Wind Turbine Condition Monitoring Based on an Improved Spline-Kernelled Chirplet Transform

Abstract

The time-varying operational conditions applied to wind turbines (WTs) not only challenge their operation but also make condition monitoring (CM) difficult. To achieve a reliable CM result, more advanced signal processing techniques, rather than the conventional spectral analyses, are urgently needed for interpreting the nonlinear and nonstationary (NNS) CM signals collected from the turbines. The work presented in this paper is an effort to meet such a requirement. Based on the proven capability of the spline-kernelled chirplet transform (SCT) in detecting the instantaneous frequencies within NNS monocomponent signals, this paper improves the SCT to enable it to detect the instantaneous amplitude of lengthy NNS multicomponent signals at a fault-related frequency of interest. The improved SCT is then applied for developing a new real-time CM technique dedicated to extracting fault-related features from WT CM signals. Experiment proves that the improved SCT has overcome existing SCT issues and is capable of correctly tracking the amplitude characteristics of NNS multicomponent signals at fault-related frequencies of interest. The new CM technique developed, based on this improved SCT, shows success in detecting both mechanical and electrical faults occurring in a WT drive train, despite the constantly varying operational conditions of the turbine. Moreover, its algorithm is efficient in computation, which not only enables it to deal with lengthy NNS CM signals but also makes it ideal for online use.