CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   What is the Better Way to Do CFD? (http://www.cfd-online.com/Forums/main/3337-what-better-way-do-cfd.html)

John C. Chien April 7, 2001 15:29

What is the Better Way to Do CFD?
 
(1). Even though everyone knows that modern society is derived from the industrial revolution based on the thermo-fluid dynamics, we are still having great difficulties in understanding the physics of fluid flows. (2). Even though computers are now nearly everywhere, and Internet can reach anywhere in the world, we are still having problems in CFD, not to mention the jobs in CFD. (3). So, it seems to me that there is a need to have a general discussion about "What is the better way to do CFD?" (4). Is CFD just to run some existing codes? What is your point of view ? (5). Who should be doing CFD? the school? or the industries? (6). I think, it is about the time to talk about the better way to do CFD, so that the student, the teachers, the schools, the research labs, the government labs, and the whole industries can all benefit from doing CFD. With this, I hope that there will be enough jobs for everyone who is really interested in CFD and we also will be able to see the benefit of doing CFD. (7). So, is there any better way to do CFD? This should cover all CFD related issues and fields.

John C. Chien April 8, 2001 01:13

Re: What is the Better Way to Do CFD?
 
(1). The better way to do CFD can be divided into groups and questions. (2). Do you know everything about CFD? If the answer is No, what do you want to know to improve your understanding of CFD? (3). Do you think your products can be improved by using CFD? If you are not sure, what do you want to know to change that? (4). Have you written a CFD code before? If the answer is No, do you think that it is important to learn how to write a cfd code? (5). If you are using CFD right now, what are you looking for from doing CFD? Do you think that your current CFD activities are the best approach? Are there rooms for improvement? (6). If you are not satisfied with your current CFD activities, then which areas require further improvement? Faster computers? better salary? more engineers? better education? or...? (7). What is your goal of doing CFD? To design a better product? To obtain the best solution? To get a higher paid position? To publish more papers? Are there better ways to achieve your goal? (8). Do you think that the market is crowded with CFD engineers and you have to work hard to stay alive? If the answer is yes, is there a better way to improve it? (9). Do you think that CFD should stay in school and research labs? So you can do researches in mesh generation, turbulence modeling, and numerical algorithm development without worrying about the time schedule? (10). If the code you are using is not producing good results, what is the better way to improve it? go back to school? find someone else to do it? (11). These are just a few questions you can try to answer. And hopefully, in the process, you will get a better picture of how to do CFD in a better way.

A.Hassaneen April 8, 2001 07:15

Re: What is the Better Way to Do CFD?
 
I believe CFD is ( money driven) just like any thing else in our life today. Although it started in schools as an academic subject some 30 years ago, it became fully commercial recently. i.e. the improvement or the way to do a better CFD is strongly affected by what the industry needs from the CFD codes. In other word, it depends on how essential CFD is for industry. Moreover, before we start this (reform) process we should know what are our klints needs. sorry if my english is not as good.

John C. Chien April 8, 2001 13:19

Re: What is the Better Way to Do CFD?
 
(1)."i.e. the improvement or the way to do a better CFD is strongly affected by what the industry needs from the CFD codes." So, what you are saying is: the industry knows what they need (the code) to do their business? (2). If that is the case, then we really don't have to worry about them. When they need more engineers to perform CFD tasks, they hire more. On the other hand, when the tasks are completed, they fire the engineers.

Greg Perkins April 8, 2001 20:07

Re: What is the Better Way to Do CFD?
 
From a company's point of view, maybe the idea of just CFD is too narrow - how to make a better, cheaper, etc. product is the broad question. CFD is one tool, albeit a complex one for a very complex set of physics.

One the other hand, for a narrowly defined, essentialy technical/academic project, there may be 'better' ways. But it might be difficult to generalise as to what they are. For some applications developing expertise inhouse and hiring PhDs etc. may be appropriate while for others using leasing a commerical code for a month and having a graduate engineer do some analysis might be 'best'.

At the moment, I think from a technical perspective it would be nice to have a freeware CFD code, like the Linux OS, which is robust, fairly complete and documented, but for which the source code is also readily and freely available. I'm quickly being converted back to writing my own codes etc., after the difficulties with some commercial codes. But that's not to say they don't have their place.

I even think that their might be room for like a commercial product that is really a collection of modules which can be linked up at will to form a customised user code. Something like Lego for CFD. Patch together your equations, boundary conditions, a matrix assembler, a matrix solver and some post processing tools within an object framework - and well maybe you can have a flexible product. ??

Greg

John C. Chien April 8, 2001 22:15

Re: What is the Better Way to Do CFD?
 
