ENSEgram: A Fault Detection Method of Repetitive Transient Impulses With Scale Spatial Frequency Band Segmentation

Repetitive transient impulse detection methods are often hindered under the severe operating conditions of real-world engineering applications. One primary issue is that the constant spectrum segmentation approach fails to adapt to the distinctive features of the spectrum, leading to the omission of crucial signatures. Another reason is that current state-of-the-art indicators are vulnerable to severe disturbances and rely on a priori knowledge of fault correlation. To deal with these limitations, a new fault detection method, ENSEgram, is proposed in this paper to detect repetitive transient impulses under severe disturbances. This approach introduces a frequency band segmentation technique based on multi-scale spatial analysis of the spectrum. It is designed to automatically divide frequency bands in accordance with the specific distribution characteristics of the spectrum. To reduce the impact of significant disturbances, a novel index has been developed using the squared envelope spectrum. This index takes into account the sparsity, complexity, and cyclostationarity of the squared envelope spectrum concurrently, effectively pinpointing the resonance band caused by repetitive fault transient impulses. In comparison with traditional signal processing techniques, the ENSEgram can precisely isolate the fault bands that contain more fault-related information while minimizing noise interference, especially in harsh operational environments.