Ultrasonic Signal Pattern Recognition of Pipeline Corrosion Defects with PSO-SVM
-
摘要: 针对金属管道腐蚀问题,提出了一种基于支持向量机(Support vector machine,SVM)与粒子群优化(Particle swarm optimization,PSO)相结合的管道腐蚀缺陷的分类方法。对预处理后的超声缺陷信号进行经验模态分解(Empirical mode decomposition,EMD),提取相应的时域无量纲参数作为特征向量;建立SVM缺陷分类模型,并采用PSO算法优化SVM参数,提高模型的缺陷分类准确率。实验证明,该方法建立的模型针对不同深度的超声缺陷信号的识别率达到87.5%,优于相同试验样本下BP神经网络和RBF神经网络的分类准确率。Abstract: In order to solve the metal pipeline corrosion, a classification method of pipeline corrosion defects based on the support vector machine (SVM) and particle swarm optimization (PSO) is proposed. The time domain dimensionless parameters are extracted by using the empirical mode decomposition (EMD) for the pre-processed ultrasonic defect signals as the eigenvectors. SVM defect classification model is established, and PSO algorithm is used to optimize the SVM parameters to improve the accuracy of defect classification. The experimental results show that the recognition rate of the present model for the ultrasonic defect signals under different depths is of 87.5%, which is better than the classification accuracy of BP neural network and RBF neural network with the same experimental samples.
-
表 1 时域无量纲参数
No. 统计参数 公式 1 斜度 2 峰度 3 峰值指标 4 清除指标 5 形状指标 6 脉冲指标 表 2 缺陷样本详情
缺陷腐蚀面积/mm 缺陷腐蚀形状 缺陷最大横截面积/mm2 无缺陷 - - 2 圆形 40 5 矩形 80 8 无规则形状1 120 无规则形状2 160 无规则形状3 无规则形状4 表 3 缺陷回波信号数据集分布
缺陷腐蚀面积/mm 缺陷样本数量 样本信号数据数量 样本信号数据编号 无缺陷 24 240 1~240 2 24 240 241~480 5 24 240 480~720 8 24 240 720~960 表 4 测试集识别结果
缺陷实际情况 无缺陷 2 mm 5 mm 8 mm 识别率/% 测试集预测结果 无缺陷 60 0 0 0 87.5 2 mm 0 58 9 9 5 mm 0 1 44 3 8 mm 0 1 7 48 敏感度/% 100 96.67 73.33 80 表 5 不同寻优算法性能对比
寻优算法 识别率/% 耗时/s GA 86.67(208/240) 84.15 GS 86.25(207/240) 73.63 PSO 87.50(210/240) 53.47 表 6 不同模型的分类性能对比
模型 识别率/% 耗时/s BP-NN 78.33(188/240) 47.55 RBF-NN 74.17(178/240) 43.97 PSO-SVM 87.50(207/240) 53.47 -
[1] Barbian A, Beller M. In-line inspection of high pressure transmission pipelines: State of the art and future trends[C]//18th World Conference on Nondestructive Testing. Durban, South Africa, 2012: 16-20 https://www.researchgate.net/publication/265999709_In-Line_Inspection_of_High_Pressure_Transmission_Pipelines_State-of-the-Art_and_Future_Trends [2] Duisterwinkel E H A, Talnishnikh E, Krijnders D, et al. Sensor motes for the exploration and monitoring of operational pipelines[J]. IEEE Transactions on Instrumentation and Measurement, 2018, 67(3):655-666 doi: 10.1109/TIM.2017.2775404 [3] 车红昆, 吕福在, 项占琴.超声检测信号多特征SVM-Bayes融合识别[J].振动与冲击, 2011, 30(12):265-269 doi: 10.3969/j.issn.1000-3835.2011.12.053Che H K, Lv F Z, Xiang Z Q. Ultrasonic signal recognition by multiple features SVM-Bayes fusion method[J]. Journal of Vibration and Shock, 2011, 30(12):265-269(in Chinese) doi: 10.3969/j.issn.1000-3835.2011.12.053 [4] Meng M, Chua Y J, Wouterson E, et al. Ultrasonic signal classification and imaging system for composite materials via deep convolutional neural networks[J]. Neurocomputing, 2017, 257:128-135 doi: 10.1016/j.neucom.2016.11.066 [5] 黄晶, 阙沛文.小波分析在管道缺陷超声检测中的应用[J].传感技术学报, 2003, 16(3):263-266 doi: 10.3969/j.issn.1004-1699.2003.03.007Huang J, Que P W. The application of wavelet analysis in ultrasonic testing of pipeline defect[J]. Journal of Transcluction Technology, 2003, 16(3):263-266(in Chinese) doi: 10.3969/j.issn.1004-1699.2003.03.007 [6] 丁攀, 吕福在, 项占琴.基于小波包分解和支持向量机的石油套管缺陷智能识别[J].钢铁研究学报, 2012, 24(5):58-62 http://d.old.wanfangdata.com.cn/Periodical/gtyjxb201205012Ding P, Lv F Z, Xiang Z Q. Intelligent flaws identification method for oil casing pipe based on wavelet packet decomposition and support vector machine[J]. Journal of Iron and Steel Research, 2012, 24(5):58-62(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/gtyjxb201205012 [7] 赵永林, 刘桂雄, 周德光, 等.应用经验模式分解法处理超声无损检测信号[J].现代制造工程, 2006(4):90-92 doi: 10.3969/j.issn.1671-3133.2006.04.032Zhao Y L, Liu G X, Zhou D G, et al. Ultrasonic nondestructive test signal processing based on empirical mode decomposition[J]. Modern Manufacturing Engineering, 2006(4):90-92(in Chinese) doi: 10.3969/j.issn.1671-3133.2006.04.032 [8] 李秋锋, 黄攀, 施倩, 等.