority date: 2000-07-01
Filing date: 2001-07-02
Publication date: 2002-01-10
Also published as: JP2004503005A; EP1316053A2; US6611736B1; EP1316053A4; CA2415165A1; WO2002003173A3; AU2001279281A1

Abstract

The present invention provides a method for solving the Navier-Stokes equation of viscous, incompressible laminar flows with moving free surfaces in complex domains. The method (100) derives discretized momentum equations using finite element procedures on a fixed mesh (40). Pressure and velocity of this discretized momentum equation is defined on all the nodes of the mesh (42). The method further provides a discretized continuity equation where velocity is defined on a Gaussian-point between the nodes of an element (105). Further, the Gaussian-point velocity is a function of the nodal velocity. The discretized momentum and continuity equations are solved to determine the pressure field (106). The velocity field is then determined from the newly determined pressure field. As the velocities and pressure fields are solved for at each nodal point, the accuracy of the solution is preserved. The predicted Gaussian-point velocity field conserves mass to machine round-off levels for element types in which the pressure gradient and velocities can be evaluated on a consistent basis.

Description

EQUAL ORDER METHOD FOR FLUID FLOW SIMULATION

CROSS REFERENCE TO RELATED APPLICATIONS

The present application is a continuation-in-part application of a co-pending and commonly owned U.S. Patent Application, entitled EQUAL ORDER

METHOD FOR FLUID FLOW SIMULATION, Serial No. 09/610,641, filed on July 1, 2000 by the same inventive entity. The referenced application is incorporated herein by reference in its entirety.

FIELD OF THE INVENTION

This invention relates to a numerical method for solving the Navier-Strokes equation and has application in computer simulation of viscous, incompressible laminar fluid flow; in particular, it relates to a method using finite element-control volume techniques, wherein the velocity and pressure of the fluid are evaluated at the same order to preserve accuracy of solution without sacrificing computation efficiency.

BACKGROUND OF THE INVENTION

Modeling incompressible, viscous flow generally involves solving the Navier-Stokes equation to obtain the velocity field. The Navier-Stokes equation describes the fundamental law of mass conservation and momentum balance of incompressible viscous flow. For a given pressure field, there is no particular difficulty in solving the momentum equation. By substituting the correct pressure field into the momentum equation, the resulting velocity field satisfies the continuity constraint. The difficulties in calculating the velocity field lies when the pressure field is unknown. The pressure field is indirectly specified via the continuity equation. Yet, there is no obvious equation for obtaining pressure. The difficulty associated with the determination of pressure has led to methods that eliminate pressure from the momentum equations. For example, in the well-known "stream-function/vorticity method" described by others and

Gosman et al. (In Heat and Mass Transfer in Recirculating Flows, Academic, New York, 1969.) The stream-function/vorticity method has some attractive features, but the pressure, which has been so cleverly eliminated, frequently happens to be a desired result or even an intermediate outcome required for the calculation of density and other fluid properties. In addition, the method cannot be easily extended to three-dimensional situations. Because most practical problems are three-dimensional, a method that is intrinsically restricted to two dimensions suffers from a serious limitation.

Numerical methods have been developed for solving the momentum equation and obtaining the pressure field. As finite-difference methods, the SIMPLE (Semi-Implicit Method for Pressure-Linked Equations by Patenkar, in Numerical Heat Transfer and Fluid Flow, McGraw Hill, 1980), and SIMPLER (SIMPLE Revised) algorithms are well-known examples of the numerical methods using finite-difference techniques. The SIMPLE and SIMPLER methods employing finite element techniques, discretize the flow domain, and transform the momentum equation into a set of algebraic equations. The pressure-gradient term is represented in the discretized momentum equations by the pressure drop between two grid points. One consequence of this representation is the non-uniqueness of the solution of the pressure field. As long as the pressure values at alternate grid points are uniform, the discretized form of the momentum equations is incapable of distinguishing a checker-board pressure field from a uniform pressure field.

However, the potential to result in a checker-board pressure field should make the solution unacceptable. Another consequence of this representation is that pressure is, in effect, taken from a coarser grid than the one actually employed. This practice diminishes the accuracy of the solution. It has been recognized that the discretizing equation does not demand that all the variables be calculated on the same grid points. The SIMPLE and SIMPLER methods resolve the checker boarding of pressure by using a "staggered" grid for the velocity component. In the staggered grid, the velocity components are calculated on a grid staggered in relation to the normal control volume around the main grid points. A computer program based on the staggered grid must carry all the indexing and geometric information about the locations of the velocity components and must perform interpolation for values between grid points. Furthermore, a "staggered" grid is difficult to apply to a flow domain that is meshed into irregularly shaped finite elements.

