A dynamic-model-based fault diagnosis method for a wind turbine planetary gearbox using a deep learning network

Protection and Control of Modern Power Systems

Table 4 Recognition result of experiments

Method	CNN	DANN	DTLN-T	DTLN-F	Method	CNN	DANN	DTLN-T	DTLN-F
A₁ → B₁	0.413 ± 0.028	0.888 ± 0.012	0.962 ± 0.009	0.961 ± 0.009	A₁ → B₂	0.314 ± 0.024	0.77 ± 0.022	0.919 ± 0.022	0.911 ± 0.011
B₁ → A₁	0.411 ± 0.015	0.885 ± 0.020	0.953 ± 0.015	0.964 ± 0.008	B₁ → A₂	0.337 ± 0.031	0.798 ± 0.042	0.915 ± 0.023	0.906 ± 0.030
A₂ → B₂	0.410 ± 0.015	0.879 ± 0.018	0.952 ± 0.010	0.951 ± 0.007	B₂ → A₁	0.332 ± 0.031	0.781 ± 0.025	0.889 ± 0.018	0.918 ± 0.006
B₂ → A₂	0.406 ± 0.016	0.891 ± 0.010	0.959 ± 0.016	0.949 ± 0.011	A₂ → B₁	0.339 ± 0.019	0.783 ± 0.027	0.890 ± 0.014	0.906 ± 0.008
Average	0.410	0.886	0.957	0.956	Average	0.331	0.783	0.903	0.910
C₁ → A₁	0.318 ± 0.027	0.723 ± 0.042	0.826 ± 0.010	0.904 ± 0.013	C₁ → A₂	0.285 ± 0.030	0.658 ± 0.049	0.767 ± 0.017	0.892 ± 0.016
C₁ → B₁	0.313 ± 0.031	0.702 ± 0.045	0.836 ± 0.008	0.914 ± 0.013	C₁ → B₂	0.293 ± 0.023	0.656 ± 0.047	0.737 ± 0.027	0.882 ± 0.019
C₂ → A₂	0.336 ± 0.027	0.688 ± 0.032	0.818 ± 0.011	0.899 ± 0.018	C₂ → A₁	0.287 ± 0.025	0.673 ± 0.049	0.768 ± 0.027	0.898 ± 0.021
C₂ → B₂	0.336 ± 0.029	0.676 ± 0.025	0.820 ± 0.016	0.918 ± 0.008	C₂ → B₂	0.288 ± 0.028	0.696 ± 0.049	0.748 ± 0.032	0.885 ± 0.013
Average	0.326	0.697	0.825	0.909	Average	0.288	0.671	0.755	0.889