Verifying hyperparameter sensitivities of optimal filter design methods for fault signature enhancement

Recent advances in vibrations, noise, and their use for machine monitoring

Open Access

Table 6

The hyperparameter sensitivity verification analysis for the GES2N is presented. A check mark (X) indicates alignment between the results in Figure 9 and Table 1.

#	HP	Measurements					Comment
		1	2	3	4	5
1	D	✓	✓	✓	✓	✓	The result is consistent with Table 1.
2	CORF	✓	✓	✓	✓	✓	Decreasing the CORF increases the amplitudes of the components. For this dataset, the cyclic order width of the harmonics are broader due to the length of the signal and by decreasing the CORF while keeping Δα_b at its baseline resulted in some parts of the targeted harmonics to manifest in the noise indicator and impeded its performance.
3	Δα_b	✓	✓	✓	✓	✓	The result is consistent with Table 1. No prominent extraneous impulses are in the signal indicator’s bands when Δα_b = 0.5 and therefore the GES2N’s performance is not impeded for measurements where the damage components were prominent (e.g., Meas. 1 and 5). The general decrease in ${\tilde{M}}_{g e s 2 n}$ for smaller Δα_b is caused by including the fault information in the noise indicator (i.e., Δα_b is too small). For Meas. 2 and 3, the fault signatures are weaker in the signal, which results in spurious components being enhanced when Δα_b = 0.5.
4	TEP	✓	✓	✓	✓	✓	The decrease of ${\tilde{M}}_{g e s 2 n}$ with TEP is expected. The fault information is not visible when TEP = 10.
5	N_h	✓	✓	✓	✓	✓	The result is consistent with Table 1.
6	*w_s*						A more in-depth analysis showed H₁ ’s first harmonic was prominent (which is expected), but the other methods H_1:5, H_- and H₊ performed similarly and therefore their results could not be verified.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.