Analysis of Film Cooling on Flat Plate- not attaining steady state.

    • Ananth Narayan
      Hello all I am working on post-graduate project on film cooling on flat plate (without rotation). I tried with various pressure-velocity coupling schemes, but even after 1.5 day my result is not attaining steady state, and the contours of temperature is not changing after certain iterations even after 3-4 hours of solving them.
      Flow will be turbulent and hence for y+ = 1, first cell spacing from wall has been made, and aspect ratio (max) is around 500, total mesh cell count around 1.3 million. Image of geometry is attached below [Paper]: (here D = 6.35 mm)

      I tried almost all possible ways I know to attain it steady state, but nothing worked. The possible combinations which I tried are listed below:
      1.     When I start to reduce the mesh size, I see that steady state convergence is not reached.
      2.     For SIMPLE, PISO residuals diverge to 1e5 and I stopped the calculations. But for COUPLED and SIMPLEC, residuals are converging but after continuity residuals 1e-3 they are not converging anymore.
      3.     For COUPLED with and without Pseudo-Transient, I am not getting steady state convergence. For both my continuity residuals attains 1e-3 and after that convergence doesn't occur. For your reference, I am attaching an image without Pseudo-Transient after 55000 iterations. Also even after so long time, I find the same contour.

      4.     For SIMPLEC, same thing happens as of COUPLED solver.
      5.     If I do transient analysis, for time step above 1e-3- solution starts diverging, but when I go below that it takes so so long time.
      For your reference, I am attaching similar kind of simulation I did with other geometry:
      Journal Paper
      The paper linked uses temperature piecewise polynomial for Cp, K, Dynamic Viscosity and density is considered as ideal gas. Since both fluid coming at different density and velocity this is one of the way to replicate the phenomenon. Else we can directly use species transport if density and other parameters are not relatable to each other.
      I have used piecewise polynomial relation so as to validate Fig 6(a) for Re = 8e4. As every parameters are know we can find mainstream velocity by, (here 6.35 mm is diameter of cooling tube)

      How do I achieve the steady state solution, am I missing something or do I need more computational power. I am using Intel i7 with 16 GB, 8 Processor CPU.

      For your reference, I am attaching case file. (Please replace .txt to .cas.h5)
      Thanks in advance

      Regards Ananth Narayan
      M.Tech Thermal Engineering

    • Swathi V. V.
      Ansys Employee
      Hi Ananth Ensure the mesh quality is atleast 0.1. Aong with residual, please check if flux imbalance is low and create monitor points to determine if the solution is converged. Refer to the following documents for more details: 32.21.1. Judging Convergence (ansys.com) and 32.15.3. Monitoring Solution Quantities (ansys.com). You can initally run steady state simulation and then switch to transient to start with a reasonable initial guesses and for fasterconvergence. If you are facing divergence, start your simulation with lower under relaxation factor and then as the solution gets stable increase it to default value.
      You can use adaptive time stepping to change time step size as the calculation proceeds and solution attains stability. Please refer to 32.13.1. Inputs for Time-Dependent Problems (ansys.com).
    • Ananth Narayan
      Hello, thanks for initiating answer for this question. Does mesh quality is orthogonal quality?. I am using ICEM CFD for meshing and min skewness is 0.5.

      If suppose I reduce the relaxation factor and run steady state and then switch to transient state, do I need to follow adaptive time step or fixed time step. What is the reasonable CFL number I should be using. If not can I use expression for time step as the time increases, time step size should decrease?

      Thanks again

      ~Ananth Narayan
Viewing 2 reply threads
  • You must be logged in to reply to this topic.