Digital-Twin-Guided Reinforcement Learning for Adaptive Visual Inspection Systems
DOI:
https://doi.org/10.36520/joofin.v1.i1.5Keywords:
Reinforcement LearningAbstract
Industrial visual inspection must balance defect-detection quality, cycle time, and inspection cost under changing product and imaging conditions. Digital twins provide a risk-free environment for testing control policies, while reinforcement learning can adapt decisions to observed state. This paper proposes VisionTwin-RL, a digital-twin-guided inspection policy that selects among fast, balanced, and high-fidelity acquisition modes according to estimated defect risk and image degradation. A controlled nine-state digital twin was constructed from three defect-risk levels and three blur levels. A tabular reinforcement-learning agent was trained for 5,500 episodes using a reward that combines detection probability, inspection cost, and the consequence of a missed defect. The learned policy was compared with two static inspection strategies. VisionTwin-RL achieved a higher detection–cost balance and the highest mean reward in the simulated evaluation. The results demonstrate the computational value of state-dependent acquisition rather than a fixed inspection setting. The study is a proof-of-concept digital-twin experiment; the detection probabilities are controlled simulation parameters and must be replaced by measurements from a real vision model and physical inspection cell before industrial deployment claims are made.
Downloads
References
[1] M. P. Ramos and A. Busboom, “Integration of digital twins with reinforcement learning in industry: A systematic review,” in Proc. IEEE ETFA, 2025, doi: 10.1109/ETFA65518.2025.11205657.
[2] P. Nguyen, M. Kim, E. Nichols, and H.-S. Yoon, “AI-driven digital twins for manufacturing: A review across hierarchical manufacturing system levels,” Sensors, vol. 26, no. 1, art. 124, 2026, doi: 10.3390/s26010124.
[3] Y. M. Alginahi, O. Sabri, and W. Said, “Reinforcement learning for industrial automation: A comprehensive review of adaptive control and decision-making in smart factories,” Machines, vol. 13, no. 12, art. 1140, 2025, doi: 10.3390/machines13121140.
[4] S. Chen et al., “Enhancing digital twins with deep reinforcement learning: A use case in maintenance prioritization,” in Proc. Winter Simulation Conf., 2024, doi: 10.1109/WSC63780.2024.10838867.
[5] H. A. Faqeer and S. H. Khajavi, “Digital twin and computer vision combination for manufacturing and operations: A systematic literature review,” Applied Sciences, vol. 15, no. 18, art. 10157, 2025, doi: 10.3390/app151810157.
[6] D. Y. Patrashko and V. Gurau, “Machine learning-powered vision for robotic inspection in manufacturing: A review,” Sensors, vol. 26, no. 3, art. 788, 2026, doi: 10.3390/s26030788.
[7] P. Link et al., “Real-time quality prediction and local adjustment of friction with digital twin in sheet metal forming,” Robotics and Computer-Integrated Manufacturing, vol. 91, art. 102848, 2025, doi: 10.1016/j.rcim.2024.102848.
[8] X. Fu, S. Li, H. Song, and Y. Lu, “Digital twin-driven multi-scale characterization of machining quality: Current status, challenges, and future perspectives,” Robotics and Computer-Integrated Manufacturing, vol. 93, art. 102902, 2025, doi: 10.1016/j.rcim.2024.102902.
[9] M. H. Zafar, E. F. Langås, and F. Sanfilippo, “Exploring the synergies between collaborative robotics, digital twins, augmentation, and Industry 5.0 for smart manufacturing: A state-of-the-art review,” Robotics and Computer-Integrated Manufacturing, vol. 89, art. 102769, 2024, doi: 10.1016/j.rcim.2024.102769.
[10] S. Wang et al., “A deep learning-enhanced digital twin framework for improving safety and reliability in human–robot collaborative manufacturing,” Robotics and Computer-Integrated Manufacturing, vol. 85, art. 102608, 2024, doi: 10.1016/j.rcim.2023.102608.
