基于时频熵特征的2.25Cr-1Mo损伤开裂声发射识别方法研究

Acoustic Emission Identification of 2.25Cr-1Mo Crack based on Time-Frequency Entropy

通过三点弯曲试验来模拟加氢反应器等高压容器在运行时的应力腐蚀开裂情况,根据力与时间的关系,将试件的受力分为弹性阶段、塑性阶段和断裂分离阶段,在每个阶段分别提取10组数据作为样本,采用信息熵和Hilbert-Huang变换相结合的方法对采集到的样本进行时频熵特征提取,以此来建立基于支持向量机(SVM)的智能识别方案。测试结果表明,采用时频熵特征提取的方法,即使是小样本,也达到了96.7%的识别精度。

The stress corrosion cracking of high-pressure vessels such as hydrogenation reactors was simulated by three-point bending test.According to the relationship between stress and time curve,the stress of the specimen was divided into elastic stage,plastic stage and fracture separation stage.Ten groups of data were extracted in each stage as samples,and the method of information entropy and Hilbert-Huang transform was applied to analyze the collected samples.Based on the feature extraction of time-frequency entropy,intelligent recognition scheme of support vector machine(SVM)was established.The test results showed that the recognition accuracy by using the method of time-frequency entropy feature extraction was 96.7%even for small samples.