CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   CFD Design...The CFD Future (https://www.cfd-online.com/Forums/main/1335-cfd-design-cfd-future.html)

John C. Chien September 26, 1999 17:22

CFD Design...The CFD Future
 
(1). Other than publishing CFD papers on technical journals and study the fluid dynamics phenomenon, the current useage and vision of CFD is rather vague. (2). Will the use of commercial CFD codes become the main activity of the CFD in the future? So, you have many CFD technicians running various CFD codes trying to get converged results. Is this the future of CFD? (3). Without going into the detailes of the future of CFD in academic institutions and in consulting business, one really must ask the basic question of : " Is CFD really useful? And if so, How do you make it useful. " (4). This question is for persons outside the academic institutions and the technical consulting business. Because for the persons inside the academic institutions, the main goal is to publish new results. The usefulness of the paper is never asked. This is because, the contents of the paper could become very useful in the future. For the persons in the technical consulting business, the need is straight forward, that is to crank out numbers for the customer and get paid by hours. The working environment is ready along with the computers and the softwares. What is required is the experience in using the software. There are various reasons why such jobs exist. In reality, persons working in this environment do not care. (5). After having covered these two areas, we now come to the question of usefulness of the CFD. And this is the future of CFD, I think. (6). This is really a strange question though, because our environment is fluid (air, water, and gas ). From the steam locomotive, power plant, the car, the airplanes, the water pumps, the electirc fans, the air conditioners, the gas stoves,....etc.....to the warfare such as fighters, bombers, rocket, missiles, launch vehicles and even the satellites, all requires the use of fluid dynamics and aerodynamics. (7). So, looking from all of these products and ask a question about the future of the modern CFD is somewhat hard to understand. What's wrong with the current state of CFD? (8). If we look a little deeper into the design processes (or the invention processes) of these products, we can easily find the use of fluid dynamics everywhere. And I would say that in many cases, these processes are computerized. I mean computer codes have been developed to solve the design problem of these products. Yet, from the current understanding of the CFD they are somehow outside the range of CFD or at best near the boarder line. (9). So, without going further into the detailes, I think, the future of CFD lies in its ability to acquire the "DESIGN" quality and produce the products directly. I think, this missing part of the puzzle (the DESIGN QUALITY) will be the key issue in the future time. You can't sell a very elegant numerical mesh to a customer, but a pump with higher efficiency using such mesh in the CFD design will be welcomed by the customer and the national power planners. (10). The whole world is waiting for you, but first, you must turn the CFD recipes into a better working product through the CFD DESIGN. How to achieve that goal, perhaps is a very important issue for the persons inside the academic institutions, the consulting business, and the rest of us. What do you think?

Aleksey Alekseev September 27, 1999 16:02

Re: CFD Design...The CFD Future
 
You are quite right! We do not need flow, we need maximum lift, thrust, minimum drag, dissipation, minimum/maximum heat flux etc. So, the future of CFD is in optimum design, control, parameter estimation - in Inverse Problems. Look at Heat Transfer- every Forward problem is surrounded by a lot of Inverse. Future codes will contain (as a minimum) a forward problem (CFD), adjoint problem for fast gradient calculation (hand made or from Automatical Differentiation) and some minimizer (conjugate gradient or quasi-Newton). Aleksey

John C. Chien September 28, 1999 12:40

Re: CFD Design...The CFD Future
 
(1). I think, the current CFD is following the foot steps of the wind tunnel testing approach, going from the models to the solutions and the data reductions. Basically, the current approach is essentially the trial-and-error approach. (2). The basic problem is that it is neither a design approach nor a learned process. (3). After so many years, the newcomer still "trying" to solve some basic problems which have been solved numerous times before. It is a waste of time and effort. And I am not sure that one will follow the same process , use the same grid and get the same results if he tries to solve the same problem twice in six months. It is not a learned process, and thus it is wasteful. (4). It is not a design process, because the current CFD results can not direct the user to make the change to the design. If you are given the 3-D transient Navier-Stokes solutions for flow over a turbine blade, can the results tell you which way to improve the blade design? (5). Design is not just running some testings or computer codes, so, the current approach to CFD must be changed.

