基于多特征提取与核极限学习机的汽轮机转子故障诊断

Fault Diagnosis of Steam Turbine Rotor Based on Multi Feature Extractionsand Kernel Extreme Learning Machine

针对汽轮机转子发生的典型故障,提出一种基于多特征提取和核主元分析的核极限学习机诊断模型。首先,对测取到的各典型故障时的振动信号进行变分模态分解,获得不同尺度固有模态函数;其次,对各固有模态函数计算特征能量和样本熵组成特征向量;最后,对特征向量采用核主元分析法进行去噪和降维,并将其作为输入进行核极限学习机(遗传算法优化)模型的训练和测试。与基于单一特征向量的模型相比,多种特征提取能够增强模型的输入特征,而核主元分析方法能够明显减少信息冗余和特征向量的相关性,且节约时间成本,在一定程度上提高模型的预测性能,为汽轮机转子故障的诊断研究提供一种新的思路。

Aiming at the typical faults of steam turbine rotors,a kernel extreme learning machine diagnostic model based on multi feature extraction and kernel principal component analysis is proposed.Firstly,the vibration signals under the typical faults are subjected to variational mode decomposition to obtain the natural mode functions of different scales.Secondly,the characteristic energy and sample entropy are calculated for each natural mode function to form the feature vector.Finally,the feature vector is obtained.The kernel principal component analysis method is used for denoising and dimensionality reduction,and it is used as input to train and test the kernel extreme learning machine(genetic algorithm optimization)model.Compared with the single feature vector based model,the kernel principal component analysis method can significantly save time cost and improve the prediction performance of the model to a certain extent,which provides a new idea for the diagnosis of turbine rotor faults.