Hello everybody!

I'm tired of employers who seem to base their whole candidate selection critera on whether or not you have run a certain commercial code. I wrote my own 3D unstructured NS solver as a PhD student - that should be more important than if I have experience from using a particular commercial code, but apparantly it isn't. Surely it wouldn't take me more than a few weeks to get used to a new commercial code.

Do you other guys have similar experience? I think this focus on commercial codes is very short-sighted.

Any comments from employers?

Dear Dr. Steven:

You opinion is very great. I have a similar experience with you. The aim is to solve a problem, no matter what you are using. If all of employers would ask candidates with some skills for commercial codes, professors in universities had better teach students in advance how to use FLOW3D, CFX, PHOENIX, FIDAP, xCFxD, ....., instead of how to solve partial differential equations.

If there is really the professor in the world, he must be the computer itself.

Ding, Yan
Doctor
Ocean Engineering Deparatment
INA Co. Ltd.
Tokyo, Japan

As a CFD Engineer working in the Gas Turbine industry I appreciate your concerns/ annoyance at present state of affairs.

I am not personally convinced that you can in fact learn how to use a code effectively in a few weeks - taking into account the pre/post processor, solver, debugging problems and, to be frank, the bugs/ frailties of a particular code. It will generally take longer than that and uses up considerable time of a colleague.

That having been said, firms have to do training to increase the skills base and cannot simply expect to poach other people's trained staff. However, since most people have more than one job to do already they will take the 'lazy' way out - I know that we have. Another issue is that the CFD world is still quite small and it may be possible to get a second opinion on the person you are hiring from a third party, which is also helpful.

In terms of ability to use a code, the ability to write a code does not 'a priori' mean that someone can necessarily use the code to produce useful engineering results, which after all is the real point of industrial CFD. Knowledge of the workings of a code is useful but so is a detailed knowledge of how flow devices work and the ability to understand when junk is being generated. The latter is not usually trivial on complex flows. Essentially what we are doing is numerical experimentation with great instrumentation and lousy ,or at best somewhat dubious, physics.

I would suggest that if you are looking for a CFD job in industry that you major on how you have used CFD to get results that were useful to a problem solution. The other difficulty you will find is that you hit HR departments that will bounce your resume if it doesn't have the 'correct' key words (perhaps another reason why codes are asked for is it does make screening easier). Try to apply to people within the organisation, never easy to do but undoubtedly more effective.

Finally good luck, as with most avenues of life breaking in is unfortunately harder than moving around having broken in.

It is very, very sad to hear your message about the employer's job requirement on commercial codes experience. It is also very, very sad to tell you that I don't have a temporary or a permanent job right now, even though I have a Phd degree and 25 years of CFD experience in aerospace industries. I wrote my 3-D Navier-Stokes code 25 years ago on IBM 360 when the name of CFD first appeared and researchers were solving 2-D problems. Over the next 25 years, I have been busy promoting CFD in industries almost every minuite. I have worked at Air Force Arnold Center, Aerospace Corporation, Science Applications International, North American Aircrafts/Rockwell, Air Force Astronautics Lab, Industrial Technology Research Institute in Taiwan, AlliedSignal Engines over the last 25 years in CFD field. I have used numerious commerical CFD codes, someone's CFD codes and have written numerious CFD codes of my own. I also have started developing Integrated CFD codes on PC since 1989 with geometry, mesh generation, graphics, solver,and animation all written in one compact and very fast code. " What I am trying to tell you is that there is nothing wrong with CFD ", I was able to stay in this field by changing companies over the last 25 years. Because I think CFD and the structure FEA will be the two most critical technology items in developing competitive products when coupled with 3-D CAD and CAM. And the only way to take advantage of this competitive edge is to hire engineers or researchers with CFD training and experience. Unfortunately, in the real world, no one seems to know how to implement it. " Codes like wind tunnels or test facilities simply can not create useful results on their own". Same is true that "Internet can not create Response by itself". It is you and me who are exchanging the experience and knowledge. Maybe the next exciting course for college professors to teach is " How to improve company earning by proper management of CFD talent". By the way, real CFD technology is extremely critical item, you are not suppose to give it away , even though you maybe still looking for a job. The defense budget cut and merging of aerospace companies are two major factors in affecting the proper utilization of CFD talent. I maybe able to tell you a lot more, but, still I need to get a job first. on the father's day, an un-employed CFD researcher: " a happy father's day to every CFD family from Westminster, California !!!".

