One of the most valuable applications of AI is certainly predictive maintenance – the forward-looking servicing of machines. The neural analysis of data streams, usually from multiple sensors, makes it possible to detect critical system states at an early stage and to intervene before any damage actually occurs. In this article, we will provide not only background information but also a practical application of the method. We will use an inexpensive development board and example software from NXP, as well as an acceleration sensor, to detect anomalies in the operation of a fan.