叶片非整阶振动的叶尖定时压缩感知重构

The compressed sensing reconstruction error analysis and comparison of 2.5 times rotating frequency blade non-synchronous vibration signals were conducted based on the condition number optimized sensor arrangement compressed sensing algorithm and the random sensor arrangement combined with the vibration equation dictionary compressed sensing reconstruction algorithm. It was found that the condition number optimized sensor arrangement algorithm was lower in amplitude accuracy than the algorithm based on the random sensor arrangement combined with the vibration equation dictionary. Under the same number of sensors, the stability of the reconstruction accuracy obtained by the generation method of the random sensor arrangement was far lower than the condition number optimized sensor arrangement algorithm. In order to improve the amplitude reconstruction accuracy of compressed sensing for non-synchronous vibration signals and obtain a more stable reconstruction error level, the condition number 1 of the recovery matrix was taken as the optimal objective. By optimizing the sensor layout through genetic variant and combining it with the vibration equation to construct a dictionary, the coefficients of the vibration equation were obtained to reconstruct non-synchronous vibration signals. Compared with the above two algorithms, the amplitude reconstruction accuracy was improved and the stability of the reconstruction error distribution was enhanced. The reconstruction accuracy and stability of three algorithms were verified through non-synchronous vibration experimental data of the compressor. The results showed that under the condition of 10 sensors, the maximum reconstruction error of the compression sensing vibration equation reconstruction method with optimized sensor arrangement was 0.021, the maximum reconstruction error of genetic optimization with small condition number was 0.025, and the maximum reconstruction error of the vibration equation with random sensor arrangement was 0.06. When the number of sensors decreased to 4, the maximum error of the vibration equation reconstruction algorithm based on sensor layout optimization for reconstructing non-synchronous vibration experimental data was 0.05, indicating that this algorithm had high reconstruction stability and accuracy.