声发射(AE)技术能用来区分发生在受载材料内的不同损伤模式，而聚类分析能在无先验知识的情况下通过揭示数据内部结构对数据进行分类。声发射波形包含了丰富的声发射源信息，而常规的特征参数并不能满足深层次的声源识别要求。文章尝试从波形的频率分布特征、形状特征和强度特征三个方面分别选取小波变换能量特征系数、波形裕度因子和幅值作为描述声发射波形的新参数。基于波形新参数的聚类分析能有效地区分加氢反应器材料2.25Cr-1Mo带裂纹和无裂纹试件拉伸过程中屈服阶段塑性变形信号、微裂纹扩展信号和断裂失稳信号。

Acoustic Emission(AE) can be used to discriminate the different types of damage occurring in a constrained metal material. And cluster analysis can separate a set of data into several classes that reflect the internal structure of the data. AE waveform contains a wealth of information about AE source, and the traditional parameters can no longer meet the higher demands of AE source identification. In this paper, we worked hard to extract new parameters from three aspects: vector of frequency band energy extracted by wavelet transformation characterizes the frequency distribution of the waveform, waveform Margin factor characterizes shape feature of AE waveform and the amplitude characterizes intensity of AE waveform. We worked on specimens of hydrogenation reactor material 2.25Cr-1Mo, subjected to tensile loading, awaiting damage modes in the material. k-mean clustering based on new parameters was used to analyze the AE signals during the total procedure, and signals of plastic deformation at yield stage, micro-crack signals and crack signals at destructed stage were finally indentified.