Parametric study

This example shows how you can use PyAdditive to perform a parametric study. The intended audience is a user who desires to optimize additive machine parameters to achieve a specific result. Here, the ParametricStudy class is used to conduct a parametric study. While this is not required, the ParametricStudy class provides data management and visualization features that ease the task.

Units are SI (m, kg, s, K) unless otherwise noted.

Perform required imports and create a study

Perform the required import and create a ParametricStudy instance.

from ansys.additive.core import Additive, SimulationStatus, SimulationType
from ansys.additive.core.parametric_study import ColumnNames, ParametricStudy
import ansys.additive.core.parametric_study.display as display

study = ParametricStudy("demo-study")

Saving parametric study to /home/runner/work/pyadditive/pyadditive/examples/demo-study.ps

Get the study file name

The current state of the parametric study is saved to a file upon each update. You can retrieve the name of the file as shown below. This file uses a binary format and is not human readable.

print(study.file_name)

/home/runner/work/pyadditive/pyadditive/examples/demo-study.ps

Select a material for the study

Select a material to use in the study. The material name must be known by the Additive service. You can connect to the Additive service and print a list of available materials prior to selecting one.

additive = Additive()
print("Available material names: {}".format(additive.materials_list()))
material = "IN718"

user data path: /home/runner/.local/share/pyadditive
Available material names: ['316L', 'Ti64', '17-4PH', 'CoCr', 'IN718', 'IN625', 'AlSi10Mg', 'Al357']

Create a single bead evaluation

Parametric studies often start with single bead simulations in order to determine melt pool statistics. Here, the generate_single_bead_permutations() method is used to generate single bead simulation permutations. The parameters for the generate_single_bead_permutations() method allow you to specify a range of machine parameters and filter them by energy density. Not all the parameters shown are required. Optional parameters that are not specified use default values defined in the MachineConstants class.

import numpy as np

# Specify a range of laser powers. Valid values are 50 to 700 W.
initial_powers = np.linspace(50, 700, 7)
# Specify a range of laser scan speeds. Valid values are 0.35 to 2.5 m/s.
initial_scan_speeds = np.linspace(0.35, 2.5, 5)
# Specify powder layer thicknesses. Valid values are 10e-6 to 100e-6 m.
initial_layer_thicknesses = [40e-6, 50e-6]
# Specify laser beam diameters. Valid values are 20e-6 to 140e-6 m.
initial_beam_diameters = [80e-6]
# Specify heater temperatures. Valid values are 20 - 500 C.
initial_heater_temps = [80]
# Restrict the permutations within a range of energy densities
# For single bead, the energy density is laser power / (laser scan speed * layer thickness).
min_energy_density = 2e6
max_energy_density = 8e6
# Specify a bead length in meters.
bead_length = 0.001

study.generate_single_bead_permutations(
    material_name=material,
    bead_length=bead_length,
    laser_powers=initial_powers,
    scan_speeds=initial_scan_speeds,
    layer_thicknesses=initial_layer_thicknesses,
    beam_diameters=initial_beam_diameters,
    heater_temperatures=initial_heater_temps,
    min_area_energy_density=min_energy_density,
    max_area_energy_density=max_energy_density,
)

Show the simulations as a table

You can use the display package to list the simulations as a table.

display.show_table(study)

show_table

Skip some simulations

If you are working with a large parametric study, you may want to skip some simulations to reduce processing time. To do so, set the simulation status to SimulationStatus.SKIP which is defined in the SimulationStatus class. Here, a DataFrame object is obtained, a filter is applied to get a list of simulation IDs, and then the status is updated on the simulations with those IDs.

df = study.data_frame()
# Get IDs for single bead simulations with laser power below 75 W.
ids = df.loc[
    (df[ColumnNames.LASER_POWER] < 75) & (df[ColumnNames.TYPE] == SimulationType.SINGLE_BEAD),
    ColumnNames.ID,
].tolist()
study.set_status(ids, SimulationStatus.SKIP)
display.show_table(study)

show_table

Run single bead simulations

Run the simulations using the run_simulations() method. All simulations with a SimulationStatus.PENDING status are executed.

