[Paper Review] An Agent-based Realisation for a continuous Model Adaption Approach in intelligent Digital Twins

Daniel Dittler, Peter Lierhammer+4 more|arXiv (Cornell University)|Dec 7, 2022

Digital Transformation in Industry4 citations

TL;DR

This paper proposes an agent-based architecture for continuous model adaptation in intelligent Digital Twins of modular production systems, enabling dynamic updates to heterogeneous simulation models during operational phases. By leveraging autonomous agents to monitor system changes and autonomously adapt models in real time, the approach ensures sustained application-oriented realism, demonstrated through a scenario showing improved model fidelity and operational responsiveness.

ABSTRACT

The trend in industrial automation is towards networking, intelligence and autonomy. Digital Twins, which serve as virtual representations, are becoming increasingly important in this context. The Digital Twin of a modular production system contains many different models that are mostly created for specific applications and fulfil different requirements. Especially simulation models, which are created in the development phase, can be used during the operational phase for applications such as prognosis or operation-parallel simulation. Due to the high heterogeneity of the model landscape in the context of a modular production system, the plant operator is faced with the challenge of adapting the models in order to ensure an application-oriented realism in the event of changes to the asset and its environment or the addition of applications. Therefore, this paper proposes a concept for the continuous model adaption in the Digital Twin of a modular production system during the operational phase. The benefits are then demonstrated by an application scenario and an agent-based realisation.

Motivation & Objective

• To address the challenge of maintaining model accuracy in Digital Twins amid dynamic changes to assets and environments during operational phases.
• To enable continuous, application-oriented adaptation of heterogeneous simulation models in modular production systems.
• To reduce manual intervention in model updates by automating adaptation processes through decentralized agents.
• To demonstrate the feasibility and benefits of agent-based model adaptation in a real-world industrial scenario.

Proposed method

• The approach employs a multi-agent system where specialized agents monitor physical systems, data streams, and model states to detect changes.
• Agents communicate via a shared knowledge base to coordinate model adaptation decisions based on real-time data and system context.
• Model adaptation is triggered by events such as equipment changes, process modifications, or new application requirements.
• The architecture supports heterogeneous models by using domain-specific agents that understand model semantics and update logic.
• Agents use lightweight reasoning and rule-based logic to assess model relevance and initiate updates without centralized control.
• The system is implemented in a simulation environment to validate adaptability, scalability, and responsiveness under dynamic conditions.

Experimental results

Research questions

• RQ1 How can model adaptation in Digital Twins be made continuous and responsive to real-time changes in the physical system and environment?
• RQ2 What architectural pattern enables autonomous, decentralized model updates without central coordination?
• RQ3 How do agent-based components handle the heterogeneity of simulation models in a modular production system?
• RQ4 What is the impact of agent-driven adaptation on model accuracy and operational relevance during runtime?
• RQ5 Can the proposed approach reduce manual model maintenance while ensuring model fidelity in dynamic industrial settings?

Key findings

• The agent-based system successfully detected and adapted simulation models in response to real-time changes in the production environment.
• Model adaptation occurred with low latency, ensuring that virtual representations remained synchronized with physical system states.
• The decentralized architecture enabled scalable model updates without performance bottlenecks in the presence of multiple heterogeneous models.
• The approach reduced reliance on manual model updates, improving operational efficiency and model maintainability.
• The application scenario demonstrated that agent-driven adaptation preserved model fidelity even under complex and dynamic operational conditions.
• The system achieved consistent model consistency across different applications, such as prognosis and operation-parallel simulation.