y+
 

In what range should the y+ value be? Is [30;300] ok? What should I do if it's too small or too big?

Thanks everybody, Roberto

Re: y+
 

Between 30 and 100 is a good tight range for standard k-epsilon with wall functions.

You can go up to 500 depending on what you are trying to simulate and what information you are looking to extract.

You will have to adjust the thickness of the near wall cell layer (bigger or smaller) to acheive the y+ values you desire.

I have not used the latest version of STAR so I do not know if there is any grid adaption available. In the past, there was not any. Therefore, you would have to manually alter the thickness of the near wall layer. If the grid is structured, you can use combinations of Vcextrude, Vgen, and Vfill depending on what your topology looks like. For complicated models it is not a pleasant experience.

Ron

Re: y+
 

If I get really high values (>1000) or really small ones (almost=0), is my model worth anything from the physical point of view?

If they are too high, should I use larger near wall cells? Or should I use smaller cells?

If I change near wall cell dimensions, do I change both the highest and the lowest y+ values?

Thanks for your interest, Roberto

Re: y+
 

I think you want to modify these values if you are looking at extracting any kind of wall info, like heat transfer coefficients.

If y+ is too big, you need smaller cells.

If y+ is too small, you need bigger cells.

Either you can try to move the grid around to change these cell sizes or you can use the Cdivide, Crefine commands to refine the cells. I also believe there is a Coarsen command. Cdivide and Coarsen will only work on structured hexahedral grids.

Ron

Re: y+
 

So if I'm not really interested in near wall data, but I want to calculate the mass flow or the flow momentum around the axis of the duct, a [30;500] range will be fine?? Is it ok if the y+ value is in that range except for very small zones where it is lower or higher?

Thanks a lot, Roberto

Re: y+
 

Roberto,

I think if the bulk of your y+ values are within the 30-500 range, and you are not looking to accurately solve for heat transfer in the duct, you should be fine.

Good luck.

Ron

Re: y+
 

(1). CFD is "scientific" although your goal is engineering. (2). To be scientific, it must be repeatable. And to be engineering, it must be within certain range acceptable to you. (3). So, the suggestion is: (a). Set your engineering acceptable range, (b). run several cases, using different meshes, and check the results. This will tell you the mesh sensitivity to your problem. (c). run a low Reynolds number model case to establish a separate reference point. Make sure that the Y+ is <=1.0 for the low Re model case. (d). check all of your results, and make your own decision based on your acceptable engineering limits. (4). You must realize that most commercial CFD codes are packages of functions, subroutines, and libraries, which can not solve your problem automatically. It is "you" who is putting together the problem and is responsible for the solution of your problem. So, it is your responsibility to set the acceptable limits of the results. (it is not easy to do because these subroutines are invisible inside the black box. And it is also your responsibility to validate the results of interest to you. The commercial CFD code is like Home Depot store where you can get the tools and materials you need for a home project. )

Re: y+
 

Ron

I suppose one ought to be a wee bit careful even with respect to the calculated mass flow and related quantities since the wall friction also depends upon y+. Or do you think the resultant mass flow is much less sensitive to y+ than the underlying wall friction?

Lars Ola

Re: y+
 

Lars,

Point well taken.

To John's point, make several variations of the grid and check the sensitivities of the quantities you are interested in.

Although, some times the old rule of thumb must prevail, if you want an answer in a "less than glacial" time frame.

Regards to all.

Ron