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  

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

作　　者：杨克己[1] 

机构地区：[1]浙江大学现代制造工程研究所,杭州310027

出　　处：《仪器仪表学报》2005年第8期813-817,共5页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金项目(50375144);国家"863"高技术研究发展计划项目(2002AA421110);浙江省自然科学基金项目(501129)资助。

摘　　要：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。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 tilter,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.

关 键 词：神经网络 自适应滤波器 超声检测 材料噪声 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] TN911.7[自动化与计算机技术—控制科学与工程]