Research Article

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.

BearingProposed methodDNNCNNLSTMMS-CNN in [15]
MAE (%)RMSE (%)MAE (%)RMSE (%)MAE (%)RMSE (%)MAE (%)RMSE (%)MAE (%)RMSE (%)

1-36.147.8317.3122.548.5010.5211.2114.686.758.98
1-419.2924.6234.1436.2726.3133.2826.7132.3724.3827.47
1-517.4623.9828.7432.1524.8432.2225.1532.0518.1522.25
1-618.1921.9632.4638.4322.8626.8124.3129.0820.3322.95
1-710.8717.9619.6832.612.0917.9616.8523.1910.3116.36
2-317.5423.2340.5146.1925.3531.3628.1834.5825.1830.25
2-429.8134.9842.7355.7232.7839.0432.6940.7026.1428.69
2-521.7128.2337.4641.0322.5629.4126.3232.7523.6230.16
2-619.2024.3529.4539.8522.2027.7729.8936.2220.7025.71
2-727.2032.6441.2044.2138.2943.0228.3733.6822.1430.26
3-313.7817.9728.0523.7518.0121.8316.8620.3715.4620.07