Our company manufactures forward inclined blower wheels and blower housings (scrolls) for use in furnaces and air handling equipment. We also manufacture propellors for A/C condensers. We would like improve the efficiency and reduce the noise of our existing designs, and products under development.

1) Would the use of a commercial CFD code generate useful results for design improvements?

2) What level of expertise (education & experience) is necessary to analyze these types of systems effectively?

3) What resources (training seminars, college programs, good books) are available to persons not knowledgeable in CFD, but wanting to learn more about it?

4) What criteria would be useful in selecting a commercial code for these applications?

I don't expect there's an out-of-the-box solution for us, other than contracting the work out to others. We'd like to learn more about what CFD can and cannot do. Any comments and/or suggestions are welcome.

Thank You.

(1). I like the questions. In early 90's, I also had an opportunity to work in this area, trying to develop programs to improve the efficiency and reduce the noise of the fan and pumps. (2).To answer your question-1, I would say "yes, the use of the commercial cfd codes can improve the design, if the original design is rather poor." For example, if the design is copied from someone's existing design, then it is likely that all of these essential parameters are not copied correctly. Then the use of cfd code can show you the general problem areas, thus offering an opportunity to improve the design. (only if you know which way to improve it) (3). In some cases, especially in the pump area, the design has been improved many times in the past using real world test data, then, it is unlikely that the current general purpose code will offer any design improvement at all. The effect could be the other way around, that is, you are likely to get poor design based on the cfd results. So, in this case, the real improvement will have to come from testing. The reason is the pump and fan are running all the time in the 3-D separated flow regime. (4).To answer your question-2, I must say that you are probably talking about the knowledge about the fan and pump design itself. So, the person must have fluid dynamics training,especially in the fluid machinery area. Unless you are developing the program for pump and fan design using cfd, you don't need to have a PhD in cfd in order to run the code. Although it is not going to hurt to have an cfd expert to work on it, you need at least a few years of hands-on experience in pump and fan design.(5). It is hard to answer your question-3, because there is no limit in learning. If you can identify one commercial cfd code which will work for you, then the vendor should be able to tell you how to get the right kind of training in order to use the code. If you can not identify the code, then learning the cfd is just like learning any subject. What I am saying is, getting an A in cfd course does not guarantee that you will be able to use cfd code to improve the design. (6). The selection of the commercial cfd code in your case is somewhat easier because you already have a product in mind. In addition, you are producing the product and know how to evaluate the performance. This is not the case for the code vendors,because they are busy in creating and selling the codes. You can set your goal and contact the vendors of the cfd code to get their response. (7). To improve the efficiency from 50% to 60%, it may be possible. If the efficiency is already in the 70% or 80% range, it is going to be difficult to improve it using the current cfd codes. But there are specialized codes available , and there is always exception. (8). Even if you can find a very accurate cfd code, the design improvement still must come from the engineer's creative thinking. (you could run a hundred parametric cases to find the better design, but in 3-D it can be very time consuming.

Commercial CFD codes will generated useful results if the CFD model was carefully-setup, however, if the results cannot be field validated, it is useless in design aspect. I would suggest you get some demo licenses and run through the tutorial and maybe setup a model for your own product and compare the results with experimental data (assuming you already have). Most CFD vendors give out demo copies these days so you can try it all. And you will also get a flavor of what kind of code you want to use.

The routine applications won't require much knowledge in CFD as most codes are intended as blackbox to their users. But some creative modeling will require both hands on experience in prodcut design and the CFD modeling background. At the least, the former you must already have, and the latter you can always get from the vendor.

