Re: What is the Better Way to Do CFD?
 

John,

I get your point and one suggestion is - please be more succint.

In my earlier post I recognised the need for standards in order to have a modular system. This is obvious, since the communication between modules will need to be well defined - a standard. But that's no major headache - all software today is built on standards - any OS, like Windows, is a prime example. Of course it needs thought about the detail, but conceptually it's a proven method. Think also of the internet!

The real question at this stage, is does this idea have potential?? Of course, you could argue that until you have a system that works, it has no value/potential. But then on what basis do you expend effort to do something new?? You need to be able to evaluate ideas, not just the ones I'm mentioning, on some basis with regard to future potential/value/utility. To say that something doesn't perfectly work until it works perfectly, and is thus a failure, highlights the complexities of cfd, but doesn't in my mind address how to solve them.

To me, its easier to build and test modules independently and then re-use them for utility. However, I was suggesting that these modules are smaller, and more interchangable and modifable by users than current codes. Thus if you want to solve a complex problem you (a) don't have to re-invent the wheel by writing a cfd code from scratch and (b) aren't completely dependent upon the code vendor, like today, since you can interchange modules and write your own more easily.

Another thing I'm puzzled about, is it was your original post which started this talk, but then in your most recent reply you don't seem to be considering 'cfd of the future' much at all.

Greg

Re: What is the Better Way to Do CFD?
 

(1). The modular approach is in the right direction. (2). Actually, if the code is "open", then in most cases, the user can add and modify the code for his problem. And there is no need to re-invent the wheel. It can also be written in such a way that the user supply and replace the original modulars in the code. (3). So, this super flexible and easy to use code basically is sound and practical. As a matter of fact, I have been using this concept for a long time in my own code development. (4). I will get into more about the future cfd development soon. But I will have to do this offline first. and copy back to this message editor. There are a couple of issues need to be discussed, such as the degree of difficulty of the problem to be solved, and the offline pre- and post-processing, especially related to the control of the program. (similar to the traffic control issue)

Re: What is the Better Way to Do CFD?
 

(1). "What is the biggest advantage of CFD?" If we define CFD as the "Numerical Analysis and Mathematical Modelling in Fluid Mechanics", as an approach to obtain accurate data in fluid mechanics, then it is quite obvious that almost all engineering design and analysis require CFD. (2). Testing is another way to obtain data in fluid mechanics. But in general, it requires test facilities, experimental setup, and sometimes it is quite expensive (power consumption,etc.) to run tests. In ceratin occasions, it is simply impossible to do lab simulations. (3). Today, you simply can't design an aircraft or jet engines without using CFD. It is part of the evolutional process in design and analysis. In daily weather prediction, it is also necessary to use CFD, especially for the hurricane tracking in the summer through fall seasons. (4). "which type of problems lend themselves to CFD?" There is really no limitation in using CFD for design and analysis in fluid mechanics related problems. After having said that, it is necessary to define the time and cost involved in doing CFD. I would say that, this is the key issue of using CFD today and in the future. (5). If we can invent CFD software and the related processes similar to a digital calculator, then the cost to own a calculator and using a calculator will be affordable to everybody including engineers. The need to do better CFD is in the need to reduce the cost and time to do CFD. And I would make this issue the biggest problem currently facing CFD. (6). Apparently, if someone is selling you a CFD software and claiming that it could solve your problems accurately, then the issue of the cost and time involved in using CFD would not exist, and you simply go ahead and use the code. But as we have seen here, it is an over-simplified approach. And we are not getting the solution this way. (7). If by using the current commercial CFD codes, we are not getting the solution to the cost and the time problem in doing CFD, then what are the better ways to do CFD. This is the subject open to discussions, and I also have a few suggestions to make soon. I think, it is related to the issues of how to make it practical rather than just create a code for the user (which as you have seen are creating more headaches for users already). (8). I understand that it is a broad issue, but if you just look at your need and also the time and cost involved in doing CFD, you can easily see the big picture. Perhaps, it is easier to understand it by asking a simple question: Is it possible to make the CFD process as simple as using a calculator? Naturally, the second question is: Is it possible for the user to accumulate his experience and be able to understand and modify the code?

