支持向量机和BP神经网络在水轮发电机轴承故障诊断中的应用

支持向量机(SVM)与BP神经网络相比各有优缺点,通过对支持向量机和BP神经网络在水轮发电机滚动轴承故障诊断中的仿真实验,来对比两者在轴承故障诊断上的泛化能力。首先通过应用经验模态分解(EMD)的方法将轴承振动信号进行分解,得到本征模函数(IMF),再将IMF的平均能量值作为故障特征向量。将这些特征向量作为支持向量机和BP神经网络的学习样本。经过仿真研究结果表明,在小样本集的前提下,支持向量机在轴承故障诊断中的精确度不但受样本数量变动的影响较小,准确度也高于BP神经网络,具有较强的泛化能力。对水轮发电机滚动轴承故障诊断模型的应优先考虑选择SVM。

Application of Support Vector Machines and BP Neural Network in the Rolling Bearing of Hydraulic Turbine Generator Fault Diagnosis

The support vector machine （SVM） and BP neural network have their own respective advantages and disadvantages, and by simulating of support vector machines and BP neural network, the generalization abilities of both are compared in the rolling bearing of hydraulic turbine generator fault diagnosis. Firstly, the beating vibration signal is decomposed by application of the empirical mode decomposition （EMD） method to obtain the intrinsic mode functions （IMF） and then the average energy values of the IMF are taken as the fault feature vectors, which in turn are taken as the SVM and the BP neural network learning samples. The simulation results show that under the premise of the small sample set, the precision of the SVM fault diagnosis is less affected by changes in the number of samples, and the accuracy is higher than the BP neural network and the generalization is also better. Therefore the SVM method should be preferred in selecting the actual the rolling bearing of hydraulic turbine generator diagnosis model.