Dns Horizontal Velocity Comparison
Horizontal Velocity Comparison
Section titled “Horizontal Velocity Comparison”This notebook compares the mean-square horizontal velocities between DNS data and the model prediction.
- Data Loading
- Function Definitions for Reusable Calculations
- Plotting and Graph Comparisons
In this notebook, we replicate and compare the collision kernels from the DNS data as presented in Figure 12 of the following reference:
Reference: Ayala, O., Rosa, B., & Wang, L. P. (2008). Effects of turbulence on the geometric collision rate of sedimenting droplets. Part 2. Theory and parameterization. New Journal of Physics, 10. https://doi.org/10.1088/1367-2630/10/7/075016
Python · section 1
import numpy as npimport matplotlib.pyplot as pltimport particula as par
from particula.dynamics.coagulation.turbulent_dns_kernel.sigma_relative_velocity_ao2008 import ( _compute_rms_fluctuation_velocity, VelocityCorrelationTerms,)
from particula.dynamics.coagulation.turbulent_dns_kernel.velocity_correlation_terms_ao2008 import ( compute_b1, compute_b2, compute_c1, compute_c2, compute_d1, compute_d2, compute_e1, compute_e2, compute_z, compute_beta,)Data Loading
Section titled “Data Loading”This section loads the necessary DNS datasets and corresponding model predictions. Verify that the file paths and dataset formats are correct.
Python · section 3
# Figure 12: Comparison of the predicted and simulated mean-square horizontal# particle velocities for droplets falling in a turbulent flow of Rλ = 72.41 and# turbulent_dissipation = 400 cm2 s−3
# droplet radius (a2, microns) vs rms_velocity (cm2/s2)
# dns_10cm2/s3: 6 rows, 2 columns (X, Y)dns_10cm2_s3 = np.array( [ [9.938118812, 26.66666667], [20.02475248, 26.41975309], [30.04950495, 26.41975309], [40.01237624, 24.69135802], [50.16089109, 22.71604938], [60.06188119, 18.51851852], ])
# dns_100_cm2/s3: 6 rows, 2 columns (X, Y)dns_100_cm2_s3 = np.array( [ [9.938118812, 84.44444444], [20.02475248, 80.98765432], [29.98762376, 77.03703704], [39.95049505, 71.11111111], [49.97524752, 59.25925926], [60.06188119, 44.19753086], ])
# dns_400_cm2/s3: 6 rows, 2 columns (X, Y)dns_400_cm2_s3 = np.array( [ [9.876237624, 166.9135802], [20.08663366, 163.9506173], [30.11138614, 150.1234568], [40.07425743, 129.1358025], [50.03712871, 100.4938272], [60.06188119, 69.62962963], ])Function Definitions
Section titled “Function Definitions”The following functions perform repeated calculations and data manipulations:
- calculate_horizontal_velocity: Computes the mean-square horizontal velocity of particles.
Python · section 5
# Define Particle Radii and Parametersparticle_radius = np.linspace(10e-6, 60e-6, 6)temperature = 273 # Temperature in Kelvinparticle_density = 1000 # Particle density in kg/m³fluid_density = 1.0 # Fluid (air) density in kg/m³air_velocity = 1e-9 # Relative velocity in m/s
reynolds_lambda = 72.41 # Example value
# Calculate dynamic and kinematic viscositydynamic_viscosity = par.gas.get_dynamic_viscosity(temperature)kinematic_viscosity = par.gas.get_kinematic_viscosity( dynamic_viscosity, fluid_density)
# Calculate Particle Settling Velocitymean_free_path = par.gas.get_molecule_mean_free_path( temperature=temperature, dynamic_viscosity=dynamic_viscosity)
# 2. Slip correction factorsknudsen_number = par.particles.get_knudsen_number( mean_free_path=mean_free_path, particle_radius=particle_radius)slip_correction_factor = par.particles.get_cunningham_slip_correction(knudsen_number)# iterative terminal settling velocityparticle_settling_velocity = ( par.particles.get_particle_settling_velocity_with_drag( particle_radius=particle_radius, particle_density=particle_density, fluid_density=fluid_density, dynamic_viscosity=dynamic_viscosity, slip_correction_factor=slip_correction_factor, ))# Calculate Particle Inertia Timeparticle_inertia_time = par.particles.get_particle_inertia_time( particle_radius=particle_radius, particle_density=particle_density, fluid_density=fluid_density, kinematic_viscosity=kinematic_viscosity,)
def calculate_horizontal_velocity(turbulent_dissipation, reynolds_lambda): """ Helper function to calculate the mean-square horizontal velocity of particles for these specific cases. """ # Calculate Fluid RMS Velocity fluid_rms_velocity = par.gas.get_fluid_rms_velocity( re_lambda=reynolds_lambda, kinematic_viscosity=kinematic_viscosity, turbulent_dissipation=turbulent_dissipation, )
