How cycles are defined for the anomaly detection module?

[LR_9796586]

Let’s suppose that I have a pump that completes various cycles (more than 5 full cycles) across a day. Under normal conditions, every time the pump completes a cycle it goes through the following states: starting; running; stopping. Depending on the amount of fluid that needs to be pumped the cycle lasts a different amount of time. Also, after each cycle the pump remains stopped for different amounts of time. 

Below an example of how it could work in a day.

The cycles depicted in the image above are “operation cycles” that don’t have the same length and are aperiodic, are these the correct unit of analysis for the anomaly detection or should be defined in a different way? 

[jgreiner · Accepted solution]

Hello,

Yes your understanding of cycles is correct for Anomaly Detection.  Two other things to consider for your use case below since the pump runs at aperiodic intervals, I would suggest only running anomaly detection when the pump itself is actually in operation.  Also since the pump runtime varies each cycle, there may be some false positives between state changes like from running to stopping.

Let me know if you have any further questions or concerns.

Warm Regards,

John