一种机械故障诊断多传感器数据融合特征提取的方法

发布时间：2018-03-30 20:24

本文选题：机械故障诊断　切入点：多传感器数据融合　出处：《西安石油大学学报(自然科学版)》2017年01期

【摘要】：为提高机械故障诊断的准确率,将多个振动传感器采集机械系统不同位置的信息进行融合,提出一种基于同源数据融合的特征提取方法。以柴油机缸盖和机身的振动信号为例,分析振动信号频谱与激励源到测量点的传输特性,构造基于频谱的高维特征向量。使用PCA方法和子带平均法降维,支持向量机进行分类验证其分类效果。结果表明,相比传统单通道传感器,此方法提取的特征不仅具有更高的可压缩性,而且其分类准确性有所提高。
[Abstract]:In order to improve the accuracy of fault diagnosis, a plurality of mechanical vibration sensor system acquisition in different position of the information fusion, this paper presents an approach to extract homologous features based on data fusion. The vibration signal of diesel engine cylinder head and body as an example, to analyze the transmission characteristics of vibration signal spectrum and the excitation source to the measurement point, high structure dimensional feature vector based on spectrum. With the average method of dimensionality reduction using the PCA method and the sub classification, verify the classification performance of support vector machine. The results show that compared with the traditional single channel sensor, this feature extraction method not only has the higher compressibility, and the classification accuracy is improved.

【作者单位】：西安石油大学机械工程学院;
【分类号】：TH17;TP212.9

【相似文献】