摘要
为使超声波无损检测准确识别摩擦焊接头的弱结合缺陷,本文利用小波包分析法分析处理超声波检测的回波信号,利用“能量-故障”法提取各检测信号的信号特征,将信号特征向量作为输入量引入到利用改进算法所构建的信号识别神经网络中,使其对接头中的缺陷进行后端分类和识别处理,从而实现了摩擦焊接头缺陷的智能识别。实验结果表明该方法具有较高的准确度。
In order to improve the veracity of testing the weak defection signal of the friction welding joints by ultrasonic non-destructive testing, the echo signal is analyzed based on modern wavelet packet analysis. Furthermore, a signal-classing neural network is constructed by improved BP algorithm, besides, the character of signal is extracted and its character vector is inputted into it. The result proved that various defections can be distinguished exactly.
出处
《微计算机信息》
北大核心
2007年第06S期283-285,共3页
Control & Automation
关键词
摩擦焊接头
超声波无损检测
小波包分析
神经网络
BP算法
friction-welding joints
ultrasonic non-destructive testing
wavelet packet analysis
neural network
BP algorithm
