Data, Code, and Supplementary Material for "Hilbert–Huang Transform and Multi-sensing for Within-Specimen Fault Diagnosis in Induction Motors"

This repository contains the data, processed features, analysis code, supplementary results, and computational resources associated with the manuscript “Hilbert–Huang Transform and Multi-sensing for Within-Specimen Fault Diagnosis in Induction Motors”. The study investigates fault diagnosis in three-phase induction motors using Hilbert–Huang Transform (HHT) descriptors extracted from line-current and vibration measurements. Three fault families are considered: inter-turn short circuit (ITSC), broken rotor bars (BRB), and distributed bearing wear. Experiments were performed using two physical induction-motor specimens, rated at 1 HP and 2 HP, under 11 supply-voltage conditions and 10 shaft-torque levels. The revised computational workflow uses the original 3 s acquisitions sampled at 16 kHz. Each acquisition is divided into ten non-overlapping 0.30 s windows for feature extraction, while training, validation, and test partitions are defined at the acquisition level. The primary evaluation employs repeated acquisition-grouped five-fold cross-validation over five random seeds, providing 25 external evaluations for each motor, diagnostic task, and signal representation. The repository includes: processed HHT feature tables derived from current and vibration signals; acquisition identifiers, fault labels, severity information, torque levels, and supply-voltage conditions; scripts and notebooks used for feature extraction, classification, and statistical analysis; repeated acquisition-grouped validation results; current-only, vibration-only, and current–vibration fusion experiments; matched time-domain and Fast Fourier Transform (FFT) reference representations; systematic held-out torque and voltage evaluations; held-out fault-realization experiments for ITSC, BRB, and distributed bearing wear; direct cross-motor transfer experiments; HHT parameter-sensitivity analyses; grouped permutation-importance results; computational-cost measurements; supplementary tables supporting the revised manuscript; software-environment and execution information required to reproduce the analyses. The repository is intended to support transparency and reproducibility of the computational analyses reported in the manuscript. The experimental database contains one physical specimen at each rated power; consequently, the data support evaluation within the measured specimens and operating domain and should not be interpreted as independent replication across a population of induction motors. Raw waveform availability is subject to the applicable institutional data-sharing conditions. When raw recordings cannot be distributed directly, the processed features, metadata, predefined validation structure, analysis code, and numerical results required to reproduce the reported analyses are provided.