多信息融合技术在船舶柴油机故障诊断中的应用

Application of Multi-Information Fusion Technology for Fault Diagnosis in Marine Diesel Engine

柴油机工作过程中的各种参数含有大量的信息,通过数据挖掘将这些参数的内在信息挖掘出来,解决以往在柴油机故障诊断上出现的诊断不准确和耗时长等问题。采用K均值聚类分析(k-means)将数据聚类,并设计BP(Back Propagation)神经网络,对柴油机的运行状态进行诊断。在此基础上,利用PCA(Principal Component Analysis)对上述算法进行优化,用PCA对原始数据简化,再进行k-means聚类,最后将聚类后的数据特征量作为BP神经网络的输入,建立柴油机的故障诊断模型。通过对两种诊断算法的结果进行分析和比较,表明优化后的算法能够更有效地提取数据特征,提高了诊断准确度,同时也减少了诊断时间。

Measurements taken from a diesel engine during its working process contain extensive information. The implicit information in the measurements can be extracted by data mining for facilitating diesel engine fault diagnosis which can be a time-consuming process otherwise. A fault diagnosis algorithm is developed based on the k-means clustering and the back propagation（ BP） neural network. The Principal Component Analysis（ PCA） algorithm is incorporated in the diesel engine fault diagnosis model to screen raw data before clustering and neural network processing. Experiments proves that the PCA algorithm improves feature extracting process and,consequently,leads to more accurate diagnoses and shorter processing time.