Hi there CFD gurus.

I am CFD Dummy and I have an aircraft aerodynamic problem. (It's really more just a question than a problem.) I know I can get it solved in a wind tunnel for as low as $300 an hour (plus building the model for some $20-30,000) and I'll have a reliable answer. Now, what can the CFD do for me other than confuse me and why should I look into it?

CFD Dummy

(1). That's a good question. And I don't think that one has to use cfd , if he can obtain reliable results from a model testing in a wind tunnel. (2). In old days, thousands of wind tunnel testing hours were required for aircraft development project. (3). The need to use cfd comes largely from the testing side, because you can't get reliable results form 1 ft x 1 ft wind tunnel. It is quite expensive to use 16x16 wind tunnel. (4). Well, everything is relative. If you are happy with wind tunnel testing results, it is perfectly all right to do so. (5). At one time, Boeing aircraft company nearly abandoned the wind tunnel testing approach, instead they decided to use cfd approach, as reported by news media. Even though it is not quite 100% cfd, using various cfd codes at different level of complexity in the aircraft development has been a long tradition in the aerospace industries.(NASA, Air Force and the government labs as well) (6). The same is true for the engine (jet engine) companies. (7). So, if it is easier, cheaper and more reliable to use testing approach, then I think, you really don't have to use cfd at all. The choice is yours.

If the design is fixed, and you are absolutely sure that the wind tunnel results will scale well to the full sized aircraft, then (I'm quite sorry to say) there isn't much reason to do CFD (except that it's likely to be a bit less expensive).

CFD really provides a benefit earlier in the design process when the design is changing. Modern CFD tools provide good connectivity to CAD packages. This allows design studies to be turned around quickly. It allows designers to assess the effects of design changes, enabling them to improve the design. Then at the end of the design process, a test will often be run to CONFIRM the CFD results.

The use of CFD in the design process results in a better design. If you were to try to use testing throughout the process, it would take longer to finalize the design and cost a lot more.

In my experience CFD is an excellent tool for saving time and money as part of the design process. It may someday completely replace testing, but it hasn't yet.

Alton

Somewhere I've read that CFD really stands for Colorful Fluid Dynamics. The results are unreliable and inaccurate and it will take another 30 years and several computer generations before CFD will become really useful. When it comes to drag prediction in aircraft design, CFD is notoriously unreliable (or at least used to be 7 years ago. Has that changed?) In one of your previous answers on CFD accuracy I've read: "Different codes will give different answers, [and] some code also will give different answers by different users at different time." So what case could Boeing possibly have for nearly abandoning wind tunnel testing and going CFD?

How is CFD less expensive? Doesn't it cost at least $25,000 for a decent code for one year license, and then I have to do the wind tunnel testing anyway?

Don't get me wrong. I love new tech toys and if they save money I am all for it. I'd love to run my CAD design on my computer through a virtual wind tunnel, but my understanding of the situation is that for a large investment in time into running a CFD code I can get only qualitative predictions that can be made most of the time by good engineering judgment and to get the hard numbers I must to go to the wind tunnel anyway.

(1). Well, the Wright brothers of a bicycle shop invented the working airplane a hundred years ago. (2). They used a 1x1 wood wind tunnel. There was no computer or cfd. (3). And I don't think you can measure the aircraft's drag accurately in a wind tunnel. There is this model scale effect, Reynolds number effect, the wind tunnel wall effect, etc, not to mention the transonic shock interference effect. (4). cfd has always been the research field at and beyond the PhD level. So, with the right kind of people and tools, I think cfd has been shown to be a viable approach to the fluid dynamic design and analysis. (5). Do you think that the reason why Wright brothers were able to invent the airplane, was because they were able to measure the wing drag accurately in their 1x1 ft wind tunnel? Do you think that people were able to fly their model airplanes, because they were able to measure the airplane drag accurately in their garage or basement ? (6). The purpose of using cfd is just another way to provide information for product design. When you fly by airplane, do you have any idea about the total airplane drag? I think, with proper validation, it is possible to predict the drag accurately. But who is interested in the accurate drag value? When you are flying by airplane, you don't even know the so-called free stream conditions. Do you think that the real free stream is always nice and uniform? (7). I think, if cfd can help in some way to design a product, someone out there will definitely give it a try.

If you are going to test only your final design, and you currently have your final design CFD will not save you time or money.

If you have in front of you a blank sheet of paper and want to optimize flow characteristics of the design, CFD will save you time and money compared to testing. Multiple CFD design iterations combined with 1 test at the end is much faster and less expensive than multiple test cycles.

Alton

Hmm,

At the $300/hr for a wind tunnel that you quoted at the start of this thread, the $25,000 code cost will buy you about two weeks (~ 83 hours) of test time. I seem to recall a fellow engineer spending 6 weeks (!) at the Air Force test facilities in Tullahoma, TN, then returning for a second visit. When he was done, he had the test data he had. And a bill for two round trip flights from California and 12 weeks of rental car, per diem, and motel costs.