Re: What is the Better Way to Do CFD?
 

CFD most certainly has shown its vital importance to BETTER AND FASTER experimentation with a much proven track record in a host of fluid flow areas.Some sectors better than others it appears.Some codes better for one circumstance than another- But thats healthy choice [only when you can afford to mess up and that in this day and age is almost zero toleranced].

You give a special mention to CFD having a vital place in the aerodynamic field which afterall is where the nurturing grounds of its development and growth and success emerged. Boeing and the like are commercial too-but how much of their CFD expertise reaches the everyday scenes.?

How much longer did it take for the automobile industry to latch on to Cfd with its vast potential too.

How much investment is targeted to CFD Development in just this one industry alone? How vital is CFD to their future success.

My own area of interest[rotodynamic pumps] too has been relatively slow to grasp the benefits. Again as an industry how much investment is targeted toward CFD development in just this special but valuable product area.

The sponsors of each and every company on this excellent forum have their work cut out in moving their product base forward - but is there a case for specialising and more selectivity toward the end user.

Moreover, is there a vital need to bring in bigger bucks as ventured in e-commerce to forward the role of CFD.

Re: What is the Better Way to Do CFD?
 

(1). As a better way to do CFD, I am proposing a method to identify the degree of difficulty of a CFD problem. (2).To make the story short, one can assign a degree of technical difficulty to each component in the CFD problem. The highest value among the components will be the degree of difficulty of the problem. (3). In addition to that, the sum of all numbers in the problem will represent the degree of complexiety. (4). As an example, for the 2-D cavity flow with a moving lid problem, we have uniform mesh(0), square cavity geometry(0), laminar flow, low RE(1), Navier-Stokes equations(1), point iterative solution(1), 2-D line contour plot(1). (5). Based on this method, the degree of difficulty of the 2-D laminar low RE cavity flow is "1". In other words, this is a standard problem for all new comers of CFD. Everyone should have the capability to solve this problem. (6). The complexity of the problem is "0+0+1+1+1+1=4". This will provide the baseline for all other CFD problems and effort, so that the relative degree of difficulty and complexity can be identified numerically.

Re: What is the Better Way to Do CFD?
 

(1). Excellant questions and comments. (2). I'll have to address these questions carefully, off-line first.

Re: What is the Better Way to Do CFD?
 

(1). The research and development of CFD in aerospace applications has been cut drastically since late 80's. And the remaining researchers are depending on their consulting work (some are cfd code vendors). In other words, there has been no serious CFD development in aerospace applications since late 80's. This covers both the airframe and engine companies, because of the continuing merger and cut back. (2). If the company is no longer there, then it is hard to find the CFD group. And even if you can find one, they are mostly non-expert types. (3). In other words, automobile applications, rotodynamic pumps, jet engine applications and aircraft applications are roughly at the same pretty poor situation. (4). The difference is: It is well recognized in aircraft and engine business to do CFD, while in the rest of the application areas, it still requires some justification. (5). But this does not mean that aircraft and engine business are in the better shape than other field. The lack of funding in the aircraft and engine business actually make the CFD development and applications more difficult. (business consideration, burden of history, political strugle in-house, fear of job security, etc... tend to make the norm CFD development and applications more difficult) (6). On the other hands, in the rest of the fields, it might be easier to do good CFD development and applications because the field is sort of brand new. (7). The key issue is: the core of CFD is researcher himself, not the code. The lack of cfd development and applications means that: there is a lack of CFD policy, detailed plan, resources and execution. (8). In other words, researchers are not doing CFD research , development and applications. I have already seen the consequence of the problem in the engine business. It is pretty serious right now in terms of CFD capabilities in aircraft and engine business. (9). This is the reason why, I think, it is the right time to look at the issue of improving the CFD development and applications. I can't tell the best way to do it, but through this exercise, you might be able to get some directions. (I will have do some offline writing first. this is just to fill the gap for now).

