CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Benchmarking FLOWORKS99 with experimental datas (

Anis June 15, 2000 10:54

Benchmarking FLOWORKS99 with experimental datas
Hi everyone.

I'm actually testing a commercial CFD code (FLOWORKS99). Finite Volume Method, rectangular grid AUTOmeshing.

I'm comparing measured mass flows through engine pipes, with those computed using FLOWORKS99.

The solver used allows two computationnal alternatives : - the incompresible one, - the compressible one.

One knows that each alternative has a specific set of computationnal algorithms : one cannot use a compressible algorithm to compute a incompressible flow.

The accuracies of the computed mass flows are not satisfying with both alternatives.


Is the fact that FLOWORKS99 allows only one of the compressible or incompresible approaches to compute a flow, while the computed field shows a mixed subsonic-transsonic-supersonic velocity field (incompressible approach) abd a totaly subsonic field (compressible approach), THE source of error or is the sofware SIMPLY unable to compute such mixed compressible-incompressible flows?


Raza Mirza June 15, 2000 12:38

Re: Benchmarking FLOWORKS99 with experimental datas
"One knows that each alternative has a specific set of computationnal algorithms : one cannot use a compressible algorithm to compute a incompressible flow."

I do not think the above statement is correct. It is true that codes do have different algorithms, but typically one code (such as FLOWORKS) would have one tupe of scheme to handle both types of flows. Common schemes are variants of SIMPLE scheme of Dr. Patankar, the so-called pressure-based schemes, and some so-called density based schemes. Both have advantages over the others in certain classes of flows.

So what if the flow is compressible in some regions and incopressible in the others. Well, one GLOBAL scheme will be used to solve for compressible flow (if the user considers the compressiblity effects to be important).

As far as the results are concerned, there are many factors. The most important is the user himself who has the responsibilty of knowing the problem and the tool. But you can check one thing: try refining your mesh (I am assuming that the AUTO meshing feature does allow at least this degree of control) and see if that brings your solution any closer to your experiments. If it does, then your solutions are not grid independent and you have to try to find grid-independence of some level first. Automatic meshing, and solution-adaptive grid refinement still need human interferece, and the user has the final responsibility.

Anis MAALEM June 22, 2000 04:19

After several simualtions, I actually handeled FLOWORKS99 in the appropriat manner.


Thanks for all.

John C. Chien June 22, 2000 19:47

Quite Good=Accurate Results ?
(1). When do we say that the benchmark test results are "Accurate results, Quite Good"? (2). It only rasies more questions, I think.

All times are GMT -4. The time now is 18:36.