Scholarlink

<p>Silicon carbide (SiC), owing to its high hardness and brittleness, often exhibits ductile–brittle transition behavior during grinding, making accurate surface roughness prediction and parameter planning challenging. This study proposes an integrated stochastic forward–inverse framework for SiC grinding that combines Uniform Design (UD), a forward multiple linear regression (MLR) model, and a Kalman filter (KF)–based recursive inverse estimation scheme. A U₆ mixed-level uniform design was first employed to conduct six grinding experiments involving spindle speed, feed rate, and grinding depth. A statistical regression model for surface roughness (Ra) was then established, achieving a coefficient of determination of R² = 0.9967, indicating excellent predictive capability. The KF was subsequently introduced to perform uncertainty-aware recursive inverse estimation, in which the target Ra serves as the input for automatic determination of the optimal machining parameters. Validation experiments showed that the identified optimal parameters (3600&#xa0;rpm, 50&#xa0;mm/min, 20&#xa0;μm) produced an Ra of 0.059&#xa0;μm, with only a 5.53% deviation from the predicted value of 0.056&#xa0;μm. Furthermore, inverse estimations for four target roughness levels (0.1–0.4&#xa0;μm) yielded experimental deviations ranging from 3.8% to 9.8%, demonstrating the robustness of the proposed framework under stochastic grinding variability. The proposed methodology significantly reduces experimental cost, improves parameter search efficiency, and provides a practical pathway toward intelligent and optimized SiC grinding process planning.</p>

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

Integrated forward–inverse framework for surface roughness prediction in SiC grinding using uniform design and Kalman filtering

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Tsung-Pin Hung,
  • Chao-Ming Hsu,
  • Wen-Yang Li,
  • Ah-Der Lin

摘要

Silicon carbide (SiC), owing to its high hardness and brittleness, often exhibits ductile–brittle transition behavior during grinding, making accurate surface roughness prediction and parameter planning challenging. This study proposes an integrated stochastic forward–inverse framework for SiC grinding that combines Uniform Design (UD), a forward multiple linear regression (MLR) model, and a Kalman filter (KF)–based recursive inverse estimation scheme. A U₆ mixed-level uniform design was first employed to conduct six grinding experiments involving spindle speed, feed rate, and grinding depth. A statistical regression model for surface roughness (Ra) was then established, achieving a coefficient of determination of R² = 0.9967, indicating excellent predictive capability. The KF was subsequently introduced to perform uncertainty-aware recursive inverse estimation, in which the target Ra serves as the input for automatic determination of the optimal machining parameters. Validation experiments showed that the identified optimal parameters (3600 rpm, 50 mm/min, 20 μm) produced an Ra of 0.059 μm, with only a 5.53% deviation from the predicted value of 0.056 μm. Furthermore, inverse estimations for four target roughness levels (0.1–0.4 μm) yielded experimental deviations ranging from 3.8% to 9.8%, demonstrating the robustness of the proposed framework under stochastic grinding variability. The proposed methodology significantly reduces experimental cost, improves parameter search efficiency, and provides a practical pathway toward intelligent and optimized SiC grinding process planning.