I have a device that came in as Fridge 2. Doesn’t surprise me I have a fridge in the kitchen and one in the garage, so I was hoping to correctly tag them. Unplugged frig in garage, device stays on. Hit the breaker for the kitchen frig, device stays on. I went down and flipped every breaker, and guess what…the device stayed on. Anyone else run into this?
Eventually the bubble disappears for me when I hit the right breaker. But your Fridge might have a really weird off signature.
MachoDrone has had the same scenario with a dehumidifier and refrigerator/freezer when using breakers/unplugging. The bubbles stayed on.
Eventually I was near them when they ran their natural course to cycle power off and on and tagged each… Just luck.
You may attempt to dial your fridge to a cooler setting so it turns on, then dial it much warmer setting to turn it off… Remember to set it to the original temp when done.
I know you’re savvy on Sense’s learning feature, but for other members reading this, this is a great example on why we cannot think we can train sense on devices… there’s just too many variables on how things ramp-up to operating conditions and then ramping-down to power off.
Turning a breaker on and off has its own signature (I broke out my vintage oscilloscope and some spare current transformers from my welserver.com power monitoring project for the first time in over a decade…and I had to relearn how to operate it…to watch), in addition to whatever loads powering up there are on the breaker. That’s why breaker cycling doesn’t help Sense learn (if indeed anything ever does…sigh). Similarly, I did some experiments with a couple of other devices (two different displays…Samsung and ViewSonic, Brother laser printer) and I found that their “turn power on” signature is quite different than the “start working” signature, another reason why such training doesn’t help Sense.
So, this is still a science experiment with customers as the Guinee Pigs.
Note, clearly Sense engineers and data scientists know the why-not, so rather than saying “training doesn’t/can’t work”, why not tell frustrated customers why not…perhaps with signature screen shots? As an electrical engineer with experience in data analytics, that would have saved me much frustration.
C’mon, @andy. You place your digs on Sense any thread you can. Your responses are created in a way that as if you never read any of the successes that the majority share on Sense Monitoring while they keep adding devices, like EVs, etc. Guinea Pigs? It’s truly a dig again including additional language.
Imma gonna look for a block button, I’ve never done that at any forum.
MachoDrone speaks the truth.
Glad you did the fieldwork and experimentation. I agree that Sense needs another in-depth blog on signature identification and machine learning that explains these kinds of things:
- Details about on and off signatures - what kind of time window is examined and what features. Gotta explain that the “obvious” signature a human sees is NOT the signature that Sense is (currently) looking for.
- Why Sense learning requires “natural” ons and offs - not breaker ons and offs. The whole idea of manual training with breakers and wall switches (unless the device is always turned on with the wall switch) is malarky.
- Why Sense needs lots of background noise conditions to learn. I get this intuitively having worked with facial recognition that requires lots of different head positions, lighting conditions and backgrounds, in order to develop high accuracy. The whole idea of training mainly on a "baseline, no-noise, power waveform, is kind of like saying that just one perfect front-on photo of a face is the key to facial recognition (hint - it is NOT).
But I can also see a challenge not to reveal too much proprietary information.
If you find a block, let me know, so I can get rid of your (seemingly endless) hype for a product that just plain doesn’t work
Then there are folks like MachoDrone (who apparently works for Sense) and who keeps claiming this science experiment works fine…except that IT DOESN’T
Yeah, we can definitely do a better job of telling/showing why not and we’re working on a video for some of that, but a deep blogpost like @kevin1 mentions would help as well. There is a risk of revealing proprietary info, so we’ll have to dance around that a bit. We are planning on doing another customer conference call hopefully in the next month or so, so keep your eyes open for that as we can answer a lot more questions there.
@MachoDrone and @andy: There is no mute/block functionality in the Discourse forum software. If you believe that someone’s posts violate the Community guidelines or the overall spirit of the Community, flag them and I will deal with it. Comments to each other in this thread are getting borderline in violation of Community rules. Please remain respectful to each other and respectful to @jasonemoyer who started this thread. The only people who work for Sense that post on this forum will be clearly identified as such.
Well, I can’t let some of these comments go without throwing in my two cents about my Sense. (Trying to rhyme here).
In my opinion and use of Sense, it is unfair to say the product does not work. Does it have problems? Yes. But is does work if you set your expectations to that of using new technology. BTW, I started using Windows back in the version 1.0 days and today, over ten versions later, it still has problems.
It is also unfair to claim anybody who likes this product works for them. That is just plain throwing stones and trying to influence others to believe your way, which is not helpful. Sharing information is helpful. Sense may have set expectations too high with their advertising, but the product does provide value and it is not outrageously expensive.
There are issues with device detection. Why others expect perfection in this new technology is beyond me. Those who have worked in technology who use Sense understand this. Those who want perfection in their life for no or very little cost, are setting themselves up for disappointment.
I have worked with the Sense support people on trying to get my kiln identified correctly and they generally are helpful, responsive, and professional. Did they get it corrected totally? Not yet. Shall I call them names and claim their product sucks? No. It is not fair to do that.
Try working with your cell phone provider, cable company, Internet Service Provider, or health care provider. These people at Sense are far better than these provider of services. So, I suggest, we all stop throwing digs at each other and become more kind and helpful.
It might just start a trend that will spread elsewhere.
I’ll climb off my soapbox now.
Thanks for chiming in, Fred. Your comments are appreciated.
I want to second that this Community is absolutely not a place for personal attacks of any sort and I will moderate that content whenever I see it. Please feel free to flag comments as you see fit and I will deal with them and the poster.