A Model-Based Approach to Automated Digital Twin Generation in Manufacturing

TL;DR

The paper addresses the need for flexible, reconfigurable manufacturing and the challenge of rapid, reliable digital twin deployment. It proposes a model-based DT generation platform that starts from AutomationML factory plans, extracts core concepts, builds a virtual world in Unity, configures a DT platform, and uses Generative AI to create production scenarios that are simulated before deployment. The architecture integrates an AML parser, Unity-based virtual twin, BPMN-based orchestration, ThingsBoard IoT, and a hybrid data store, coordinating with middleware that translates commands to ROS and Modbus TCP. Validation on a five-component production line demonstrates real-time monitoring, virtual commissioning, and automated reconfiguration, while highlighting scaling and AI semantics challenges in complex systems.

Abstract

Modern manufacturing demands high flexibility and reconfigurability to adapt to dynamic production needs. Model-based Engineering (MBE) supports rapid production line design, but final reconfiguration requires simulations and validation. Digital Twins (DTs) streamline this process by enabling real-time monitoring, simulation, and reconfiguration. This paper presents a novel platform that automates DT generation and deployment using AutomationML-based factory plans. The platform closes the loop with a GAI-powered simulation scenario generator and automatic physical line reconfiguration, enhancing efficiency and adaptability in manufacturing.

Figures (5)

• Figure 1: Digital Twin Architecture (This image has been designed using resources from Flaticon.com)
• Figure 2: Business Process Generation depicting the transitions between various stages of the methodology
• Figure 3: Physical Production line Demonstrator Environment
• Figure 4: Digital Twin of the production line
• Figure 5: Generated Process