Y+ and convergence

PremSagar · February 14, 2017, 09:04

Hey!

Im trying to solve for a compressible flow boundary layer phenomena. The Y+ average reaches from ~30 to 2 in 2000 iterations exponentially. But then it's taking forever to go to one. It's decreasing for sure but in the order of 0.0001 per iteration. The boundary layer is not resolved. How can I make it reach below 1 faster enough to resolve the boundary layer?

Any suggestions would be appreciated

LuckyTran · February 14, 2017, 15:42

The y+ is a property of your grid and flow solution. Sitting around waiting does not make it go down. Either change your grid or change your flow.

It starts at some value because of your initial condition. It changes with each iteration because your solver is iterating the solution field, updating the solution with each iteration. If your solution has converged, then your y+ will be a fixed value. To resolve the boundary layer you need to make sure there are enough grid points in that region of your flow field, not wait for the solver to "solve for the boundary layer".

PremSagar · February 14, 2017, 23:01

I agree. I used this online calculator for the first grid spacing. https://geolab.larc.nasa.gov/APPS/YPlus/
But it never seems to go below 1 even with 10 times smaller grid spacing I got from this calculator. :/ I was under the impression probably it's the solver taking more iterations to resolve each grid layer close to to wall. Correct me if I'm wrong.

Thank you

LuckyTran · February 14, 2017, 23:32

Read the readme's!

"This page provides a worksheet for estimating the normal spacing to be used in a viscous CFD grid. Estimates are for a turbulent flat plate in free air and are based on the Sutherland formula for viscosity. "

"For a given Reynolds number, Mach number and desired y+ value, this page will estimate the spacing normal to a solid surface required to yield 1 grid point in the laminar sublayer."

The tool is not a y+ calculator, it's a tool to help you estimate or guess the spacing you need to get a y+. The calculation is actually very simple to do yourself. Nowadays digital calculator are available everywhere (the page was last updated in 1997). The crux of the tool is that you need a way to estimate the boundary layer thickness.

The y+ you get from your simulation is your best guess at the y+. Even when you make the grid sizes smaller, you may end up with the y+ because y+ depends on the flow, it is not something you know beforehand. When you change the grid, you may change the flow just enough so that you end up getting something different than what you would expect if the flow did not change.

Also don't become obsessed with getting a y+ of 1. A y+ less than 5 is already in the viscous sublayer. Having X number of grid points within the boundary layer is much more important than getting a particular value of y+ even if you end up with a y+ of 1.2 or 2 or 3 or 3.14159. What you are trying to resolve usually is the gradient next to the wall (the slope of velocity, temperature, etc. to satisfy the impenetrable boundary condition), you're not trying to just fit a cell right next to the wall.

PremSagar · February 15, 2017, 01:42

Thank you!

I was only concerned because I used the area average of y+ as the monitor. With the average at about 2, there were regions where the y+ was about 5-6. I guess I'll see my boundary layer profile and call the shots.

February 14, 2017, 09:04	Y+ and convergence	#1
PremSagar New Member Prem Sagar Join Date: Dec 2016 Posts: 7 Rep Power: 10	Hey! Im trying to solve for a compressible flow boundary layer phenomena. The Y+ average reaches from ~30 to 2 in 2000 iterations exponentially. But then it's taking forever to go to one. It's decreasing for sure but in the order of 0.0001 per iteration. The boundary layer is not resolved. How can I make it reach below 1 faster enough to resolve the boundary layer? Any suggestions would be appreciated

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence	Centurion2011	FLUENT	48	June 15, 2022 00:29
In the case of convergence	aja1345	FLUENT	1	July 31, 2015 04:58
CFX overwhelming Fluent in mass convergence of boundary layer separation case	Pierre1	FLUENT	7	March 26, 2015 22:43
Force can not converge	colopolo	CFX	13	October 4, 2011 23:03
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 02:17

February 14, 2017, 15:42		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,761 Rep Power: 66	The y+ is a property of your grid and flow solution. Sitting around waiting does not make it go down. Either change your grid or change your flow. It starts at some value because of your initial condition. It changes with each iteration because your solver is iterating the solution field, updating the solution with each iteration. If your solution has converged, then your y+ will be a fixed value. To resolve the boundary layer you need to make sure there are enough grid points in that region of your flow field, not wait for the solver to "solve for the boundary layer".

February 14, 2017, 23:01		#3
PremSagar New Member Prem Sagar Join Date: Dec 2016 Posts: 7 Rep Power: 10	I agree. I used this online calculator for the first grid spacing. https://geolab.larc.nasa.gov/APPS/YPlus/ But it never seems to go below 1 even with 10 times smaller grid spacing I got from this calculator. :/ I was under the impression probably it's the solver taking more iterations to resolve each grid layer close to to wall. Correct me if I'm wrong. Thank you

February 14, 2017, 23:32		#4
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,761 Rep Power: 66	Read the readme's! "This page provides a worksheet for estimating the normal spacing to be used in a viscous CFD grid. Estimates are for a turbulent flat plate in free air and are based on the Sutherland formula for viscosity. " "For a given Reynolds number, Mach number and desired y+ value, this page will estimate the spacing normal to a solid surface required to yield 1 grid point in the laminar sublayer." The tool is not a y+ calculator, it's a tool to help you estimate or guess the spacing you need to get a y+. The calculation is actually very simple to do yourself. Nowadays digital calculator are available everywhere (the page was last updated in 1997). The crux of the tool is that you need a way to estimate the boundary layer thickness. The y+ you get from your simulation is your best guess at the y+. Even when you make the grid sizes smaller, you may end up with the y+ because y+ depends on the flow, it is not something you know beforehand. When you change the grid, you may change the flow just enough so that you end up getting something different than what you would expect if the flow did not change. Also don't become obsessed with getting a y+ of 1. A y+ less than 5 is already in the viscous sublayer. Having X number of grid points within the boundary layer is much more important than getting a particular value of y+ even if you end up with a y+ of 1.2 or 2 or 3 or 3.14159. What you are trying to resolve usually is the gradient next to the wall (the slope of velocity, temperature, etc. to satisfy the impenetrable boundary condition), you're not trying to just fit a cell right next to the wall.

February 15, 2017, 01:42		#5
PremSagar New Member Prem Sagar Join Date: Dec 2016 Posts: 7 Rep Power: 10	Thank you! I was only concerned because I used the area average of y+ as the monitor. With the average at about 2, there were regions where the y+ was about 5-6. I guess I'll see my boundary layer profile and call the shots.