Solver Control

cfdengineer · June 20, 2011, 11:55

I have read in other threads about Upwind and high resolution. From what I have read it looks like if you select high resolution CFX with try to solve 2nd order but if there is some numerical instabilites it will use blending to get convergence. I find this misleading since when you check the box one would expect a purely 2nd order solution or attempt at a 2nd order solution. I see where there is an option for a blending scheme and it looks like if you want to enforce CFX to run a purely 2nd order(high resolution) run that you should use the blending option and set the number to 1. Why would CFX give a high resolution option when your results may or may not end up 2nd order? I find this confusing. In your results how do you know if your results are 2nd-order accurate if you use the high resolution option?

stumpy · June 20, 2011, 12:26

Why would you expect "High Resolution" to mean purely 2nd order accurate? To me, this is just a name given to a scheme that presumably has good resolution of gradients. It doesn't imply any particular order of accuracy, 1st, 2nd or 3rd etc. Remember that 2nd order accurate is a mathematical description and does not always mean physically more accurate results. In the end you are best trying the High Res versus the pure 2nd order scheme and see which works best for your case. The pure 2nd order scheme may not remain bounded. In my experience the High Res scheme gives gradient resolution very close to that of a fully 2nd order scheme but remains bounded.

cfdengineer · June 20, 2011, 12:57

I guess I was confused in the sense that high resolution was 2nd-order when it isn't. I now see it means that CFX will try to make the results as high order as possible. Basically I was looking for the option where I want CFX to enforce a 2nd order solution and the way to do that is set the blending to 1.0. Basically the customer I am doing work for was asking about this and they were looking for a 2nd order accurate solution.

Another question, When starting a solution is it a best practice to set the advection and turbulence to high resolution rather than starting at Upwind and then switching to high resolution?

stumpy · June 20, 2011, 17:02

I would just start with whatever you need for your final results. There's usually no need to start from Upwind.

June 20, 2011, 11:55	Solver Control	#1
cfdengineer New Member Stephen Join Date: Mar 2010 Posts: 21 Rep Power: 16	I have read in other threads about Upwind and high resolution. From what I have read it looks like if you select high resolution CFX with try to solve 2nd order but if there is some numerical instabilites it will use blending to get convergence. I find this misleading since when you check the box one would expect a purely 2nd order solution or attempt at a 2nd order solution. I see where there is an option for a blending scheme and it looks like if you want to enforce CFX to run a purely 2nd order(high resolution) run that you should use the blending option and set the number to 1. Why would CFX give a high resolution option when your results may or may not end up 2nd order? I find this confusing. In your results how do you know if your results are 2nd-order accurate if you use the high resolution option?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Creating New Solver: For particle-laden compressible jets	sankarv	OpenFOAM Running, Solving & CFD	17	December 3, 2014 20:41
Creating New Solver: For particle-laden compressible jets	sankarv	OpenFOAM	0	April 4, 2010 19:06
Control function in Elliptic solver	Hoang Anh Duong	Main CFD Forum	4	July 9, 2007 04:32
maintaining a logarithmic velocity distribution	Morten Andersen	CFX	1	January 8, 2007 12:37
SST in 2D?	Jesper Sørensen	CFX	16	December 23, 2006 09:40

June 20, 2011, 12:26		#2
stumpy Senior Member Join Date: Apr 2009 Posts: 531 Rep Power: 21	Why would you expect "High Resolution" to mean purely 2nd order accurate? To me, this is just a name given to a scheme that presumably has good resolution of gradients. It doesn't imply any particular order of accuracy, 1st, 2nd or 3rd etc. Remember that 2nd order accurate is a mathematical description and does not always mean physically more accurate results. In the end you are best trying the High Res versus the pure 2nd order scheme and see which works best for your case. The pure 2nd order scheme may not remain bounded. In my experience the High Res scheme gives gradient resolution very close to that of a fully 2nd order scheme but remains bounded.

June 20, 2011, 12:57		#3
cfdengineer New Member Stephen Join Date: Mar 2010 Posts: 21 Rep Power: 16	I guess I was confused in the sense that high resolution was 2nd-order when it isn't. I now see it means that CFX will try to make the results as high order as possible. Basically I was looking for the option where I want CFX to enforce a 2nd order solution and the way to do that is set the blending to 1.0. Basically the customer I am doing work for was asking about this and they were looking for a 2nd order accurate solution. Another question, When starting a solution is it a best practice to set the advection and turbulence to high resolution rather than starting at Upwind and then switching to high resolution?

June 20, 2011, 17:02		#4
stumpy Senior Member Join Date: Apr 2009 Posts: 531 Rep Power: 21	I would just start with whatever you need for your final results. There's usually no need to start from Upwind.