>Well, just to spice it up... there is an implicit assumption >in your assessment that 2+4=6. I can argue that 2+4=sqrt(20)

Not a very good example - two vectors added together give another vector which implies what you mean is |2i + 4j|=sqrt(20). The 2i and 2j are different entities and so when added give the vector 2i+4j whose length is sqrt(20).

A better example would be to use modulo arithmetic (e.g. take the remainder after division by 5 of the result) then

2 + 4 = 1 and 2*4 = 3.

This type of calculation preserves all the usual rules for addition and multiplication found in algebra.

You're still thinking in the same "space" with the modulo example, even though your example, nevertheless, proves (part of) my point.

>>two vectors added together give another vector

Just like two scalars added together give another scalar? Imagine that! Is adding two apples with three oranges acceptable? But this is clearly a scalar field operation...

Again, we go back to the point of the implicit assumptions. Anyway, I only meant to interject a "caution" statement to the discussions on this thread, which is sort of one-sided )

Adrin Gharakhani

Hi,

This is an interesting topic and some very good points have been made!

It is interesting to note that over the years I have seen the same question arise for any of the individual "tough problems" out there....turbulence modeling, radiation, aero-acoustics, rarefied flows, etc

Since most industrial combustion systems are high Reynolds number, then often the limiting issue is the capability of modeling turbulent combustion, which really is a specific case of closing the turbulence problem.

The main issue in practical RANS combustion modeling is getting sources in the equations in terms of mean temp, mass fractions, etc,...ie the closure problem.

The often surprising ability of simple turbulence models for other applications (cold flows with no reactions) leaves people very optimistic when first looking at combustion modeling. Even a very simple model such as eddy dissipation (which reduces the whole complexity of combustion locally to one time scale: k/eps) looks promising when applied to an appropriate problem.

However, many combustion applications (or what we need to know in a specific application ie NOx, peak temp, local features, etc) do not fit nicely in the range of applicability of the simple models. If you do not get those sources in the closure right.....you will get USELESS results!

People are in general, overly optimistic about what the simple models can provide because they do not understand what is at the heart of the problem.

The research community, clearly undrstands the challenges in the problem and has responded with an abundant number of models which are often very complex, possibly not widely applicable nor robust in a solver. The industrial community, however, does not have the resources or knowledge to use these models...is there a way out of this cycle????

I have really been amazed over the years just how many industrial simulations are being done with k-eps and eddy dissipation...particularly in north america. Europe seems to be a bit more daring? And yes I have seen some really useful and insightful results....and some that are totally inappropriate. The trick is that the number of people who can tell the difference (in a predictive sense not just after they have compared to the rig)....is pretty small!

So yes the models are of use in predictions if your exectations do not exceed your ability to "self judge" your simulations.

Regards,

Bak_Flow