SAE TECHNICAL PAPER SERIES

961195

3D Vortex Simulation of Intake Flow in a Port-Cylinder with a Valve Seat and a Moving Piston

Adrin Gharakhani and Ahmed. F. Ghoniem

Massachusetts Institute of Technology

Reprinted from: Modeling Techniques in SI and CI Engines (SP-1178)

The Engineering Society For Advancing Mobility Land Sea Air and Space

INTERNATIONAL

International Spring Fuels & Lubricants Meeting Dearborn, Michigan May 6-8, 1996

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (412)776-4841 Fax:(412)776-5760

The appearance of the ISSN code at the bottom of this page indicates SAE's consent that copies of the paper may be made for personal or internal use of specific clients. This consent is given on the condition however, that the copier pay a $7.00 per article copy fee through the Copyright Clearance Center, Inc. Operations Center, 222 Rosewood Drive., MA 01923 for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. SAE routinely stocks printed papers for a period of three years following date of publication. Direct your orders to SAE Customer Sales and Satisfaction Department. Quantity reprint rates can be obtained from the Customer Sales and Satisfaction Department. To request permission to reprint a technical paper or permission to use copyrighted SAE publications in other works, contact the SAE Publications Group.

GLOBAL MOBILITY DATABASE All SAE papers, standards, and selected books are abstracted and indexed in the Global Mobility Database.

No part of this publication may by reproduced in any form, in an electronic retrieval system or otherwise, without the prior written permission of the publisher.

ISSN 0148-7191 Copyright 1996 Society of Automotive Engineers, Inc. Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in SAE transactions. For permission to publish this paper in full or in part, contact the SAE Publications Group. Persons wishing to submit papers to be considered for presentation or publication through SAE should send the manuscript or a 300 word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.

Printed in USA

96-0049

961195 3D Vortex Simulation of Intake Flow in a Port-Cylinder with a Valve Seat and a Moving Piston
Adrin Gharakhani and Ahmed. F. Ghoniem
Massachusetts Institute of Technology
Copyright 1996 Society of Automotive Engineers, Inc.

ABSTRACT
A Lagrangian random vortex-boundary element method has been developed for the simulation of unsteady incompressible flow inside three-dimensional domains with time-dependent boundaries, similar to IC engines. The solution method is entirely grid-free in the fluid domain and eliminates the difficult task of volumetric meshing of the complex engine geometry. Furthermore, due to the Lagrangian evaluation of the convective processes, numerical viscosity is virtually removed; thus permitting the direct simulation of flow at high Reynolds numbers. In this paper, a brief description of the numerical methodology is given, followed by an example of induction flow in an off-centered port-cylinder assembly with a harmonically driven piston and a valve seat situated directly below the port. The predicted flow is shown to resemble the flow visualization results of a laboratory experiment, despite the crude approximation used to represent the geometry.

INTRODUCTION
Computational fluid dynamics is fast becoming a serious alternative to the empirical cut-and-try approach in the design and analysis of IC engines. Numerous newlydeveloped methodologies and their applications have appeared in the literature, ranging from the 2-D/ axisymmetric flow in port-cylinder configurations [1-5], with and without a poppet valve, to the more realistic three-dimensional flow in the intake ports and the cylinder of the engine, including moving valves [6-9]. However, despite the enormous progress achieved so far, several computational and physical modeling difficulties remain unresolved. In grid-based computational methods, the accurate, efficient and automatic volumetric meshing of typical timevarying engine configurations remains a challenging task. To obtain a well-behaved solution, the mesh generation algorithm must resolve the widely varying geometric length scales in the engine accurately. In addition, it must be capable of capturing the high velocity gradients within the thin concentrated jets around the intake valve, as well as the large scale turbulent flow structures in the cylinder. Moreover, due to the highly convoluted nature of the flow, care must be exercised in order not to create degenerate 179

meshes. The most adverse consequence of a poor-quality meshing is the introduction of false diffusion into the flow field which, for the engine problem, leads to weaker large eddy vortical structures and faster than expected decay rates. False diffusion may be reduced by increasing the mesh resolution and/or applying adaptive gridding technology, both of which can become computationally too expensive in a complex engine geometry. Alternatively, false diffusion may be minimized by aligning the streamwise side of the finite volume along the local streamline. This can best be achieved using a Lagrangian motion of the mesh nodes. However, given the complexity of engine flows, this approach may eventually lead to degenerate finite volumes. In this paper, an alternative approach is presented based on the random vortex-boundary element method for the simulation of unsteady flow in 3-D geometries with moving boundaries of the type encountered in engines. The Navier-Stokes equations are expressed in the vorticity transport formulation, and are discretized using a collection of spherical vortex elements. The element velocity is expressed as a superposition of a vortical component evaluated by the Biot-Savart law, and a potential component obtained from the solution of a 3-D Neumann problem over the domain. The convection and stretch of vorticity are evaluated in the Lagrangian frame of reference (without the need for grids) and its diffusion is described stochastically by the random walk method. The boundary element method is used to solve a 3-D Laplace equation that defines the potential flow and imposes the normal flux boundary condition on the boundary surfaces without having to discretize the domain interior. The noslip boundary condition is satisfied by generating vorticity tiles at the boundaries. Within a thin prespecified region near the boundary, the tiles convect and diffuse in the Lagrangian frame of reference according to the Prandtl equations. Beyond this region, the tiles are converted to spherical vortex elements. The Lagrangian vortex-boundary element method is grid-free in the fluid domain. As a result, the complex task of volumetric meshing of the interior of an engine is reduced to the far simpler task of meshing its surfaces. The method is self-adaptive and capable of dynamically concentrating computational elements in regions with significant velocity

