Adaptive sub-Nyquist sampling and its application in rotating machinery fault detection

In rotating machinery fault detection, oversampling induces three forms of data redundancy: (1) redundancy resulting from conventional Nyquist-based sampling, (2) redundancy induced by changes in the degradation state of the rotating machinery's components, and (3) redundancy caused by operating over a wide range of rotational speeds. To resolve these data redundancies, this paper proposes a novel adaptive sub-Nyquist sampling (ASNS) technique. ASNS precisely aligns the sampling strategy with the shaft's rotational period and dynamically adjusts the number of samples per rotational period based on condition awareness. We applied ASNS to fault detection and proposed an ASNS-based method for detecting faults in rotating machinery. The ASNS-based fault detection method does not require signal reconstruction; instead, it analyzes the ASNS data to extract fault features. The effectiveness of the ASNS-based fault detection method is validated using simulated and experimental data. The results demonstrate that the ASNS-based fault detection method improves data acquisition performance and significantly reduces the sampling frequency while maintaining detection reliability and accuracy. This advantage effectively reduces the costs of data acquisition, transmission, storage, and processing, thereby optimizing predictive maintenance decisions. Additionally, the ASNS-based fault detection method has low computational complexity and supports real-time fault detection.