Scaled residuals vs Mass flow rate

Gobbi · February 23, 2019, 15:54

Hello,

While simulating steady flow through a 3D model of a vein I've been struggling to get convergence. It's my first time doing these kind of simulations in Fluent. While monitoring the scaled residuals I left the default tolerance at 0.001, yet, they roughly get close to 1e-2 and then show an asymptotic behavior. Even playing with the boundary conditions, refining the mesh, nothing seems to help. The continuity residuals in particular were the highest.

But turns out that when I check the mass flow between the inlet and outlet, the difference between them is in the order of the 1e-6 (which is pretty good). I thought that the residuals of continuity not decreasing meant that the mass flow in and the mass flow out were different. I thought these quantities were related that way, but apparently I'm wrong.

I know this may be a beginner's question, but I'm still wondering: why are the residuals not decreasing and thus the simulation not converging?

LuckyTran · February 23, 2019, 20:09

1) Imbalances are calculated at each cell, averaged over the entire domain, and then scaled. You can have global mass balance and still have one or two cells with terrible imbalance that keeps the residuals high, but also vice versa. You can have low reported residuals and still not satisfy the global balance.
2) Scaled residuals are scaled by the worst residual during the first 5 iterations (the default is 5, but you can set this to any number). If the worst residual encountered in this period is low, you get large scaled residuals. If the worst residual encountered in this period is stupidly high, you get small scaled residuals. Continuity is always scaled no matter what. Momentum, energy, and so on do not (and by default they are not scaled this way).

Gobbi · February 24, 2019, 19:09

Thanks for your response LuckyTran!

I have a better understanding now. So for particular cases like this one, since the residuals are likely to be having this not very convenient behavior due to imbalances in some cells, the best approach is to refine the mesh in the most critic areas (there're areas with vortexes, for example) in order to reduce the imbalances in those cells?

LuckyTran · February 25, 2019, 12:04

How do you know there is any unbalance?

With scaled residuals, if you reset the residuals and run it again (and if your calculation is already converged) the scaled residuals will be stuck at 1. That doesn't mean there isn't any imbalance, but looking at the scaled residual output doesn't really tell you this information.

Gobbi · February 25, 2019, 12:47

Here I attach an image of the scaled residuals and the mass flow balance between the inlet and outlet.

The imbalance in the mass flow is within the tolerance for what I'm doing, yet the solution is not converging (I'm using Fluent's default convergence settings), and hasn't converged for any configuration despite I've changed the boundary conditions, done some refinements in the mesh, even changed the model just in case the flow started to become turbulent in some areas that I hadn't considered.

I know that in the image the simulation ran only for about 250 iterations, but in other tries it got close to 2000 and the behavior of the residuals was the same.

LuckyTran · February 25, 2019, 15:51

Residuals are not a measure of convergence

Gobbi · February 25, 2019, 16:53

While investigating more about it found this Convergence old entry in the forum which covers the subject in a very thorough way, clears many doubts and gives very good advice on the topic.

Regardless, thank you for your help!

February 23, 2019, 15:54	Scaled residuals vs Mass flow rate	#1
Gobbi New Member Stefano Gobbi Join Date: Jan 2019 Posts: 5 Rep Power: 7	Hello, While simulating steady flow through a 3D model of a vein I've been struggling to get convergence. It's my first time doing these kind of simulations in Fluent. While monitoring the scaled residuals I left the default tolerance at 0.001, yet, they roughly get close to 1e-2 and then show an asymptotic behavior. Even playing with the boundary conditions, refining the mesh, nothing seems to help. The continuity residuals in particular were the highest. But turns out that when I check the mass flow between the inlet and outlet, the difference between them is in the order of the 1e-6 (which is pretty good). I thought that the residuals of continuity not decreasing meant that the mass flow in and the mass flow out were different. I thought these quantities were related that way, but apparently I'm wrong. I know this may be a beginner's question, but I'm still wondering: why are the residuals not decreasing and thus the simulation not converging?

February 23, 2019, 20:09		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	1) Imbalances are calculated at each cell, averaged over the entire domain, and then scaled. You can have global mass balance and still have one or two cells with terrible imbalance that keeps the residuals high, but also vice versa. You can have low reported residuals and still not satisfy the global balance. 2) Scaled residuals are scaled by the worst residual during the first 5 iterations (the default is 5, but you can set this to any number). If the worst residual encountered in this period is low, you get large scaled residuals. If the worst residual encountered in this period is stupidly high, you get small scaled residuals. Continuity is always scaled no matter what. Momentum, energy, and so on do not (and by default they are not scaled this way). Gobbi likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
boundary condition with pressure AND mass flow rate	tsi07	FLUENT	3	April 14, 2022 12:32
Pressure Outlet Targeted Mass Flow Rate	LuckyTran	FLUENT	1	November 23, 2016 10:40
Calculating mass flow rate at multiphase flows	Kuslo187	OpenFOAM Post-Processing	1	August 21, 2015 18:11
mass flow rate in DPM steady tracking	Roule	FLUENT	4	June 1, 2015 10:44
Mass flow rate	sepidecent	CFX	0	August 9, 2011 00:15

February 24, 2019, 19:09		#3
Gobbi New Member Stefano Gobbi Join Date: Jan 2019 Posts: 5 Rep Power: 7	Thanks for your response LuckyTran! I have a better understanding now. So for particular cases like this one, since the residuals are likely to be having this not very convenient behavior due to imbalances in some cells, the best approach is to refine the mesh in the most critic areas (there're areas with vortexes, for example) in order to reduce the imbalances in those cells?

February 25, 2019, 12:04		#4
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	How do you know there is any unbalance? With scaled residuals, if you reset the residuals and run it again (and if your calculation is already converged) the scaled residuals will be stuck at 1. That doesn't mean there isn't any imbalance, but looking at the scaled residual output doesn't really tell you this information.

February 25, 2019, 15:51		#6
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	Residuals are not a measure of convergence

February 25, 2019, 16:53		#7
Gobbi New Member Stefano Gobbi Join Date: Jan 2019 Posts: 5 Rep Power: 7	While investigating more about it found this Convergence old entry in the forum which covers the subject in a very thorough way, clears many doubts and gives very good advice on the topic. Regardless, thank you for your help!