jonathanwvd/awesome-industrial-datasets

What this project is

Awesome Industrial Datasets is a carefully curated list (an "awesome list" — a common format in open source where someone collects and organizes links and resources on a topic) of high-quality industrial datasets. The project lives on GitHub and organizes ~190 datasets across manufacturing, energy, oil & gas, chemicals, aerospace, and related sectors.

Its purpose is simple: make it easier for people who need real-world industrial data to find it. Instead of scouring the web, researchers, engineers, and data scientists can browse this one catalog, see what each dataset offers, and jump straight to the source.

What you'll find inside

Every dataset in the list gets a row in a large table (and its own detail page in the repo's markdown/ folder). Each row describes the dataset across 10 dimensions:

Domain — the industry sector (e.g., Oil & Gas, Manufacturing & Production, Energy & Power)
Asset / Process — what physical thing the data comes from (e.g., turbines, bearings, PCB boards, chemical processes, solar panels)
Modality — the data format (images, time series, tabular data, vibration signals, 3D point clouds, X-ray/CT scans, etc.)
Task — what machine learning problems the data can support (anomaly detection, defect classification, predictive maintenance, forecasting, quality prediction, remaining-useful-life estimation, and more)
Annotation — what labels come with the data (e.g., pixel masks for defects, fault type labels, scalar targets for regression)
Source Type — where the data came from (real production systems, lab testbeds, simulations, public records)
Access — where to download it (UCI, Kaggle, Hugging Face, official portals, repositories, etc.)
Size — the number of samples or records, when stated
Year — publication year, when known
License — usage terms, when known (many entries note "information not available")

Real examples from the catalog

The table gives you a flavor of the breadth:

Image-based defect detection: datasets like 3CAD (27,000+ PCB images with pixel-level defect masks) and Casting product image data (7,348 images for quality inspection)
Predictive maintenance: datasets like C-MAPSS Aircraft Engine Simulator Data and CNC Mill Tool Wear, where the goal is predicting failures or tool wear
Energy forecasting: datasets like Electricity Load Diagrams and Brent Oil Prices
Process monitoring: datasets like Condition monitoring of hydraulic systems and Tennessee Eastman Process simulations
Benchmark suites: some entries are collections of scenarios or tasks rather than single datasets (e.g., AssetOpsBench)

Who maintains it and how to use it

A single maintainer (jonathanwvd) curates the list. The README includes a link to an HTML version of the same catalog at indatlas.com for easier browsing. Readers are also invited to star the repository if they find it useful.

What it's not

The project does not host the datasets themselves — it's a directory. Each entry points to the original source (Kaggle, UCI, official portals, etc.), and some of those sources require login or have license restrictions. Also, metadata completeness varies: many entries explicitly state that size, year, or license information is not available.

Bottom line: If you need industrial data for a machine learning or analytics project, this is a one-stop directory that tells you what exists, what it's good for, and where to get it.