3.5 years overtake by Ansys: Where are the enhancements in grid generation
I still think that the taking over of Fluent by Ansys 3.5 years ago, was not a goal for the Fluent Users. Especially on the field of grid generation stagnation is the right word for what is happening. The "cfd-for-all" (like Bill Gate's MS OS and MS Office for all) strategy of the ANSYS officers seems to neglect existing users and the thinking is an FEM or CSM way of thinking respectively, where the time to result is in the order of minutes or even seconds and not hours or days. CFD aspects seem to be reduced to CFX-users' needs.
Who has already "real" experience with the new cfd-meshing tool and the new CAD-modeller? Am I just conservative when I think that GAMBIT is still the better alternative even if we all know that was never perfect or at least very good?:mad:
Behind the curtain...
Wow time flies… I didn’t realize it had already been more than 3.5 years. (June 2006)
Thanks for starting this thread, I am interesting in seeing the responses, I am sure there will be some who like ANSYS Meshing and some who don’t. I hope that people will give us specifics so we can solve the remaining issues and not just give vague complaints.
In the mean time, let me give you a look behind the curtain so some of the decisions can make more sense…
Gambit was already sunsetting before ANSYS purchased Fluent. It had many bottlenecks and the Fluent development team had already started replacing it with a new tool developed from scratch (new sizing function algorithm, new platform, etc.) ANSYS, with its much greater resources, should have sped up this effort, but there were some delays as we integrated teams and are refactoring the ANSYS platform (one step back so we can launch ahead without restrictions). There was actually a last release of Gambit in 2007 (after the ANSYS buyout), but that was already planned.
If you were in charge of ANSYS would you want all your acquired meshing technology development to continue as more than 5 separate teams competing or would you combine the teams and go forward with your best technology in each area?
After the purchase, ANSYS combined all the developers into larger teams and focused on creating the best meshing tool possible with the largest team of experienced meshing technology developers in the world. Many Fluent people took senior positions and key positions and we began to work together. It was decided that the GAMBIT developers would instead focus on the new product to get it ready for Gambit user migration as soon as possible. Honestly, I don’t think they thought it would take this long, but it was more important to do it right than fast. Key Themes included looking after the high end TGrid and ICEM CFD users (we are working hard to improve the technology underlying these tools because they will eventually go into the ANSYS MESHING Platform) and making sure that the new ANSYS Meshing tool was a much better tool than Gambit as soon as possible. At 12.0, it is still missing a few things (like meshing on faceted data or working well with zero thickness baffles), but in many ways, it is already much better than Gambit (parametric updates, better sizing function, better sweeping, etc.). We expect the technology migration to be complete for Gambit at 13.0. In the mean time, we have had a lot of positive feedback from users who have made the transition. You can see about the latest development by watching the recorded product update webinars on www.ansys.com...
The eventual plan (still a release or two away) is a single meshing platform with two product levels. The lower level, ANSYS Meshing, will support automated meshing routines, etc. but without all the bottle necks that Gambit has. We expect that Gambit users, CFX Mesh users and perhaps some ICEM CFD Tetra users will end up using “ANSYS Meshing”. The higher level tool will be built on the same interface (sold as an add-on), but will allow for much more interactive controls (more like TGrid or ICEM CFD). You will be able to easily move between ANSYS Meshing and its extended functionality.
If you would like to discuss this in much more detail, please contact me privately.
some new questions and statements
I am really surprised that someone directly from Ansys is responding. But this is a good and hopeful sign..
Let me first say that I am a "veteran" user of tgrid and gambit (and of course fluent). As I already mentioned I do not think that GAMBIT is or was the best tool for meshing one could imagine. I just got accustomed to it during the time I used it. Me and other users had several wishes of enhancement like a simple measuring tool or a working boundary layer algorithm like in tgrid and last but not least the possibility of switching between GAMBIT (CAD based) and a tool like tgrid (mesh based).
The last, if I understood you correctly, will be apart of the Ansys's new meshin suite? I ask you because I heard something else in the presentations during the last years: CFD- mesh will still remain a geometry based mesher. Doesn't it?
The most important at the end: I must admit that all my knowing about the new meshing suite is from second hand. But this is not my fault, since until now I as a Linux User had no opprtunity to use or test it. That leads me back to my first postage where I said s.th about Microsoft and the way of thinking inside ANSYS... It's still a windows world with small cases...
