Scholarlink

This study investigates the single-objective optimization of surface roughness (Ra) in ultrasonic vibration-assisted milling (UVAM) of external cylindrical surfaces on hardened 90CrSi steel. The machining process focuses on precision finishing of cylindrical features, which are common in high-performance mechanical components and demand tight surface quality. Four key process parameters were considered: ultrasonic amplitude (A), cutting speed (Vc), feed rate (Vf), and radial depth of cut (ae). Two modeling approaches—Response Surface Methodology (RSM) and Gaussian Process Regression (GPR)—were applied to predict Ra. The RSM-based quadratic model yielded a moderate determination coefficient (R2 = 0.8674), while the GPR model demonstrated superior accuracy (R2 = 0.9976), with predictions closely matching the experimental data. The results suggest that GPR effectively captures the nonlinearities in the UVAM process and provides a powerful tool for optimizing surface quality in the precision milling of cylindrical surfaces.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Single-Objective Optimization of Surface Roughness in Ultrasonic Vibration-Assisted Milling of 90CrSi Steel Using GPR and RSM Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dinh Van Thanh,
  • Le Thu Quy,
  • Vu Ngoc Pi,
  • Pham Thanh Cuong

摘要

This study investigates the single-objective optimization of surface roughness (Ra) in ultrasonic vibration-assisted milling (UVAM) of external cylindrical surfaces on hardened 90CrSi steel. The machining process focuses on precision finishing of cylindrical features, which are common in high-performance mechanical components and demand tight surface quality. Four key process parameters were considered: ultrasonic amplitude (A), cutting speed (Vc), feed rate (Vf), and radial depth of cut (ae). Two modeling approaches—Response Surface Methodology (RSM) and Gaussian Process Regression (GPR)—were applied to predict Ra. The RSM-based quadratic model yielded a moderate determination coefficient (R2 = 0.8674), while the GPR model demonstrated superior accuracy (R2 = 0.9976), with predictions closely matching the experimental data. The results suggest that GPR effectively captures the nonlinearities in the UVAM process and provides a powerful tool for optimizing surface quality in the precision milling of cylindrical surfaces.