How often do you use of CFD (Computational Fluid Dynamics) ?

Maybe I would share another brief statement from my personal experience.
I am one of this PhD specialist that has been exposed to everything from towing tanks, wind tunnels and CFD for americas cup, formula 1 etc etc.
I earn my money doing this etc. I would like to say a few things that might help some misconceptions.

CFD is expensive
Not true. CFD can range from practically free to expensive, but even still is way cheaper than some alternatives. What is expensive is paying the guy who knows what he is doing. He will be expensive, I can promise you that. If he is not then maybe he is not the right guy.
This guy should not be a pure CFD specialist. Otherwise you get into the **** described in the other posts.
He should an engineer/fluid dynamisier/ naval architects/ aerospace.
He needs to understand the real life scenarios and the limitations of his tools. If he doesn’t he is a waste of time and money.

i will also say the following.
I have worked an multimillion €€€ projects just for CFD, doing everything.
When I come to design something for myself (sailing boat for example) I rarely do CFD. Not because of the expense. Actually that is very low. But because to really answer the real life question you need to take into account so many conditions (turbulence, waves, operator input, geometric uncertainty, structural deflections etc) that it ends up being a multimillion multi year task.
Instead I use the lessons learned from all those other projects and I design from my PERSONAL experience what I think will fit all the requirements.
Occasionally I might run 1 or 2 to check some trade offs, but usually I trust my experience and stick to it.

This is what you pay for. The experience of the people.
My experience would not be what it is without CFD. I would not have the detailed understanding of endless projects I have now.
But my experience includes wind and towtank tests, real life tests etc, just as much as CFD

in short. Know you tool and use it correctly.

edit. I describe myself as an engineer and not a CFD specialist because I consider so much more. A combination of experience and CFD is the right mix. If you take the dumb approach of just looking at the numbers then as discussed you need a lot more simulations

 

Very well put, fastwave, thank you! I myself use CFD on a regular basis during r &d work on turbomachinery (including marine propulsion), their applications as well as for break-down and trouble-shooting analysis. But NEVER as a standalone tool, always together with other tools, with great scepsis and as far as possible using real world data for validation. Also, I never trust/use the CFD tool for strict quantitative analysis, but use it for qualitative analysis and comparison. In some cases, however, customers explicitly demand a CFD analysis in spite of available, well proven and validated results from one- or twodimensional methods, only because of the pedagogic qualities of the colored images.....(that's my lucky day then, I know I can charge without blushing...-.)

It must be stressed over and over again, that CFD is just one tool of several, and just as useless as all the others if you do not understand or master the fluid mechanics as such. Learning and thinking are laborious and time consuming activities. As Ad Hoc said, gaining a ten year experience will take aaah, humm, about ten years and as we are notoriously lazy and greedy, it is tempting to "buy time" and get the whole shebang with a press on the button.

This is where the educated rookie with fresh computer skills comes onboard with just basic academical understanding of the physical world of fluids; the company can say that hey, we use the latest of the late, put your money in our pockets! I have seen this scenery repeatedly over the last twenty years; CFD beeing applied without background knowledge and resulting in seriously wrong decisionmaking. The Biiig tech companies are no exception at all, rather the contrary!!!! I have seen and corrected a number of these entrepreneur mistakes during the recent upgrading of power stations, f.i. where cavitating condensate pumps are limiting the turbine output. The Biiig company CDF-experts insist on their conclusions, but the real world says it will break...oh sh**... it just broke!

The most frequent mistakes I see refer to faulty definition of boundary conditions, or lack of understanding of the physical consequenses to the external system connected to the boundaries. To avoid these mistakes, a genuine engineering understanding of the system behaviour is necessary; to gain this understanding it takes learning, thinking, working; the essence of experience. So, Alan C.: the selection of tools is not purely a question about technical "fit-for-the-purpose" in all quarters, the bragging factor often has to be brought into the picture.

