how cfd results can be trusted in a very complex geometry
 

In my work we perform CFD analysis for heavy duty gas turbine enclosure ventilation. CFD gives the flow field in the enclosure and we do the design work according to those results.

The geometry of the computational domain is very complex, I cannot upload a pic due to the company policy, linked picture is from internet which can give you an idea, enclosure is not shown, but you can see the huge turbine, plenty of pipings, tubes...

http://www.zjzszx.cn/imagematerial/u...8WLFOQUR2S.jpg

The computation covers heat transfer inside the enclosure(temperature of turbine and pipes surface are high), and the heat transfer through the GT enclosure to outside atmosphere(considering the thickness of enclosure insulation). There are 1-2 inches hot pipes, and huge turbine casing in the model, i don't even know how to calculate the Reynold Number. Since there are small geometries and the domain is quite large, the computational grid number is >20Million, grid independence validation cannot be done.. and of course, it's steady state simulaiton..

We spend more than a month for each project, but the problem is: I do not have any confidence about the results I get.

any one have any experience on this? how cfd results can be trusted in a very complex geometry like this? is there any literature talk about this, any experiment?

Colorful pictures is not what we want, is it?

i am not a industry guy but in a academic area. I heard that at the CTR(center for turbulence research) which is the most famous center in turbulence, they performed gas turbine simulation with LES at the combustor and RANS at the compressor and turbine. But people in turbine society do not usually trust that results. I think that is very hard to predict combustion and heat trasfer in turbine regarding flow phenomena and boundary condition.. but flow at compressor is quite reasonable. This is just my thingking. good luck.