(1). I think, that is one area where the human brain, the thinking logic, the computer system, and the computer language must work together. I said "must". (2). The computerized automatic optimization can be done. But the reality is more jobs will be eliminated. (3). Currently, the design systems are configured to fit the human activities. The simplest example is the use of CAD. Stored data base in parametric form can easily eliminate the CAD engineer, not mentioning the smart CAM programs to create optimized machine tool path automatically to create the parts. (4). It is always important to think whether the future will be "machine-centered" or "human-centered".

(1). The future of CFD depends on whether it can serve the human being or not. (2). To serve human being effectively, CFD must take in a proper form. (3). One of the form is a CFD code. (4). Take for example of an automatic IC plant, where an IC chip is automatically produced through several hundred of different steps. When the final chip has one defect in it, it is garbage. You can not say that one error out of one million semi-conductors on the chip is only 0.000001 . (5). So, sometime in the future, when this automatic fully-optimized cfd system produces a product(say a blade geometry) which is obviously wrong, are you going to use another system to correct it? Or at that point, you still have some talented engineers around to find the source of errors and fix it? Or just like the bad ICs, they simply ended up in the garbage cans. (6). You can build many redundant systems into this automatic optimization system. But in the end, like space shuttle systems, launch vehicle systems,... it still depends highly on human being's brain. (7). Can you write a code to automatically check out a CFD code and fix the bugs? So far, most commercial codes or in-house codes can only stop the execution when the solution diverges.

Human brain is the biggest supercomputer. The computerized automatic optimization is designed by human. It seems like in future only the creative people in their field of specialization will be in top. I agree with you that in reality more jobs will be eliminated with modern computerization. Those who are average or below average in their field will do only the same kind of work on a daily basis. Those scientists or engineers who have facilities to use supercomputers to solve CFD problems in reality will be in much better positions than the CFD analysts who only have access to commercial softwares. In future,(since commercial softwares have limited applications and memory allocation) commercial CFD softwares will be used by average and below average CFD analysts to solve the same kind of problem on a daily basis.

(1). In some companies, it is already in this way. (2). Similar to Henry Ford's concept, there will be CFD assembly lines. And there also will be researchers and scientists working on the more fundamental issues like numerical algorithms, turbulence modelling, anmation and visualization, integrated design information systems. (3). Recently, someone has already modified a PC CPU to run at 800MZ or 1000MZ. So, everyone will have access to the super-computer class PC workstations. (4). It is therefore very important for the engineering managers or company CEO to have a clear picture of this dynamic development of both the hardware and the CFD. Failure to do so will create great chaos in the company's engineering department and unable to produce competitive products in time to meet the market need. (5). This complex technological environment will pose a big challenge to the future comapny planners. I would say, from the use of the computer systems alone, today's system of networked distributed computing environment is much less efficient and reliable than the old centralized main frame computers. The routine network down problem is fairly common to many morden companies. What I am trying to say is: many companies are running CFD problems on workstations (or networked computers), while these problems should be run from the begining on the super-computers. The mis-match between the computer and the cfd problem has a bad consequence of un-reliable and un-useable results due to inadequate mesh points and poor models.

Are you kidding me John! I mean Cray, Teraflop etc. etc. .... !

(1). I mean old Cray.

"NAS" Cray system

(1). The information forwarded to me about the 1000MHz and 800MHz PC CPU from my son ( who has a popular website on libretto. http://www.cerfnet.com/~adorable/overclocking.html ) is : (2). http://www.hardwarecentral.com/hardw...torials/718/1/ (3). http://www6.tomshareware.com/cpu/99q3/990823/index.html (4). The old Cray I used has one Meg words fast RAM only. We could only ran a small 3-D Navier-Stokes problem. A mini-computer then was using virtual memory and can handle large 3-D problems. (5). If you are using the current super-computer, you are working in the paradise.

John, you are trying to undermine the supercomputer used by Nick Georgiadis along with his group at NASA.

(1). The old Cray computer I used was belong to a private company. In early 80's, because of the need to do CFD work, we were able to get the company to buy a Cray computer. (2). There are many many companies in the world using super-computers. I think, you are right that NASA also uses super-computers. I don't know what computers they use. I have no connection with any NASA before. (3). I am talking about my personal experience using the super-computer and PCs. The only software I have used which were developed by a contractor at a NASA lab is Plot3D (and FAST). I don't see your logic about the connection between my experience and the NASA. (4). You are still free to create your own story.

i posted to this thread a few days ago saying that the future of CFD was in all sorts of automatic design techniques (optimisation , inverse design , neural networks etc) but i've come to the realisation after reading through the thread "extremely simple problem... can you solve it properly" posted by some guy named Mikhail that the future of cfd indeed all of engineering is proper education and a real knowledge of physics. this man was going on and on about some mathematically well posed but physically impossible problem that is best solved by hand. it troubled me that this person might be the man designing my gas stove or something. i think the best thing we can do is to pass on our knowledge

Don't worry about such posted message as "extremely simple problem... can you solve it properly". The thinking process of such message is reversed. We have firstly the problem of physicas then we model it mathematically. The reversion has no sense. It is somewhat like the confusing drawings of Escher: The structures he drew can not exist in the reality but can exist optically on paper.

About the "expert system", I mean a design software system containing experience and wisdom of million researchers and engineers working on fluid dynamics. There has been a software named as "Invention machine" containing the experience of the engineers of the aerospace and millitary industry of the ex Soviet Union.

X. Ye

oh yeah i know about expert systems. but the point i was trying to make before was that most of the time one optimisation is notdemonstrably better than the other ie each has advantages in certain areas and disadvantages in another. so a combination of these is better than one. if you'd like could we start a new thread to discuss this. i don't like hanging on a thread if i'm discussing the original problem