基于小波神经网络的管道腐蚀缺陷定量识别研究

Quantitative Study of Pipeline Corrosion Defects Based on Wavelet-neural Network

漏磁检测是目前广泛采用的油气管道检测方法,通过漏磁检测仪获得腐蚀缺陷漏磁场切向分量的漏磁场信号,提取与腐蚀缺陷外形长、宽、深有关的漏磁场信号波形特征量,结合小波分析和神经网络的优势形成小波神经网络,分析设计了小波神经网络的结构,给出了网络训练算法,利用网络的非线性逼近性能,对腐蚀缺陷外形进行定量评价,给出预测评价曲线,试验验证方法有效可行。

At present, magnetic flux leakage test is widely used for oil gas pipeline defect detection. The tangential component of magnetic flux leakage testing signal around corrosion defect is detected by magnetic detector. Some features of magnetic flux leakage signals are extracted to evaluate and calculate the length, width and depth of defects. The wavelet neural network is a combination of wavelet analysis and neural network. The architecture of the wavelet neural network was designed, and then training algorithm was obtained. According to the non-linear property of wavelet-neural network, the quantitative shape assessment and predicted curve of corrosion defect can be given out. The method is reliable according to experimental verification.