【作者】 张春辉；

【导师】 李伟；

【作者基本信息】 东北石油大学 ， 化工过程机械， 2012， 硕士

【摘要】 在石油工业生产过程中，工业设备的安全问题始终处于重要地位，而金属材料的腐蚀现象是生产设备安全的最大隐患。金属设备所接触的复杂环境加速了腐蚀现象的发生，进而增加了生产事故的发生机率。一旦在生产过程中发生事故，必然会给国家的财产造成损失，对社会造成恶劣的影响。因而世界各国每年都会在腐蚀防护与监测上投入大量的人力与物力。金属储罐底板的腐蚀现象包括点蚀和均匀腐蚀两部分。两种腐蚀现象都能对金属储罐底板造成严重的破坏，引发裂纹产生、腐蚀穿孔，进而导致介质泄漏，引起严重的事故。均匀腐蚀的发展速度虽然缓慢，但腐蚀面积大，并且均匀腐蚀过程中的气泡产生过程和金属溶解过程所产生的声发射信号具有代表性。Q235钢为金属储罐常用底板材料，对其均匀腐蚀过程中产生的声发射信号进行研究，探讨腐蚀声源的产生机制，有利于进行腐蚀防护与监测。本文经在查阅大量国内外腐蚀声发射研究文献后，建立了一套提取均匀腐蚀声发射信号的实验系统。分别对均匀腐蚀过程中气泡产生、破裂过程和金属溶解、生成物剥落过程进行声发射信监测。利用声发射参数分析法，分析了Q235均匀腐蚀气泡产生过程和金属溶解过程的声发射特征。运用小波分析在信号去噪和特征提取中的相关理论知识，结合实际声发射信号的特点采用阈值去噪方法剔出了噪声信号的干扰，并对降噪处理后的信号进行四层小波包分解，提取分解后各节点能量作为神经网络的输入。在结构上选取“紧致型”的小波神经网络，以Morlet小波函数作为隐含层激励函数，网络学习训练过程基于误差的逆向传播，按照梯度下降方向调整网络参数，同时避免了网络陷入局部最优解中,达到了对均匀腐蚀过程中金属溶解与气泡破裂过程的模式识别。更多还原

【Abstract】 In the production process of the oil industry, the security of industrial equipment hasalways been an important position. And for the safety of production equipment, the corrosionof metal materials is the biggest hidden trouble. the complex environment that metalequipment contact with accelerated corrosion phenomena, and the accidents in productionincreases. The occurring of an accident in the production process is bound to cause lossing theproperty of the state as well as the adverse impact to the society. And a lot of manpower andresources were put into the corrosion protection and monitoring every year in the wholeworld.In general, the corrosion of the metal tank bottom includes two parts, the pitting and theuniform corrosion. Both of the two corrosions can cause serious damage to the metal tankbottom. For example causing cracks and corrosion perforation, which can led to a media leakand causing a serious accident. Although the rate of the development of uniform corrosion isrelatively slow, the corrosion area is relatively large. And AE signals Collected from theprocess of bubble formation process and the metal dissolution process is representative in theuniform corrosion. The studying of acoustic emission signals generated in the process ofuniform corrosion will enable us to understand the corrosion mechanism better and better, anddo well in the protection and monitoring of corrosion.After reading a large number of domestic and foreign research literature on the corrosionacoustic emission we establish our own experimental system of extraction of uniformcorrosion on the acoustic emission signals. The experiment monitored the rupture process ofHydrogen bubble, the metal dissolution and the process of resultant flaking. According to theanalysis by the acoustic emission parameters, we study into the acoustic emissioncharacteristics of the bubble formation process and the metal dissolution in the metal uniformcorrosion. According to the theoretical knowledge of the wavelet analysis and thecharacteristics of the actual acoustic emission, we exclude the interference of the noise signalby using the threshold noise reduction method. We use the four-layer wavelet packet methoddecompose of the signal after the noise reduction processing, and extract the energy of eachnode as input of neural network. And we select the “compact” Wavelet Neural Network. Wemake morlet wavelet function as the hidden layer activation function. According to theapplication of the reverse spread error network learning and training process, we adjust thenetwork parameters following the direction of gradient descent. While avoiding the network into the local optimal solution, we have reached the pattern recognition between the processof metal dissolution and bubble rupture of uniform corrosion.更多还原