摘要
声发射技术能够监测焊接冷裂纹,但由于焊后冷却过程中干扰信号很多,并且监测时间长,使得人工分析、评价困难较大。以5个典型声发射信号参量为输入单元、开裂信号和噪声信号特性为输出单元,建立了一个能识别焊接冷裂纹开裂信号的BP神经网络。通过对SPV490Q钢平板刚性拘束焊接裂纹试验的数据进行训练和测试,验证了该网络的可行性。
Acoustic emission technique is capable to monitor cold cracks. But in the long-time coolingprocess after welding, there exist a host of interference signals which make it harder to analysis and evalu-ate the signals. Thus, a BP neural network which can recognize welding cold crack signals was estab-lished. Its input unit was constituted by 5 typical parameters of the acoustic emission signals, as well asoutput unit was constituted by characteristics of crack signals and interference signals. Through trainingand testing the data from the experiment of SPV490Q steel plate with rigid restraint, the feasibility of theneural network was confirmed.
出处
《压力容器》
2016年第3期51-55,共5页
Pressure Vessel Technology
基金
黑龙江省博士后科研启动项目(LBH-Q14031)
