Abstract
Effective fault detection and diagnosis (FDD) for rotating machinery is always a focus issue in improving the prognostic and health management (PHM) of the equipments. The existing usage of similarity measurement has been widely spread in searching the homologous fault responses from vibration signals, but most of them are just suitable for stable speed and cannot be applied in all variable speed conditions. In order to improve measurement performance, a fast-meshed phase portrait (FMPP) frame combining the phase-space technique and box-scoring calculation is proposed. Firstly, the variable-speed signal is divided into multiple undetermined fragments according to fault characteristic orders (FCOs). Secondly, the undetermined fragments are reconstructed into corresponding phase-space trajectories to overcome the time-delay matching inconsistency of the variable speed fragments. Thirdly, the phase-space trajectories are mapped into meshed phase portraits via box-scoring calculation. Such decisive calculation can effectively transforms the diverse unequal fragments into the phase diagrams with same size, which saves time for the subsequent similarity measurement. Finally, the proposed FMPP is tested both for accuracy and timeliness on a self-built bearing bench, where the order tracking (OT) and dynamic time warping (DTW) methods are used for the comparisons. The experiments proved the effectiveness of the proposed method.
Highlights
Time-domain waveform clearly and intuitively reflects the inherent characteristics of machine vibration, which is a key technique widely used in mining nonobvious fault information from vibration signals
Georgoulas et al [8] introduced a novel bearing fault detection method that uses the symbolic aggregate approximation (SAX) framework and associated related intelligent to represent the diagnostic representation of the extracted vibration record. e fault diagnosis is considered as a classification problem, the vibration signal and its feature vector are classified, and the classification accuracy is significantly improved
We find that as τ varies from one to n − 1 sample periods, the cross-correlation value does not show a local minimum, so the optimal τ cannot be determined. τ calculated by autocorrelation method (AUD) is 0.075 ms, and the signal sampling period of the fault response fragments is 0.05 ms
Summary
Time-domain waveform clearly and intuitively reflects the inherent characteristics of machine vibration, which is a key technique widely used in mining nonobvious fault information from vibration signals. The energy of the vibration signal is different, which is reflected in the difference in signal amplitude due to the SAX and cosine similarity methods cannot measure time series similarities of different lengths. The high time complexity of DTW is difficult to meet the needs of high-frequency computing for the era of big data For this case, it is necessary to find a low time complexity way to calculating vibration signals’ similarity, which overcomes the influence of variable speed and maintains the essential characteristics of fault response. When the length of a single response fragment is n, the response fragments group after signal slice
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