When modeling incompressible viscous flow with finite-element methods, reduced order and reduced integration methods have been suggested to overcome the checker-board problem in the predicted pressure field. The reduced order method suggests that pressure be evaluated in a coarser grid, e.g. at alternate nodal positions of the finite element mesh (Baliga and Patankar, A Control Volume Finite-Element Method for Two Dimensional Fluid Flows and Heat Transfer, in Numerical Heat Transfer, 6, 245-261 (1980)). The reduced integration method suggests to use a lower-order interpolation function to describe the pressure profile within each element resulting in less than full integration for one pressure term. Either the reduced order method or the reduced integration method diminish the accuracy of the pressure field, and therefore compromises the accuracy of the subsequent mass flux calculations. These methods have shown some success in specific applications, but there are many limitations to their general application. Especially, when moving free surfaces exist, the accuracy and resolution of the predicted location of the free surfaces rely heavily on how well the mass conservation is satisfied. Since pressure field is closely coupled with mass conservation, reducing the order of pressure, relative to the velocity means that the accuracy of mass conservation and the predicted free surface location also decreases proportionally.

There is therefore a need for a method to solve the momentum equations of incompressive, viscous fluid flows such that accuracy of predicted velocity, pressure and free-surface location are preserved without sacrificing computation efficiency. SUMMARY OF THE INVENTION

Accordingly, one preferred object of the present invention is to provide a numerical method for solving the Navier-Stokes equation of viscous incompressible, laminar fluid flow and wherein the velocity and pressure of the fluid are evaluated at the same order to preserve accuracy of solution without sacrificing computation efficiency.

Another preferred object of the present invention is to provide a numerical method for solving the momentum equation and providing a solution of such accuracy that the global mass conservation can be maintained to machine round-off tolerances.

One embodiment of the present invention is a numerical method for solving the coupled momentum and mass conservation equation. The method preferably operates within a processor environment. The method discretizes the flow domain into a mesh of finite elements. In order to maintain same accuracy in velocity, pressure and mass conservation, the velocity and pressure are defined at each nodal point and the momentum equation is discretized with typical finite-element procedure. The checker-board problem in predicted pressure field is avoided by defining a Gauss-point velocity field with same order as the pressure gradient within each element and the continuity equation is discretized with the Gaussian- point velocity, instead of the nodal velocity. Gaussian-points are the sampling points where a multi-dimensional high-order polynomial can be integrated numerically. Details of Gaussian-points and the numerical integration method, called Gaussian quadrature, can be found in Finite Element Structural Analysis by T. Y. Yang, Prentice-Hall Inc. The reason why the proposed method can eliminate checker-board problem without sacrificing the solution's accuracy can be illustrated most clearly by comparing its effects on linear (first order) triangular element with Patankar' s staggered-grid method. In a linear triangular (three-node) element with velocity and pressure defined on each node, the pressure gradient is a constant. According to the proposed method, the Gauss velocity should be a constant, which is equal to the average of the velocities on the three nodal points. Conceptually, this single-point Gauss velocity is "centered" inside the pressure field defined at the nodes, similar to the effects of using staggered points for velocity in Patankar' s method in finite-difference procedure. More details on the mathematical justification are included in the description of the embodiment. As the velocities and pressure fields are solved for at each nodal point, the accuracy of the solution is preserved. The predicted Gauss-point velocity field conserves mass to machine round-off levels for element types in which the pressure gradient and velocities can be evaluated on a consistent basis. One aspect of the present invention is to provide a method of solving the governing equations of an incompressible viscous flow in a flow domain comprising: providing a discretized mesh of said flow domain of said fluid flow, wherein said mesh comprises elements having nodes; deriving a discretized momentum equilibrium equation based on a finite-element formulation, wherein pressure and velocity are defined at every nodes of elements of said mesh; deriving discretized mass conservation equations based on said finite-element formulation, wherein pressure gradient is defined over each element and velocity is defined on a Gaussian-points inside each element, thereby said Gaussian-point velocity having the same order as said pressure gradient; determining velocity and pressure at each nodes of said flow domain; repeating said deriving and said determining until a convergence test of said velocity values is satisfied. Another aspect of the present invention is a method operating within a processor environment for simulating incompressible viscous fluid flows comprising: inputting material properties of said fluid and boundary conditions; discretizing a flow domain of said fluid flow into a mesh of finite elements; each elements having nodes; defining pressure and velocity on said nodes; deriving a discretized momentum equation based on said velocities and pressures defined on the nodes; defining a Gauss-point velocity inside each of said elements, wherein said Gaussian-point velocity have same order as pressure gradient in said element and is a function of said velocities defined on said nodes; deriving a discretized continuity equation based on said Gaussian-point velocity; determining values of pressure and velocity at each said nodes of said finite element; and repeating said deriving and said determining until a convergence test of said velocity values is satisfied.

A further aspect of the present invention is a numerical method operating within a processor environment for solving the momentum equation of a viscous, incompressible, laminar flow within a flow domain comprising: (a) discretizing said flow domain into a mesh of finite elements having nodes and fluxing surfaces; (b) forming momentum equations, wherein velocity and pressure are defined on each nodes of said finite elements; (c) defining a Gaussian-point velocity inside each of said finite elements, wherein said Gaussian-point velocity is computed from said velocities defined at the nodes of each same said finite elements; (d) updating velocity; (e) repeating said forming, computing, calculating and updating until a convergence test of said velocity is satisfied.