Re: What is the Better Way to Do CFD?
 

(1)."You give a special mention to CFD having a vital place in the aerodynamic field which afterall is where the nurturing grounds of its development and growth and success emerged. Boeing and the like are commercial too-but how much of their CFD expertise reaches the everyday scenes.? " "Moreover, is there a vital need to bring in bigger bucks as ventured in e-commerce to forward the role of CFD." These are your questions. And I think, they are related to the planning and implementation of CFD in general. And these are very important. Instead of waiting for my off-line answer, I am going to give you some quick answers. (2). Throughout the industries and government laboratories for the last many many years, I have not seen any of my managers at work place who are CFD experts and have hands-on experience in geometry/mesh generation, turbulence modeling/physical modeling, solution algorithm development and computer graphics. Among the exception is the company which called "Sciences Applications International Corporation"(SAIC). This is a very big consulting type company. I had work for this company for a very short period of time of a couple of years. ( I remember Dr. Tony Hirt used to work at the headquarter of the company in La Jolla, California, before he created his own company. That was long time ago.) (3). So, you see, the biggest problem is in the management, regardless of whether it is an aircraft company, an engine company, a governmnet laboratory or a so-called non-profit organization working for government. It is hard to find a manager who is an expert in CFD and has hands-on experience in all essential phases of CFD operations. (4). This is a very serious issues throughout US industries and government labs as well. The CFD experts become the long term slaves in the company working very hard in front the computers, and chasing after jobs from company to company. (5). In the end, they try to work as consultant, or form their own consulting firm. So, I have nothing against the commercial CFD codes vendors, because some of them used to be the CFD experts working for a company. (6). So, in order to see the benefit of CFD, the management structure must be changed first. After that, there will be big bucks available to do CFD research, development and applications. (7). Otherwise, the experts will find ways to protect themselves by hiding their trade secrets to themselves. As you know, a single error can destroy the solution accuracy and convergence. The consequence is invisible and very serious. (8). I can only say this from the bottom of my heart that, if this is not changed, you will not going to see any progress in CFD, not even the technology which depends heavily on CFD. (9). The management solution to hire someone to run CFD codes (in-house or commercial) without solving this fundamental management problem can only speed up the death process. (10). In other words, CFD prohibits the promotion of an engineer to a higher management level. Without that management responsibility, you become a hard working computer slave. Do you see the future? (11). I definitely would like to hear from the CFD expert who is currently in the high management position.(make sure that you are the hands-on expert, not the one who runs the codes) (12). I hope that this provides you a quick answer for the moment.

Re: What is the Better Way to Do CFD?
 

John the problem is I believe Global and its also historical.The problem still emanates from the boffin syndrome of classical based science.

Indeed, the role of research and development has been historically much the same.

Management is too often commercially inspired with getting results and this often out features how the results were obtained.

Leave it to the capable CFD guy-that`s why we pay him strategy.The absence of CFD or high ranking Technical personnel in higher executive management positions also finds its way into governments.

I have just taken another look at a Joe Thompson 90`s review on CFD issues and it does marry well into your original question.

Thankfully I am retired and away from the skirmishes but the element of holding something back up your sleeve featured in my old company of pump makers where the chief Design Engineer carried so much in his head that was vital and he led commercial management and executives a merry dance-he could never be pinned down as no one knew enough to do so. That was during the very early years of CFD.

Keep pitching and help get over the basics and complexities to appreciative forum readers and questioners.

Re: What is the Better Way to Do CFD?
 

