Methods to Enable Evolvable Digital Twins for Flexible Automated Production Systems

Abstract

Changing requirements cause flexible automated Production Systems (aPS) to evolve over decades. Digital Twins (DT) of the different hierarchy levels and design steps ease this evolution, e.g., by enabling requirement analysis and compatibility checks ahead of any physical changes. To ensure up-to-date models and integrate additional knowledge, information gained during operation is included in DTs. Consequently, evolvability, decomposability, control software modularity, and learning during operation are identified as four requirements to achieve such evolvable DTs. Concepts to realize every requirement are introduced and exemplarily validated using a demonstrator machine. AutomationML (AML), the XML-based vendor neutral language for information modeling and exchange in between different disciplines and their tools and product classification systems like ECLASS that specify components attributes vendor neutral enable evolvability during the design phase. Decomposability is achieved by assembling DTs of components according to ISA 88 levels from control unit to facility. A control primitive concept that realizes control software modularity is introduced and validated. Based on data analytics and operation data the DT can be updated by using the versioning mechanism of AML. Thereby, the DT for the next machine generation is improved with knowledge from operation and represents the already existing machine more precisely.