Remaining Useful Life Prediction of Rolling Bearings Based on Multiscale Convolutional Neural Network with Integrated Dilated Convolution Blocks

Table 3

Performance comparisons with different methods.

Bearing	Proposed method		DNN		CNN		LSTM		MS-CNN in [15]
	MAE (%)	RMSE (%)	MAE (%)	RMSE (%)	MAE (%)	RMSE (%)	MAE (%)	RMSE (%)	MAE (%)	RMSE (%)
1-3	6.14	7.83	17.31	22.54	8.50	10.52	11.21	14.68	6.75	8.98
1-4	19.29	24.62	34.14	36.27	26.31	33.28	26.71	32.37	24.38	27.47
1-5	17.46	23.98	28.74	32.15	24.84	32.22	25.15	32.05	18.15	22.25
1-6	18.19	21.96	32.46	38.43	22.86	26.81	24.31	29.08	20.33	22.95
1-7	10.87	17.96	19.68	32.6	12.09	17.96	16.85	23.19	10.31	16.36
2-3	17.54	23.23	40.51	46.19	25.35	31.36	28.18	34.58	25.18	30.25
2-4	29.81	34.98	42.73	55.72	32.78	39.04	32.69	40.70	26.14	28.69
2-5	21.71	28.23	37.46	41.03	22.56	29.41	26.32	32.75	23.62	30.16
2-6	19.20	24.35	29.45	39.85	22.20	27.77	29.89	36.22	20.70	25.71
2-7	27.20	32.64	41.20	44.21	38.29	43.02	28.37	33.68	22.14	30.26
3-3	13.78	17.97	28.05	23.75	18.01	21.83	16.86	20.37	15.46	20.07