Code for sound field predictions in domains with impedance boundaries. Used for generating results from the paper

Last update: Dec 17, 2022

Overview

Authors:

Nikolas Borrel ([email protected])

Allan P. Engsig-Karup ([email protected])

Cheol-Ho Jeong ([email protected])

Code for sound field predictions in domains with Neumann and impedance boundaries. Used for generating results from the paper "Physics-informed neural networks for 1D sound field predictions with parameterized sources and impedance boundaries" by N. Borrel-Jensen, A. P. Engsig-Karup, and C. Jeong.

Run

Train

Run

python3 main_train.py --path_settings="path/to/script.json"

Scripts for setting up models with Neumann, frequency-independent and dependent boundaries can be found in scripts/settings (see JSON settings).

Evaluate

Run

python3 main_evaluate.py

The settings are

do_animations = 
do_side_by_side_plot =

">
id_dir = <unique id>
settings_filename = 'settings.json'
base_dir = "path/to/base/dir"

do_plots_for_paper = <bool>
do_animations = <bool>
do_side_by_side_plot = <bool>

The id_dir corresponds to the output directory generated after training, settings_filename is the name of the settings file used for training (located inside the id_dir directory), base_dir is the path to the base directory (see Input/output directory structure).

Evaluate model execution time

To evaluate the execution time of the surrogate model, run

python3 main_evaluate_timings.py --path_settings="path/to/script.json" --trained_model_tag="trained-model-dir"

The trained_model_tag is the directory with the trained model weights trained using the scripts located at the path given in path_settings.

Settings

Input/output directory structure

The input data should be located in a specific relative directory structure as (data used for the paper can be downloaded here)

base_path/
    trained_models/
        trained_model_tag/
            checkpoint
            cp.ckpt.data-00000-of-00001
            cp.ckpt.index
    training_data/
        freq_dep_1D_2000.00Hz_sigma0.2_c1_d0.02_srcs3.hdf5
        ...
        freq_indep_1D_2000.00Hz_sigma0.2_c1_xi5.83_srcs3.hdf5
        ...
        neumann_1D_2000.00Hz_sigma0.2_c1_srcs3.hdf5
        ...

The reference data are located inside the training_data/ directory generated, where the data for impedance boundaries are generated using our SEM simulator, and for Neumann boundaries, the Python script main_generate_analytical_data.py was used.

Output result data are located inside the results folder

base_path/
    results/
        id_folder/
            figs/
            models/
                LossType.PINN/
                    checkpoint
                    cp.ckpt.data-00000-of-00001
                    cp.ckpt.index
            settings.json

The settings.json file is identical to the settings file used for training indicated by the --path_settings argument. The directory LossType.PINN contains the trained model weights.

JSON settings

The script scripts/settings/neumann.json was used for training the Neumann model from the paper

{
    "id": "neumann_srcs3_sine_3_256_7sources_loss02",
    "base_dir": "../data/pinn",
    
    "c": 1,
    "c_phys": 343,
    "___COMMENT_fmax___": "2000Hz*c/343 = 5.8309 for c=1, =23.3236 for c=4",
    "fmax": 5.8309,

    "tmax": 4,
    "xmin": -1,
    "xmax": 1,
    "source_pos": [-0.3,-0.2,-0.1,0.0,0.1,0.2,0.3],
    
    "sigma0": 0.2,
    "rho": 1.2,
    "ppw": 5,

    "epochs": 25000,
    "stop_loss_value": 0.0002,
    
    "boundary_type": "NEUMANN",
    "data_filename": "neumann_1D_2000.00Hz_sigma0.2_c1_srcs7.hdf5",
    
    "batch_size": 512,
    "learning_rate": 0.0001,
    "optimizer": "adam",

    "__comment0__": "NN setting for the PDE",
    "activation": "sin",
    "num_layers": 3,
    "num_neurons": 256,

    "ic_points_distr": 0.25,
    "bc_points_distr": 0.45,

    "loss_penalties": {
        "pde":1,
        "ic":20,
        "bc":1
    },

    "verbose_out": false,
    "show_plots": false
}

The script scripts/settings/freq_indep.json was used for training the Neumann model from the paper

{
    "id": "freq_indep_sine_3_256_7sources_loss02",
    "base_dir": "../data/pinn",

    "c": 1,
    "c_phys": 343,
    "___COMMENT_fmax___": "2000Hz*c/343 = 5.8309 for c=1, =23.3236 for c=4",
    "fmax": 5.8309,

    "tmax": 4,
    "xmin": -1,
    "xmax": 1,
    "source_pos": [-0.3,-0.2,-0.1,0.0,0.1,0.2,0.3],
    
    "sigma0": 0.2,
    "rho": 1.2,
    "ppw": 5,

    "epochs": 25000,
    "stop_loss_value": 0.0002,
    
    "batch_size": 512,
    "learning_rate": 0.0001,
    "optimizer": "adam",

    "boundary_type": "IMPEDANCE_FREQ_INDEP",
    "data_filename": "freq_indep_1D_2000.00Hz_sigma0.2_c1_xi5.83_srcs7.hdf5",

