Deviations found before they become stops.
Anomaly detection watches sensor data with established statistical methods and standard machine learning models, and lets your experts label what they see.
public/images_site/anomaly-sensors.pngWhat it does
Sensors watched in real time
Detected anomalies are shown on the trend of the measure, as they occur.
Statistics first
Rolling z-scores flag values that drift away from the recent behaviour of the measure.
Standard machine learning
Unsupervised models such as Isolation Forest score a run after it ends, and similarity with clusters of past runs adds context.
Models for a process
A dedicated model covers welding quality.
Experts in the loop
Experts label anomalies and add their insight, and the model improves with their feedback.
The data it stands on
Like every module, it reads the shared ISA-95 model. Nothing is copied into a separate database.
- Sensor variables subscribed from each machine
- Equipment from the site model
- Labels added by experts
What it measures
- Anomalies detected and labelled
- Anomaly score per run
- Statistics of each measure: mean, deviation, minimum, maximum and trend
See DigiFactor on your own data.
Book a demo with our engineers. We show the platform, the modules and the agents on a case close to yours.
Book a demo We answer within one business day