A feature sample screening and local outlier factor fusion method for detecting tidal stream turbine blade impact fault

Abstract

The impact fault caused by ocean creatures poses a risk to the safe operation of tidal stream turbine blades. However, it is difficult to detect the weak impact fault directly because the collected signal is disturbed by the waves, turbulence, and continuously variable flow velocity. To solve this problem, a fusion method of feature sample screening and local outlier factor is proposed in this paper. This method consists of three main parts. First, the Teager–Kaiser energy operator and the sliding window technique are introduced to extract the envelope statistical features from the current signal. Second, a parameter optimized density-based spatial clustering of applications with noise (DBSCAN) algorithm is proposed to perform feature sample screening before detecting. Notably, the conventional DBSCAN algorithm is sensitive to the parameter selection and lacks the capability of adaptive screening, so this paper proposes an adaptive sand cat swarm optimization algorithm to optimize the parameters. Finally, the local outlier factor is utilized to detect faults based on the screened feature samples. The experimental results show that the proposed method stands out in reducing the false alarm rate compared with traditional methods. Specifically, within the flow velocity ranges of 1.0–1.3 m/s and 1.3–1.6 m/s, the false alarm rates can reach 0% and 0.17%, respectively.