CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   HARPOON Evaluation (http://www.cfd-online.com/Forums/main/11304-harpoon-evaluation.html)

Tang April 25, 2006 11:35

HARPOON Evaluation
 
This is just a discussion about HARPOON. I have read that HARPOON is fast, blah, blah, blah. Who cares how fast it is if it cannot produce the kind of mesh you want. Personally it has a long way to go to compete with the quality type of meshes created by ICEM, GAMBIT, or HYPERMESH. These meshers are industrial proven. I am currently trying to mesh a case where there is a jet plume exhausting into ambient air. Now, as you can see this is a very compressible case. The problem with HARPOON is that it can create a really coarse mesh in the interior. It also isn't a pure hex mesher. You have some surfaces that HARPOON automatically creates tets for. And creating boundary layers has proven to be a challenge, meaning they don't come out how I like them. Also, there is the memory problem. I can only get to 3.4 million cells after which I get a box stating to use the -winmem option. After selecting this it won't even attempt to mesh. It crashes instantly. We also meshed a combustor swirler that turned out good and matched well with experimental data, but that is a incompressible case so we don't have to deal with the boundary layers. HARPOON uses the octree method like ICEM to create the mesh which I don't care for since the meshed surfaces can have unequal mesh sizes. The octree method basically grows cells from a predescribed box inward to your geometry once that is done the outside mesh to your geomtery is eliminated the the mesh inside your geometery is refined and smoothed to whatever transition you gave. It doesn't help that my boss thinks this is the next greatest thing and will replace ICEM and the other meshers we use. Now, we come to the discussion on geometry import. I'm sorry but importing stl files is just wrong. It is the worst geometry file out there, in my opinion. Then you have to go and separate faces to get what you want and the fun is if you have two faces that are at a high angle and HARPOON separates a zillion faces. You then have to go and find those faces which can take a lot of time. I have gotten to point where I just create a decent ICEM mesh, save it as a .msh file and then I import that into HARPOON. That way I have the families and surfaces I want. Once I get my HARPOON mesh I export it as a FLUENT mesh file, import that into ICEM along with my .tin geometry file and project the nodes to the correct geometery. I have heard HARPOON is working on getting a package that is able to import parasolids. This may make it alot easier to seprate the faces you want. HARPOON has a ways to go to compete with ICEM and other meshers. Though if you are a grad student working on simple geometry cases for classes or thesis then I would say this is a pretty good package, but for industrial use it has a ways to go. Currently, I would give the mehser a C+.

Charles April 25, 2006 15:17

Re: HARPOON Evaluation
 
Have you given this feedback to the developers, and what was their reply?

Jim_Park April 25, 2006 15:30

Re: HARPOON Evaluation
 
"We also meshed a combustor swirler that turned out good and matched well with experimental data, but that is a incompressible case so we don't have to deal with the boundary layers."

Are you saying that incompressible means no boundary layers?

Ben April 25, 2006 15:38

Re: HARPOON Evaluation
 
I havent used it myself but I have seen a lot of harpoon meshes in action (first hand) and I am yet to see a good one, yes its very fast but the cell quality is terrible from what I have seen. Personally I like the new STAR-CCM+ mesher, its nice and quick, very customisable gives good quality mesh and can do polys. I have seen ICEM in action and it looks good do although I havent seen ICEM hexa and I am yet to hear anything good about gambit

Richard Bardwell April 26, 2006 05:41

Re: HARPOON Evaluation
 
Thank you for your comments.

My response to your points are below (S-statement; A-answer):

S:"it can create a really coarse mesh in the interior"

A: User error. This is controllable using the max volume size option or by modifying the cell growth rate. This is all described in the user guide.

S:"It also isn't a pure hex mesher"

A: Correct. If a 100% hex mesh is required Harpoon is not the tool. Use ICEM-hexa and spend a lot of time meshing

S: "You have some surfaces that HARPOON automatically creates tets for"

A: I do not understand this point

S: "And creating boundary layers has proven to be a challenge, meaning they don't come out how I like them"

A: BLs will be comparable to ICEM if 100% tet mesh is used. If hex dominant, the BL will be smaller because it is much harder to control BL growth with quads in the mesh

S: "I can only get to 3.4 million cells after which I get a box stating to use the -winmem option. After selecting this it won't even attempt to mesh. It crashes instantly"

A: All bugs reported to us are fixed. Clearly this has not been reported to us as a bug

S:"HARPOON uses the octree method like ICEM to create the mesh which I don't care for "

A: Octree has pros and cons. If the user wants to create your own surface mesh then that is fine.

S: "Now, we come to the discussion on geometry import. I'm sorry but importing stl files is just wrong"

A: Not according to ICEM. Their preferred import is also STL. This stops Harpoon being blamed for poor IGES files. We have had many years experience of data import/export. We decided that the STL route was the most robust.

S:"Then you have to go and separate faces to get what you want and the fun is if you have two faces that are at a high angle and HARPOON separates a zillion faces"

A: Separate twice with different angle settings will solve this problem

S:"Though if you are a grad student working on simple geometry cases for classes or thesis then I would say this is a pretty good package, but for industrial use it has a ways to go"

A: Absolutely not true. We have a large number of commercial companies who use Harpoon in aerospace, automotive and environmental industries.