A memorandum from an old peer of mine on how to distinguish between characters:
"The rookie will brag about his successful design, that surpassed expectations and the SOR with great margin, not understanding that this is a failure per se, but the experienced guy will tell you a story about what was learned from his project that went wrong."

 

Wise comments from an experience guy.
It make me laugh that all the experienced guys come to the same conclusions.
Managers making what they thing are smart decisions to save many ending with costing the company millions and years to correct.
Most of the time these idiots have moved to the next company bragging about their achievements before the **** hits the fan at the previous company.
Sad but that is the reality of the world.
Fully agree with all your comments

 

Since you quoted my "meaningless for analysis" statement about your table, I will respond by saying your leading question..."Here are the tank test results. How should the shape be modified in order to achieve targets ?" ...renders any ability to do analysis moot because there is no shape data to modify. The table you put up is spurious relative to the usefulness of CFD for adjusting hull forms; i.e. the CFD program did not change the hull form, someone looking at all the data did. As to the second portion of the query above; no, CFD visualizations of the hull-air-water interface from the stem through the forebody still have significant computational issues (even though today methods are much better than the issues that lead to the coke-bottle SWATHS) whereas an actual model test, though with their own issues, provides significant insight as to where to focus improvements.

Years and years of RDT&E for the US Navy, from things the size of my hand to full size ships. From preliminary design to contract spec to lab testing to fully instrumented sea trials the best I ever saw as ~2% predicted/measured in the lab and ~4% at sea....actual real seawater is totally different than CFD water (and yes, I typically stuck 5% in my hip pocket for "sea service"). And then there is the actual items...scatter among "identically machined" propellers powered by "identical" motors is a significant. Hell, have you actually seen the data that comes out of MIL-SPEC bolt lot testing?

I bet if you let me instrument her I think I could prove what I would bet the problem is. Often people pushing on the envelope don't look behind them; the devil in in the details.

Unfortunately too many just pull the shrink-wrap off the code, toss the manuals, and dig in. One of the things we always did before starting a big project was to do a variable sensitivity analysis on the code we were going to use for the regime we were interested in. This can help prevent the "self-licking ice cream cone" as one contractor once aptly described some problematic code.

Remember, those "companies" are those ex-researchers; and as was implied earlier, it is the accountants and managers that are making the buy decisions. These buy decision are based on the the same factors that got "everyone" on a computer...reduced floorspace and salary costs by getting rid of drafting tables and drafters. Read Admiral Rickover's "Doing a Job" speech for some insight into industrial shipyard decision making. Doing a Job - The Management Philosophy of Adm. Hyman G. Rickover https://govleaders.org/rickover.htm

No, it just a tool with its own quirks; it is not a panacea for all things Naval Architecture.

Well this has taken a while, so now I'll read what got x-posted since I started.

EDIT AFTER READING X-POSTS: I agree with most of what fastwave and baeckmo said: CFD is just one tool and is a force multiplier in the hands of experience, but not a substitute for experience.

 

Of course not. What I think is that what some CFD artists sell, applied to "normal" boats, not competition ones, in many cases YES is a farce.
Nor should we forget that the improvements that can be obtained with CFD only represent a minimal part of the problems that the designer must solve to achieve the project that best adopts the SOR. So saying that, for example, the hull has been checked through expensive CFD studies, may say nothing, or very little, about the quality of the design as a whole.
Having said all this, I would like to say that, in my opinion, until all marine designers learn hydrodynamics seriously and gain practical experience, this wonderful tool will be of almost no use (if we are talking about technique and not marketing).

 

Fully concur with Fastwave and Baeckmo and JEH.

The same as noted above can be attributed to FEA.
I have been using FEA for 30 years, and have witnessed time and time again the same reactions of those speaking in hushed voices and prostrating at the glossy colour image as if staring into the face of God, and never questioning the image, it is the truth, as all those that follow the religion require such blind faith of what is presented.

It is a tool, just a tool, and one must learn to master it, otherwise you will be the slave.
And when in that position, there is no hope…

 

But what about the use of CFD in aerodynamics, not just looking at drag, but calculating the size of the actual lift and drag in Newtons from a wing or the size and direction of the aerodynamic force in Newtons and degrees from a sail? Is CFD able to do that?

