基于空化辐射噪声的检测方法实验研究  

Experimental research on pump cavitation detection based on acoustic radiation

作　　者：刘进[1,2] 阎兆立[2] 程晓斌[2] 杨军[2] 乔文孝[1] 

机构地区：[1]中国石油大学(北京)理学院,北京102249 [2]中国科学院噪声与振动重点实验室(声学研究所),北京100190

出　　处：《应用声学》2014年第1期60-65,共6页Journal of Applied Acoustics

基　　金：国家自然科学基金委员会机械工程学科资助项目(51205404)

摘　　要：空化空蚀严重影响水利设备的正常工作和使用寿命,为此本文研究了基于水听器信号的水泵空化检测方法,从空化噪声辐射基本特性出发,分析并选取强度、脉冲以及频谱结构等特征参数组成最优分类特征向量,使其既有较强的稳定性,也有较好的灵敏度。以此特征向量训练获得的支持向量机(SVM)分类器进行水泵空化状态识别准确率平均能有99.6%。检测其他结构水泵的空化,识别准确率也在96.5%以上。在较高环境噪声单一能量特征无法识别的情况下,依然有较好的识别准确率。表明该方法对不同结构的水泵及较高环境噪声具有一定的鲁棒性,有较好的应用价值。Cavitation is a dynamic phenomenon typically associated with liquid. Researches on acoustic methods for monitoring pump cavitation have been doing for the purpose of regular work and working life. A method of pump cavitation detection based on hydrophone signal was studied. An optimal feature vector, including en- ergy, pulse, spectrum structure, was extracted based on the essential features of cavitation physical phenome- non and noise radiation in different degrees, which showed strong stability and good sensitivity. The experiment shows that the recognition accuracy average rate for cavitation condition is over 99.6% and the accuracy rate is over 96. 5 % when the classifier is used to identify cavitation of other pumps. The cross-validation experiment shows that the proposed classifier has a good generality. The classifier is competent even in strong background noise. It shows that the method is robust for different pumps and different noise condition which has a better practical value.

关 键 词：空化检测 声辐射特征 分类准确率 

分 类 号：O427.4[理学—声学]