FastBEMT

A fast approximate BEMT (Blade Element Momentum Theory) solver for propeller aerodynamic/aeroacoustic analysis.

Overview

FastBEMT provides a high-performance Python package for analyzing propeller aerodynamics and acoustics using Blade Element Momentum Theory combined with acoustic source models. The package leverages PyTorch for GPU acceleration, enabling efficient computation of:

• Aerodynamic Analysis: Blade Element Momentum Theory (BEMT) calculations with Prandtl loss corrections
• Tonal Noise: Farassat 1A compact source formulation for thickness and loading noise
• Broadband Noise: Brooks-Pope-Marcolini (BPM) model for broadband noise prediction
• Structural Analysis: Blade stress calculations under centrifugal and bending loads

Package Structure

src/FastBEMT

Propeller.py

Propeller data and aeroacoustic workflow. Handles:

• Propeller geometry and environmental parameters
• F1A acoustic source computation (monopole and dipole noise)
• BPM broadband noise prediction in third-octave bands
• Integration of results into output acoustic spectra and time histories

BEMT.py

Complete blade element momentum theory analysis. Handles:

• Operating-point inputs including RPM and freestream velocity
• BEMT aerodynamic solution across blade sections
• Section-level result storage
• Integrated thrust, torque, coefficients, and figure of merit

Section.py

Blade Element Momentum Theory solver for individual propeller sections. Provides:

• Iterative solution of momentum and blade element equations
• Prandtl tip and hub loss factor computations
• Airfoil coefficient interpolation using pre-built aerosandbox.Airfoil objects
• Mach and Reynolds number effects on aerodynamic coefficients
• Local inflow angle and force distribution calculations

BPM.py

Brooks-Pope-Marcolini broadband noise prediction model (PyTorch implementation). Implements five distinct noise sources:

• Turbulent Boundary Layer (TBL): Noise from turbulent pressure fluctuations on blade surfaces including suction-side, pressure-side, and separated-flow components
• Laminar Boundary Layer (LBL): Instability noise from laminar boundary layers at low frequencies
• Trailing Edge Bluntness (TEB): Scattering of incoming vorticity by blunt trailing edges
• Tip Vortex (TV): Noise from unsteady loading fluctuations induced by tip vortex
• Turbulence Ingestion (TI): Interaction of ingested turbulence with blade surfaces

Features include GPU-accelerated Strouhal number and correction factor calculations, and third-octave band spectrum generation.

F1A.py

Farassat 1A acoustic formulation (PyTorch implementation) for rotating sources. Handles:

• Thickness (monopole) source noise from blade volume displacement
• Loading (dipole) source noise from aerodynamic forces
• Compact source approximation for efficient far-field calculation
• GPU-accelerated tensor operations for time-domain pressure computation
• Observer position and blade angle handling

Environment.py

Immutable physical properties:

• Air density, speed of sound, and dynamic viscosity
• Acoustic reference pressure

Simulation.py

Numerical and temporal settings:

• Number of simulated revolutions and samples per revolution
• PyTorch device specification
• RPM-dependent source times and observer time range

DataLoader.py

Utility functions for data input/output:

• Loading propeller geometry dictionaries from pickle files in the Datasets directory
• Repository root detection and path management
• Access to project figure output directories

Plotter.py

Visualization utilities for acoustic analysis results:

• Time-domain pressure histories for monopole, dipole, and total pressure
• Frequency-domain acoustic spectra (Sound Pressure Level)
• Blade passing frequency harmonic indicators
• Overall A-weighted Sound Pressure Level (OASPL) display
• Multi-observer comparison plots

Stress.py

Blade structural analysis tools computing:

• Centrifugal stress distribution along blade span
• Bending stress from thrust and torque loads
• Moment of inertia calculations for arbitrary airfoil sections
• Combined stress field at each blade section

Requirements

• Python ≥ 3.12
• PyTorch (for GPU acceleration)
• NumPy, SciPy (numerical computing)
• AeroSandbox (airfoil aerodynamics)
• Plotly, Matplotlib (visualization)
• scikit-learn (data processing)
• And additional dependencies as specified in pyproject.toml

Installation

uv sync
pip install -e .

BEMT Sweep

from FastBEMT import BEMT, Environment, Propeller, Simulation

environment = Environment(
    a_inf=343.0,
    rho=1.225,
    mu=1.81e-5,
    p_ref=2e-5,
)
simulation = Simulation(
    revolutions=1,
    num_obs_times_per_rev=100,
    device="cpu",
)

propeller = Propeller(
    geometry=geometry,
    environment=environment,
    simulation=simulation,
)
bemt = BEMT(
    propeller=propeller,
    environment=environment,
    rpm=[3000, 7000],
    v_inf=[0.0, 10.0],
)

This evaluates the four-point Cartesian product. Section results are stored in bemt.solution_data with a (rpm, v_inf, section) MultiIndex. Integrated results are stored in bemt.performance with a (rpm, v_inf) MultiIndex. Use bemt.solution_for(7000, 0.0) to select one radial solution.

Alternatively, specify advance ratio J instead of v_inf:

bemt = BEMT(
    propeller=propeller,
    environment=environment,
    rpm=[3000, 7000],
    J=[0.4, 0.8],
)

For each RPM and advance-ratio pair, the freestream velocity is computed as v_inf = J * (rpm / 60) * propeller_diameter.