Convergence

Contents

Convergence#

This example demonstrates how to retrieve, process, and visualize convergence data from a completed Flow360 simulation using the Python API. It outlines the procedure for setting up a basic steady-state case, running it, and subsequently accessing and plotting convergence metrics such as nonlinear residuals and CFL numbers.

  1from pylab import show
  2
  3import flow360 as fl
  4from flow360.examples import OM6wing
  5
  6OM6wing.get_files()
  7
  8project = fl.Project.from_volume_mesh(
  9    OM6wing.mesh_filename,
 10    name="Convergence results from Python",
 11)
 12
 13vm = project.volume_mesh
 14
 15with fl.SI_unit_system:
 16    params = fl.SimulationParams(
 17        reference_geometry=fl.ReferenceGeometry(
 18            area=1.15315, moment_center=[0, 0, 0], moment_length=[1.47602, 0.801672, 1.47602]
 19        ),
 20        operating_condition=fl.AerospaceCondition(velocity_magnitude=286, alpha=3.06 * fl.u.deg),
 21        time_stepping=fl.Steady(
 22            max_steps=5000, CFL=fl.RampCFL(initial=1, final=100, ramp_steps=1000)
 23        ),
 24        models=[
 25            fl.Fluid(
 26                navier_stokes_solver=fl.NavierStokesSolver(
 27                    absolute_tolerance=1e-9, linear_solver=fl.LinearSolver(max_iterations=35)
 28                ),
 29                turbulence_model_solver=fl.SpalartAllmaras(
 30                    absolute_tolerance=1e-8, linear_solver=fl.LinearSolver(max_iterations=25)
 31                ),
 32            ),
 33            fl.Wall(surfaces=vm["2"]),
 34            fl.SlipWall(surfaces=vm["1"]),
 35            fl.Freestream(surfaces=vm["3"]),
 36        ],
 37        outputs=[
 38            fl.SurfaceOutput(surfaces=vm["1"], output_fields=["Cp", "CfVec"]),
 39            fl.VolumeOutput(output_fields=["Cp", "Mach", "qcriterion"]),
 40        ],
 41    )
 42
 43case = project.run_case(params, "Convergence case from Python")
 44
 45
 46# wait until the case finishes execution
 47case.wait()
 48
 49results = case.results
 50
 51# nonlinear residuals contain convergence information
 52nonlinear_residuals = results.nonlinear_residuals.as_dataframe()
 53print(nonlinear_residuals)
 54
 55nonlinear_residuals.plot(
 56    x="pseudo_step",
 57    y=["0_cont", "1_momx", "2_momy", "3_momz", "4_energ", "5_nuHat"],
 58    xlim=(0, None),
 59    xlabel="Pseudo Step",
 60    secondary_y=["5_nuHat"],
 61    figsize=(10, 7),
 62    title="Nonlinear residuals",
 63)
 64
 65max_residual_location = results.max_residual_location.as_dataframe()
 66print(max_residual_location)
 67
 68max_residual_location.plot(
 69    x="pseudo_step",
 70    y=[
 71        "max_cont_res",
 72        "max_momx_res",
 73        "max_momy_res",
 74        "max_momz_res",
 75        "max_energ_res",
 76        "max_nuHat_res",
 77    ],
 78    xlabel="Pseudo Step",
 79    xlim=(0, None),
 80    ylim=(-25, None),
 81    secondary_y=["max_nuHat_res"],
 82    figsize=(10, 7),
 83    title="Max residual location",
 84)
 85show()
 86
 87cfl = results.cfl.as_dataframe()
 88print(cfl)
 89
 90cfl.plot(
 91    x="pseudo_step",
 92    y=["0_NavierStokes_cfl", "1_SpalartAllmaras_cfl"],
 93    xlim=(0, None),
 94    xlabel="Pseudo Step",
 95    figsize=(10, 7),
 96    title="CFL",
 97)
 98show()
 99
100results.set_destination(use_case_name=True)
101results.download(
102    nonlinear_residuals=True,
103    linear_residuals=True,
104    cfl=True,
105    minmax_state=True,
106    max_residual_location=True,
107)

Notes#

  • The script utilizes case.wait() to pause execution until the simulation completes, ensuring that results are available for post-processing.

  • Convergence history, including nonlinear residuals (results.nonlinear_residuals), maximum residual locations (results.max_residual_location), and CFL values (results.cfl), are accessed via the results object. These are converted to pandas DataFrames using .as_dataframe() for convenient analysis and plotting.

  • Specific convergence-related files can be downloaded using the results.download() method by specifying the desired data types (e.g., nonlinear_residuals=True, cfl=True).

Contents