Identifying Benchmarks for Failure Prediction in Industry 4.0

Industry 4.0 is characterized by the availability of sensors to operate so-called intelligent factory. Predictive maintenance, in particular, failure prediction, an important issue cut costs associated with production breaks. We studied more than 40 publications on predictive maintenance. point out that they focus various machine learning algorithms rather selection suitable datasets. In fact, most consider a single, usually non-public, benchmark. More benchmarks are needed design and test generality proposed approaches. This paper first define requirements these benchmarks. It highlights there only two can be used for supervised among six publicly available ones we found literature. also illustrate how such benchmark deep successfully train evaluate prediction model. raise several perspectives research.