9611953D Vortex Simulation of Intake Flowin a Port-Cylinder with a ValveSeat and a Moving Piston
Adrin Gharakhani and Ahmed. F. Ghoniem
Massachusetts Institute of Technology
Copyright 1996 Society of Automotive Engineers, Inc.
A Lagrangian random vortex-boundary elementmethod has been developed for the simulation ofunsteady incompressible flow inside three-dimensionaldomains with time-dependent boundaries, similar to ICengines. The solution method is entirely grid-free in thefluid domain and eliminates the difficult task of volumetricmeshing of the complex engine geometry. Furthermore,due to the Lagrangian evaluation of the convectiveprocesses, numerical viscosity is virtually removed; thuspermitting the direct simulation of flow at high Reynoldsnumbers. In this paper, a brief description of the numericalmethodology is given, followed by an example of inductionflow in an off-centered port-cylinder assembly with aharmonically driven piston and a valve seat situateddirectly below the port. The predicted flow is shown toresemble the flow visualization results of a laboratoryexperiment, despite the crude approximation used torepresent the geometry.
Computational fluid dynamics is fast becoming aserious alternative to the empirical cut-and-try approach inthe design and analysis of IC engines. Numerous newly-developed methodologies and their applications haveappeared 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 realisticthree-dimensional flow in the intake ports and the cylinderof the engine, including moving valves [6-9]. However,despite the enormous progress achieved so far, severalcomputational and physical modeling difficulties remainunresolved.In grid-based computational methods, the accurate,efficient and automatic volumetric meshing of typical time-varying engine configurations remains a challenging task.To obtain a well-behaved solution, the mesh generationalgorithm must resolve the widely varying geometriclength scales in the engine accurately. In addition, it mustbe capable of capturing the high velocity gradients withinthe thin concentrated jets around the intake valve, as wellas 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 degeneratemeshes.The most adverse consequence of a poor-qualitymeshing is the introduction of false diffusion into the flowfield which, for the engine problem, leads to weaker largeeddy vortical structures and faster than expected decayrates. False diffusion may be reduced by increasing themesh resolution and/or applying adaptive griddingtechnology, both of which can become computationallytoo expensive in a complex engine geometry.Alternatively, false diffusion may be minimized by aligningthe streamwise side of the finite volume along the localstreamline. This can best be achieved using a Lagrangianmotion of the mesh nodes. However, given the complexityof engine flows, this approach may eventually lead todegenerate finite volumes.In this paper, an alternative approach is presentedbased on the random vortex-boundary element methodfor the simulation of unsteady flow in 3-D geometries withmoving boundaries of the type encountered in engines.The Navier-Stokes equations are expressed in thevorticity transport formulation, and are discretized using acollection of spherical vortex elements. The elementvelocity is expressed as a superposition of a vorticalcomponent evaluated by the Biot-Savart law, and apotential component obtained from the solution of a 3-DNeumann problem over the domain. The convection andstretch of vorticity are evaluated in the Lagrangian frameof reference (without the need for grids) and its diffusion isdescribed stochastically by the random walk method. Theboundary element method is used to solve a 3-D Laplaceequation that defines the potential flow and imposes thenormal flux boundary condition on the boundary surfaces -without having to discretize the domain interior. The no-slip boundary condition is satisfied by generating vorticitytiles at the boundaries. Within a thin prespecified regionnear the boundary, the tiles convect and diffuse in theLagrangian frame of reference according to the Prandtlequations. Beyond this region, the tiles are converted tospherical vortex elements.The Lagrangian vortex-boundary element method isgrid-free in the fluid domain. As a result, the complex taskof volumetric meshing of the interior of an engine isreduced to the far simpler task of meshing its surfaces.The method is self-adaptive and capable of dynamicallyconcentrating computational elements in regions withsignificant velocity