深度卷积神经网络在混沌振动识别中的应用研究

Application of deep convolution neural network in chaotic vibration identification

针对混沌振动信号识别中,混沌特征指数计算量大、运算耗时长,难以满足实时性的要求,提出一种基于深度卷积神经网络的智能混沌识别方法。首先通过相空间重构技术,得到不同振动信号的吸引子图;在此基础上,优化设计了经典网络模型AlexNet的结构参数并进行训练;最后将改进后的模型用于混沌信号的智能识别。仿真和实测信号的结果表明,该方法是可行的,为混沌在线识别提供了有益参考。

In recognition of chaotic vibration signals,the calculation amount of chaotic characteristic index is large and time-consuming,so it is difficult to meet the real-time requirements.Here,an intelligent chaotic recognition method based on deep convolution neural network was proposed.Firstly,attractor graphs of different vibration signals were obtained using the phase space reconstruction technique.Then,structural parameters of the classic network model AlexNet were optimized and trained.Finally,the improved model was applied in intelligent recognition of chaotic signals.The results of simulated and actually measured signals showed that the proposed method is feasible,it can provide a useful reference for on-line chaos recognition.