This Week in CFD

CEI Is The Exclusive Visualization Sponsor of FLOW-3D Events

CEI will be the exclusive visualization sponsor of Flow Science’s FLOW-3D European Users Conference and FLOW-3D World Users Conference. The announcement of this exclusive deal comes shortly after the announcement that CEI’s EnSight postprocessing software will be integrated with FLOW-3D. At each event CEI will offer training on EnSight. The integration of FLOW-3D and EnSight is scheduled for a beta release this autumn.

Barriers to CFD adoption poll results from Symscape.

What’s Holding Back CFD?

Virtually half of respondents to Symscape’s poll about the barriers to further CFD adoption cited CFD software price making it the top culprit. Second with 17% was software complexity.

Two thoughts spring to mind. First, I wonder how the survey respondents would rate the complexity of commercial versus open source CFD tools. Second, I can’t help but think of Clayton Christensen’s The Innovator’s Dilemma. You can watch a brief video illustration of the dilemma here.

Do you agree with the poll results?

News in Brief

Leadership Changes at Tecplot

Tecplot announced the promotion of former COO Rich Stillman to President and CEO, replacing company co-founder Mike Peery who moved into chairman of the board and “product czar ” roles.

According to Tecplot, the change is in response to current and anticipated growth due to the recent release of Tecplot Chorus, a new product that created the field of simulation analytics. Stillman has extensive experience from prior executive positions at Coinstar and Ivey Imaging.

Simulating Sky Diving

This sky diver simulation was performed by Creare, Inc. using ANSYS Fluent. Image is from FlightGear.

The U.S. Army has a keen interest in knowing whether their airborne troops can parachute safely to the ground under a wide variety of circumstances including the amount and type of gear they carry. As you can imagine, the wake downstream of a sky diver is affected considerably by the configuration of packs they carry and in turn the parachute’s inflation is affected by that wake.

Creare, under contract to the Army, developed a Fluent-based toolkit for simulating a skydiver in freefall. The results of those simulations were visualized using FlightGear, an open source flight simulator.

Print Your CFD Solution in 3D

Read this article and watch this video about 3D printing and then think about printing your CFD solution in 3D. Would that be a valuable capability in FieldView, Tecplot, and EnSight?

This entry was posted in Events, News and tagged , , , , , , , , , , , . Bookmark the permalink.

27 Responses to This Week in CFD

  1. Jeff Waters says:

    I would guess that the Symscape results could be skewed based on the open-source focused audience attracted to the Symscape site. (By the way, I am personally a Symscape and OpenFoam fan, too).

    If you consider software to be $2 worth of compact disc material, and immediately relate it’s price to the cost of your car, you will have also have a skewed view.

    If you look at it from a a beancounter perspective, I think it still often makes fine business sense. How much can you save in physical prototyping, windtunnel testing, and time to market in a year? $500k-$2M worth? Even at $200k in software licensing, I’d take that deal.

    Here’s my highly scientific poll of 1 (me) for “What’s the biggest barrier to CFD adoption?”:

    35% Slow overall turnaround time
    35% Skill and time invested in geometry cleanup and meshing
    20% Not putting specific milestones for CFD in project plans / Gantt charts
    10% All others

    Having said all that, this is definitely an industry ripe for the Innovator’s Dilemma. Price will likely play a roll in that, but not the biggest.

  2. John Chawner says:

    Jeff. You’re right – the results are likely skewed toward Symscape’s audience. That’s not a bad thing because I don’t think anyone is claiming this to be a statistically valid poll with quantifiable margins of error.

    Regarding perspective, another one is to measure the value of CFD software relative to the amount of money in my wallet. Sometimes you just don’t have enough.

    And thanks for making meshing one of the top barriers to CFD. I guess I’ll have a job for a few more years. 😉

    • Jeff Waters says:

      Agreed, that is in fact another valid perspective.

      Yes, anything that can cut time/effort on the meshing side will directly impact that value in a CFD implementation.

      Finally, yikes… I wish there was an edit button here. I just reread my bit above. Atrocious grammar and spelling. Oops.

  3. Rich Smith says:

    Looks like the party started without me 🙂

    As Jeff and John have already said, I think the poll results are skewed towards the Symscape audience – otherwise I’ll need to have words with our marketing department – i.e., prospects looking for affordable and easy to use CFD. However, I don’t think the poll suggests that respondents think a fair deal would be to reduce CFD costs to less than a working wage. I believe what the poll reflects is the perception that current mainstream CFD (especially parallel solving) prices are out of step with the perceived benefits (that Jeff cites) to cost sensitive users.