# Calculate Turbulence Scales taylor_microscale = par.gas.get_taylor_microscale( fluid_rms_velocity=fluid_rms_velocity, kinematic_viscosity=kinematic_viscosity, turbulent_dissipation=turbulent_dissipation, ) eulerian_integral_length = par.gas.get_eulerian_integral_length( fluid_rms_velocity=fluid_rms_velocity, turbulent_dissipation=turbulent_dissipation, ) lagrangian_integral_time = par.gas.get_lagrangian_integral_time( fluid_rms_velocity=fluid_rms_velocity, turbulent_dissipation=turbulent_dissipation, ) normalized_accel_variance = ( par.gas.get_normalized_accel_variance_ao2008( re_lambda=reynolds_lambda ) ) kolmogorov_time = par.gas.get_kolmogorov_time( kinematic_viscosity=kinematic_viscosity, turbulent_dissipation=turbulent_dissipation, ) lagrangian_taylor_microscale_time = ( par.gas.get_lagrangian_taylor_microscale_time( kolmogorov_time=kolmogorov_time, re_lambda=reynolds_lambda, accel_variance=normalized_accel_variance, ) )
z = compute_z(lagrangian_taylor_microscale_time, lagrangian_integral_time) beta = compute_beta(taylor_microscale, eulerian_integral_length)
# Calculate v'² Values vel_corr_terms = VelocityCorrelationTerms( b1=compute_b1(z), b2=compute_b2(z), d1=compute_d1(beta), d2=compute_d2(beta), c1=compute_c1(z, lagrangian_integral_time), c2=compute_c2(z, lagrangian_integral_time), e1=compute_e1(z, eulerian_integral_length), e2=compute_e2(z, eulerian_integral_length), )
return _compute_rms_fluctuation_velocity( fluid_rms_velocity, particle_inertia_time, particle_settling_velocity, vel_corr_terms, )Running the Code
Section titled “Running the Code”Here we run the code for each configuration.
Python · section 6
model_rms_10cm2_s3 = calculate_horizontal_velocity( turbulent_dissipation=10 * par.util.get_unit_conversion("cm^2/s^3", "m^2/s^3"), reynolds_lambda=reynolds_lambda,)model_rms_100cm2_s3 = calculate_horizontal_velocity( turbulent_dissipation=100 * par.util.get_unit_conversion("cm^2/s^3", "m^2/s^3"), reynolds_lambda=reynolds_lambda,)
model_rms_400cm2_s3 = calculate_horizontal_velocity( turbulent_dissipation=400 * par.util.get_unit_conversion("cm^2/s^3", "m^2/s^3"), reynolds_lambda=reynolds_lambda,)Plotting and Graph Comparisons
Section titled “Plotting and Graph Comparisons”The graphs include:
- DNS Data: Raw experimental or simulation data.
- Model Predictions: Analytical or simulated forecasts.
Python · section 7
fig, ax = plt.subplots(figsize=(8, 6))
# Case 1: R_lambda = 72.41, epsilon = 10 cm²/s³ax.scatter( dns_10cm2_s3[:, 0], dns_10cm2_s3[:, 1], label=r"DNS: $R_\lambda=72.41$, $\varepsilon=10$", color="blue", marker="o",)ax.plot( particle_radius * 1e6, model_rms_10cm2_s3 * 1e4, label=r"Model: $R_\lambda=72.41$, $\varepsilon=10$", color="blue",)
# Case 2: R_lambda = 72.41, epsilon = 100 cm²/s³ax.scatter( dns_100_cm2_s3[:, 0], dns_100_cm2_s3[:, 1], label=r"DNS: $R_\lambda=72.41$, $\varepsilon=100$", color="green", marker="^",)ax.plot( particle_radius * 1e6, model_rms_100cm2_s3 * 1e4, label=r"Model: $R_\lambda=72.41$, $\varepsilon=100$", color="green",)
# Case 3: R_lambda = 72.41, epsilon = 400 cm²/s³ax.scatter( dns_400_cm2_s3[:, 0], dns_400_cm2_s3[:, 1], label=r"DNS: $R_\lambda=72.41$, $\varepsilon=400$", color="red", marker="s",)ax.plot( particle_radius * 1e6, model_rms_400cm2_s3 * 1e4, label=r"Model: $R_\lambda=72.41$, $\varepsilon=400$", color="red",)
# Set labels, title, legend, etc.ax.set_xlabel("Particle Radius (µm)")ax.set_ylabel(r"$<(v'_x)^2>$ (cm²/s²)")ax.set_ylim(0, 180)ax.set_xlim(5, 65)ax.set_title("Mean-Square Horizontal Velocity Comparison")ax.legend(loc="upper right")ax.grid(True)plt.subplots_adjust(bottom=0.2)plt.show()
Summary
Section titled “Summary”Overall we have a similar comparison as Ayala et al. (2008) for the horizontal velocity of particles in a turbulent flow. The model predictions align well with the DNS data, confirming the validity of the model.
The curves are very sensitive to the temperature, and you can increase the temperature to move the curves up.
Explore this topic: Cloud-droplet collision kernels