Scholarlink

<p>This paper introduces an auxiliary-information-based conditional aggregate mean (MD-CAM) chart for efficient and robust process monitoring, especially useful when auxiliary variables exhibit instability. The MD-CAM chart integrates a modified difference statistic with a conditional resetting mechanism to dynamically manage shifts in auxiliary data. A key innovation is the development of a tuned XGBoost machine-learning model, which optimally selects the auxiliary weight to balance detection sensitivity and false alarm robustness. The Robbins–Monro algorithm is subsequently employed to precisely calibrate control limits, guaranteeing targeted average run-length performance under adverse auxiliary conditions. Extensive simulation results reveal that the proposed MD-CAM chart consistently outperforms existing methods particularly under shifts in auxiliary variables. The paper further explores Phase I estimation effects, providing practical guidance for optimal data collection and parameter estimation. A comprehensive case study using real-time particulate matter (PM<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(_{10}\)</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mn>10</mn> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation>) air-quality data highlights the practical utility of the MD-CAM chart, demonstrating improved early detection of shifts and reduced false alarms. An interactive web application accompanies this work, facilitating practical implementation and real-time parameter tuning of the proposed chart.</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

An XGBoost-optimized conditional aggregate mean chart for enhanced air-quality surveillance

  • 期刊论文

  • 发表日期：

    2026年2月28日

查看全文
  • Nasir Abbas

摘要

This paper introduces an auxiliary-information-based conditional aggregate mean (MD-CAM) chart for efficient and robust process monitoring, especially useful when auxiliary variables exhibit instability. The MD-CAM chart integrates a modified difference statistic with a conditional resetting mechanism to dynamically manage shifts in auxiliary data. A key innovation is the development of a tuned XGBoost machine-learning model, which optimally selects the auxiliary weight to balance detection sensitivity and false alarm robustness. The Robbins–Monro algorithm is subsequently employed to precisely calibrate control limits, guaranteeing targeted average run-length performance under adverse auxiliary conditions. Extensive simulation results reveal that the proposed MD-CAM chart consistently outperforms existing methods particularly under shifts in auxiliary variables. The paper further explores Phase I estimation effects, providing practical guidance for optimal data collection and parameter estimation. A comprehensive case study using real-time particulate matter (PM \(_{10}\) 10 ) air-quality data highlights the practical utility of the MD-CAM chart, demonstrating improved early detection of shifts and reduced false alarms. An interactive web application accompanies this work, facilitating practical implementation and real-time parameter tuning of the proposed chart.