Predictive models and anomaly detection
Models that anticipate drift, expiries and failures from calibration and monitoring series.
An instrument that drifts between two calibrations, a chamber that loses stability, a machine that gives signals before it fails: the information is already in the calibration and monitoring series. We build models that read it and warn in time.
What we deliver
| Anomaly detection | Classification of deviations in operation and early warning on thresholds learnt from each machine's history. |
|---|---|
| Drift between calibrations | Estimation of an instrument's drift from the history of its certificates, to anticipate an out-of-tolerance result. |
| Predictive maintenance | Failure prediction and classification models with sensor and automation data. |
| Deployment | The versioned model, served as a multi-user application in containers, in the cloud or on site. |
Where it already works. A predictive maintenance prototype: vibration indicators, a failure classifier and a monitoring dashboard, deployed in containers.
A model is only as good as its data. The models start from measurements by instruments calibrated by AKRIMET: calibrations traceable to the SI within the recognised scope.
One unit of an ecosystem
Each AKRIBIS One unit serves one part of the measurement cycle; together they cover it end to end.
A question about your measurement
A specialist replies with the solution, the instrument and the service that fit.