Fluids

Fluids

Effect of Momentum equation spatial discretization scheme on the solution

    • snr1
      Subscriber
      This query is mainly intended to know the effect of momentum equation spatial discretization scheme on the solution and convergence history.nA simulation was converged with 1st order upwind momentum discretization scheme but it is not converging with the other high order schemes.nWhat might be the reason for this? how to get good convergence with the high order schemes in this situation?nCan we consider the lower scheme order solution as final solution of the simulation?n
    • YasserSelima
      Subscriber
      Decrease the under relaxation factors. And check the box, use higher order relaxation. If you are doing unsteady simulation, decrease the time step and increase the number of iterations in every time step. If running steady, Pseudo transient is a good option.nDepending on the application. If you are using this to test theory or to make a draft design, yes ... but if you want to publish this data or to optimise a design, no. n
    • snr1
      Subscriber
      Dear YassernI decreased the URF, enabled the higher order relaxation and pseudo transient formulation. But, still my residuals are not decreasing. In present case, whenever i change the material properties, the residuals are not decreasing. Means the material properties are influencing the convergence history. In this case how we can make the case to converged solution?n
    • YasserSelima
      Subscriber
      When you run First-order with a set material properties, you get a good preliminary solution. Then, when you use higher order terms, you are refining the solution.nWhen you change the properties after obtaining the preliminary solution, the problem becomes a different problem. And your preliminary solution might not be valid.nCan you show your mesh? and the residuals plotn
    • snr1
      Subscriber
      After getting the preliminary solution (i.e. converged solution) with first order scheme, i used second order schemes, which leads to non-convergent solution.nI am not changing the properties at all while switching the discretization schemes. i.e. the properties are same for both first order and high order schemes.nMesh:nnn.Residualsnn5.Temperature profilenn
    • DrAmine
      Ansys Employee
      Actually you should never on an arbitrary mesh publish or quantify results coming from a first order runs. Saying that one also needs to note that high order schemes are more prone to wiggles and stability issues but this can be mitigated by for example starting from first order run results and continuing iterating or using HOTR in Fluent or even relaxing a bit the the blending between first order and second order schemes.nn
    • snr1
      Subscriber
      I am not using an arbitrary mesh and i have tried the above said methods except the blending. In all the cases, the results are not converging with higher order schemes whereas solution is getting converged in first order scheme.n
    • YasserSelima
      Subscriber
      This is converged solution. You have residuals converging and reaching constant value. The monitored value reached steady state.nThe reason you are not getting lower residuals could be the nature of the problem. When you try steady simulation of a case that is unsteady by nature.nn
    • DrAmine
      Ansys Employee
      The residual you are using are global scaled which are level dependent and also depend on the URF's etc ( I recommend to use local scaling for residual to have better judgment in general). But as stated by Yasser: if all other quantities are converging (monitor points, imbalances,..) then you might have already a good solution. Residual are only showing how the solution might change relative to solution of previous iterations and it does contain a lot of scaling. You might perhaps also consider looking into the raw residual without any scalingn
    • snr1
      Subscriber
      From the pictures, we can say that the simulation was converged. Even, the monitored value also reached steady state value. Apart from the temperature, I am also monitoring another value (Swirl number), which is varying based on the iteration number though it has reached steady state as seen from the above. How to tackle this issue?n
Viewing 9 reply threads
  • You must be logged in to reply to this topic.