Data and code underlying the publication: Transfer learning for improved generalizability in causal physics-informed neural networks for beam simulations

DOI:10.4121/724d8188-2434-49a5-bdb3-2391667290bd.v1
The DOI displayed above is for this specific version of this dataset, which is currently the latest. Newer versions may be published in the future. For a link that will always point to the latest version, please use
DOI: 10.4121/724d8188-2434-49a5-bdb3-2391667290bd
Datacite citation style:
Kapoor, Taniya (2024): Data and code underlying the publication: Transfer learning for improved generalizability in causal physics-informed neural networks for beam simulations. Version 1. 4TU.ResearchData. dataset. https://doi.org/10.4121/724d8188-2434-49a5-bdb3-2391667290bd.v1
Other citation styles (APA, Harvard, MLA, Vancouver, Chicago, IEEE) available at Datacite

Dataset

### Research Objective


The main objective of the research is to improve the generalizability of causal physics-informed neural networks (PINNs) for simulating the dynamics of beams on elastic foundations. This is achieved by integrating transfer learning into the PINN framework to address the limitations of conventional PINNs, particularly in simulation for large space-time domains and varying initial conditions.


### Type of Research


The research is an applied study focused on the development and validation of advanced computational methods in structural engineering. It combines elements of theoretical development (modification of the PINN loss function) and empirical validation (numerical experiments on Euler-Bernoulli and Timoshenko beams).


### Method of Data Collection

For validating the proposed methodology closed analytical form is utilized. This closed form solution is exciplicitly mentioned as exact solution function. These analytical solution is utilize for simulating the dynamics of beams on elastic foundations using both the traditional PINNs and the proposed transfer learning-based causal PINN framework.


### Type of Data


The type of data used in this research includes:

   1. Training Data for Sequential experiments: Solutions (Data) to partial differential equations (PDEs) modeling the dynamics of Euler-Bernoulli and Timoshenko beams are simulated using physics informed neural networks.


### File type/extension included in the folder


All codes are implemented using python jupyter notebook(.ipynb), Trained model files (.pkl, .pth), Log files (.log) showing the results at every iteration, and (.sh) files to execute on the cluster, .pdf and .png are figures which are used in the main paper.


For Causal PINN experiments well-posed physical equations of Euler-Bernoulli and Timoshenko is utilized.


History

  • 2024-06-05 first online, published, posted

Publisher

4TU.ResearchData

Format

python jupyter notebook(.ipynb), Trained model files (.pkl, .pth), Log files (.log), (.sh) files to execute on the cluster, .pdf and .png figures

Organizations

TU Delft, Faculty of Civil Engineering and Geosciences, Department of Engineering Structures

DATA

Files (1)