**Yichuan Fu 1, Zhiwei Gao 1,\*, Yuanhong Liu 2, Aihua Zhang <sup>3</sup> and Xiuxia Yin <sup>4</sup>**


Received: 4 May 2020; Accepted: 25 August 2020; Published: 1 September 2020

**Abstract:** In response to the high demand of the operation reliability and predictive maintenance, health monitoring and fault diagnosis and classification have been paramount for complex industrial systems (e.g., wind turbine energy systems). In this study, data-driven fault diagnosis and fault classification strategies are addressed for wind turbine energy systems under various faulty scenarios. A novel algorithm is addressed by integrating fast Fourier transform and uncorrelated multi-linear principal component analysis techniques in order to achieve effective three-dimensional space visualization for fault diagnosis and classification under a variety of actuator and sensor faulty scenarios in 4.8 MW wind turbine benchmark systems. Moreover, comparison studies are implemented by using multi-linear principal component analysis with and without fast Fourier transform, and uncorrelated multi-linear principal component analysis with and without fast Fourier transformation data pre-processing, respectively. The effectiveness of the proposed algorithm is demonstrated and validated via the wind turbine benchmark.

**Keywords:** fault diagnosis; fault classification; fast Fourier transform (FFT); multi-linear principal component analysis (MPCA); uncorrelated multi-linear principal component analysis (UMPCA); additive white Gaussian noises (AWGN); wind turbine systems
