J. Eur. Opt. Society-Rapid Publ. 21, 30( 2025) 321
While the calculated Pearson correlation coefficient of 0.315 confirms a measurable trend, the absolute reliability of this correlation is affected by the interferometric and alignment uncertainties. Among all error sources, the interferometric system – due to cavity limitations and repositioning requirements – represents the dominant contributor to measurement inaccuracy. The overall experimental setup is sufficient for trend validation but not yet optimized for precise predictive modeling. Future improvements should focus on increasing metrology integration, real-time data capture, and pressure monitoring to further reduce variance and improve reliability.
Compared to the reviewed literature, the dominant sources of uncertainty in this study differ notably in nature and origin. While Killinger and Thiess also experienced spatial misalignment and interferometric deviation when transitioning from manual to robotic polishing, their setup benefited from integrated measurement in a closed robotic cell. As such, their positioning error was less influenced by re-clamping or cavity limitations but still showed a significant initial jump in form deviation. In contrast, the present work relies on external metrology, making the cavity size and spatial referencing the most critical contributors to error.
Veselý et al. focused on motion modeling rather than form measurement, with error mainly arising from sensor noise and kinematic simplifications. Their system did not include removal data, and therefore the error had no direct impact on surface accuracy evaluation. Similarly, Chen’ s iteration-based controller addresses motion control accuracy but does not involve physical material removal or surface metrology, making their error domain primarily dynamic and internal.
The more recent feedback-oriented works by Li and Zhou integrate force and compliance sensors, allowing direct process monitoring and thus reducing external measurement dependency. However, these systems require substantial hardware integration, which the current study deliberately avoids. The reinforcement learning approach by Cramer introduces simulation-based error reduction but remains mostly in virtual environments, where physical measurement limitations such as interferometer cavity size are not present.
In summary, the primary error source in this work – interferometric measurement constraints due to part size – is specific to the production-oriented setup and reflects the challenge of transferring academic control concepts into practical environments. While other studies optimize for dynamic or control accuracy, this investigation reveals the practical importance of integrating suitable metrology into the process chain. The findings underline the need for improved in-situ measurement or larger-cavity systems if model-based polishing control is to be scaled to industrial optics manufacturing.
Funding
The work is part of a funded by ASML Germany GmbH. The authors would like to thank the company for funding this project. This research received no external funding.
Conflicts of interest The authors have nothing to disclose
Data availability statement
Data associated with this article cannot be disclosed due to legal reason.
Author contribution statement
Conceptualization, M. S., Methodology, M. S..; Software, A. S., U. A. and M. S.; Validation, M. S.; Writing – Original Draft Preparation, M. S., Writing – Review & Editing, J. M.; Project Administration, M. S.; Funding Acquisition, J. M.
References
1 Cook LM, Chemical processes in glass polishing, J. Non- Cryst. Solids 120( 1 – 3), 152 – 171( 1990). https:// doi. org / 10.1016 / 0022-3093( 90) 90200-6.
2 Evans CJ, Paul E, Dornfeld DA, Lucca DA, Byrne G, Tricard M, Klocke F, Dambon O, Mullany BA, Material removal mechanisms in lapping and polishing, CIRP Ann. 52( 2), 611 – 633( 2003). https:// doi. org / 10.1016 / S0007-8506( 07) 60207-8. 3 Becker E, Chemisch-Mechanische Politur von optischen Glaslinsen, Dissertation, Shaker Verlag, 2011.
4 Schneckenburger M, Machine learning Modell für die Abtragsvorhersage in der Roboter-Glaskeramik-Politur, Dissertation, Universitätsverlag Imenau, 2021. https:// doi. org / 10.22032 / dbt. 52077.
5 Preston FW, The theory and design of plate glass polishing machine, J. Soc. Glass Technol. 11( 44), 214 – 256( 1927).
6 Luo J, Dornfeld DA, Integrated modeling of chemical mechanical planarization for sub-micron IC fabrication( Springer, Heidelberg, 2004). https:// doi. org / 10.1007 / 978- 3-662-07928-7.
7 Pal RK, Garg H, Sarepaka RGV, Karar V, Experimental investigation of material removal and surface roughness during optical glass polishing, Mater. Manuf. Process. 31, 1613 – 1620( 2016). https:// doi. org / 10.1080 / 10426914.2015. 1103867.
8 Vesely L, Matousek O, Vit T, Mueller M, Contribution on the kinematics of the polishing process on a polishing machine with horizontal overarm, EPJ Web Conf 264, 01047( 2022). https:// doi. org / 10.1051 / epjconf / 202226401047. 9 Kaller A, Elementarvorgänge im Wirkspalt beim Polieren,
( Bauwesenverlag, 1980), p. 34 – 40.
10 Steinhilper W, Hennerici H, Britz S, Kinematische Grundlagen ebener Mechanismen und Getriebe( Vogel Fachbuch, Würzburg, 1993).
11 Chen ZH, Wang Y, Ouyang P, Huang J, Zhang W, A novel iteration-based controller for hybrid machine systems, Robotica 29, 317 – 324( 2011). https:// doi. org / 10.1017 / S0263574710000159.
12 Sun Y, Ge W, Zheng J, Dong D, Design and evaluation of a prosthetic knee joint using the geared five-bar mechanism, IEEE Trans. Neural Syst. Rehabil. Eng. 23( 6), 1031 – 1038( 2015). https:// doi. org / 10.1109 / tnsre. 2015.2401042.
13 Killinger S, Thiess H, Setting up an industrial robot for automated overarm polishing( SPIE, 2024), p. 13221. https:// doi. org / 10.1117 / 12.3033079.
14 Veselý L, Matousek O, Vit T, Novosad J, Optimization of the polishing process on a polishing machine with horizontal