SemML: Facilitating development of ML models for condition monitoring with semantics

Monitoring of the state, performance, quality operations and other parameters equipment production processes, which is typically referred to as condition monitoring, an important common practice in many industries including manufacturing, oil gas, chemical process industry. In age Industry 4.0, where aim a deep degree automation, unprecedented amounts data are generated by this enables adoption Machine Learning (ML) approaches for monitoring. Development such ML models challenging. On one hand, it requires collaborative work experts from different areas, scientists, engineers, experts, managers with asymmetric backgrounds. there high variety diversity relevant Both factors hampers modelling work, we address these challenges empowering ML-based monitoring semantic technologies. To end propose software system SemML that allows reuse generalise pipelines conditions relying on semantics. particular, has several novel components relies ontologies ontology templates task negotiation feature annotation. also instantiate parametrised annotation industrial data. With SemML, users do not need dive into scripts when new datasets studied application scenario arrive. They only annotate then will be constructed through combination reasoning modules. We demonstrate benefits Bosch use-case electric resistance welding very promising results.