Datasets for uncertainty-aware surrogate modelling of a permanent-magnet synchronous motor (CP-GE-PCE)

This dataset accompanies a master’s thesis on uncertainty-aware surrogate modelling of a two-dimensional permanent-magnet synchronous motor. It supports the construction and evaluation of polynomial chaos expansion (PCE) and gradient-enhanced polynomial chaos expansion (GE-PCE) surrogates, combined with dimensionality reduction and conformal prediction. The simulation data were generated using a nonlinear magnetostatic isogeometric analysis model implemented in MATLAB with the AMIGA framework. The varied inputs describe rotor geometry, rotor position, and electrical excitation. The archived data include sampled input parameters, magnetic vector-potential coefficient vectors, electromagnetic torque, and first-order sensitivities with respect to the model parameters. Reference mass and stiffness matrices are provided for evaluating field approximation errors. The upload also includes datasets used to assess empirical coverage of conformal prediction regions and trained autoencoder models used for comparison with proper orthogonal decomposition (POD). The files support the numerical experiments presented in the thesis, including comparisons of surrogate accuracy, dimensionality reduction, and the coverage and size of prediction regions. Simulation outputs are stored in MATLAB .mat format, and sampled input parameters are provided as CSV files. The accompanying code and reproduction instructions are available at:https://github.com/lh2011/CP-GE-PCE Extract the data archive so that the data folder is located in the root directory of the code repository. Preserve the archived filenames and folder structure, as these are used by the accompanying scripts.