    I’d like to think the CFD market is ripe for an Innovator’s Dilemma revolution, in fact the Symscape business model is in hot pursuit of such an event. Whether Symscape is to the mainstream CFD vendors what Solidworks was once to CATIA remains to be seen. However, I fully believe that there is a vast number (an order of magnitude more than currently use CFD) of under served CFD prospects who see mainstream CFD prices as a major barrier to further CFD adoption.

    As an aside, I’ve just posted a new poll “What is the most difficult part of CFD?” and straight into the lead is “Creating Mesh” – matching Jeff’s list and confirming John’s excellent career choice 🙂

  4. I guess I’m the odd man out. For me, by far, the difficulty is inadequate physics modeling and the difficulty in capturing the physics, even when the CFD code can model it. To me, turnaround time and gridding is just a byproduct of that.

  5. John Chawner says:

    Martin: Can you go into more detail about inadequate physics modeling and difficulty in capturing the physics?

    • OK, to start off I would like to say that “inadequate physics modeling” was not fair since, lets assume, DNS can capture everything. But, for a full geometry that is currently, and for the foreseeable future, impossible.

      The difficulty in capturing physics boils down to things like, transition, eddies, wakes, separation and/or reattachment points, tracking vortices, natural dissipation (i.e. viscosity), and artificial dissipation (discretization error). That’s for subcritical flow. Once you go supercitrical, i.e. shocks, that’s another level of difficulty. Sure, we can capture isolated shocks with maybe some shock on shock, or shock on boundary layer affects, but start mixing shocks and eddies or wakes, and you are in it deep.

      But, sticking to subcritical flow, I believe it has been only a few years since we’ve been able to capture drag on a cylinder over the entire Reynolds number spectrum. But I’m not really sure. I believe about 10 years back LES made inroads with capturing the drag crisis that occurs with a cylinder, but capturing drag at higher Reynolds numbers was an issue back then. But LES for an entire geometry is not practical. Unfortunately the circular cylinder is a poster child for some of the tricky aerodynamics.

      Of course we can not capture drag (without using DNS) on an airfoil at low angles of attack due to transition. So the drag bucket is another example. Of course it takes a lot of effort to capture the trailing edge stall and any leading edge bubble correctly.

      Another issue is tracking a vortex from one empennage/surface to another, or predicting the vortex sheet, (street) coming off a forebody past 5 degrees angle of attack. Agile missiles are an example of this. Here is a fun tidbit, as told to me by a pilot. The high angle of attack performance of an F18 A/B is affected by small manufacturing differences in the nose cones. The nose cone had to be flight certified for the F18. If an aircraft performed poorly the nose cone was thrown away and replaced with a new one. The flow field around the F18, under certain circumstances, was that sensitive. CFD (RANS) can not capture this. RANS is just not that sensitive to geometry changes due to lots of eddy viscosity. And we can not (I believe) do LES around an F18 at flight Reynolds numbers.

      Then there is the issue of calculating the loads on something which is embedded in the wake of another object or surface.

      Oh, cavities are a big pain and require LES I believe.

      Another difficult area is blunt objects with small fineness ratios (due to leading edge bubble reattachment, or lack thereof), Pipes/channels with multiple bends is challenging due to turbulence modeling (more than one single 90 degree bend). Any sort of object or restrictor in a pipe or channel is very challenging due to eddies and recirculation regions. Boattails are also very challenging (again LES territory). Then there is rotating helicopter blades. One can google HELIOS to find papers with rotorcraft comparisons. Again, challenging stuff.

      • John Chawner says:

        Martin: Thanks for all those details. One might get depressed reading it if you didn’t also keep in mind that we might call those problems the fringe of the envelope. There’s still a huge sweet spot where CFD is robust and accurate. And that sweet spot keeps growing over time. Because of that I’d like to think that people aren’t adopting CFD because they think it’s inaccurate. On the other hand, what you might be saying is that all the important stuff is still in that fringe area.

  6. What’s holding back CFD? Maybe it is the lack of funding (either external or internal) to expand into new markets (the underserved areas were we find lots of new customers). Perhaps the economy is forcing potential new customers (yacht builders, light airplane manufacturers, builders of new homes, sporting goods people) out of business (active or competitive) or to delay their development of new products until the return of their customer base. The goal of making a CFD user out of every CAD user might be delayed by the current global economy and manufacturing situation.
    I think that if (when) brought to their attention, the potential CFD customer would be amazed by the latest developements in personal/upfront CFD made over the last year or so.

