摘要
本文提出了一种基于声学信号功率谱分析和聚类支持向量机的检测管道损坏导致气体泄漏的新方法。该方法以现场采集的声学信号功率谱作为聚类支持向量机的输入,首先分析了环境变量对信号频谱的影响,然后,通过对现场不同环境下采集的信号采用自组织特征映射网络算法进行聚类分析,将检测空间分为若干独立子空间,再针对每一子空间构造一个支持向量机检测模型,并以对应子空间的样本集训练支持向量。基于某炼化厂采集数据所进行的实验表明,该方法在不同的现场环境下均具有较好的检测性能。
This paper proposes a novel method for detecting gas leakage from cracks in pipes using acoustic signal power spectrum analysis and clustering support vector machine （CSVM）. After sampling the acoustic signal in the fields, the data are processed using FFT method, and the influences of environmental factors on the frequency char- acteristics of the signal spectra are discussed, then CSVMs are built specifically for each subsets partitioned intelli- gently by adopting the self-organizing feature map （SOFM） clustering algorithm. This feature makes learning tasks for each CSVM more specific and simpler. Experimental results performed in a petroleum refining plant show that the proposed method can achieve good detection performances under different circumstances.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2007年第11期2028-2033,共6页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(50477010)
日本文部省科学研究费补助金(基盘研究B16360199)资助项目
关键词
气体泄漏
检测
声学信号
聚类支持向量机
gas leakage
detection
acoustic signal
clustering support vector machine