I am looking forward to hear from you again.
Several of us at ANSYS try to keep an eye on these boards when we are not completely buried. We are very interested in feedback on our work, and well detailed criticisms can be very useful for us. It is not the ideal environment because the users are mostly anonymous, but it is still valuable.
Thousands of people have spent time mastering these “legacy” meshers (including ICEM CFD, Prep-7, etc.), We understand that it will be a hassle for users to migrate, but that is not really enough reason to keep them all going as they are. It would just be too fragmented, inefficient, etc. for ANSYS to justify. Instead, we have taken on the challenge of making the unified tool so powerful and automated that it will be easy to learn and worth switching too. There is also an option to switch to the paid-up versions of these legacy tools so users who really love them can keep them forever.
At 12.1, ANSYS Meshing is still geometry based because the controls, contact, BoCos, loads, etc. are all placed on the geometry. So if you were generating a Tetra/Prism mesh, it would start a bit like Gambit where the surface mesh would be generated using the global sizing parameters, combined with entity parameters and/or a sizing function. But then the Prism or Volume mesh are generated using TGlib for Prism and Tetra generation. This is all seamless to the user. If your geometry is so complicated that you really need to edit the surface mesh before prism and/or edit the prism mesh before Tetra, then TGrid or ICEM CFD would be your best bet, just as they were previously. However, this is ongoing development. We are focusing on providing better automatic meshing to reduce your need for interactive editing and we are working on implementing “Extended Meshing” tools that will allow for things like interactive mesh editing, interactive hexa blocking, meshing from surface mesh, etc.
ANSYS has always supported a much wider range of operating systems than just Windows for most of its tools. This windows dependency issue was limited to our Workbench interface. This was a great time saver (development kit) when we thought of it as a designer tool, but it grew to much more than that. Since it is now our central framework, we have completely redone it (now Workbench II (2.0) at ANSYS 12.0) to support a number of advanced paradigms. Of course, this included removing that embarrassing Windows dependency. We are now finished removing the remaining windows dependencies from the applications; 12.1 (due out in mid November) supports several Linux platforms. The remaining Linux platforms will be certified by 12.1.1, due out several months later.
Our customers tell us that UNIX usage is all but faded out (less than a few percent still download Unix versions) and the remainder intend to switch to Windows or Linux soon. Many of the stand alone ANSYS inc. tools, such as the Solvers (Fluent, CFX, ANSOFT, ANSYS, etc.), ICEM CFD, and so on, will continue to support a number of UNIX platforms for a while longer, but will taper off on the less used platforms (we give notice at each release, send out letters looking for feedback, etc.). Mac users have occasionally requested builds, but it is just too rare to make a business case for. I expect that most engineering software will consolidate resources on Windows and Linux.
Well here is some thing in India.Three and half years back when Ansys. tookover Fluent Many and many veterans from Fluent started moving out. Now what would say about it.This is all about real fact.
Looking back at Fluent as a company the technical support and eagerness of people to resolve customer problems was Priority. Once Ansys acquired it i real don't understand what went wrong. Things have changed.
My personal view Ansys has a culture more over like a CAD Company. where once you sell your product get you r money never think of any further assistance. I am sorry if it hurts you. Second point as a Fluent user i feel Ansys just wants kill Fluent and its support softwares slowly by poisoning (Adding up all the Fluent feature to CFX and then gradually stop Fluent as a software and make people use work bench) That might be the Company vision in creating a highly user friendly and bundled multiphysics software.
But i would say it is really hard for Fluent users to accept this change.
Point by point...
Let’s take these point by point…
First, whenever one company buys another, some changes are inevitable and some people always leave :(. Many leave over what they think will happen, even if *it* doesn’t. Others leave because they feel like they lost status, etc. I was part of ICEM CFD when we were acquired by ANSYS. That was the first acquisition ANSYS made, so it was a bit rough and I understand being on that end of the irrational fears about what “might” happen, however, I am very glad I stuck around. The Fluent acquisition was much smoother, ANSYS had learned a lot about how to properly acquire a company (practice) and so it went much more smoothly, also Fluent was so big that it felt much more like a merger than an acquisition. The fluids people like to joke now that they outnumber the structural people at ANSYS now.