 

Sailor Al; what makes you believe that simulations in "non-aerodynamic" or marine disciplines are restricted to "just looking at drag"?

 

Ok. Maybe I’m wrong about hydrodynamics, but my question about aerodynamics stands.

 

The "F" in CFD stands for FLUID. Gases are fluids as are liquids.

 

Yes, I know, but you haven’t addressed my question: But what about the use of CFD in aerodynamics for calculating the size of the actual lift and drag in Newtons from a wing or the size and direction of the aerodynamic force in Newtons anddegrees from a sail? Is CFD able to do that?

 

Thank you very much, everyone. A sketch of answer can be drawn. From what I understand :

"CFD designates numerical tools, directly based on approximations of physical principles, or, indirectly; based on statistical representations. These tools should be used by qualified operators with significant experience in the application for which CFD is intended, or, at least, in a similar field. CFD should be used inside a procedure, that defines the appropriate formulation, in accordance with the application, and, takes into account the intrinsic limitations of the choosen formulation. CFD should be used in combination with other methods, for the results to be compared and critized."​

What I don't really get is why CFD integrates in the design loop of Naval Architecture, in the first place.

- If the choosen CFD tool is based on statistical representations already well known in Naval Architecture : Its utility could be questionned, because, in a sense, the result of the CFD study is already known. The validation of the CFD results can be done according to the experience of naval architects themselves, having the experience of these statistical representations. Here, CFD brings nothing.

- If the choosen CFD tool is based on physical principles : Two cases may be distinguised.

- The object of the study is inside the bounds of already well known statistical representation : The validation of the CFD results can be done using the statistical representation, or, according to the experience of naval architects. Here again, CFD brings a double check.

- The object of the study is outside the bounds of already well known statistical representation : The validation of the CFD results (should be/should already have been) done with another tool with higher precision, in the domain of application.

- CFD results have been already validated in the domain of application : The use of CFD is exploratory and aims at finding local maxima or minima that could have been missed in previous studies.

- CFD results have not been already validated in the domain of application : The CFD itself is being validated, to be used in further design explorations.​

I'm very curious to know more about the strategies that were developped around the use of CFD in your design loops, for your respective projects. Because it seems to me that :

The following proposition should be wrong.

CFD is expensive, because either it brings nothing and paying for nothing is expensive. Or you are using it as a double check that can be easily realized by other methods, by hand. Or you have already paid a considerable amount in validation of your CFD in a certain domain, that you dig desesperatly again with the same tool with a systematic approach this time, with the faint hope that it will give you a solution of more than 5% better than the actual, when making your boss believe that you work hard. Or you wish to validate your CFD procedure, and in that case, you will be the guy crying on the boss desk to spend the rest of the budget in more costfull validation procedure. Not to mention the high price your boss have to take out of his pocket, just to hear you cry, since you're such a hot shot.

No wonder why so much people have bad experience with CFD...

 

It is not so much the "experience" per se, but those with less knowledge on the subject and theory behind it, with their expectation of what it delivers.

 

How do opinions about model testing, both tow tank and self powered models, align with or differ from opinions about CFD?

 

A physical model provides a very good representation of the proposed design.
The science behind the testing is quantitatively documented and strict procedures given to test tanks for such consistency.
But these test houses are also venerable to errors and poor procedures if not checked ...one must select a tank that is suitable for the proposed design...

The results are like anything else, a 'tool' and a data point for the entire design. And using the data from the test tank gives a high degree of confidence of what could happen, if the testing programme is done correctly.

Since test houses only do what you ask them... so if you don't ask for an LCG chase, for example, you may not know that the design's LCG is critical and a minor movement fwd or aft can cause serious issues.

So, the final results of a tank test is only as good as testing standards and the procedure under which it was done.
But, most of the systematic error(s) is generally known and mitigated by the better test houses to give results with a high degree of confidence.