The continuous and efficient monitoring of humans and machines demands sensors with edge intelligence, but for predictive analytics, the intelligence cannot be limited to a recognition and decision logic. It must also include a learning logic capable of modeling the novelties without interrupting the always-on recognition. Indeed new objects or events represent critical information. Their detection cannot just call for a warning or actuation signal. Recording them would generate too much data, and it is not possible on a chipset affixed to a ball bearing or inside a wrist band.
This is where a NeuroMem® neural network becomes a true problem solver. Unlike a Deep Learning engine, it can admit when it does not know, which is the information of interest in predictive maintenance, and learn in real-time (including new categories) to allow the classification of novelties.
This application note explains how you can take advantage of a NeuroMem neural network to learn new conditions while running an always-on recognition, making it a unique enabler for predictive maintenance.
Read the full application note @