Learning Forum

[Sergio92]

I am new to ANSYS, and I wanted to use ANSYS FLUENT to simulate water flowing past a sphere. To decrease time of simulation, I am using 2D axisymmetric mesh. The mesh, together with the boundary conditions I used for this problem, are shown below.

 

When I run my simulation, the residual errors for continuity and velocity are 0 (zero) in the first 5 iterations. Only after that I get non-zero residuals (3.26e-17 for continuity, 1.43e-4 for x velocity and 7.39e-5 for yvelocity in iteration 6). The residuals decrease afterwards, and the results seem correct if I let the simulation running for a sufficiently large no. of iterations.

Does anyone know what is causing the issue in the first 5 iterations? It is a bit of an inconvenience to me because I cannot set a convergence criterion under Residual monitors (if I do so, the simulation ends after 1 iteration saying the program has converged).

 

A bit of additional information that may or may not be useful:

• I am using ANSYS student 2020 R1


• I am using the Workbench as my framework


• All algorithms (SIMPLE, SIMPLEC, PISO, Coupled) yield this same behavior

[RK]

Hello, 

Have you checked the "Normalize" option in the residual monitors?

Regards,

Rahul 

[Sergio92]

Hi,

 

No, I had not checked the "Normalize" option. But the behavior persists even if I do so.

[DrAmine]

At the beginning there is nothing so zero: appears when starting natural convection case and it also depends on initial condition

[Sergio92]

Hello Amine,

 

Thank you for your reply. But I am afraid I do not understand it entirely. Shouldn't the velocity be able to "diffuse" from the inlet starting from the first iteration? Also, shouldn't the pressure gradient (the source term in the momentum balance eqs) be non-zero at least in the second (and subsequent) iterations?

 

Some extra info:

• I am using Viscous (Laminar) model


• The issue persists regardless of how I initialize my problem (I have tried using Hybrid Initialization, Standard Initialization from inlet and Standard Initialization setting all zones to 0)

[DrAmine]

Again is not an issue at all if you understand how this terms are calculated and it appears in several cases where flow settlement requires some steps.

[Sergio92]

It IS an issue regardless since the user can no longer set a convergence criterion (program thinks the simulation has converged after 1 iteration). Also, I am not quite clear on what do you mean by "flow settlement".

I may not be a CFD expert, but I do understand how the SIMPLE algorithm works. The velocity profile we get after solving the momentum balance equation (MBE) in general does not obey the continuity equation (CE). So the residual for continuity should not be 0.

Because mass balance is not satisfied, we calculate a pressure correction term and a velocity correction term. The new velocities now "obey the CE" but do not obey the MBE, so we re-compute them by solving the MBE with the new pressure field. And so on...

 

Residuals being 0 from iteration 1 means that the pressure profile remains unchanged. If that is the case, none of the terms change in the system of equations corresponding to the MBE.

The only (bizarre) explanation for having residuals equal to 0 would be that, because of the relaxation factor used to compute the velocity, the velocity profile "takes time" to violate mass balance (for some reason it satisfies CE in the beginning, but as the under-relaxation velocities converge to the final solution a discontinuity arises that changes the pressure profile). But changing the under-relaxation factors does not influence on this behavior. I have decreased both pressure and velocity relaxation to 0.1, and my first 5 iterations still have residual equal to 0, with the 6th iteration having residuals different from 0.

 

 

I understand this may be a trivial topic to more experienced students/researchers, but that is not my case. I would be really greatful if you could provide a few more details in your answer, or share a thread/book chapter/article from where I could get a more detailed answer.

[chitresh29]

Hi Sergio92,nI am facing the same issue while solving a problem in fluent (Ansys). Were you able to solve the problem that you mentioned? Please suggest. n

[Rob]

What boundary conditions are set? A zero residual can occur if nothing happens, ie zero flow over a boundary means there is no flow so the change is zero etc. It typically sorts itself out after a few iterations. Turning the monitors checking off can bypass the problem but it's best to understand what's going on too. n

[chitresh29]

Hi Rob,nI am simulating unsteady flow (Re <1) in a 0.3 mm dia. pipe. The flow is driven by a pressure-gradient in x-direction. The inlet-outlet boundary conditions are set to periodic. The walls are set to no-slip boundary condition.nThe first 5 iterations are exact zero. The absolute convergence was set to 1e-5. The convergence criteria fails as the first iteration is zero. Please suggest how can I solve this problem.n

[Rob]

I'm pretty sure there is an ignore first some iterations for the parametric models, so have a look at those. The option I tend to favour is to not check for convergence, I monitor it (ie watch the residuals) but don't check it:set Convergence Criterion to none in the Advanced section of the Residuals panel. n