摘要
应用有限元分析和BP神经网络分析主动加热式红外无损检测中影响缺陷定量的因素。对背部开有不同深度、不同直径沉孔缺陷的平板表面采用恒定热流加热。使用有限元分析软件Ansys计算得到表面温度分布云图和温升。将计算数据作为样本训练用于缺陷定量分析的BP神经网络,测试表明神经网络方法对于未知缺陷深度的识别相当有效。研究表明该信号处理技术对红外无损检测具有工程应用价值。
Finite element analysis combined with back propagation（BP） neural active heating infrared nondestructive testing for determing the influencing factors network were applied to the on defect quantitation. Finite element analysis software Ansys, was used to calculate the temperature field of the plates with back sinking hole defects of different depth and diameter under constant heating condition, and surface temperature cloudy map and temperature rise were obtained. The data obtained were used as the sample to train the BP neural network for quantitative evaluation of defects. It was showed by testing that neural network was very effective in predicting the depth of defects. The signal processing method was valuable for engineering application to infrared nondestructive testing.
出处
《无损检测》
北大核心
2006年第1期2-7,共6页
Nondestructive Testing
基金
海军工程大学科学研究基金资助(E344)
关键词
红外热成像
有限元分析
神经网络
信号处理
缺陷定量
Infrared thermography
Finite element analysis
Neural network
Signal processing
Quantitative evaluation of defect
