高级检索

基于PSO-SVM的点焊接头强度自动分类

吴刚, 陈天, 余靓辉, 柳志鹏

吴刚, 陈天, 余靓辉, 柳志鹏. 基于PSO-SVM的点焊接头强度自动分类[J]. 焊接学报. DOI: 10.12073/j.hjxb.20220829001
引用本文: 吴刚, 陈天, 余靓辉, 柳志鹏. 基于PSO-SVM的点焊接头强度自动分类[J]. 焊接学报. DOI: 10.12073/j.hjxb.20220829001
WU Gang, CHEN Tian, YU LiangHui, LIU Zhipeng. Research on automatic classification of spot welding joint strength based on PSO-SVM[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION. DOI: 10.12073/j.hjxb.20220829001
Citation: WU Gang, CHEN Tian, YU LiangHui, LIU Zhipeng. Research on automatic classification of spot welding joint strength based on PSO-SVM[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION. DOI: 10.12073/j.hjxb.20220829001

基于PSO-SVM的点焊接头强度自动分类

详细信息
    作者简介:

    吴刚,博士,副教授,硕士研究生导师;主要研究方向为智能焊接与无损检测;Email: 2623800082@qq.com

  • 中图分类号: TG 453.9

Research on automatic classification of spot welding joint strength based on PSO-SVM

  • 摘要: 点焊是汽车零部件的主要连接方式之一,其接头的拉伸和剪切强度是评价连接质量的重要因素. 在制备大量点焊试样的基础上,对各试样接头进行超声信号检测. 运用信号处理方法获得时域、频域和小波包特征值,通过对点焊试样进行拉剪试验,建立了点焊接头拉剪强度的分级标准. 根据试验数据训练了BP神经网络和基于粒子群优化支持向量机(particle swarm optimization support vector machines,PSO-SVM)的神经网络分类器. 最后通过输入不同数据量的特征值参数,比较两种神经网络模型对点焊强度分类的精度. 结果表明,结合9个超声信号特征值的PSO-SVM神经网络点焊强度分类精度达到95%.
    Abstract: Spot welding is one of the main connection methods of automobile parts, and the tensile and Shear strength of its joints are important factors to evaluate the connection quality. On the basis of preparing a large number of spot welding samples, this investigation conducted ultrasonic signal detection on the joints of each sample. Using signal processing methods to obtain time-domain, frequency-domain, and wavelet packet eigenvalues, a grading standard for the tensile and shear strength of spot welded joints was established by conducting tensile and shear tests on spot welded samples. According to the test data, BP neural network and neural network classifier based on Particle swarm optimization support vector machines (PSO-SVM) are trained. Finally, the accuracy of two neural network models for spot welding strength classification was compared by inputting feature value parameters with different data set. The experimental results show that the PSO-SVM neural network combined with 9 ultrasonic signal eigenvalues has a spot welding strength classification accuracy of 95%.
  • 图  1   点焊试样示意图

    Figure  1.   Schematic diagram of spot welding sample

    图  2   点焊试验样本

    Figure  2.   Spot welding test sample

    图  3   超声探头检测点焊试样

    Figure  3.   Ultrasonic testing probe for spot welding sample

    图  4   超声检测仪及软件

    Figure  4.   Ultrasonic tester and testing software

    图  5   点焊超声检测信号

    Figure  5.   Ultrasonic testing signal of spot welding

    图  6   点焊超声信号三层小波包变换子带信号

    Figure  6.   Three layer wavelet packet transform subband signal of spot welding ultrasonic signal. (a) S1; (b) S2; (c) S3; (d) S4; (e) S5; (f) S6; (g) S7; (h) S8

    图  7   点焊拉剪失效模式

    Figure  7.   Failure mode of spot welding tensile shear. (a) base material tearing; (b) nugget extraction; (c) interfacial tearing

    图  8   载荷—位移曲线

    Figure  8.   Load— displacement curve

    图  9   点焊试样拉剪强度分布

    Figure  9.   Tensile and shear strength distribution of spot welding sample

    图  10   基于PSO算法优化的SVM神经网络模型流程图

    Figure  10.   Chart of SVM neural network model optimized based on PSO algorithm

    图  11   BP神经网络测试集分类结果

    Figure  11.   Classification results of BP neural network test set. (a) time domain characteristic value;(b) frequency domain characteristic value;(c) wavelet packet domain characteristic value;(d) all characteristic value

