Dataset Viewer
The dataset viewer is taking too long to fetch the data. Try to refresh this page.
Server-side error
Error code:   ClientConnectionError

FixFaults Industrial Fault Codes — Sample

A public sample (currently 2,808 — refreshed weekly from the live catalog) of the FixFaults industrial repair encyclopedia. Each row pairs a manufacturer fault code with a human-readable description and recommended repair action, ready to load for diagnostic-LLM fine-tuning, evaluation, or retrieval.

Full corpus: 70,000+ codes across 82+ manufacturers (Caterpillar, Cummins, Siemens, Fanuc, ABB, Bobcat, Haas, Yaskawa, Daikin, OBD-II, J1939 and more). This sample is CC-BY-NC for non-commercial use. For commercial / production access, see the live REST API below.

🚀 Quick start

from datasets import load_dataset
ds = load_dataset("avneetsingla/industrial-fault-codes-sample")
row = ds["train"][0]
print(row["code"], "—", row["description"][:80])
# P0001 — Fuel Volume Regulator Control Circuit/Open

# Filter by brand
cummins = ds["train"].filter(lambda r: r["brand_slug"] == "cummins")
print(len(cummins), "Cummins codes")

📊 Schema

column type description
code string The fault / error code (e.g. P0420, A.022, SPN 4364 FMI 5)
brand string Manufacturer / standard (e.g. Cummins, OBD-II, Siemens)
brand_slug string URL-safe brand identifier (lowercase, hyphenated)
description string Human-readable description of the fault — what is suspect
solution string Recommended repair / next-step diagnostic action
equipment_type string Equipment category (e.g. diesel, hvac, cnc, servo-drive, plc)
severity string Severity tier: low / medium / high / critical
permalink string Slug of the source article on fixfaults.com (e.g. cummins-111-fault-code-solution)
source_url string Canonical URL of the source article — every row has a long-form diagnostic guide

🎯 What this is useful for

  • Diagnostic copilots & repair LLMs — fine-tune Mistral / Llama / Phi on (code, brand) → (description, solution) to build a chatbot that mechanics can query in natural language.
  • RAG over technical docs — embed the rows and use them as ground-truth retrieval for "what does this code mean" questions, with source_url as the citation back-link.
  • Severity classification — train a model to predict severity from description + solution.
  • Multi-brand mapping — disambiguate similar-looking codes across brands (e.g. A.022 on Yaskawa vs ABB).
  • Multilingual seed — use English descriptions as canonical seeds for translation pipelines.

🔌 Want the full 70,000+ corpus? Use the live API

The full database is available via REST API with a free tier of 500 requests/month, no credit card required.

Example API call:

curl -H "Authorization: Bearer ff_live_xxx" \
  "https://www.fixfaults.com/api/v1/codes/P0420?brand=Cummins"

📜 License

CC-BY-NC 4.0 — free for non-commercial research, evaluation, and personal use, with attribution to FixFaults (https://www.fixfaults.com).

For commercial use — training a product, embedding in a SaaS, redistribution, fine-tuning a model you plan to sell — please contact support@fixfaults.com for a commercial license, or just use the live API which is commercially licensed by default.

📚 Citation

@misc{fixfaults2026,
  title        = {FixFaults Industrial Fault Codes — Sample},
  author       = {Singla, Avneet},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/avneetsingla/industrial-fault-codes-sample},
  note         = {Sample of the FixFaults industrial repair encyclopedia (https://www.fixfaults.com)}
}

🔗 Links


Maintained by Avneet Singla — Founder & Editor, FixFaults. Auto-refreshed weekly from the live catalog.

Downloads last month
56