多信息融合技术在船舶动力装置故障诊断中的应用  

Application of multi-information fusion technology in fault diagnosis of ship power plant

作　　者：叶树璞 孙俊[1] YE Shupu;SUN Jun

机构地区：[1]武汉理工大学能源与动力工程学院,湖北武汉430063

出　　处：《中国修船》2022年第1期45-48,共4页China Shiprepair

摘　　要：船舶动力装置工作过程中会产生大量多域故障信号,通过收集、挖掘隐藏的关联信号,可以解决船舶动力装置在故障诊断中面临的诊断时长问题。文章采用K-均值聚类算法(K-means)对数据进行聚类,聚类结果输入BP神经网络进行模型训练,并在此基础上,设计了主成分分析法(PCA)对模型进行优化。结果显示,2种算法都能有效降低网络诊断时长,而且经PCA优化的算法更能有效地提升神经网络诊断的收敛速度和准确性。说明PCA能为智能故障诊断算法提供可行的优化方案。A large number of multi-domain fault signals are generated during the operation of a ship power plant.The diagnosis time problem in the fault diagnosis of the ship power plant can be solved by digging and collecting hidden correlated signals.In this paper,the K-means clustering algorithm is employed to cluster the data,and the clustering result is input into the back-propagation(BP)neural network for model training.On this basis,a principal component analysis(PCA)method is designed to optimize the model.The results show that both the two algorithms can effectively reduce the network diagnosis time and that the algorithm optimized by PCA can improve the convergence speed and accuracy of neural network diagnosis more effectively,which means that PCA can provide feasible optimization schemes for intelligent fault diagnosis algorithms.

关 键 词：K-均值聚类算法 数据挖掘 主成分分析法 BP神经网络 故障诊断 

分 类 号：U672[交通运输工程—船舶及航道工程]