(1). Very good ideas. (2). I am sure that one of the major difficulty in CFD is the creation of a super flexible, easy to use cfd code. Along this line, I think, there is still a bright future. The code itself definitely can be improved in various ways. (3). But I guess, in order to achieve that goal, there must be some practical way to do it. (4). From my point of view, to make a code public, itself is not a problem. I have seen many of these codes in the past.(with some restrictions such as export control. But that is because those codes were developed by government agencies or under government contracts for specific use.) (5). I think, it is not very good idea to withheld information from the users, just because the code owner is trying to protect his financial interest. The consequence is the long period of suffering by the users of the code, and also negative impact on his project or products design. (6). I always think that the management has to understand how to implement CFD, because CFD is the other side of testing. The management probably has a better understanding of the testing. So, that is another area for improvement.

A.Hassaneen April 9, 2001 05:00

Re: What is the Better Way to Do CFD?
 
What I'm saying is that CFD development at the Universities or research lab in most cases is trying to find an answers to industry questions. That is (in my opinion) why they do CFD these days even if these answers are for questions still to come. i.e. CFD at the universities depends in most cases on the funds from industry or people who needs the results.

A.Hassaneen April 9, 2001 06:09

Re: What is the Better Way to Do CFD?
 
Who will develop this complete super flexible, easy to use cfd code for free???. I'm thinking about collaboration between the vendors and universities to produce that aforementioned cfd code at a reasonable price (less than $1000 for the whole code). The university will do the solver part and keep improving it according to suggestions from the users. The vendor do the pre and postprocessor parts and also keep improving it. Let us make it a target:"a complete super flexible, easy to use cfd code for less than a $1000", who say yes???

reza April 9, 2001 06:57

Re: What is the Better Way to Do CFD?
 
hi i believe before thinking of a code we must think of an analytic and theoretical base to be provided for all the people coming into CFD and more generally the complete scope of fluid and gas flow and interaction studies...

best regards

ken elms April 9, 2001 11:45

Re: What is the Better Way to Do CFD?
 
I think this question deserves a much wider playing field. What is reassuring is that interest in CFD maintains a good momentum with benefits to all sectors continuing.

The difficulties of fundamental science translation has been due to an inertia of proving its full worth. Our whole gambit of knowledge from basic theory to standards,units and measurements coupled with how best to use it remains still a conundrum.

The diversification of problem areas-space flight,harnassing energy,product fluid interactions,complex shapes,heat,mass and chemical transfer.It goes on.

Can you image when engineers used regularly factors of ignorance [for want of a better term] with material strength having 5 times more than a worst senario could throw up. Look at how FE tackles successfully that sort of issue saving material and cost benefits.

CFD has never been [or will in my lifetime] be a standalone all purpose method of solution. Each field hones its attributes -mixes or matches workable codes against the welter of current knowledge versus dare I say it-areas where were we really don`t fully know what goes in the real physics or mathematics.

If we were to develop[given all the resources] a unifying code what future awaits the next and future generations. Robot coding ,Robot design and Manufacture. Thankfully the real situations remain ahead of the best models we can throw at them. We are getting closer but the future remains a mystery still.

The pioneers of CFD used imagination,flair,met setbacks resolutely,never promised to deliver a unique standalone science-but adapted to demonstrate the worth of their scientific based work for problems of the era like defence or leading their respective fields and staying ahead of the global competition, winning markets.


John C. Chien April 9, 2001 19:36

Re: What is the Better Way to Do CFD?
 
(1). It's not going to be written by the existing commercial cfd codes for sure. So, at the moment, we will have to assmue that it will be written by someone (a human being). I am not saying it's easy or practical at all. (2). Between vendors and universities? Unfortunately, Verdors are for-profit organizations, and universities are non-profit organizations. Both can not think at all. They sure can spend money at very fast rate. (3). I think, that is one idea, but who in the company or university should create these code or tools? Along this line, nothing can be done, until the person(s) can be identified. (4). Before that, one need to identify what this person(s) should do to create this super flexible code(s). In other words, what are the essential parts of this super flexible code(s)?

John C. Chien April 9, 2001 19:58

Re: What is the Better Way to Do CFD?
 
(1). Well, good thinking. But it is not 100% correct. (2). First of all, the graduate students need the degree, so they need to write a thesis or a dissertation. For that, they need a subject. In fluid dynamics in general, there are three approaches, one is analytical, another is experimental, and the last is CFD. (3). With the subject selected, the student can start working on it. If the student need the financial aid, he can get a loan, or scholarship. And if the professor is working on the same subject through a project, the student can also work on it and get paid. (4).It is possible that sometimes a company will contact a professor for his opinion or give him a small research money to work on it. This seem to be a good place to start looking. What are these companies? Do they have the money to support the university research? Do they have the infrastructure to learn, absorb and use the technology developed? I mean, if the student work on the project, what is the possibility of getting a job when he graduats from the school? (5). In this case, the future seems to be more important than doing the research or getting the degree. You don't want to increase the jobless rate, right? So, what is the other end of the pipe-line? A perfect cfd code selling for $1000, with no one buying?

