| 基于CARS-CNN-GRU的发动机尾焰红外光谱浓度求解方法 |
| 投稿时间：2025-02-27 修订日期：2025-03-30 点此下载全文 |
| 引用本文： |
| 摘要点击次数: 2 |
| 全文下载次数: 0 |
|
| 基金项目:国家自然科学基金(61602321)资助 |
|
| 中文摘要:发动机尾焰组分浓度是决定红外光谱辐射强度的重要因素之一,它不仅影响到尾焰的热力学状态,还对军事和民用航空领域的隐身技术、环境监测以及燃烧效率评估等方面有着深远的影响。文中创新性地提出了CARS-CNN-GRU发动机尾焰红外光谱浓度求解模型,该模型通过将CARS算法与CNN-GRU深度学习算法相融合,实现了对尾焰红外特征波段的精准选择与多尺度尾焰光谱特征的有效提取。在本研究中,CARS算法被用来筛选出尾焰红外光谱中的关键波长,这些波长包含了关于尾焰成分浓度的关键信息。另一方面,CNN-GRU模型则凭借其独特的架构结构,不仅能够捕捉到特征的局部细节,还能有效处理序列数据中的长程依赖问题。仿真结果显示,CARS-CNN-GRU模型相比经典模型在红外光谱浓度求解方面具有更高的精度。具体而言,对于H2O和CO2浓度求解的均方根误差(RMSE)分别低至0.0014和0.0017,R2值高达0.999和0.998,平均绝对误差(MAE)分别为0.0011和0.0014。这表明该模型不仅在准确性上有了显著提升,而且在稳定性和可靠性方面也表现出色。 |
| 中文关键词:发动机尾焰 红外光谱 CARS-CNN-GRU模型 浓度求解 |
| |
| Engine tail flame based on CARS-CNN-GRU Infrared spectral concentration solving method |
|
|
| Abstract:The concentration of engine exhaust plume components is one of the critical factors determining the intensity of infrared spectral radiation. It not only influences the thermodynamic state of the plume but also has far-reaching implications for stealth technology in military and civilian aviation, environmental monitoring, and combustion efficiency assessment. In this paper, an innovative CARS-CNN-GRU model for solving the concentration of engine exhaust plume infrared spectra is proposed. This model integrates the CARS (Competitive Adaptive Reweighted Sampling) algorithm with the CNN-GRU (Convolutional Neural Network-Gated Recurrent Unit) deep learning algorithm, achieving precise selection of infrared feature bands in the plume and effective extraction of multi-scale spectral features. In this study, the CARS algorithm is employed to screen out key wavelengths from the infrared spectrum of the exhaust plume, which contain crucial information about the concentration of plume components. On the other hand, the CNN-GRU model, leveraging its unique architectural structure, not only captures local details of the features but also effectively handles long-term dependencies in sequential data. Simulation results show that the CARS-CNN-GRU model exhibits higher accuracy in solving the concentration of infrared spectra compared to classical models. Specifically, for H2O and CO2 concentration determination, the root mean square error (RMSE) is as low as 0.0014 and 0.0017, respectively, with R2 values reaching 0.999 and 0.998, and mean absolute errors (MAE) of 0.0011 and 0.0014, respectively. These results indicate that the model not only achieves significant improvements in accuracy but also excels in stability and reliability. This advancement represents a powerful analytical tool that promises to enhance real-time monitoring of engine emissions, support the enforcement of environmental regulations, and aid in the development of cleaner and more efficient combustion technologies. Overall, the CARS-CNN-GRU model marks a significant step forward in addressing complex spectral data analysis challenges. |
| keywords:Engine exhaust plume Infrared spectroscopy CARS-CNN-GRU Model Concentration determination |
| HTML 查看/发表评论 下载PDF阅读器 |
|
|
|
|
|