基于局部线性嵌入算法的柴油机故障诊断研究

为提高柴油机故障诊断准确率和效率，提出了改进局部线性嵌入算法的柴油机诊断系统。应用小波包能量谱分析方法提取某柴油机振动信号的特征值，将提取的高维特征向量映射到低维空间上，能将高维特征向量进行优化，即特征值的二次提取。该改进算法可模糊化近邻点k的选择，从而提高计算的速度，并应用SOM-BP神经网络进行故障识别。实验表明，经过局部线性嵌入算法的特征值优化，能减少SOM-BP神经网络的输入节点，可在一定程度上提高故障识别的效率和准确率。

Diesel engine fault diagnosis based on local linear embedding algorithm

In order to improve the accuracy and efficiency of diesel engine fault diagnosis system of diesel engine, improve the locally linear embedding algorithm. Spectrum feature analysis method to extract the value of a diesel engine vibration signal by wavelet packet energy, and then maps high-dimensional feature vectors extracted into low dimensional space and optimize the high-dimension feature vector, the improved algorithm can fuzzy neighbor point k, therefore it can increase the speed of calculation and SOM-BP neural network is applied for fault recognition. Experiments show that, by the locally linear embedding algorithm optimization, can reduce the input nodes of SOM-BP neural network, and can improve the efficiency of fault identification and accuracy to some extent.