Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
59
datasets available to search
ShareScore release 0.9.0
Dataset results
59 results for “Surrogate Model”
GPR surrogate model dataset for remnant black hole properties using perturbation theory and NR
<p>This is the data-set used in <strong><code>BHPTNR_Remnant</code></strong> which is an easy-to-use python package to efficiently predict the remnant mass, remnant spin, peak luminosity and the final kick imparted on the remnant black hole directly from the gravitational radiation using GPR fits. These fits have been built on the remnant data calculated from numerical relativity informed black hole perturbation theory based waveforms.</p>
Dataset: Multidimensional surrogate modelling for Airborne TDEM data
<p>DataHF contains the synthetic data for a SkyTEM 304 TDEM system, flying at 40 m height for a two-layered model with an interface at an angle, described in parameters.csv, via 3D simulations. DataLF contains the 1D data (without angle) with a 1D analytical forward model.</p> <p>Look out for the published PhD dissertation "Improving Airborne Time-Domain Electromagnetic Imaging with Applications to Groundwater Salinity Mapping" for a description of the dataset in Chapter 5.</p>
Simulation data and surrogate model for the DTU 10MW reference wind turbine including down-regulation, power boosting and individual blade control
<p>This contribution provides the simulated data and surrogate models for the DTU 10 MW reference wind turbine in an onshore configuration simulated with FAST v8.16.00. The dimensions include mean wind speed, turbulence intensity, and power level, as well as the application of an individual blade control (IBC) loop. Down-regulation up to 50% is considered using two controller trajectories. The <em>constTSR</em> trajectory considers only pitching for down-regulation, maintaining a constant tip speed ratio, and the <em>lin70</em> trajectory considers both pitch and rotational speed reduction to achieve down-regulation. Power boosting is performed up to 130% power level by following the optimal Cp trajectory until the requested power level is reached.</p> <p>The regression is done with two methods: a spline-based interpolation and a Gaussian Process Regression (GPR). The raw data, smoothened data, and the trained GPR models are provided along with scripts for generating the surrogate model's predictions with both methods. A short description of the simulation parameters and variables considered is given in the supplementary pdf file.</p> <p>The dataset is part of the doctoral thesis 'Wind Turbine Operational Optimization Considering Revenue and Fatigue Objectives' by Vasilis Pettas at the University of Stuttgart (<a href="http://dx.doi.org/10.18419/opus-13959">http://dx.doi.org/10.18419/opus-13959</a>) and the journal publication 'Surrogate Modeling and Aeroelastic Analysis of a Wind Turbine with Down-Regulation, Power Boosting, and IBC Capabilities' <a href="https://doi.org/10.3390/en17061284">(https://doi.org/10.3390/en17061284</a>). Detailed analysis of the controller design and validation of the surrogate models can be found in these publications. </p>
Data of "Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: application to history-dependent homogenized response of heterogeneous materials"
<h1>Development of the Self-Consistency reinforced Minimum Recurrent Unit (SC-MRU)</h1> <p>This directory contains the data and algorithms generated in publication<sup><a href="#fn-1-5292">1</a></sup></p> <h2>Table of Contents</h2> <ol> <li><a href="#dependencies-and-prerequisites">Dependencies and Prerequisites</a></li> <li><a href="#structure-of-repository">Structure of Repository</a></li> <li><a href="#part-1-data-preparation">Part 1: Data preparation</a></li> <li><a href="#part-2-rnn-training">Part 2: RNN training</a></li> <li><a href="#part-3-multiscale-analysis">Part 3: Multiscale analysis</a></li> <li><a href="#part-4-reproduce-paper1-figures">Part 4: Reproduce paper[^1] figures</a></li> </ol> <h2>Dependencies and Prerequisites</h2> <p> </p> <ul> <li> <p>Python, pandas, matplotlib, texttabble and latextable are pre requisites for visualizing and navigating the data.</p> </li> <li> <p>For generating mesh and for vizualization, gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) is required.</p> </li> <li> <p>For running simulations, cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>) is required.</p> </li> </ul> <h3>Instructions using apt & pip3 package manager</h3> <p>Instructions for Debian/Ubuntu based workstations are as follows.