基于数据可视化的复杂系统信号时序识别方法

针对复杂系统研发及运行过程中产生的大量信号可以表征系统运行的时序健康状态这一特性,提出了一种基于数据可视化及卷积神经网络(convolutional neural networks, CNN)智能识别的时序特征识别方法;该方法使用数据可视化技术将信号的时序特征映射至图像,通过训练好的特征识别模型对信号可视化图像进行时序特征的识别,可实现系统运行时的实时智能状态监测;选取了三种典型信号的正常及异常特征,通过模型构建及测试分析,验证该方法对复杂系统信号的时序特征有良好的识别效果,可应用于对时序要求较高的复杂系统进行状态监测及故障诊断。

Time Series Feature Recognition of Complex System Signalsbased on Data Visualization

In view of the fact that a large number of signals generated in the process of complex system development and operation can represent the time sequence health state of system operation, a time sequence feature recognition method based on data visualization and convolutional neural networks (CNN) intelligent recognition is proposed. This method uses data visualization technology to map the time sequence features of the signal to the image, and uses the trained feature recognition model to recognize the time sequence features of the signal visualization image, which can realize the real-time intelligent state monitoring of the system. The normal and abnormal characteristics of three typical signals are selected. Through model construction and test analysis, it is verified that the method has good recognition effect on the time sequence characteristics of complex system signals, and can be applied to condition monitoring and fault diagnosis of complex systems with high time sequence requirements.