基于SOM神经网络的船舶分油机故障诊断研究

SOM Neural Networks-based Fault Diagnosis for Marine Separator

为了实现对船舶分油机故障的智能诊断,提出一种基于SOM神经网络的诊断方法。首先,在分析分油机典型故障及特征参数的基础上,提取故障特征向量并建立学习样本。其次,建立了SOM网络模型,通过样本数据集进行训练,获取了输入与输出间的非线性映射。最后将建立的SOM网络应用于分油机的故障分类和诊断。实验验证表明:该方法诊断准确度高和对不同故障识别的适应性强,是一种可行有效的分油机故障智能诊断方法。

SOM neural networks-based method is proposed to implement intelligent diagnosis for marine separator faults. Fistly, standard faults feature vectors, as learning sampling, are built based on separator classic faults and the related feature parameters. Next, the nonlinear map between inputs and outputs of SOM is constructed by training with data sampling. And then, the constructed SOM is used for diagnosis and classification of separator faults. According to tests, this method is a feasible and effective one with high accuracy and flexibility for identification of different faults.