clifford bradford September 28, 1999 19:51

Re: CFD Design...The CFD Future
 
these are some deep issues that i think affect about half the readers here (ie practising engineers). (1) one questions is: Is CFD adequate as a design tool? yes and no. yes because i think we have tools (algorithms, solvers, mesh generators, optimisers and inverse design codes etc) to make CFD a design tool. no because of two reasons. (a) these tools have not been integrated so that industrial users can take advantage of them (say in a commercial code). every package for CFD has strengths but every one have enough weaknesses to preclude it from being an adequate (not even excellent) design tool. what i'm trying to say is that we're not integrating the best of breed in each area of fluid design into our CFD codes (a code may have a good solver but poor turbulence models for example) (b) in general the industrial users of CFD do not have the knowledge to use CFD codes as design tools in general engineering design. this is a problem of knowledge dessemination (the originators of knowledge aren't efectively communicating that knowledge to the user community) and acquisition (the user community is not actively pursuing knowledge). these problems spring from the fact that in the field the main source of edification is a wide variety of journals which are not suited to acquisition of knowledge from the basic to the advanced level in a smooth progression and so users cannot make out the forest for the trees. the lack of introductory, intermediate and advanced texts on the subject make it difficult for the user to gain knowledge in the logical progression required for deep understanding.

i would say that many industrial processes and devices are beyond the present capabilities of design level CFD. however CFD allows the knowledgeable designer to make assumptions and use the CFD results as a better approximation to the design than what he was getting before. i'd say in the industrial world that there are 'haves' and 'have nots' regarding CFD use in design. the 'haves' have good codes and a institutional knowledge of the benefits and limitations of what their CFD can do for them. as a result they make best use of what their CFD can do for them even if it is very approximate. the 'have nots' have little or no idea of the limitations of their CFD or ways in which to use what CFD they have as best they can. as a result they don't get value for the money and time invested in CFD

John C. Chien September 28, 1999 23:13

Re: CFD Design...The CFD Future
 
(1). This is a heavy weight issue, I'll take time to read your message a few times before make a comment. I think, you probably agree with me that we shouldn't turn the CFD into something like "key-punch operator" job. (2). Today's engineers are invading the secretary's job. In old days, we used to have a secretary for every 10 to 15 engineers. Now, it is really very hard to find a secretary in the work place. (3). You know what? they should invade CFD instead of word processing. (4). I'll come back to answer your message later.

Tareq Al-shaalan September 29, 1999 08:05

Re: CFD Design...The CFD Future
 
CFD is not complete withought the power of human being brain.

CFD is only a tool; it will be beneficial and excellent tool when it is used wisely. CFD needs knowledge not only in computation (C) but also in understanding the physics of the problem (FD).

For industry to gain full advantage of CFD they needs to hire a person whom well knowledgeable of both Computation and Fluid.

Scott Rubinstein September 29, 1999 10:33

Re: CFD Design...The CFD Future
 
(1)First a little background: I've been using a commercial code for a year. My previous CFD experience was two courses in graduate school. I've periodically read posts on this forum, but this is my first posting. My job is mostly CFD and FEA analysis, and I've been involved in a number flow/erosion tests.

(2)I find it very difficult to have an impact on designs by using CFD. Often my analysis is sort of after the fact, and just provides a pretty picture. Often the fast turn around time required on some analyses leaves me very uncomfortable as to the meaning and accuracy of the results. Since I am the only person involved in CFD at this location, I am termed a "CFD Expert" (kind of scary)

(3) Having read previous posts about the future of CFD, I am certainly concerned if I am making a mistake devoting some much time to running commercial code.

(4) On the brighter side, I've been working on some simple models trying to develop some basic erosion models. There is an abundance of emperical erosion formulations around, and maybe I'll be able to use this as my means of improving designs.

(5) Thanks for listening.


Dian Georgiev September 30, 1999 02:24

Re: CFD Design...The CFD Future
 
This problem is clearly methodological. The widespread use of scince in all the areas of life brings many achievements, but also a negative back impact on scientific methodology, due to the commercial, not consistent criteria of social life. This fact is well illustrated in the field of CFD: so many modeling methods and calculations and so little consecutive, clear and verified results for the benefit of knowledge. And since the inherent complexity of fluid dynamics, such lack of clarity makes the heap of numerious CFD modelings obscure and often useless for the advancing of hydrodynamics, and also design, which depen on it.

