In your experience which is the most user friendly commercial CFD code in the market?

(1).NONE. (2).Convergence alone will make you sick. (3). In most cases,it is difficult to generate the right mesh, not to mention the creation of the geometry. (4). It is ideal for research work, where the schdule, the results are not critical. (5). For design purposes and mission critical, write your own code. Otherwise you are spending way too much time in getting an answer over and over again. (6). The special purpose codes may perform better than general purpose codes. (7). If user-friendly is defined as the use of GUI, then the above comments do not apply.

Hi Waquar,

I would define the appropriate parameters for "user-friendy" as:

1. The ability to set-up and modify solution parameters such as boundary conditions and block-offs, fluid properties, physical models (turbulence, combustion, multi-phase), discretization parameters, initial and restart conditions, etc

2. The ability to solve the problem as specified above. This is where the issue of convergence as John had mentioned comes in. The issues here include the type of linear solver ie. is it a multi-grid or a single grid solver and how robust is it. Is it a coupled or a segregated algorithm: the coupled solvers require more RAM for a given problem but will, in general, take less iterations.

3. Does the solution capture what you require to solve your problem. This is now an issue of accuracy of the discrete solution on a grid that you can afford. Also of concern is whether the physical models capture the essence of the physics in your problem.

So a really quick set-up with a cool GUI that leads to a solution that does not converge or very poor accuracy on your grid is in the end not productive.

So contact some of the vendors and talk about these issues and the types of problems that you are trying to solve.

I have no idea what John is talking about when he says that you should write your own code for design problems. The testing of new algorithms is the domain of research problems where a research code is written specifically to analyze, test, and understand specific properties and behaviors. Often, this is done on simplified test problems and the codes are therefore not useful for industry problems. Writing a 3D flow solver including all of the flexibility to solve real design problems is not practical for most design offices.

Regards.............................Duane

My definition of "user friendly" is a little bit different:

Most of the older versions (1-2 years ago) general-purpose CFD codes already has the ability to set-up and solve a problem with acceptable solutions. However, the primitive GUI is limiting the effective usage of the functions of the program and that's not user-friendly, even though it has all the capabilities of CFD.

"user-friendly" is more to the users who have not used the code before. Imagine a person never has used computers in his life, but he can now quickly take on Windows and write up his articles using MS-Word. The learning curve defines the "user-friendly".

A "user-friendly" CFD code should allow the user a easiest, most comfortable and yet the shortest time to master the code and apply it to engineering problem quickly. The convergence and accuracy of the code only serves as a side dish for the user-friendly-design of the code.

In-house CFD codes are not purposed for general usage and might not be user-friendly but they sure will generate the best results for the specific problem.

(1). This issue is very important. It is especially important in packaging and marketing the products. (2). Take for example the CAD program, it used to be the "line command" approach, and there was no mouse to use. (3). For experienced users, this is a faster approach. But for the new comers, it is easier to use windows and icons approach. (4). The windows approach tends to divide the task into many more steps of mouse clicks, and thus slow down the operation. (5). Beyond this user interface stage, it going to take a lot of thinking to make the program user-friendly. (6). A good user-friendly designed program will guide you from the begining to the end without much effort from the user. It will guide the user to the next logical step and ask only a couple of questions. And this can be done either using the line command approach or the windows approach. (7). Currently, no such cfd program exist to meet this requirement. Most simply provide the options in terms of the menu items or options. It is like providing a 21 speed standard shift on a passenger car. In this case, the user-friendly solution is a 4-speed automatic transmission. The users don't even have to set the automatic transmission at all. And it achieves the same goal. (8). So, making several turbulence models available, or several algorithms available is not to be considered as user-friendly feature. (9). There are many areas where a cfd program can become more user-friendly. One such area is the mesh generation. In old days, most of in-house cfd codes could automatically generate the mesh. (for the special purpose applications) (10). For general purpose cfd codes, this kind of user-friendly features are hard to implement. So, the users end up spending a lot of time doing the tedious work of manipulating the geometry, mesh, the relaxation factors, etc.... (11). The best one can say is that modern commercial cfd codes have much more built-in features and options included. But, this does not mean that it is easier for the user to get the final correct answer.

(1). You would be surprised to know that many companies are using the commercial cfd codes to replace the testing, and also are using these codes for product design and analysis. (2). In many cases, the users are not even trained to use these codes properly. They learned how to use the codes by reading the user's guides. Or at most received a week training. (3). I am talking about turbomachinery field in general.

Hi Kai,

You said: "The convergence and accuracy of the code only serves as a side dish for the user-friendly-design of the code."

Wow I am glad that you said it!

At the end of the day, a buyer who accepts crappy results but got it really fast or worse, it never converged to a result is either:

1) looking for pretty pictures and should not really be using CFD.

or 2) unhappy and will soon be in tough shape for business or atleast their reputation.