(1). To do or not to do, that is the question. (2). When one decided to do, the door will be opened, and the rest of it is easy. (3). In many cases, the failure is related to the unwillingness to learn something new. (4). I was shocked recently when a project engineer of a 100 year old turbomachinery company asked me the question:"what is cfd?". When I mentioned the story to a friend of mine, his comment was that the person still has the courage to ask. (5). If you can write a story on paper using pencil, you can also write the story using a typewriter. It is getting more difficult to find a typewriter. So, one can use a computer and a word processor to write the story. The same is true for the design, analysis and testing. (6). Design is not drafting, it is an iterative process to create something useful based on the existing rules and information (in chart and equation form). If you can design something using paper and pencil, you can also use the computer and the CAD. One can even program the equation or chart into the calculator, the programmable calculator. The design manuals are just charts and equations for designing the products. (7). Beyond this stage, we need to do analysis and testing. The cfd belong to the analysis part, which has the same goal as the testing to produce the information to aid the design. In other words, cfd and testing results must be analyzed and transformed into useful design information. The machine and the computer can not handle this part yet. (8). Currently, there is a big understanding about cfd in terms of the nature and operation of the activities. This is because the designer using CAD tends to think that cfd is the extension of the CAD. And in many cases, the cfd code needs the input from the CAD. The consequence of this misunderstanding is that people tend to put a deadline or a schedule on the cfd activities. (9). You can't do it this way!!! And you are going to fail!!! (10). The proper way to handle the cfd activities is to treat them as if you are planning for the testing. This is the trade secret for the successful implementation of cfd in any organization. (11). In the testing, you need the wind tunnel or test facility. So, for the cfd, you also needs the computer facility and supporting staff. You can't say that we already have many PC's doing paper work why not just use them. That's the rule number one. The manager must be able to say that this is the computer to do the cfd, and why. And that is the rule number two. (12). To do cfd, you need a plan just like the test plan. You need the time to do the data reduction. I think everybody involved in testing know that it is not trivial. It not only requires manpower, but also needs the data reduction program. And these must be prepared ahead of the time. But many people think that somehow the cfd code can crank out beautiful pictures instantaneously. It is hard for the manager to understand the computer graphics, thus they tend to avoid getting familiar with it. If you ask the manager to learn how to run the post-processor first, it is easier for him to estimate the time required to do the post-processing or the data reduction. (13). Well, back to the model design, model making, model instrumentation. These steps are just like creating the model geometry on computer, making the topology blocking, and the mesh generation. These steps are interlinking. To measure the surface pressure, one needs to drill a hole and connect it to a tubing. So, the model has to be designed in such a way that one can do so easily. The same is true for the cfd geometry model, because it will affect the accuracy of the results. If one create a model in such a way that only coarse will be generated in the region of interest, then the results obtain are not going to be useful. If this is not planned, then a hole will be drilled in the area which prevent the accurate measurement of the flow variables. (14). But what about the solver itself? It is equivalent to the operation of a wind tunnel. A wind tunnel is not a structure with air running inside. There are subsonic wind tunnels, transonic wind tunnels, supersonic wind tunnels, tri-sonic wind tunnels, hypersonic wind tunnels, special test cells for jet engine testing, for rocket engine testing, etc... In the same way, the are different kind of cfd algorithms and programs to solve the governing equations. (15). It is true that there is no need to understand the design of a wind tunnel. But if one does not understand the principle of wind tunnel operation, then it is going to be very difficult to write a testing plan. If one does not understand the cfd operations, he is not going to know how to make a plan for the cfd work. (16). I am taking time to write down this seemingly common sense information, because there are really many stupid idiots out there. (17). So, the principle is this: cfd is identical to the wind tunnel testing, the difference is that this wind tunnel is virtual wind tunnel which is inside the computer. Do not even think that cfd is linked to CAD or a word processor. I know that some people are actually using the word processor and spread sheet to prepare the chart and the report (or even to do the data reduction from the wind tunnel results.) (18). If a manager is asking that can you get the results back next Monday for the early morning review meeting, you know that his company is dying. For the test, they normally have the schedule planned well ahead of the time. (19). You can still survive in the future without using cfd. In that case, you must know how to do the testing and be able to do it in your facilities. (20). Thinking about outsourcing cfd to a developing country?

(1). There are a couple of typing errors: "understanding" should be "misunderstanding", "coarse" should be "coarse mesh". (2). Sorry for that. I think, it is obvious for our smart readers. But still, for accuracy, I need to make the correction.

