Abstract: The security condition of oil storage tank can be assessed without opening pot by acoustic emission technology, and acoustic emission signal recognition is the basis of analysis of the corrosion status for oil storage tanks. Against deficiencies of the analysis method by parameters, a new acoustic emission signal recognition method was proposed based on wavelet transform and RBF neural networks. Acoustic emission signal was decomposed to 6 layer by db2 wavelet, and feature vector from acoustic emission signal was composed of the space energy based on 6 layer detail feature. RBF neural network was designed combining the characteristics of acoustic emission signal. The RBF network was trained by using of the acoustic emission signal which pattern has been known.Corrosion, crack and condensation acoustic emission signal were studied by RBF neural network, respectively.The results showed that the recognition rate of RBF neural network reached 93.3%, and the RBF network displayed superiority in identification of the acoustic emission signal, which had a certain significance for quantitative analysis on oil storage tank safety situation.

Key words: Acoustic emission,    ,  Wavelet,    ,  Feature,     ,  RBF neural network,   ,  Recognition

摘要： 声发射检测技术不需开罐就能对储油罐安全性在线评估, 声发射信号识别是储油罐腐蚀状况分析的 基础, 针对现有参数分析法的不足, 提出一种基于小波变换特征提取与 R B F神经网络识别的声发射信号识别方法。 利用d b 2小波对声发射信号6层分解, 将6层细节特征空间的能量作为声发射信号特征向量; 结合声发射信号特点 设计R B F神经网络, 利用已知模式声发射信号训练 R B F网络; 用 R B F神经网络对腐蚀、 裂纹和冷凝声发射信号进 行分类测试。实验结果表明, R B F网络的识别率达到9 3. 3%, 显示了R B F网络识别声发射信号的优越性。对储油罐 安全状况的定量分析具有一定意义。

关键词: 声发射,  , 小波, 细节特征, R B F神经网络, 识别