gradients, such as in shear layers and jets. Most importantly, due to the Lagrangian evaluation of convection, artificial diffusion is minimized (theoretically, it is zero.) This makes the scheme an excellent candidate for the direct simulation of flow at high Reynolds numbers; it also enables one to ascertain the correctness of subgrid scale viscosity or other turbulence models with more confidence. The potential advantages of the vortex-boundary element method have been demonstrated using a simulation of the unsteady flow over an impulsively started cubic bluff body [10], and flow inside a simplified cylinder connected to an off-centered port and driven by a harmonically moving piston [11]. A massively parallel version of the method has also been completed to demonstrate its inherent parallel advantages [12]. In this paper, we present a brief description of the 3-D vortex-boundary element method; the detailed formulation and the parametric study of its accuracy are given in [1315]. We also show results from the simulation of induction flow caused by a harmonically driven piston inside a cylinder and an off-centered port, containing a valve seat. At present, the boundary surface may be approximated using plane rectangular boundary elements and tiles only, analogous to the staircase discretization in 2-D. The extension to the body-fitted discretization of the boundary is under development. To reduce the computational cost, the Reynolds number based on the piston diameter and its maximum speed was set at 350. The choice for the low Reynolds number was also based on a parametric study, using the visualization results of flow in an axisymmetric engine with a poppet valve [16], which concluded that the large scale features of the vortical structures during induction are quite independent of the Reynolds number. In addition, in an unpublished work using an axisymmetric random vortex-finite element simulation of flow in the same geometry, we have observed similar results at lower Reynolds numbers of 350 and 10,000. The objectives of the present simplified engine flow simulation are (1) to demonstrate the potential advantages of using an internally grid-free simulation technique, and (2) to explore, qualitatively, the initiation and growth of large unsteady vortical structures in the cylinder. The simulation is shown to closely reproduce the main characteristics of a similar flow visualization experiment, despite differences in geometry and flow conditions between the two cases.

normalized by L/U; Re = UL/v is the Reynolds number, and v is the kinematic viscosity. (x, t) = (x, y, z) = ^u is the vorticity vector and ^ represents the vector product. On the boundary, the velocity is expressed in terms of the local orthogonal coordinate system, --n, where = (x, y, z,) and = (x, y, z) are the unit tangents to the boundary, and n = (nx, ny, nz) is the unit outward normal to it. The vorticity field in Eq. (1) is discretized using a collection of Nv vortex elements; each centered at xj, with element volume Vj and vorticity vector j:

1 x where g (x) = ----- g ----- is a smoothed delta function having

3

(t) = (t)V is the a spherical core with radius , and j j j volumetric vorticity. We used the second order core 3 function g(x) = -----[1 - tanh2( x 3)], selected from an 4

available list [17]. In Eqs. (1), the velocity and its gradients at a point are evaluated as a superposition of a vortical field, u, in free space and a potential flow, uP, inside the domain, such that continuity and the normal flux boundary condition are satisfied. Given the smooth vorticity distribution (2) and using the Biot-Savart law, the discrete vortical velocity and its gradients at the vortex element centers are evaluated as follows:

FORMULATION
THE INTERIOR - The Navier-Stokes equations for incompressible flow inside a three-dimensional domain, D, and with boundary surfaces, #D, are expressed in the following vorticity transport form:

x K ( 0 ) = ------------ The accuracy and convergence properties 3 4

of the vortex element method are found in [18-23]. The potential component of the velocity and its gradients at a point are obtained by solving the following 3D Neumann problem for the interior: 2(x) = 0 q(x0) = - uP(x0).n = (u(x0) - u(x0)).n xD (4a) x0 D (4b)

where x = (x, y, z) is the position vector normalized by a reference length, L; u(x, t) = (u, v, w) is the velocity vector normalized by a characteristic speed, U; t is the time 180

where is the potential distribution, the gradient of which yields the potential velocity, and q is the normal flux. A piecewise linear variation of the boundary potential and its normal flux is assumed over a set of M boundary elements discretizing the boundary surface. To preserve the grid free character of the vortical flow solution, the direct boundary

element method is used to obtain the potential flow. The details of the accurate solution of this problem are given elsewhere [13, 15]. Having obtained the potential distribution on the boundary, the potential velocity and its gradients at the vortex element locations are evaluated using the following regularized formulations [13, 24]:

THE BOUNDARIES - In what follows, the processes of vorticity generation on a rectangular wall and its evolution within a thin region near the wall are presented. (The approach is applied to all walls simultaneously.) All variables are defined with respect to the local coordinate system, z is selected normal to the wall and into the flow interior, and z = 0 represents the wall surface. The application of the normal flux boundary condition (4b) to the flow field yields a tangential velocity on the domain boundary that is different from the no-slip boundary condition. This velocity jump on the wall is equal to the amount of surface vorticity that must be generated to satisfy the no-slip condition. We discretize the jump distribution on the wall using a set of rectangular vortex tiles - each with sides h xi and h yi , and surface vorticity
(xi, yi, 0, t) at its center, (xi, yi, 0). The tile vorticity is linked to the velocity jumps at the center of the boundary elements, and is obtained by summing the area-weighted jump contributions from all boundary elements that are shadowed by the vortex tile: where MB is the number of boundary elements on a wall
t t

segment. Each element is defined by its sides h xm and

1 where G(x0, xi) = ------------------------- , j, l, m and p indices indicate 4 x0 xr h ym and an area-averaged velocity jump ukm at the
b

direction with respect to the global coordinate system and follow the Einstein rule; (.), represents differentiation with respect to x0 in the r-th direction, jp is the Kronecker delta and jpl is the antisymmetric permutation tensor. Sk(x0) is the surface area of the k-th element. The first and second degree tangential derivatives of the boundary potential and flux distributions are assigned piecewise linear variation across the elements. See [15] for details. Applying viscous splitting algorithm to the Lagrangian formulation of the Navier-Stokes equations [25], and using the vorticity element distribution (2) yields the following set of ordinary differential equations for the grid-free prescription of the vorticity field evolution:

center, (xm, ym, 0). The area-weighted velocity jump contributions from the boundary elements to the tiles are
xi xi xj xj given by (xi, xj, h xi , h xj ) = ------------------------ - Max ------------------------, 2 2 t b xi xj Min( x i x j , ------------------------ )))/ h xi (Indices 1 and 2 used with the

(h + h )

(h + h )

(h + h ) 2

surface vorticity and the velocity jump denote the x and y directions, respectively.) To maintain a finer discretization of the flow in the direction normal to the wall, the tiles are split into Ns.i = ( (xi, yi, 0, t)l/max + 0.5) stacks of tiles (xi, yi, 0, t) = (xi, yi, each with surface vorticity equal to 0, t)/Ns.i [27-30]. max is a user-specified maximum surface vorticity. The evolution of the wall-generated vortex tiles within a thin region near the boundary Dh is assumed to be locally two-dimensional and is approximated by the Prandtl equations [31]:

where i describes the trajectory of the i-th vortex element i,o, t is the time it is initially at i,o and with vorticity 0 introduced into the domain, t = (tk+1 - tk) is the integration timestep, and F[.] represents the integration scheme. The second order modified Euler method was used for the present study. Eq. (6c) is the stochastic approximation for the diffusion process, and i = (x, y, z)i are random variables in each of the three coordinate directions, selected independently from a Gaussian distribution with zero mean and variance equal to 2t/Re. Note that the random walk method is not a model for diffusion, but an "exact" solution. See [26] for details of the method and the parametric tests of accuracy and convergence. 181

v u where (x, t) = (x, y, z) -----, -----, 0 . (U, V, 0) is z z

the velocity at the edge of the boundary layer as seen by the wall, and is obtained from the Navier-Stokes solution in the interior. b = BLTC 2 t Re is a user-specified boundary

thickness, within which the Prandtl approximation is applied. Note that b is unrelated to the physical boundary layer thickness, and remains constant in space and time during the simulation. BLTC is assigned in the range 1.0 3.0 such that the tiles will jump into the flow interior in a few timesteps with relatively high probability [4, 27]. Within layer b, the discretized u and v velocity components at the tile centers (xi, yi, zi) are derived by directly integrating the approximate definitions for y and x, respectively, and applying velocity boundary conditions (8d) and (8e):

signifying the end of the computational timestep. If a tile jumps into the flow interior, it is converted into a vortex
t (x, t) = (x, t) h t element with volumetric vorticity x h y and

core radius = Max( h x h x ). This sets the beginning of the next timestep. See [15] for the details of the vortexboundary element method.

RESULTS
Results from the simulation of intake flow in an offcentered port-cylinder assembly with a valve seat and a harmonically moving piston are presented herein. The circular cross-section of the engine was approximated using a staircase idealization, as depicted by the schematic diagram in Fig. 1. The flow was generated by the harmonic displacement of the piston from rest at the top dead center (TDC) to rest at the bottom dead center (BDC) positions. A uniform velocity profile was assigned at the inlet of the port and its value obtained by applying the no-flux boundary condition to the remaining boundary surfaces and satisfying the continuity requirement. The no-slip boundary condition was applied to all solid boundaries. All length scales were normalized by the cylinder diameter, Dc, and their values declared in Fig. 1. The velocity was normalized by the maximum piston speed at 90 crank angle, Vm. The Reynolds number based on Dc and Vm was set at 350. The induction process was discretized by 200 equal timesteps, corresponding to a 0.9 crank angle. For the present study, the simulation was stopped at 150 timesteps or equivalently at 135 crank angle. Additionally, max = 0.5 and BLTC = 1.5 were set. Time-varying boundary elements were used to discretize the cylinder wall along its length, Lc, which changes in time due to the piston motion. For this purpose, a maximum number of boundary elements, Nmax, was used to discretize the maximum length of the cylinder wall, Lmax, corresponding to the piston position at BDC. The instantaneous number of elements, N, was obtained using the relation N = Max (1, NmaxLc/Lmax + 0.5). The same procedure was repeated for the vortex tiles. In addition to saving CPU time, this approach maintains a nominal elemental mesh size in the order of Lmax/Nmax; which, for the present study, was set to 0.05 and 0.1 for the boundary elements and tiles, respectively. In contrast, selecting a fixed value for N would yield extremely thin elements in the direction of piston motion at the beginning of the computation, and would lead to inaccurate flow predictions. For all other surfaces, the boundary element and tile sizes were fixed at 0.05 and 0.1, respectively. 1,392 boundary elements were used to discretize the domain at TDC, which increased to 2,512 at 135 crank angle. The velocity jump discretization required 404 and 740 tiles, respectively. The number of vortex elements reached 23,191 at the end of the simulation. It is interesting to note that a simulation of flow in the same port-cylinder geometry and using an identical set of computational parameters, but without a valve seat, led to the generation of only 9,095 vortex elements at 135 crank angle. This is the first indication that the majority of vortex elements are generated within the shear layers around the intake 182

where l(xi, yi) = (xi, xl, h xi , h xl ))(yi, yl, h yi , h yl ), NT is the total number of vortex tiles, s is the Heaviside step function, and (U 1., U2.) (U, V). The w component is obtained by satisfying continuity:
t t t t

where the divided difference approximate the derivatives, and

rule

is

applied

to

The grid-free evaluation of (8a), using a first order Euler time integration, is given as follows:

where ( xi, yi, zi) denote the tile trajectory in this context. Note that Eq. (12b) simulates diffusion normal to the wall and into the domain interior by reflecting tiles that jump below the wall back into the field [30]. During each timestep, vortex elements convect, stretch and diffuse in the interior. An element is eliminated, if it leaves the domain or jumps into the boundary region. (The trajectory of the elements is such that the normal flux boundary condition is imposed on the flow.) The tiles in the boundary domain are convected concurrently with the evolution of vorticity in the interior. New tiles are then created on the solid walls to satisfy the no-slip boundary condition. Finally, all tiles are diffused normal to the wall -

jet at the port boundary and the seat circumference. The presence of a seat increases the jet velocity and hence the shear on both sides of the jet.

Figure 1: Schematic diagram of the port-cylinder geometry. Integers refer to units of length 0.05. Fig. 2 depicts the side view of the trajectory of the vortex elements, and their velocity, within a slice of volume with 0.15 thickness and plane of symmetry lying on the plane of symmetry of the engine - region B in Fig. 1. Fig. 3 depicts a similar visualization within region A in Fig. 1. The velocity vectors are represented by sticks, and their origins at the vortex element locations are depicted as solid circles. In Figs. 2 and 3, the envelope of the vortex elements in the port depicts the development of the familiar boundary layer on the port wall. Note that as the piston accelerates from TDC to its peak speed at midstroke, and decelerates thereafter to rest at BDC, the effective Reynolds number in the port increases and decreases correspondingly. As a result, the boundary layer on the intake port wall experiences a decrease and increase in its thickness, respectively. Furthermore, the absence of vortex elements in the inner region of the intake port points to the existence of a wide potential core there. The extension of this core into the chamber, which bifurcates around the valve seat, marks the jet core as it is issued from the port into the chamber (Figs. 2 and 3.) The 183

