Smart Process Failure Analysis based on Bayesian Networks and Knowledge Graphs

In production, when it comes to quality defects or production disruptions, expert knowledge for root cause analysis is essential. This knowledge can be captured in a machine-readable format using ontologies and can then be probabilistically expanded in the next step. Bayesian networks can be employed to provide an estimate of the probability describing the cause of a production error. The aim of this work is, therefore, the development of a process diagnosis tool for an existing knowledge graph of a robot-based system, which relies on a Bayesian network to estimate potential error probabilities. For this purpose, the concept of the tool is developed and validated using the use case of robot-based glass panel assembly.