地震检测波的分形神经网络模式识别

基于分形理论提出了尺度窗分配维的概念,与单一分形维相比,尺度窗分配维集合能更好地对自然界中的无规分形进行全面、细致的描述,形成其分形特征向量。将分形特征向量作为输入矢量,通过标准样本集的学习,建立基于分形分析的复杂非线性系统 RBF 神经网络辨识模型。以地震检测波为分析对象,尝试将该模型应用于混凝土地下连续墙的区域物性预测,实例证明它的可行性。该方法将分形理论与神经网络的优势联合,提出了一条非线性系统检测识别的新技术路线,在无损检测等多个领域具有应用价值。

PATTERN IDENTIFICATION OF SEISMIC DETECTING WAVES BY FRACTAL NERUAL NETWORK

Scale window divider dimension is proposed based on fraetal theory.The set made up of the dimensions is better than single fractal dimension at depicting the irregular fractal and becomes the fractal emblematic vector.The RBF neural network model to identify different complex systems is created taking the set as imputed vector and by studying of standard examples.The reasonable result obtained by applying the model in the investigation of concrete diaphragm wall shows the feasibility to identify complex systems.By combining the advantages of fractal theory and neural network,the technology illuminated in this paper becomes a novel way to deal with the detection and identification tasks of nonlinear systems and is of practical value in nondestructive field etc.