</p> <h3>python, pandas and dependencies</h3> <div> <pre><code> sudo apt install python3 python3-scipy libpython3-dev python3-numpy python3-pandas</code></pre> </div> <h3>matplotlib, texttabble and latextable</h3> <div> <pre><code> pip3 install matplotlib texttable latextable</code></pre> </div> <h3>Pytorch (only for run with cm3Libraries)</h3> <ul> <li>Without GPU</li> </ul> <div> <pre><code> pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu</code></pre> </div> <ul> <li>With GPU</li> </ul> <div> <pre><code> pip3 install torch torchvision torchaudio</code></pre> </div> <h3>Libtorch (for compiling the cells)</h3> <ul> <li>Without GPU: In a local directory (e.g. <code>~/local</code> with <code>export TORCHDIR=$HOME/local/libtorch</code>)</li> </ul> <div> <pre><code> wget https://download.pytorch.org/libtorch/cpu/libtorch-shared-with-deps-2.1.1%2Bcpu.zip unzip libtorch-shared-with-deps-2.1.1%2Bcpu.zip</code></pre> </div> <ul> <li>With GPU: In a local directory (e.g. <code>~/local</code> with <code>export TORCHDIR=$HOME/local/libtorch</code>)</li> </ul> <div> <pre><code> wget https://download.pytorch.org/libtorch/cu121/libtorch-shared-with-deps-2.1.1%2Bcu121.zip unzip libtorch-shared-with-deps-2.1.1+cu121.zip</code></pre> </div> <h2>Structure of Repository</h2> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/All_Path_Res">All_Path_Res</a>: results of the direct numerical simulations used as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE">ConstRVE</a>: script to run direct numerical finite element simulations, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale</a>: scripts to run and visualise the multiscale analyses, see details in <a href="#part-3-multiscale-analysis">Part 3: Multiscale analysis</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU</a>: implementation of the RNN and scripts to train them, see details in <a href="#part-2-rnn-training">Part 2: RNN training</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a>: scripts to collect, normalise and truncate the RVEs direct simulation results as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>. The director also contained the storred processed data used in <sup><a href="#fn-1">1</a></sup>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a>: scripts to generate the different loading paths for the direct numerical simulations used as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> </ul> <h2>Part 1: Data preparation</h2> <h3>Generate the loading paths</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/testGenerationData.py">TrainingPaths/testGenerationData.py</a> is used to generate random walk paths, with the options <ul> <li><code>Rmax = 0.11</code> # bound on the final Green Lagrange strain</li> <li><code>TimeStep = 1.</code> # in second</li> <li><code>EvalStep = [1e-4,5e-3]</code> #Bounds on the Green Lagrange increments</li> <li><code>Nmax = 2500</code> #maximum length of the sequence</li> <li><code>k = 4000</code> # number of path to generate</li> <li>The path are storred by default in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>. The path has to be existing before launching the script. You can change the name in line 123 <code>saveDir = '../ConstRVE'+'/Paths/'</code>.</li> <li>Examples of generated paths can be found in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE/PathsExamples">ConstRVE/PathsExamples/</a></li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> </li> </ul> <div> <pre><code>(mkdir ../ConstRVE/Paths) #if needed python3 testGenerationData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/generationData_Cyclic.py">TrainingPaths/generationData_Cyclic.py</a> is used to generate random cylic paths, with the options <ul> <li><code>Rmax = [np.random.uniform(0.,0.04),np.random.uniform(0.,0.06),np.random.uniform(0.0,0.09),0.12]</code> # bound on the final Green Lagrange strain is random</li> <li><code>TimeStep = 1.</code> # in second</li> <li><code>EvalStep = [1e-4,5e-3]</code> #Bounds on the Green Lagrange increments</li> <li><code>Nmax = 2500</code> #maximum length of the sequence</li> <li><code>k = 2000</code> # number of path to generate</li> <li>The path are stored by default in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>. You can change the name in line 123 <code>saveDir = '../ConstRVE'+'/Paths/'</code>.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> </li> </ul> <div> <pre><code>(mkdir ../ConstRVE/Paths) #if needed python3 generationData_Cyclic.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/countPathLength.py">TrainingPaths/countPathLength.py</a> gives average, minimum and maximum lengths of the generated paths and the distribution of the <code>\Delta R</code>. By default the paths are read in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a> but the directory can be given as an argument. The file can be used to read <ul> <li>either the generated loading paths</li> </ul> </li> </ul> <div> <pre><code> python3 countPathLength.py '../ConstRVE/PathsExamples'</code></pre> </div> <ul> <li> <ul> <li>or the results of the <a href="#generate-the-rves-direct-simulation-results">simulations</a></li> </ul> </li> </ul> <div> <pre><code> python3 countPathLength.py '../All_Path_Res/Path_Res9'</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/graphData.py">TrainingPaths/graphData.py</a> generates illustrations from randomly picked paths in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a> and generate png figures.