I agree with all your points 100 percent.

I usually don't give personal accounts but this one is interesting (I think)

Unfortunately, I have also noticed that from the early 80's when nobody would believe CFD at all (thinking it is only a toy) through today, there has been a full swing in the opposite direction in favor of cfd when everyone (not seriously knowledgable about the field) now thinks all their engineering problems (that is the actual thinking part) can be solved by CFD. This is bound to create alot of disappointment to many, if it hasn't already.

In the mid-late 80's I was in charge of the CFD part of a joint experimental-CFD analysis of acoustic damping studies for a rocket engine company. There was tremendous resistance by the old timers (and some younger ones as well) to believe any result that CFD would "spit out". We had gone through the entire experimental exercise and had gathered shock tube data. However, my CFD results would not match those of the experiments. I am talking huge differences in terms of orders of magnitude not percentages. Well, needless to say that everyone was dumping on the CFD toy (and how useless waste of time that was). However, I knew better, because I had done everything correctly. So my task was to prove that the experimental data was wrong, and indeed with the help of the experimentalist we proved that the experimental data were completely flawed on physical grounds. The entire experimental setup had to be re-fabricated, etc. This was a switch! Unfortunately, this experience raised the expectation of the skeptics of cfd unrealistically too high, to the point that they wanted to do away with experiments. And they wanted to do cfd of the rocket engine combustion! So, then I found myself in the awkward situation of defending experiments.

The point is that I believe (and I think "fair" people with many years of experience) believe and recognize the need for both experiments and cfd as a tool only. The actual design task still remains on the engineer's shoulders.

A side lesson for the novice: Don't always think that you need to blindly match your cfd results with the experiment. Look at the physics and try to make sense of the results, both experimental and cfd. And, don't take anyone seriously who suggests you need only one of the three main ingredients for a successful product design: 1-Good/experienced engineer, 2-cfd, and 3-experiment.

Adrin Gharakhani

Our company MCube specializes in the numerical simulation of aeroacoustic phenomena and we work mostly on the fan, exhaust, intake and components CAA problems. Our code, RADIOSS-CFD is a compressible explicit solver with an explicit time integrator able to capture precisely pressure waves propagation and hence treat both CFD and acoustics at the same time. We recently refined our LES turbulence sub-grid model and obtained a broad band noise within 3 dB of the experiment for an automotive centrifugal fan. Modal behavior was also extremely well captured (blade passing noise, cavity modes...). You can check our web site www.mcube.fr for examples in this domain.

(1). I think, your story and experience worth a lot of money. (assuming that the reader can understand it and be able to follow the suggestions, he is going to save a lot of money and be able to accomplish a lot.) (2). If the company has a group of experienced engineers, who know how to do the testing, then with the help of the cfd, they can improve the productivity and the quality of the products or services. Very simple guidelines, yet, in many cases, it is very difficult to follow. (3). Like the old saying, "A craftman wishes to improve his work, he must first sharpen his tools." At the cfd coding level, it can be the development of more accurate mesh generation method and solution algorithm. At the working level, it can be the right planning of cfd activities along with the testing. At the company level, it can be the essential organizational structure for the experienced engineers, testing and cfd activities. (4). One more suggestion,: if the test results change with the size of the probes used, his results are not reliable. The same principle applies to cfd, if the results change with the mesh size used, then the cfd solution is useless. ( In the linear equation, the coarse mesh solution may not be very far away from the real solution, but, in non-linear equation, the coarse mesh solution is likely sitting on the opposite side of the real solution. That's why the name of non-linear is given.)

I notice that there is not a lot of discussion of one commercial code versus another... My company needs a commercial code to solve incompressible, steady air flow problems with heat transfer. A high priority is cost of the software, it would not be used extensively and our budget is tight. Would like to be able to construct new tube bundle type geometries easily. Any reccomendations? Anyone have experience with CFD 2000 from Adaptive Research?

Thanks