Connected Components-based Colour Image Representations of Vibrations for a Two-stage Fault Diagnosis of Roller Bearings Using Convolutional Neural Networks

Chinese Journal of Mechanical Engineering

Table 4 Comparison with the results obtained using 2-D grayscale-based CNN

Training size (%)	2-D grayscale vibration images-based CNN classification accuracy	Our proposed RGBVI-CNN classification accuracy
40	86.8% ± 13.1	99.3% ± 0.7
50	88.1% ± 11.6	99.8% ± 0.1
60	91.7% ± 8.2	99.9% ± 0.1
70	92.6% ± 7.4	100% ± 0.0