[CFD grass]

I have a question. I'm doing an CFX transient analysis; lobe pump

I made the 3 domains([Lobe1], [Lobe2], [BODY]), and also boundary conditions(inlet, opening, wall).
Lobe1, 2 contains rotating information(ex. angular velocity, axis definition etc.).

I think there is no problem, but I got an error message : In analysis 'Flow Analysis 1' - Domain 'BODY': transient analyses require that initial conditions are specified unless an initial valuew file is specified at run-time.

so, I clicked the domain:BODY>Initialization and check the "Domin Initialization". I set the initial condition, but the error message exist yet!

Is there any missing condition? or does it need any extra setting in "Immersed solid"

cf. I setted that the lobe1,2 condition is immersed solid

Pleas help me

[Gert-Jan]

Normally, you start a transient calculation from a steady state simulation. So, the steady state solution is your initial guess.

If you have immersed solids, you normally start with zero velocity. So what you should set in Pre are values for velocity, pressure, temperature, and turbulent variables. Better use 'Values. than 'Automatic with Value'.

Since you already set this, I don't know where it goes wrong. There are two places where you can set initial guess. I would use the Global initialization instead of the domain initialization, espcially with Immersed solids,These solids should not require anything. Did you try this?

[CFD grass]

Thanks for answering. I understand what you're saying.

I resolved my problem with many trials. Finally I can solve it.
However, I got a new curiosity.
I could solve this problem with setting sequence which has different with the previous way.

Is there any difference with
(1) First of all sequence, changing the Analysis type(steady state → transient), and proceed and make the other setting.
(2) Except changing the analysis type(steady state → transient), proceed other settings and the last step, makes type transient.

In real, with changing the order that set the analysis type, I could solve my problemI

cf. Every setting is same with each other but the order of changing "analysis type"

[Gert-Jan]

In principle, a converged steady state flow solution should not depend on the intial guess. But since you are running transient I'm a little lost. So, I do not understand what you are saying or what you are doing.

[ghorrocks]

There are many cases when the final converged steady state result depends on the initial conditions. This is called bifurcation and fluid flows are good examples of this. Examples include jets running along walls (which either stick to the wall due to the Coanda effect or not depending on the initial condition of the jet).

You are correct in that most flows I have seen have only a single converged solution. But there are many cases where there are multiple converged solutions.

[Gert-Jan]

For a Lobe pump, this does not hold.
For flows where jets either stick on the wall or not, due to a Coanda effect, I consider this as proof for a transient problems, poor converged, incorrect BC, ill posed problem, etc.
Or even better: it is a poor design which should be avoided in real industrial life. Better design something that always give the same result. Then you know what you get.....

[ghorrocks]

For flows where jets either stick on the wall or not, due to a Coanda effect, I consider this as proof for a transient problems, poor converged, incorrect BC, ill posed problem, etc.

Transient: No, bifurcations can be steady state.
Poorly converged: No, bifurcations can be fully converged.
Wrong BC: No, bifurcations can have the correct BC applied.
Ill posed: Nope.

Bifurcations are real and they do happen. It is an example of a metastable state - it is locally fully stable and real, and to kick it out of that state and into another equally valid state you need to apply a reasonably large external forcing function.

I have personally seen it happen in hydraulic oil flows in modelling of power steering valves. The valves were not poorly designed - with the constraints of what is required to make it function as a power steering valve there are configurations where the exit oil jet sticks to one or other of the walls depending on initial conditions. Here is a recent publication on it: https://www.sciencedirect.com/scienc...42727X19304084

You can get much more complex bifurcations when things like non-Newtonian fluids, multiphase flows and other complex physics exists.

But OK I agree that for a lobe pump this is probably not applicable