Reproducible Python Code and Experimental Results for AI-Based Bearing Fault Classification Using Vibration Sensor Data: Effects of Signal Representations and Decision Thresholds

This repository provides the complete Python implementation and verified experimental results associated with the research article entitled “AI-Based Bearing Fault Classification Using Vibration Sensor Data: Effects of Signal Representations and Decision Thresholds.” The study investigates the effects of dimensionless vibration features, signal representations, physical-time dynamics, and source-derived decision thresholds on artificial intelligence-based bearing fault classification. Experiments were conducted using 59 source recordings from the MCC5–THU three-phase asynchronous motor dataset and 32 target recordings from the University of Ottawa Electric Motor Dataset. Five predefined feature and temporal configurations, 11 classification models, and four decision rules were evaluated. The models include physical-time spiking neural networks, a memoryless SNN counterpart, conventional machine learning classifiers, and ensemble methods. Model training, calibration, and decision threshold selection were performed exclusively on source data using nested stratified group cross-validation with three outer folds and two inner folds. The trained models were evaluated on the Ottawa recordings without target-side parameter updates. The Ottawa analyses include exploratory comparisons at the recording level. The archive contains: The complete Python source code and software dependencies. Publication Figures 5–10 in PNG and SVG formats. Publication Tables 5–8 in CSV format. Source cross-validation results and record-level predictions. Ottawa classification scores, calibration results, and decision thresholds. Group-based bootstrap results and supporting feature statistics. Experimental protocol, raw-data checksums, software environment information, and reproducibility verification reports. Numerical results were checked against the manuscript at the reported precision. The archive also includes independent verification of metrics recalculated from serialized prediction files. The original raw vibration recordings are not redistributed. They remain accessible through the original dataset repositories. Original datasets: MCC5–THU: https://doi.org/10.17632/6s3dggj9mw.1 University of Ottawa Electric Motor Dataset: https://doi.org/10.17632/msxs4vj48g.2 This repository supports transparent evaluation, independent verification, and reproducibility of the reported experimental findings.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24370284
Primary Topic
Machine Fault Diagnosis Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Reproducible Python Code and Experimental Results for AI-Based Bearing Fault Classification Using Vibration Sensor Data: Effects of Signal Representations and Decision Thresholds

Ramazan Güngüneş
Zenodo (CERN European Organization for Nuclear Research)
Machine Fault Diagnosis Techniques
article

Reproducible Python Code and Experimental Results for AI-Based Bearing Fault Classification Using Vibration Sensor Data: Effects of Signal Representations and Decision Thresholds

Ramazan Güngüneş
article en

Abstract

This repository provides the complete Python implementation and verified experimental results associated with the research article entitled “AI-Based Bearing Fault Classification Using Vibration Sensor Data: Effects of Signal Representations and Decision Thresholds.” The study investigates the effects of dimensionless vibration features, signal representations, physical-time dynamics, and source-derived decision thresholds on artificial intelligence-based bearing fault classification. Experiments were conducted using 59 source recordings from the MCC5–THU three-phase asynchronous motor dataset and 32 target recordings from the University of Ottawa Electric Motor Dataset. Five predefined feature and temporal configurations, 11 classification models, and four decision rules were evaluated. The models include physical-time spiking neural networks, a memoryless SNN counterpart, conventional machine learning classifiers, and ensemble methods. Model training, calibration, and decision threshold selection were performed exclusively on source data using nested stratified group cross-validation with three outer folds and two inner folds. The trained models were evaluated on the Ottawa recordings without target-side parameter updates. The Ottawa analyses include exploratory comparisons at the recording level. The archive contains: The complete Python source code and software dependencies. Publication Figures 5–10 in PNG and SVG formats. Publication Tables 5–8 in CSV format. Source cross-validation results and record-level predictions. Ottawa classification scores, calibration results, and decision thresholds. Group-based bootstrap results and supporting feature statistics. Experimental protocol, raw-data checksums, software environment information, and reproducibility verification reports. Numerical results were checked against the manuscript at the reported precision. The archive also includes independent verification of metrics recalculated from serialized prediction files. The original raw vibration recordings are not redistributed. They remain accessible through the original dataset repositories. Original datasets: MCC5–THU: https://doi.org/10.17632/6s3dggj9mw.1 University of Ottawa Electric Motor Dataset: https://doi.org/10.17632/msxs4vj48g.2 This repository supports transparent evaluation, independent verification, and reproducibility of the reported experimental findings.

Zenodo (CERN European Organization for Nuclear Research)
Kırıkkale University (TR)
Machine Fault Diagnosis Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon