【作者】 范晓静； 张来斌； 梁伟； 王朝晖；

【Author】 Fan Xiaojing Zhang Laibin Liang Wei Wang Zhaohui (College of Mechanical and Electronic Engineering in the China University of Petroleum Beijing 102249 China)

【机构】 中国石油大学(北京)机电工程学院；

【摘要】 针对小样本情形下难以建立可靠的管道泄漏检测识别模型,提出了基于支持向量机(SVM)的管道泄漏负压波检测方法,SVM可较好的解决非线性数据分类问题,在小样本和二元分类方面优势突出。本文将SVM方法用于小样本情况下管道泄漏信号检测,实验表明,SVM方法在学习速度、分类效果、泛化能力方面均优于传统神经网络方法,非常适合于管道泄漏检测。更多还原

【Abstract】 When the number of leakage samples is small,it is difficult to establish a reliable model to detect leakage. In this paper a new method based on SVM to detect leakage is proposed.SVM resolves classification of non-linear data well and has advantages in classification of few samples.In this paper,SVM is used for negative pressure wave leak detection.Compared with neural network method,the experiment results show that SVM classifier has higher classification accuracy and better generalization performance.Therefore,it is more applicable to pipeline leak detection.更多还原