This readme file was generated on 2024-10-22 by Richard Sheridan ------------------- GENERAL INFORMATION ------------------- Title of Dataset: Data from: ViscoNet: a lightweight FEA surrogate model for polymer nanocomposites viscoelastic response prediction Author Contact Information (Name, Institution, Email, ORCID) Principal Investigator: Cate Brinson Institution: Duke University Email: cate.brinson@duke.edu ORCID: 0000-0003-2551-1563 Co-Investigator: Anqi Lin Institution: Duke University Co-Investigator: Richard Sheridan Institution: Duke University Co-Investigator: Binyin Hu Institution: Duke University *Date of data collection (range): 2021-2023 *Funding and grant numbers (if applicable): NSF-CSSI (OAC-1835677) -------------------- DATA & FILE OVERVIEW -------------------- This is the data for reproducing and training ViscoNet from https://doi.org/10.1016/j.jmps.2024.105915. To train ViscoNet, you only need the contents of the JSON, TSV, and TXT files -- the TSV files contain outputs, and the JSON files map the outputs to the input parameters, including the matching TXT input file. To do image-to-feature learning, you would use the input NPY data of interphase images as an additional input. File list (filenames, directory structure (for zipped files) and brief description of all data files): Each *.tar file is a TAR archive that contains XZ compressed TAR archives. Each archive nested archive should be extracted into a folder without the archive file extensions for use. The total uncompressed size of the dataset is over 1TB, so consider only extracting specific files as needed! result_tsv_tar.xz_files.tar contains output result (*.tsv) data of storage and loss modulus derived from nanocomposite FEA simulations. microstructure_mat_tar.xz_files.tar contains input (*.mat) data of microstructures for nanocomposite FEA simulations without interphase rules applied. intph_img_npy_tar.xz_files contains input (*.npy) data for of microstructures nanocomposite FEA simulations with interphase rules applied. abaqus_inp_tar.xz_files contains ABAQUS input (*.inp) files generated from the interphase images and JSON metadata. These would mainly be useful for reproducing specific simulations if necessary. Text files of matrix master curves data. JSON files that contain experiment metadata. *Relationship between files, if important for context: There are 5 JSON files that summarize the input and output of every single FEA simulation: "all_data.json" for PC "ps_all_data.json" for PS "pmma_all_data.json" for PMMA "ep2_all_data.json" for epoxy "pet_all_data.json" for PET Here is an example of one record in the JSON: "ps_no_intph_no_agglm_4_results/50_20_500_0_0.0_0.1_1_11-youngs.tsv": {"ParRu": "50.00", "ParRv": "20.00", "pix": "500.00", "NumAgl": "0.00", "VfAglm": "0.00", "VfFree": "0.10", "scale": 0.002, "seed": "11.00", "VfAfl": [0], "mtx_density": 1.2e-15, "master_curve": "PS_mc.txt", "mtx_poisson": 0.49, "fil_density": 2.65e-15, "fil_youngs": 73000.0, "fil_poisson": 0.17, "layers": [0], "periodic_intph": true, "intph_density": 1.2e-15, "intph_poisson": 0.49, "intph_shift": 1.0, "intph_l_brd": 1.0, "intph_r_brd": 1.0, "displacement": 0.005, "disp_BC_dof_first": 1, "disp_BC_dof_last": 1, "num_freq": 30, "fmin": 0.0001, "fmax": 1000.0, "ms_filename": "./doe_no_agglm_4/50_20_500_0_0.0_0.1_1_11.mat", "ms_mat_var": "MS", "reverse": true, "force_layer_assign": true, "intph_img": "./ps_no_intph_no_agglm_4/50_20_500_0_0.0_0.1_1_11.npy", "microstructure": "doe_no_agglm_4/50_20_500_0_0.0_0.1_1_11.mat"}​​​ The key "ps_no_intph_no_agglm_4_results/50_20_500_0_0.0_0.1_1_11-youngs.tsv" is the relative path to the output VE response (frequency, E', E''). Everything else is input data to the viscoelastic FEA simulation. The other relative path keyed by "intph_img" or "ms_filename" is the microstructure image, which could be loaded by numpy or matlab. The "master_curve" points to one of the top level text files of matrix master curves data. Some keys refer to files that do not exist in the dataset due to oversights during generation. These are listed in "missing.json", which can be re-generated by check.py. -------------------------- METHODOLOGICAL INFORMATION -------------------------- Description of methods used for collection/generation of data: *Software- or Instrument-specific information needed to interpret the data, including software version numbers, packages or other dependencies: The files are in portable formats that can be viewed with any number of software packages. -------------------------- DATA-SPECIFIC INFORMATION -------------------------- Variable/field list ParRu: Average particle radius in pixel units. ParRv: Standard deviation of particle radius. pix: Microstructure matrix size in pixels (pix*pix). NumAgl: Number of agglomerations in the microstructure. VfAglm: Volume fraction of agglomerations in the microstructure. VfFree: Volume fraction of free particles in the microstructure. VfAfl: Unused. scale: Length scale of a pixel, in micrometers. seed: Random seed for microstructure generation. mtx_density: Matrix density in kg/micrometer^3 mtx_youngs: Matrix long-term Young's modulus in MPa. If not provided, the storage modulus at the lowest frequency in the master curve will be used as the long-term Young's modulus. master_curve: Filename of the master curve for the matrix. The master curve is expected to have a three-column format: frequency, storage, loss. The data should be sorted by frequency. mtx_poisson: matrix long-term Poisson's ratio. fil_density: Filler density in kg/micrometer^3. fil_youngs: Filler Young's modulus in MPa. fil_poisson: Filler Poisson's ratio. layers: A list specifying the number of interphase layers. If an integer is passed, it will automatically be wrapped into a list. periodic_interph: a flag to assign periodic interphase. Set to True to enable periodic interphase conditions. intph_density: Interphase density in kg/micrometer^3. intph_youngs: Interphase Young's modulus in MPa. intph_poisson: Interphase Poisson's ratio. intph_shift: Shifting factor for interphase. intph_l_brd: Left broadening factor for interphase. intph_r_brd: Right broadening factor for interphase. displacement: Distance to stretch the microstructure in micrometers. disp_BC_dof_first: The first degree of freedom to which displacement is applied. (0, 1, 2) disp_BC_dof_last: The last degree of freedom to which displacement is applied. (0, 1, 2 but greater than or equal to first) fmin: The lower limit of simulated frequency range. fmax: The upper limit of simulated frequency range. num_freq: The number of frequency interval (logarithmically spaced). scale: Scale per pixel in nm. ms_filename: Filename of the microstructure data. ms_mat_var: The variable name of the microstructure matrix, if it is in a .mat file. reverse: Set to True if the matrix is represented as 0 and the filler as 1 in the microstructure matrix. force_layer_assign: If set to true, the final microstructure after applying interphase must have its matrix or interphase layer with at least 1 pixel assigned. Value/attribute list Include units of measure, codes or symbols used Missing data treatments (null, -99, na, etc.) ------------------------- USE and ACCESS INFORMATION -------------------------- Data License:Creative Commons CC0 1.0 Universal https://creativecommons.org/publicdomain/zero/1.0/ Other Rights Information: To cite the data: Brinson, C., Lin, A., Sheridan, R. J., & Hu, B. (2024). Data from: ViscoNet: a lightweight FEA surrogate model for polymer nanocomposites viscoelastic response prediction. Duke Research Data Repository. https://doi.org/10.7924/r4g166t5p