    "__comment0__": "NN setting for the PDE",
    "activation": "sin",
    "num_layers": 3,
    "num_neurons": 256,

    "impedance_data": {
        "__comment1__": "xi is the acoustic impedance ONLY for freq. indep. boundaries",
        "xi": 5.83
    },

    "ic_points_distr": 0.25,
    "bc_points_distr": 0.45,
    
    "loss_penalties": {
        "pde":1,
        "ic":20,
        "bc":1
    },

    "verbose_out": false,
    "show_plots": false
}

The script scripts/settings/freq_dep.json was used for training the Neumann model from the paper

{
    "id": "freq_dep_sine_3_256_7sources_d01",
    "base_dir": "../data/pinn",

    "c": 1,
    "c_phys": 343,
    "___COMMENT_fmax___": "2000Hz*c/343 = 5.8309 for c=1, =23.3236 for c=4",
    "fmax": 5.8309,

    "tmax": 4,
    "xmin": -1,
    "xmax": 1,
    "source_pos": [-0.3,-0.2,-0.1,0.0,0.1,0.2,0.3],
    
    "sigma0": 0.2,
    "rho": 1.2,
    "ppw": 5,

    "epochs": 50000,
    "stop_loss_value": 0.0002,

    "do_transfer_learning": false,

    "boundary_type": "IMPEDANCE_FREQ_DEP",
    "data_filename": "freq_dep_1D_2000.00Hz_sigma0.2_c1_d0.10_srcs7.hdf5",
    
    "batch_size": 512,
    "learning_rate": 0.0001,
    "optimizer": "adam",

    "__comment0__": "NN setting for the PDE",
    "activation": "sin",
    "num_layers": 3,
    "num_neurons": 256,

    "__comment1__": "NN setting for the auxillary differential ODE",
    "activation_ade": "tanh",
    "num_layers_ade": 3,
    "num_neurons_ade": 20,

    "impedance_data": {
        "d": 0.1,
        "type": "IMPEDANCE_FREQ_DEP",
        "lambdas": [7.1109025021758407,205.64002739443146],
        "alpha": [6.1969460587749818],
        "beta": [-15.797795759219973],
        "Yinf": 0.76935257750377573,
        "A": [-7.7594660571346719,0.0096108036858666163],
        "B": [-0.016951521199665469],
        "C": [-2.4690553703530442]
      },

    "accumulator_factors": [10.26, 261.37, 45.88, 21.99],

    "ic_points_distr": 0.25,
    "bc_points_distr": 0.45,

    "loss_penalties": {
        "pde":1,
        "ic":20,
        "bc":1,
        "ade":[10,10,10,10]
    },

    "verbose_out": false,
    "show_plots": false
}

HPC (DTU)

The scripts for training the models on the GPULAB clusters at DTU are located at scripts/settings/run_*.sh.

VSCode

Launch scripts for VS Code are located inside .vscode and running the settings script local_train.json in debug mode is done selecting the Python: TRAIN scheme (open pinn-acoustics.code-workspace to enable the workspace).

License

See LICENSE

Code for sound field predictions in domains with impedance boundaries. Used for generating results from the paper

Related tags

Overview

Run

Train

Evaluate

Evaluate model execution time

Settings

Input/output directory structure

JSON settings

HPC (DTU)

VSCode

License

Owner

DTU Acoustic Technology Group

A super lightweight Lagrangian model for calculating millions of trajectories using ERA5 data

Hypersearch weight debugging and losses tutorial

DNA-RECON { Automatic Web Reconnaissance Tool }

A modular domain adaptation library written in PyTorch.

Improving Compound Activity Classification via Deep Transfer and Representation Learning

QuakeLabeler is a Python package to create and manage your seismic training data, processes, and visualization in a single place — so you can focus on building the next big thing.

Official PyTorch implementation of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image", ICCV 2019

Some bravo or inspiring research works on the topic of curriculum learning.

A simple Rock-Paper-Scissors game using CV in python

Script that attempts to force M1 macs into RGB mode when used with monitors that are defaulting to YPbPr.

PyTorch Implementation of Spatially Consistent Representation Learning(SCRL)

Codes and models for the paper "Learning Unknown from Correlations: Graph Neural Network for Inter-novel-protein Interaction Prediction".

Fully Automatic Page Turning on Real Scores

EqGAN - Improving GAN Equilibrium by Raising Spatial Awareness

This app is a simple example of using Strealit to create a financial data web app.

Interactive web apps created using geemap and streamlit

SpanNER: Named EntityRe-/Recognition as Span Prediction

Source Code for ICSE 2022 Paper - ``Can We Achieve Fairness Using Semi-Supervised Learning?''

GULAG: GUessing LAnGuages with neural networks

Conflict-aware Inference of Python Compatible Runtime Environments with Domain Knowledge Graph, ICSE 2022