    图  12   PSO-SVM神经网络测试集分类结果

    Figure  12.   Classification results of PSO-SVM neural network test set. (a) time domain characteristic value;(b) frequency domain characteristic value;(c) wavelet packet domain characteristic value;(d) all characteristic value

    图  13   PSO-SVM 寻找最佳参数的适应度曲线

    Figure  13.   Fitness curve of PSO-SVM to find the best parameters. (a) partial eigenvalue; (b) all eigenvalue

    表  1   点焊超声检测时频特征值及最大拉剪强度

    Table  1   Time domain and frequency domain characteristic value and maximum tensile shear strength of spot welding ultrasonic testing

    序号时域特征值频域特征值小波包特征值拉剪强度
    $ \bar S/mm $$ K $$ {f_m}/MHz $$ {A_m} $$ Kurt $$ \bar E $$ {S_E}^2 $$ \bar \alpha (dB/mm) $$ {S_a}^2 $$ L/KN $
    12.740.1361.1720.6030.4090.1380.3212.520.6816.67
    22.650.1581.1720.4560.3890.6560.8152.820.3518.72
    32.650.0421.1720.6640.5270.9500.2351.750.6719.06
    42.640.0671.1720.9740.4950.8650.3100.961.9214.32
    52.560.3541.1720.6550.4930.5150.8272.131.1713.14
    62.570.431.1720.6610.4770.2190.8260.961.1611.62
    72.790.2071.3670.5660.4420.9940.9441.970.3915.18
    82.860.4091.1720.5050.3410.8770.5843.140.2617.81
    $\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $$\vdots $
    992.650.2230.7800.3780.3250.6320.3843.722.6416.16
    1002.530.1541.3670.4890.3660.8530.2261.490.8515.72
    下载: 导出CSV

    表  2   不同神经网络分类器识别正确率

    Table  2   Recognition accuracy of different neural network classifiers

    特征值类型BP神经网络 PSO-SVM神经网络
    正确识别
    数量n/个
    正确率
    a(%)
    正确识别
    数量n/个
    正确率
    a(%)
    时域特征值1050 1155
    频域特征值9451050
    小波包特征值13651680
    全部特征值14701995
    下载: 导出CSV
  • [1] 王斌, 李升, 朱富慧, 等. 基于超声无损检测的扩散连接界面缺陷尺寸评估[J]. 焊接学报, 2020, 41(8): 34 − 38. doi: 10.12073/j.hjxb.20200323004

    Wang Bin, Li Sheng, Zhu Fuhui, et al. Evaluation of interface defect size in diffusion bonding based on ultrasonic non-destructive testing[J]. Transactions of the China Welding Institution, 2020, 41(8): 34 − 38. doi: 10.12073/j.hjxb.20200323004

    [2]

    Abbas Moghanizadeh. Evaluation of the physical properties of spot welding using ultrasonic testing[J]. The International Journal of Advanced Manufacturing Technology, 2016, 85: 535 − 545. doi: 10.1007/s00170-015-7952-y

    [3]

    Esmaeil Mirmahdi. Numerical and experimental modeling of spot welding defects by ultrasonic testing on similar sheets and dissimilar sheets[J]. Russian Journal of Nondestructive Testing, 2020, 56: 620 − 634. doi: 10.1134/S1061830920080069

    [4] 赵大伟, 王元勋, 梁东杰, 等. 基于功率信号动态特征的钛合金电阻点焊熔核直径预测[J]. 焊接学报, 2022, 43(1): 55 − 59. doi: 10.12073/j.hjxb.20201026001

    Zhao Dawei, Wang Yuanxun, Liang Dongjie, et al. Prediction of nugget diameter in titanium alloy resistance spot welding based on dynamic characteristics of power signals[J]. Transactions of the China Welding Institution, 2022, 43(1): 55 − 59. doi: 10.12073/j.hjxb.20201026001

    [5] 杨鑫华, 贾昕, 朱平, 等. 基于信息增益率的点焊接头疲劳性能影响因素分析[J]. 焊接学报, 2020, 41(10): 73 − 78.

    Yang Xinhua, Jia Xin, Zhu Ping, et al. Analysis of factors affecting the fatigue performance of spot welded joints based on information gain rate[J]. Transactions of the China Welding Institution, 2020, 41(10): 73 − 78.