(1). If a PC system is a tool, then CFD is not. (2). This understanding is very important. I don't think, CFD will become a tool like a PC in the near future. (3). If we look at a level below the PC system, you will see the monitor, the keyboard, the mother board, the hard drive, etc... These can be called subsystems. (4). Another level below, you see the circuits, the chips, the resistors, the ic's, etc.... These are components. (5). Below this level, you have the design of IC's, CPU, etc... Now you see that, every piece must be working correctly. This is not the case in the current CFD. (6). For this reason, there is no way one can develop a general cfd code and make it a design or analysis tool. (7). If one has figured out how to solve a cavity flow with a moving lid, then all he can do is to show other engineers exactly how to repeat the same process and obtain the right solution. Without this step, even with the source code, there is no guarantee that another engineer will be able to repeat his result. (8). So far, the only practical way to obtain the right solution to a new problem is for a person who has the training in all phases of CFD, to try it out, from the begining of the geometry construaction to the converged solution. If he is lucky, then he will have the converged solution. (9). I don't want to say that cfd is reserved for the best expert who think that he knows how to solve a cfd problem. But in reality, even the best cfd expert does not know how to solve a simple problem,say a separated diffuser flow problem. (10). With the chemical reactions involved, suddenly it becomes a research project, which no one can claim that he knows how to solve it. (unless he has done exactly the same problem before and was able to obtain a good solution.)(11). The need to find the better ways to do CFD, comes from the fact that for a PC, the system is put together from different levels where the technology and implementation is validated and guaranteed. On the other hand, in most cases, no one really knows whether the solution actually exist, or how to obtain the solution systematically. (12). So, CFD is in a way, still look like the trade secret of an old fashion chief designer who developed his know-how by doing. (13). I personally do not think that a cfd code developed by someone else can be useful for another person successfully to solve other problems. (14). A validated cfd code can only be used for the original problem with a very narrow range of parametric variation.(due to strong non-linear nature of the governing equations). (15). Sure, if you are dealing with the linearized, inviscid, irrotational potential flow, or other simplified equations, then the likelyhood to obtain the good solution will be much higher. But then the solution will not be useful for competitive design.

Re: What is the Better Way to Do CFD?
 

The technology of a mass produced motor car centres on one model being produced ,a batch of motor cars with a different model could possibly be produced in the same processes of manufacture but with a proviso that changes were contemplated for inclusion.

The concept of mass production by Henry Ford was new and compared to our technology now, only the concept can be said to fully apply.

The PC has also historically developed and continues to evolve with BINARY[0,1-on/off] the operative tool.

CFD has similarly historically emerged with the emphasis on computational deriving from Numerical Solutions of Partial Differential Equations.Enter related sciences- mechanics of solids,liquids and gases,dynamics,heat,chemistry,electrical and magnetic phenomenum and our ability to apply mathematics giving us more analytical tools.

We can derive Navier Stokes equations in our concepts of understanding the laws of physics-energy,conservation,momentum etc.We can rationalise to finite-difference,finite volume and finite element. We can rapidly approach real situations to give us best possible results - usable,comparitive and also highly accurate in lots of cases.

CFD has tool like behaviour for specific uses but because it fails to handle all the generalisations we can throw at it then it is not yet a full tool.However, the science of CFD still strives toward that workable objective.

Therefore,better can mean more user friendly for existing and emerging packages.

Better can mean a wider area of activity and application.

Better can mean that reproducabilty does take place regardless of the person doing the CFD.

CFD is an ongoing process,maybe not a perfect process but what else is there to match it or even replace it.

Having come through-log tables,slide rule,desk mechanical operated calculators,room sized computers and magnetic tapes to punched ribbons-I think the PC can help CFD push both the learning and application barriers within reasonable expectancy.

CFD must therefore harness the full potential of the computer. I look forward to the day when a pump manufacturer stands up and confidently says like BOEING did on aircraft-CFD from start to finish of the project.

Re: What is the Better Way to Do CFD?
 

