基于漏磁信号分析的储罐钢板缺陷尺寸评估  

Dimensional Evaluation of Steel Storage Bank Defect Based on Analysis in Magnetic Leakage Signal Characteristics

作　　者：陈建升 杨涛[1] 张宏杰[1] 巩文东 CHEN Jian-sheng;YANG Tao;ZHANG Hong-jie;GONG Wen-dong(School of Mechanical Engineering,Tianjin Polytechnic University,Tianjin 300387,China)

机构地区：[1]天津工业大学机械工程学院,天津300387

出　　处：《仪表技术与传感器》2018年第9期114-118,共5页Instrument Technique and Sensor

摘　　要：针对钢制储罐腐蚀缺陷的在线检测和评价,首先建立了以ARM单片机和磁场检测传感器为核心的漏磁信号采集系统,并且针对一系列局部磁化的人造凹坑缺陷进行二维漏磁数据同步检测;为了实现凹坑缺陷几何尺寸的预测,创新性地关注了漏磁信号的几何学图形特征,利用Fourier级数展开拟合了漏磁信号曲线,提取了曲线包络面积、周长、图形重心等几何特征;随后针对40个缺陷凹坑样本,建立了表征凹坑缺陷的多维数据集;最后利用BP神经网络建立了凹坑缺陷几何尺寸的预测模型。研究结果表明在小数据样本情况下,利用漏磁信号几何学图形化特征,能够有效挖掘漏磁信号中隐含特征参量,该方法可行、可信,丰富和完善了钢制储罐腐蚀缺陷尺寸评价方法。Aiming at on-line detection and evaluation of corrosion defects in steel storage tanks,firstly,a magnetic flux leakage signal acquisition system based on ARM microcontroller and magnetic field detection sensor was established.For a series of artificial pits locally magnetized,two-dimensional MFL signals were measured synchronously.To realize the size estimation of the pit defect mentioned above,the geometric graphic features of the MFL signals were concerned creatively.The MFL signal curves were fitted through the Fourier series expansion technique and some geometrical characteristics were extracted,such as envelop area,perimeter,geometric gravity center,etc.Then,forty pit defect samples were selected and a multi-dimension data set was constructed.Finally,the BP artificial neural network technology was used to develop the size prediction model for the pit.Test results show that the proposed method,which combines the features in the time-domain and graphics of the MFL signals,can mine the concealed feature within the MFL signal and it is feasible and reliable,especially for the small samples condition.The newly developed method also can enrich and perfect the size evaluation for the corrosion defect of the steel storage tank.

关 键 词：BP神经网络 三维霍尔磁场传感器 钢板凹坑缺陷 漏磁检测 定量化预测 

分 类 号：TM93[电气工程—电力电子与电力传动]