Fluent Vs Solid-works Flow Simulation
 

Hello,

I am going to simulate a Thyristor and its heat sink with one of commercial CFD packages. I would appreciated if any body could suggest suitable CFD software for this job and also compare SolidWorks Flow simulation with Fluent in terms of accuracy and simulation time.

Thanks

Morteza

I suggest you Fluent,because:
1) It's easy to work
2) It has different models to satisfy your need
3) There are many people who know about the software and can help you
4) You can adjust setting to solve your problem according to your need.

Thank you for your nice recommendation

Forget about Fluent, CFX, CFDesign, Flotherm, Icepak, etc. Use Floworks (if you use Solidworks), EFDPro (if you use proE), or FLOEFD if you use another CAD package. These are the the most user friendly, most intuitive and stable (meshing, convergency) packages for Electronic cooling.

Hi Morteza,

first keep onething in mind before thinking of finalising one software. As far as i am aware of, fluent, icem, iepak, cfx are the professional level softwares. which are very accurate and precise if proper boundaryconditions and case setup are used. Where as solidworks flow or any other simulation option in a CAD package is also good enough but only for the sake of rough calculations. If it is for a college project or some assignment it is sufficient if you use simulation option from a CAD package. But if it is for some research work, better go for professional level softwares. Initially u might find a little difficulty, but as soon as u get habituated, u seem like playing with the tools...


All the best..
Regards
Pramodh

Hi CFD FAN,

I am verry sorry that my statement made you laugh. Frankly speaking i have very little idea about the flow works from solid works. I am a CFD professional in electronic cooling simulation works. I have been using icepak, CFX and CFDesign. I just have shared my experience. There were some cases involving complex physics, for which the simulation results were found to be nearly 98% close to the physical test. Thats the reason why i have suggested for ICEPAK. And thankyou for enlighting me with the information that the accuracy of the results depends on te case setup and the boundary conditions ( being a professional i am aware of that). Apart from that there will be some other capabilities like model selection and colving capability of the software which also matters.
http://www.lboro.ac.uk/departments/c...dices/appa.pdf
I think the stuff in the link works.. And please let me know if there is any additional information about Flow Works .

regards
pramodh

Dear Pramoth
The .pfd file is a)outdated and b)written by an academic person. I wouldn't trust the opinion of any of the academic people on topics involving commericial designs. I am not a professional in CFD but an electronic Eng who uses CFD once every 3 months or so when the design has to be thermally optimized. I don't have any affiliations with Floworks and the reasons I am speaking so highly of them is because their software has made my life so much easier. I have to add that Solidworks is the CAD package of our company and the transfer of mechanical assemblies and pcb designs into Floworks models is trouble free even for very complex geometries. In the end that's what an electronic designer wants from the CFD package – since the geometry is already done by the PCB or the mechanical group he just have to enter the materials, power loss, and the fan's curves to evaluate the temperatures performance of the product. One would also like to move some of the components here and there to reduce the temperature stress and expect these geometry changes to be propagated without meshing or other model problems in the CFD package. Floworks in combination with Solidworks does this very fast and trouble free (and better than anybody you've mentioned). If one is a professional CFD engineer doing CFD simulations all the time, yet he draws the geometries by himself, knows the tricks of geometry transfers and cleaning as well as the limitations of the meshers, then Fluent, CFX, Icepak and the other “heavy lifters” are probably a good choice. If you an electronic Eng (that was my impression about the guy who started this topic) end especially if Solidworks is the CAD package of your company, then Floworks is the best choice by far. And please don’t talk about the level of professionality of Floworks if you haven’t worked with it. Here is what I found on Youtube about Floworks (there is a lot more there): http://www.youtube.com/watch?v=Ehbnm...eature=related. Please nore that the setup and the geometry changes in the CFD model are all done under Solidworks (i.e. no geometry translators/editors/modelers are needed)

Mr CFDfan with all my respects as ur nick sais you are a fan in this world of CFD. I strongly recommend you to read a little abaut what is a turbulence model, what kind of modelizations exist, and why we NEED TO MODELIZATE a turbulence, also it would be nice for you to know what is and eddy scale and the correspondence with the grid... you can start with those really basics comcepts of how this world works. Honestly you are not in place for arguing about wich CFD software provides a better solution.
At this point i have to say that Mr Pramoth was correct in his ausmption of "rougth results" for those aplications of flow analisys in CAD soft.

PD: The mathematical solution of a turbulence phenomena envolves multiple coupled differential eqquations (Navier-Stokes equattions) wich need to be solved for each scale of the turbulence.

