Precise feature extraction from wind turbine condition monitoring signals by using optimised variational mode decomposition

Abstract

Reliable condition monitoring (CM) highly relies on the correct extraction of fault-related features from CM signals. This equally applies to the CM of wind turbines (WTs). Although influenced by slowly rotating speeds and constantly varying loading, extracting fault characteristics from lengthy, nonlinear, non-stationary WT CM signals is extremely difficult, which makes WT CM one of the most challenge tasks in wind power asset management despites that lots of efforts have been spent. Attributed to the superiorities to empirical mode decomposition and its extension form Hilbert-Huang transform in dealing with nonlinear, non-stationary CM signals, the recently developed variational mode decomposition (VMD) casts a glimmer of light for the solution for this issue. However, the original proposed VMD adopts default values for both number of modes and filter frequency bandwidth. It is not adaptive to the signal being inspected. As a consequence, it would lead to inaccurate feature extraction thus unreliable WT CM result sometimes. For this reason, a precise feature extraction method based on optimised VMD is investigated. The experiments have shown that thanks to the use of the proposed optimisation strategies, the faultrelated features buried in WT CM signals have been extracted out successfully.

Machine Learning Models Coupled with Variational Mode Decomposition: A New Approach for Modeling Daily Rainfall-Runoff

“…The EMD has a disadvantage called the mode mixing [53], which results in incorrect time-frequency representation and consequently degrades the accuracy of time series processing. Furthermore, since the EMD is a recursive algorithm, the error of envelope estimation can be enlarged more and more, and the efficiency can be decreased [54]. The stopping criteria and end-point effect also affect the decomposition process [53].…”

“…As compared to the EMD, the VMD is more robust to sampling and noise, and has excellent performance in frequency search and separation. Furthermore, the VMD can extract the time-frequency features accurately since it can alleviate the mode mixing through yielding narrow-banded modes [54]. Due to these advantages of the VMD, the development of hybrid MLMs based on the VMD has been accomplished successfully in various fields, including renewable energy, financial and economic fields [56][57][58].…”

A New Method of Two-stage Planetary Gearbox Fault Detection Based on Multi-Sensor Information Fusion

“…(3) multivariate data cannot be processed at the same time; and (4) the mode mixing problem [35]. These shortcomings greatly affect the accuracy of signal processing [36].…”

A Novel Detrended Fluctuation Analysis Method for Gear Fault Diagnosis Based on Variational Mode Decomposition

“…erefore, central frequency of each mode will be gradually demodulated to the corresponding base band, which mitigates mode mixing [11]. By comparison analysis, it is concluded that VMD overcomes the disadvantage of lacking theoretical basis and noise sensitivity of EMD when analyzing nonlinear and nonstationary signals.…”