It is this type of an attitude that has given a lot of people a bad taste for CFD. The real aim should be giving a reasonable solution to the physical problem.

Regards..............Duane

Thank you all for the interesting replies. What I am really after is a CFD code that can be used for undergraduate projects/class room instruction/ CFD lab work as well as by graduate students. Correct results are a must but so is ease of use, if a student has to search the online help and jump from one place to another to find out about a topic - that would be too bad. A user friendly code means easy to use , well documented , easy to set up the problem etc.Robustness,accuracy and convergence are also a must but they do not neccesarily make a code difficult to use. waqar

Hi you all,

I work with STAR-CD (3.102) and CFX (5.3). I must say that the user interface of both codes are very nice. If you know anything about CFD it is easy to set up a problem. Mostly grid generation is somewhat of a problem. We use ICEM tetra (which gave us a lot of problems and requires an expert user) and the internal grid generator of CFX which is quite easy to handle but doesn't have the overall capabilities of ICEM. The grid generator of CFX is sufficient in our case (transsonic internal flows). Convergence is always a problem. In our case 90% of the problem is in creating a good initialisation of the flow field. After that convergence is not a problem. The latest solver of CFX (coupled multigrid) is very fast and saves a lot of time in our design process. However it still laks the physical models (user coding) STAR has. STAR-CD will have a coupled solver in the near future. If that happends you have a userfriendly code with a LOT of capabilities and which is fast. Example: in our case we run a transsonic flow problem on a 2.4 million tet/prism grid with boundary layers and all with CFX in a day on a SUN (4 processors 2Gb RAM). This makes it fast enough for our purpose. I wrote my own 3D Euler code when I did my PhD but could not afford to spent time on a GUI. This is a big problem with in-house developed codes. You can use it yourself but nobody else can.

If you are looking for codes for use in classroom/execises. I would recommend STAR-CD. The user interface is really intiutive and you can build your own structured grid within the code, which should be fine with students. Do not start with unstructured grids as they require more cells to get the same results and are much more difficult to get a converged solution on. Biggest problem is to get the budget for a couple of licenses (I think STAR is cheaper than CFX)

(1). To maintain the math standard, keep the electronic calculators out of the classroom. (2). The same is true for cfd, keep the black box cfd codes away from the classroom. (3). After you have learned the cfd from your professors, you are always free to use any cfd codes.

Thanks for your comments.

I believe in a CFD company, the accuracy and robustness of a CFD code depends on the engineers/mathematician who's developing and implementing the numerical algorithms; the user-friendly design is part of the work for those software engineers where their job is to make the code easy to use, and runs smoothly on all kinds of platforms. The back-core and the frond end of the program are completely two different things.

If you ever had a chance to look at the newly-released Airpak from Fluent, you will probably know what is HVAC-"user-friendly" and what is non-HVAC-"user-not-friendly".

agree with both Bart and JCC:

A commercial CFD code only helps to solve the problem but not to help you gaining knowledge on the CFD. I wrote my own SIMPLE(R) program and FEA code before picking up those commerical CFDs/FEAs, and felt it quite straightforward in working with those program even if it is not user-friendly (just spending more time in learning the input format).

A good start with the structured mesh, talking about North, South, East, West, Top, and Bottom

As the eltronic calculators are allready in the classroom, why not let a commercial CFD package in? I noticed that the developers at the CFD software companies are people with the same background as the 'expert users' They mostly have a PhD in fluid mechanics and so on. So why the distrust? (I'm NOT working for a CFD software developer, by the way, but I don't think they deliver more junk than anybody writing their own code would)

(1). There is nothing against the commercial cfd codes. (2). I developed large scale cfd codes, user-friendly codes. I used someone else's codes. I am using several commercial cfd codes listed here. (3). The only purpose of going to a school is for training from teachers. (4). Sure, you can read books with answers attached. You can also read cfd source codes. But there is no substitute for the formal cfd courses and training. (5). Yes, you can use the commercial cfd codes any time, when you are taking the formal cfd courses. But the commercial cfd codes are no substitute for the formal cfd courses. (6). I can give you a real example. An engineer-A learned how to use a commercial code-C and used it for a year. He got some answers. He then started using the commercial code-T because the company decided to change the code from code-C to code-T for political reasons. The engineer-A think that code-T is better than code-C, so he tried to promote it. After six months, the engineer-A started working for division-C, and was asked to use the commercial code-T. He worked very hard to learn the other commercial code-I in order to get the model created. After several months, the model was done, but the commercial code-T could not give him the converged solutions. He decided to apply another job in other division. At the same time, he started complaining about the commercial code-T, which was his favorate at the begining. In the end, he used another commercial code-F, and was able to get the converged solution and finished the assignment. After that, he move on to work for another group, where he again started promoting this commercial code-F. He then worked on another problem using this commercial code-F, he was able to get a converged solution, but the solution was not accurate enough to provide the useful information to guide the design. At the same time, in he old department, someone was able to use the commercial code-T and commercial code-I to get useful and converged solution for the same problem. (7). I am trying to write the story in a long fashion. The point I am trying to make is that a well-trained cfd researcher or engineer with a solid foundation in cfd "CAN OBTAIN GOOD SOLUTIONS BY USING MOST ANY COMMERCIAL CFD CODES". On the other hand, a person with a couple of years experience running commercial cfd codes without formal cfd training at school, will likely end up in the same situation as the engineer-A in this real story. (and if you are the vendor of the cfd codes, you definitely would like to have your codes used by the real professionals, because they can appreciate the effort put in the codes. On the other hands, the non-professionals will complain about the code if they are not getting the right solutions they wanted.)

