One of the key benefits of an industrial Internet of Things-based system is predictive maintenance. Based on the metrics ingested by the connected devices, a machine learning algorithm detects anomalies to predict potential failures. This not only saves a few million dollars for enterprises but it also ensures minimal downtime involved in dealing with defective equipment.
This tutorial series will help developers apply predictive maintenance to devices connected to the PubNub Data Stream Network (DSN), a global third-party streaming network service. It will explore the concepts of the recently announced PubNub support for MQTT, along with Functions, the serverless environment of PubNub. It will highlight how to integrate with third-party cloud services such as Microsoft Azure.
The scenario for this tutorial involves using a simulated turbine that ingests key metrics such as the fan speed, vibrations, noise levels, and temperature. Each metric ingested into PubNub DSN will be sent to Azure ML to predict possible anomalies. A PubNub Function invokes the REST endpoint exposed by an Azure ML model that’s trained to predict the anomalies. When the Web Service repeatedly reports anomalies, the PubNub Function sends a message to a simulated alarm. If the alarm goes unattended for a predefined amount of time, it automatically shuts down the turbine in question.
Read the entire article at The New Stack