Autonomous Model Lifecycle Management for Digital Twin-Based Manufacturing Control

Zhengyang; Gu; Thomas Cook; Fredaljohn Rohrbaugh; Joseph E. Hernandez; Chris Couch

cs.LGSep 19, 2026

Autonomous Model Lifecycle Management for Digital Twin-Based Manufacturing Control

Authors: Zhengyang, Gu, Thomas Cook, Fredaljohn Rohrbaugh, Joseph E. Hernandez, Chris Couch

Organizations: Cissy · Liveline Technologies, Livonia, MI, USA

Abstract

Manufacturing AI systems must autonomously adapt to continuous distributional shift from raw-material variability, ambient changes, and equipment aging, under strict safeguard and operator-trust requirements where model failures risk physical damage. This paper presents a closed-loop Cyber-Physical System (CPS) for autonomous model lifecycle management in automotive manufacturing, deployed since 2023. The system manages product-specialized model pairs: a sequence-to-sequence physics model (LPP) serving as a digital twin, and a deep Reinforcement Learning (RL) control policy (LCP) trained against it. Per retraining cycle, multiple model variants spanning architecture families and RL algorithms compete; only the best-scoring candidate advances. A Conductor orchestrator autonomously manages plant-wide model inventories with dependency-aware retraining and Proportional-Integral-Derivative (PID) fallback. Reflecting the principle of Human-Centric Intelligence, the LCP composite score embeds an operator-trust gate penalizing policies deviating from established practice; without it, 23% of policies are rejected by operators despite passing accuracy thresholds. Across multiple facilities, LCP-controlled processes achieve process stability improvements of 28-45% over uncontrolled baselines with zero safety incidents.

Figures & tables

Fig. 1: Seven-stage closed-loop MLOps lifecycle. LPP backtest (Stage 3) gates LCP training; on failure, data packages are re-selected (max 2 attempts). If both fail, the system defaults to PID. LCP backtest (Stage 5) gates deployment. PID fallback remains active regardless of ML state.

Fig. 2: Competitive model selection. $N$ LPP variants and $M$ LCP variants are trained per cycle. Only the best-scoring candidate advances; others are discarded.

LPP Metrics	Mean $\pm$ Std
Spearman $\rho$ $\uparrow$	0.72 $\pm$ 0.13
RMdSE $\downarrow$	0.05 $\pm$ 0.02
MDA $\uparrow$	0.78 $\pm$ 0.15
$S_{\text{LPP}}$ $\downarrow$	0.20 $\pm$ 0.10
Control Metric	Uncntrl.	PID	LCP (ours)
Avg $C_{pk}$ $\uparrow$	2.63	3.05	3.37-3.81

TABLE I: LPP Accuracy and LCP Control Performance. Averaged across 3 facilities;

\uparrow

higher better,

\downarrow

lower better.

RL Control

Reinforcement learning control uses reward-driven learning to make decisions in dynamic systems, from robots and vehicles to energy and infrastructure.

See all 276 papers

FacProcessTwin: An LLM-Based System for Process Twin Development
Jun 16, 2026Yash Pulse, Yong-Bin Kang, Abhik Banerjee +1Manufacturing Human-in-the-Loop AI
Swinburne University of Technology Melbourne, Australia
A Continual Validation, Updating, and Decision-Making Framework for Self-Adaptive Digital Twins via Robust Model Predictive Control: A Case Study in Additive Manufacturing
Jul 20, 2026Yi-Ping Chen, Ying-Kuan Tsai, Vispi Karkaria +3Neural Surrogate Modeling Concept Drift
Department of Mechanical Engineering, Northwestern University · Department of Industrial Engineering and Management Science, Northwestern University
Closed-Loop Control with Rule-Aligned Small Language Models and Multi-Agent Self-Correction
Jun 24, 2026Yuchen Wang, Javal Vyas, Tong Liu +1Language Model-Based Control Small Language Models
School of Computer Science University of Sheffield Sheffield, UK · Department of Chemical Engineering Imperial College London London, UK

Jun 16, 2026J/K move · Enter open · S save

FacProcessTwin: An LLM-Based System for Process Twin Development

Yash Pulse, Yong-Bin Kang, Abhik Banerjee, Prem Prakash Jayaraman

Swinburne University of Technology Melbourne, Australia

Manufacturing Human-in-the-Loop AI Digital Twins

Open paper AskPDF arXiv

Process twins provide real-time representations of entire production processes. By capturing how process steps interact, rather than monitoring a single machine in isolation as an asset-based digital twin does, they have the potential to drive efficiency gains across the whole process. However, developing a process twin is costly. It requires accurately modelling the entire production process: its process steps, the equipment and product-specific settings each step uses, and its process variations. The resulting model must then be bound to live operational data. We present FacProcessTwin, a system that leverages a large language model (LLM) to reduce this development time, building a process twin from a plant's process documentation and natural-language input from an operator. FacProcessTwin generates this complete process model and then automatically binds its process steps to live operational data. The generated model and its data bindings are rendered as an interactive process diagram through which manufacturing personnel can monitor and correct the system's autonomous decisions, such as resolving uncertainty at safety-critical binding steps. We evaluate FacProcessTwin through a real-world case study of an Australian food manufacturer, covering 16 production process flows that span chilled, frozen, and aseptic shelf-stable product categories and include process variations within the same product. The results show that FacProcessTwin generates these process models accurately (a mean F1 of 95.2% against ground truth) and builds each twin in roughly a sixth of the manual time. Its human-in-the-loop governance then keeps the safety-critical bindings correct: at ambiguous tags where a single-pass baseline silently mis-binds 75.0% of the time, FacProcessTwin defers to the operator and mis-binds none.

Abstract

RL Control

Explore similar work

FacProcessTwin: An LLM-Based System for Process Twin Development

A Continual Validation, Updating, and Decision-Making Framework for Self-Adaptive Digital Twins via Robust Model Predictive Control: A Case Study in Additive Manufacturing

Closed-Loop Control with Rule-Aligned Small Language Models and Multi-Agent Self-Correction