high-velocity vortex elements on the outer surface of the jet core depict the jet shear layer. Also notice that near the walls, the vortex elements are separated from the former by a very thin region. This region is the so called boundary domain, within which the Prandtl boundary layer approximation is used to evaluate the evolution of the wall-generated vortex tiles. (The tiles are not shown in the figures for clarity of presentation.) Note that the boundary region is thinner than the physical boundary layer (depicted by the outer envelope of the vortex elements near the wall.) In addition, unlike the physical boundary layer, which is controlled by the instantaneous Reynolds number, the thickness of the boundary region is fixed in time and specified by the global Reynolds number. As the piston proceeds toward BDC, the valve seat directly below the intake port deflects the incoming jet from the port and fans it out radially into the cylinder. The high velocity gradients, caused by the interaction between the inner surface of the conical jet in the cylinder and the side-walls of the valve seat, generate vorticity on the latter instantly. During the early stages of the induction process, the cylinder contains minimal vorticity and the flow within it may be viewed primarily as potential. Consequently, dictated by the potential flow properties and as seen in Figs. 2 and 3 (27 and 36 crank angles,) the valvegenerated elements convect sharply under the seat toward the centerline and initiate the development of a primary (or valve) toroidal eddy. Concurrent with the development of the valve eddy, the outer surface of the conical jet separates from the port due to the sudden expansion at the interface of the port and the cylinder head, and begins to roll up at the top corner of the cylinder into a secondary toroidal eddy. In what follows, for the sake of simplicity, references to left and right eddies imply the left and right cross-sections of the toroidal eddy, respectively. Similarly, left and right jets refer to the corresponding cross-sections of the conical jet. During the early stages of the induction process, the left and right eddies in regions B and A (Figs. 2 and 3, respectively) are basically circular and equal in diameter. In both cases, the eddies remain centered symmetrically with respect to the valve axis, in the proximity of the valve seat edges, until about 50 crank angle. As the flow develops further (until 90) the left and right jets continuously supply fluid to, and grow the size of their respective eddies. The latter evolve into ellipses whose major axes are parallel to the jets that drive them. The eddies in region B (Fig. 2) experience an unequal growth in size. From TDC through the mid-stroke, the left eddy grows such that its center stays essentially stationary below the left edge of the valve seat. At 90, the left eddy center is below the valve seat by approximately a fourth of the gap between the valve bottom and the piston. On the other hand, due to the geometric asymmetry, the right eddy in region B (Fig. 2) rolls radially outward and axially downward, away from the valve seat, such that by 90 crank angle it is centered half-way between the right edge of the valve seat and the cylinder wall on the right. Moreover, the vertical position of the right eddy center is at approximately two thirds of the valve-piston gap below the valve seat. The eddies in region A (Fig. 3) remain, for all practical purposes, symmetric with respect to the plane of symmetry of the

Figure 2: Vortex element trajectories within volume B in Fig. 1. Normalization in each time frame is based on the instantaneous maximum speed in the field. 184

Figure 2: (Continued)

185

Figure 3: Vortex element trajectories within volume A in Fig. 1. Normalization in each time frame is based on the instantaneous maximum speed in the field.

186

Figure 3: (Continued)

187

engine. Indeed the flow structure closely resembles that observed in axisymmetric valved engines [16]. In this case, both left and right eddy centers slowly migrate radially out and axially down. By mid-stroke, the centers are positioned half-way between the valve seat and the piston, and away from the valve edges by a third of the distance between the seat edge and the cylinder wall. Note that the vertical position of the eddy centers in Fig. 3 project onto the axis of symmetry of the inlet port (and the valve) in Fig. 2. Connecting the three center positions in Fig. 2, one can reconstruct the evolution of the toroidal eddy as follows. During the initial stages of induction, the core (or cross-section) of the primary toroidal eddy is azimuthally uniform - starting with a circular shape and slowly growing into an elliptical core. The initial diameter of the toroidal eddy is smaller than the valve diameter, and its axis of rotation is normal to the piston face. During this stage, the dynamics of the torus is quite similar to the behavior of ring vortices in free space. As time progresses, the conical jet continuously supplies fluid to the primary eddy, so that the core of the latter continues to grow. In the mean time, due to the combined effects of the geometric asymmetry of the flow in the xz plane and the expansion caused by the piston motion, the diameter of the torus expands radially out while its axis of rotation begins to tilt in the positive x direction. The tilting of the torus proceeds as if a section of it is hinged or anchored near the valve seat (left eddy in Fig. 2.) The tilt angle of the toroidal axis of rotation with respect to the zaxis is 20-30 at mid-stroke. Beyond mid-stroke, a weak counter-rotating toroidal eddy develops at the corner between the piston and the cylinder wall. In addition, the primary toroidal eddy begins to lose its coherence. In particular, the flow in region B (Fig. 2) appears to break up into multiple smaller scale vortical structures. The flow in region A (Fig. 3) remains weakly symmetric and displays a mild coherence. We will return to this topic shortly. In addition to creating the primary toroidal vortex, the conical jet traps a secondary toroidal vortex at the top of the cylinder. The structure of this weak torus remains stable during the entire simulation, although its core size and shape change in time due to the unsteady dynamics of the entrapping jet. As shown in Figs. 2 and 3, the core size and shape of the torus around its azimuth are primarily defined by the distance between the port and the cylinder walls. Note that, within the limited range of parameters tested, the break up of the primary toroidal eddy after midstroke, into smaller less coherent vortical structures was found to be independent of the temporal and spatial resolutions of the predictions. Therefore, we propose that the observed break up is due to the inherent instability whose mode is probably selected by perturbations caused by the numerical discretization of the engine surfaces. Before we elaborate on this proposition, we refer to a series of flow visualization experiments conducted by Ekchian and Hoult [16], on a valved axisymmetric engine and an engine with an off-centered valve. (The geometry of the latter case is similar to that of the present simulation.) Fig. 4 is a reproduction of the flow visualization results for the case with the off-centered 188

