AI-based identification of connectors

Object detection as a starting point for the digital twin

Abstract

We describe a method for automatically generating the digital twin of a connector system in order to address existing challenges for production arising due to digitization. A system is presented that enables the identification of components and their specific features using hierarchical object detection strategies based on deep learning. The recognised components are linked
to the semantic knowledge of the company and a procedure is explained for deriving the digital connector twin. Finally, we present a procedure in which synthetic data is used to train object detection networks. In practice, this opens up possibilities for cost reduction. These options are discussed with reference to previously conducted experiments.