欢迎访问林业科学,今天是

林业科学 ›› 2010, Vol. 46 ›› Issue (6): 176-181.doi: 10.11707/j.1001-7488.20100628

• 论文 • 上一篇    下一篇

基于模态分析和BP神经网络的红松方材孔洞定量检测

王立海,徐华东,邢涛,倪松远   

  1. 东北林业大学工程技术学院 哈尔滨 150040
  • 收稿日期:2008-07-15 修回日期:2009-05-07 出版日期:2010-06-25 发布日期:2010-06-25

Quantitatively Determining of Hole-Defects in Korean Pine Lumber Based on Modal Analysis and BP Neural Network

Wang Lihai;Xu Huadong;Xing Tao;Ni Songyuan   

  1. <i>College of Engineering & Technology, Northeast Forestry University Harbin</i>150040
  • Received:2008-07-15 Revised:2009-05-07 Online:2010-06-25 Published:2010-06-25

关键词: 模态分析, 固有频率, BP神经网络, 孔洞缺陷, 定量检测

Abstract:

Inner defects of timber caused a great loss of wood resources, especially valuable wood resources. Therefore, it was great significant to quantitatively detect wood defects for high-efficient utilization of wood resources. In this paper, the modal experiment was carried out, using AD-3651-02 FFT analyzer and other necessary instruments, under the normal circumstances in the laboratory. And the first three orders of intrinsic frequency of 15 Korean Pine specimens with different positions or diameters of hole-defects were obtained. After that, two parameters <i>ζ</i> <sub>1</sub>and <i>ζ</i> <sub>2</sub> were constructed with the aid of intrinsic frequency. Here parameter <i>ζ</i> <sub>1</sub> was sensitive to hole-defect's position and parameter <i>ζ</i> <sub>2</sub> was sensitive to its size. And then <i>ζ</i> <sub>1</sub> and <i>ζ</i> <sub>2</sub> were introduced as input vectors of BP neural network, so the location network and the quantitative recognition network were constructed to realize training and testing of specimens data. The results showed that whether Korean Pine specimens with hole-defects or not could be diagnosed directly by variation of intrinsic frequency. At the same time, the positions and sizes of hole-defects could be recognized effectively by using the location network and the quantitative recognition network.

Key words: modal analysis, intrinsic frequency, BP neural network, holedefects, quantitatively determining