Of course, every time you make a design change and go back to the wind tunnel, you make a new model. Am I wrong in thinking that changing a virtual model is cheaper and quicker than changing a piece of hardware?

Using CFD, in addition to the code, you need a computer ($25K for a workstation [?] or $5K for a loaded PC), printers and supporting software ($10K). But when you've successfully completed your project, you still have the code, the computer, and the infrastructure. To be used again. It's really hard to say how long a CFD simulation series might take. Anywhere from a few hours to well over a year ...

Whether testing or calculating, the designer/analyst is the big cost, (I guess) $150-250K per year including fringes. It's possible that your choice of method will depend on the skills of your staff. As Dr. Chien points out, with the right person doing the work, either approach can lead to very usable hardware. Without the right person, good luck.

Sure it's difficult to do rigorous analysis using CFD, but just think about what you're trying to do: Mathmatically model the flow of fluids, which are often turbulent and of different phases. Then approximate solutions to those models using numerical methods. It's a tough proposition. Often, however, the amount of rigor involved depends on the objectives of the study. It is in this respect, that CFD can be useful. It's not meant to be a substitute for experimental investigation, but rather a supplement. Most importantly, as John pointed out, the skills and modeling "know-how" of the CFD analyst often determines whether the CFD study is a success or not.

(1). You can always get the cfd code for "free" , that is zero cost. (2). You can always get the vendor interested in your problem and have them solve your problem for free, if you have the connection. (3). Can you get your wind tunnel model made for free? Can you run the wind tunnel for free ? Unlikely. (4). And in most cases, cfd engineers are paid to run the code which is paid for by the company already. It is not only free to the engineer, but also he is getting paid for using the code. (5). So, if you have to run the wind tunnel test, get someone else to pay the bill for you.

(1). This is a real world case related to the design modification of a highly complex inlet duct used in the power generation equipment. (2). The inlet is round, and is connected to long pipe. the exit is a quarter of an annular shape, that is, a quarter of the space between two cylinders. (3). Because of the tight space available, the inlet first make a slight left turn, say, 5 degrees, then followed by a somewhat sharp 40 degree turn looking from the front view. (4). On the final 40 degree plane, the duct then go through a 135 degree U-turn, and is connected to the exit duct. So, it looks like a highly twisted distorted stomach. (5). In the first month, available commercial codes were surveyed in house. The geometry and mesh generation code and the solver code were identified (there are several other packages available as backup). The user's guides and tutorials were then printed page by page from the HELP menu of the code. With this information, a straight duct with round to annular transition were modelled and the complete process validated with result obtained. This is just to validate the process.(6). In the second month, a new transition duct construction method was developed and applied to the existing transition duct design. In less than a month, the geometry is on the computer with structured mesh generated. With some minor modification and fine tuning to the mesh, it is ready for the solver step, which will take typically a couple of days work and CPU time. (7). At this point, the design is out for casting work, and the final piece should be available in six months. That means, currently the hardware is not available. (8). For the testing, there is a question mark as whether such large air supply is available in the company's facility. But the inlet condition could be changed and test can still be conducted. (9). As for the measurement, one could only measure the flow at the exit plane, because it is a very thick casting piece. It is also the final product, and one would not want to drill holes in it. So, there will be no flow field information from the inlet to the exit.(except for the exit plane from outside). (10). The biggest problem is still the uncertainty of the availability of the test facility six months from now. Assuming that the testing can be done and done quickly,it probably will not affect the delivery date (this is the real piece, not a model). (11). For this real world problem, in the third months, the first cfd 3-D calculation with several hundred thousand mesh cells will be available. The new construction method will allow the variation of the new design to be computed, because the geometry and mesh generation have been worked out under this method in the first month. So, the geometry and mesh can be easily changed and produced. (12). By the end of the third month, the 3-D results of the current design and several new designs will be available for review and incorporated in the future products. At that time, the casting is still waiting in the foundary to be casted. (13). So, long before the real product is manufactured and tested, the 3-D cfd analysis of the current design and the future design modification are already done and available. (14). When finally the real piece of the product is tested to provide the data at the exit plane, it can be used to validate the cfd result of the current design. Thus the improvement or characteristics of the new designs can be addressed accurately. (15). In this case, it is impossible to do the design improvement based on the detailed flow field survey, becuase the flow field survey simply can not be done on the real product. This is based on the available tools on the system. On can definitely propose a research project to build an instrumented piece (model) for more complete flow field survey. (16). If there is a need to design this product frequently, a dedicated cfd program can be developed and written to further reduce the total turn-around time in the cfd loop. That all depends on the management of the company, and the vision of the decision maker. The new cfd program sometimes will proceed forward smoothly, and sometimes it will encounter technical problems, just like any other research and development project. But if the company just sit there and do nothing, then I am sure that nothing will be done. (This will be John Chien's law) (17). So, the principles used in cfd analysis are war principles. You can't ask your enemy to stop the attack because you are not convinced that you can win the war. (18). By the way, to survive, one needs to identify the potential enemy first. To win the war, one needs to be well prepared ahead of the time. Who is going to invest in this part? A real good question. The answer is "those who have the strong desire to stay alive will have to invest in it" . How do you do that. It doesn't matter how, just do it. The cost is important only when you are alive and are making money. When you are nearly dead, or not making money, the only way to become alive again is to invest in it first. (19). The other law of John Chien is: It is impossible to change a person once he has grown up. (this is because the grow up process is "irreversible". to change that, you will have to know where in the brain memory to kill what program in it))

