Fault Detection Method Based on the Condition Indicators of Compressive Sampling for the Gearbox in Helicopter

In the field of condition monitoring and fault diagnosis of rotating machinery, it is always necessary to monitor the component uninterruptedly for a long period of time. This process will generate masses of original data which would burden the storage and transmission. In data analysis, we can decode the compressed data we saved to recover the original vibration data and implement fault detection and diagnosis etc. with relevant available methods. The majority of the original data would be dropped, which should be an enormous waste of resources. Since the signal can be compressed, we should consider whether we can obtain compressive sampling data directly and omit much data containing useless information at the same time. This idea is the rudiment of compressed sampling. The general method for gearbox fault detection is achieved by using vibration signals which sampled in the frame of Shannon sampling theory. For this issue, a fault detection method based on compressed sampling theory is proposed in this paper. A condition indicator is extracted based on the measurements of the reconstruction deviation, and these reconstruction signal sets are all based on the normal redundant dictionary. The availability and effectiveness of the proposed method is validated through the simulation data sets with different faults seeded.