论文部分内容阅读
为了准确地检测城市燃气管道泄漏,提出了一种基于广义概念的管道泄漏检测定位方法。声发射技术对于管道泄漏的检测、定位是一个极好的工具,但由于泄漏源的传播容易受到周围背景噪声以及复杂工况的影响,其定位误差较大。基于时延估计的互相关信号处理方法被广泛用于管道泄漏检测定位,但由于泄漏应力波传播通道的动态特性,使得源信号在传播过程中会产生波形变化,给互相关函数峰值位置的确定带来困难。由此引入广义相关分析方法,通过对信号进行前置滤波,在一定程度上减少了传播通道动态特性因素对泄漏点定位的不利影响,得到了更为准确的时延估值。在此基础上,通过模拟实验,编写Matlab神经网络代码,构造GRNN模型,进一步预测定位。结果表明,GRNN预测的声发射检测值、互相关定位值以及广义相关定位值,相比之前定位精度分别得到提高,其中基于广义相关的延时估计方法定位最为精确,将该方法用于工程实际中,可以更加精确地定位出泄漏点。
In order to accurately detect urban gas pipeline leaks, a pipeline leakage detection and localization method based on generalized concept is proposed. Acoustic emission technology for pipeline leak detection, positioning is an excellent tool, but because of the spread of leakage sources are easily affected by the ambient background noise and complex conditions, the positioning error. The cross-correlation signal processing method based on time-delay estimation is widely used in pipeline leak detection and localization. However, due to the dynamic characteristics of the leaky stress wave propagation channel, the waveform of the source signal is changed during the propagation process and the peak position of the cross-correlation function is determined Bring difficulties Therefore, the generalized correlation analysis method is introduced. By filtering the signal pre-filtering, the unfavorable influence of the dynamic characteristics of the propagation channel on the location of the leakage point is reduced to a certain extent, and a more accurate delay estimation is obtained. On this basis, through simulation experiments, the preparation of Matlab neural network code, construct the GRNN model, and further predict the location. The results show that the acoustic emission detection value, cross-correlation positioning value and generalized relative positioning value of GRNN prediction are respectively improved compared with the previous ones. The method of delay estimation based on generalized correlation is the most accurate and the method is applied to engineering practice In, you can locate the leak more accurately.