Nieke Roos
9 May

NXP and Microsoft have developed an edge-to-cloud machine learning solution for identifying anomalous behavior in IoT and industrial equipment. Applications include predictive maintenance for rotating components, presence detection and intrusion detection. Combining the chipmaker’s offline ML capability and embedded processing with the software giant’s cloud expertise, the two companies aim to bring productivity optimization and system-level safety to users of the Azure IoT platform.

The jointly developed solution consists of a small form factor, low power system-on-module integrating NXP’s I.MX RT106C crossover microcontroller and a robust set of sensors. Running up to 600 MHz, the MCU is capable of collecting and analyzing the sensor data in real time locally at the edge. The solution seamlessly connects to the Azure IoT platform, providing users an easy way to transfer the data to the cloud.

The edge-to-cloud machine learning solution consists of a small form factor, low power system-on-module integrating NXP’s I.MX RT106C crossover microcontroller and a robust set of sensors. Photo: NXP

With the solution comes an associated Anomaly Detection Toolbox. This utilizes various ML algorithms such as Random Forest and Simple Vector Machine, to model normal behavior of devices, detect anomalous behavior through combined local and cloud mechanisms. This requires much lower cloud bandwidth requirements while maintaining full online logging and processing capabilities at a fraction of the cost.

“Preventing failures and reducing downtime are key to enhance productivity and system safety,” says Denis Cabrol, executive director and general manager of IoT and Security Solutions at NXP. “We partnered with Microsoft to combine the power of Azure IoT with local intelligence running on NXP’s embedded technology to unlock innovation for the IoT – as part of our continued efforts to bring cognitive services down to the silicon.”