Solution does not converge for 7 bar pressure.

  • 222 Views
  • Last Post 30 August 2018
nav.shirur posted this 29 August 2018

Hello there,

I know basics of Ansys Fluent and little bit of theory behind it.

I am simulating a simple compressible gas flow in ANSYS 19.1 Fluent. My problem details are as follows:

Inlet : 6 bar (gauge)

Outlet : 0 (Gauge)

Spalart Allmaras Model

Time step size : 1e-5

Max cell size : 0.7 mm

Inlet size: 2mm

Outlöet dia: 25mm

Convergence conditions; 1e-5

Explicit formulation

Hybrid Initialization

Second order Upwind

My questions here are:

1. Residuals after 1e-3 or 1e-4 follow horizontal line nad do not converge anymore. Should i consider solution at 1e-3 or 1e-4?

2. How to analyse residuals?

3. Should i trust solution at 1e-3 or 1e-4?

 

Attached Files

Order By: Standard | Newest | Votes
Kremella posted this 29 August 2018

Hello,

Couple of questions / suggestions:

  • What is your set residual criteria? Also, how many iterations per time-step are you running? Are you converging every time-step?
  • Please try and monitor other physical parameters such as velocity or mass flow rate at the outlet. Are you seeing a steady state behavior as a function of time? You can also plot this as a function of iteration and check if you are reaching a constant value (roughly) every single time-step.
  • If you feel you are not getting the desired convergence you require, you might want to check your mesh quality. What is your min orthogonal quality and max skewness values?
  • Please elaborate a little more on your model. Why are you using a pressure inlet boundary condition as opposed to mass flow rate inlet?
  • You might also want to think about slowly ramping up your pressure in small increments. Since you are solving a transient problem, you might want to be careful. But perhaps, you might want to start at a lower inlet pressure and let your simulation run to a steady state. Then without re-initializing, you might want to increase your pressure to a slightly higher value and again let the simulation run to a steady state. You continue this process until you get to the pressure condition you want to investigate. This method is extremely useful for steady state simulations, but when employed to transient, it would stretch your overall simulation time.

I hope these points help.

Please let us know what you find.

Thank you.

Best Regards,
Karthik

nav.shirur posted this 29 August 2018

What is your set residual criteria? Also, how many iterations per time-step are you running? Are you converging every time-step?

Ans: Residual is absolute 1e-5 for X, Y Z, Continuity and Energy. Initially i started with only one time step and 5000 iterations to know where solution would converge. Solution reaches 1e-4 at approximately 2000 iterations and they goes straight. In this case i do not understand if i have to belive results or not. Could u please let me know how to monitor physical property by plotting against iteration and how to judge conservation (mass and energy?)

 

If you feel you are not getting the desired convergence you require, you might want to check your mesh quality. What is your min orthogonal quality and max skewness values?

From theory i understood acceptable results would be 1e-5 residuals,but is there a way to decide what should be residual value for a specific problem?Mesh quality is as follows,

Max Skewness: 0.797

Orthogonal Quality minimum: 0.2022

Max Aspect Ratio: 9.57

Cell size: 0.7mm

 

Please elaborate a little more on your model. Why are you using a pressure inlet boundary condition as opposed to mass flow rate inlet?

Since it is a problem of flow from high pressure container, i used pressure inlet and pressure outlet with compressible flow.  Moreover, i used transient phenomena so thought pressure is easy.

 

You might also want to think about slowly ramping up your pressure in small increments. Since you are solving a transient problem, you might want to be careful. But perhaps, you might want to start at a lower inlet pressure and let your simulation run to a steady state. Then without re-initializing, you might want to increase your pressure to a slightly higher value and again let the simulation run to a steady state. You continue this process until you get to the pressure condition you want to investigate. This method is extremely useful for steady state simulations, but when employed to transient, it would stretch your overall simulation time.

I don't know to how simulate using ramp concept and how to solve without reinitializing. Could u please throw somw light on this?

 

rwoolhou posted this 30 August 2018

To plot values during the simulation you need to create a Report Definition and then plot it.

Alternatively use the Execute Commands tools to save an image every some iterations, then create a movie of these images: this will show any instabilities in the flow. I'd recommend using the TUI to create the commands as it's more reliable, press <Enter> in the text window to see a list of commands, typing the command will take you to the next level etc.  To move back up a level use q

A sample command to display velocity contours (single phase) would be:

/display/contour/velocity 0 10      where 0 10 is the range I want to plot.  I'll leave you to find the save-picture syntax. 

Close