Causal Models and a Prototypical Implementation of a Pipeline for Causal Analysis in Manufacturing
摘要
Questions of cause and effect are fundamental to an understanding of processes. In the manufacturing context, the knowledge of causal relationships can help to avoid errors, improve processes, increase efficiency and reduce costs. In this work we pursue two aspects: Firstly, the prototypical implementation of a causal analysis software service that reveals the causal structures of the manufacturing facilities across single processes, detects the causes of errors, identifies key factors among the process parameters and estimates the effects of these factors on the outcome. Secondly, the generation of synthetic process data based on a given causal model, establishing a so-called causal emulation. Therefore, this tool helps manufacturers to simulate different scenarios and optimise operations without incurring the costs and risks associated with physical experimentation. The services developed in this work are connected to an asset administration shell to form a pipeline that is geared towards the high-rate production of electrolysers.