系统仿真学报 ›› 2016, Vol. 28 ›› Issue (8): 1863-1869.

• 仿真系统与技术 • 上一篇    下一篇

基于奇异谱分析—SVM的木构件缺陷识别

周国雄1, 陈爱斌1, 周先雁2   

  1. 1.中南林业科技大学计算机信息工程与学院,长沙,410004;
    2.中南林业科技大学土木工程与力学学院,长沙,410004
  • 收稿日期:2015-01-21 修回日期:2015-08-27 出版日期:2016-08-08 发布日期:2020-08-17
  • 作者简介:周国雄(1980-),男,湖南嘉禾,博士后,副教授,硕导,研究方向为智能控制、无损检测;陈爱斌(通信作者 1971-),男,湖南攸县,博士,教授,研究方向为图形图像处理。
  • 基金资助:
    国家948项目(2014-4-09),国家林业公益性行业科研专项(201304504-3)

Wood Structure Nondestructive Detection Based on Singular Spectrum Analysis and SVM

Zhou Guoxiong1, Chen Aibin1, Zhou Xianyan2   

  1. 1. School of Electricity & information Engineering, Central South University of Forestry & Technology, Changsha 410004, China;
    2. College of Civil Engineering and Mechanics, Central South University of Forestry and Technology, Changsha 410004, China
  • Received:2015-01-21 Revised:2015-08-27 Online:2016-08-08 Published:2020-08-17

摘要: 针对木构件缺陷的未知性,提出一种基于奇异谱分析—SVM的木构件缺陷识别方法,采用超声波测试仪对木材试件进行测试,获取测试信号,为消除探伤时由于测试仪增益调节及缺陷尺寸、角度的变化对测试缺陷回波波高的影响,采用奇异谱分析,过滤异常随机波动,并从中提取出表征原始信号的特征参数,采用改进的SVM算法对特征参数进行网络训练,识别木构件缺陷类型。测试结果表明该方法区分标准试样和胶缝试件的准确率为97.5%,在识别死节试件时也达到了95%,具有较高的准确率。

关键词: 缺陷识别, 奇异谱分析, SVM, PSO

Abstract: In view of unknown defect of wood component defect, a method of wood structure nondestructive recognition based on Singular spectrum analysis and SVM was proposed. The wood specimen was tested to obtain the test signal by ultrasonic testing instrument. In order to eliminate the testing effect of the tester gain control and defect size, angle variation on the test defect echo amplitude, the abnormal fluctuation was to filter and characteristic of the original signal was extracted by singular spectrum analysis, and SVM could train the parameters and distinguish the wood defects. Simulation results show that the proposed method can distinguish standard samples and glue joint with accuracy of specimen for 97.5%, and recognition of knots specimens also reaches 95% with high accuracy.

Key words: defect recognition, singular spectrum analysis, SVM, PSO

中图分类号: