A Multimodel Decision Fusion Method Based on DCNN-IDST for Fault Diagnosis of Rolling Bearing

Table 5

The comparison test results of several network models.

Number	Model	Feature learning from raw data (%)	Time-domain features (%)	Manual feature extraction frequency-domain features (%)	Time-frequency-domain features (%)
Vibrating sensor1	BPNN	89.17	86.25	87.55	88.78
	RBFNN	87.08	65.83	70.35	73.33
	DCNN	83	72.50	78.33	80.62
Vibrating sensor2	BPNN	89.58	80.42	81.25	85.00
	RBFNN	91.42	77.92	85.66	88.66
	DCNN	90.92	80.51	81.55	83.33