Continuity Convergence Issue

    • fuad87

      Hi all,

      I know there is already enough discussion on how to control or analyze continuity residuals. One suggestion states that "keep the imbalance less than 1%". And also nobody explicitly expresses what will happen if the fluid is compressible and shock is entering in the domain. So let me explain what I am doing and what specific questions I hope to get the answer for.

      My problem is multiphase, transient, compressible. The domain is 5mm x 5mm axisymmetric. x-axis is the 'axis', while there is a pressure inlet at right and pressure outlet in the left, and the top is the wall. whole domain is initialized with 101325 pa (P_op = 0 Pa) and then a vapor bubble of spherical shape is patched. Pressure inlet is defined such that 2GPa shock enter with Mach 2. I want to collapse the bubble with shock interaction. Now after some iteration, if I check the "mass imbalance" from contour plot I see it varies from -1E-7 to +1E-7. This is definitely cell based value. Now If I calculate mass flow rate from 'report' and select inlet and outlet and I see the net mass flow in the order of, O(1E3).

      1. So my first question is how to compare the net mass flow rate in the domain (inlet minus outlet) to the cell value of mass imbalance? 

      the second problem, since my domain experience mass influx, but due to the compressibility (incompressible means infinite sound speed and inflow immediately will push outflow) the inflow shock (which also brings mass into the domain) is not felt by the outlet. And I realize that in each cell it might be also true that one face is experiencing inflow while the other side face might not have any outflow. Does it mean that the cell based mass convergence will never be achieved? 

      2. So my second question is, in my problem I am getting continuity residual below 1e-7 for first couple of iteration but as soon as shock enter couple of cell distance, the continuity residuals are always in the order of, O(1) to O(10). is it acceptable? given that all other residuals remain below O(1E-6). 


    • DrAmine
      Ansys Employee

      The residuals which are available for post-processing are cell based residuals and are not scaled. 


      1/You will need to do that balance for each single cell as the reported residual value is cell based. That is why this step is almost never used. Tthe difference between inflow and outflow is the accumulation in the domain and this would be similar to the sum of the mass imbalance. For multiphase this might be more difficult due to density biasing.

      2/Sometimes it is appropratae to check unscaled residuals as well as much better to monitor the run. Check the paragraph juding convergence in the Fluent documentation


      Sincerly I do not really understand the problem you are trying to describe. Perhaps more information or pictures will help us understand more.

    • MHDJ


      I have the following question and I'd be thankful if you could help me with this regards,

      Flow pattern in a cylindrical pipe was studied whereby there is flow into the pipe but no outflow from the other end of the pipe. The simulation was modeled with the following boundary conditions: Velocity inlet, and the outlet was modeled as wall (since there is no outflow). Fluent solved this scenario and gave the velocity profile and pressure profile result.

      Since there is no outflow in the pipe, there should be no flow in the pipe, while the outflow is closed and the fluid is incompressible the change in the contour of velocity field appears. Also, the result obtained does not explains the mass conservation in the system and amazingly says the continuity is converged and the continuity residual reaches the below residual criteria 1e-6.

      Thank you in advance.

    • DrAmine
      Ansys Employee
      Fluent will just propagate the momentum flux from inlet. The case will converge by residuals even if it's providing garbage as solution. Better to check mass imbalance if the system.
      This kind of runs should accoubt for compressibility and transient pressure build-up
Viewing 3 reply threads
  • You must be logged in to reply to this topic.