Converged variable with changing residuals query

em17 · November 10, 2017, 09:36

Hi all,

I am wondering why the residuals change between iterations despite a variable having reached convergence.

In the example below (and from my understanding - correct me if I'm wrong), at Time 199, the solution for Ux has stopped as shown by 'No Iterations 0'. What I am struggling to understand is why the residual for Ux at time 199 and at time 120 are different? If Ux is meeting its residual target and the solution has stopped, why would the residual change between iterations? Does this mean that the value for Ux is still changing with each iteration?

Thanks in advance for your help!

Time = 119

smoothSolver: Solving for Ux, Initial residual = 7.53813e-07, Final residual = 7.53813e-07, No Iterations 0
smoothSolver: Solving for Uy, Initial residual = 9.21586e-07, Final residual = 9.21586e-07, No Iterations 0
smoothSolver: Solving for Uz, Initial residual = 0.0386961, Final residual = 0.00146457, No Iterations 2
GAMG: Solving for p, Initial residual = 7.01485e-07, Final residual = 7.01485e-07, No Iterations 0
time step continuity errors : sum local = 1.22429e-06, global = 2.23222e-07, cumulative = -0.00030492
ExecutionTime = 2.89 s ClockTime = 5 s

Time = 120

smoothSolver: Solving for Ux, Initial residual = 7.43496e-07, Final residual = 7.43496e-07, No Iterations 0
smoothSolver: Solving for Uy, Initial residual = 9.02652e-07, Final residual = 9.02652e-07, No Iterations 0
smoothSolver: Solving for Uz, Initial residual = 0.0164174, Final residual = 0.00109311, No Iterations 2
GAMG: Solving for p, Initial residual = 7.81812e-07, Final residual = 7.81812e-07, No Iterations 0
time step continuity errors : sum local = 1.36448e-06, global = 2.39555e-07, cumulative = -0.000304681
ExecutionTime = 2.91 s ClockTime = 5 s

Zeppo · July 28, 2019, 09:23

The reason why your Ux residual isn't steady is that the whole system of NS-equations is non-linear, i.e. the coefficients of Ux equation are dependent on the Uy Uz and p values. So while Uy Uz and p are updated Ux equation is updated too. In matrix form A Ux = b, A and b are changing throughout the iterations for Uy Uz and p, even though there is no iterations for Ux. And the residual r = b - A Ux is changing too. Wait a bit more till Uz gets steady and residual for Ux will get steady as well.

November 10, 2017, 09:36	Converged variable with changing residuals query	#1
em17 New Member Emily Join Date: Nov 2017 Posts: 24 Rep Power: 8	Hi all, I am wondering why the residuals change between iterations despite a variable having reached convergence. In the example below (and from my understanding - correct me if I'm wrong), at Time 199, the solution for Ux has stopped as shown by 'No Iterations 0'. What I am struggling to understand is why the residual for Ux at time 199 and at time 120 are different? If Ux is meeting its residual target and the solution has stopped, why would the residual change between iterations? Does this mean that the value for Ux is still changing with each iteration? Thanks in advance for your help! Time = 119 smoothSolver: Solving for Ux, Initial residual = 7.53813e-07, Final residual = 7.53813e-07, No Iterations 0 smoothSolver: Solving for Uy, Initial residual = 9.21586e-07, Final residual = 9.21586e-07, No Iterations 0 smoothSolver: Solving for Uz, Initial residual = 0.0386961, Final residual = 0.00146457, No Iterations 2 GAMG: Solving for p, Initial residual = 7.01485e-07, Final residual = 7.01485e-07, No Iterations 0 time step continuity errors : sum local = 1.22429e-06, global = 2.23222e-07, cumulative = -0.00030492 ExecutionTime = 2.89 s ClockTime = 5 s Time = 120 smoothSolver: Solving for Ux, Initial residual = 7.43496e-07, Final residual = 7.43496e-07, No Iterations 0 smoothSolver: Solving for Uy, Initial residual = 9.02652e-07, Final residual = 9.02652e-07, No Iterations 0 smoothSolver: Solving for Uz, Initial residual = 0.0164174, Final residual = 0.00109311, No Iterations 2 GAMG: Solving for p, Initial residual = 7.81812e-07, Final residual = 7.81812e-07, No Iterations 0 time step continuity errors : sum local = 1.36448e-06, global = 2.39555e-07, cumulative = -0.000304681 ExecutionTime = 2.91 s ClockTime = 5 s

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
changing a variable (turbulence dissipation rate) formolation	m.beh	Fluent UDF and Scheme Programming	6	November 7, 2016 05:53
changing scheme variable with UDF?	Wagemakers	FLUENT	17	September 30, 2015 14:34
Changing the unknown(target) variable in convection;From WallTemperature to HTC!	fshak92	STAR-CCM+	1	April 28, 2012 08:14
Question about making a variable function of proprities of onother variable ?	rafiktharwat	CFX	0	March 26, 2011 14:23
Accuracy and convergence	pawan1989	FLUENT	5	April 15, 2010 17:17

July 28, 2019, 09:23		#2
Zeppo Senior Member Sergei Join Date: Dec 2009 Posts: 261 Rep Power: 21	The reason why your Ux residual isn't steady is that the whole system of NS-equations is non-linear, i.e. the coefficients of Ux equation are dependent on the Uy Uz and p values. So while Uy Uz and p are updated Ux equation is updated too. In matrix form A Ux = b, A and b are changing throughout the iterations for Uy Uz and p, even though there is no iterations for Ux. And the residual r = b - A Ux is changing too. Wait a bit more till Uz gets steady and residual for Ux will get steady as well.