Other objects, embodiments, forms, aspects, features, advantages, and benefits will become apparent from the description and drawings provided herein.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram of a machine to implement the present invention.

FIG. 2 is a flowchart schematically showing one embodiment of a method of the present invention.

FIG. 3 is an example of a triangular finite element similar to those used in the illustrated embodiment of the method of the present invention.

FIG. 4 is an example of a triangular finite element mesh including a control volume and associated sub-control volumes and fluxing surfaces, preferred by the illustrated embodiment of the method of the present invention.

FIG. 5 is an example of a tetrahedral finite element preferred by an embodiment of the method of the present invention.

FIG. 6 is an example of a two-dimensional flow domain showing a boundary of a computational domain.

DESCRIPTION OF THE PREFERRED EMBODIMENT For the purposes of promoting an understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended. Any alterations and further modifications in the illustrated embodiment, and any further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates are also included. The illustrated embodiment of the present invention is a method applicable for solving the momentum and mass conservation equations to obtain the velocity and pressure distributions of an incompressible fluid flow in a complex domain. The method introduces a Gauss-point velocity field which prevent the checkerboard problem in pressure and enables pressure and velocity be evaluated at the every nodal points, hence the name equal order method. The method provides accurate solutions to the velocity and pressure of the fluid, such that the mass conservation of the fluid can be satisfied to the truncation limit of the computational device (machine round-off levels). An embodiment of the method is designed for incorporation in computer simulations of molten material flows. These simulations have advantages in the prediction of casting behavior, improving the efficiency of mold design, and determining the appropriate processing conditions such as temperature and injection speed control. Most importantly, an accurate mass conservation is essential to simulate the formation of shrinkage porosity. As the density increase of alloys during solidification is small, e.g. 6% for Al 357, a small error in the mass conservation could be magnified and lead to significant error to the predicted porosity. Furthermore, since pressure field is coupled with continuity equation, any deviation from mass conservation would cause large, unphysical fluctuations in the predicted pressure field. As a result, the model would not be able to predict the required injection/dwell pressure or the corresponding clamp force on the mold. The method of the present invention is generally applicable to Newtonian or Non-Newtonian fluid for steady or unsteady problem with or without moving free surfaces.

The preferred embodiments of the present invention are designed to operate within a processor based computation device. Typically, the logic routines of the present method resides on a storage medium, e.g. a hard drive, an optical disk, magnetic tapes or any state of art storage medium which is accessible by the processor. Referring now to FIG. 1, there is shown a typical processor system which is capable of carrying out the method of the invention. A conventional computer system 2 including a processor 4 which having access to read only memories (ROMs) and random access memories (RAMs) and operatively connected by cables and other means to the input and output devices. As illustrated, these input and output devices may include a display 5, a keyboard 6, a mouse 7, a printer 8, an input/output storage device 9, or any combinations thereof. Processor 4, in cooperation with an operating system, executes the logic routines of the present method, receives and stores data from the input devices, performs computations, and transmits data to the output devices. Alternatively, the executable program, the input data and the output data could be stored in the local system or in a remote system that can be accessed through network. While a conventional computer system is illustrated, other connection means and input and output devices may be substituted without departing from the invention. Specifically, the illustrated embodiment of the method operates within a Silicon Graphic system using a UNIX operating system. However, other systems having similar computation capability, memory capacity and graphic displays can also be used.

The method of the present invention is developed based on a finite-element procedure where the flow domain is subdivided into a mesh of finite elements with nodes (or grid points), where variables, e.g. velocity, pressure and front-location parameter, are defined. The finite element mesh can be one, two or three- dimensional depending on the flow domain being modeled. The term "flow domain" hereinafter is defined as the physical or theoretical boundary within which a fluid may flow, e.g., a mold. Since a flow domain may be partially filled by liquid, it would not be efficient to solve those variables in the empty region. In order to improve computational efficiency, a "computational domain" is defined as a sub-set of elements within the flow domain, where the space is filled by the liquid. Variables outside the boundary of the computational domain will not be solved. For unsteady problems, the boundary of the computational domain is adjusted in each time step as the fluid flows through the flow domain. The variable parameters are evaluated at the nodal points of the finite element mesh. These variable parameters include scalars, e.g. pressure and temperature, and vectors, e.g. velocity, in two/three-dimensional problems. Following typical finite element methodology, the variable parameter in any spatial coordinates can be interpolated using its nodal values and the so-called shape functions. For each element, a set of momentum shape functions, -VjS, and a set of mass shape functions, Mβ, may be defined, such that the functional value of a variable,/, inside the element can be expressed as:

where Fi is the value of the variable on the I^th node, ,- is either the momentum shape function Nt or the mass shape function _r- associated with the I^th node, n is the total number of nodes in the element.