</li> </ul> <h3>Generate the RVEs direct simulation results</h3> <ul> <li>Uses the loading paths existing in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/rve.geo">ConstRVE/rve.geo</a> is the RVE geometry file that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/rve.msh">ConstRVE/rve.msh</a> is the RVE mesh file that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/utilsFunc.py">ConstRVE/utilsFunc.py</a> contains python tools to be used.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Rve_withoutInternalVars.py">ConstRVE/Rve_withoutInternalVars.py</a> is used to run all the RVE simulations: <ul> <li>This requires cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>).</li> <li>All the ouptus are stored in <code>All_Path_Res/Path_Res12</code>, you can change the name in line 71 <code>Path_Res = '../All_Path_Res/Path_Res12/'</code>. The results of RVE simulations are saved as the sequence (one configuration per line) of the Green-Lagrange strains and Second Piola-Kirchhoff stress (in column). One example can be found in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_Res1/data_path1000.csv">All_Path_Res/Path_Res1/data_path1000.csv</a>.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE">ConstRVE</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Rve_withoutInternalVars.py</code></pre> </div> <h3>Collect, normalised and truncate the RVEs direct simulation results as training and testing data</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/CheckNanData.py">TrainingData/CheckNanData.py</a> is used to check the integrity of the direct numerical simulations results <ul> <li>DNS results are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res1">All_Path_Res/Path_res1</a> to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res11">All_Path_Res/Path_res11</a> subdirectories.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 CheckNanData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/CollectData.py">TrainingData/CollectData.py</a> is used to gather all the direct numerical simulations results <ul> <li>DNS results are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res1">All_Path_Res/Path_res1</a> to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res11">All_Path_Res/Path_res11</a> subdirectories. It can be changed in line 31 <code>for ll in range(11):</code>.</li> <li>It saves the bounds and raw data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Bounds_GS">TrainingData/Processed_Data/Bounds_GS</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Origin_GS">TrainingData/Processed_Data/Origin_GS</a>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 CollectData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Normalization.py">TrainingData/Normalization.py</a> is used to normalise the gathered direct numerical simulations results <ul> <li>Bounds and raw data are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Bounds_GS">TrainingData/Processed_Data/Bounds_GS</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Origin_GS">TrainingData/Processed_Data/Origin_GS</a>, respectively.</li> <li>It saves the normalized training (75%) and testing (25%) data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Train">TrainingData/Processed_Data/Normalized_GS_Train</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Test">TrainingData/Processed_Data/Normalized_GS_Test</a>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Normalization.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/GaussCollectData.py">TrainingData/GaussCollectData.py</a> is an alternative using Gaussian normaliation and is used to gather all the direct numerical simulations results.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/GaussNormalization.py">TrainingData/GaussNormalization.py</a> is an alternative to normalise following a Gaussian the gathered direct numerical simulations results.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Data_Padding.py">TrainingData/Data_Padding.py</a> is used to pad and trim the normalised data <ul> <li>Normalized training (75%) and testing (25%) data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Train">TrainingData/Processed_Data/Normalized_GS_Train</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Test">TrainingData/Processed_Data/Normalized_GS_Test</a>, respectively.</li> <li>The final length of the sequence (including zero padding and trimming) is given by <code>N = 200</code>.</li> <li>It saves the trimmed and padded normalised training and testing data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data">TrainingData/Processed_Data/</a> as <code>'GS_Train'N</code> and <code>'GS_Test'N</code>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Data_Padding.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Tool.py">TrainingData/Tool.py</a> is a list of function used to normalise dat.</li> </ul> <h2>Part 2: RNN training</h2> <h3>Available rnn</h3> <ul> <li>The different cells are in the following directories <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a>: Neural network with SMRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SCMRU_T">SC_MRU/MGRU/NNW_SCMRU_T</a>: Neural network with SC-MRU-T recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_MGRU">SC_MRU/MGRU/NNW_MGRU</a>: Neural network with orginal MGRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_MGRU_M">SC_MRU/MGRU/NNW_MGRU_M</a>: Neural network with modified MGRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/FNN_LeakyReLU">SC_MRU/FNN_LeakyReLU/NNW_<code>X</code>Fw</a>: Neural network with SC-MRU-I recurrent cell using <code>X</code> Feed Forward non-linear transition layers.