基于时域和小波域稀疏表示的发动机状态数据压缩感知

The Compressive Sensing of Engine Status Data Based on Signal Sparse Representation in Time Domain and Wavelet Domain

提出了基于统计分析的时域信号稀疏化算法,对发动机的转速等在时域内表现出一定聚集性的信号进行稀疏化处理;利用提升小波变换对油温、水温等在时域内表现出趋势性的信号进行稀疏化处理,并对小波系数进行阈值量化,得到了发动机状态数据的稀疏表示。采用工程上更易实现的广义循环测量矩阵对稀疏信号进行采样观测。对5种实测信号的压缩重构效果表明,重构误差在0.02以内,信号的压缩比均在0.3以内,有良好的压缩效果。

The time-domain signal sparse algorithm based on the statistical analysis was proposed in order to process engine speed signal with time-domain aggregation characteristics.The lifting wavelet transform was used for the sparse processing of trend signals such as the oil temperature and water temperature and the threshold quantization of wavelet coefficients was carried out.Finally,the sparse representation of engine status data was acquired.The sparse signal was sampled and observed with the generalized cyclic measurement matrix which can be realized easily in engineering.The results of compression and reconstruction for 5kinds of test signals show that the reconstruction error is less than 0.02and the signal compression ratio is within 0.3.