Table 4 Performance comparison of different methods under varying signal-to-noise ratios.
From: RS-SCBiGRU: a noise-robust neural network for high-speed motor fault diagnosis with limited samples
Methods | 0 | 2 | 4 | 6 | 8 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
Accuracy (%) | F1 (%) | Accuracy (%) | F1 (%) | Accuracy (%) | F1 (%) | Accuracy (%) | F1 (%) | Accuracy (%) | F1 (%) | |
Proposed method | 57.65 | 53 | 72.00 | 68.89 | 84.49 | 83.28 | 91.80 | 91.31 | 95.04 | 94.84 |
SCBiGRU | 51.91 | 45.75 | 64.44 | 60.45 | 76.05 | 73.59 | 83.54 | 82.08 | 88.13 | 87.22 |
BiGRU | 43.11 | 33.10 | 53.94 | 47.03 | 67.06 | 62.31 | 77.69 | 74.71 | 85.47 | 83.60 |
BiLSTM | 41.56 | 31.66 | 52.98 | 45.29 | 64.51 | 57.84 | 74.27 | 69.06 | 81.27 | 77.83 |
DAMN | 28.13 | 15.91 | 31.82 | 19.81 | 35.45 | 23.84 | 37.71 | 26.53 | 45.29 | 36.52 |
MSCNN | 10.00 | 1.82 | 10.00 | 1.82 | 10.09 | 1.99 | 13.73 | 6.28 | 22.87 | 15.72 |
MSCNN-LSTM | 27.76 | 18.35 | 30.36 | 20.11 | 33.02 | 23.04 | 36.00 | 26.80 | 41.35 | 33.57 |
RNN-WDCNN | 58.18 | 53.01 | 71.84 | 69.86 | 81.31 | 79.95 | 87.25 | 86.62 | 90.04 | 89.49 |
WDCNN | 47.51 | 41.11 | 56.04 | 51.30 | 62.78 | 59.76 | 69.09 | 67.41 | 73.95 | 72.77 |