Worldwide, vast majority of Fluent people stayed. My manager, the VP above him and more than half the people in my product management group are from the former Fluent organization. Across the company, you can find “Fluent” people in many key positions and the Fluent culture is definitely part of the ANSYS DNA now as we are all pulling together. India is a special case because it is so easy to find another job that it was easier to leave than adapt, but again, in India, our beautiful Fluent Pune office was kept and the ANSYS office people (much fewer people) were moved there. I work with ANSYS (Fluent) India regularly, and got to visit two of the offices last year. The remainder of our India staff seem to have acclimated to the new company and are pushing ahead with good customer support again (sorry if there was a bumpy spot ;^).
You would have to get the numbers from our corporate reporting (publically traded company), but most of ANSYS revenue is also from recurring business. ANSYS is now permeated with “Fluent” people (who, after more than 3 years, are now ANSYS people ;^) Also, the global head of Support and Services is also from the Fluent organization and has made many changes to improve the ANSYS Fluent customer support.
Last but not least… ANSYS does NOT want to kill Fluent. ANSYS paid $600 million for Fluent, more than 10 or 15 times what they paid for CFX. The plan is not to replace Fluent with CFX… Rather, the plan was to acquire the biggest and best group of CFD developers and along with the most developed CFD code in the world and to link that code with our other codes so that our customers could easily perform accurate multi-physics solutions. There has been a lot of cross pollination of our CFD development which has significantly improved both CFX and Fluent. Like the meshing, you could guess that ANSYS wouldn’t want to keep both codes running next to each other forever… But the final situation (still years away) won’t look anything like “ANSYS Killing Fluent.” It will simply be a closer and closer integration of the tools (along with extensive ongoing R&D), each focusing on their strengths and working together as one large team.
Also, to clear this up… ANSYS doesn’t expect you to “stop using Fluent and use Workbench”… Workbench is free. That would be bad for business.;) Workbench is a free framework that can link all the ANSYS tools (and other tools) in an easy-to-use parametric and persistent way… Inside workbench, Fluent becomes a piece of the puzzle. Within that piece, we have put in a better graphics window and a model tree, but your old Fluent controls are exactly the same along the top (two sets of controls, old and new). This is all connected to CFD Post which is much better than Fluent Post.
You can see demos of Workbench in the www.ansys.com/demoroom/
Somthing really changed with Fluent after the theacquisition!
Yeah, that's really true, udayrg, it is hard for Fluent users to accept this change.
Concerning CFX I feel the same. It seems to me that the former CFX- responsibles, who are now the chief managers for ANSYS Europe, somehow enjoy that the "cfd-market-battle" was won after all by ANSYS-CFX despite the hopeless situation they were faced to during the years before the overtake: CFX Look, CFX mesh, CFX Post CFX turbulence models CFX philosophy: What is really the advantage of saving results, loading them in another tool in order to assess whether I can /should go on with my run or not? Imagine you run at the RAM-limit of your machine: You have to save the results + case + stop the session; read them into a post-processor and stop that session; lead result and case again to continue the calculation. You don't like that then you don't like CFX.
If european right had been deciding the merger would not be permitted since CFX and Fluent together share clearly more than 50% of the european market. This is the best way on achieving a monopoly. And this isn't really good for customers, since competition gets lost.
Ansys' CEO's vision is: ANSYS products on every PC (like windows). Everyone shall be able to carry out CSM/CFD calculation like users work with Excel.... Multiphysics (what does it realy mean, cfd is already much more multiphysics than all parts which now belong to the ansys portefolio) is the way to go..
But who cares on all those engineers fighting with "simple:D" "single"-physic problems of meshes with million of cells which need more than 4 GB RAM, need good graphical performance and at first need fast and robust and usable mesh-generators? (why to the hell is "grid" changed for "mesh "? Is this an advantage for me? Mesh seems to be s.th. surface-like, not a volume; at all a result of CAD or FEM thinking?) - There is too much simple case holy world thinking at ANSYS at the moment.
I hope that all those 3 yeras before promised new technologies will finally work and really help us in our work. I am yet not convincd they will.
