Ability to detect / identify time of day and length of time used

Strange anomaly here has brought an idea. Perhaps you guys have thought about this already, but worth mentioning.

I have two devices, a tea kettle and instant pot. Apparently the signatures are identical in wattage and startup. Whenever the instant pot is on “keep warm” it cycles with two pulses back to back every so many minutes in a very predictable pattern over time.

Sense identifies these pulses as the tea kettle running.

The only difference seems to be the pattern of usage over time. The tea kettle runs a solid wattage from on to off, always much longer run time than the pulses of the instant pot. Even though the traces themselves look identical, there are a few differences:

  1. The kettle runs once for several minutes and cuts off. Any subsequent runs would be a random time later.
  2. The kettle, at the end of a run, does pulse on and off as it approaches set temp, but that is after a brief run of longer duration
  3. The instant pot has a repeated pattern over long periods of time. Sometimes all night in a repeated pattern of wait, run, of, run, wait.
  4. The kettle runs randomly throughout the day while the instant pots repeats the pattern perfectly every few minutes for hours.

I wonder if sense could pick up on usage patterns over time to detect a difference here?

Perhaps there could even be some way to set a sort of “profile” for discovered devices that says how often or how long or how randomly they run to help sense differentiate between them? At some point it becomes obvious that a pattern seen can not be the device “found” becuaee the pattern over time is impossible.

“Progressive Device Detection” should help with things like this when it rolls it out.
2021 Data Science Update - YouTube