Research On Measurement Information Fusion Of Multi-Electrode Sensors Based On Complex Network Theory

Complex system consists of lots of nonlinearly interactive elements.Many complex systems usually can't be analyzed in terms of mathematical models,therefore how to characterize such complex systems represents a significant challenge of great importance.Time series analysis brings light to such challenging problem in that the complex dynamical behaviors can be characterized from the observation time series.However,when system complexity increases,the observation data increase from univariate time series to multivariate time series.Consequently,how to realize multi-information fusion remains to be solved.Recently,complex network analysis of time series has attracted more and more attention,which allows investigating complex systems from multivariate time series.This novel methodology provides a new and effective way to deal with the above challenging problems.Multiphase flow is a typical complex system in real life.Especially the oil-in-water two-phase flow is a common occurrence in oil industry.However,the complicated flow behaviors underlying flow transitions are still elusive,which can't be satisfactory solved by mathematical model.Characterizing the flow behaviors from experimental measurements becomes quite important and necessary.We conduct the vertical oil-water two-phase flow experiment to obtain the multivariate time series from high-speed cycle motivation conductance sensor for different flow patterns.Then we use multivariate time-frequency analysis to probe the typical features of three flow patterns from the perspective of energy and frequency.Moreover,we infer complex networks from multi-channel measurements in terms of phase lag index,aiming to uncovering the phase dynamics governing the transition and evolution of different oil-in-water flow patterns.The results indicate that our approach yields quantitative insights into the flow behaviors and dynamical mechanism governing the transition and evolution of different oil-in-water flow patterns.Furthermore,in order to characterize the dynamical flow behaviors of oil-in-water bubbly flows,this dissertation proposes a novel methodology for inferring multilayer network from multivariate time series.Particularly we employ macro-scale,meso-scale and micro-scale network measures to characterize the generated multilayer networks,and the results suggest that the macro-scale network measure allows distinguishing different flow states,the meso-scale and micro-scale network measures enable to uncover the local flow structures associated with configuration of oil droplets in the transition from oil-in-water bubble flow to VFD flow(very fine dispersed oil-in-water bubble flow).Brain is one of the most complicated systems,and mathematical model usually is unable to reveal its complicated behaviors.In recent years,traffic accidents frequently happen everyday due to fatigue driving,however the mechanism underlying brain fatigue is still elusive.Therefore,acquiring the EEG signals during fatigue driving and then detecting and characterizing brain fatigue state become an important research topic.This dissertation designs an experiment to obtain the 30-channel EEG signals,and then infer the brain functional network in terms of Spearman rank correlation coefficient.The results indicate that this novel method allows identifying brain states of normal and fatigue during human driving.In addition,this dissertation analyzes the brain networks in four frequency bands and find that the brain networks inferred from normal and fatigue states present different topological characteristic,and the difference is more obvious in the Theta frequency band.