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/Quadratic_NLT">SC_MRU/Quandratic_NLT/NNW_<code>X</code>Q, NNW_Q_Fw, NNW_Fw_Q</a>: Neural network with SC-MRU-I recurrent cell using <code>X</code> quadratic or mixed feed-forward-quadratic non-linear transition layers.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/ReferenceRNN">SC_MRU/ReferenceRNN/NNW_<code>X</code>Layers</a>: Neural network with SC-LMSC recurrent cell using <code>X</code> non-linear transition layers.</li> </ul> </li> </ul> <h3>Compile the neural network models</h3> <ul> <li>In the adequate directory, e.g. <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a> for the Neural network with SMRU recurrent cell.</li> </ul> <div> <pre><code> cd build rm -rf * cmake -DCMAKE_PREFIX_PATH=$TORCHDIR .. make</code></pre> </div> <ul> <li>This create the RNN_<code>CELL</code> model in the build directory</li> </ul> <h3>Train the neural network models</h3> <ul> <li>In the adequate directory, e.g. <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a> for the Neural network with SMRU recurrent cell.</li> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN model.</li> <li>The file <code>Train.py</code> <ul> <li>Uses the RNN_<code>CELL</code> model compiled in the <code>build</code> directory.</li> <li>Uses the trimmed and padded normalised training and testing data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data">TrainingData/Processed_Data/</a> as <code>'GS_Train'N</code> and <code>'GS_Test'N</code>, respectively.</li> <li>Can be modified to use the requested ratios of training and testing sequences of different lengths to prepate the mini-baches, e.g.: <ul> <li><code>ratio = [0.2,0.02,0.6,0.06,0.01,0.9]</code></li> <li><code>PathIn1 = ['../../../TrainingData/Processed_data/GS_Train200','../../../TrainingData/Processed_data/GS_Train2500','../../../TrainingData/Processed_data/GS_Test200','../../../TrainingData/Processed_data/GS_Test2500']</code></li> <li><code>PathIn2 = ['../../../TrainingData/Processed_data/GS_Test200','../../../TrainingData/Processed_data/GS_Test400','../../../TrainingData/Processed_data/GS_Test2500','../../../TrainingData/Processed_data/GS_Test400']</code></li> </ul> </li> <li>Saves the mini-batches in <ul> <li><code>PathOut = "TrainingData.pt"</code></li> </ul> </li> <li>Saves the model <code>module-checkpoint.pt</code> and <code>module-checkpoint-optimizer.pt</code>, and loss evolution <code>Loss.txt</code> in <ul> <li>Module/H<code>n</code> with <code>n</code> the number of hiden variables.</li> <li>Output directory can be changes in NNW_<code>CELL</code>.cpp of the cell name <code>CELL</code> (and recompiling).</li> <li>Warm start can be disabled in by commenting <code>torch::load(net, "./Module/H120/module-checkpoint_CM0.pt");</code> and <code>torch::load(optimizer, "./Module/H120/module-optimizer-checkpoint_CM0.pt");</code> in NNW_<code>CELL</code>.cpp of the cell name <code>CELL</code> (and recompiling).</li> <li>Is executed with</li> </ul> </li> </ul> </li> </ul> <div> <pre><code> python3 Train.py</code></pre> </div> <h3>Convert trained c++ models for pyTorch (in view of multiscale simulations)</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Save_RU.py">SC_MRU/CheckLoss/RU.py</a>: functions used to read and use the rnn models.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Save_RNN_script.py">SC_MRU/CheckLoss/Save_RNN_script.py</a>: <ul> <li>Converts c++ trained models to pyTorch models</li> <li>Reads the trained models in SC_MRU/CELL_<code>Kind</code>/NNW_<code>CELL</code>/Module/H<code>N</code>/module.pt</li> <li>Save the converted models to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> and to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model/</a></li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 Save_RNN_script.py</code></pre> </div> <h3>Vizualize loss evolution and testing results</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Data_prepare.py">SC_MRU/CheckLoss/Data_prepare.py</a> contains fucntion used for testing.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/LossVS_insertN.py">SC_MRU/CheckLoss/LossVS_insertN.py</a> tests the effect of testing data augmentation: <ul> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>repeat = 10</code> #number of tests</li> <li><code>dataType ='random'</code> #'even' or 'random'</li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> <li><code>reevaluate= False</code> # <code>False</code> to use the saved loss values and <code>True</code> to evaluate the loss values</li> <li><code>fastshifting=False</code> # <code>False</code> to vizualize before fast shifting (Fig. 11) and <code>True</code> after (Fig. 12). When <code>reevaluate== True</code>, this has no effect: the training with fast-shifting or not has to be done manually, see <a href="#train-the-neural-network-models">details</a>.