基于TimeGAN数据增强的复杂过程故障分类方法

杨磊,何鹏举,丑幸幸

TimeGAN data augmentation-based fault classification method for complex processes

Lei YANG,Pengju HE,Xingxing CHOU

表 3 不同故障模式下的故障分类性能评价指标

Tab.3 Performance evaluation indexes of fault classification under different fault modes

提取子空间 所用数据 P/% R/% F1/% 故障1 故障2 故障6 故障7 故障12 故障14 故障1 故障2 故障6 故障7 故障12 故障14 故障1 故障2 故障6 故障7 故障12 故障14 不平衡故障样本 95.56 76.35 78.26 85.37 61.70 73.21 99.50 79.50 76.50 84.62 64.25 71.37 97.49 77.89 77.37 85.00 62.95 72.28 数据增 强后故 障样本 WGAN-GP 98.70 99.48 100.00 99.03 87.04 98.26 95.25 96.50 100.00 89.75 98.25 98.88 96.95 97.97 100.00 94.16 92.31 98.57 SMOTE 95.56 98.71 95.85 99.03 86.93 98.51 99.50 95.75 98.12 89.75 98.12 99.12 97.49 97.21 96.97 94.16 92.19 98.82 TimeGAN 99.12 100.00 100.00 99.87 92.33 100.00 98.88 97.75 100.00 95.13 99.38 99.62 99.00 98.86 100.00 97.44 95.73 99.81 全部真实故障样本 99.87 100.00 100.00 99.87 91.58 100.00 99.88 97.88 100.00 98.25 99.75 98.12 99.87 98.93 100.00 99.05 95.49 99.05