The Analysis Of Multiphase Flow And EEG Signals Based On Complex Network-based Multi-information Fusion Theory

Complex systems widely exist in our life.Mathematical analysis usually cannot reveal their dynamical mechanism.Complex network is an abstract description of complex systems,and complex network-based multi-information fusion theory has become a hot topic in the research of nonlinear complex systems.The multiphase flow system is a typical nonlinear complex system,and how to reveal the dynamical mechanism from flow signals remains to be solved.Besides,brain is one of the most complex systems and characterizing the brain behavior from EEG signals has become a cuttingedge research area.How to extract intrinsic features of EEG signals and realize the fusion of information in different brain regions is of great significance for the research of brain behavior.To deal with the challenges,we in this dissertation propose some novel methods based on complex network-based multi-information fusion for the analysis of multiphase flow and EEG signals.The main work are as follows:1)We propose a novel wavelet multiresolution complex network(WMCN)for the analysis of multivariate nonlinear time series from the oil-water two-phase flow experiment.First,we decompose the measured signals through the wavelet decomposition technology,and then infer a weighted network in terms of the distance between feature vectors.Next,we calculate the weighted average clustering coefficient and the weighted average shortest path length to characterize the topology of the derived networks.The results suggest that WMCN method allows characterizing the chaotic and nonlinear flow behaviors underlying the transitions of different oil-water flows at different resolutions.2)We propose a novel wavelet relative entropy complex network(RWECN)to analyze the EEG data collected under simulated driving experiments at different states(alert and fatigue).First,we infer a directed weighted networks in terms of the wavelet relative entropy between two channels for obtaining the intrinsic and effective features related to the characteristics of brain networks.Then,we calculate the RWECN statistical measures and combine them with the wavelet entropy to form a feature vector,aiming to realize the classification of different states through the Fisher linear discriminant analysis.The results suggest that RWECN method enables to improve the classification accuracy of EEG-based fatigue driving.3)We propose a novel transfer entropy complex network(TECN)to characterize the different brain cognition states from EEG signals.First,we carry out an experiment to obtain the EEG data of the subjects under different mental arithmetic tasks.Next,we infer the directed weighted networks in terms of the transfer entropy between different channels.Then,we calculate the weighted average clustering coefficient to characterize the nonlinear dynamical behavior from EEG signals underlying the derived brain networks.The results suggest that TECN method allows revealing the brain cognitive behavior and information transition between different brain regions.