Dear Sir;

I don't know how to say after I read your words. I think you are one of the million victims of the gap between university and company. Unfortunately, I'm one of them.

We are educated by our professors that we are the only people who can develop 'a perfect' CFD code, who can make breakthrough on the numerical simulation of NS equations. We are stuffied with a huge amount of knowledges on solving Partial differential euqations, we are indulged in the competition of program optimization. Almost all of the professors around the world are doing the same thing, why? Because they are professors, they need to be promoted, they need funds. If they just ask their PhD's to use commercial software, where the papers come from?

I think this is not a matter for professor, but the matter of whole post-graduate educations of the world. It is the time for university presidents to think about how to educate the graduates correctly. They should know that it is only small fraction of graduates who will be profs later.

BTW, Sir, don't worry about being un-employed. In Sunshine-always CA, it should be difficult to get the computer related job.

God bless you, bless us, poor CFD 'developers'.

Fayi
PhD student
Minnesota Supercomputer Institute
and University of Alberta

Steve,

It is too bad that employers look for previous commercial code experience. There are many highly qualified CFD engineers coming from college that could be very productive if given the proper training. But, let me also give you a perspective from the employers side.

First, most all companies will be turning to used commercial software in the future (this is true for the auto industry which I am mainly involved). It is very costly to keep developers on staff when a package can already be bought. It's not just code development, but also user support, documentation and all the other hastles that come with keeping a code up to date. That being said, if you want a job doing code development, you should try and focus on a particular subject. Development of upwind differencing schemes is pretty complete and people that can do it are a dime-a-dozen. But those who can do spray, combustion, and other special topics are in higher demand.

Second, while you believe that your code development experience would make it easy for you to learn a commercial package, this is not the case. It is not really important that you know how an upwind differencing scheme is written, but more how it should be applied and what difficulties it can create. Most importantly, how it will affect your results. Frankly speaking, nobody cares how the code works, just as long as it can provide dependable, accurate results. That is the main job of the CFD engineer. That being said, you should assume that it will take you just as long to learn how to use a CFD package as it would to learn to do CAD using a commercial package. Code development is not a prerequisite for learning which buttons to click and what the shortcuts are. This only comes with experience.

Third, if you really want a job doing CFD applications, you should focus your resume on what problems you have solved using CFD. Obviously, when you have completed you code development, you did so with a purpose (a problem to solve). Tell potential employers about it. If you can convince them that you can solve problems using CFD, they will hire you. Trust me, this is not a light decision. When an employer hires someone they will have to train, it is a big expense on their behalf. They know that not only will they have to pay someone to train you (whether it be someone internally that will lose productive work time or an external source which is also expensive), but it will probably be 4-6 months before you are really productive and capable to do a project on your own.

You should also remember that not all people with your experience become good analysts. Sometimes people get bogged down with things like differencing schemes, artificial dissipation, and other details that take away from the big picture. When an employer gets this type of person, he loses alot of time and money on that person and then needs to repeat the process all over again. From this standpoint, if they can find someone with experience, it cuts the risk by at least a factor of 5.

Now, I'll finish my short narration with a positive note. This is a great place to be right now. CFD technology is really starting to be accepted and companies are really starting use it. There are very few engineers available that have experience with commericial codes (at least in the auto sector). Companies will have to start training people more in order to keep their labor costs down. I believe that your future will be bright if you just stay with it. Don't let requests for experience stop you from applying for a job, because I guarantee that not all openings will be filled with experienced cantidates. There aren't enough out there.

Good Luck.

Tim

Thank you very much, Jim. You taught us, the CFD developer a good lesson on how to adapte to the job market of CFD. I think it is not a good idea to encourage people to develop their own codes. Instead, we should encourage people how to use the commercial code smartly, how to use code to solve your own problem. Leave the coding and numerical scheme verification to a limitted groups.

Thanks again,

I'm not sure I agree with this. You have to remember that CFD is still a field of engineering where you can't trust the results your code produces, even if it is the latest and most modern commercial code. Knowing and understanding the weaknesses of your code is therefore essential. I am not saying that you have to be a top numerical analyst to be a good CFD engineer. But having worked with numerics and modeling is definitely a valuable experience for every CFD engineer. In the long term, I would say, a more valuable experience than having used the latest pre-processor interface or so - there is plenty of time to do that once you get out to industry. Of course you need to know the application, you need to know how to sort out what is important and what is relevant, and you need to be a good engineer in general. You also need to be a stubborn and smart job-seeker to land that first job ;-) - talk about the engineering problems you have solved and your analytical skills etc. to the "boss" and talk about more details like codes, numerics, turbulence modeling to the "engineers".

