Tacit AI runs one risk engine for your plant: the rules, every asset and a master failure library. Each failure report refines it. Each FMEA is assembled from it on demand, sourced and consistent.
What counts as a system, equipment, component or part. What defines each equipment class, which characteristics matter and which failure modes apply. Guardrails keep the rules from drifting.
Every asset ID with its level, parent, category, components and every characteristic found so far, straight from your maintenance system.
Every failure mode held once, with its causes, mechanisms and indications, and linked to every category it applies to. A bearing failure from shaft voltage is the same on a pump, a compressor or an agitator.
Each new failure report is read with the current rules. Reports without enough detail go back for a data quality review.
Do the asset and component fit the hierarchy? Do the failure, mode and impact fit the failure rules? If everything fits, the library updates quietly.
If something does not fit, the system proposes a rule change with the reason and the source reports behind it.
An approved change applies wherever the rule is used, including past reports. Every FMEA reflects it the next time it is assembled.
Each engineer gets 5 to 10 proposed rule changes a week. Approve, edit or reject each one. As confidence grows, fewer proposals need review, and review time is tracked to prove it.
Which failure modes apply is decided by the rules, not by the model. Language models and statistical tests propose and refine the rules. Your engineers approve them.
The FMEA is not stored. It is built fresh from the current rules for any asset or category, with the source reports and the rule behind every line.
A vertical pump fails on a dry steady bearing. The system proposes a new category with the bearing as a component. Once approved, every vertical pump FMEA includes it and no horizontal pump FMEA does.
Starts from what you already have: existing FMEAs, manuals, P&IDs, failure history and standard-aligned failure libraries. Connects to 100+ systems.
We warm-start the engine from your FMEAs and failure history. First FMEA in under 3 weeks, similar ones in days.