    [6]

    Elangovan S, Anand K. , Prakasan K. Parametric optimization of ultrasonic metal welding using response surface methodology and genetic algorithm[J]. The International Journal of Advanced Manufacturing Technology, 2012, 63: 561 − 572. doi: 10.1007/s00170-012-3920-y

    [7]

    Hyo-Sun Yu, Byung-Guk Ahn. A study on ultrasonic test for evaluation of spot weldability in automotive materials[J]. KSME International Journal, 13, 775-782.

    [8]

    Siljama Oskar, Koskinen Tuomas, Jessen-Juhler Oskari, et al. Automated flaw detection in multi-channel phased array ultrasonic data using machine learning[J]. Journal of Nondestructive Evaluation, 2021, 40(3): 67 − 80. doi: 10.1007/s10921-021-00796-4

    [9] 张佳莹, 丛森, 刚铁, 林尚扬. 基于频率–相位编码信号激励的焊缝超声检测分析[J]. 焊接学报, 2018, 39(7): 7-11, 41.

    Zhang Jiaying, Cong Sen, Gangtie, Lin Shangyang Analysis of Weld Ultrasonic Testing Based on Frequency Phase Encoded Signal Excitation[J]. Transactions of the China Welding Institution, 2018, 39 (7): 7-11, 41.

    [10]

    Wang Ting, Xu Chunsheng, Ma Guocheng, et al. Analysis of ultrasonic C-scan image features and image processing for spot welding[J]. China Welding, 2015, 24(2): 52 − 56.

    [11]

    Liu J, Xu G C, Gu X P, et al. Ultrasonic test of resistance spot welds based on wavelet package analysis[J]. Ultrasonics, 2014, 56: 557 − 565.

    [12]

    Zhao Yiying, Huang Qingjiu. Image enhancement of robot welding seam based on wavelet transform and contrast guidance[J]. International Journal of Innovative Computing, Information and Control, 2022, 18(1): 149 − 159.

    [13]

    Yang P, Li Q. Wavelet transform based feature extraction for ultrasonic flaw signal classification[J]. Neural Computing & Applications, 2014, 24(3-4): 817 − 826.

    [14] 吴刚, 关山月, 汪小凯, 等. 薄板点焊超声检测信号特征分析与缺陷识别[J]. 焊接学报, 2019, 40(4): 112 − 118.

    Wu Gang, Guan Shanyue, Wang Xiaokai, et al. Signal feature analysis and defect identification of ultrasonic testing for thin plate spot welding[J]. Transactions of the China Welding Institution, 2019, 40(4): 112 − 118.

    [15] 吴刚, 关山月, 汪小凯, 等. 基于超声信号增益补偿的电阻点焊熔核直径评估算法[J]. 中国测试, 2018, 44(8): 13 − 19.

    Wu Gang, Guan Shanyue, Wang Xiaokai, et al. Evaluation algorithm for nugget diameter in resistance spot welding based on ultrasonic signal gain compensation[J]. China Measurement & Test, 2018, 44(8): 13 − 19.

    [16]

    Xu Mengxi, Shi Jianqiang, Chen Wei, et al. A Band selection method for hyperspectral image based on particle swarm optimization algorithm with dynamic sub-swarms[J]. Journal of Signal Processing Systems, 2018, 90: 1269 − 1279. doi: 10.1007/s11265-018-1348-9

    [17]

    Fei Ye. Evolving the SVM model based on a hybrid method using swarm optimization techniques in combination with a genetic algorithm for medical diagnosis[J]. Multimedia Tools and Applications, 2018, 77: 3889 − 3918. doi: 10.1007/s11042-016-4233-1

  • 期刊类型引用(2)

    1. 侯金保,赵磊. SiC_f/SiC与MX246A钎焊接头界面组织及性能分析. 焊接学报. 2021(04): 74-78+100-101 . 本站查看
    2. 许欣星,董红刚,陈晶阳. 镍基高温合金用钎料研究进展. 机械制造文摘(焊接分册). 2018(02): 1-9 . 百度学术

    其他类型引用(1)

图(13)  /  表(2)
计量
  • 文章访问数:  104
  • HTML全文浏览量:  34
  • PDF下载量:  11
  • 被引次数: 3
出版历程
  • 网络出版日期:  2023-06-24

目录

    /

    返回文章
    返回