IoT predictive analytics is used to predict any impending equipment failure. It utilizes predictive algorithms that leverage multivariate data collected from several IoT sensors mounted on the same equipment. For example, for a refrigeration system, the set of values could be case temperature, compressor suction pressure, voltage, etc.
The initial step is to detect if there are any abrupt changes in equipment behavior, which can be done using IoT time-series data. A new developer pattern titled “Predict equipment failure using IoT sensor data” steps through the process. Once the anomaly is confirmed using a change-point detection method, the logical next step is to predict the possibility of equipment failure in the near future. This prediction step is the focus for this pattern. Predictive algorithms implemented in Python 2.0 on the IBM Data Science Experience platform is used with data stored in a database.
All components are open-sourced and designed as modules to enable reuse of components individually or as a whole. The entire flow is made configurable so that multiple iterations of the predictive model can be run by merely reconfiguring parameters and initiating a re-run of the flow.
Check out our “Predict equipment failure using IoT sensor data” pattern and let us know what you think.