基于BP神经网络的电场指纹点蚀信号补偿系数

Compensation factor on pitting corrosion via field signature method based on BP neural network

电场指纹技术是一种依据被测对象表面微小电压变化,对金属结构的缺陷、裂纹、腐蚀及其扩展情况进行检测的无损检测技术,但在点蚀信号的辨识方法上依然存在不足。通过对电场指纹点蚀信号进行数值模拟和仿真建模,根据其特征信号对点蚀进行准确定位,基于BP神经网络建立关键点蚀参数的补偿系数模型,由补偿系数计算电场指纹系数,得出实际腐蚀深度和剩余壁厚。结果表明通过试验验证提出的补偿系数模型计算方法得到的最终结果误差较小,解决了传统经验系数中普适性差、精准度不高的问题。

The field signature method(FSM)is a new nondestructive detection method.It mainly detects flaw,crack,corrosion of metal structure based on the tiny voltage change of the measured object surface.However,some shortages still exist in the identification method of the pitting signal.In the present study,through the numerical simulation and modeling on the electric field fingerprint pitting signal,the pitting was accurately located according to its characteristic signal.Then the compensation coefficient model of the key point corrosion parameters was established based on the BP neural network method.The actual corrosion depth and residual wall thickness were obtained by the compensation coefficient to calculate the fingerprint coefficient value.The results show that the relative error between the proposed compensation coefficient by experimental verification and the simulation results is small,which solves the problems of poor universality and low precision on the traditional empirical coefficient.