This repository contains all materials for the PMLG talk on using Maximum Entropy Spectral Analysis (MESA) and Wavelet Transforms for predictive maintenance of rotating machinery.
| File | Description |
|---|---|
Wavelet_MESA_PdM_PMLG.ipynb |
Main Jupyter notebook with full analysis, code, and visualizations |
pmlg_talk_slides.html |
Reveal.js HTML presentation (view in browser) |
pmlg_talk_slides.pptx |
PowerPoint version of the presentation |
extract_data.ipynb |
Data extraction and preprocessing notebook |
download.mp4 |
MESA spectrum animation across bearing run-to-failure |
PoC_PdM_steel_mill.pdf |
Proof of Concept: Predictive Maintenance for Steel Mills |
- The Problem — Why spectral analysis matters in predictive maintenance
- MESA — Maximum Entropy Spectral Analysis: theory, intuition, implementation
- Wavelet Transform — Time-frequency analysis beyond Fourier
- Application — Bearing vibration data (NASA IMS Dataset)
- Animation — Watching degradation unfold in the frequency domain
- Real-World Use Case — Predictive maintenance at a steel pipe factory
- Summary & References
- Python 3.10+
numpy,scipy,matplotlibmemspectrum(MESA implementation)pycwt(Continuous Wavelet Transform)jupyter(for notebooks)
- Burg (1967). Maximum Entropy Spectral Analysis
- Torrence & Compo (1998). A Practical Guide to Wavelet Analysis
- NASA IMS Bearing Dataset
Papers, datasets and other resources for Vibration Analysis.
- MAFAULDA. Machinery Fault Database [Online] (13Gb)
- NASA Bearing Dataset (6Gb):
-
- Example of this dataset in a notebook ⭐
- Another NASA Bearing Dataset (1Gb)
- Bearing Data Center. Seeded Fault Test Data
- Vibration of rolling bearings under widely varying speed conditions: This dataset contains observations of vibration of test stand under widely varying speed conditions.
- Bearing Vibration Data under Time-varying Rotational Speed Conditions: The data contain vibration signals collected from bearings of different health conditions under time-varying rotational speed conditions.
- Planetary gearbox vibration data
- Condition Monitoring for Handpumps - Vibration Data
- Unbalance Classification Using Vibration Data (Vibration Measurements on a Rotating Shaft at Different Unbalance Strengths): This dataset contains vibration data recorded on a rotating drive train. This drive train consists of an electronically commutated DC motor and a shaft driven by it, which passes through a roller bearing.
- AI4I 2020 Predictive Maintenance Dataset Data Set: The dataset consists of 10 000 data points stored as rows with 14 features in columns, it is also avaiilable at https://www.kaggle.com/shivamb/machine-predictive-maintenance-classification
- Microsoft Azure Predictive Maintenance: This dataset was available as a part of Azure AI Notebooks for Predictive Maintenance. But as of 15th Oct, 2020 the notebook link is no longer available. More information in here and exploratory data analysis
- Electric Motor Vibrations Dataset
- EXPERIMENTAL DATABASE FOR DETECTING AND DIAGNOSING ROTOR BROKEN BAR IN A THREE-PHASE INDUCTION MOTOR
- A Review of Feature Extraction Methods in Vibration...MDPI ⬇️
- Vibration Analysis Techniques included in the Predictive Maintenance Plan of a Paper Machine
- Machine Learning-Based Unbalance Detection of a Rotating Shaft Using Vibration Data
- Learning to See the Vibration: A Neural Network for Vibration Frequency Prediction
- Fault detection analysis in rolling element bearing: A review
- Fault Detection in Induction Motor Using Time Domain and Spectral Imaging-Based Transfer Learning Approach on Vibration Data
- An Animated Introduction to Vibration Analysis by Mobius Institute 🎞️
- Vibration Analysis - Bearing Failure Analysis by Mobius Institute 🎞️
- NASA Bearing Feature Extraction 📓
- IMS Bearing Dataset: Extracting Failure modes from vibration signals 📓
- Machine Learning Based Unbalance Detection of a Rotating Shaft Using Vibration Data
- Vibration in Cooling Towers ⬇️
- Vibration Analysis: FFT, PSD, and Spectrogram Basics [Free Download, Notebook, Examples]
- https://blog.endaq.com/top-vibration-metrics-to-monitor-how-to-calculate-them