For each velocity component on a nodal point, a linear algebraic equation can be derived from the momentum equation, and expressed with the momentum shape function and nodal velocity and nodal pressure variables. Likewise, the continuity equation is discretized with mass shape function and Gauss-point velocity variables. By solving these algebraic equations, the approximate solution to the Navier-Stoke equation can be obtained. The discretized equations may be derived following several conventional finite element procedures, e.g. Ritz, Galerkin, least-squares, collocation or, in general, weighted-residual finite-element methods in "An Introduction to the Finite Element Method" by J. N. Reddy, McGraw-Hill International Book Company or the control volume finite-element method in: "Control- Volume Based Finite- Element Formation of the Heat Conduction Equations", Progress in Astronomical and Aeronautical Engineering, Vol. 86, pp. 305-327, 1983). As an example, in Galerkin or control-volume finite-element method, a set of shape functions, are used to describe how a variable varies over an element. The shape function could be linear or higher order. In the preferred embodiment, the momentum equation is discretized following one of the above procedures. The mass conservation (continuity) equation is discretized with a specially defined Gauss velocity which has the same order as the pressure gradient in each element.

There are many ways to solve the resulting discretized momentum and mass-conservation equations. Any solver that could generate converged solution within an acceptable error is applicable with the proposed method. However, when the number of variables is large, e.g. simulating fluid flow and heat transfer of molten alloys in a large, complicated cavity, the computation process may become inefficient if all of the equations are solved at once directly. In the preferred embodiment, the discretized momentum equations are inserted into the continuity equation to form a Poisson pressure equation. The Poisson pressure equation can be solved either directly or iteratively. An iterative solver requires less storage and may require less computational steps than a direct solver. However, the iteration may not converge if the matrix is in ill condition. Once the new pressure field is obtained, each velocity component is updated successively from the corresponding momentum equation. The above procedure is repeated until the incremental change in pressure and velocity field is negligibly small, as the processor's truncation limit.

Referring now to FIG. 2 which shows a flowchart of an embodiment of the method of the present invention developed based on a control- volume finite- element method and where the coupled mass and momentum equation is solved iteratively. In the illustrated control- volume finite-element formulation, the Navier-Stoke momentum differential equation is integrated over each control volume. Piecewise profiles expressing the variation of the variable parameter between the nodes are used to evaluate the required integrals. The result is that the discretization equations containing the values of the variable parameters of a group of nodes. The discretized equations obtained in this manner expresses the conservation principle for the variable parameter for the finite control volume. One attractive feature of the control- volume based finite-element formulation is that the resulting solution would imply that the integral conservation of quantities such as mass, momentum, and energy can be satisfied over any group of control volumes and over the whole computation domain within the truncational limit of the computing processor.

(a) Creation of the finite element mesh, stage 101. At stage 101, the flow domain is discretized or subdivided into a mesh of finite elements having nodes (or grid points), and uniquely defined control volumes are constructed surrounding each of the nodes. The elements adjacent to each of the nodal points accommodate a portion (a sub-control volume) of the control volume and the corresponding control volume fluxing faces. Any type of element mesh may be employed, for example, a regular grid similar to a finite difference method or an irregular mesh created according to a typical finite element method. An irregular mesh provides more flexibility in fitting irregular domains and in providing local grid refinement. The proposed Equal-Order method is applicable to either moving or fixed finite element meshes. A moving mesh method could generate more accurate prediction on the location of moving free surface.

However, moving mesh method often suffers from mesh distortion and heavier computation due to limitation in time step. A fixed mesh is preferred in the present invention mainly for its flexibility to handle free surface flow in complex domain. The finite element mesh may be created by any conventional procedure. Such suitable procedures have been disclosed, for example, in U.S. Patent No. 5,010,501 issued on April 24, 1991 to Arakawa and entitled "Three-Dimensional Geometry Processing Method and Apparatus Therefor", and in U.S. Patent 5,553,206 issued on September 3, 1996 to Meshkat and entitled "Method and System for Producing Mesh Representations of Objects".

For modeling a two-dimensional problem, where a two-dimensional finite element mesh would be required, a mesh of triangular finite elements similar to that shown in FIG. 3 and 4 is preferred, again for its geometric flexibility. However, a quadra-lateral element is also applicable. Referring now to FIG. 3, the triangle finite element 10 has three vertex nodes 12. The variable parameters of the fluid, i.e. temperature, velocity and pressures are evaluated at nodal (grid) points 12 of element 10. Particular of this invention, a Gauss-point velocity, V_g and a

Gauss-point pressure gradient VP_g are defined inside each element so that they have the same order and are computed from the variables at the nodal points 12.

Finite element 10 can be divided into three sub-control volumes 18 by joining the mid-side points 14 to the centroid 16. Each sub-control volumes 18 is bounded by four fluxing surfaces 30 where material can flow from one sub-control volume to another. Those sub-control volumes 18 surrounding a vertex node 12 form the control volume for that node. For example, as shown in FIG. 4, vertex node 22 is surrounded by six sub-control volumes 23-28 and the shaded area represents control volume 20 of vertex node 22. While it is illustrated that control volume 20 is composed of six sub-control volumes 23-28, it is possible that only some of the sub-control volumes are filled or partially filled by liquid. Therefore, only those filled and partially filled control volumes are included in the computational domain. For a given mesh, defining the filling status in each SCV can provide more flexibility and better resolution to the predicted moving free surfaces than defining filling status only in the whole control volumes. However, defining filling status in the whole control volume simplifies the program's structure.

For modeling a three-dimensional problem, a three dimensional finite element mesh would be required. Tetrahedral element is preferred over hexahedral elements for the convenience of available automatic tetrahedral mesher. As shown in FIG. 5, a linear tetrahedral element 40 has four vertex nodes 42 and four surfaces. Finite element 40 can be divided into four sub-control volumes 48 by joining the surface centroid 44 (centroid at each surface of the tetrahedral element 40) to the volume centroid 46 (the centroid of the tetrahedron). Each of the sub- control volumes 48 is a hexahedron having six fluxing surfaces. While triangular and tetrahedral finite elements are illustrated and preferred, it should be understood that other geometric shaped finite elements or a combination thereof might be used. In addition, linear finite elements are illustrated for its simplicity. It is, however, contemplated that the finite elements may be of any order. Preferably, a computational domain is defined to exclude unneeded variables and to reduce computational time. The boundary of the computational domain may be drawn according to Wang, et al., as disclosed in U.S. Patent No. 5,572,434; and the disclosure of which is expressly incorporated herein by reference in its entirety. Wang draws the boundary of a computational domain according to the filling status of the element in the flow domain. Wang defines, for a flow domain of triangular elements as shown in FIG. 6, that an empty element 62 is an element having all of its sub-control volume empty; a filled element 64 is an element having all of its sub-control volumes filled, and a partially filled elements 66 is an element having at least one, but less than all of its sub-control volume filled. Further, Wang designates the nodes of the flow domain as interior or exterior nodes. If a node is surrounded by only filled sub-control volumes, it is designated as an "interior node" 67. All other nodes are "exterior nodes." These exterior nodes are further divided into several groups: exterior nodes associated with filled and partially filled sub-control volumes are known as "free surface nodes" 68, those immediately adjacent to free surface nodes, but not themselves associated with a filled sub-control volume are called "driven nodes" 69. The boundary 70 of a computational domain is defined as the line linking those exterior nodes which are associated with empty sub-control volumes and adjoining driven nodes. While Wang's method of defining a computational domain is illustrated, any other methods of defining a computational domain which allows a meaningful solution to be obtained may also be used without deviating from the scope of the invention.

(b) Inputing initial parameters, stage 102.

At stage 102, the finite element mesh, control volume's filling status, the properties of the fluid, the initial values of variable parameters, e.g., temperature, pressure, and velocities, and the boundary constraints of these variables are input. These values of the variables will be updated as the iterative computation progresses.

(c) Deriving the discretized momentum equations, stage 103.

The general equations governing the conservation of mass and momentum of the fluid flow can be expressed by equations (1) - (4)

Conservation of mass:

(1)

Conservation of momentum:

du ,_{τ π} . dP

P-—+ P(V • Vn) - --— + pgχ+ V(μVw) στ ox

(2)

P ~+ P(V • Vv) = -~+ Pgy + V( Vv) ότ dy

(3) 9w „, „ N dP + pgz + V(μVw) aτ όz

(4)

where:

X,y, Z = Cartesian coordinate variables

U, V, W = magnitude of the velocity vector V in the

X-,y-, andz-directions, respectively,

t = time p - density of fluid

P = pressure

V = flow velocity vector μ = viscosity

Sx' £y_> Sz ⁼ magnitude of gravity in the X-, y-, and z-directions, respectively.

The momentum equations (2) - (4) above describe the fundamental conservation law of incompressible viscous flow. Equations (2) - (4) are the summation of several terms each representing a physical phenomenon pertaining to fluid flow. Examples of these terms include inertia, advection, diffusion, body force and pressure gradient. Generally, the terms pertinent to the problem being solved are selected and the remaining terms are grouped and collectively defined as the source term. Using a control volume-finite element formulation, an elemental matrix can be computed to approximate the terms of interest. These matrices are then assembled into a global momentum matrix (a set of discretization momentum equations) which can be solved, iteratively, explicity or implicity or the combination thereof, to approximate a solution to the Navier-Stoke equation.

The illustrated embodiment of the method follows the methodology presented by Schneider and Zedan for assembling the global momentum matrix. ("Control- Volume-Based Finite Element Formulation of the Heat Conduction

Equations", Progress in Astronomical and Aeronautical Engineering, Vol. 86, pp. 305-327, 1983). A "linear shape function" or profile assumption is used to describe how a dependent variable varies over an element. The governing equation is integrated over the control volume and includes contributions from each sub- control volume to the integral conservation for the control volume.

While Schneider and Zedan' s methodology is illustrated, other control- volume based procedure or Galerkin method may be used.