Good point. I see the same thing. I notice that an experienced CFD user will try strategies not mentioned in the manual of a commercial CFD code.

Hi, My name is Phan Cong Thuong, My age is 24, My nationality is Vietnamese, I am an Aeronautical Engineer of The Technical University HCMC, VietNam. I am researching on The multi-phase flows. You are a member of CFD forum. I have just happened to see your name in the web of CFD Forum. Could you help me to look for the document about The multi-phase flow!!!! In my country, so far nobody research about this domain, so it is lack of the information about this. Please, help me!!! Thank you very much!!!! BYE, THUONG. My email: thuongartist@dme.hcmut.edu.vn My home add: 182 Hoang Dieu 4th District, Ho Chi Minh City, Viet Nam.

(1). You can visit the Internet bookstore such as amason.com. It is quite useful looking into this website, because it not only give you reviews by the readers, but also tell you what other books the reader are buying at the same time. (2). From there, you will get a whole picture about what other readers are reading in the same field. It saves you a lot of time to find out which book is worth reading. (3). From the book, you can narrow down your area of interest, check the references in the book, get copies of the paper published in the technical journals (there are more than one journal in multi-phase flows). (4). This should get you going, and you need additional help, the best place is the librarian in your library.

If you're looking for educational software it's probably best to get a free/share -ware. from your email address I'm guessing that you are Myanmar if so you'd be looking for a cost effective approach I'm from a "third world" country too and I know it can often be difficult to access high tech products. I'd recommend the foll webpage which has links to free and low cost software some of which can be directly downloaded www.capella.colorado.edu/~laney follow the software link. i guess most of the codes don't have fancy GUIs etc but i've used a code downloaded from their and have been able to use it pretty well. the one i got is NSC2KE which is a 2D NS/Euler code which along with the BAMG and EMC2 mesh generator forms a complete system. all of these codes are linked on this webpage and originally come from INRIA in France. the CFD code is written F77 which is a disadvantage since for most students/young engineers (like me) it is ancient. but maybe you make it an exercise to convert it to F90 or C. EMC2 is a GUI based 2D CAD/mesh gen software and BAMG is a mesh generator with some gui abilities, it's best feature is that it can be integrated w/ the CFD code to give an adaptive mesh capability.

there are lots of other codes available, many are 3-D research codes from NASA etc many of which are parallel codes whose availability to non-US citizen is limited but they'd be too advanced, hard to use and/ or resource intensive for your purposes anyway. there are enough nice 2-d codes for you to pick from and surprisingly quite a few grid gen and post processing (less of these) codes to choose from.

John this true, but at the same time I think Waqar wants to supplement his students formal CFD coursework with hands on use of a CFD code.

I think it's best to work with a prewritten code available as source code rather than a commercial code because students can go in and for example see how a runge kutta scheme is implemented in code and at the same time run the code and get results without spending a lot of time coding.

many cfd courses at universities started in the 80's when cfd was new and very few codes were out there. so much of the work in class was to write code (afer all the students would have to write a code as part of their research anyway). today though, most schools will have even a 2-d research code written by grad student available for students to use.

one should never underestimate the power of that first "pretty picture" to motivate students (even if it's wrong!). when I took CFD in grad school I had three semesters of it: one in basic algorithms applied to linear equations, one for transonic flows using time iterative methods and one for turbulence modelling. during that time i only once got the opportunity to use a 2-d code but it was good becuase i got to see the actual effect of artificial dissipation on shocks and see BL separation etc.

one of the drawback of fluid mechanics is that it isn't easy to see what the fluid is doing in say a wind tunnel but CFD allows us to visualize the flow and sight is our most visceral sense. John Anderson made this point in his CFD book. students leaning CFD must be able to run code and see their own solutions and frankly nowadays thers just too much theory to cover in class to spare time to write a full on code