</li> </ul> </li> <li>Generates new Loss_<code>CELL</code>.txt and TrainingData.pt files in case <code>reevaluate== True</code>, read the existing one if <code>reevaluate== False</code>.</li> <li>Generates Figs. 11 and 12 from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> </ul> </li> </ul> <div> <pre><code> python3 LossVS_insertN.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Plot_GS.py">SC_MRU/CheckLoss/Plot_GS.py</a> test the different NNW on RVE testing paths: <ul> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> </ul> </li> <li>Generate Figs. 13, 14 and 15 from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 Plot_GS.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Plot_GS_step.py">SC_MRU/CheckLoss/Plot_GS_step.py</a> tests the different pyTorch NNWs on the RVE testing paths: <ul> <li>Requires to have <a href="#convert-trained-c-models-for-pytorch-in-view-of-multiscale-simulations">converted</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> </ul> </li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 Plot_GS_step.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_HiddenV.py">SC_MRU/PlotLoss_HiddenV.py</a> shows the loss evolution for the different number of hidden variables --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for the SMRU cell: <ul> <li>Generates Fig. 8.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_HiddenV.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_3MRU.py">SC_MRU/PlotLoss_3MRU.py</a> shows the loss evolution for the different recurrent units --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for the 120 hidden variables: <ul> <li>Generates Fig. 9.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_3MRU.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_RefNL_H120.py">SC_MRU/PlotLoss_RefNL_H120.py</a> shows the loss evolution for the different SC-LMSC and SC-MRU-I cells --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for 120 hidden variables: <ul> <li>Generates Fig. 10.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_RefNL_H120.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_MGRU.py">SC_MRU/PlotLoss_MGRU.py</a> shows the loss evolution for the original and modified MGRU --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. A.18.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_MGRU.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_NL.py">SC_MRU/PlotLoss_NL.py</a> shows the loss evolution for the different non-liner transition layers (quadratic and Leaky ReLU) of the SC-MRU-I cell --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. B.19.</li> <li><code>Case=1</code> # 0 for quadratic transition blocks and 1 for Leaky ReLU transition blocks</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_NL.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_DiffNL_H120.py">SC_MRU/PlotLoss_DiffNL_H120.py</a> shows the loss evolution for the different non-liner transition layers (hybrid) of the SC-MRU-I cell --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. B.20.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <h2>Part 3: Multiscale analysis</h2> <h3>Trained surrogate models</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model</a>: contains the different rnn <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/Bounds_GS">MultiScale/Model/Bounds_GS</a>: Bounds.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/DInpFullModel.pt">MultiScale/Model/DInpFullModel.pt</a>: trained rnn with <code>SC-MRU-T</code> recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/IncrementModel.pt">MultiScale/Model/IncrementModel.pt</a>: trained rnn with <code>SC-MRU-I</code> recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/FullModel.pt">MultiScale/Model/FullModel.pt</a>: trained rnn with <code>SMRU</code> recurrent cell.</li> </ul> </li> </ul> <h3>Run mutiscale simulations using the surrogates</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.geo">MultiScale/2D_MultiScale/model.geo</a>: geometry of the macro-scale model that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.msh">MultiScale/2D_MultiScale/model.msh</a>: mesh of the macro-scale model that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.py">MultiScale/2D_MultiScale/model.py</a>: is used to run all the multiscale simulation: <ul> <li>This requires cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>).</li> <li>Uses the bounds and trained models in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model</a>.</li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li><code>factorStep= 100</code> # is the number of steps x 20 during the reloading stage between points B and C.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 model.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2">MultiScale/FE2</a>: <ul> <li>Contains the reference displacement-force results of the FE2 simulation.