基于双谱分析特征提取的汽轮机故障智能诊断

摘要

Due to the fault signals with non-linear characteristic for rotary machinery, the traditional liner methods are not suitable for signal processing.Compared with higher-order statistics (HOS) for non-linear signal processing, the bispectrum method possesses the advantages of high computational effectiveness.Firstly, an intelligent turbine fault diagnosis method is proposed based on bispectrum feature extraction by extracting such features as rub, unbalancing and misalignment.Then, the selected feature vectors are used as training data for the trained support vector data description (SVDD) models, whereas the remained feature vectors are used as testing data and are input into (SVDD) models.Finally, the effectiveness and feasibility of this approach are verified via experiments.%旋转机械发生故障时,其信号往往呈现出非线性特征,故常规的线性信号处理方法不再适用于旋转机械故障信号的特征提取.高阶谱分析方法是基于高阶统计量(Higher Order Statistics,HOS)的一种非线性信号处理方法,其中的双谱分析方法具有高阶统计量的一切优点,并且具有较低的阶数,便于计算.提出基于双谱特征提取的汽轮机故障智能诊断方法:用双谱分析方法分别对汽轮机的碰摩故障信号、转子不平衡故障信号及转子不对中故障信号进行特征提取,用提取到的部分特征作为支持向量数据描述(Support Vector Data Description,SVDD)的训练数据,部分特征向量作为测试数据.分析结果验证了所述方法的有效性及可行性.