valve, within regions A and B in Fig. 1. Note the similarity between the experiment and the present simulation, despite the detailed differences in the geometries and flow conditions. The differences include the Reynolds number, the valve lift and its exact off-center position, the staircase discretization of the engine geometry, and the valve simplification by a seat. The two relevant conclusions made by Ekchian and Hoult [16] - as a result of moving the valve from the center of the engine cylinder toward its edge - were: (1) "the planes of the axes of rotation of the vortices are no longer perpendicular to the cylinder axis but are tipped at an angle to it," and (2) "the lifetime of the vortex is shorter."

Figure 4: Flow visualization in a motored engine with an off-centered valve (Ekchian and Hoult.) A and B correspond to the planar views in Figure 1. (Bore/Stroke = 1.35, R.P.M. = 750, Lift = 0.21) We proceed with our proposed explanation of the observed break up of the flow coherence near the cylinder top by focusing on the dynamics of the primary toroidal eddy as a single three-dimensional structure. As we mentioned earlier, the dynamics of the primary toroidal eddy in its simplest form - during the initial stages of the induction step - closely resembles the dynamics of a vortex ring moving in free space. The latter has been analyzed extensively [32, 33], as well as investigated experimentally [34, 35] and computationally [36]. In fact, Ekchian [37] was able to predict vortex breakdown in axisymmetric engines with poppet valves, using linear theories of the inviscid instability of ring vortices. In a

parametric study by Knio and Ghoniem [36], vortex rings were excited by a sinusoidal perturbation in the radial direction, using an integer number of waves around the azimuth of the rings. The computations predicted the development and growth of streamwise vortices around the ring azimuth, whose stability mode depends on the nature of the perturbation as well as the ring. More specifically, the most unstable wavenumbers were found to be proportional to the ratio of the ring radius to the core size. Additionally, outside a narrow band around the most unstable wavenumbers, the rings were shown to oscillate stably. In long time computations, these perturbations grow leading to the break up of the ring. In the present computation, the circular cross section of the port and the valve seat are discretized by 20 unequal staircase segments, as depicted in Fig. 1. As a result, the contour of the discretized circles imparts small perturbations with multiplicity of wavenumbers on the inner and outer surfaces of the conical jet, as well as the perimeter of the primary torus. The growth of these perturbations, via the mechanism mentioned earlier, leads to the development of streamwise vortices around the azimuth of the torus, accompanied by initially mild oscillations. The tilting and the radial expansion of the toroidal vortex lead to further stretch and the additional amplification of the waves, such that eventually break-up is observed. While the origin of the initial perturbation is numerical (e.g., the discretization of the circle into a staircase-shaped polygon,) in practice one encounters other perturbations due to turbulence that lead to a similar break up of the vortex - as verified by the experiment in Fig. 4. Fig. 5 depicts the evolution of = 0.18 surfaces, where is the local vorticity, normalized by the instantaneous maximum vorticity in the field. The local value of vorticity was evaluated on a uniform grid using its definition as the curl of the velocity. corresponds to the enstrophy, which can also be regarded as a measure of the energy dissipation rate per unit volume, and can be used to detect regions of high energy dissipation, shear and mixing within a flow. The constant value of 0.18 was selected to focus on the flow in the vicinity of the valve seat. Note that since normalization in each time frame is done with respect to its instantaneous maximum, the displayed iso-surfaces do not correspond to the evolution of a particular enstrophy value. For example, the larger volume of vorticity at 135, as compared to say 108, does not necessarily imply a corresponding increase in vorticity. Notice the development of relatively strong vorticity patches below the valve seat (from 27 to 45 crank angles.) The patches correspond to the streamwise vorticity in the primary toroidal vortex. The development of streamwise vorticity near the corners of the seat correspond to the growth of perturbations caused by the approximate representation of the valve seat. Beyond 50 crank angle and until approximately the mid-stroke, an iso-surface representing a coherent toroidal vortex with a non-circular core is distinctly visible. The perturbations associated with the observed waviness of the iso-surface grow and lead, in the later stages, to the break up of the torus, and the formation of a "spotty" vorticity. 189
2 2

