论文部分内容阅读
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。
Aiming at the unknown of wood component defects, a method of wood component defect identification based on singularity spectrum analysis (SVM) is proposed. The wood samples are tested by ultrasonic tester and the test signal is obtained. In order to eliminate the flaw of the tester, Size and angle of the test defect echo wave height, using singular spectrum analysis, filtering abnormal random fluctuations, and extracted from the original signal characterization parameters extracted, the use of improved SVM algorithm for network training of characteristic parameters to identify wood components Defect type. The test results show that the accuracy of the method is 97.5% for distinguishing standard specimens and seams, and 95% for recognizing dead-end specimens with high accuracy.