If you have a problem with Harpoon, you should contact your distributor. Any bugs you complained about will not be fixed by discussing them on cfd-online.


opaque April 26, 2006 09:44

Re: HARPOON Evaluation
 
Dear Richard,

I understand Tang is an upset user at this point, and it is using the wrong route to get the problems fixed.

However, I hope your statement

"If you have a problem with Harpoon, you should contact your distributor. Any bugs you complained about will not be fixed by discussing them on cfd-online."

meant that you (Harpoon developers/support) do not monitor cfd-online; therefore, it is not the forum to get your problems fixed. But, that you (as a Harpoon group) would take a look at those already mentioned bugs, and assuming they are legitimate bugs you will fix them..

Thanks, Opaque

PS.. By the way, I am not a Harpoon user at this time.


Richard Bardwell April 26, 2006 10:06

Re: HARPOON Evaluation
 
Opaque

Of course. We fix all bugs when we get them. Why would we want Harpoon to have bugs in? In most cases, a detailed description and model is required to replicate any issues


Kent Misegades April 26, 2006 10:37

Re: HARPOON Evaluation
 
Dear Mr. Tang and others:

We are the distributor and provider of technical support for Harpoon in the United States.

Harpoon is an excellent tool for a certain class of problems, however we would be the last to claim that it handles all models equally well. Harpoon has been a great commercial success for its developer, SHARC, and for ourselves, probably because we are careful not to recommend it to those for which there does not exist a good match between their requirements and its abilities.

It is unfortunate that Mr. Tang did not contact our technical support group for assistance; his misunderstanding of the code's features could likely have been avoided.

We also do not know of the other two respondents to this thread, and we can only assume that they are not active Harpoon users.

We'd be happy to work with anyone to explore how Harpoon can be used to gain dramatic improvements in modelling efficiencies for CFD and FEA.

Sincerely,

Kent Misegades CEI Inc. Apex, NC, USA kent@ensight.com Tel. +1.919.363.0883 www.ensight.com www.sharc.co.uk

Tang April 26, 2006 13:31

Re: HARPOON Evaluation
 
I apologize if I have offended the makers of HARPOON. I searched this forum and found no opinions about the software so I decided to create one. I just want to hear the opinions from other users. And if they had similar problems and were able to solve them ,then I as well as others can benefit. I understand where the HARPOON makers are coming from. The only way to make a product better is to communicate to the maker. I also understand HARPOON is an infant when it comes to competing with other meshers that have been out longer. I am impressed with the speed, and through time and customer input this product will get better. The one thing I have difficulty beliving is that ICEM prefers stl files. I have not read this statement in any of their user manuals. I would think they would prefer parasolids.

Kent Misegades April 26, 2006 14:32

STLs, etc.
 
On the issue of STLs - we all love to hate them in a way. The format is simple and ubiquitous, which is what ISVs like ourselves prefer. Of course, they are also very limited, since they simply describe triangular facets. We have learned that most codes that export STL files, in particular the leading CAD packages, can, when the proper parameters are selected, produce very good faceted STL representations of the surfaces. Unfortunately, the default settings in some CAD packages lead to poor STL files, with very high aspect ratios and other typical flaws. You will find an FAQ on our web site dealing with this issue, "How to get the best STL for Harpoon" http://www.ensight.com/FAQ/faq.0086.html

Some industries make very heavy use of STL files. In fact, one of the world's largest auto makers has standardized on STL files for all their geometry as imported into CAE codes. This has required the development of good interfaces betweenm CAD and quality STL, however it appears to work well for them.

The aerospace industry, where one is likely to find the heaviest use of ICEM, remains a major user of IGES files, in spite their antiquity and many known issues. However they also have some two decades of experience dealing with IGES, or as we say "you know what you know".

For further assistance using Harpoon, we recommend you direct your requests to harpoon_support@ensight.com My colleagues will be pleased to assist you.

Sincerely,

Kent Misegades President, CEI Inc.

Tang April 26, 2006 14:44

Re: STLs, etc.
 
Thank you. This has been very helpful.

Richard Bardwell April 27, 2006 02:01

Re: HARPOON Evaluation
 
Tang

I heard the statement that "ICEM prefers stl data" from an Icem european sales guy at a NAFEMS meeting about a year ago. It caused some debate at the time since both Fluent and Star were pushing their direct CAD interfaces. I think the sentiment was that the STL would always import correctly, whereas the various CAD interfaces would be problematic especially on large, complicated data

jason April 27, 2006 14:18

Re: HARPOON Evaluation
 
I would normally not respond in a thread like this, as the original post appears to be the product of the standard mesh/grid generation frustration when things are not working as one thinks they should. But what exactly does

"We also do not know of the other two respondents to this thread, and we can only assume that they are not active Harpoon users."

have to do with anything? I hope that you are not suggesting that I shouldn't respond to this thread unless I've used your program. This is a discussion forum - the original post was maybe a little over the top, but I would bet that a lot of the regular visitors to CFD-Online know exactly how the original poster feels. I know that I do. Whether people have used the program or not is irrelevant.

Jason


All times are GMT -4. The time now is 22:41.