Scholarlink

<p>This work assesses a POD/SVD-based sparse sensing framework for thermochemical field reconstruction in the LUCY cyclonic combustor operating under MILD methane combustion conditions. A reduced-order representation is built from CFD data and combined with QR decomposition with column pivoting (QRCP) to design sensor layouts and infer reduced coefficients from sparse measurements. Two truncation criteria are compared, the optimal hard threshold of Gavish et al. and a 99.5% cumulative variance, together with Auto-scaling preprocessing, to quantify their impact on reconstruction accuracy and out-of-sample prediction. Beyond unconstrained QRCP, practical deployment constraints are investigated through distance separation and region-limited candidate sets, and the performance of a predefined thermocouple layout is evaluated by augmenting it with additional optimally placed sensors. Results are reported for both training and testing operating conditions, complemented by an uncertainty-propagation analysis based on experimental measurement variability. The study shows that robust reconstruction is achieved with a limited number of sensors when the balance between retained rank and measurements is respected, while spatial constraints can preserve near-optimal performance and improve physical feasibility. The framework provides a computationally efficient soft-sensing strategy to support monitoring-oriented applications in confined MILD combustors.</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

Sparse Sensor Placement for a Cyclonic Burner under MILD Combustion Conditions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月9日

查看全文
  • Vincenzo Rosati,
  • Giovanni Battista Ariemma,
  • Giancarlo Sorrentino,
  • Raffaele Ragucci,
  • Mara de Joannon

摘要

This work assesses a POD/SVD-based sparse sensing framework for thermochemical field reconstruction in the LUCY cyclonic combustor operating under MILD methane combustion conditions. A reduced-order representation is built from CFD data and combined with QR decomposition with column pivoting (QRCP) to design sensor layouts and infer reduced coefficients from sparse measurements. Two truncation criteria are compared, the optimal hard threshold of Gavish et al. and a 99.5% cumulative variance, together with Auto-scaling preprocessing, to quantify their impact on reconstruction accuracy and out-of-sample prediction. Beyond unconstrained QRCP, practical deployment constraints are investigated through distance separation and region-limited candidate sets, and the performance of a predefined thermocouple layout is evaluated by augmenting it with additional optimally placed sensors. Results are reported for both training and testing operating conditions, complemented by an uncertainty-propagation analysis based on experimental measurement variability. The study shows that robust reconstruction is achieved with a limited number of sensors when the balance between retained rank and measurements is respected, while spatial constraints can preserve near-optimal performance and improve physical feasibility. The framework provides a computationally efficient soft-sensing strategy to support monitoring-oriented applications in confined MILD combustors.