Voiceprint Representation of Motor Bearing Acoustic Signals under Fault Conditions

This dataset contains approximately 14,200 time–frequency representation images (224 × 224) generated from one-second audio recordings sampled at 44.1 kHz. The recordings were collected from a 3 kW induction motor, focusing on the acoustic signals produced by the motor bearing under different conditions (healthy state and faulty conditions).

For each audio segment, four complementary acoustic representations were generated to capture different spectral and temporal characteristics of the signal:

- Mel-spectrogram– A perceptually motivated frequency representation based on the Mel scale.
- Cochleogram – A biologically inspired time–frequency representation modeling human cochlear behavior.
- Tempogram – A temporal representation capturing periodic, rhythmic, and impact-related patterns over time.
- Chromagram– A harmonic representation that distributes spectral energy across 12 pitch-class bins.

The dataset covers five bearing (6206/C3) conditions, providing a comprehensive benchmark for fault diagnosis and machine learning applications:

- H (Healthy): Normal bearing condition
- B (Ball Fault): Defect in the rolling element (ball)
- C (Cage Fault): Defect in the bearing cage
- I (Inner Race Fault): Defect on the inner race
- O (Outer Race Fault): Defect on the outer race

Each one-second audio file was transformed into the four aforementioned representations, enabling multi-domain feature extraction for:

- Fault classification
- Condition monitoring
- Predictive maintenance

By combining perceptual (Mel), human cochlear behavior modeling (Cochleogram), time-based rhythmic modulation (Tempogram), and harmonic structure (Chromagram), this dataset provides a rich and diverse feature space for researchers working on acoustic signal processing, deep learning, and intelligent fault diagnosis in rotating machinery. For more details on the methodology and model development, please refer to our article published in Results in Engineering. If you use this dataset, please cite both this dataset and the following article in your scientific work.

"A High-Precision Voiceprint Recognition for Fault Diagnosis of Motor Bearings through Lightweight Dynamic Convolutional Neural Network"
🔗 https://doi.org/10.1016/j.rineng.2025.108463