In this situation of lack of general methodological criteria, the only way of the designer firms is to create its own methodology by consultations, "model, compute and test" etc. It would be very good if some of them meanwhile obtain more general results and have also the good wish to teach others to them, as such things happen in science. In this sense I agree that "design quality" could be very usefull, aldough partial criterion of CFD modeling. However, the more general solution should be introducing of standards and relevant appraisement by special commities of CFD achievements. In this way an impartial basis will be established to replace the existing self-sufficient appraisement, which proves to be inefficient.


John C. Chien September 30, 1999 14:51

Re: CFD Design...The CFD Future
 
(1). I am going to summarize the four issues you mentioned. (2). We need a super CFD code for design purpose, with the best solver, turbulence model in it. A good idea, but who is going to do it. As a ongoing school project? or .....(3). Nobody is really teaching CFD and the industries really don't care. It is a failure of the whole system, the school, the teacher, the student, and the industries. And the journal papers are useless because it does not tell you the exact steps you need to get the answer. I would say, the school only teach the numerical part of CFD which is still very far away from the design applications. They need to teach how to design a turbine blade using CFD from the begining to the end. And the final blade must be a working one. Is there anyone teaching such course? How can one design a turbine without such a course? (they hope that somewhere in the computer system in industry, there is a magic code which will take care of the blade design.) (4). The real problem is too complicated and too difficult for CFD. I guess, in the post-super computer era, it is going to be very difficult for man to re-visit the moon. Simple problems have been solved long time ago, the remaining problems are very difficult. If CFD is no good for real world problems, then it is no good. (5). Some companies were able to take advantage of CFD while others were not. Are there such companies? If so, who are they? .....

John C. Chien September 30, 1999 15:48

Re: CFD Design...The CFD Future
 
(1). I think, the real problem is the lack of real understanding of the CFD technologies and the lack of ability to use it to solve design problems. (2). Running CFD computer codes really do not change the picture at all. This is because each component of CFD technology, whether it is geometry, mesh topology, mesh distribution, coefficients in a turbulence model, numerical solution algorithms, control of convergence,... all have direct impact on the final solution. That solution is just the begining of the design process, yet, it contains so many unknowns already in the solution. (3). We really must start screening the CFD technology in terms of its usefulness in the design process. Otherwise, CFD will become the slowest moving technology in the world. (4). In short, the CFD technology is not in its useful form today. It is in an abstract form of what could be solved. Something like a problem could be solved by several different types of meshes, turbulence models, and numerical methods, instead of the relationship of the result and the design goal. It is like saying that there are many ways to obtain two by adding two numbers, 1+1=2, 0+2=2, 0.5+1.5=2, 0.1+1.9=2 and so forth. For the design purpose, 1+1=2 is the answer.

John C. Chien October 3, 1999 21:50

Re: CFD Design...The CFD Future
 
(1). It is quite right. (2). CFD is a way of solving CFD problems, whether it is a design problem or a physics problem. (3). CAD is also a way of solving drawing and design problems. Even though CAD is quite mature in terms of the technology, we are still talking about CAD capability of a company per seat (per person). CAD has not replaced people. And sometimes, it is still much faster to use pencil and paper to create original design concepts and answers. The only good thing about the CAD is the files can be sent to the CAM programs and then create the parts on the numerical machines. (4). The CFD code is still far from the status of a "TOOL". It is not there yet. If you insist on saying that it is a tool, then I think it is a very primitive hand tool which always create unique answers. Don't feel bad about it, it is just my personal point of view. I had visited some Internet Pen sites, and they always say that the hand-crafted wood pen is always unique, no two are alike.

John C. Chien October 3, 1999 22:14

Re: CFD Design...The CFD Future
 
(1). As an engineer, there is no set methods one should use to generate quality design. I think, 1-D,2-D and 3-D codes can be used to help the design processes if necessary. But still the most important part is the common sense and the professional training. Even in the CAD field, there are many programs to choose from. I must say that, these codes including CAD are programs designed to help the engineer to make the life easier. (2). The hard part of real world problem is that it is always has the time and resources limit. Since the currently available CFD codes are mainly general purpose codes ( a groups of modules and libraries), it is not efficient for specific tasks under tight schedule and budget. So, there are lots of rooms for improvement.(3). I have written codes for IBM/360, CDC, VAX, CRAY,...PC.., and I think the software environment is like the computer itself, it is changing all the time. (4). My feeling is, a CFD code in general must satisfy the continuity and momentum equations, so, the results should be useful to some extent. And it is up to you to make use of it and interpret the results. That part of design integration in most cases is still missing. Just like a wind tunnel or a test cell, it can not solve the problem for you, unless you know how to use it in the right way. And in most cases, you have to modify it before you can use it.

