基于经验小波变换的流体管网泄漏多方向多模态声发射时频定位方法研究

The modal analysis models are developed based on mechanism of leakage-induced acoustic emission (AE) in combination with guide wave theory and vibration theory of cylindrical elastic thin shell, to predict modal types, displacement distribution, dispersive curves and attenuation characteristics of leakage-induced AE in fluid-filled pipe network under different conditions. The leakage-induced multi-directional AE data acquisition and fusion method is proposed according to modal displacement distribution to solve the problem of low signal-to-noise ratio caused by attenuation characteristics of AE. The multi-modal leak AE signals are decomposed into intrinsic mode set by the improved empirical wavelet transform (IEWT). The mechanisms of intrinsic modes selection are established based on leakage characteristics identification. The time-frequency distributions of selected intrinsic modes are obtained by the improved normalized Hilbert transform (INHT) to determine the time difference of arrival (TDOA) of different modes and corresponding modal frequency. The sound velocity of corresponding modes can be obtained by the modal frequency and theoretical prediction of dispersive curves. Hence, the distance between the leakage and sensors can be determined using the TDOA and sound velocity of two modes. Combining the distance between two sensors, the judgment criteria of leakage direction will be built to determine the leakage between and outside the two sensors. The existing leak location schemes in fluid-filled pipe network will be supplemented and perfected through the research in this project, which provides new solution to the problem of large leak location error due to multi-modal, dispersive and attenuation natures of leakage-induced AE signal.

从泄漏声发射产生机理出发，结合壳体振动和导波分析理论构建流体管网泄漏声发射模态分析模型，通过实验修正理论模型并优化数值分析方法，预测不同情况下泄漏声发射模态类型、位移分布、频散曲线及衰减特性。根据模态位移分布提出泄漏多方向声发射数据采集及融合方法，解决泄漏信号因衰减特性导致信噪比低的问题。采用改进的经验小波变换将泄漏多模态声发射信号分解为本征模态集，根据泄漏特征辨识构建本征模态筛选机制，采用改进的归一化Hilbert变换对筛选的本征模态进行时频分析，确定不同模态的到达时间差和对应的频率，结合频散曲线确定频率对应的模态声速，利用两模态到达时间差和模态声速确定泄漏点与传感器之间的距离，结合两传感器之间的距离建立泄漏点方向判定准则，确定两传感器之间和之外的泄漏位置。本课题的研究是对现有流体管网泄漏声发射定位方法的补充和完善，为解决泄漏信号多模态、频散及衰减特性导致定位误差大的问题提供新的思路。

本项目针对流体管网泄漏声发射信号的频散、多模态特性导致泄漏定位误差大的问题，采用模态分析理论与信号处理方法研究了流体管网泄漏辨识与定位方法，并开发了供水管网泄漏定位仪器系统。具体研究内容包括：1）研究流体管道不同方向的泄漏声发射信号模态特性，提出提出了基于拾振方向选取的单一模态泄漏定位方法，同时提出了基于不同振动方向的泄漏信号的互相关辨识方法，使相对定位误差稳定在1.38%左右。2）提出的流体管网泄漏声发射时频定位方法，与传统互相关方法相比误差减少了6倍；3）提出基于变分模态分解与支持向量机的多源信息融合辨识方法准确率达到98.75%，提出基于变分模态分解与互谱分析的供水管网泄漏定位方法，在低信噪比下实现泄漏定位误差稳定在2.53%以内；4）提出基于改进经验小波变换及互谱相位差谱的供水管道泄漏定位方法，采用小波包分解得到不同尺度的信号能量谱，根据小波包能量谱局部极小值的分布自适应确定频带分割区间，解决了传统经验小波变换中频谱划分问题，在低信噪比下供水管道泄漏定位误差减少到0.36%；5）提出基于四阶累积量及自适应滤波的供水管道泄漏信号时延估计方法，有效解决了相关噪声下互相关时延估计误差大的问题；6）研发了基于物联网的供水管网泄漏监测系统得到示范应用，对城市供水管网安全运行提供重要保障。