Table 2 Comparison of recognition results based on different pooling methods.
Working condition recognition methods | Measured dynamometer card | Measured electrical power card | Measured dynamometer card and measured electrical power card | |||
|---|---|---|---|---|---|---|
Accuracy (%) | F1-score | Accuracy (%) | F1-score | Accuracy (%) | F1-score | |
Different pooling methods under single-source | ||||||
S-WP-db4-ResNet-50 | 98.81 | 0.9879 | 98.32 | 0.9828 | – | – |
S-WP-bior3.3-ResNet-50 | 98.87 | 0.9883 | 98.36 | 0.9834 | – | – |
S-LWP-db4-ResNet-50 | 98.70 | 0.9868 | 98.39 | 0.9837 | – | – |
S-LWP-bior3.3-ResNet-50 | 98.92 | 0.9889 | 98.44 | 0.9839 | – | – |
S-DTCWP-ResNet-50 | 98.87 | 0.9886 | 98.32 | 0.9829 | – | – |
S-CP-ResNet50 | 99.16 | 0.9915 | 98.47 | 0.9839 | – | – |
Different pooling methods under multi-source | ||||||
AMMFFE-WP-db4-ResNet-50 | – | – | – | – | 99.36 | 0.9935 |
AMMFFE-WP-bior3.3-ResNet-50 | – | – | – | – | 99.24 | 0.9924 |
AMMFFE-LWP-db4-ResNet-50 | – | – | – | – | 99.19 | 0.9918 |
AMMFFE-LWP-bior3.3-ResNet-50 | – | – | – | – | 99.39 | 0.9937 |
AMMFFE-DTCWP-ResNet-50 | – | – | – | – | 98.97 | 0.9897 |
AMMFFE-CP-ResNet-50 (ours) | – | – | – | – | 99.52 | 0.9952 |
