An Enhanced Gated Recurrent Unit-Based Adaptive Fault Diagnosis of Rotating Machinery

Table 5

The framework of CNN for ablation experiment.

Parameter profile	Framework A	Framework B	Framework C	Framework D
Input	1024	1024	1024	1024
Convolution	32, 3, 1, same	32, 5, 1, same	32, 3, 2, same	48, 3, 1, same
Maxpooling	2	2	—	2
Convolution	64, 3, 1, same	64, 5, 1, same	64, 3, 2, same	96, 3, 1, same
Maxpooling	2	2	—	2
Convolution	128, 3, 1, same	128, 5, 1, same	128, 3, 2, same	192, 3, 1, same
Maxpooling	2	2	—	2
Fully connection	1024	1024	1024	1024
Fully connection	9	9	9	9
Keep_prob	0.9	0.9	0.9	0.9
Learning rate	0.1	0.1	0.1	0.1
Accuracy	95.18% ± 1.64	94.22% ± 1.84	93.16% ± 1.86	92.52% ± 1.98