Digital Twin-University of Ottawa Induction Motor Synthetic Dataset (DT- UOIMSD)

This synthetic dataset was generated using a physics-based digital twin of the University of Ottawa Electric Motor Dataset - Vibration and Acoustic Faults under Constant and Variable Speed Conditions (UOEMD-VAFCVS) test rig. The purpose of this dataset is to evaluate whether synthetically generated motor fault data can be used to train machine learning models that successfully transfer to real laboratory measurements, without requiring real data during training. The digital twin replicates the vibration signatures of eight induction motor health conditions healthy, rotor unbalance, rotor misalignment, stator winding fault, voltage unbalance, bowed rotor, broken rotor bars, and faulty bearings under four constant speed conditions (15 Hz, 30 Hz, 45 Hz, and 60 Hz) at both unloaded and loaded states. Synthetic signals were calibrated against the real UOEMD-VAFCVS measurements using spectral analysis, with signal amplitudes and dominant frequency content matched to the real dataset.

The dataset contains 128 CSV files, with two synthetic replicates per operating condition. Each file contains 420,000 samples collected at a sampling frequency of 42,000 Hz, equivalent to 10 seconds of data. Each file contains four columns: Acc 1 (m/s²), Acoustic (V), Acc 2 (m/s²), and Acc 3 (m/s²), corresponding to the three accelerometer channels and microphone of the original test rig. Files are named using the convention S-{fault}-{speed}-{load}_{index}.csv, for example S-B-R-2-0_01.csv represents the first synthetic replicate of a bowed rotor fault at 30 Hz unloaded.
This dataset is intended to be used alongside the real UOEMD-VAFCVS dataset available at https://data.mendeley.com/datasets/msxs4vj48g to evaluate zero-shot domain transfer from synthetic to real motor fault data.