Machine Learning Predictive Maintenance on Data in the Wild
Adrian Binding, Nicholas Dykeman, Severin Pang · 2019
In this paper, we report on our real-word experiences in forecasting machine downtime based on real-time predictions of imminent failures. Predictions are based on the use of a machine learning classification algorithm trained on historical machine data. This is constrained by the available sensor equipment. We report on our recent collaborative work with a machine builder of premium printing equipment for purposes of predictive maintenance. We describe our data analytics approach with a view towards processing unstructured data, show initial results, discuss issues and lessons learnt.