基于神经网络的自适应滤波技术及其在超声检测中的应用

为满足超声无损检测和评价在新材料和新构件上应用的需求,提出了一种基于神经网络的自适应滤波器,使其兼具自适应滤波和非线性处理的能力。该滤波器是在线性自适应滤波器中引入一非线性隐含层而构成,在对该滤波器的结构和收敛性进行详细研究的基础上,实现了一种自适应噪声消除器并应用于超声检测中的材料噪声消除。实验结果证实:利用这种滤波器构成的材料噪声消除器比采用线性自适应滤波器具有更强的降噪能力。

An Adaptive Filter Based on Artificial Neural Net and Its Application in Ultrasonic Testing

In order to meet the needs of the application of ultrasonic nondestructive testing and evaluation to new materials and structures,an adaptive filter based on artificial neural net (ANN),that possesses both adaptively filtering and nonlinearly processing abilities,is put forward.This filter is constituted by introducing a nonlinear hidden layer into the conventional linear adaptive filter,on the basis of studying on this adaptive filter's structure and its convergence property detailedly,an ANC (adaptive noise canceller) by means of this filter is successfully realized and used to cancel the material noise in the field of ultrasonic testing.The practical results testify that the ANC composed of this filter is able to cancel noise more effectively than using conventional linear adaptive filters.