However, my experience from Sweden is that most companies do value your scientific background even if you haven't used commercial codes. I'm just about to finish my PhD and I were looking for a job for about half a year, before I signed with my future employer. In my research I have mainly used my own codes. Once or twice did I encounter this attide that the first question they ask is if you have run Fluent, and when you haven't they loose interest. This was only with consultant companies though, and only when talking to people who did not know me and who were not working as engineers. I think this attitude is much more common with consultant companies. The reason is probably that they have a customer who demands to get a guy with for example two years experince from using Fluent. Your experience is therefore important for them in a different way. All bigger companies that I have talked to have not been so centered around commercial codes, although they are using them more and more. If a company is very focused on me having experience from a certain code then I see that as a sign that they don't need someone with my scientific background. At first it made me irritated but after having thought about it it just made me look for another employer.

I would also like to say that I don't think that it is bad that Unveristies are using their own codes, although I realise that it is becoming more an more difficult to remain in the "front line" using only in-house codes.

Should also say that my experince is from Sweden. I think these things vary a lot depending on the current job-cituation in your region. Here in Sweden there is a big need for CFD engineers right now.

(1).If you ask a group of 20 students each to solve a problem using a commerical code, you are likely to get 20 different answers. (2).If you ask one student to solve a problem using 20 different commerical codes, you are likely to get 20 different answers also. (3). If you ask one student to solve a problem 20 times at different locations and time, you are likely to get 20 different answers. ( unless he copied the input file 20 times.) (4). If you educate students using black box, then the question is who will be responsible for the crash of an airplane, the failure of an engine, when he starts his design career. (5). The truth is , so far we are not sure whether we know how to find a solution by using any CFD code. (6). If different turbulence models produce different answers, and if different mesh arrangements produce different flow fields, then you must be very curious about " What's really going on there ?" (7). It's up to you to find the answer. (Hint: You may want to read the recent issues of AIAA Journal on CFD validation. Then, you are likely to reach the conclusion:" CFD is not for everybody ".

Not everyone has to drive from A to B in a Rolls-Royce Silver Shadow or a TVR Chimera. Often a Ford Escort or Vauxhall Astra will do. They're cheaper and often more reliable than their more expensive counterparts.

If you after results < 0.5% then start saving for a Roller. If your BCs are only known to within 10% then you migth as well go and drive an Astra.

I agree with you only if the driver knows the exact location of point A and point B. He also needs to know the performance of the car. And a map is also essential to get him from A to B. The road condition will be an important factor also. After all, once the driver knows how to get from A to B, he can give it to someone else. The problem is considered solved. If you don't have a solution yet, CFD approach will always be attractive to many people. Because most people don't know how to get from point A to point B. Maybe they can find some clues here.

[lakeat]

Originally Posted by John C. Chien
;118

I agree with you only if the driver knows the exact location of point A and point B. He also needs to know the performance of the car. And a map is also essential to get him from A to B. The road condition will be an important factor also. After all, once the driver knows how to get from A to B, he can give it to someone else. The problem is considered solved. If you don't have a solution yet, CFD approach will always be attractive to many people. Because most people don't know how to get from point A to point B. Maybe they can find some clues here.

Well said!

[jchawner]

Employers should hire for intelligence, not skill set. Having a good understanding of fluids represents intelligence. Knowing how to run a particular CFD code is a skill. An intelligent person can learn a new skill.

But, there are always exceptions to this rule. Suppose I'm a small CFD consultant and my biggest client requires that I run their jobs using Fluent. In that case, I'd make experience with Fluent a job requirement (probably because the training expense would eat into profits). At the same time, I'd still try to hire the smartest person I could find.

[seyedashraf]

yes its weird.
I think employers have to do that, cause they don't have to extent a scientific method to write an article, etc... they just want to have the numerical results ASAP
and I think much more man-hour has been spent on coding Fluent, Flow3D and similar commercial softwares than your own code, plus they have many solvers, boundary condition support, etc.. subsequently these softwares can be used in almost every definable geometries.

Between I myself am writing my own 2D-3D SWE solver in FEM as a M.Sc. student and I'm sure it's much more difficult than using a Commercial CFD software. but it's much sweet to see your own code running and giving the right results rather than any other commercial software.

