Predict the health condition and time to failure of an industrial duct fan using MATLAB® and ThingSpeak™. Use MATLAB to develop predictive maintenance algorithms based on measured vibration data from an instrumented fan. Simulate a variety of failure conditions, including a blocked fan and a fan with dust build-up. Extract features from the vibration data and build and train a machine learning model to diagnose the different types of failure. With Predictive Maintenance Toolbox™, create a model that estimates the time to failure of the fan. To collect the vibration data, a Particle Photon with attached accelerometer is mounted to the fan. The Particle Photon is an Internet-connected device that connects to ThingSpeak over wi-fi, enabling you to stream the vibration signal to the ThingSpeak IoT analytics platform in the cloud.
The data processing, feature extraction, and training of machine learning and predictive maintenance (condition-based maintenance) models are performed offline using MATLAB tools. The same code used for offline training and the trained model are uploaded to the cloud and used to predict the condition of the fan using the built-in MATLAB Analysis app on ThingSpeak.
On ThingSpeak, you can execute predictive algorithms on the data as the data streams in. The channel display shows the current condition of the fan and can be viewed from any Internet-connected web browser or mobile device. You can also configure ThingSpeak to send an SMS and email alert when the time to failure is predicted to be less than a certain threshold value.
Quickly prototype condition-monitoring algorithms with ThingSpeak and MATLAB!