I will pass these comments about CFX/Fluent to the Fluent product manager. I am sure that as fun as MultiPhysics between tools is, there are still thousands of customers who stay within Fluent or CFX and are interested in optimizing just their task.
As for a Monopoly, there are still plenty of options out there, including Star, Exa, OpenFoam and hundreds of others, to keep ANSYS on its toes. We are #1 because customers have chosen us, but we can't assume we can stay there without constantly working at it.
Making this software takes time, hundreds of man years and hundreds of millions of dollars have already been invested, and it gets better each release.
Keep pushing the software and keep pushing us.;)
As you said previously about ICEM and GAMBIT, that it was "too fragmented", why this point of view dosent apply to FLUENT and CFX case (CFD softwares)?
I am a user of fluent for quite a while now, and I am really not happy with this new version. At first I was open to the idea of something new, I was constantly looking for beta versions and trying it, but at the release I was really disappointed.
I was the first in our lab to "upgrade" to the new version and ANSYS new installation "bureaucracy" really made me mad... It was so easy to install FLUENT before, but now or you donwload a bunch of pieces (which really dosent work very well, at my experiences) and put them together or you have to download the whole dvd...
The upgrade in the GUI was something nice, made some tasks easier, but why change the name or place of some things?
Other point I am REALLY concerned is about documentation. I already used CFX some years ago (coincidently to move some routines from CFX to FLUENT) and I had to go through CFX "manual" and the FORTRAN reference and it really is poor. I dont known how many times I went in cyclical researches, how many times I had to call the local support (happily a friend) because I just couldnt find what I wanted... FLUENT on the other hand has a documentation that is near a novel. It is clear, straight to the point, has lots of examples, mostly complete. It improved my english (no kidding). It would be really disappointing if FLUENTs documentation starts to get poor too.
Still at the previous point, do you known about any plan about a scheme documentation? It has been years without a official one, old rumors said that it was because they were switching to python (what I thing is a good idea, btw) thats why they didnt developed any. Does it proceed?
To Draw some inference from this discussion Ansys should seriously take note of the users opinion before they decide to Scrap out things which are unique in its own credentials of application and features.
Its going to be like more :rolleyes: unsatisfied users making the user base sink..:eek:
And further on................
I hope these discussions do reach to the ears of policy makers for "Ansys Product Designers"
I agree in all three points with you.
The fluent download was easy and efficient (except for one server..). Yet I still cannot download from the Ansys' download center but have to ask my provider to put the files on his ftp-server or even better tell him to send the whole CD.
ANSYS' download page is completly confusing. You have to download things you'll never need. The installing process is s.th for Linux experts. Up to Fluent 6 and Gambit 2.4, tgrid 5 installing was not realy more than running a simple script which made nothing more than extracting files of a tar archive. On the windows platform up to fluent 6.3 the solver software came as a "mobile" version where you need the "EXE"-files only and no registry entry was necessary nor admin rights. Now everything changed.
2. The new GUI: grids are now meshes and scm,jou do not work anymore..:eek:. The postscript hardcopy comes with a bug. The preview doesn't really help. The autosave saves case where it should not...
The devellopers seem to think that fluent users never would like to have two graphic windows of different session one beside the other in order to compare results. Or for easier copy and paste: arrange the open office window adjacent to the windows TUI... The question must always be: What functions better? And not what looks better?
Something positive: OpenGL finally works on Linux 64 bit. Only zooming by box seems to be slow. That makes it easier to handle large models.
3. I am really pacified that someone thinks the same of Fluent's and CFX's documentation like me. There is another thread here somewhere (unfortunatly I did not bookmark ist) where the big majority really maintained that the CFX-Documentation is the better one. I supposed them to be from the former CFX offices or somehow related to CFX.
I had a look on the CFX-Post manual. That was sufficient to get an image.. It is really confusing from my point of view.
Decided to remove.
Bravo, Barvo! Haven't read the Fluent thread for while and now I see the posts from old Fluent users regarding the lousy new ANSYS paradigm. Totally agree with your points and finally found a place to release my frustration.
It is absolutely right that ANSYS is taking away Fluent's strengths (simple to install, use and find) by replacing them with the old CFX paradigm (complex, difficult and frustrated). It seems like a perfect storm for Fluent's future. The funny part is that all new users only have the choice for the new paradigm so they will never know the better old ways of Fluent. This apparently will set a lower bar for new users and thus can be easily improved upon, ha!
