IoT: Maintenance

Device maintenance is probably one of the lowest hanging fruits on the IoT tree. It is easy: if a device is connected, it potentially can inform about the state of its health.  Or even if it cannot, a loss of contact with it may just be a signal that it has just died and needs to be checked / replaced. This REACTIVE - but automated - maintenance is a huge improvement when compared to non-connected devices, which have to be periodically checked by humans if they work or not. The concept is very broadly applicable, from lights in hotel rooms to soil moisture sensors.

Upping the game, the next step is PROACTIVE maintenance. We can track parameters like run-time hours and knowing a device is nearing its rated lifetime, send a service team over to replace it before it fails. Proactive, while better than reactive, is not ideal. Very often a device (or a part of it) is replaced too early and sometimes it is replaced too late, after it fails. Because rated lifetime is a projected average, that usually does not take the operating conditions and potential variations in materials or subsystems: some parts may wear faster and some may withstand more. But in most cases proactive is better than reactive.

The ultimate goal in maintenance is PREDICTIVE. Which means not just mechanically tracking run hours or miles traveled or on/off cycles. predictive means understanding symptoms indicating an upcoming failure. They may be some vibration patterns or current spikes or a combination of variety of factors. Predictive is where the real value kicks in and that requires IoT to meet big data. Connectivity allows capturing of data points Once many of them are captured and properly labelled, machine learning techniques can consume the data and produce failure patterns that can be matched against.