基于特征金字塔卷积循环神经网络的故障诊断方法

刘秀丽, 徐小力

A Fault Diagnosis Method Based on Feature Pyramid CRNN Network

LIU Xiuli, XU Xiaoli

表6 DNN、CNN和LSTM诊断准确率

Tab.6 Diagnostic accuracy of DNN, CNN, and LSTM

组号 模型 ε/% K=1 K=2 K=3 K=4 K=5 1 LSTM 95.91 95.23 95.02 96.14 95.06 CNN 85.03 87.10 87.05 86.21 86.69 DNN 59.87 65.35 63.54 61.52 63.18 2 LSTM 95.35 96.26 96.19 95.33 96.16 CNN 84.92 86.97 87.80 86.49 87.35 DNN 64.14 61.47 65.20 62.80 58.48 3 LSTM 95.88 94.62 96.24 95.56 95.56 CNN 87.17 86.92 87.55 85.91 84.34 DNN 59.42 63.64 65.38 62.00 63.91