May be I am too sarcastic, but one can definitely tell ANSYS is way more bureaucratic and don't care as much about customer support as long as they are paying for it. That's why Star CD is jumping on the Fluent-decaying wagon and grapping Fluent's customers like crazy. Now I kind of regretting to say "Yes" to the questionaire Fluent sent out to users when they were going to be purchased by ANSYS. This is the first-hand experience dealing with a potential monopoly in the CFD market. Users beware!
I agree with some of Your points, especially if we compare Fluent/Ansys tutorials/books/materials... I have never read all of them, but they was/are excellent source of knowledge....
I do not know anything about red tape or bureaucracy (frankly I do not know what that means in CFD surrounding ;) cause I am only an user of software not a buyer) so I cannot agree or disagree with any of You.
But... when we/I compare Gambit and new Workbench enviroment there is a great gorge between them, ok I spent last year working on journal files, and probably I cannot use it here, but the simplicity with creating geometry and changing it without hassles is the biggest advantages for me and my colleagues.
I am gratefull for the new look and maybe there are some headache with "red tape" but they do not bother me so much.
I am keep on watching this thread regularly and at last sharing my opinion also. As we are using commercial code, there is no chance to escape. Ansys is planning to merge all meshing package and make one big suite. However, personally I think that there are lot of benefits by merging and switching over to single package. All meshing softwares (ICEM CFD, Gambit, Tgrid, Ansys meshing) have own merits and demerits. If all good features are added into the ansys meshing, then it is really good to use.
This is an old thread, but I was recently asked about it again. So a brief unofficial update...
ICEM CFD is doing well. Some time in Spring of 2010, it was decided that ICEM CFD was a very powerful tool set that would be very difficult to replace. We also had a number of very big customers argue for keeping ICEM CFD as it is...
As a result, ICEM CFD is a long term ANSYS Product. I have rough plans stretching out for 5 years (which is as far as we plan). Of course, for the benefit of those who do want to switch to ANSYS Meshing, we are still transferring plenty of technology to that product also.
There were a number of enhancements for 14.0 (released in Nov 2011) which you can read about in the update presentation available on the customer portal. Many of those enhancements were in the area of "MultiZone", but there were also a larger than usual number of interactive Hexa improvements.
For this next release (14.5), we are working on a number of interesting things, including the option of using ICEM CFD in the Workbench schematic and bringing advanced methods from ANSYS Meshing (including the sizing function and perhaps Cutcel) into ICEM CFD. Again some hints about these plans are included at the end of the 14.0 update.
As for naming... We still have our traditional ICEM CFD products, ICEM CFD Tetra/Prism and ICEM CFD Hexa. These each have a lot of overlap in terms of geometry tools, surface meshing tools and mesh editing tools. However, the Tetra/Prism one excludes the hexa tools and vice versa. These are each stand alone tools and each include the output options to over 100 solvers (CFD and FEA).
The full tool, known previously as simply "ICEM CFD" has been superseded by a new product called "ANSYS Extended Meshing" which gives you everything in ICEM CFD plus everything in TGrid. This new tool is an add-on to other ANSYS products (can be used with just about any other ANSYS key). It can also be used with an ANSYS Meshing key and that particular combination (ANSYS Meshing + ANSYS Extended Meshing) gives stand alone ICEM CFD capability for the same price as the previous ICEM CFD product. (Superseded = more capability (nothing missing, plus TGrid) for the same price as before).
Anyway, have fun with it.
So title can be changed to 5.5 years overtake by ANSYS :D.
I would like to admit one thing, that spending many years with GAMBIT (at least six years), it was difficult to learn new software. But it was must do as GAMBIT lacked many meshing techniques which are necessary to create high quality meshing such as O-grid (full O-grid, Half O-Grid aka C grid and quarter O-grid aka L-type or Y-Type) and easily adjusting to the underlying shape. Also it is lot easier to change the shape of the edge (What a coincidence ICEM also use same terminology- the edge for line). It is easily to slide the vertex, whereas in GAMBIT every thing is fixed. Also you can adjust the geometry to meshing, meaning it is not necessary to use the same topology as geometry forces specially at leading edge of airfoil or blade.
On the other hand we have lost several things, which I hope shall be there soon.
1. some geometry options such as reflection in the local direction with few clicks.
2. Net surface command.
3. Surface meshing (In fact with these three process, We were able to recreate the whole geometry again in the GAMBIT with accuracy equivalent to original model. This process is not shown in any tutorial and invented by our team, thereby we were able to avoid the geometry cleanup and virtual geometry creation).
4. Visibility, in ICEM we are lost in so many lines, faces coming here to there.
But after all every thing has to change except change itself.
I never properly addressed the original question of this thread because it was so far off, increasing the number to 5.5 is just further off. There have been many many enhancements in the meshing technology at ANSYS over the past years. Yes, Gambit had some good stuff, but the new technology is better. Look at all the TGrid Cutcel or shrinkwrap development. Look at all the Delaunay surface meshing improvements or the MultiZone development. There have been some necessary paradigm shifts that some people have trouble absorbing, but I am sure people said the same thing when airplanes took over from trains.
Those same people who developed Gambit have been working for years on this new technology, along with many other developers from the other ANSYS meshing products. The vast majority of Gambit users have switched over (I can see this by looking at the licensing migration from legacy Gambit to ANSYS Fluent).
It would be a mistake to think that a big company like ANSYS (the largest simulation company) wouldn't be putting many man years of development effort into each release. I think the conservative estimate for R14 was 600,000 man hours over R13. Of course, that includes solver development, but that is still a very large number; roughly equivalent to 300 man years of development (ANSYS actually employees ~2000 people (not including distributors and partners), but I think just under half are in development. Of those, we can't count time spent on email or in meetings as "development time".) It also doesn't include the many hundreds of thousands of hours of testing, planning, designing, etc.
If you actually want to know what has gone into each release, you can find the update presentations on the ANSYS Customer Portal or you can attend one of our many UGM's held around the world...
As for your particular concerns...
1) Reflection is no problem in ICEM CFD... Go to Transform. You can reflect, copy/rotate, transform, scale, etc. It only takes a few clicks. When you do a transformation on the blocking you can have it automatically do the same for the associated geometry...
ANSYS Meshing has its own options (I am less familiar) under the "Create" menu. It can also handle "instances" automatically when they are specified in the CAD system or DM. You don't need to even specify the transformation, it just happens automatically.
2) I don't recall what "Net Surface" is, but perhaps if you tell me a bit more, I can tell you the equivalents.
3) ANSYS Meshing has inherited much of the surface meshing technology from its underlying sources. Gambit users can use the descendant of their sizing function (plus years of R&D to give beter control and quality) with the descendant of their delaunay triangulation surface meshing (lower memory requirements, higher quality, robustness and speed). And of course, prism has had many man years of development since Gambit (but that old gambit preview was really nice).
ICEM CFD has surface meshing with a variety of different methods. Maybe I am just not sure what aspect of surface meshing your are talking about. One comment users have made is that they would like to use a sizing function with our blocking to control distribution in a more global way... That is planned for 14.5. The Sizing function does already work with our patch independent surface meshing in ICEM CFD.
4) ICEM CFD experts know how to control visibility. I usually only display what I need. Gambit simply didn't have any way to separate the mesh topology from the geometry topology. ICEM CFD does. This has many advantages (which you hinted at). Of course if you turn it all on at the same time, you have more on the screen. But your don't need to turn it all on. It frustrates me to see new users displaying all the geometry (curves, surfaces, points) along with all the blocking (faces, blocks, edges, etc.) when trying to associate edge to curve... In that instance, all you need to see are edges and curves, turn everything else off. If you are still having trouble seeing, you can turn of curve parts or adjust the index control reduce the edges that are visible... You can also right click on any geometry entity and blank that individual entity if it is in your way.
ANSYS Meshing also has some nice visibility options, such as the clip plane.
I have been using GAMBIT and FLUENT for more than 8 years, now I am starting to learn CFX. I have one question, why there are so many very special words in the software settings? for example "inflation" in DM meshing why don't use boundary layer? since every CFDer is first a fluid dynamicer:D and BL is a much familiar word!
another thing is that why not combine the good points of gambit and DM?
|All times are GMT -4. The time now is 05:38.|