Embedding domain expertise in machine learning-based fault detection development: A collaborative framework

Modern manufacturing systems benefit from digitalisation technologies in various ways, including digital twins and artificial intelligence. Machine learning can support in the field of reactive fault detection. However, the potential of including domain expertise during development of machine-learning based fault detection is not fully explored. To enhance reactive fault detection and to encourage a collaboration between machine learning and domain experts, a collaborative framework gets introduced in this work. It embeds the machine learning lifecycle into the V-model and extends each phase by expedient domain-driven concepts, like modularisation of machine learning, based on manufacturing equipment, and digital twins. Due to its strong focus on the domain, this framework strongly emphasises the involvement of domain experts throughout the whole machine learning lifecycle, which shall lead to a more targeted development and an easier familiarisation of those, when deployment takes part. Optimisation potentials are discussed and will be further investigated in the future.