I computed one squarepipe model,the whole model consiste of two partsne part is a small squarepipe,the other part is a big squarepipe,they are coupled by an arbitrary match. when I check the result(casename.info),I found some warning messages:there are negative turbulent kinetic energy in some cells,and those cells lie in the connective surface of the two parts. what is the reason that cause the problem? Whether the problem will influence the final results?

I am very grateful for your kind help about that!

Best Regards, Lee

If the cells are arbitrary coupled, then there can still be regions where the coupling fails at the edge where the small square duct interfaces with the big duct. Especially if you are not using partial arbitrary couples. Then the actual area of the interface is not equal with the cross sectional area of the small duct. And i suspect that starCD views a couple of the cells that are not coupled as walls in the middle of nowhere and tries to apply wall treatment which results to negative k. If you make a couple of section views at different angles and choose to view only the surface then you should see the regions where coupling fails as walls. Will it influence the results. Sure, if you have a smaller cross sectional area at the arbitrary interface than you should have, the mean velocity through the arbitrary coupled cross section should increase. If the blocked area is a big percentage of the interface cross sectional area, then the results will be affected.