"Using a CFD soft its really easy, knowing what you are really computing not "

You assumed automatically here that Floworks (Floefd) doesn't model properly the turbulence and therefore is not accurate. I don't know where you got this from, certainly not from the developer's site. My feeling also is that you have never used the code in question but were nevertheless doing bold assumptions about its inaccuracy.
I hope somebody from floefd/FW developers would take part in the discussion as well and explain you how the turbulence is modelled there.

Finally please don't forget that the whole discussion started by a guy involved in electronic cooling area where the turbulence is not so prominent anyway.

This has nothing to do with electronic cooling, but may be useful for anyone who finds this thread via google.

A small company I worked at previously was using FloWorks to model FSI and aero loads on an aircraft at the PDR phase of design. Before I came onboard (2006-ish), our aero group wanted to check the software, so they tried to match the wind tunnel results of a NACA 0012 airfoil with FloWorks. They ended up finding the results were off by ~20%, and I forget which way. They found it to be a consistent enough error that they were comfortable using that difference as a correction factor for some of their results. The work was done by 2 CFD analysts that had 5-10 years of experience and M.S. degrees in aerospace engineering from decent schools. I don't know if they were able to compare that result to CFX/Fluent/etc in a double blind test.

I found FloWorks very easy to use and very intuitive, but I was not able to comment directly on its accuracy as I did not perform any studies testing that. (I was on the structures side and was looking at thermal profiles.) The team at SolidWorks seemed to be actively developing and improving the software with every release.

In electronic cooling I find the correspondence with the reality within 2-5% I've done the comparison for 7 or 8 models. This percentage is pretty good in our field because we don't try to push the devices to their thermal limit anyway. We usually target 15-25% thermal margin "to keep the smoke inside the devices" :)

Some feedback from the "development" side of FloEFD and SolidWorks Flow Simulation
 

Hi everyone,

first to Mortezab: In General I wouldn't use Fluent for something fluent is not really meant to be used. You also wouldn't use a SUV to race against a Ferrari, would you?
The tool you should use is if you’re going into more detail in the electronics cooling, so to use thermal network representation of the component such as Delphi or others. There are just a few tools that are highly specialized on electronics cooling. FloTHERM is one of them but Fluent or SolidWorks Flow Simulation are not. They certainly can do a component with a heat source and you can model it more detailed but it’s much easier to use the other tools. For FloTHERM for example you often get FloTHERM compact models from the vendors of the component, sometimes on their website.

@mashinsazi:
Have you ever used any other tool than Fluent?
1) It seems you never used a tool that is really easy to work with but fluent is not one of them
2) That’s valid for many codes depending on what you want to do, in what detail and what physics you want to include.
3) There are also many people who know the other software too and can help but in general a good training is more worth than self-training. A good vendor provides you the training suited to your application when you buy the software and you are trained by those who know the software inside out.
4) See point 2)

I would highly recommend to state the exact task you have, the complexity, where you get your data from and the future in your work. For example if you just want to use a Thyristor as a block with a heat source on a nicely designed CAD model heatsink then FloEFD or SW Flow Simulation (SWFS) would fit depending on which CAD system you use (FloEFD for Catia, ProE, Creo, NX and many others and SWFS if you already have SW inhouse). If it is a more complex representation of the Thyristor with a detailed PCB design from a PCB design software such as from Mentor, Zuken or others it is more suitable for the electronics cooling specialized tools such as FloTHERM and Icepak that CFDfan mentioned. Especially if you want to do more e-cooling work in detail in future you will benefit from those tools.
Of course I would suggest FloTHERM but you should test the software yourself with which one you feel more comfortable, will suite your application better or fit in your processes better. The vendors can tell you all the features of the software but the user is the one who will has to decide at the end and as always the personal feel and taste for a product and the processes in the different companies are different so you can get some ideas here but I strongly recommend to test it yourself and don’t buy because someone who doesn’t know you or your work tells you to buy a certain product. You do test drive a car you buy yourself, don’t you? You don’t listen to a stranger and buy one blindly.