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a>: contains the macro-scale displacement and stress fields. They can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_<code>STEPS</code></a>: <ul> <li>Contain the reference displacement-force results of the rnn-based multiscale simulations for the different recurrent cells <code>CELL</code> and steps number <code>STEPS</code> during the reloading stage between points B and C.</li> <li>For the cases <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, the macro-scale displacement and stress fields are also available and can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> </ul> <h3>Vizualize mutiscale simulations results</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/plot_force.py">MultiScale/2D_MultiScale/plot_force.py</a>: is used to plot the multiscale simulations results: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li>The multiscale simulations results to be plotted are saved in the directories <code>CELL_120_Step</code>, where <code>Step</code> # is the number of steps during the reloading stage between points B and C.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 plot_force.py</code></pre> </div> <ul> <li>To vizualize the macro-scale displacement and stress fields distributions: <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a>: contains the macro-scale displacement and stress fields. They can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, the macro-scale displacement and stress fields are also available and can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> </ul> <h2>Part 4: Reproduce paper<sup><a href="#fn-1">1</a></sup> figures</h2> <ul> <li>Fig. 7: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> <div> <pre><code> python3 countPathLength.py '../ConstRVE/PathsExamples'</code></pre> </div> <ul> <li>Fig. 9: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_HiddenV.py</code></pre> </div> <ul> <li>Fig. 10: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_3MRU.py</code></pre> </div> <ul> <li>Fig. 11: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <ul> <li>Figs. 12 and 13: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 LossVS_insertN.py</code></pre> </div> <ul> <li>Figs. 14, 15 and 16: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> </ul> <div> <pre><code> python3 Plot_GS.py</code></pre> </div> <ul> <li>Figs. 17(b)(c)(d): The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is, see <a href="#vizualize-mutiscale-simulations-results">details</a>:</li> </ul> <div> <pre><code>python3 plot_force.py</code></pre> </div> <ul> <li>Figs. 18-23: Need gmsh to vizualize the results stored in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, see <a href="#vizualize-mutiscale-simulations-results">details</a></li> <li>Fig. A.24: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_MGRU.py.py</code></pre> </div> <ul> <li>Fig. B.25: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_NL.py</code></pre> </div> <ul> <li>Fig. B.26: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <h2>Disclaimer</h2> <p>This project has received funding from the European Union’s Horizon Europe Framework Programme under grant agreement No. 101056682 for the project “DIgital DEsign strategies to certify and mAnufacture Robust cOmposite sTructures (DIDEAROT)”. The contents of this publication are the sole responsibility of ULiege and do not necessarily reflect the opinion of the European Union. Neither the European Union nor the granting authority can be held responsible for them.</p> <ol> <li> <p>The work is described in:<br>"<em>Wu, L. and Noels, L. (2024).</em> <strong>Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: application to history-dependent homogenized response of heterogeneous materials</strong> 424: 116881, <a href="https://doi.org/10.1016/j.cma.2024.116881" target="_blank" rel="nofollow noreferrer noopener">doi: 10.1016/j.cma.2024.116881</a>" which can be downloaded. We would be grateful if you could cite this publication in case you use the files. <a href="#fnref-1-5292">↩</a> <a href="#fnref-1-2">↩<sup>2</sup></a> <a href="#fnref-1-3">↩<sup>3</sup></a></p> </li> </ol>
dataset for Machine Learning Surrogates for Surface Complexation Model of Uranium Sorption to Oxides
<p>The data sets are used to train, validate, and test SCM surrogates for uranium sorption on oxide surfaces. </p> <p>train_norm.csv, val_norm.csv, and test_norm.csv are used to train, validate, and evaluate DNN-SCM surrogate. </p> <p>test_set1.csv and test_set2.csv are two additional datasets to evaluate the model's performance. </p>
4D-Var data assimilation experiment of the Lorenz 96 model using an adjoint model of a neural network surrogate model
<p>These data are the output of the 4D-Var data assimilation experiment of the Lorenz96 model using an adjoint model of a neural network surrogate model.<br> The details are described in Nishizawa (2022).<br> </p>
Surrogate-modelling & machine learning dataset : finite element stress analysis of biaxial specimen with random elastic properties - 1000 samples
<p>Dataset finite element stress analysis of biaxial specimen with random elastic properties</p> <p>Unzip and execute dataset.py to visualise data samples. PyVista is needed.</p>
