基于多通道残差深度网络的往复压缩机故障诊断模型研究

Research on Fault Diagnosis Model of Reciprocating CompressorBased on Multi-Channel Residual Depth Network

往复压缩机故障机理复杂,状态监测信号类型较多,故障诊断难度大。提出一种基于多通道残差网络的往复压缩机故障诊断模型,包含多个独立的输入层通道,分别处理不同传感器监测数据;经过多次卷积、池化、批归一化、激活和展平操作,将多源传感器监测数据的深层抽象特征在全连接层进行融合,实现不同参数之间关联信息提取。应用实验台与实际往复压缩机故障数据进行测试验证,诊断模型取得了良好的应用效果。

The fault mechanism of reciprocating compressor is complex, there are many types of condition monitoring signals, and the fault diagnosis is difficult.A fault diagnosis model of reciprocating compressor based on multi-channel residual network is proposed, which includes multiple independent input layer channels, simultaneous interpreting different sensor monitoring data.After multiple convolutions, pooling, batch normalization, activation and flattening operations, the deep abstract features of multi-source sensor monitoring data are fused in the full connection layer to extract the correlation information between different parameters.The fault data intest and actual of reciprocating compressor is tested and verified, and the new diagnosis model has achieved good application results.