[xisluke]

Hi guys, I'm not sure to be in the right place of the forum by posting this thread, but I was wondering about the possible effectiveness of applying entropy analysis to common CFD industrial cases.

I'll try to explain better the concept. Usually, when efficiency evaluations of a fluid system need to be accounted, we rely on total pressure, total temperature and other "similar" quantities variations to estimate performance. Working with CFD at industry level, I never heard about entropy (or exergy) analysis applied to common calculations in order to evaluate entropy production rates due to heat fluxes and viscous/turbulence dissipations; matter of fact, my knowledge of these fascinating topics stops at academic level. At the same time, I do not know any commercial CFD code which includes a numerical solution for an entropy transport equation.

I red in some papers that there are possible ways to include, in the post-processing stage, some custom functions to estimate entropy production; does it worth it? I mean, in terms of convection heat transfer mechanisms or losses due to excessive turbulence or viscous effects, what the "plus" could be in your opinion by implementing this not so common approach?

Thanks

[andy_]

Not sure I wholly understand the question but compressible flow simulations solve the transport equations for a pair of independent thermodynamic variables and so entropy is known and widely evaluated and plotted. Some compressible codes choose to use entropy as one of the transported variables which enables a further degree of control over numerical errors related to entropy. It can be useful for example where acoustics is a significant concern but it is not a common choice.

[xisluke]

Originally Posted by andy_

Not sure I wholly understand the question but compressible flow simulations solve the transport equations for a pair of independent thermodynamic variables and so entropy is known and widely evaluated and plotted. Some compressible codes choose to use entropy as one of the transported variables which enables a further degree of control over numerical errors related to entropy. It can be useful for example where acoustics is a significant concern but it is not a common choice.

I meant that Entropy is commonly evaluated starting from other variables; for example, within Fluent I think I can plot entropy, but it should be derived from pressure and static temperature values, no transport equation is solved. If one wants to plot the entropy production, due to dissipation or heat transfer (last two terms in the rh side of the equation), it is necessary to implement a custom field function in post-processing. I heard about thermoacoustics applications, I was wondering if this concept could be used also for other losses evaluations

[LuckyTran]

For those unaware, entropy generation minimization and exergy destruction are concepts popularized by Adrian Bejan and especially prevalent in his Advanced Engineering Thermodynamics textbook.

There are a number of issues with it that make it not suitable, even for academic purposes.

One issue is that it doesn't scale, you see the phi/T term which means that it depends on the operating condition, which also means that you need proper thermal BC's and the correct geometric scale. This is at odds with most CFD'ers who like to simplify their problems by assuming adiabatic BCs and what not, or scaling their models.

Other issue is, it's highly dependent on what is happening in the dissipative range, which is highly modeled for turbulent flows. Basically, you need a closure model for entropy generation due to turbulence, e.g. a turbulent Schmidt number for entropy. There's already enough models in turbulence modeling that adding more models on top of it does not really shed any more light onto what is already being modeled. It sounds like a mouthful because it really is.

Furthermore, entropy generation is very impractical from a design standpoint. It doesn't tell you how many cycles til failure this component will last and just doesn't fit into a multi-objective optimization problem. Its application are too niche for modern heat transfer. It tells you where you need to go and tries to steer you into sometimes impossible solutions when as an engineer, you just want to know the range of operation your device can take. Entropy generation is minimum when your device has melted, try taking that to your boss!

Finally, it is simply more work, more colors. It's already hard enough to get people to plot basic profiles at 15 sections in the flow and properly document it.

tl;dr you run one CFD case, you get an entropy generation field. And then what!?

[andy_]

I'm still not sure I wholly understand your question but people simulating compressible flows have always evaluated and studied reversible and irreversible losses. Understanding and quantifying the processes in the flow in order to guide the design process is often, probably usually, the main reason an engineer performs a CFD simulation. The losses may need a bit of interpretation in the light of numerical errors and possible differentiation and/or integration that is different from what is in the code.

Or are you asking where in a particular post-processor various kinds of losses are evaluated? Or are you asking if looking at the losses in a flow is useful and informative for an engineer? Or perhaps am I still missing the point?

