基于复杂网络Newman快速算法的船舶柴油机故障诊断

Fault Diagnosis for Marine Diesel Engine Based on Newman Fast Algorithm in Complex Network

将船舶柴油机故障诊断中的聚类问题转化为复杂网络社团发现问题,在定义线性相似度、反比相似度、指数相似度和椭圆相似度函数的基础上,构造以相似度权重为边权,以样本点为节点的加权无向网络,提出了利用Newman快速算法中的准则函数作为聚类的准则函数,逐步寻找网络中的社团结构的故障诊断方法。以自主研发的轮机模拟器主机系统故障数据为例进行故障分析与诊断,验证新方法的有效性,并分析阈值和相似度系数变化对方法性能的影响。研究结果表明:新故障诊断方法具有计算量小和准确性高且运算时间短的特点,能够达到在线诊断的要求,有识别未知故障的能力,解决了聚类中必须找到类的问题。

Fault classification and diagnosis process of marine diesel engine was transformed into process of the data sample cluster. Defining functions of linear similarity, inverse similarity, exponential similarity and ellipse similarity and taking similarity as the weighted edge and each sample point as the node, a weighted and undirected network model was constructed. A fault diagnosis method using criterion function of Newman fast algorithm as clustering criterion function to detect gradually the community structure of the network was proposed. Using the fault data of a self-developed marine engine simulator as an example of fault diagnosis, fault analysis and diagnosis were conducted to verify the method and analyze impacts of thresholds and similarity coefficient on the performance of the algorithm. Results show that this method has advantages of effectiveness, computation time short, and satisfying online diagnosis requirements. Moreover, it can recognize unknown fault patterns, solving the problem to find the cluster in cluster analysis.