CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Multiphase flow - simulation crashing (

yschati November 14, 2010 14:06

Multiphase flow - simulation crashing
I am doing a multiphase flow simulation, with water and air (main phase) as the phases.
I have a duct. Air enters the duct (I have specified total temperature and total pressure as the inlet condition, setting air volume fraction = 1 and water volume fraction = 0). Inside the duct, I inject water through small holes (typically 2 mm in diameter). At the holes I give an inlet boundary condition of a lower total temperature but higher total pressure (water volume fraction = 1, air volume fraction = 0). Basically I want the water to cool the air. At the outlet, I have specified bulk mass flow rate. At the remaining surfaces I have the adiabatic wall (no slip for air, free slip for water droplets) conditions and the rotating periodic conditions. I am using k-epsilon turbulence model, buoyant flow, etc.
The problem is that the simulation is crashing. Either it gives a notice of very high Mach number or sometimes it just crashes without displaying any error. Microsoft gives an error that solver.hpmpi has crashed. I have tried changing the mesh, boundary conditions, turbulence models, but to no avail.
Can someone please advise what to do and what can be the possible reasons of failure?

ghorrocks November 14, 2010 17:57

Can you give a drawing and flow rates to describe what you are doing?

yschati November 14, 2010 22:19

1 Attachment(s)
I have attached a pic of the geometry and BC.
The boundary conditions are as follows:

Type: inlet
Total pressure: 237.38 kPa
Total temp.: 456.8 K
Air vol. frac. = 1
Water vol. frac. = 0

Type: inlet
Total pressure: 262 kPa
Total temp.: 370 K
Air vol. frac. = 0
Water vol. frac. = 1

Type: outlet
Bulk mass flow rate = 0.338 kg/s

Other BC: rotating periodic on the sides (as seen), remaining surfaces: adiabatic wall (no slip for air, free slip for water droplets).

The timescale is set to auto.
I have done this simulation earlier when instead of water, the holes were injecting cold air. It converged then and gave results.
I hope this helps.


AliTr November 15, 2010 03:31

did you try it without water spray?
if it works that way then start with low pressure and use an expresion to increase the water pressure by time.

another point, you have a fixed mass flow at outlet which should be equal to the total mass of water and air, are you sure about that? maybe using fixed mass flow at inlet and an open boundary at outlet also helps to get better convergence.

yschati November 15, 2010 04:08

It works without the water spray. The problem is that I cannot reduce the water pressure too low. It cannot go below the pressure of air surrounding the holes or else the water will not be injected (liquid can go from higher to lower pressure only).
As regards the outlet, I have also tried putting static pressure conditions at the outlet but to no avail. That too crashes.
I'll try this opening condition though. However isn't specifying a mass flow rate at inlet and opening at outlet a less robust condition? I thought a pressure condition makes the simulation robust.

Thanks for the help and please keep helping and advising. It's kinda urgent.

AliTr November 15, 2010 04:16

if it works without water spray, then it is likely that the initial condition is responsible for crashes. did you try to run it without water spray and then use the results to initilise the main model? if this doesn't work, try to use low pressure and increase it gradualy.

yschati November 15, 2010 04:21

I am using the same initializing conditions (on velocity) as I have used for the simulation without the water spray.

AliTr November 15, 2010 04:24

then use the result of converged simple model (no spray) for intilising the complex model.

yschati November 15, 2010 06:29

Is the error coming because of only wrong initialization? Or can there be some other reason? Why I ask this is because shouldn't a good steady state result be independent of the initialization conditions?
Also currently I'm giving only the air Cartesian velocity as the initialization condition. Rest all conditions (pressure, temperature, water velocities, volume fractions) are set to automatic.
Please help. I'm stuck :(

hosseinhgf November 15, 2010 17:37

when i use smaller physical timestep in my pre the mach problem solved.(0.0001)
when mach number became large(6.47) solver display much is it?

AliTr November 15, 2010 18:26


Originally Posted by yschati (Post 283472)
Is the error coming because of only wrong initialization?

In a multiphase model you need proper initial condition. If you don't have an existing model result to use as initial condition, still you need to avoid automatic setup and provide your best approximate of pressure and velocity.

yschati November 16, 2010 06:38

I tried giving the results of the basic model as initial conditions.
What I did was that, in the solver manager, I gave the results file of the converged simulation without the water spray and then ran the simulation.
However, it is again crashing, due to very high Mach numbers.

My flow is not at all supersonic.

yschati November 16, 2010 09:49

I tried the following things. All failed.
a. Initialization using the results without water spray.
b. Reducing the physical timestep by 2 orders of magnitude (1e-8).
c. Specifying opening (and giving static temperature and pressure conditions) in place of inlet conditions.
d. Reducing the total pressure at which water is being injected (262 to 250 kPa).

What to do now?????

ghorrocks November 16, 2010 17:35

Work through the hints listed here: and let us know how you go.

hosseinhgf November 16, 2010 18:29

change your air if it is ideal gas.
before solver crash stop it and see where mach is high.

All times are GMT -4. The time now is 02:59.