John C. Chien April 9, 2001 20:06

Re: What is the Better Way to Do CFD?
 
(1). Well, I have noticed that many engineers,without good training in CFD, are learning how to use commercial cfd codes. (2). Should they be sent back to school first, so that formally trained CFD engineer can have the opportunity to solve the problem more efficiently? (3). What courses should be provided in school to properly train a CFD engineer or researcher? And to what degree? (4). Can we list some fundamental courses for CFD students? What do you think should be the minimum requirement for a cfd student?

John C. Chien April 9, 2001 20:22

Re: What is the Better Way to Do CFD?
 
(1). There is definitely a huge difference between the structure/stress analysis and CFD. This comes mainly from the convection term of the governing equations. (2). So, apparently the effect of the convection term varies greatly from problem to problem. But that does not mean that there is not solution to each problem. (3). It simply says that each problem (definitely includes the geometry) will have to be addressed individually. (4). Following this line, the idea of the super flexible and easy to use cfd code(s), seems to have a lot of technical problems to solve before it can become practical. (5).It seems to me that we need to address the technical problem of the flow to be solve first, then write the code to solve it. Not the other way around like what some people are doing right now, that is first develop a super flexible and easy to use code, then find an application problem for the code. Is this what you are thinking?

Greg Perkins April 10, 2001 03:59

Re: What is the Better Way to Do CFD?
 
I think the idea of universal application is too ambitious and more importantly, not generally worthwhile.

Individual users/companies normally have a small set of related problems to solve. No individual/company etc needs one tool to solve all of these small related problems - it just doesn't make sense.

Having said that, there is a trend which will continue to add more complex phsyics and geometry to problems to be solved numerically. This is driven by a) faster computers b) more available codes and c) some new knowledge of the physics (although this is generally pretty small/secondary). a) and b) will continue and in some sense feed off each other. c) grows slowly thus most of these available codes have fairly primitive models inbuilt, wrt to the state of the art - despite what they might say in the brochure.

You just have to look at some of the old research codes from Lawrence Livermore lab., to realise they had unstructured grid adaption nearly two decades before commerical codes! And the current ones are where cfd is headed.

I suppose the need for general purpose codes is to a large degree driven by economies of scale. Its probably easier and cheaper to implement (initially), maintain and support one code than ten. For some problems, like combustion, you need a lot of these sub-models to get any solution. On the other hand, making all the sub-models work together is a significant problem.

Just look at the yet-to-be-released Fluent 6. That's been mooted for well over 1 year now and will still don't have a release date. There must be some major difficulties getting it all to work, which is very interesting given they already had most of the models implemented in an older, but structured code.

The idea of integration with fancy A.I. stuff sounds good. I want to do something in this area too, but really I think its more important for engineers to develop their own intuitive ability for specific problems than seek a brute-force all in one solution. Having said that, it seems to me that more automated, partially optimising methods are a good way to give engineers more time to spend thinking about the problem. I really didn't go to uni to spend 30% of my time creating mesh nodes, and recovering from software crashes!!!!

Greg


A.Hassaneen April 10, 2001 05:39

Re: What is the Better Way to Do CFD?
 
We cann't build a super flexible code to solve all the problem, why? because of the new theories (every day) which explain some of the most complicated problem for example: turbulence. That is why I believe that we should build it in a most complete shap and keep improving it.

A.Hassaneen April 10, 2001 05:42

Re: What is the Better Way to Do CFD?
 
we meight achieve this by good marketing people for our target price to be accepted by the profit companies.

A.Hassaneen April 10, 2001 05:48

Re: What is the Better Way to Do CFD?
 
I believe that the schools should focus on the mathematics part of CFD.i.e. the theories, the mathematics, numerical analysis, the modules and the solver as a whole.

A.Hassaneen April 10, 2001 06:00

Re: What is the Better Way to Do CFD?
 
No, if there is a SFUF (super flexible user friendly) code for $1000, this will promote the companies to explore the approach and if they get rewarded they will invest more on CFD and hire more engineers. It is like the big stores when they sell for half price and they make profits more than 500% because they sell 5 or ten times. What do you think?

John C. Chien April 10, 2001 06:19

Re: What is the Better Way to Do CFD?
 
(1). I'll be away for a few days, at the same time, keep writing, and see whether we can get a rough model of how to approach CFD in a better way.


All times are GMT -4. The time now is 06:23.