[11] H. Ni et al., “Digital twin-driven virtual commissioning for robotic machining enhanced by machine learning,” Robotics and Computer-Integrated Manufacturing, vol. 93, art. 102908, 2025, doi: 10.1016/j.rcim.2024.102908.
[12] X. Liang et al., “Adaptive shape imitation and selective semantic guidance for industrial surface defect detection,” Expert Systems with Applications, vol. 279, art. 127334, 2025, doi: 10.1016/j.eswa.2025.127334.
[13] C. Wang, X. Wei, and X. Jiang, “An automated defect detection method for optimizing industrial quality inspection,” Engineering Applications of Artificial Intelligence, vol. 127, art. 107387, 2024, doi: 10.1016/j.engappai.2023.107387.
[14] D. Bai et al., “Surface defect detection methods for industrial products with imbalanced samples: A review of progress in the 2020s,” Engineering Applications of Artificial Intelligence, vol. 130, art. 107697, 2024, doi: 10.1016/j.engappai.2023.107697.
[15] “Robust weakly supervised product surface defect segmentation based on guided cropping and inpainting extension,” Applied Soft Computing, vol. 184, art. 113661, 2025, doi: 10.1016/j.asoc.2025.113661.
[16] Q. Qin et al., “Robot digital twin systems in manufacturing: Technologies, applications, trends and challenges,” Robotics and Computer-Integrated Manufacturing, vol. 97, art. 103103, 2026, doi: 10.1016/j.rcim.2025.103103.
[17] “Micro-defect Varifocal Network: Channel attention and spatial feature fusion for turbine blade surface micro-defect detection,” Engineering Applications of Artificial Intelligence, vol. 133, art. 108075, 2024, doi: 10.1016/j.engappai.2024.108075.
[18] “Test-time feature caching network for cross-domain multilayer ceramic capacitors defect detection,” Applied Soft Computing, vol. 179, art. 113380, 2025, doi: 10.1016/j.asoc.2025.113380.
[19] C. Latsou et al., “A unified framework for digital twin development in manufacturing,” Advanced Engineering Informatics, vol. 62, art. 102567, 2024, doi: 10.1016/j.aei.2024.102567.
[20] C. Zhang et al., “Hybrid mechanism and data-driven digital twin model for assembly quality traceability and optimization of complex products,” Advanced Engineering Informatics, vol. 62, art. 102707, 2024, doi: 10.1016/j.aei.2024.102707.
[21] A. Gautam et al., “IIoT-enabled digital twin for legacy and smart factory machines with LLM integration,” Journal of Manufacturing Systems, vol. 80, pp. 511–523, 2025, doi: 10.1016/j.jmsy.2025.03.022.
[22] J.-P. Kaiser et al., “Adaptive acquisition planning for visual inspection in remanufacturing using reinforcement learning,” Journal of Intelligent Manufacturing, vol. 36, pp. 4867–4893, 2025, doi: 10.1007/s10845-024-02478-0.
[23] X. Fu, H. Song, S. Li, and Y. Lu, “Digital twin technology in modern machining: A comprehensive review of research on machining errors,” Journal of Manufacturing Systems, vol. 79, pp. 134–161, 2025, doi: 10.1016/j.jmsy.2025.01.005.
[24] B. Kang, C. Park, H. Kim, and S. Hong, “A digital twin calibration for an automated material handling system in a semiconductor fab,” Journal of Manufacturing Systems, vol. 80, pp. 1013–1028, 2025, doi: 10.1016/j.jmsy.2025.04.015.
[25] X. Tong, J. Bao, and F. Tao, “Co-evolutionary digital twins: A multidimensional dynamic approach to digital engineering,” Advanced Engineering Informatics, vol. 61, art. 102554, 2024, doi: 10.1016/j.aei.2024.102554.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Ahmad Taufi Al Afkari Siahaan (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
