Scholarlink

<p>Accurate and reliable combustion state monitoring is a key requirement for the development of future rocket-based combined cycle (RBCC) engines.&#xa0;The rapid advancements in deep learning technology have rendered data-driven combustion state sensing a possibility, thus contributing to the realization of intelligent and efficient combustion organization. This paper proposes a multi-path convolutional neural network model that is suitable for&#xa0;the reconstruction of two-dimensional temperature fields. The impact of diverse model architectures on the precision of reconstruction outcomes is examined. The results of the reconstruction of the entire test set demonstrate that the MPFC-CNN model exhibits superior accuracy and extrapolation generalization ability compared to the SP-CNN and MPU-CNN models. The overall test dataset demonstrates an average reconstruction error of 2.84%, a linear correlation coefficient of 0.9901, and a structural similarity index of 0.8842. Validation of the reconstruction was conducted for additional combustor temperature fields with varying strut placements. The reconstruction results also basically largely satisfied the requirements. Additionally, the MPFC-CNN model has fewer parameters, which can provide a reliable basis for combustion state recognition and monitoring.</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

Investigation on temperature field reconstruction for the forced ignition process of rocket-based combined cycle engine

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月25日

查看全文
  • Yi Gao,
  • Baocong Ge,
  • Bing Liu,
  • Shaohua Zhu,
  • Fei Qin,
  • Jian An

摘要

Accurate and reliable combustion state monitoring is a key requirement for the development of future rocket-based combined cycle (RBCC) engines. The rapid advancements in deep learning technology have rendered data-driven combustion state sensing a possibility, thus contributing to the realization of intelligent and efficient combustion organization. This paper proposes a multi-path convolutional neural network model that is suitable for the reconstruction of two-dimensional temperature fields. The impact of diverse model architectures on the precision of reconstruction outcomes is examined. The results of the reconstruction of the entire test set demonstrate that the MPFC-CNN model exhibits superior accuracy and extrapolation generalization ability compared to the SP-CNN and MPU-CNN models. The overall test dataset demonstrates an average reconstruction error of 2.84%, a linear correlation coefficient of 0.9901, and a structural similarity index of 0.8842. Validation of the reconstruction was conducted for additional combustor temperature fields with varying strut placements. The reconstruction results also basically largely satisfied the requirements. Additionally, the MPFC-CNN model has fewer parameters, which can provide a reliable basis for combustion state recognition and monitoring.