Does anyone have an information / experience on the use of CFD in concurrent engineering - i.e bringing CFD into the design loop with CAD/CAM?

I am currently researching an article on this.

Regards, Dave

Yes but I am not sure much of it is printable! The gap between what appears to be supported in the large expensive commercial packages and what is usable by engineers is quite wide. The problems seem to come from all directions.

Real designers tend to construct models to solve the task of design. (This assumes you can persuade/coerce them to use parametric solid models in the first place!). Fair enough but the sets of parameters/associations are rarely in a useful form for subsequent analysis and the model needs reworking. The problems of loading a designer with constraints that have no benefits for the task they solving are obvious.

Even if the engineer constructs the model themselves for the purposes of analysis subsequent modifications are not necessarily going to line up with what they guessed they might be when they choose the set of parameters/associations.

Some of the tasks specific to analysis are not well supported (if at all) by the CAD packages. Constructing fluid volumes from solid models and, particularly, automating the filtering of detail to avoid defeating the grid generator (also a failing of most grid generators which cannot work with small unresolved surfaces).

Passing parametric solid models between different proprietary packages almost always results in the loss of the parameters/associations. Neutral files like IGES and STEP fail more often than one might expect. One is often forced to develop an interface for a specific package using their particular access routines.

...

Perhaps I am being a bit too negative but the irritation has arisen from experience albeit from a few years ago now.

If you have not done so already, I would strongly recommend you approach companies like Rolls-Royce and British Aerospace who have worked with CFD and CAD/CAM for a long time. I am sure they can supply some positive real examples.

Hi,

I have done some test work on parametric CFD meshing. For example, I have a case setup of a car that you can globally change the roof profile (by extreme amounts) and regnerate automatically a hex mesh. It takes about 20 seconds to make each new hex mesh. It has never been used in the `real world' but if you are interseted in more information then send me an email.

regards

David

Concurrent Engineering is what MDICE delivers. MDICE or the Multi-Disciplinary Computing Environment from CFD Research corporation ( http://www.cfdrc.com )provides an environment in which several engineering analysis programs run concurrently and cooperatively to perform a multi-disciplinary design, analysis, or optimization problem.

Using MDICE, engineers are able to couple inherently dissimilar disciplines and programs from a variety of sources, performing distinct tasks such as geometry modeling, grid generation, CFD and structural analysis, and post processing into a single software system.

A large number of commercial and public domain codes are already integrated into the MDICE environment.* With the MDICE developer's toolkit, your application can become MDICE compliant as well!

If you want to learn more about MDICE, please visit our website, http://www.cfdrc.com. You can also email me directly at rmm@cfdrc.com.

(1).CFD has been linked to the design loop for over 20 years. This is especially true for the aerospace and engine industries. And the CAD/CAM integration has been shown practical among different computers in network for ten years.(2). At the begining, the in-house CAD design programs were used to provide the geometry for CFD analysis. This approach is still being used today because the geometry associated with the product is very specific to the industries involved,for example, the wing, fuselage, inlet, nozzle,turbine blade etc. There is no particular need to use the general purpose CAD. (3). The in-house CAD or geometry generation program also provides specific CFD analysis code geometry input data (formatted files). And over the years, this practice has become sort of the standard in the design loop within a company. It is hard to change it with modern 3-D, photo-realistic CAD. So, the interface is, in most cases, still not fully electronic. It requires human interaction (It could be bad because of human errors in data transfer or modification. It also could be good because of extra steps to check the possible errors. Fully automatic does not mean that it is fully error free. A poor design could be easily spotted by human eyes and stopped for possible design revision. This is especially the case for new engineers without much experience. A fully automatic system could produce a lot of useless work before the error is identified.) (4). So, the use of modern 3-D CAD is in most cases for the creation of plastic models or for the casting purposes, that is model verification. (5). For analysis of non-standard items, such as scroll, volute, seal, secondary flow, integrated aircraft configuration, etc.., a general purpose CAD is used ( whether it is a stand alone program or a module provided by CFD vendor ). The geometry is then exported to a CFD geometry/mesh generation program for the creation of CFD compatible geometry model and nodal distribution. The final mesh generation can go through another step (through another code) depending on the final need of the mesh and the code capability. So, there are many CFD mesh codes on the market today ( they are being used in industry ). (5). The graphic post-processing is not a critical problem today, because there are stand alone code or standard from government and private companies. And most popular commercial codes have integrated graphics modules. (6). But in general, the CFD in design loop is still very labor intensive. This is mainly because the design process by definition is iterative, while the current CFD is at most one-way. The integration of potential flow solvers, Euler solvers in the design loop maybe in a much better shape than the so-called Navier-Stokes solvers, but, what we're really after is the solutions to the Navier-Stokes equations. At that level, the integration is rather primitive, including the output from the Navier-Stokes codes ,not mentioning the possible second round of geometry change. (7). So, the technology is there, the data transfer mechanism is there, but, it requires a lot of human attention. This is not because of the lack of technology, but because of the lack of design knowledge. ( Before you can use the solution of Navier-Stokes equations in the design, you need to know how to interpret the results,which includes the mesh size effect naturally. At least you need to know the difference between the absolute total pressure and the relative total pressure, before you can analyze the turbine CFD results)

Flotherm by Flomerics has strong integration capacities to both MCAD and EDA environments. Our MCAD integration tool (FLO/MCAD) has the ability to import IGES, STL and SAT formats. It uses the ACIS solid modelling kernal from Spatial Technology. The success rate of import is therefore higher for SAT compared to IGES but not as bad as some have made out on this thread.

Flotherm users are already (and have been for some time) using FLO/MCAD in conjunction with Flotherm to perform conjugate heat transfer analyses of electronic systems at all stages of the design process. Most efficient thermal designs are ones that have been thermally optimised at the intial design phase. To this end FLO/MCAD also has the ability to take geometry from Flotherm and export it out as IGES or SAT for import back into the MCAD tool.

This area of integration is of paramount importance to Flomerics. The vision is that mechanical engineers should be able to perform thermal analysis at any stage of the design process with as little pain as possible. Tight integration of CFD tools into current design environments has, and always will be a primary goal for commercial CFD vendors.

Robin Bornoff

Senior Consultant Engineer

Flomerics Ltd.

www.flomerics.com