摘要
采用基于奇异值分解和人工神经网络的多传感器数据融合方法对喷水推进泵的空化状态进行了分类识别研究。首先利用基于奇异值分解的权值估计算法分别对水声信号和振动信号在时间上进行数据级融合,提取出各自的特征,然后将所有特征组合起来作为神经网络的输入,利用BP网络和RBF网络进行特征级融合和分类识别。分析结果表明:基于多传感器数据融合的分类识别结果优于单传感器分类识别结果;采用基于奇异值分解的数据融合方法后,分类识别率显著提高,对空化初生微弱特征的识别效果尤佳。
The classification of cavitation in a waterjet pump was studied with multi-sensor data fusion method based on singular value decomposition (SVD) and artificial neural networks (ANN). Time data fusion was used to fuse acoustic and vibration signals based on SVD firstly, then the respective features were extracted, they were composed and used as inputs of an ANN, a BP network and a RBF network were applied to fuse the features for classification. The results showed that the classification performance based on multi-sensor data fusion is better than that based on a single sensor, and the recognition rates are obviously improved after using data fusing based on SVD, that is superior for weak feature detection and recognition of cavitation inception.
出处
《振动与冲击》
EI
CSCD
北大核心
2012年第18期93-95,121,共4页
Journal of Vibration and Shock
基金
国家自然科学基金资助项目(51009144)