Putting the situation in terms of having the final design versus starting with a blank sheet of paper is not really representative to a real life situation. How does a design get final?
Let's just say that I have a design that I hope is good and final but I may be surprised. Now I have to make a decision on whether to spend time and money on doing CFD analysis or just going straight into a wind tunnel. Can you make a case for not going straight into a wind tunnel?

Who is interested in accurate aircraft drag value? Well, I would think that the aircraft designer is, if for no other reason than at least to know how big engine to put on the thing, or what the fuel consumption is going to be. And I do expect to get accurate results from a wind tunnel (for $300/hour you can get more than a garage or basement wind tunnel.) Can I get accurate results for the same or lower cost from the CFD?

I am sure that for academic types CFD has a completely different value than for practicing engineers. My inquiry is about its value for engineers.

(1). It is unlikely that you will get anything from the commercial cfd codes. These are the so-called general cfd codes. You can visit the vendors websites for the samples of various applications, ranging from automobile industries, chemical industries, to electronic cooling industries, and combustions. (2). I think, these are engineering problems. Your wind tunnel can only test aircraft models. So, I think, you have come to the wrong place. (3). The right place would be the airframe companies. There used to be seven of them a decade ago. Now the number has dropped down to two. (Boeing and Lockeed Martin I think) (4). So, in order to get the right information, I would suggest that you contact the airframe companies for further information. (I remember, long time ago, at least 15 years ago, Boeing has a cfd laboratory for aircraft design. (5). If you can pay only 300 dollars, even the wind tunnel will soon disappear for lack of customers, not to mention the money to develop the cfd code for wind tunnel simulations. (6). And I don't think that the size of the jet engine is determined by the aircraft drag. It is similar to a car, the size of the engine in a car is not determined based on the car's aerodynamic drag. So, I can say that you are not an experienced engineer in the aircraft desin. (7). Well the best place to get the right information is the airframe companies. And I am sure that they have something similar to the wind tunnel on the cfd side.

(1). Can we pick a wind tunnel for you? (2). Or do you have to use that particular wind tunnel?

Yes, even when you already have a working product like aero-engine, you still want to use CFD to understand and possibly improve it. I just had opportunity to meet Prof. Moin of Standford university - his team is doing some very exciting CFD work for combustors using LES. In fact the current results for a model can combustor are so exciting for both modellers and experimentlists, Standford is asked to run the real aero-engine combustion chamber (of course using DOE super computers)

John -- I have not seen too many airplane specs showing acceleration from 0 to 60, or car specs showing the maximum or the best cruising speed. (Well, maybe German editions of Porsche and Ferrari are an exception.) Engines in commercial transport airplanes are generally sized for max. efficiency at cruise. Engines in cars are sized for ... well -- to make an impression? Of course, the cruising speed depends not only on the engine size (power) but also, wouldn't you know, the aircraft DRAG. So you see, in an airplane aerodynamic design the drag reduction is pretty much the name of the game and how the CFD can be helpful in that is exactly what I am trying to determine.

I don't quite understand your questions. I can use any wind tunnel listed at http://www.niar.twsu.edu/sata/members/members.htm or http://www.worthey.net/windtunnels/, or perhaps some others. Why?

Quite right. But what in my particular situation -- evaluation of a new aircraft design aerodynamic characteristics -- is going to help me understand it better, faster and cheaper? CFD or a wind tunnel?

The answer is (as with many things in life) that it depends on many factors. There are several advantages to CFD: 1. In general you can get more information about the flow field from a simulation than from a test. In John's example of internal flow using the finished part for the test it is especially true, but it applies to external aerodynamics as well. The CFD solution is (more-or-less) continuous. There are physical limitations to how you can instrument a model, and you only have data where the instruments are located.

2. The CFD analysis will have a shorter turn-around time. In a CAD based design environment you already have the geometry you want to test. The geometry simply needs to be transfered to a CFD code and meshed. To test, you need to develop drawings of the model, and then it needs to be manufactured.

3. The CFD analysis will be less expensive. To do a wind tunnel test, you are not going to purchase a wind tunnel of your own, and learn how to use it. If you expect that the design you have will meet all the criteria you have (thus requiring only 1 test), then why do you treat CFD differently? If you expect to not require a design cycle, then there is no need to invest time and money in purchasing a code and learning how to use it. You can contract out the simulation services to an experienced company and expect to get the results back for less than the cost of a license.

4. It also depends on what you expect to get back. If you expect that the test results will show that your design meets all the criteria that you established, then you can go straight to testing. If the criteria are not met, and you have to re-design, then CFD saves time and money.

Alton