重庆大学机械传动国家重点实验室
国家自然科学基金资助项目(51975075)；重庆市技术创新与应用示范专项资助项目(cstc2018jszx-cyzdX0146)
State Key Lab of Mechanical Transmission, Chongqing University
Supported by the National Natural Science Foundation of China(51975075)；Supported by the Chongqing Technology Innovation and Application Program(cstc2018jszx-cyzdX0146)
针对离心鼓风机故障识别过程中单一传感器信号故障信息有限，传统的卷积神经网络（CNN）在处理多源高维数据时特征提取能力不足的问题，提出一种基于多源信息融合和自适应深度卷积神经网络（ADCNN）的离心鼓风机故障诊断方法。首先，基于相关性方差贡献率法实现离心鼓风机多源同类信息的数据层融合，建立多源信息融合框架；然后，利用ADCNN自适应地提取各异类信息的特征并完成特征融合，建立融合多源信息的ADCNN故障诊断模型；最后，将此方法应用于离心鼓风机转子故障诊断上，并与传统的融合模式以及CNN、反向传播神经网络（BPNN）、支持向量机（SVM）方法进行对比，试验结果表明：提出的方法在诊断精度与鲁棒性上均优于其他方法。
The fault information of the single sensor signal is limited in the process of centrifugal blower fault recognition, and the traditional convolutional neural network (CNN) has poorer feature extraction abilities when processing multi-source high-dimensional data. To address this problem, this paper proposes a fault diagnosis method based on multi-source information fusion and adaptive deep convolutional neural network (ADCNN) for centrifugal blowers. Firstly, data-level fusion of multi-source homogeneous information of centrifugal blower is realized based on correlation variance contribution rate method, and the multi-source information fusion framework is established. Then, ADCNN is used to adaptively extract the features of heterogeneous information and complete feature fusion, an ADCNN fault diagnosis model that integrates multi-source information is constructed. Finally, the proposed method is applied to the fault diagnosis of centrifugal blower rotor and compared with the traditional fusion method, CNN, BPNN, SVM. The experiment results show that the proposed method has superior diagnosis accuracy and robustness than other methods.