@ Pramodh:
Not true! SWFS which is the same as FloEFD just distributed by SW, instead of Mentor, is very capable to deliver faster and more accurate values than Fluent and CFX in some cases. Just because it is embedded in CAD doesn’t mean it’s based on rule of thumb methods. It is a 3D CFD code based on navier-stokes equations just like Fluent and CFX. There is nothing rough about it. Of course just like with every code, nothing is suitable for everything. For no reason there are codes mostly used turbo machinery and others for external aerodynamics or under hood simulations. Each one has its strengths and weaknesses.
I know polymer extrusion companies where CFX engineers were struggling to create a mesh and gave up after 45 minutes and FloEFD ran through the model automatically meshing it in less the time and delivered results equal to measurements. And there are many other examples.
So careful with such statements if you haven’t tried a product.
Your Link to the CFD tool selection is based on a CFD project for HVAC simulation and you cannot compare the selection criteria with electronics cooling criteria, especially not the results of this projects decision. The basic method to put basic criteria in place that have to be fulfilled and select typical test cases in order to evaluate the tools is good but it should be based on your task (PCB import from EDA tools, CAD import, detailed thermal network modeling of components etc.) and the benefits and results (easy to use, necessary interfaces maybe to stress simulation or so, optimization study tools that are embedded to not create additional costs, solver speed and accuracy compared to measurements). And always never compare CFD with CFD results but CFD with measurements. But all that depends on the problem you want to solve, one might have HVAC problems others have e-cooling problems etc.
If you want to know more about Flow Works you can go to the website www.mentor.com/mechanical and check for some of the FloEFD white paper. There is for example one on meshing technology and another in turbulence modeling.
Meshing:
http://www.mentor.com/products/mechanical/resources/overview/advanced-immersed-boundary-cartesian-meshing-technology-in-floefd-6ec65867-0711-4932-968a-3d355f0012ce
Turbulence modeling:
http://www.mentor.com/products/mechanical/resources/overview/enhanced-turbulence-modeling-in-floefd-44ea7365-b16e-4d1f-a912-4f40ce43c862

@ CFDfan:
You are right regarding CAD embedded, stability and usability. Regarding result accuracy and especially speed in larger more detailed electronics cooling simulations FloTHERM or FloTHERM XT is much better suited. But as I said each software has its strength. FloEFD or SWFS is very good where you work on CAD geometries you want to change constantly to optimize the design and FloTHERM is more e-cooling specialized with not such a good CAD connection like SWFS. FloTHERM XT bridges much better as it uses a CAD modeler and you can use SW models directly and has the electronics capabilities of FloTHERM. It is even based on the same meshing and solver so you will notice the same background for complex CAD models but can leverage the e-cooling of FloTHERM. So depending on where your priorities are you will find a suitable software.
I don’t quite agree with SWFS is the best package for e-cooling with SW models as CAD model. FloTHERM XT is better electronics cooling suited with more detail to that than SWFS is. SWFS has a e-cooling module but regarding component modeling or interfaces to EDA tools FloTHERM XT is better. SWFS for example has only 2-Resistor models or simply a heat source on a block with a package material but FloTHERM can use detailed package thermal models.
Have a look at this video:
http://www.mentor.com/products/mechanical/multimedia/electronics-cooling-simulation-step-change-by-incorporating-the-eda-and-mcad-design-flows

@ Alancran46:
Before you agree with someone that CAD embedded CFD is “rubbish” regarding accuracy you should test it first. Why do you think that CAD embedded does automatically mean that there is no turbulence modeling? The type GUI does not automatically exclude a turbulence modeling in the solver.


@ Ace37:
Just some additions to what you wrote.
The accuracy for aerospace aerodynamics has improved since back then.
Also SW is not developing the CFD code, it is done by Mentor Graphics. The former developer was NIKA which got bought by Flomerics (developer of FloVENT and FloTHERM) and in 2008 they got bought by Mentor Graphics.

So to come back to the electronics cooling task:
Define what you want to do and in what detail. Also think about the future, will it be just Thyrisors on a heat sink or several of them on a PCB with the PCB design coming from other EDA tools and with a relatively simple box like enclosure or a really neat CAD designed housing. There are often more factors to select a CFD tool than just Thyristor on a heat sink.

Regards,
Boris

PS: Yes, I do work for Mentor Graphics and no this is not tweaking the facts towards our products being good. I only talk from my experience with our customers. And no, they are not all laymen in CFD, in fact there are some of the best CFD experts trusting our products at Top 10 automotive OEMs, Tier 1 and the same fro aerospace, electronics and other industries. You probably use a lot of products designed with FloEFD and FloTHERM.

