Anomaly Detection on IoT Data
This workflow performs anomaly detection on sensor data of rotor machines using a control chart. It first calculates the "normal conditions" as the cumulative average +/- 2 times the corresponding standard deviation.
Then, it raises a 1st level alarm, if a sensor exceeds the band for normal conditions on a single frequency band, and raise a 2nd level alarm if this happens on at least 25% of the frequency bands of all sensors at the same time.
Finally, it visualizes the 2nd level alarms in a line plot w.r.t the rotor breakdown, and triggers an action if a 2nd level alarm is active.