PRISM-MPWM: a process world model for milling tool-wear diagnostics

A.A. Lobko

Abstract


This paper presents PRISM-MPWM (Physics-Residual Invariant Support-aware Machine Process World Model), a diagnostic process model for estimating the flank wear (VB) of an end mill. The model does not predict wear from scratch. It refines the prediction of a baseline force and torque model: a trained block predicts only the residual error of that baseline, the size of the correction is bounded, and its weight decreases when the transition is poorly supported by training data or the predictive uncertainty is high. A physical constraint forbids a decrease of accumulated wear. The model is evaluated on the public Denkena, Klemme and Stiehl milling dataset: training and tuning use machines 1 and 2, and machine 3 is reserved for the final evaluation. PRISM-MPWM reduces the one-step root-mean-square error of wear prediction from 0.486 to 0.445 µm, an 8.45% improvement over the baseline. The result is supported by a grouped bootstrap over tools, lower error in the high-wear zone, a smaller underprediction of high wear, zero monotonicity violations, prediction with abstention on difficult transitions and permutation controls. The result concerns wear diagnostics from observed signals and does not extend to cutting parameter selection.


Full Text:

PDF (Russian)

References


B. Denkena, H. Klemme, and T. H. Stiehl, “Multivariate time series data of milling processes with varying tool wear and machine tools,” Data in Brief, vol. 50, Art. no. 109574, 2023, doi: 10.1016/j.dib.2023.109574. [Online]. Available: https://doi.org/10.1016/j.dib.2023.109574

M. Assran, Q. Duval, I. Misra, P. Bojanowski, P. Vincent, M. Rabbat, Y. LeCun, and N. Ballas, “Self-supervised learning from images with a joint-embedding predictive architecture,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2023, pp. 15619–15629. [Online]. Available: https://arxiv.org/abs/2301.08243

M. Assran et al., “V-JEPA 2: Self-supervised video models enable understanding, prediction and planning,” arXiv:2506.09985, 2025. [Online]. Available: https://arxiv.org/abs/2506.09985

N. Hansen, H. Su, and X. Wang, “TD-MPC2: Scalable, robust world models for continuous control,” in Proc. Int. Conf. Learning Representations (ICLR), 2024. [Online]. Available: https://arxiv.org/abs/2310.16828

J. Wang, Y. Li, R. Zhao, and R. X. Gao, “Physics guided neural network for machining tool wear prediction,” J. Manuf. Syst., vol. 57, pp. 298–310, 2020, doi: 10.1016/j.jmsy.2020.09.005. [Online]. Available: https://doi.org/10.1016/j.jmsy.2020.09.005

A. Sokol, N. Moniz, and N. Chawla, “Conformalized selective regression,” arXiv:2402.16300, 2024. [Online]. Available: https://arxiv.org/abs/2402.16300


Refbacks

  • There are currently no refbacks.


Abava  Кибербезопасность ИТ-КОНГРЕСС ВМК МГУ 2026 СНЭ

ISSN: 2307-8162