John,

Some points are:

Should you consider a general cfd code a single tool, or a collection a tools - a tool box?? I think the later is more appropriate, because invariably these codes are made up of many subcomponents which must work together at the same time (ie. source term modes, matrix solving) or in succession (ie. pre, solve, post-process).

One can also say this about a PC - a mouse is 'a tool' for moving a pointer, a 'printer' is a tool for producing outputs etc. etc., however until you have all these 'tools', ie. a PC, you can't do any word-processing. So depending upon what level of abstraction is most appropriate, we can consider a PC a single tool or a collection of tools - a toolbox. Same for CFD.

One difference, which highlights some of my earlier points, is that in a PC the subsystems are connected in a standardised manner. You can replace you video card, hard-drive and upgrade the CPU easily, and using components from various suppliers. You can't do this with CFD codes . . . so there is a long way to go for CFD development at the infrastructure/design/coding level.

Also, while we accept the idea of a 'generalised tool', we know their must be limits to the notion of 'general'. My PC can't run high-end database programs or enterprise systems. It doesn't have an uninterruptable power supply. But I don't need those. Other computers do.

On the other hand, I agree that there is still a lot of uncertainty re: is there a solution, repeatability etc., which mean CFD falls down compared a PC etc. as you mentioned.

Greg

Re: What is the Better Way to Do CFD?
 

(1). I had not worked for Boeing Aircraft company, so I don't know whether their CFD technology can be easily used in the pump design or not. (2). My feeling is it is unlikely that their codes can be used directly to design the water pump. This is because, the series of codes they are using are all linked to the aircraft configurations, and the flow is almost always external flows. At the transonic flow there will be shock/ boundary layer interactions, and there will be boundary layers on the wings and fuselage. (3). To feed this series of codes into the water pump design is not possible, based on my intuition. (4). We have to be practical, even though one can say that in principle, one can change the configuration, the boundary conditions and the flow properties in the codes and convert the codes to do the pump design. We are not talking about using the steam engine technology in the aircraft engine design, even though they are all related to the thermodynamics. (5). I also must say that one can not just convert the technology of the desktop PC into the laptop PC design and manufacturing. I am just trying to point out that, the future is there, but one has to be very specific in order to make it work. (6). I think, it is not possible for a pump designer to sit and wait for the CFD technology to reach his desktop. It is not going to happen this way. He is not going to wait until the rocket designer has the time to convert his flow over a missile cfd codes into a pump design codes. Even if the rocket designer had the time to do so because of the reduction in force, I don't think the rocket designer will have the knowledge about how to design the water pump.(except for those very few who are constantly working in various fields of fluid mechanics) (7). The invention of the transistor had changed the car radio technology right away, but for the computer controlled fuel injection in a car, it was still the business of the auto-industries. (8). In other words, each industry will have to do its own CFD planning. (9). As I have said before, the CFD technology developed twenty years ago is still being used today. So, it is not the technology issue, but more a planning and implementation issue. (if you are thinking about using the codes developed from the aircraft industries in the water pump design, it would be absurd to think th other way around. That is, take the codes in water pump design for the aircraft design.) And if the old fashioned chief designer is not telling every his trade secret, then there is no way the cfd engineer can put the codes properly for him to use. That's the generation gap in technology. So, the cfd engineer will have to learn his own trade, and do his own research in the water pump design, then if he is still there, then there is a possibility of inventing the water pump cfd codes for his own use.

Re: What is the Better Way to Do CFD?
 

(1). I think, you are right in thinking. (2). I tend to think that this is similar to the begining of the industrial revolution. That is, the steam engine was just invented, and people are thinking about using it in every possible ways. (3). You can also think it the other way around, that is, this is just before the invention of the airplane. You have the modules such as the bicycle technology, the propeller technology and the internal combustion engine technology, but you are waiting for the aircraft technology. (4). If what you are saying is true (it will be somedays), then in principle, you can put together (a big assumption) these geometry and mesh generation pre-processors, turbulence models, solvers, and post-processors, and make your own CFD codes to solve some of your problems. (5). Naturally, the test will be the accuracy of the solution from the codes. But , in general, we are not there. So, I think, something is missing, and more need to be done. (6). So, what are missing? from which module? Or, we are simply waiting for someone like Wright brother to appear?

many are using code, very few understand it.
 