[LuckyTran]

You are missing the point. The question is why isn't anyone doing it Bejan's way. Bejan promises that the design is optimal when entropy generation is minimized. Any yet nobody doing CFD is using this obvious parameter. So why not? Btw Bejan proposed this formulation in the 80's, it's nothing new. So where is the disconnect?

[xisluke]

Originally Posted by LuckyTran

Finally, it is simply more work, more colors. It's already hard enough to get people to plot basic profiles at 15 sections in the flow and properly document it.

tl;dr you run one CFD case, you get an entropy generation field. And then what!?

LuckyTran, I was asking because I found papers where simple custom field functions in the post-processing stage can be added (I did it in minutes) able to estimate the entropy generation, including the mean and the oscillating terms modeled though turbulent parameters, like Turbulent Dissipation Rate and Turbulent Prandtl Number. This model works well until the y+ values remain very low, if not it would be necessary to provide wall functions (included in the papers) and I think that it must be done by exploiting some UDFs.

At this point, the question is exactly what you stated: how the results could be interpreted? My original question was related to the fact that the authors stated that, especially for a heat transfer problem, there are ways to improve heat transfer that also increase the pressure drop in the system; looking at the entropy generation is an approach theoretically able to find an optimum, and I was wondering if there are other possible (and useful) applications rather than the simple case treated in the paper.

http://www.sciencedirect.com/science...42727X05000329

[andy_]

Without quoting LuckyTran's post.

Perhaps I am missing the point but the OP seems to have got a pair of apparently conflicting answers which is interesting. I am not familiar with the details of what Bejan proposed and was answering a more general question about making an effort to understand and quantify the physics of reversible and irreversible losses in a CFD simulation.

The importance of understanding various loss generating mechanisms is going to change from flow to flow. For fan, compressor and turbine stages in aeroengines for example it is going to be important given the amount of effort invested in bringing about very small improvements. For fans in bathrooms or computers probably not so much. This is one possible source of disagreement.

Another is perhaps being sufficiently familiar with the physics of the loss generating mechanisms in the flow of interest to know when a detailed assessment is required and when it is not given the information needed from the simulations. This can be relevant when comparing flows where the conditions on the boundaries are not quite the same and small apparent improvements in efficiency may not be real. Taking the time to understand and quantify the loss mechanisms may be required if a reliable assessment is to be performed. I was involved in an example of this 30 years ago when assessing whether to continue funding some research on aeroengine turbine geometries and the extra information provided reversed a provisional decision.

So perhaps the answer to the OP's question is that it can be important in some circumstances but not in others. Knowing when it would be useful and how to extract it from a simulation is likely to be knowledge that has value for a fair few engineers working with CFD simulations.

[xisluke]

Originally Posted by andy_

OK so including the quote of post 6 at the top of my post was causing a permission denied error. Why?

I'm not sure about the reason why you had such issues in posting your answer, I never saw anything like that in writing post with quotes and so on...

By the way, you both brought interesting points. Probably my original question was more related to the point raised by LuckyTran, but it could be also meaningful to be aware that an accurate analysis of the problems to be solved can give an insight about whether it is necessary to go for detailed analysis of losses or not. Considered that almost nobody looks at entropy generation in CFD, even if only a simple coding exercise is requested in post processing, it could mean that those results are not so necessary, except from few isolated cases. I will probably look for other papers to see how these analysis could be applied, I am not an expert in this field either.

[LuckyTran]

Just a disclaimer I've tried this approach for turbine heat transfer and you get no where fast

Originally Posted by xisluke

there are ways to improve heat transfer that also increase the pressure drop in the system

That we have already known from the Reynolds analogy! So it provides nothing new.

This theory is applicable to virtually all heat transfer problems. In my personal opinion though, it is useful only when the device has few number of moving parts, and by few, I mean 0 (case in point, heat exchangers). In heat exchangers there are few penalties for making a device bigger or smaller. As soon as you have 1 moving part, it makes the product unsellable (e.g. a gas turbine). And it works better when you are a systems-level designer able to modify the entire device, not a component level design where you have upstream and downstream constraints.