Imbalanced Fault Classification of Bearing via Wasserstein Generative Adversarial Networks with Gradient Penalty

Table 3

Different training data settings using real samples and generated samples and the classification results.

	A	B	C	D	E	F	G	H	I
Raw samples	2000	0	0	0	0	0	2000	2000	2000
Generated samples	0	1500	2000	3000	4000	5000	1000	2000	3000
Accuracy (%)	94.03	89.05	93.96	98.09	98.65	99.03	98.08	98.65	99.03