  7. John Chawner says:

    Pat: Don’t you think that the current state of the economy is a relatively recent and (hopefully) temporary issue? One bit of conventional wisdom is that when times are good people buy a lot of CAE software because they can and when times are bad they buy a lot of CAE software because they have to (doing more with fewer people, being more efficient, etc.). Also, I’m not sure that every CAD user needs to be a CFD user – how many products really need to optimize their fluid performance?

    Your last comment points to the issue of perceived value. People think CFD is too expensive relative to their perceived value of it. As you say, making everyone aware of the latest advances may be a way to get them interested again.

    Thanks for your comments.

  8. John, you are correct. Hopefully the economy will improve soon and we can see more growth. Your point about the number of CAD installations being larger than the number of CFD/(other CAE) seats is well-taken.

    Do you have any recent numbers on the growth rate of CFD (over the next few years)? How does that compare to other CAE & CAD?

    I enjoyed your link to the video about the Innovator’s Delimma. In CFD, I think that CFDesign was such a company. It did not compete directly (on all grounds) with Ansys, Fluent et. al. but experienced a rapid rate of growth over the years. With CFDesign now under Autodesk, do you think the market is ripe for yet another disruptive innovator?

    • John Chawner says:

      Here are two recent sources of market size info. According to 01consulting, Mechanical CAE is a $6.5 billion market with a compound annual growth rate of 12% over the past five years. It’s called the most dynamic segment of the PLM market.

      According to a paper at the NAFEMS World Congress, CFD is a $700 million market that’s been growing at 13% per year since 2000. The authors predict a potential future CFD market of $2.9 billion. Also, they estimate there are 250,000 CFD users out of 10,000,000 engineers worldwide.

  9. I stand my original post at Symscape: The biggest obstacle for CFD or any new technology for that matter is cultural! The arrogance of management is, “I got where I am because I know what to do!” My retort to that is, “You don’t even know what you don’t know!” Exposure to FEA and CFD will change the way you think about everything without even looking at any numerical results! Most of the companies I encounter are still “using” simulation as a reactive tool instead of proactive one! As the new blood gets some experience and moves into management hopefully this will change!

  10. John Chawner says:

    Hi Woody. Thanks for your comment. What you illustrate about managers’ reluctance to try CFD does dovetail with the price/value argument. They don’t feel they need it because they don’t perceive it as having value beyond what they’ve already been doing. You are correct – once people see how CFD can add value to their engineering, then we’ll get more adoption of the technology.

  11. I somewhat disagree here. Sometimes manager got where they are because they know what they don’t know. It is the people who are below them that are sometimes “arrogant” because they do not know what they don’t know.

    Here are some examples of the uncertainties that muddle my thought process.

    Sorry Woodruff, I’m going to use the Bonneville Roadster as an example since the aerodynamics of such vehicles can be very cool! The top speed on the data chart says 325 miles per hour. The altitude at Bonneville is 4218ft. On a standard day 325 miles is equal to Mach 0.43 (Mach 0.42 at sea level). The area ratio for sonic flow (i.e. A/A_star) is 0.67. Of course I would be scared of any analysis! But then I don’t know the ins and outs of such racers. But I think this vehicle should be solved with a compressible solver. Yet I don’t see this mentioned on the design dreams web site. And I know if there is a shock on the bottom of this vehicle the aero is going to be unpleasant and any unsteady flow can destroy the vehicle, more or less, instantaneously.

    Next, I’m going to use Symscape’s images of solar panels on a roof top. I just don’t get it. Anyone who stands on a roof top knows the airflow is all over the place. The physics are tough. And I know CFD has a tough time predicting such flow features. So, where is the documentation (Engineers Handbook of Rooftop CFD) that shows how well CFD does and when and where it can be used? Personally, I thought civil engineering had all sorts of these regulations and handbooks. And, if one doesn’t follow those, lawyers tend to have a field day. Better to be safe than sorry. But, I don’t know the ins and outs of this.

    Yesterday I went to Mentor’s home page to look at their software. OMG, to find technical information on the code is a pain. When I type in “theory floefd” I get links to what seems like fluff embedded marketing white papers. I didn’t actually download any because to do that I had to register. Then when I dig around the web pages I could not find much real substance. It is a huge marketing machine. What engineer manager has time to read all this stuff? I thought I would puke. Personally, I found the experience very disrespectful.

