[Guille1811]

Hi, im doing a standard simulation of an axial fan. The simulation results are less than 5% different from the technical specifications of the fan (power, max outlet velocity and average outlet velocity) and the relevant variables of the system (velocity, absolute pressure and fan power) they all converged perfectly to a 100% stable value.

That being said, the simulation imbalances didn´t converge to zero at all (see image attached, final imbalances on the right side).

My question is: can i overlook this imbalances situation due to the fact that the actual results are accurate enough for my needs? If yes, then why it can be overlooked?

Thanks in advance!

[Gert-Jan]

In general, you cannot neglect this. Also, mostly the imbalances go to zero quite quickly.

On th other hand, the imbalance is relative to the flow coming into the domain (if I'm correct). So, if you don't have an inlet, then the behaviour you observe is normal.
Then, the imbalance is relative to the interface flow, which is very small in your case. That could also explain the behavior your observe.

[ghorrocks]

Global imbalances are just that - for mass, it is the sum total of all mass entering or leaving boundaries, and any creation or destruction of mass (such as source terms), and the accumulation/loss of mass (transient only). If you have a large mass imbalance it means that these do not balance out so some mass is being gained or lost somewhere. Same goes for momentum, heat and all the other variables.

The imbalance % is calculated by the global imbalance divided by some number which indicates how much fluid is going through the domain - it might be the flux at the largest boundary (need to check that). You have to be a bit careful about imbalances in simulations which only have one boundary such as flows surrounded by ambient air which commonly only have a single opening boundary to the surrounding atmosphere. As the imbalance calculation in this case becomes the boundary flow divided by the absolute value of the boundary flow the imbalance will flip between 100% and -100% and never converge to zero, even for a converged solution. Therefore the imbalance is not always an appropriate measure of convergence, and that is why you are able to turn it off.

In these cases have a look at the magnitude of the imbalance compared to the magnitude of the total mass in the domain. You will have to calculate this manually. It will give you an idea of whether the magnitude of the mass gain/loss is significant relative to the total mass.

[Guille1811]

Well the imbalance is in the order of magnitude of 1E-6 kg and the total mass of that domain is around 1,5 kg. That being said, is a good enough argument to say that this imbalance situation is not significant due to corresponding to such a tiny percentage of the domain mass?

[ghorrocks]

That is problem dependent - for some purposes that is unacceptable and for some they can get away with it. You need to assess in your case.

If you can't work out directly what imbalance you are happy with then do a sensitivity analysis. Do a simulation with imbalances 10x tighter and 10x looser and see if it makes a difference big enough that you care about it.