This question is for users (and developers) of commercial codes for engine flow simulation.

For a _practical_ diesel engine flow simulation with combustion (or any other engine if you care to comment), what assumptions are you willing to tolerate (and what assumptions are used in the code you use)?

Specifically, will the following equations be acceptable (which include quite a few assumptions) If not, then why not?

Mass:

Dr/Dt + r(Del.u) = 0 (r=density)

Assume anelastic flow of a thermally perfect gas:

Dr/Dt = -r Beta DT/Dt (Beta=isothermal compressibility)

Momentum:

rDu/Dt = -Del(p) + mu[ Del^2(u) ] (constant viscosity mu)

Energy:

(r.c_p)DT/Dt = Del.[ Lam Del(T) ] + Thet

Lam = conductivity Thet = source term corresponding to the rate of generation of energy per unit volume

Reaction:

Shvab-Zel'dovich with all its corresponding assumptions - Arrhenius type, fast single step reaction

Thanks in advance

Adrin Gharakhani

(1). Diesel engine simulation means a). high pressure fuel injector, b). fuel spray formation and penetration, c). fuel spray combustion in turbulent flow. (2). So, the primary issue is in the spray combustion in turbulent flow. In this case, it is somewhat transient in nature. In other words, one can first approach the steady state problem and then move on to the transient problem.

Good points John!

The question/concern for me is not steady/unsteady, but whether the assumptions in the combustion (and energy) part(s) in my message are acceptable. As for fuel spray injection, I didn't mention it because I wanted to know whether people use some sort of global heat release model or actually use fuel spray models (of the kind used, for example, in KIVA). The simplest of spray models brings down the computations to a serious halt, so what do users prefer? A fast simple heat release model, or a slow spray model? Of course, the question regarding the other assumptions still remains (e.g. constant viscosity, anelasticity, fast reaction, etc.)

Adrin Gharakhani

(1). The best way to do is to assume that the assumptions and the equations are valid. Then you can proceed to simulate the diesel engine flow field. (2). And somewhere along the line, you will probably come back to the single fuel particle combustion, or the fuel spray formation. (3). Anyway, it is the field which requires the supercomputer capability, when chemical reactions or spray combustion is involved. (4). I am not currently involved in spray combustion, so, don't pay too much attention to my message. There are probably many experts out there who are interested in making some comments.

your viscosity shouldn't be constant as the temperature range is high use sutherland's law (mu= mu(T) ) .some people at INRIA have done some work on IC engines that looks nice. i can't find it on the page anymore but you may be able to search for it. try: http://www.inria.fr/cgi-bin/MULTIMED...S;debut=;indx=

Hi!

First of all, the flow in a cylinder is a very low Mach number flow. Using a density based formulation like the one you are suggesting is not the best method for such flows. I would strongly recommend a pressure based method. However, if you are developing a solver for more than one application and want to stick to such a formulation, use pre-conditioning. You will not regret the extra time spent in coding the stuff. (Look for references from Dr. Merkle - he used to be at penn state).

Secondly, like some one suggested, you should definitely use at least Sutherland's law for temperature dependance of viscosity. Also, I would use the full energy equation with the assumption of a calorically perfect gas (c_p depends on temperature) rather than a thermally perfect gas. This way, you avoid two problems : 1) Very strong temperature gradients occur in the flow and hence strong mixture property variations can occur. Neglecting this can make substantial difference to your results. 2) Occurence of a source term in the energy can cause stiffness problems while solving it numerically.

Thridly, also related to the viscous terms is the fact that you are missing the viscous work terms in the energy equation. These terms can be very important in such a confined flow with walls everywhere.

I also have question for you. How do you plan to do things like piston and valve movement?

Regards Srinivasan

Thanks for the good comments. I agree with (almost) all the issues you raise. However, the focus of my question really is not what is physically realistic, but what is _acceptable_ to the average user in the engine community.

There are two options: (1) to develop a "fully-equipped" code that does "everything" for you, (2) to develop a more simplified, targetted code that performs the most important tasks and covers the most significant physics. I'd like to take the second route - that's why I spelled out even the engine type. So, I needed to get feedback as to what are the minimum acceptable conditions for current users.

As for the piston and valve motions, I see no particular difficulties there - it has already been done!

Adrin Gharakhani

Hi All,

My understanding is that the variation in viscosity due to temperature variations is insignificant in IC engines. This is due to the turbulent component of the total viscosity being many times (around 100 times from my experience) the laminar viscosity.

You seem to say the temperature variations are significant. I can't see how. Have I missed something?

Glenn

i didn't think of it that way. i just figured that since the gas is being highly compressed/expanded and heated to high temperatures by combustion that the molecular viscosity would vary. the turbulent viscosity is almost always at least ten times greater than the molecular viscosity. (in the Baldwin Lomax turb model transition is specified when mu(visc) = about 14*mu(molec) for example). do people in the turbine combustor field use this approximation as well.