[Martin Hegedus]

LOL, and OMG, in my personal opinion, using CFD is not about running or developing codes. It is about fluid dynamics. And I'm not laughing at anyone on this thread, I'm laughing at the ENTIRE CFD community! Given all the advances in computer speed and user interfaces, CFD seems to be even more witchcraft than science when applied to fluid dynamics. And if the employers don't realize this, so be it. They will be shooting themselves in the foot and spending goobs of money and time on poor results.

And here is an inside "joke" which shouldn't be as inside as it seems to be. On one side there are the experts who can really truly leverage CFD but they do a poor job of passing this experience (in the form of best practices and reference docs) to those who are new and inexperienced. Yes, they write plenty of papers and if one knows what to look for and how to read in between the lines you will get the information you need. But to an inexperienced user, it is not transparent. And frankly, in general, they are not interested in making it transparent. Yes, there are those who do try to do it (and a HUGE thank you to them from me), but it is not on a big scale. If ones brings up the subject, in general, one gets agreement that it should be done and the "no time" excuse.

On the other hand there are those who do not know how to use CFD, or to make fluid dynamics comparisions to experiment, but who are doing it anyway. And, there are more of them than experts. This is wasting lots of time and muddying up the waters. Of course, in the end, this comes back to bite the experts.

Case in point. There is this blog posted by John Chawner on Pointwise's site (http://blog.pointwise.com/) "According to this study of wind loads on rooftop solar arrays “CFD is not currently in a state where it can be used reliably to estimate peak pressures on low-rise building roofs, let alone solar arrays on such buildings, and should not be used as a basis for design at this point in time.”" This paper was also referenced at the symscape blog. My frustration is directed at the authors. The authors took the complex field of CFD and dumbed it down to fit into a side note. They didn't even TRY to expand on it. Nor did they even provide any references for this statement. OK, I'll grant them that a RANS code will need to be used with caution and will have difficulty. But how about a LES code? OK, a LES code, in general, will take significantly more CPU time. But how much more? LOL, maybe the authors are just trying to sell wind tunnel time. After all, the subtitle to the paper is "A Primer on Using Wind Tunnel Testing as a Basis for Code Compliant Design per ASCE 7" Though I would love for them to tell me how they match Reynolds number and compressibility for 100+ mph winds on a big building in a small WT. LOL, it is also interesting to read all the challenges in regards to modeling buildings in a WT and compare that to the CFD blurb. What a mess of a CFD statement the authors made. But this is what employers read. And, even more importantly, this is what the building inspectors and designers read.

What's my point? CFD is more politics and fluid dynamics magic than the science or art of creating a code or an input deck. What is most important is understanding the application art and politics.

[lakeat]

Okay, ironically, and to tell you the truth, this man is one of the "big figure" in wind engineering. I am from wind engineering too. And I am also from civil engineering.

And I apologize for my Canadian coworkers who has "muddied up the waters", sometimes this is unintentional, and I also must agree with you that sometimes this is political.

I am not willing to talk about politics here, that is complicated...

But I am going to defense for the "unintentional" cases. This is not about politics, this is about education system,

1) Everyone, EVERY one, his time is limited, in many cases, person A is better in CFD than person B is simply because he has spent more time on it. Listen, this doesn't mean the wind tunnel engineering students are lazy than the CFD students, we worked day and night sleepless to analyze many complex fluid issues, like flutter, buffeting and other vibration problems, and we worked very hard on them. If given more time on CFD, these students would be no worse than those pure CFD guys.

2) We try our best to use good time to educate ourselves CFD (Of cause, fluid mechanics first), but we also need those good CFD guys come to educate the civil engineers in our circle all around the world! I just came back from Netherlands attending a wind engineering conference, where many CFD application are being presented. You might be going to laugh at these simulations and conclusions they got, but what should we do? So many undeveloped countries they are looking forward to more productive..... And the good CFD guys usually do not come often into our field (where are they? most in AME, or in auto industry, or in military? I don't know.)

So, my point is, this is a learning process, pls don't blame us too much if it is unintentional, we hope to make the (CFD) world better just as you do. I ask for your sympathy, please.


But anyway, Martin, thanks for your input, it is helpful.

[lakeat]

And I been working on LES, wall resolved LES for a long time, you know it's very tedious process. It's valuable and very reliable in giving the peak pressure, but people don't buy it in this circle, employers don't buy it in this circle, not many people are interested in it, in a simulation takes more than a month. Even in some big projects, the employers are not sure that they are able to afford it, like the design of Dubai tower. You can imagine the Re number of it, and the mesh it is going to use. So wind tunnel, still has its value.