(1). The consequence of this non-CFD management situation is: many are running CFD code, but very few understand it. (2). Based on my experience in turbomachinery industries, most codes are developed by outsider and then modified or improved in-house to handle the mesh generation or post-processing. Some are derived from the company merger. And the rest are from the commercial cfd codes. (3). Unless one has the access to the original code and has the knowledge to understand and modify the code, the rest of the activities are simply wasted. (4). In other words, the whole organization is now involved and controlled by this unknown code. This is the most important discovery of the industrial secret so far. I must say that the problem is wide-spread. (5). The persons who brought in the codes will naturally present the good part of the code validation (especially on the pressure distribution for the simple blade passage flow problems). The idea was to get more research fund to continue on the modifications. Unfortunately, somewhere along the line, the code must also be released to the designer or engineer for use in the real world problem. There, the designer and engineer were forced to use it blindly, as long as the code is producing some kind of solution. (6). If the designer is not the code developer, he is not going to worry about the solution accuracy or validation at all. His job is to push the computed results out of his project schedule or milestone. Even though on some occasions, there will be testing done, the puzzle will be on the shoulder of the manager to handle. And in all cases, he is going to cover it up from the rest of the management, because no one in the organization can solve the puzzle for him in the CFD related technology. Even if there is this source code, to just understand the code, it will take an expert about six months to read through the listing, if he has the document about the subroutines. In most cases, even the cfd expert will give up the task of reading the source code. (7). This failure of in-house modified commercial codes or codes developed by some other organization, will then push the real work outside the organization. This becomes the contract work carried out by the subcontractors. In this way, the responsibility can be shifted to someone outside the organization. (8). The last phase of operation is the outsourcing of the cfd work. This is routinely done by the vendor of the commercial CFD codes for the code users, when the user is not able to get any meaningful results. The last step in this line is: outsource the cfd work to another consulting organization, this could be another cfd code vendor. In this way, the responsibility is completely outside the organization. (9). In this cfd life cycle, the real problem was not solved, but the designer, engineer, and the management escaped the responsibility. (10). The real problem is: all of these so-called cfd codes, are at one time, developed by someone who was under pressure to write the code to meet the deadline (believe me that it takes a long time to write a cfd code). There was not enough time to do the validation on the methods or turbulence models. Unfortunately, these codes got pushed into the production line in order to show the progress. I must say that, in most cases, even the original code developer didn't know whether his code is any good at all for the unknown problem. (11). Some of these codes were generated in the universities and were given free to other organization and industries. These codes are just the research products of a professor and his several generation of students. There is no way of making sure that the code will solve the industries' problem, so it is free. (or at very low cost) The codes can also come from the government lab for the same reason, and normally does not come with good documentation. This is obvious, for the protection of the original author. (12). I think, I have given you enough reasons why the current CFD activities is not working. And if you still think that an aircraft giant like Boeing can easily use its current cfd codes to design new generation of aircrafts, then, I would say that you are really too naive to jump into the conclusion.(they have the potential, but not until it is demonstrated) The same applies to other industries. (it maybe adequate to design the old engine and aircraft, but definitely not so for any new configurations or designs.) (13). I am not talking about the reacting flow at all. There, everything is research, from my point of view. (14). Recently I heard that a Mach ten spacecraft is going to be flight tested soon. The original version of Mach 25 space plane was promissed back in 80's. The news comment says that even if the test is successful, it will likely take another 25 years to make it a reality. That tells you why the reacting flow cfd in high Mach number flow is still many many years away. (14). AT the same time, you are free to run codes day and night. Who cares? That is another reason why I am looking for a better way to do CFD.