Scholarlink

<p>To effectively monitor nonlinear multivariate processes, this paper proposes a robust Multivariate Adaptive Cumulative Sum (MACUSUM) control chart utilizing Support Vector Regression (SVR) residuals. The proposed method combines Principal Component Analysis (PCA) with a memory-type control mechanism to boost sensitivity to small and moderate changes in the processes. Its efficacy is assessed by Average Run Length (ARL), Standard Deviation of Run Length (SDRL), Standard Error of Run Length (SERL), and Quantile Run Length (QRL). Monte Carlo simulations conducted in MATLAB demonstrate that the proposed method enables faster, more accurate detection of aberrant signals. It effectively captures essential information from operational datasets to wind turbines in energy production systems, part manufacturing datasets concerning operators’ performance, and film thickness measurements in semiconductor manufacturing. The MACUSUM method, built on PCA, aims to enhance process performance in energy production by utilizing wind turbine output data. Additionally, MACUSUM, in combination with SVR, is designed to improve industrial processes. This methodology focuses on capturing operators’ performance by applying it to Part-manufacturing data and Film-thickness datasets in semiconductor manufacturing systems.</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

A multivariate adaptive cumulative sum control chart based on machine learning approach, support vector regression with applications

  • 期刊论文

  • 发表日期：

    2026年7月7日

查看全文
  • Fozia Tauqeer,
  • Babar Zaman,
  • Muhammad Riaz,
  • Irshad Ahmad Arshad

摘要

To effectively monitor nonlinear multivariate processes, this paper proposes a robust Multivariate Adaptive Cumulative Sum (MACUSUM) control chart utilizing Support Vector Regression (SVR) residuals. The proposed method combines Principal Component Analysis (PCA) with a memory-type control mechanism to boost sensitivity to small and moderate changes in the processes. Its efficacy is assessed by Average Run Length (ARL), Standard Deviation of Run Length (SDRL), Standard Error of Run Length (SERL), and Quantile Run Length (QRL). Monte Carlo simulations conducted in MATLAB demonstrate that the proposed method enables faster, more accurate detection of aberrant signals. It effectively captures essential information from operational datasets to wind turbines in energy production systems, part manufacturing datasets concerning operators’ performance, and film thickness measurements in semiconductor manufacturing. The MACUSUM method, built on PCA, aims to enhance process performance in energy production by utilizing wind turbine output data. Additionally, MACUSUM, in combination with SVR, is designed to improve industrial processes. This methodology focuses on capturing operators’ performance by applying it to Part-manufacturing data and Film-thickness datasets in semiconductor manufacturing systems.