Research On Multiple Memristor Composite Circuits With Fault Diagnosis

Due to the distinctive characteristics of memristive devices,such as nanoscale dimensions,nonvolatility,variable conductivity and so on,research interests on memristors are gradually increasing.Up to now,a variety of single or multiple memristor structures have been integrated with advanced CMOS technology,which promises efficient implementations of synaptic connections in neuromorphic computing systems,or computing elements in signal processing systems.Moreover,memristors have the application potential in many fields,especially for memristor combination circuits.Since the memristance variation depends on individual memristor polarity and initial state,the behavior of multiple memristor circuit becomes complicated and is difficult to predict.So in order to better understand the behaviors and characteristics of memristor composite circuits,it is great important to investigate different memristor connections.However,geometry variations of the memristor have a significant impact on its memristance and electrical characteristics.Therefore,in order to improve the accuracy and effectiveness of memristors in various applications,the fault diagnosis of memristor circuits should be considered.In this dissertation,the electrical characteristics of multiple memristor composite circuits are studied,and several synaptic circuits based on memristors are also discussed.In addition,combining sensitivity analysis with fuzzy analysis,the faults of memristor composite circuits and memristor-based synaptic structures are diagnosed.Specifically,the main contents of this dissertation can be divided into four parts,as follows:(1)First of all,from the perspective of the physical analysis,the physical mechanisms of three kinds of memristor models are introduced,including TiO2-based memristor,WOx-based memristor and Ag2S-based memristor.Then,for mathematical modeling,two kinds of mathematical modeling methods of memristor are also introduced,namely the conduction channel length or area as state variable.Furthermore,memristor models based on the mathematical modeling methods are simulated based on Matlab software.(2)The composite characteristics of multiple memristor composite circuits are investigated,including series,parallel and hybrid structures.To be specific,transient behaviors,steady behaviors and specific conditions to reach steady state of memristor circuits are deeply analyzed.Then three kinds of different memristor connections are analyzed in detail,and the relationships between the flux and the charge are obtained by mathematical derivation.Moreover,the relationships between the memristance and the flux or charge are obtained.On this basis,the corresponding conclusions of specific structures are extended to structures with n memristors.Meanwhile,corresponding circuits are simulated to demonstrate the validity of relevant theories.(3)Based on the theoretical derivation of memristor composite circuits,several memristor-based synapses are analyzed and compared.The relationships between memristance and input stimulus of these synaptic structures are deduced.Then it is analyzed whether the realized synaptic weights are linearly updated,which is verified by simulations.Furthermore,the needed time of the memristance transformation is presented.Additionally,the impact of thickness variation on memristive characteristics is analyzed,and then its impact on memristor synapse is studied.(4)A special feedback-control doublet generator is designed,which can control the switching pulses based on the output feedback,and automatically cut off the input of the excitation.Furthermore,fault diagnosis hypotheses are presented,and based on the fuzzy concept,fuzzy expressions of faulty diagnosis are deduced.Moreover,using the method of combining sensitivity analysis with fuzzy analysis,the fault of memristor composite circuits can be diagnosed.Finally,two concrete examples are demonstrated to verify the effectiveness of the fault diagnosis method in memristor circuits.