CONCLUSIONS
A three-dimensional vortex-boundary element method has recently been developed for the simulation of flow in the time-dependent geometry of engines. A brief description of the approach was presented in this paper, followed by an example of intake flow inside a circular chamber, containing a valve seat and fitted with an offcentered port at one end and a harmonically moving piston at the other. The circular cross-section of the engine was represented by a staircase idealization. The Reynolds number based on the piston diameter and its maximum speed was set at 350. In order to save CPU time and to eliminate the generation of high aspect ratio elements, a time-varying number of surface meshes was used to discretize the cylinder walls. 404 vortex tiles and 1,392 boundary elements were used to mesh the domain at TDC, which grew to 740 and 2,512, respectively, at 135 crank angle. Results show that the flow inside the chamber is dominated by a primary toroidal eddy, which is generated due to the interaction of the valve seat with the jet issued from the intake port. The axis of rotation of the torus is initially normal to the piston. However, due to the asymmetry of the engine geometry and the expansion caused by the piston motion, the torus continues to expand radially out while its rotational axis begins to tilt to a 20-30 angle with respect to the normal to the piston face. The vortex structure remains coherent until 90 crank angle. However, beyond the mid-stroke, the geometry-induced perturbations experienced by the toroidal vortex grow and eventually lead to the break up of the vortical structure into smaller, less coherent eddies. This process was corroborated by experimental evidence. In addition to the primary torus, the jet separation was seen to give rise to a stable secondary vortex torus at the top corner of the chamber. The size and shape of the latter were seen to be dependent on the relative position of the port with respect to the chamber. The oscillations which are observed around the azimuth of the primary toroidal eddy result from perturbations introduced in the numerical discretization of the valve seat. In practice, the origin of these perturbations could be the turbulence present in the incoming flow.

ACKNOWLEDGMENTS
This project was funded by FORD Motor Company. The numerical experiments were performed on the Cray C90 at the Pittsburgh Supercomputing Center.

REFERENCES
1 Butler, T. D., Cloutman, L. D., Dukowicz, J. K. and Ramshaw, J. D., "CONCHAS: An Arbitrary Lagrangian-Eulerian Computer Code for MultiComponent Chemically Reactive Fluid Flow at All Speeds," Los Alamos Scientific Laboratory Report No. LA-8129-MS, 1979. 2 Gosman, A. D., Meling, A., Watkins, A. P. and Whitelaw, J. H., "Axisymmetric Flow in a Motored Reciprocating Engine," Proc. Instn. Mech. Engrs., Vol. 192, pp. 213-223, 1978. 3 Gosman, A. D., Johns, R. J. R. and Watkins, A. P.,

"Development of Prediction Methods for In-Cylinder Process in Reciprocating Engines," Mattavi, J. N. and Amann, C. A., eds, Combustion Modeling in Reciprocating Engines, Plenum, New York, pp. 69129, Plenum Press, 1980. 4 Martins, L.-F. and Ghoniem, A. F., "Vortex Simulation of the Intake Flow in A Planar Piston-Chamber Device," Int. J. Numer. Methods Fluids, Vol. 12, pp. 237-260, 1991. 5 Morse, A. A., Whitelaw, J. H. and Yianneskis, M., "The Influence of Swirl on the Flow Characteristics of A Reciprocating Piston-Cylinder Assembly," J. Fluids Engrg., Vol. 102, No. 4, pp. 478-480, 1980. 6 Errera, M. P., "Numerical Prediction of Fluid Motion in the Induction System and the Cylinder in Reciprocating Engines," SAE Technical Paper No. 870594, Detroit, MI, 1987. 7 Khalighi, B., "Multidimensional In-Cylinder Flow Calculations and Flow Visualization in a Motored Engine," J. Fluids Engrg., Vol. 117, No. 2, pp. 282-288, 1995. 8 Naitoh, K., Fujii, H., Urushihara, T., Takagi, Y. and Kuwahara, K., "Numerical Simulation of the Detailed Flow in Engine Ports and Cylinders," SAE Technical Paper No. 900256, Detroit, MI, 1990. 9 Taghavi, R. and Dupont, A., "Multidimensional Flow Simulation in An Inlet Port/Combustion Chamber Assembly Featuring A Moving Valve," ASME Internal Combustion Div., ICE v6, pp. 9-15, 1988. 10 Gharakhani, A. and Ghoniem, A. F., "Vortex Simulation of the Three-Dimensional Wake Behind A Cube," AIAA-96-0170, 34th Aerospace Sciences Meeting, Reno, Nevada, Jan. 1996. 11 Gharakhani, A. and Ghoniem A. F., "3D Vortex Simulation of Intake Flow in An Off-Centered Port and Cylinder with A Moving Piston," ASME International, Fluids Eng. Conference & Exhibition, San Diego, CA, July 1996. 12 Gharakhani, A. and Ghoniem, A. F., "Massively Parallel Implementation of A 3D Vortex-BoundaryElement Method," 2nd International Workshop on Vortex Flows and Related Numerical Methods, Montreal, Canada, Aug. 1995. 13 Gharakhani, A. and Ghoniem, A. F., "BEM Solution of the 3D Internal Neumann Problem and A Regularized Formulation for the Potential Velocity Gradients," to be published in Int. J. Numer. Methods Fluids. 14 Gharakhani, A. and Ghoniem, A. F., "Three Dimensional Vortex Simulation of Time Dependent Incompressible Internal Viscous Flows," Submitted to J. Comput. Phys. 15 Gharakhani, A., "3-D Vortex-Boundary Element Method For The Simulation Of Unsteady, High Reynolds Number Flows," Sc.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA., 1995. 16 Ekchian, A. and Hoult, D. P., "Flow Visualization Study of the Intake Process of an Internal Combustion Engine," SAE Paper No. 790095, International Congress and Exposition, Detroit, MI, 1979. 17 Beale, J. T. and Majda, A., "High Order Vortex Methods with Explicit Velocity Kernels," J. Comput. Phys., Vol. 58, pp. 188-208, 1985.

