摘要
人工神经网络技术在涡流检测中的应用越来越广泛。文中将多层神经网络BP算法应用到主极裂纹涡流无损检测的信号处理中,并对具体的神经网络的结构进行了设计,主要包括输入和输出层的设计,网络数据的准备,网络初始权值的选择,隐含层数及隐含层节点数的设计,网络的训练、检测及性能评价等。结果表明所设计的网络能够对主极裂纹信号进行有效的自动识别,而且识别的准确度很高。
Neural network technology is used more and more widely in eddy current testing. In this paper multiplayer neural network with BP algorithm is utilized to eddy current signal processing of the electrode. The structure of the concrete neural network is designed, mainly including the design of input layers and output layers, the preparing of network data, the choose of initial weight of network, number design of the hidden layers and the hidden layers' nodes, training , testing and judging of the network. The result proves the designed network can identify the crack signals of electrode automatically and effectively. Moreover the accuracy is very high.
出处
《计算机测量与控制》
CSCD
2004年第9期816-818,825,共4页
Computer Measurement &Control
