DAMADICS (real plant actuator data)
Public industrial actuator FDI benchmark used for our real-plant first look (partial/loss published; not spun as universal plant win).
- Dataset / project page: IAIR — DAMADICS (Instytut Automatyki i Robotyki, Politechnika Warszawska / Lublin Sugar Factory data)
- Key paper (cite the benchmark): Bartyś, M., Patton, R., Syfert, M., de las Heras, S., & Quevedo, J. (2006). Introduction to the DAMADICS actuator FDI benchmark study. Control Engineering Practice, 14(6), 577–596. doi:10.1016/j.conengprac.2005.03.003
- Our claim-safe write-up: Evidence § industrial · Phase 5 + DAMADICS note
How we used it: frozen Phase 5 MAIN evaluator, CV→demand / X→measured windows, 3 actuators × 3 days — evaporator juice level valve partial, juice flow valve partial, boiler water level valve loss.
Public eval package (gsrf-bench)
-
Live on site:
/eval/README.md· LICENSE-EVAL.md (non-production) · gsrf-bench product page -
Public GitHub (papers + eval only):
github.com/CBoon99/gsrf-papers
— PDFs under
papers/gsrf/and evaluation package undereval/. Full site / Workbench trees stay private. Site-hosted/eval/mirrors the public eval surface. - Try without install: Python snippet
Technical PDFs
Rewritten 2026-08-08 to the claim ladder. Contact on all docs: info@boonmind.io. Older “can’t overshoot / medical false-alarm detector” voice is retired.
- GSRF framework (safety-critical applications)
- GSRF Safety Filter
- GSRF Safety Brake
- Index: Papers library
Older PDFs may predate the trust rebuild. Live locked numbers always win on Evidence if wording differs.
Other public datasets mentioned
- Tennessee Eastman — simulated chemical plant XMVs (not field SCADA). Used as stress; losses published under Phase 5 MAIN rule.
- Synthetic Phase 5 / adversarial Tier-1 — generated under workbench protocols; scoreboards on Evidence + industrial note.
How to cite Zero Overshoot
Zero Overshoot / BoonMind (2026). Gradient-Stabilized Recursive Filtering (GSRF Practical) — locked claim ladder. https://www.zeroovershoot.com/evidence.html
Prefer also: llms-full.txt for agent context; methodology for how packs are locked.