In the illustrated embodiment, the terms for advection, diffusion, and pressure gradient are selected. After the assembly process, each momentum equation left hand side can be represented by a coefficient matrix operating on a velocity component. On the right hand side, there are the source terms and the pressure gradient terms.

Mathematically, if the integrand is considered constant, then equation (5) results as follows:

(5)

For the advection terms, a streamline upwind operator is used, as represented by the following equation (6):

(6) where 0 is any dependent variable, S is distance along a streamline, subscript V denotes volume and subscript S denotes the streamline function.

For the diffusion terms, Green's theorem is used to convert the volume integral to an integration of fluxes through the bounding surface, as modeled by equation (7):

jV VφdV

= j j * άω v σ σ

(7)

where ϋ, V denote surface and volume integral, respectively, and 0 is any dependent variable, J is the flux contribution and CO is the area. The flux contribution for a facej from a node i, is expressed by equation (8) as follows:

(8) where:

Λ^- is the shape function, #n — Λ₃₃ are the cofactors of the Jacobian matrix, and 0)_xj, CO_yj, CO_y are the components of the area CO, and a is diffusivity. After the assembly process, each term at the left-hand side of the momentum equations is represented by a coefficient matrix operating on a velocity component. Considering a two-dimensional case for simplicity, the assembled global momentum equations can be modeled by equations (9) and (10) as follows:

∑ a ij u j = - J dA

A a x

(9)

(10)

where the ay terms make up the influence coefficient matrix and n is the total number of fluxing faces of the neighboring sub-control volumes. Pulling the pressure gradients outside the area integrals, equations (9) and (10) is recast as equations (11) - (15) as follows:

d P

U t = U i - K d x

(11)

and

d P

V ,^• = - K d y

(12)

where,

K i = — f dA , a a ^J

(13) Λ u a. ij u j , a a ι ≠ J

(14)

Λ

V = - aijVj i ≠ J

(15)

The nodal hat velocities, ύ. and ₍. are composed of nodal velocities U_j and V_j of neighboring sub-control volumes, and contain no pressure component.

It should be understood that while the illustrated embodiment of the present invention applies to a two-dimensional flow with linear elements, the method is applicable to lower and higher dimensional flows with higher order elements.

(d) Compute nodal hat velocities ύ_i , V. , and K_f stage 104.

At stage 104, the coefficient matrix ay is first calculated from the estimated velocity field provided initially. Using these newly calculated values of the coefficient matrix and the nodal velocity values, U_j and V , interpreted from the neighboring sub-control volumes, the nodal hat velocities, U j , ,-, and Ki are computed following equations 13-15.

(e) Compute the Gauss-point hat velocities and Gauss-point coefficient, 105

At stage 105, the present invention defines a Gauss-point velocity vector and its associated coefficient as element average quantities. The Gauss-point hat velocity U _g, V _g and the Gauss-point coefficient term K_g are the average value of their nodal counterparts, as provided by the following equations (16)-(18): 1 u g — 2_J u j ,

H e ; = ι

(16)

(17) and

K _g = ∑ K j n j = l

(18)

where n_e is the number of fluxing faces within the element. The Gauss-point velocities can be expressed as: d p u _g = u K x

(19)

v _g = v g - K d y

(20)

In the above two-dimensional expression for the Gauss-point velocities U_g and ^, the pressure gradients όP/σX, oP/by are constant over the fluxing surface. For consistency, the terms U_g, V_g and K_g have to be constant over the fluxing surface as well. Traditionally, the interpolation of the Gauss-point terms would have been computed using the element shape functions, V_f. However, computations involving shape functions would give Gauss-point velocities a linear variation across the element, while the pressure gradient is constant over the element. This inconsistency will result in computational schemes that can not conserve mass to machine round-off levels and lead to checker-board pressure.

(f) Form and solve the Poisson pressure equation, stage 106.

At stage 106, by inserting the expression for the Gauss-point fluxing velocities U_g and V_g into the continuity equation (21),

u + = 0 y

(21) a Poisson equation for pressure is obtained. The Poisson pressure equation (22) is expressed as:

(22)

Integrating this equation over an area A, the fluxing surface of the element, the pressure field can be expressed as:

(23)

Using Green's theorem, these area integrals are converted to line integrals, as represented by the following equation (24):

(24)

where σ is the line path. Using the value of the Gauss-point hat velocities and Gauss-point coefficient, ύ_g, V _g, and k_f^ previously calculated at stage 105, the Poisson pressure equation can be solved to yield the pressure gradients at each element. While it is illustrated that the area integrals are converted into line integrals before the pressure field expression (23) is solved, other methods which employ different mathematical manipulations, may be used. For example, a Galerkin formulation which solves the original area integrals can be used. Furthermore, the momentum equation and mass conservation equation can be combined into a set of linear matrix equation and solved by using direct solver, e.g. matrix-inverse method, or iterative solver, e.g. conjugate-gradient method. .