clifford bradford October 4, 1999 12:37

Re: CFD Design...The CFD Future
 
i agree with you fully i think it stems from the fact that many cfd codes/algorithms/mesh generators are not being developed with design in mind. if this was the case then we wouldn't have people developing codes for structured single block grids - after all can you mesh an airplane with a single block structured mesh. most people working in CFD just want to get their name in J. fluid mech etc. in reply to my previous post John asked who are the companies using CFD properly. my answer is that they're not telling. you don't see too success stories from industrial world posted here.

i'll give an example of a code developed for design and not just making results for AIAA journal. in the seventies NASA wrote this FEM code for aircraft structures. do you know what one of the tests was. they designed a wing with the code and flew it on an experimental plane with strain gauges and a real pilot inside. i doubt seriously whether anyone has designed an airplane with CFD only and flew it with live person inside. incidentally, now NASTRAN is one of the most widely used FEM codes. that is real industrial code development

John C. Chien October 4, 1999 14:22

Re: CFD Design...The CFD Future
 
(1). A very good example indeed. (2). But, today, with budget cut, even NASA is having trouble with units used in the Martian probe computer program. (3). I agree with you that it can be done, and has been demonstrated. But it requires dedication from people involved. (assuming that they are willing to invest in it. Both the money and the effort.)

clifford bradford October 5, 1999 15:45

Re: CFD Design...The CFD Future
 
john i think it is a classic case of to make money you have to spend money (wisely). one of the most popular CFD codes is CFL3D NASA has spent years and mucho dinero developing the code. a professor in my intro to CFD course told us that someone at NASA estimated that they'd spent $15million developing CFL3D. of course now that money almost seems wasted as CFL3D is a serial code which runs well on old vector Crays. so now NASA is balking at laying out big $ again to develop a new (say) parallel code. I say though that the outlay is worth it. US industry has benefitted far in excess of $15 million from CFL3D so it isn't a waste even if that outlay would have to be made again on a new code. gotta run

John C. Chien October 5, 1999 17:03

Re: CFD Design...The CFD Future
 
(1). A computer game nowadays takes some 10 programmers and artists , two to three years of time, and several million dollars to write. (2). The end product is complete and costs only around 40 dollars for a user. For that one can start playing the game. (3). NASA should be doing things which will last longer and have greater impact on human being. (4). With the current rapid change in computer hardware and software, it is really not a practical proposal for NASA to develop a long lasting CFD code. (5). My feeling is that they should be dealing with difficult problems in aeronautical sciences and space technology. Making money is not their experties. Perhaps, they should find a solution to design a better wing and aircraft. Do they know how to design a better turbine blade? I would say that most problems are still largely un-solved. They have to use the facilities and people wisely.

Erich F. October 5, 1999 19:13

Re: CFD Design...The CFD Future
 
I agree with John, and maybe a bit off the subject - has anyone seen the article on Nasa funding poetry? While using funding wisely has never been a governmental strong point, I wonder if a new cfd code might get more support if put in ryhme...


Erich F. October 5, 1999 19:14

Re: CFD Design...The CFD Future
 
I agree with John, and maybe a bit off the subject - has anyone seen the article on Nasa funding poetry? While using funding wisely has never been a governmental strong point, I wonder if a new cfd code might get more support if put in ryhme... Maybe they could hand out metric to english conversion charts, too. ouch


clifford bradford October 6, 1999 12:37

rhyming CFD
 
i'll give it a try:

I think of a CFD code of mine, If it were TVD that would be sublime, but I find artificial dissipation works fine, but I think i'd rather use PUMA, than be a code tuner, so there'll be no CFD code of mine.

Jonas Larsson October 6, 1999 12:57

Re: rhyming CFD
 
Heh, reminds me of another nice CFD poem:

ODE to CFD


All times are GMT -4. The time now is 01:18.