Demonstrating Automated Generation of Simulation Models from Engineering Diagrams

Abstract

Digital twins are powerful tools for analysis and representation of physical systems. However, the creation of digital twins including simulation models remains a challenging and time-consuming task, requiring expertise in the domain. Previous work showed algorithms for digitizing engineering diagrams, which is a necessary step for automated simulation model generation. In this paper, we present a comprehensive pipeline for automatically generating simulation models through the digitization of engineering diagrams, exemplified by a simple hydraulic system. For the digitization, we employ several computer vision techniques and deep learning models trained on synthetic data. The demonstrator showcases necessary steps and modules to create a simulation model, and which data has to be available at each working step. Potentially, our approach accelerates the adoption and utilization of digital twin technologies, reducing the time and manual work needed to create simulation models.