    Here is another example from cfd-online of someone using FloVENT to model a HVAC system of a clean room. It illustrates, to a degree, the issue people have with Reynolds number, boundary layers, turbulence, and eddies. (NOTE: I’ll post the link next since I think it is messing up the post)

    I’ve also dealt with the FAA certification process for aircraft (outer moldline) components. Here is a fun tidbit. The people who sign off only like certain CFD codes and they do not explicitly, and up front, tell you which codes they like and, even more important, the codes which they hate. I feel sorry for the engineer manager who uses the wrong code, or the wrong analysis house, for their analysis. Kiss A LOT of money and time goodbye. The certifier will make you go back and do it again! And it is not a committee or anything. It is ONE person! And they know they have the power. And, I assume this is true for other fields. Oh, I know it is true in the aerospace industry where a technical person must sign off on an analysis. The manager really does not have much choice but to go with what that technical person says. (Well, the manager can sign off on it and accept personal responsibility) Anyway, that technical person, in general, knows what they are doing. They are not dumb. And, if one wants to see an example of arrogance (i.e., superior manner toward inferiors), you’ll see it there. And, frankly, a lot of times it is well placed.

    And, to a degree, all this may illustrate my own ignorance. And some would say I have a lot of experience with this stuff. But, I’m well aware that experience is sometimes yesterday’s news. However, eduction is a time consuming business. I’m certainly not up to date with everything, or even anything. Especially when one needs to wade through all the fluff and pretty, but possibly meaningless, pictures.

    So, everyone feel free to change my opinions on the above examples. I love to educated! I don’t mind being disagreed with. Show me explicitly, concisely, and in engineering undergraduate mathematical and layman’s terms, where CFD can and can not be used. And it is OK to draw sketches showing the physics. It doesn’t need to be all words and equations. Also, if I don’t see sketches, I gather the writer may not really be able to describe, or understand, the problem.

    Don’t get me wrong. I believe in CFD. I am even creating my own CFD code. So I can’t think it is bad! Maybe my problem is I have too much experience. Ah, to be young and dumb again!

    So, sure, CFD is very important. But, it is also a very sharp knife. You can cut and dissect a problem with it, or you can cut yourself. Sorry, I am repeating myself from past posts. So, some users will get very good results, some users will get garbage, and some users will get qualitative, rather than quantitative, results. And, it is up to the managers, and those they have to rely on, to decide which category they are in. I think this is a tough job, considering all the noise out there.

    Personally, I don’t see a big tidal wave of users coming. Sure, the user base will increase. But, I don’t think it will be jaw dropping experience.

  12. Forums/main/97589-what-raynolds-number-transition-laminar-turtulent.html

    Sorry, you’ll need to add www . cfdonline . com in front.

  13. John Chawner says:

    Martin: Your FAA example made me laugh because it’s so very like Uncle Sam. They’ll never tell you how to do things right, but they’ll certainly point out every time you do it wrong.

    • It’s not funny! 😉 OK, probably a laughter of frustration! NEAR, did get the code selection right by using OVERFLOW. But there are some big commercial CFD names which are not on the list. One of the reasons is discretization and handling of dissipation. And so much is just word of mouth. Very frustrating. And who knows if some of that is true or not. Unfortunately some of these codes are based on legacy knowledge/coding and people don’t dare mess with it. Pointwise has first hand knowledge of what it takes to update a code. Of course I’m not going to say anything negative about a CFD code in public because I don’t want a bent out of shape exec/lawyer giving me a call. :p And, I’m not even sure if my knowledge is up to date.

  14. Ooops, I made a mistake. I misread the Bonneville Roadster data. The max flowfield velocity is 325 m/h, not the speed of the car. I’m not sure what the speed of the car is.

    Sorry Woodruff!

    You see, I make mistakes! But, I did catch this one on my own. Of course not till after I opened my mouth.

    • Martin, you make some excellent points but if you look again at the first page of my webpage you will see two quotes: “All models are wrong but some are useful!” G.E. Box (1967) and one by me, “Without simulation you likely don’t know what you don’t know!”™ Anyway how precise does an approximation have to be to be useful? I think that is our real quandary!

      Getting customers to do empirical testing to correlate is another one! Trying to find verification and validation for real world examples is a real struggle! These days time to market seems to be the driver not necessarily engineering expertise!

      For the roadster, I am working with an Indy car designer with 15 years wind tunnel experience and drivers with equal seat time. So far they like the results which BTW are not on the webpage those are just the “Colors For Directors”! Even if we get into the record book that does not mean that we have proved that CFD is correct – only useful! We will sneak up on the top speed with on board data acquisition! And just to keep me honest I get a turn behind the wheel!

