I am using RSM model to calculate the flow velocities at different cross sections of a pipe. The pipe has an inner diameter of 0.7m, and it has 2 out-of-plane bends. It is about 35m long. Re~3*10^7. I have about
2,500,000 cells throughout the pipe volume. Anybody has a sugestion about how to do an error analysis for this model? I appreciate everyone's advice. Thanks!
Structural modeling is my thing. But these general ideas should apply to what you're doing.
Double or half the mesh size to see if the mesh density is enough.
Try building a simpler model that's related to what you're doing for comparison to test data, hand calculations or something.
Jeff, thanks for your suggestions. I cannot do a hand calculation because it is a 3-d turbulent flow, and the friction loss from the bends is unknown. However, I did tried different mesh sizes: 0.035,0.03,0.0275, 0.025. The results didn't change much.
As you said I can double or half the mesh size: If I reduce the size to 0.0125 and 0.006125, and let's say their results are the same. Then I compare the result of the 0.025 mesh to the 0.0125mesh result and find 0.005% error for the 0.025mesh. Does it mean my mesh is fine enough if I only want the error to be within
0.01%? If I want to know the error of the result for each mesh size, should I just compare the results of the other mesh sizes with the 0.0125mesh result?
I was thinking of a model with the different size pipe and no bends or a model of 1 bend with the a constant diameter. That'd still require test results or something to compare to, which you may not have.
With modeling, a change of less than 1% for a mesh doubling/halfing the mesh is great. I'd go with one of the middle mesh sizes provided that does not make the model difficult to run and work with.
PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.