Scholarlink

<p>Optimization plays a critical role in modern manufacturing industries for both productivity and quality considerations. However, manufacturers still face challenges in achieving stable and efficient milling operations due to complex trade-offs among productivity, machining cost, tool life, and part quality. Traditional optimization approaches, such as metaheuristic algorithms and nonlinear programming (NLP) techniques, often suffer from high computational cost, lack of physical interpretability, and limited scalability when multiple constraints are involved. Addressing these limitations, this study introduces a hybrid method based on Bayesian Optimization combined with Gaussian Process Regression, tailored to meet industry-oriented production demands. In this respect, the optimization process aims to achieve two primary objectives, which are maximizing the Material Removal Rate (MRR) and minimizing the machining cost, while also incorporating critical constraints related to power consumption, torque limit, chatter stability, and tool breakage. For the finishing process, form error was considered a key constraint due to its impact on dimensional quality, and tool life was incorporated to ensure each part is machined with a single cutting tool for consistency. Cutting forces—used to estimate power, torque, form error, and bending stress—were predicted through a physics-based ML model that integrates an analytical force model with machine learning algorithms, reducing dependency on large experimental datasets. The proposed method demonstrated a performance improvement of 18–26%, optimizing for MRR and machining cost while satisfying all the mentioned constraints. In addition, sensitivity analysis using the Shapley value algorithm provided interpretable insights into the influence of input parameters. This innovative approach addresses current industry needs for smarter, faster, and more reliable process optimization in milling, making it highly applicable in digital twin and adaptive manufacturing environments.</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

Machine learning-based optimization to achieve stable milling conditions with maximum productivity

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Farzad Pashmforoush,
  • Arash Ebrahimi Araghizad,
  • Gamze Balcik,
  • Erhan Budak

摘要

Optimization plays a critical role in modern manufacturing industries for both productivity and quality considerations. However, manufacturers still face challenges in achieving stable and efficient milling operations due to complex trade-offs among productivity, machining cost, tool life, and part quality. Traditional optimization approaches, such as metaheuristic algorithms and nonlinear programming (NLP) techniques, often suffer from high computational cost, lack of physical interpretability, and limited scalability when multiple constraints are involved. Addressing these limitations, this study introduces a hybrid method based on Bayesian Optimization combined with Gaussian Process Regression, tailored to meet industry-oriented production demands. In this respect, the optimization process aims to achieve two primary objectives, which are maximizing the Material Removal Rate (MRR) and minimizing the machining cost, while also incorporating critical constraints related to power consumption, torque limit, chatter stability, and tool breakage. For the finishing process, form error was considered a key constraint due to its impact on dimensional quality, and tool life was incorporated to ensure each part is machined with a single cutting tool for consistency. Cutting forces—used to estimate power, torque, form error, and bending stress—were predicted through a physics-based ML model that integrates an analytical force model with machine learning algorithms, reducing dependency on large experimental datasets. The proposed method demonstrated a performance improvement of 18–26%, optimizing for MRR and machining cost while satisfying all the mentioned constraints. In addition, sensitivity analysis using the Shapley value algorithm provided interpretable insights into the influence of input parameters. This innovative approach addresses current industry needs for smarter, faster, and more reliable process optimization in milling, making it highly applicable in digital twin and adaptive manufacturing environments.