基于压缩感知的超燃冲压发动机欠采样试验数据处理方法研究

项目名称： 基于压缩感知的超燃冲压发动机欠采样试验数据处理方法研究

项目编号： No.51306012

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 能源与动力工程

项目作者： 童晓艳

作者单位： 北京航空航天大学

项目金额： 25万元

中文摘要： 本项目针对超燃冲压发动机试验测量点多、数据量大、难以采用高频测量以及基于拟合和插值的数据处理方式可能遗漏有用信息的问题，采用基于压缩感知方法，对超燃冲压发动机试验中的欠采样数据进行稀疏分解，再通过求解稀疏优化问题进行恢复，从有限的采样点中高概率地准确重建原始信号。 本项目将利用小波基、傅里叶变换基、过完备基等稀疏字典分析超燃冲压发动机试验数据的稀疏性，建立数据特征和采样次数间的内在数学关系；基于随机矩阵的受限正交条件和非相干性条件，从数学上证明确定性测量矩阵在压缩感知中的可行性，并确定适用于超燃冲压发动机试验数据采集的测量矩阵；将超燃冲压发动机设计、计算与试验中的先验知识与基于压缩感知的试验数据处理相融合，建立融合先验的压缩感知恢复算法模型。 本项目将提出超燃冲压发动机试验的采样次数准则，为试验设计提供科学依据；试验数据处理的新方法将解决采样空白区域数据的可靠构建问题。

中文关键词： 压缩感知；稀疏分解；超燃冲压发动机；火箭发动机；数据重构

英文摘要： There are a lot of measurements and data in scramjet experiments. High frequency measurement is also a challenge. The traditional data processing such as interpolation or regression may ignore useful information. This research project will process the undersampling scramjet experimental data by using the compressed sensing method. The sparse decomposition and sparse optimizaiton will be used to accurately reconstructure the original data of scramjet from limited samples. The wavelet basis, Fourier transform basis and over complete basis will be used to analyze the sparsity of scramnet experimental data. The internal relation between the data characters and the sample numbers will be built. Based on the restricted isometric property and incoherence of stochastic matrix, the feasibility of the deterministic measurement matrix will be verified and designed for scramjet measurements. The prior information on scramjet design, calculation and experiment will be coupled to data processing based on compressed sensing. The recovery model coupled compressed sensing and prior information will be built. This project will propose the sampling rule on scramjet experiments to provide the scientific foundation for experiment design. The novel data processing method will provide solution to data reliable recovery in non-

英文关键词： Compressed sensing；Sparse decomposition；Scramjet engine；Rocket engine；Data reconstruction