CFD modelling of a water distribution system
 

Hello

I am trying to simulate pipe flow through a water distribution system. The objective is to obtain the pressure head and flows at each outlet.

Please guide me for the right approach for the simulation.

1. The geometry is 100 m in length. The diameter of the main pipe is 250 mm. It branches to 190 mm and 160 mm at 50 m length and branches to 100 mm and 50 mm at 100 m length. (z=0)

2. The working fluid is water at 25-degree Celcius and its properties are used. The density is 998.2 kg/m3 and viscosity is 0.001003 kg/(m.sec).

3. The condition at the inlet is the pressure head of 50 m and 6000 LPM (litre/min) and I am not sure about the condition at the outlet of the system. Also, I have been told velocity at inlet is 2 m/s. ( When the Reynolds number is calculated, the value is 5x10^5 !!!) . Is this expected?

I have the following questions.

1. What should be the correct boundary condition? Should I give pressure head at the inlet and atmospheric condition at outlets? Or mass flow at the inlet and atmospheric condition at the outlet?

2. I am using the k-e model for the simulation. What should be the wall condition? No-slip condition? Is there a y+ value requirement to use no-slip condition?

If there is any recommendation while carrying out the simulation. Please suggest.

Thank you

1) Both should work, the one to be chosen tipically depends from what you know at inlet. If you know both pressure and mass flow, I suggest using mass flow at inlet.

2) Unless you know otherwise, wall is always no slip, independently from the turbulence model and y+ value; this should be all handled by the code, internally (i.e., Fluent does it).

3) The Re number seems correct, at least at the inlet (1000 * 0.25 * 2/0.001 = 5e5)

However, besides this, and the specific choice of the turbulence model (not sure if I would bet on std k-e), let me highlight the fact that you are embarking on a pipe simulation with length L = O(400D), or probably more, where D is the diameter. It's not necessarily going to be that huge as it sounds, but you are probably going to need O(1e6) cells. If I had to do this for work, I would carefully check if a simpler 1D method would be a better fit for me or my client.

Thanks for your reply. I have set up my simulation. I used a no-slip wall condition for the wall.

Can you please explain the following results to me.

1. When I used mass flow inlet at 250 mm diameter pipe and atmospheric condition at the outlets, the simulation converges well. The mass flow is conserved as is as given by m1v1 = m2v2 + m3v3 + m4v4. But the pressure head at the outlets is quite low. For example, at inlet shows around 45 m pressure head and 0.1 m head at the outlets.

2. When I use a pressure head of 50 m head at inlet and mass flow rate at the outlets, the simulation converges well. The mass flow rate is conserved. Also, I get an expected pressure head at outlets around 30 m head at both the outlets.

3. When I used pressure head of 50 m head at the inlet and atmospheric pressure at the outlet. The simulation has difficulty to converge. ( why pressure head at the inlet and atmospheric condition at outlets is difficult to converge? Is this not the right boundary condition for my simulation?

I can't say without, at least, knowing which code you are using.

What I don't inderstand is if you have one or multiple outlets and one or multiple inlets. For example, with one inlet and multiple outlet you can't specify mass flow at outlet because you wouldn't typically know that.

Also, for a given network, equal mass flows should correspond to equal pressure drops.

The problem with only pressure at boundaries is that, for such a long domain, it is much longer to converge, because the mass flow has to establish (as opposed to be fixed). But, again, this depends from the specific code as well