两种基于模式识别的枝状燃气管网泄漏定位方法

燃气管网的泄漏检测及定位研究一直是关注的焦点，但至今尚无较好的办法来实现管网的泄漏检测和定位。为此，介绍两种基于模式识别的枝状燃气管网泄漏检测与定位方法。利用小波分析对枝状燃气管网各端点获取的压力波信号进行消噪处理和奇异性分析，获取各端点压力突变点；再分别用加权最小距离分类器和神经网络法对泄漏点进行识别和定位。加权最小分类器法是基于标准化欧式距离和各属性权值计算基础上，将待检时间向量判定为与之距离最近的类，从而实现对泄漏点的定位；神经网络法分别选择BP网络和线性网络来判定泄漏点发生的管段及在该管段中的具体位置。仿真计算结果表明，两种方法获得的定位结果一致，证明了两种方法的有效性和精确性。

TWO LEAKAGE POSITIONING METHODS FOR DENTRITIC GAS PIPELINE NETWORK BASED ON PATTERN RECOGNITION

The leakage inspection and positioning have always been the focus of attention. However, there still are not good methods by means of which to realize the leakage inspection and positioning of pipeline network. Thus, this paper introduced two kinds of leakage inspection and positioning methods for the dentritic gas pipeline network on the basis of pattern recognition. Wavelet analysis was applied to conduct noise abatement and singularity analysis on the pressure wave of the each endpoint of the dentritic gas pipeline network. The pressure catastrophe point of each endpoint was acquired. Then the leakage points were recognized and positioned via weighted minimum distance classifier and neural network. Based on the calculation of standardized Euclidean distance and attribute weight, the weighted minimum distance classifier defined the unexamined time arrow as the nearest class, thereby to realize the positioning of leakage point. The neural network method determined the pipe section where the leakage took place and positioned the specific location of the pipe section. The simulated calculation findings showed that the positioning results of the two methods were identical, which manifested the validity and the accuracy of the two methods.