Learning Forum

[Dato]

Hi,

I have model below and simulating natural convection. Two phases, namely water-liquid and water-vapor, are considered. The lower piper on the right is heating wall. At initial condition, whole pipe was filled with water except that upper half of water tank was filled with water-vapor. Water was heated and would boil in rising section because saturation temperature decrease with decreasing water pressure. Therefore, a natural circulation is formed. 

In the case, saturation temperature used in Lee model is specified as pressure dependent. At initial few seconds, all seems run well. But residual suddenly ascended and diverged.  Fluent console shown turbulent viscosity turbulent viscosity limited to viscosity ratio of 1.000000e+05 and limited temperature of 5K. 

It's temperature before divergence. The strange situation keep exsiting with increasing flow time. The vapor in upper tank will be affected by heated water due to decreasing density and boiling. Therefore adverse flow warning keep existing during simulation. But the high temperature is unanticipated. I don't know if solution diverged because of vapor temperature in upper tank. 

What's happending during simulation?  Any suggestions given is highly appreciated!!

[Rob]

How are you adding heat, and how well resolved is the mesh? If you're boiling the water at the bottom how are you cooling it elsewhere?

[Dato]

Constant temperature is applied at heat-wall as shown. The others wall is adiabatic. Heat is mainly removed by heat conduction and partial heat would be used to heat water-vapor at upper tank because flow of water is drived by density change. But there is no way that vapor temperature is higher than water in the tank. 

for mesh, all radial division is 5 and axially strench with maximum aspect ratio is 8. Maybe radial mesh need to be increase slightly. But it would  lead to large amount of mesh. 

I'm wondering whether I need to expand numeric outlet region to avolid adverse flow in physical outlet. But it would also increase mesh number. 

 

 

 

[Rob]

For a single phase model 5 cells across the tubes would be considered coarse. If you expect vapour there too you need a lot more mesh. With a fixed temperature the liquid should reach that value, as will the whole domain in time, that's probably not the intended result? 

[Dato]

Yes, vapor will appear near the wall. Actually, I haven't do mesh independence analysis because accuracy is not the concern now. It's also because the huge amount of mesh due to the large scale model with 10mm pipe diameter while total length is around 20 meters. It's also a transient situation. Initial temeprature of water in the domain is 293K and it's expected to be heated to generate vapor.  

The whole domain reaches the fixed temperature is intended final result. But It takes more than 1 thounsand seconds to obtain desired result, which is calculated by other system code.  But it diverged after about hundred seconds as illustrated above. During the calculated period, vapour was generated. Do you mean mesh should be refined to achieve converge in the case?  But time consumption is really a great consideration. 

I think that one possible reason causing divergence is the pressure-outlet boundary condition. Sorry that I haven't make my case condition clear. Initially, whole domain is 293K, and upper volume is filled with air whose lower surface is level with the pipe extending into the tank. Then the water is heated and natural circulation is forming. Air will fluctuate irregularly. Whole device is assumed to be  placed at room temperature so I have set adverse flow temperature to 293K but actually it isn't. That's maybe the reason causing divergence with water heated. 

Do you have any advice? Thank you so much.  

[Rob]

If you generate any vapour you need to resolve that. So, it's not just a case of mesh independence but of actually capturing the physics. Failure to do so may cause stability issues. Whilst we can run models with those dimensions and for those durations you really need to rethink what you're wanting from the models. 

As vapour moves in the pipes you may also experience flow regime change: look up the Baker Chart for some background reading. That can also result in some very rapid changes in the flow field, that may require smaller time steps. That may also influence the water level. 

[Dato]

Yes, it's necessary to refine the mesh to capture some critical physics. But now the most important thing now for me is to achieve convergence in flow time for more than one thounsand seconds. Insufficient mesh density may be one reason. I understand flow regime changes in pool boiling or flow boiling, such as slug, bubbly, mist and the like. What I want to do now is to obtain a rough solution with less time consumption because deadline is approaching. 😂

But the biggest problem, I think, is highly related to the appearance of reversed flow. And it would appear actually. While the total temperature of reversed flow setting in pressure-outlet will increase because the air at upper tank will be heated, which is confirmed in my calculation before divergence, I have no idea that how total temperature of reversed flow change and specify it as constant initial temperature. It's time when superheated fluid on the right pipe touched the air at lower temperature that caused some problems, I think.  

What could be possible solution? Extend calculation domain near the outlet? Or have a preliminary theoretical calculation and specify a time dependent reversed total temperature? 

Thanks for you patient reply.