A remote health condition monitoring system based on compressed sensing

Data-driven health condition monitoring has received increasing attentions. However, the bandwidth of transmission channels imposes the limit on the amount of sensor data to be used in remote condition monitoring systems in real-time applications. In this paper, a remote health condition monitoring (RHCM) method based on compressed sensing (CS) is proposed for machine state classification and signal reconstruction. Compressed sensor signals can be directly used to identify different machine states based on a pre-constructed dictionary without the need of traditional feature extraction process. Alternatively, the complete signals can also be reconstructed from the compressed signals and traditional classification approaches can be applied. A case study based on rolling bearing is used to show that the proposed RHCM method can effectively recognize and classify the machine states under different operation conditions using low-volume sensor signals, and the reconstructed signals are accurate enough for post-evaluation or quality assessment of on-site machine process.