Surrogate-modelling & machine learning dataset : finite element stress analysis of biaxial specimen with random elastic properties - 100 samples
<p>Dataset finite element stress analysis of biaxial specimen with random elastic properties</p> <p>Unzip and execute dataset.py to visualise data samples. PyVista is needed</p>
Data-driven surrogate model for wind turbine damage equivalent load
<p>There are four zip files in this data set:</p> <ul> <li>PythonCode_OpenFAST: The code used to generate 32768 OpenFAST fst files to build the database.</li> <li>ML_TrainingCode: The code that used to train the TCN-FCNN and FCNN models for both free stream and wake</li> <li>Trained_Models: All the trained models are saved in Keras format. The models with max in their filenames were trained on maximum values. The models with XY in their naming were trained on wind in the X and Y directions.</li> <li>data: It includes all the CSV files for training and testing.</li> </ul>
Surrogate models to assess the induction of arrhythmias during virtual programmed ventricular stimulation
<p>These videos show a simulation of the Programmed Ventricular Stimulation (PVS). The first video represents the entire mesh simulation, and the second video the surrogate model.</p>
Surrogate Model Optimisation of a 'micro core' PWR fuel assembly arrangement using deep learning models - Figures
<p>Figures for Physior 2020 paper</p>
Limitations of using surrogates for behaviour classification of accelerometer data: refining methods using random forest models in Caprids
<p>Animal-attached devices can be used on cryptic species to measure their movement and behaviour, enabling unprecedented insights into fundamental aspects of animal ecology and behaviour. However, direct observations of subjects are often still necessary to translate biologging data accurately into meaningful behaviours. As many elusive species cannot easily be observed in the wild, captive or domestic surrogates are typically used to calibrate data from devices. However, the utility of this approach remains equivocal. </p> <p>Here, we assess the validity of using captive conspecifics, and phylogenetically-similar domesticated counterparts (surrogate species) for calibrating behaviour classification. Tri-axial accelerometers and tri-axial magnetometers were used with behavioural observations to build random forest models to predict the behaviours. We applied these methods using captive Alpine ibex (Capra ibex) and a domestic counterpart, pygmy goats (Capra aegagrus hircus), to predict the behaviour including terrain slope for locomotion behaviours of captive Alpine ibex. </p> <p>Behavioural classification of captive Alpine ibex and domestic pygmy goats was highly accurate (> 98%). Model performance was reduced when using data split per individual, i.e., classifying behaviour of individuals not used to train models (mean ± sd = 56.1 ± 11%). Behavioural classifications using domestic counterparts, i.e., pygmy goat observations to predict ibex behaviour, however, were not sufficient to predict all behaviours of a phylogenetically similar species accurately (> 55%).</p> <p>We demonstrate methods to refine the use of random forest models to classify behaviours of both captive and free-living animal species. We suggest there are two main reasons for reduced accuracy when using a domestic counterpart to predict the behaviour of a wild species in captivity; domestication leading to morphological differences and the terrain of the environment in which the animals were observed. We also identify limitations when behaviour is predicted in individuals that are not used to train models. Our results demonstrate that biologging device calibration needs to be conducted using: (i) with similar conspecifics, and (ii) in an area where they can perform behaviours on terrain that reflects that of species in the wild.</p>
SIMCor - Demo of fast-to-evaluate surrogate models used for device implantation
<p>This video illustrates a fast to evaluate model for TAVI device implantation based on isogeometric analysis. The tissue is modelled in a rigid manner, meaning that no deformation can be caused by the device. The device deployment is modelled using an expanding cylinder and assessing the contacts between device and tissue. This approach allows rapid evaluation of the risk of paravalvular leakage for different device orientations.</p>
Dataset for the SIAM MPI23 project "Model inversion for complex physical systems using low-dimensional surrogates"
<p>This dataset contains 20,000 synthetic simulations of a simplified two-dimensional confined aquifer model of the Hanford Site. The inputs are the Kosambi-Karhunen-Loève expansion (KKLE) coefficients of the input log-transmissivity field. The outputs are the corresponding stationary pressure responses observed at 323 observation wells. Also included are the arrays necessary to reconstruct the log-transmissivity inputs from the KKLE coefficients.</p>
Real Time 3D Imaging and Surrogate Bone Model
ClinicalTrials.gov study NCT02204007. IPD Sharing: NO. Countries: 1. Publications: 15.