(g) Update velocity field, stage 107. The nodal velocities, M,- and v,-, and the Gauss-point velocities, U_g and V_g, can be computed by inserting the newly derived pressure field and the values M,-,

V i, Kt, ύ_g, V_g, and .K_g into equations 11, 12, 19 and 20.

Once the pressure and velocity field is updated, other variables, e.g. the filling status of each SCV's and temperature which influence the rheological and thermal physical properties of the fluid, are determined using the update velocity field. For those variables which do not influence the flow field, their value is preferably determined after the flow field has converged.

(h) Check for the velocity converging, stage 108. An iterative process is said to have converged when further iterations will not improve the accuracy of the dependent variables. In practice, the iterative process is terminated when the desired accuracy is obtained. As the momentum equations are non-linear and the fluids properties may change with the velocity field and temperature, iterations are required until the incremental change is negligibly small. In this illustrative embodiment of the present method, the flow equations are iteratively solved until all velocity values converge to a user supplied tolerance. While the above convergence criteria was chosen for the illustrated embodiment, other convergence criteria, e.g. range or threshold; may be used without deviating from the spirit of the present invention.

At this stage, the method is completed if the thermal flow field is steady. Otherwise, the method returns to stage 103 to form the set of equations using the updated variables for the following time step.

Numerical tests have revealed that, typically, the velocity field converges rapidly for highly viscous flows during the iterations. However, for fluids having lower Reynolds numbers solving the coupled mass and momentum equations simultaneously may aid in obtaining a more rapid convergence.

In addition, it is observed that for making the conservation of mass and momentum equations consistent at the element level, the Gauss-point fluxing velocity field does indeed conserve mass to machine round-off levels. Furthermore, pressure checker-boarding is eliminated. It should be understood that using Gauss-point velocity and pressure fields to solve the coupled mass and momentum equation is not limited to the methodology illustrated. The invention is applicable to other discretized forms of the momentum equations derived by other weighted-residual formulation, e.g. Galerkin' s method. In addition, the present invention is also applicable to methods which solve the equation implicitly, e.g. the penalty function method.

While the invention has been illustrated and described in detail in the drawings and foregoing description, the same is to be considered as illustrative and not restrictive in character, it being understood that only the preferred embodiment has been shown and described and that all changes and modifications that come within the spirit of the invention are desired to be protected.

Claims

What is claimed is:

1. A method of solving the governing equations of an incompressible viscous flow in a flow domain comprising: providing a discretized mesh of said flow domain of said fluid flow, wherein said mesh comprises elements having nodes; deriving a discretized momentum equilibrium equation based on a finite- element formulation, wherein pressure and velocity are defined at every nodes of the elements of said mesh; deriving a discretized mass conservation equation based onsaid finite- element formulation, wherein pressure gradient is defined over each element and velocity is defined on a Gaussian-points inside each element, thereby said Gaussian-point velocity having the same order as said pressure gradient;determining velocity and pressure at each nodes of said flow domain; repeating said deriving and said determining until a convergence test of said velocity values is satisfied.

2. The method of claim 1, wherein for each of said elements, the Gaussian-point velocity is a function of said nodal velocities.

3. The method of claim 2 wherein said determining comprises inserting said discretized momentum equation into said discretized mass conservation equation thereby forming a Poisson pressure equation; and solving said Poisson pressure equation for a pressure field.

4. The method of claim 3 wherein said Poisson pressure equation is solved iteratively.

5. The method of claim 3 wherein said Poisson pressure equation is solved simultaneously.

6. The method of claim 1, wherein said finite element formulation is a control- volume based finite element procedure.

7. The method of claim 1, wherein said finite element formulation is a weighted-residual finite-element method.

8. The method of claim 1, wherein said determining further comprises defining a computational domain as a sub-set of elements of said flow domain, and determining said pressure and said velocity only on nodes of elements within said computational domain.

9. The method of claim 1, wherein said mesh is a fixed finite element mesh.

10. The method of claim 1, wherein said flow domain is two- dimensional, and wherein said elements are triangular finite elements.

11. The method of claim 10, wherein said finite elements are 3-node linear triangular elements.

12. The method of claim 11, wherein said Gaussian-point velocity is an average of the said velocities defined at said nodes of same said element.

13. The method of claim 1, wherein said flow domain is three- dimensional, and wherein said elements are tetrahedral finite elements.

14. The method of claim 13, wherein said finite-elements are 4-node linear tetrahedral finite elements.

15. The method of claim 14, wherein said Gaussian-point velocity is average of four said velocities defined at said nodes of same said element.

16. The method of claim 1, wherein said elements are finite elements, having linear shape functions.

17. The method of claim 1, wherein said elements are finite elements having higher than 1^st order (linear) shape functions.

18. The method of claim 1, wherein said flow domain of is two- dimensional, and wherein said elements are quadrilateral finite elements.

19. The method of claim 18, wherein said finite-elements are 4-node linear quadrilateral elements.

20. The method of claim 19, wherein said Gaussian-point velocity is an average of said four velocities defined at the nodes of same said element.

