A surface roughness prediction method based on physical guidance and hybrid integrated driving

By combining a hybrid ensemble approach that integrates physical constraint neural networks and multiple data-driven models, the robustness of surface roughness prediction under data noise and unseen working conditions is solved, achieving high-precision and highly robust prediction results and improving the interpretability and stability of the model.

CN122432470APending Publication Date: 2026-07-21SHENYANG GOLDING NC & INTELLIGENCE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENYANG GOLDING NC & INTELLIGENCE TECH CO LTD
Filing Date
2025-11-03
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing surface roughness prediction methods are not robust enough in the face of data noise and unseen operating conditions, and pure data-driven models lack physical interpretability, making it difficult to guide process optimization.

Method used

A physical-guided and hybrid ensemble-driven approach is adopted, which uses physical constraints and standard machine learning models to train and integrate them in parallel. By using physical laws as 'hard constraints' and combining them with the powerful fitting ability of data-driven models, the final surface roughness prediction value is generated.

Benefits of technology

While maintaining high accuracy, it significantly improves the model's robustness and generalization ability in complex industrial environments, and enhances the interpretability and stability of the prediction results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122432470A_ABST
    Figure CN122432470A_ABST
Patent Text Reader

Abstract

The application discloses a surface roughness prediction method and system based on physical guidance and hybrid integration, and the method comprises the following steps: acquiring cutting process parameters of a workpiece to be predicted; processing the process parameters through a physical constraint neural network (PGNN) to generate a physically guided prediction sub-result, wherein the PGNN embeds a surface roughness physical theory formula as a structured constraint in a forward propagation calculation process thereof; in parallel, processing the process parameters through at least one standard machine learning model to generate a data-driven prediction sub-result; and integrating the physically guided prediction sub-result and the data-driven prediction sub-result to obtain a final surface roughness prediction value. By hard-constraining physical laws in a neural network structure and combining with integrated learning, the robustness and generalization ability of the prediction model under noise interference and unseen working conditions are significantly improved, and reliable technical support is provided for industrial intelligent manufacturing.
Need to check novelty before this filing date? Find Prior Art