基于小波包和RBF神经网络的木材空洞诊断研究

Empty Diagnosis for Wood Based on Wavelet Packet and RBF Neural Network

采用小波包与径向基神经网络(RBF)松散结合的方法,对健康和空洞蒙古栎试件进行了研究,利用小波包变换对应力波检测信号进行5层小波包分析,构造8维特征向量,然后利用特征向量训练径向基神经网络和建立诊断模型。结果表明:所建模型的辨识正确率达到90.80%,能有效的评估木材的性质,为应力波无损检测仪器的设计提供了参考依据。

The health Mongolian Oak specimens and the Mongolian Oak specimens with empty were studied by the method of wavelet packet and RBF neural network loosely bound. The wavelet packet transform was used to ana-lyze the stress wave detection signal to carry on the 5-layer wavelet packet analysis, and the 8 dimensional feature vector was constructed. Then the feature vector was used to train the RBF neural network and establish the diagnosis model. The experimental result showed that the identification accuracy rate of the model was up to 90. 80%, which could effectively evaluate the nature of wood, and a theoretical reference for the design of stress wave nondestructive testing instrument was provided.