Datadory notebook
Bearing Vibration Dataset Download Data: Dataset Structure and Field Coverage
1,744 datasets. Pick your catch. Every guide here is built on what the catalog can actually prove.
1,744 datasets. Pick your catch.
Which datasets actually contain bearing vibration data?
The first is a bearing vibration test-stand collection: per-cycle sensor readings from bearings run to failure on a lab test rig, which is exactly the experimental shape condition-monitoring engineers want for fault diagnosis - healthy operation degrading into known failure modes. The second is the FEMTO bearing accelerated degradation archive, where accelerated stress shortens time-to-failure so more degradation trajectories fit into one experiment campaign.
Both sit among the repository's 21 run-to-failure datasets, which range from about 12 MB (C-MAPSS v1) to about 15.8 GB (C-MAPSS v2) and total tens of gigabytes. The experiments behind them were collected 2008-2019 and published as fixed archives, so every team benchmarking against NASA data pulls identical bytes.
Why does the Kaggle mirror not answer a bearing query?
Because it packages only the engine benchmark. Kaggle: NASA Turbofan Jet Engine (C-MAPSS) is the community-standard remaining-useful-life corpus - four subsets (FD001 100 train / 100 test units, FD002 260/259, FD003 100/100, FD004 248/249) totalling 25,888,576 bytes uncompressed, with train_FD001 alone holding 20,631 rows of per-cycle sensor readings - but no bearings appear anywhere in it.
The mirror is still useful next to a bearing project in three specific ways:
Who uses bearing vibration data, and what limits it?
Datadory tagged five personas against the NASA Prognostics Data Set Repository, and the top three all touch the bearing archives at relevance 3 of 3: Data Scientists & ML Engineers, who train fault-diagnosis models on the run-to-failure trajectories; Developers & Data-Product Builders, who build ingestion and labelling tooling on free S3-hosted zips; and Journalists, Academics & Students, who cite per-dataset donor citations in prognostics research. Competitive-intelligence teams and market researchers carry it at relevance 1, testing vendors' predictive-maintenance claims against a public baseline.
Four constraints recur alongside those use cases, and all four are cheaper to design around than to discover mid-project.
For procurement-side context on the physical bearings themselves - indicative prices, MOQs and audited suppliers - Made-in-China.com remains this slice's only supplier-level source, with listing prices quoted as indicative offers rather than binding ones.
Pick up where this leaves off
Every one of these ships with sample rows before you commit to anything.
NASA Prognostics Data Set Repository
Kaggle: NASA Turbofan Jet Engine (C-MAPSS)
JMTBA machine tool statistics Japan data
6 documented core fields · further breakdown fields on request · uchinaiju …+3 more
Made-in-China.com
7 verified core fields · category placement · product_title …+6 more
Want rows instead of a pitch? Name the datasets.
API, files, or your warehouse. Daily, weekly, or hourly.
Get a sampleQuestions worth asking
Do I need an account or integration key to get the NASA bearing data?
No. The per-dataset links point into the phm-datasets.s3.amazonaws.com bucket, and Datadory verified in August 2026 that this pattern serves HTTP 200 with Content-Type application/zip without authentication. Only seven repositories-wide datasets - none of them the bearing sets named here - currently require an email request.
Is there a bearing dataset on Kaggle like C-MAPSS?
Not in this slice. The Kaggle mirror behrad3d/nasa-cmaps carries only the C-MAPSS turbofan subsets FD001-FD004 - 25,888,576 bytes uncompressed under commercial delivery terms - with no bearing archives. For bearings, use NASA's own repository; use the Kaggle mirror to shake down your remaining-useful-life pipeline before loading larger files.