      I guess I should not indict all managers (since I have been one, too) as I have worked with some great ones but in my experience they have been the exception and not the rule!

      Now should I muddle on with the Mentor products or wait for your code to hit the market? 🙂

  15. John Chawner says:

    Summarizing the discussion so far…

    Symscape’s poll said the top barrier to CFD adoption was software price. Other culprits have been identified such as process issues (turnaround time, including meshing), technical issues (inaccuracy at certain conditions, modeling) and cultural issues (management reluctance, lack of user education).

    But Rich still wins because he’s the one with data while the rest of us just have opinions. 😉

    • Sorry, got to ask. What is your personal vote?

      • John Chawner says:

        As people at Pointwise will tell you, I’m a master at evading direct answers. But, I did go back to Symscape’s original poll and I think I would’ve voted for Slow Turnaround Time putting me (and Jeff Waters) in the vast minority of respondents (4%). My opinion is based on CFD being complex (another choice on the poll) which means there’s a big learning curve. I’m also biased toward simulations upward of 100 million cells which tends to complicate things. I think people hear and read about that and get scared off. Good, fast, or cheap – choose 2 of 3.

  16. I am discounting one of my CFD software over the month of March (to coincide with an incremental upgrade). If we sell more packages at the lower price, I send Rich a Starbucks gift card.

  17. Yes, I read the quotes. Can’t really miss it. But, I’ll be honest, I’m probably reading it differently than you. If all models are wrong and some are not useful, how does one know which ones are useful? Also, sometimes a simulation teaches one something one should not have been taught. Sorry, that’s just me and the first question is only rhetorical. Obviously, CFD requires an expert in the physics to answer these issues. And, therefore, to a degree, the growth of CFD requires the growth of experts. And, IMO, these type of experts are not exactly being pumped out of the universities.

    “Getting customers to do empirical testing to correlate is another one! Trying to find verification and validation for real world examples is a real struggle! These days time to market seems to be the driver not necessarily engineering expertise!”

    Exactly, except I’d like to add to the validation part. There are plenty of papers on the various aspects of the physics. One just needs to know what to look for and how to extract the information about “bad” CFD runs from the papers. In general, people don’t brag about the things that went wrong. Unfortunately the process is sometimes hard, for me a least, since sometimes I’m not sure what they are talking about until I hit the problem myself. An “oh! that’s what they meant” moment. Or maybe it just doesn’t sink in for me. Yes, I wrote the “Ivory Tower” email to John which was mentioned a few posts back.

    I like your example of the roadster. I don’t know what your design is, what you are trying to achieve, or what your knowledge base is. So I’m not really commenting directly on it. I can’t.

    But this is what I think I know, and this is not saying that you don’t know it and I’m definitely not lecturing anyone. But, it is an example of my thought process. So, lets say there is the possibility of a shock under the the car. It seems that the undercarriage is being used to generate downward force, thus the door is being opened for a shock. A shock is a discrete event, the only way to really sneak up on it is to know it is there. Unless you have instrumentation on the underside of the car which measures Mach number. Also, there is no way to really sneak past a shock. Every step past, is a step of uncertainty. Extending the envelope is pain. The existence of a shock on the under carriage means choked flow. You’ll go from suction on the underside to compression. This could happen very fast and the vehicle dynamics could be such that the system is unstable, at any time or any condition. I have serious doubts that you have wind tunnel time with this type of flow. Anyway, once a shock develops, a wind tunnel must be specifically designed to handle it, or I should say the choked flow. Solid walls will just make the situation worse. Unfortunately slotted walls also give erroneous results, though significantly better than solid walls. I am also skeptical seat time has prepared anyone for a shock on the underside. In general, I don’t think Indy cars experience it, though they might a little, at higher speeds Anyway, how does anyone prepare for it? One second everything is going fine and the next things are in pieces. Your car is not that much different than some stores which have been attached to aircraft. Of course other cars have gone very fast on the salt flats, so there must be a knowledge database there. Though, I’m not sure how similar those cars are to yours.

    Maybe you and your team will not hit that part of the envelope, but someone will. Of course, if something goes wrong, CFD might be blamed. That’s what Virgin Racing did, at least to a degree. But, maybe this is how the CFD envelope needs to be defined for those outside of the aerospace community. Not that the aerospace community is any smarter or anything. We’ve just been through the process already.

    So, why don’t you put some information on your web page that instructs people and shows off your skills with understanding the physics? Maybe that would help hook some of the smart managers.

  18. Pingback: Another Fine Year in Blogging About Meshing | Another Fine Mesh

Leave a Reply