EUVIMEDEUVIMED entdecken
Crossref

SURFACE ROUGHNESS PREDICTION USING GAUSSIAN PROCESS REGRESSION DATA AUGMENTATION BASED ON LHS SAMPLING AND MATHERON POSTERIOR

Haifeng Chen, Shan Li, Lubing Shi

Ukrainian Journal of Mechanical Engineering and Materials Science · 2026 · 12 (2) · 95-108

DOI: 10.23939/ujmems2026.02.095 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Ukrainian Journal of Mechanical Engineering and Materials Science, 2026. Verfasst von Haifeng Chen und 2 weiteren Autor:innen. Die Arbeit nennt 15 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To address the problems of limited experimental samples, complex nonlinear relationships between process parameters and surface roughness, and insufficient prediction accuracy of traditional models in gear grinding, this paper focuses on the gear grinding of 12Cr2Ni4A steel. Wheel speed, grinding depth, and ultrasonic vibration amplitude were selected as input parameters, and a Gaussian Process Regression (GPR) data augmentation framework integrating Latin Hypercube Sampling, adaptive noise scaling, and Matheron posterior sampling was constructed to generate expanded samples. Subsequently, the Multilayer Perceptron (MLP), LightGBM, and XGBoost models were employed to predict surface roughness. The results demonstrate that the proposed augmentation method can significantly improve prediction accuracy; the double-augmentation strategy exhibits better stability, and there exists an optimal range for the number of augmented samples—excessive augmentation reduces generalization performance. The MLP model shows excellent stability, while the XGBoost model achieves the highest accuracy when the number of augmented samples ranges from 1500 to 1800, with the coefficient of determination (R2) reaching 0.93. The combination of effective data augmentation and nonlinear models enables high-precision prediction of gear-grinding surface roughness under small-sample conditions.

Zitieren

APA

Chen, H., Li, S., & Shi, L. (2026). SURFACE ROUGHNESS PREDICTION USING GAUSSIAN PROCESS REGRESSION DATA AUGMENTATION BASED ON LHS SAMPLING AND MATHERON POSTERIOR. Ukrainian Journal of Mechanical Engineering and Materials Science, 12(2), 95-108. https://doi.org/10.23939/ujmems2026.02.095

Vancouver

Chen H, Li S, Shi L. SURFACE ROUGHNESS PREDICTION USING GAUSSIAN PROCESS REGRESSION DATA AUGMENTATION BASED ON LHS SAMPLING AND MATHERON POSTERIOR. Ukrainian Journal of Mechanical Engineering and Materials Science. 2026;12(2):95-108. doi:10.23939/ujmems2026.02.095

BibTeX
@article{Chen2026,
  author = {Chen, Haifeng and Li, Shan and Shi, Lubing},
  title = {SURFACE ROUGHNESS PREDICTION USING GAUSSIAN PROCESS REGRESSION DATA AUGMENTATION BASED ON LHS SAMPLING AND MATHERON POSTERIOR},
  journal = {Ukrainian Journal of Mechanical Engineering and Materials Science},
  year = {2026},
  volume = {12},
  number = {2},
  pages = {95--108},
  doi = {10.23939/ujmems2026.02.095}
}
Kostenlos starten →