Internet Explorer is not a supported browser for TI.com. For the best experience, please use a different browser.
Video Player is loading.
Current Time 0:00
Duration 5:44
Loaded: 0%
Stream Type LIVE
Remaining Time 5:44
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected

[MUSIC PLAYING]

Hello, everyone. I'm Adrian.

I'm Nick.

And welcome to another episode of Connect. Today we're bringing in J.D. Crutchfield to talk about condition monitoring with our MSP432 devices. So welcome to the show.

Thanks, Adrian.

So can you talk a little bit about what condition monitoring is, and why we should care about it?

Sure. So condition honoring is, as it sounds, it's a very, very simple concept. Really you're just trying to monitor the condition or, let's say, the health on a system.

Cool. And when you say system, are there particular things that we're keeping an eye on?

Yeah, I mean this can be anything, like, you could say a factory or your home, promoters to pumps. But, really, we're trying to predict the maintenance of those devices, right? All those devices are going to fail at some point. But we really want to know when they're going to fail ahead of time as long as possible.

OK. And so, if you're talking about monitoring the condition or the health of a motor, what are some of the methodologies for achieving that?

Sure. So all motors have a predictable life cycle, right? They're going to fail in a very common pattern. So the very basic way to predict the health of it is maybe temperature. Right? When a motor is hot to the touch, right--

It's not a good thing.

Yeah, it's not a good thing. It's really close to failing, right? If it's squealing and screaming as it's running, obviously, you're going to break any moment now. So you don't have much of a heads up on those types of detection algorithms.

OK.

What we're actually pursuing is vibration condition monitoring. So vibration gives us the longest lead time of all the other, let's say, methods out there. This gets-- we predict we can get anywhere from three to six months of heads up before the system actually fails.

OK. And so, I guess, vibration monitoring-- we're focusing on that, because it gives you the biggest window of lead time before failure. Can you explain a little bit how we're actually doing the vibration monitoring?

Sure. So today's factory rates is getting more and more smart every day. So what we've actually developed is we have a EVM, it's a wireless condition monitor, wireless vibration condition monitor. So it's using the simply link products, right, for the wireless connection. It's using the MSP432 low power MCU. And then we have a accelerometer on the front end, a 3x accelerometer that we're using.

The reason that we're using wireless is it makes it very, very easy for the factories to deploy it. Right. So instead of having to run wires everywhere, they can just go around and stick these on each of their motors or pumps or even in the house, stick it on all of the systems. And then that can just monitor it locally, and then send it back to some sort of gateway.

And are you doing that-- that monitoring locally, at the edge?

Yeah. So we're actually doing all the processing and decision-making in the sensor itself. The reason that we're doing this, and this is what we call edge processing, is because it actually gives you a lower average power. It gives you much longer battery life. Historical systems, you know, they might do some just measuring the data, and just send that raw data out over the network. But that really kills the battery life having the radio on constantly streaming this data.

So with the-- with the MSP432, we have a phony point unit where we can do all that processing locally, make the decision, and only turn on the radio and transmit that when there's an actual fault detected.

Gotcha. And just to make sure I understand. So there is a motor that's vibrating, I guess, at a known good frequency. And based on that accelerometer data, we can kind of detect if there are any shifts that start to go beyond that known good frequency.

Yeah. You can get very, very complex in the algorithms you choose to use. But, yeah, at the very beginning and basic one is, basic thresholding. Right. You're looking-- you know what a good signature looks like. And if it starts moving beyond that, something's wrong. You can take that even further, especially when you have a library of historical data, to say, all right, with this type of signature, this is the fault that's actually occurring.

OK.

And so you mentioned things like low power and how the vibration monitoring is happening. Adrian mentioned up front that this is all done on the MSP432P4. Can you tell us a little bit about that device specifically? And why it's suitable for this application?

Sure. So the MSP432P4 is a low power Cortex M4 microcontroller. So it also has a very high precision ADC on it. So that high precision ADC allows us to get very accurate measurements from the vibration sensor, in this case.

And then we have the Cortex M4 core plus a floating point unit where we can do all that processing. On top of that, it's a very low power microcontroller. So it allows us to very efficiently do all that processing detection locally on the system, and keep ours-- do it very quickly and get back to sleep, and keep our average power very low.

OK.

And then at that point, we can take advantage of some of the wireless devices in the sampling platform. So what are some common ways of adding this to a network?

Sure. So there's common-- common ways this is deployed. It can either be added to a network, like you said. This can be a sub on gig network, a Bluetooth mesh network, or Wi-Fi system. There's also kind of just a direct point-to-point systems where we only put Bluetooth on the sensor, and a technician walks through the factory and connects to each one, all right, and downloads and checks that data. So that's something that we can definitely do with all of the simple wireless products.

Cool. So, J.D., if people watching are interested in kind of getting started with this solution or learning more, are there resources on the web that they can start with? Yeah, of course, Nick. Yeah, we have several blogs talking about the different aspects of these types of systems. We have a TI design out there that has all the hardware and software references already available, as well as several demo videos available on the TI.com.

OK. Great. Yeah. Thank you for joining us, J.D. This is a very interesting topic. Especially for me, this was brand new to learn about. I'm sure our viewers enjoyed it. Thank you, guys, for watching us and joining us on this episode of Connect. Be sure to tune in next time. We're going to have a hands-on demo with some of the products we're talking about. And if you have any feedback or suggestions of topics you'd like us to cover, don't forget to tweet us @sensortocloud. Thanks, guys.

[MUSIC PLAYING]

This video is part of a series
  • Connect
    video-playlist(87 videos)