[Martin Hegedus]

Sorry, I guess I wasn't clear. However you drove the point home very well, though you may not realize it. I don't expect the civil engineers, race car individuals, solar cell people, etc. to be experts.

However, the professors, whether that is at Stanford, MIT, U.C. Davis, Georgia Tech., etc. and various government labs and agencies are the experts. And I gather, in general, they don't seem motivated to come out and interact with the other groups that are not experts.

So here is the chain: 1) core CFD experts don't say much, 2) well known authors in other fields get burned by CFD and write bad things, 3) decision makers think poorly of CFD, 4) the world in which the CFD experts live in does not grow, i.e. funding stagnates.

OK, take civil engineering. I'm an aerodynamicist, so I apologize if I get this wrong. However, I gather their are various books and such that I can go to that spell out a "code". Some of books are also reference material. For example, like Roark's Formulas for Stress & Strain. OK, maybe that is not civil engineering, it just came to mind. It is probably more relevant than Abbott and von Doenhoff's Theory of Wing Sections.

Does something like this exist for CFD? I'm not aware of it. And I'm not a novice at this. Yes, there are some papers on best practices out there. But usually they are top level stuff that shows what works. I don't know of anything that actually nails down the things that do not work. Personally, I would rather see papers on "worst practices" than "best practices"

Or another example. A few months ago there was a CFD conference in Hawaii. Yet for all those people that went to that, I hardly see anyone from that group on this forum. Why? I don't know. I have no idea.

Maybe John can correct me and tell me I'm completely off base. Frankly, I would not mind.

Am I, as someone with a bit of experience in CFD, frustated with this? Yes I am.

[Martin Hegedus]

Originally Posted by lakeat

And I been working on LES, wall resolved LES for a long time, you know it's very tedious process. It's valuable and very reliable in giving the peak pressure, but people don't buy it in this circle, employers don't buy it in this circle, not many people are interested in it, in a simulation takes more than a month. Even in some big projects, the employers are not sure that they are able to afford it, like the design of Dubai tower. You can imagine the Re number of it, and the mesh it is going to use. So wind tunnel, still has its value.

I totally agree with your WT statement. In regards to LES, I personally think it would be wonderful if governments (U.S., European, etc.) would dedicate supercomputer time to developing reference material for the application of LES to problems such as this or even just supplement reference material with RANS/LES studies. And I would assume that they do, but I don't know for sure, since it is not the circle I travel in.

Unfortunately, even in my field, the core field, aerodynamics of flight vehicles, there seems to be more uncertainty than there needs to be. If this is true for flight vehicles, it has got to worse for other fields.

So for those entering the world of CFD, and this is my opinion, there is more involved than just math and coding.

[lakeat]

Good words, Martin!

1. Fluid is everywhere, so you cannot stop people using CFD, or trying CFD. {{{What I am going to do for the next several years now, perhaps is to develop some platforms that will "muddy up the waters" even more . Our lab, we are aiming to (1) Let more engineers get their hands dirty, (2) Alert them to be very careful to draw their conclusions, (3) Build database in our circle to accumulate both the worst practices and best practices.}}}
2. Interdisciplinarity! Interdisciplinarity!
3. Are we really ready to develop a good "worst practice" right now?
4. Fluid structure interaction is so sensitive to the outer shape, eg. flutter and buffeting, note we don't have streamlined shape often. But most these problems challenge both CFD as well as Wind Tunnel as well.

Believe me, for many civil project (super long bridges or super tall rise building or super span roofs, or even surges along the coastlines, those houses that suffer storms and tornadoes a lot) we met in recent years, here is the situation.
If I am the team leader and engineer in a design company, given me say 100 days, I would devote most of my time to wind tunnel testing, and a few time to CFD. And PIV? "It's good, very useful, but go to hell." This is engineer problem, we need make best use of our money and time. For most of the nonlinear problems we met, CFD is still not ready to solve them. That's one of the reason why most "decision makers" think of CFD in a insufficient way.

We really hope to improve the quality of CFD in our civil engineering circle.

Sometimes it is related to funding, Americans and Japanese have the big funding possibilities, but the civil construction is not growing anymore, but auto and aircraft are keeping growing. In China, we have a vast and crazy growing trend of civil engineering, but our scientific research for some historical reasons are weak.

And also, sometimes people are talking as if they are in different worlds. I think even Nature journal doesn't have many CFD articles published, except Moin's one. Does it mean that CFD expert are not better than them? I don't think so. And are civil engineers not better than the CFD experts? I don't think so. We all have our own job and priority to do. In a generation like this, we all need each other.