18

19 20

21 22 23 24 25 26

27 28 29 30 31 32 33 34 35

36

37

Beale, J. T. and Majda, A., "Vortex Methods for Fluid Flow in Two or Three Dimensions," in ContemporaryMathematics, Amer. Math. Soc., Vol. 28, pp. 221-229, 1984. Beale, J. T. and Majda, A., "Vortex Methods. I: Convergence in Three Dimensions," Math. Comput., Vol. 39, No. 159, pp. 1-27, 1982. Beale, J. T. and Majda, A., "Vortex Methods. II: Higher Order Accuracy in Two and Three Dimensions," Math. Comput., Vol. 39, No. 159, pp. 29-52, 1982. Goodman, J., "Convergence of the Random Vortex Method," Commun. Pure Appl. Math., Vol. 40, No. 2, pp. 189-220, 1987. Hald, O., "Convergence of Vortex Methods for Eulers Equations," SIAM J. Numer. Anal., Vol. 16, No. 5, pp. 726-755, 1979. Long, D.-G., "Convergence of the Random Vortex Method in Two Dimensions," J. Amer. Math. Soc., Vol. 1, No. 4, pp. 779-804, 1988. Sladek, V. and Sladek, J., "Non-Singular Boundary Integral Representation of Stresses," Int. J. Numer. Methods Eng., Vol. 33, pp. 1481-1499, 1992. Beale, J. T. and Majda, A., "Rates of Convergence for Viscous Splitting of the Navier-Stokes Equations," Math. Comput., Vol. 37, No. 156, pp. 243-259, 1981. Ghoniem, A. F. and Sherman, F. S., "Grid-free Simulation of Diffusion Using Random Walk Methods," J. Comput. Phys., Vol. 61, No. 1, pp. 1-37, 1985. Fishelov, D., "Vortex Methods for Slightly Viscous Three-Dimensional Flow," SIAM J. Sci. Stat. Comput., Vol. 11, No. 3, pp. 399-424, 1990. Ghoniem, A. F. and Ng, K., "Numerical Study of the Dynamics of a Forced Shear Layer," Phys. Fluids, Vol. 30, No. 3, pp. 706-721, 1987. Laminar Recirculating Flow," J. Comput. Phys., Vol. 68, No. 2, pp. 346-377, 1987. Puckett, E. G., "A Study of the Vortex Sheet Method and Its Rate of Convergence," SIAM J. Sci. Stat. Comput., Vol. 10, No. 2, pp. 298-327, 1989. Chorin, A. J., "Vortex Sheet Approximation of Boundary Layers," J. Comput. Phys., Vol. 27, pp. 428-442, 1978. Saffman, P. G., "The Number of Waves On Unstable Vortex Rings," J. Fluid Mech., Vol. 84, No. 4, pp. 625639, 1978. Widnall, S. E., Bliss, D. B. and Tsai, C-Y, "The Instability of Short Waves on A Vortex Ring," J. Fluid Mech., Vol. 66, pp. 35-47, 1974. Maxworthy, T., "Some Experimental Studies of Vortex Rings," J. Fluid Mech., Vol. 81, No. 3, pp. 465495, 1977. Sullivan, J. P., Widnall, S. E. and Ezekiel, S., "A Study of Vortex Rings Using A Laser-Doppler Velocimeter," AIAA Jl, Vol. 11, No. 10, pp. 13841389, 1973. Knio, O. M. and Ghoniem, A. F., "On the Formation of Streamwise Vorticity In Turbulent Shear Flows," AIAA-88-0728, 26th Aerospace Sciences Meeting, Reno, Nevada, Jan. 1988. Ekchian, A., "Flow Visualization Study of the Intake Process of An Internal Combustion Engine," Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1978.

190

Figure 5: Iso-surfaces of . = 0.18, where is the vorticity vector normalized by the instantaneous maximum field vorticity.

191