study.run_simulations(additive)

2024-02-21 03:22:39 Completed 0 of 31 simulations
2024-02-21 03:23:00 Completed 1 of 31 simulations
2024-02-21 03:23:14 Completed 2 of 31 simulations
2024-02-21 03:23:26 Completed 3 of 31 simulations
2024-02-21 03:23:38 Completed 4 of 31 simulations
2024-02-21 03:23:51 Completed 5 of 31 simulations
2024-02-21 03:24:04 Completed 6 of 31 simulations
2024-02-21 03:24:16 Completed 7 of 31 simulations
2024-02-21 03:24:28 Completed 8 of 31 simulations
2024-02-21 03:24:41 Completed 9 of 31 simulations
2024-02-21 03:24:53 Completed 10 of 31 simulations
2024-02-21 03:25:11 Completed 11 of 31 simulations
2024-02-21 03:25:22 Completed 12 of 31 simulations
2024-02-21 03:25:33 Completed 13 of 31 simulations
2024-02-21 03:25:47 Completed 14 of 31 simulations
2024-02-21 03:25:58 Completed 15 of 31 simulations
2024-02-21 03:26:10 Completed 16 of 31 simulations
2024-02-21 03:26:25 Completed 17 of 31 simulations
2024-02-21 03:26:37 Completed 18 of 31 simulations
2024-02-21 03:26:48 Completed 19 of 31 simulations
2024-02-21 03:27:01 Completed 20 of 31 simulations
2024-02-21 03:27:15 Completed 21 of 31 simulations
2024-02-21 03:27:30 Completed 22 of 31 simulations
2024-02-21 03:27:44 Completed 23 of 31 simulations
2024-02-21 03:28:05 Completed 24 of 31 simulations
2024-02-21 03:28:26 Completed 25 of 31 simulations
2024-02-21 03:28:39 Completed 26 of 31 simulations
2024-02-21 03:28:54 Completed 27 of 31 simulations
2024-02-21 03:29:08 Completed 28 of 31 simulations
2024-02-21 03:29:29 Completed 29 of 31 simulations
2024-02-21 03:29:45 Completed 30 of 31 simulations
2024-02-21 03:29:56 Completed 31 of 31 simulations

Save the study to a CSV file

The parametric study is saved with each update in a binary format. For other formats, use the to_* methods provided by the DataFrame class.

study.data_frame().to_csv("demo-study.csv")

Load a previously saved study

Load a previously saved study using the static ParameticStudy.load() method.

study2 = ParametricStudy.load("demo-study.ps")
display.show_table(study2)

show_table

Plot single bead results

Plot the single bead results using the single_bead_eval_plot() method.

display.single_bead_eval_plot(study)

single_bead_eval_plot

Create a porosity evaluation

You can use the insights gained from the single bead evaluation to generate parameters for a porosity evaluation. Alternatively, you can perform a porosity evaluation without a previous single bead evaluation. Here, the laser power and scan speeds are determined by filtering the single bead results where the ratio of the melt pool reference depth to reference width is within a specified range. Additionally, the simulations are restricted to a minimum build rate, which is calculated as scan speed * layer thickness * hatch spacing. The generate_porosity_permutations() method is used to add porosity simulations to the study.

df = study.data_frame()
df = df[
    (df[ColumnNames.MELT_POOL_REFERENCE_DEPTH_OVER_WIDTH] >= 0.3)
    & (df[ColumnNames.MELT_POOL_REFERENCE_DEPTH_OVER_WIDTH] <= 0.65)
]

study.generate_porosity_permutations(
    material_name=material,
    laser_powers=df[ColumnNames.LASER_POWER].unique(),
    scan_speeds=df[ColumnNames.SCAN_SPEED].unique(),
    size_x=1e-3,
    size_y=1e-3,
    size_z=1e-3,
    layer_thicknesses=[40e-6],
    heater_temperatures=[80],
    beam_diameters=[80e-6],
    start_angles=[45],
    rotation_angles=[67.5],
    hatch_spacings=[100e-6],
    min_build_rate=5e-9,
    iteration=1,
)

