Development of Deep Convolutional Neural Network with Adaptive Batch Normalization Algorithm for Bearing Fault Diagnosis

Table 3

Recognition rate of the WFK-CNN model under noise environment.

Convolution kernel size	SNR (dB)
	−4 (%)	−2 (%)	0 (%)	2 (%)	4 (%)	6 (%)	8 (%)	10 (%)
16	27.24	40.65	55.46	72.11	85.61	94.62	98.53	99.25
24	35.15	52.84	70.32	84.83	94.32	98.53	99.66	99.85
32	42.32	57.77	72.71	86.57	95.39	98.46	99.42	99.61
40	46.73	63.09	77.60	90.16	97.03	99.18	99.61	99.86
48	50.22	66.32	80.34	92.15	97.75	99.36	99.78	99.89
56	51.78	66.75	80.79	92.32	97.80	99.19	99.72	99.72
64	51.89	67.21	82.08	93.05	98.10	99.37	99.76	99.83
72	53.85	68.67	82.41	92.95	97.88	99.44	99.69	99.80
80	56.17	69.43	84.37	94.79	98.74	99.49	99.87	99.85
88	56.13	71.78	85.43	95.09	98.42	99.33	99.79	99.84
96	64.43	78.87	89.93	96.90	99.06	99.55	99.85	99.86
104	62.77	79.31	90.44	97.68	99.36	99.69	99.84	99.83
112	66.97	80.92	90.62	97.09	98.79	99.51	99.82	99.80
120	61.66	77.70	90.37	97.30	99.05	99.60	99.80	99.85
128	60.73	77.43	89.65	97.17	99.21	99.51	99.84	99.82
The average	66.97	80.92	90.62	97.68	99.36	99.69	99.87	99.89