21. The method of claim 1, wherein said flow domain is three- dimensional, and wherein said elements are hexahedral finite elements.

22. The method of claim 21, wherein said finite-elements are 8-node linear hexahedral finite elements.

23. The method of claim 22, wherein said Gaussian-point velocity is an average of said eight velocities defined at the nodes of same said element.

24. A method operating within a processor environment for simulating incompressible viscous fluid flows comprising: inputting material properties of said fluid and boundary conditions; discretizing a flow domain of said fluid flow into a mesh of finite elements; each elements having nodes; defining pressure and velocity on said nodes; deriving a discretized momentum equation based on said velocities and pressures defined on the nodes; defining a Gaussian-point velocity inside each of said elements, wherein said Gaussian-point velocity have same order as pressures gradient in said element and is a function of said velocities defined on said nodes; deriving a discretized continuity equation based on said Gaussian-point velocity; determining values of pressure and velocity at each said nodes of said finite element; and repeating said deriving and said determining until a convergence test of said velocity values is satisfied.

25. The method of claim 24 wherein said determining further comprises couphng said discretized momentum equation and said discretized continuity equation to form a Poisson pressure equation, and solving said Poisson pressure equation for a pressure field.

26. The method of claim 25, wherein said Poisson pressure equation is solved iteratively.

27. The method of claim 24, wherein said discretized equations are derived following a control- volume based finite-element formulation.

28. The method of claim 24, wherein said discretized equations are derived following a weighted-residual based finite-element formulation.

29. The method of claim 18, wherein said finite element mesh is a fixed mesh.

30. The method of claim 29, wherein said simulation is based upon a

Navier-Stokes fluid model.

31. The method of claim 24, wherein said finite-element formulation having linear shape functions.

32. The method of claim 24, wherein said determining further comprises defining a computational domain as a sub-set of finite elements of said flow domain, and wherein said pressure and velocity are only determined on nodes of said sub-set of finite elements of said computational domain.

33. The method of claim 23, wherein said flow domain is two- dimensional, and wherein said elements are triangular finite elements.

34. The method of claim 33, where said finite-elements are 3-node linear triangular elements.

35. The method of claim 34, wherein said Gaussian-point velocity an the average of three side velocities defined at said nodes of same said finite element.

36. The method of claim 24, wherein said flow domain is three- dimensional, and wherein said elements are tetrahedral finite elements.

37. The method of claim 36, wherein said finite-elements are 4-node linear tetrahedral finite elements.

38. The method of claim 36, wherein said Gaussian-point velocity is an average of four said velocities defined at the nodes of same said elements.

39. A device operable within a processor environment for modeling a fluid flow comprising: a medium adapted to receive and store logic executable within a processor to perform a method as disclosed in claims 1 to 23.

40. The device of claim 39, wherein said device is an input/output storage device.

41. The device of claim 40, wherein said processor environment is a conventional computer system.

42. The device of claim 41, wherein said processor environment is a remote network system accessible from local systems.

43. A numerical method operating within a processor environment for solving the momentum equation of a viscous, incompressible, laminar flow within a flow domain comprising:

(a) discretizing said flow domain into a mesh of finite elements having nodes and fluxing surfaces;

(b) forming momentum equations, wherein velocity and pressure are defined on each nodes of said finite elements; (c) defining a Gaussian-point velocity inside each of said finite elements, wherein said Gaussian-point velocity is computed from said velocities defined at the nodes of each same said finite elements;

(d) updating velocity; ;

(e) repeating said forming, computing, calculating and updating until a convergence test of said velocity is satisfied.

44. The method of claim 43, wherein said mesh is composed of finite elements.

45. The method of claim 44, wherein said mesh is a fixed mesh.

46. The method of claim 45, wherein said finite elements are triangular elements.

47. The method of claim 45, wherein said finite elements are tetrahedral elements.

48. The method of claim 46, wherein said forming of momentum equations comprises deriving discretized momentum equations based on a control volume based finite element formulation, wherein said discretized momentum equations are represented as follows:

49. The method of claim 48, wherein said forming momentum equations further comprises recasting said discretized momentum equations as follows: d P d P_

U i — U i — K i and V i = V i K ι dx d y

where,

K ^{a i}J J

50. The method of claim 49, wherein said Gaussian-point velocities, U_g, V_g, are expressed as:

Ug = U -g - K τrg — ^d —^P an Ad v_g = v₈ - K τri ^{d P} dx dy and wherein said Gaussian-point velocities are element average quantities, such that:

51. The method of claim 50, wherein said updating velocity further comprises deriving discretized mass conservation equations based on said Gaussian-point velocity.

52. The method of claim 51 , wherein said updating velocity further comprises coupling said discretized momentum equations and said discretized mass conservation equation to form a Poisson pressure equation as follows:

V k V p + x y

and, integrating said Poisson equation over areas of said fluxing surfaces of said finite elements as:

and converting area integrals via Green's theorem to line integrals as:

and solving said Poisson equation iteratively.