支持向量机在船舶电力推进系统故障诊断中的应用

针对人工神经网络收敛速度慢、容易陷入局部最优解等不足,本文采用支持向量机技术建立船舶电力推进故障诊断系统。确定支持向量机的核函数和分类方法,结合训练样本,采用基于网格搜索的K重交叉验证法进行核函数的参数优化,从而得到支持向量机故障诊断模型。利用支持向量机工具箱函数,在MATLAB中进行故障诊断模型的仿真计算,结果表明基于支持向量机所建立的故障诊断模型有较强的诊断准确性和泛化推广能力,从而提高船舶的安全性。

Applications of Support Vector Machine to Fault Diagnosis for Marine Electric Propulsion System

Recognizing the shortcomings of slow convergence of intelligent neural-network as well as local optimization, a new method of fault diagnosis of marine electric propulsion system using support vector machine （SVM）is used in this paper, which determines the kernel function and classification method. Using training sampling and K-multiple principal component analysis to optimize parameters of the kernel function, a model of fault diagnosis compatible with SVM is obtained. Simulation using MATLAB shows that the model has high precision, suitable for generalization as well as improved ship safety.