储罐底板钢点蚀过程中声发射信号的聚类分析

Cluster analysis of acoustic emission signals during tank bottom steel pitting corrosion process

联合采用声发射和电化学技术研究储罐底板钢试样在w(NaCl)=3.0%,pH=2.0的酸性溶液中的点蚀特征,基于K-均值聚类算法对点蚀声发射信号特征参数进行聚类分析,从而提取各类信号的自身特征。将分类后的信号作为样本训练BP人工神经网络,成功对平行试验采集的声发射信号进行识别。研究结果表明,底板钢在酸性条件下的点蚀过程主要产生氢气泡、膜破裂和蚀坑生长这3类典型的声发射信号,通过聚类方法可以区分这3类信号,并能用神经网络对声发射源进行有效识别。这对现场常压金属储罐底板腐蚀声发射检测结果的解释和评价具有指导意义,有助于提高检测结果可靠性,降低储罐运行风险,保证其运行安全。

The pitting characteristics of tank bottom steel sample were studied by combined acoustic emission( AE) and elec-trochemical techniques in acidic NaCl solution (w=3. 0%, pH=2. 0). The AE signals characteristic parameters were classi-fied using K-means clustering algorithm and each cluster signal characteristic was also extracted. The classified signals were trained using BP artificial neural network,and the AE signals from parallel experiments were successfully identified. The re-sults show that the oscillation, movement and burst of hydrogen bubbles, breakage of passive film, growth and propagation of pit are the typical AE sources in pitting, which could be effectively classified using cluster analysis and identified by artificial neural network. It has guiding significance for interpreting and evaluating the AE on-site testing result of bottom corrosion of atmospheric storage tank, improving the reliability of testing result, reducing risk and ensuring the safety of tank.