Predictive Maintenance - Deployment
This workflow deploys a previously trained auto-regressive model for anomaly detection in sensor data of rotor machines.
The workflow first selects the date for deployment (e.g., two months of data prior to the selected date must be available), loops over each frequency column and applies the previously trained auto-regressive model to the data.
Next, it calculates 1st level alarms based on the prediction errors, and 2nd level alarms as the moving average of the 1st level alarms. Finally, it triggers an email notification, if a 2nd level alarm is activated.