基于经验模态分解去噪的粗晶材料超声检测[J].应用基础与工程科学学报, 2014, 22(3):566-573 http://www.cnki.com.cn/Article/CJFDTotal-YJGX201403015.htmLi Q F, Huang P, Shi Q, et al. Ultrasonic testing of coarse-grained materials based on EMD denoising method[J]. Journal of Basic Science and Engineering, 2014, 22(3):566-573(in Chinese) http://www.cnki.com.cn/Article/CJFDTotal-YJGX201403015.htm [9] 陈天璐, 阙沛文, 金涛, 等.管道缺陷多超声传感器检测数据的神经网络融合[J].传感技术学报, 2004, 17(3):403-406 doi: 10.3969/j.issn.1004-1699.2004.03.012Chen T L, Que P W, Jin T, et al. Neural network-style fusion of pipelines' defect data scanned by multi-ultrasonic sensors[J]. Journal of Transcluction Technology, 2004, 17(3):403-406(in Chinese) doi: 10.3969/j.issn.1004-1699.2004.03.012 [10] 唐东林, 魏子兵, 潘峰, 等.基于PCA和SVM的管道腐蚀超声内检测[J].传感技术学报, 2018, 31(7):1040-1045 doi: 10.3969/j.issn.1004-1699.2018.07.011Tang D L, Wei Z B, Pan F, et al. Ultrasonic internal detection of pipeline corrosion based on PCA and SVM[J]. Chinese Journal of Sensors and Actuators, 2018, 31(7):1040-1045(in Chinese) doi: 10.3969/j.issn.1004-1699.2018.07.011 [11] 戴波, 赵晶, 周炎.基于支持向量机的管道腐蚀超声波内检测[J].化工学报, 2008, 59(7):1812-1817 doi: 10.3321/j.issn:0438-1157.2008.07.038Dai B, Zhao J, Zhou Y. Ultrasonic in-line inspection of pipeline corrosion based on support vector machine[J]. Journal of Chemical Industry and Engineering (China), 2008, 59(7):1812-1817(in Chinese) doi: 10.3321/j.issn:0438-1157.2008.07.038 [12] 刘清坤, 阙沛文, 郭华伟, 等.基于支持向量机和特征选择的超声缺陷识别方法研究[J].中国机械工程, 2006, 17(1):9-12 doi: 10.3321/j.issn:1004-132X.2006.01.003Liu Q K, Que P W, Guo H W, et al. Support vector machine and chaos-genetic optimization based ultrasonic flaw identification[J]. China Mechanical Engineering, 2006, 17(1):9-12(in Chinese) doi: 10.3321/j.issn:1004-132X.2006.01.003 [13] 熊庆, 张卫华.基于MF-DFA与PSO优化LSSVM的滚动轴承故障诊断方法[J].振动与冲击, 2015, 34(11):188-193 http://d.old.wanfangdata.com.cn/Periodical/zdycj201511033Xiong Q, Zhang W H. Rolling bearing fault diagnosis method using MF-DFA and LSSVM based on PSO[J]. Journal of Vibration and Shock, 2015, 34(11):188-193(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zdycj201511033 [14] 姜久亮, 刘文艺, 侯玉洁, 等.基于内积延拓LMD及SVM的轴承故障诊断方法研究[J].振动与冲击, 2016, 35(6):104-108 http://d.old.wanfangdata.com.cn/Periodical/zdycj201606019Jiang J L, Liu W Y, Hou Y J, et al. Bearing fault diagnosis based on integral waveform extension LMD and SVM[J]. Journal of Vibration and Shock, 2016, 35(6):104-108(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zdycj201606019 [15] Huang N E, Shen Z, Long S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings of the Royal Society of London. Series A:Mathematical, Physical and Engineering Sciences, 1998, 454(1971):903-995 doi: 10.1098/rspa.1998.0193 [16] Tsao W C, Li Y F, Du Le D, et al. An insight concept to select appropriate IMFs for envelope analysis of bearing fault diagnosis[J]. Measurement, 2012, 45(6):1489-1498 doi: 10.1016/j.measurement.2012.02.030 [17] 徐卓飞, 刘凯, 张海燕, 等.基于经验模式分解和主元分析的滚动轴承故障诊断方法研究[J].振动与冲击, 2014, 33(23):133-139 http://d.old.wanfangdata.com.cn/Periodical/zdycj201423025Xu Z F, Liu K, Zhang H Y, et al. A fault diagnosis method for rolling bearings based on empirical mode decomposition and principal component analysis[J]. Journal of Vibration and Shock, 2014, 33(23):133-139(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zdycj201423025 [18] Cortes C, Vapnik V. Support-vector networks[J]. Machine Learning, 1995, 20(3):273-297 http://d.old.wanfangdata.com.cn/Periodical/hwyhmb200803006 [19] Moosavian A, Khazaee M, Najafi G, et al. Spark plug fault recognition based on sensor fusion and classifier combination using Dempster-Shafer evidence theory[J]. Applied Acoustics, 2015, 93:120-129 doi: 10.1016/j.apacoust.2015.01.008 [20] 温正, 孙华克.MATLAB智能算法[M].北京:清华大学出版社, 2017Wen Z, Sun H K. MATLAB intelligent algorithm[M]. Beijing:Tsinghua University Press, 2017(in Chinese)