Limitations of using surrogates for behaviour classification of accelerometer data: refining methods using random forest models in Caprids
Open the record for dataset details and reuse information.
SMART FUS: Surrogate model of attenuation and refraction in transcranial focused ultrasound
Open the record for dataset details and reuse information.
Input Data for A Fast Surrogate Model for 3D-Earth Glacial Isostatic Adjustment using Tensorflow (v2.8.0) Artificial Neural Networks
<p>Training datasets for the manuscript A Fast Surrogate Model for 3D-Earth Glacial Isostatic Adjustment using Tensorflow (v2.8.0) Artificial Neural Networks. Two separate datasets are contained for training the ANNs: the 3D-spherically-symmetric (SS) rate-of-change of relative sea level (ROCRSL) and the 3D-SS rate of change of radial displacement (ROCRAD) as a function of SS profiles. Two other datasets contain RSL projections from the explicit (i.e. Seakon 3D - Seakon SS + NMSS ) model and the NMSS model, labelled Seakon_plus_NMSS_RSL and NMSS respectively.</p> <p>Filenames denote the structure of the SS profile: </p> <p>???_?.??_??.*.csv = LT_UMV_LMV.*.{csv,nc}<br> </p> <p>LT = elastic lithosphere thickness (km)</p> <p>UMV = upper mantle viscosity (1E21 Pa s)</p> <p>LMV = lower mantle viscosity (1E21 Pa s)</p> <p>i.e. 96_0.5_10.seakon_S40RTS_lr18-SS.rrad.roc.r360x180.P5.density_wSSRRADROC.csv.bz2 has the SS profile</p> <p>96km elastic lithosphere, 0.5E21 Pa s upper mantle viscosity, 10E21 Pa s lower mantle viscosity</p> <p> </p> <p>The columns of the input files are as follows:</p> <p>LT, UMV, LMV, longitude, latitude, time(t=0), ice(t=0), SS_ROC_RSL (t=0), time(t=-1), ice(t=-1), time(t=-2), ice(t=-2), time(t=-3), ice(t=-3), time(t=-4), ice(t=-4), 3D-SS_ROC_RSL(t=0)</p> <p>units for the above are as follows:</p> <p>km, 1E21 Pas, 1E2 Pas, degrees east (0->360), degrees (-180->180), days since 2000, m, mm/year, days since 2000, m, days since 2000, m, days since 2000, m, days since 2000, m, mm/year</p> <p>where 'days since 2000' assumes exactly 365.25 days per year.</p>
Uncertainty quantification in cerebral circulation simulations focusing on the collateral flow: Surrogate model approach with machine learning
<p>Data and code underlying the findings reported in the paper titled "Uncertainty quantification in cerebral circulation simulations focusing on the collateral flow: Surrogate model approach with machine learning."</p>
Surrogate modelling for the forecast of Seveso-type atmospheric pollutant dispersion
<p>Online resource 1 - Test-data response for GIM model.</p> <p>Online resource 2 - Test-data response for RGI model.</p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.