Run porosity simulations

Run the simulations using the run_simulations() method.

study.run_simulations(additive)

2024-02-21 03:30:00 Completed 0 of 15 simulations
2024-02-21 03:30:50 Completed 1 of 15 simulations
2024-02-21 03:31:27 Completed 2 of 15 simulations
2024-02-21 03:31:59 Completed 3 of 15 simulations
2024-02-21 03:32:53 Completed 4 of 15 simulations
2024-02-21 03:33:32 Completed 5 of 15 simulations
2024-02-21 03:34:05 Completed 6 of 15 simulations
2024-02-21 03:35:14 Completed 7 of 15 simulations
2024-02-21 03:35:57 Completed 8 of 15 simulations
2024-02-21 03:36:30 Completed 9 of 15 simulations
2024-02-21 03:38:29 Completed 10 of 15 simulations
2024-02-21 03:39:17 Completed 11 of 15 simulations
2024-02-21 03:39:52 Completed 12 of 15 simulations
2024-02-21 03:42:47 Completed 13 of 15 simulations
2024-02-21 03:43:59 Completed 14 of 15 simulations
2024-02-21 03:44:37 Completed 15 of 15 simulations

Plot porosity results

Plot the porosity simulation results using the porosity_contour_plot() method.

display.porosity_contour_plot(study)

porosity_contour_plot

Create a microstructure evaluation

Here a set of microstructure simulations is generated using many of the same parameters used for the porosity simulations. The parameters cooling_rate, thermal_gradient, melt_pool_width, and melt_pool_depth are not specified so they are calculated. The generate_microstructure_permutations() method is used to add microstructure simulations to the study.

df = study.data_frame()
df = df[(df[ColumnNames.TYPE] == SimulationType.POROSITY)]

study.generate_microstructure_permutations(
    material_name=material,
    laser_powers=df[ColumnNames.LASER_POWER].unique(),
    scan_speeds=df[ColumnNames.SCAN_SPEED].unique(),
    size_x=1e-3,
    size_y=1e-3,
    size_z=1.1e-3,
    sensor_dimension=1e-4,
    layer_thicknesses=df[ColumnNames.LAYER_THICKNESS].unique(),
    heater_temperatures=df[ColumnNames.HEATER_TEMPERATURE].unique(),
    beam_diameters=df[ColumnNames.BEAM_DIAMETER].unique(),
    start_angles=df[ColumnNames.START_ANGLE].unique(),
    rotation_angles=df[ColumnNames.ROTATION_ANGLE].unique(),
    hatch_spacings=df[ColumnNames.HATCH_SPACING].unique(),
    iteration=2,
)

Run microstructure simulations

Run the simulations using the run_simulations() method.

study.run_simulations(additive)

2024-02-21 03:44:39 Completed 0 of 15 simulations
2024-02-21 03:46:52 Completed 1 of 15 simulations
2024-02-21 03:48:32 Completed 2 of 15 simulations
2024-02-21 03:49:56 Completed 3 of 15 simulations
2024-02-21 03:52:08 Completed 4 of 15 simulations
2024-02-21 03:53:50 Completed 5 of 15 simulations
2024-02-21 03:55:15 Completed 6 of 15 simulations
2024-02-21 03:57:32 Completed 7 of 15 simulations
2024-02-21 03:59:15 Completed 8 of 15 simulations
2024-02-21 04:00:40 Completed 9 of 15 simulations
2024-02-21 04:02:57 Completed 10 of 15 simulations
2024-02-21 04:04:44 Completed 11 of 15 simulations
2024-02-21 04:06:11 Completed 12 of 15 simulations
2024-02-21 04:08:32 Completed 13 of 15 simulations
2024-02-21 04:10:18 Completed 14 of 15 simulations
2024-02-21 04:11:49 Completed 15 of 15 simulations

Plot microstructure results

Plot and compare the average grain sizes from the microstructure simulations using the ave_grain_size_plot() method.

display.ave_grain_size_plot(study)

ave_grain_size_plot