Thank you Boris for your comprehensive reply.
Yes I agree that Flotherm is better suited for electronic cooling because it was created as an electronic cooling package in a first place. What I didn't like in Flotherm (prior its XT version) was the clumsiness in dealing with complex geometries and the need to "clean" them for successful meshing.
We have an Engineer in our company who does CFD for living and he has tested the new Flotherm XT. He said its interface and integration with Solidworks is very similar to SWFS and is much better in terms of user friendliness than the previous versions of Flotherm. He said that if we were to order a CFD package now we would have probably ordered Flotherm XT.
He also compared Flotherm XT results with experimental data he had of a product he designed, modeled and measured (it was a large enclosure with a lot of electronics inside generating heat and cooled in a natural convection way). The only thing he didn't like in FT XT was that in order to achieve very good correspondence with the measurements he had to use the finest meshing option of FT. He said in this case the amount of mesh cells generated was much larger than of SWFS (with a mesh resolution factor of 4 or 5) and the computing time was longer (than of SWFS). There was a noticeable difference in the results when using the courser mesh options of FT XT. He said the SWFS results didn't vary so much with the mesh resolution factor (and didn't deviate so much from the real measurements). His summary about FT XT was as follows - very big improvement (compared with the previous versions of FT) in terms of user friendliness and CAD integration, stable and fast, but should one always go for the finest mesh option to get realistic results.

Yes, FT had its weaknesses with complex CAD geometries that's a reason why there is FT XT now. But for anything like Server racks which are square shaped and all the electronics in it FT is unbeatable in scalability to the whole rack and Speed.
Well it's the very first Version of FT XT and it will improve for sure ;-)
But it is nice to hear that he liked it :-)

Very strange comment to make. Ease of use is very personal matter, having used CFX4, CFX5, Fluent, Cobalt, Tascflow, Starccm+ and few more codes, I think Fluent 6.3 is the easiest code to work with. I have not used their latest version but I would never put fluent in 'not easy' to work in category.
Could you give example of 1 or 2 solvers that are really easy to work with. And also why do you think they are easier to work with. I am interested in reasoning as to what makes them easy to work with while doing the same with Fluent would be difficult.

You are right, ease of use is a matter of perception but I think you would agree that there is a difference of using a cell phone from back in 2000 and an iPhone. Just for the user friendliness of how to use it. Also if you have had a windows phone with Windows mobile 6.x and an iPhone or Android from today you would also know that there is a big difference.
Both work but the one is just easier to use and more intuitive to get to results, wouldn’t you say?

Also when talking about Fluent and CFX we will have to distinguish between the old code and GUI and the Ansys Workbench. No doubt the Ansys Workbench was a big step in user friendliness and that’s exactly my point. FloEFD or SWFS had a nice GUI since the beginning in mid-end of the 90’s where Fluent and CFX (compared to that) looked like a beta from the GUI point of view and this is where “easy to use” comes into play. Easy to use means the GUI which means Graphical “USER” Interface so that’s what the user is actually using and if there is a user interface that is not really user friendly then it is hard to call it easy to use. Yes, it is still better than manual coding like one would do when using OpenFOAM (not using any GUI programmed by any company).

So to explain what easy to use really means…
It is to get from the first model you get to the (quality) results without a lot of work and struggle.
In FloEFD and SWFS you use the original CAD model and don’t have to use the negative fluid volume. You run through some basic settings such as unit system, what physics to consider (heat conduction, radiation etc.) some initial and default conditions and the overall project definitions are done. Then you apply solid material, heat sources and local meshes etc. to the CAD parts and you are basically done. The software fluid volume on its own by having an enclosed cavity where the in and outflow boundary conditions are applied to the corresponding surfaces. So depending on how large and how many boundary conditions the model needs a project is usually setup within a few minutes to a few hours in 80-90% of the cases. Some cases need some work on the geometry or adding some geometry that can be referenced as local meshes or to close gaps that are not really interesting for the simulation and therefore avoid unnecessary cells.
If the project setting is done you start the mesher and it meshes automatically. When he’s done you check if the mesh is good enough, especially in the important regions such as nozzle openings to accurately resolve important geometries or flow areas. That might take 2-3 mesh runs, but all automatically, and you are done with the mesh. The only thing the user has to do is to set a higher mesh resolution for the region that was not good enough meshed and then start the mesher again.
The solver delivers an extreme stable solution.

So really easy to use means to get from A to B without much work and time spend on the project. As I said I know a lot of OEMs and Tier 1 where experts as well as non CFD experts (in terms of numerical and programming skills) use the software and used also other CFD codes but get faster to results in a production environment than with the other tools.

So if that’s not easy to use then what is it for you? What do you understand under easy to use?
Of course if you are not a person that knows how to use a CAD system it would mean additional learning of the CAD system but FloEFD and SWFS is not designed for people who don’t use CAD systems but for those who do and know how to work with them. That’s the whole point of being CAD embedded, to leverage the very close connection to the original design and the quick changes you can do.
I’m not saying the software is for everyone or that a 10 year old can do your job now.