Analysis And Diagnosis Of Intermittent Connection Fault In Controller Area Network

Controller Area Network(CAN),which is the fieldbus that effectively supports distributed and real-time control,has been applied widely in areas of critical safety and reliability require-ments,such as automobile communication,aerospace systems,and industrial control,etc.Phe-nomena including cable fatigue,insulation wearing and loose connections are inevitable in the process of industrial production or equipment operation,which will lead to frequent occurrence of intermittent connection(IC3)fault in the CAN network.The communication between the in-site devices and the network is interrupted by the IC fault transiently and repetitively,which not only results in the decline of system quality of service,but also brings great risks to the reliability and security of the system.Therefore,the real-time evaluation of network performance status and the on-line accurate and efficient diagnosis of IC faults in CAN networks become increasingly urgent.Based on the error response mechanism of the CAN network to the IC fault,the causality between IC fault and error response is determined and the estimation method of IC fault feature parameters is established.The state transition relationship of the CAN network caused by the IC fault is described,and the performance of CAN network is evaluated in real-time based on the network operation information and the physical meaning of each state.The data-link layer error information acquisition sensor is designed,whose measurement results form the basis of the development of the systematic IC fault diagnosis method applied to the CAN network with simple topology.In the light of the diagnosability analysis of IC faults in CAN network with complex topology,the optimal sensor deployment strategy which can accomplish the diagnosability of IC faults is designed and various types of IC faults are diagnosed efficiently and accurately.Multiple experiments have been conducted on the CAN network semi-physical experimental platform,and the effectiveness of the proposed method have been confirmed by the experimental results.This dissertation consists of six chapters and is organized as follows.In chapter 1,the research background is investigated from the aspects of CAN network and IC fault characteristics.In the introduction of the state-of-art of the fault and error response in the network,the shortcomings of existing methods in describing IC fault and its impacts are analyzed.In the introduction of the state-of-art of the network performance evaluation and analysis,the drawbacks of applying existing methods to CAN network are analyzed.In the introduction of the state-of-art of the fault diagnosability,intermittent fault diagnosis and the fault diagnosis in CAN network,the characteristics and the disadvantages of the existing methods in the diagnosis of IC faults are analyzed.The research objectives and significance of this dissertation are put forward,and the main contents as well as the internal relationship are summarized.In chapter 2,the error response mechanism of the CAN network to the IC fault and the real-time estimation of IC fault parameters are studied.Firstly,based on the analysis of the error re-sponse mechanism,the time and probability description models of error response are established.Then the causal relationship between IC fault and error response characteristics is expressed by modeling the counting process of error frame in light of stochastic theory,and the estimation of IC fault characterization parameters is completed.Finally,an experiment platform is constructed,and the effectiveness of the error response mechanism model and the IC fault parameter estimation method is verified by experiments.In chapter 3,the influence trend of IC fault on CAN network operation performance and its real-time quantitative evaluation are studied.Firstly,all possible states of CAN network when IC fault occurs are illustrated,and the deterministic and stochastic Petri net is applied to describe the transition relationship between the states.Secondly,the model parameters are calculated based on the network operation information and the estimated values of the IC fault parameters obtained in chapter 2.Then,the availability parameters of the network are defined and calculated in real-time according to the physical meaning and probability distribution of each state.Finally,the validity of the method for real-time evaluation of CAN network availability is verified by experiments.In chapter 4,the diagnosis of IC faults in the CAN network with simple topology is studied.Firstly,the sensor that collects the data-link layer error information and generates the error event when the IC fault occurs is designed,and the error event tuple mode corresponding to different types of IC fault is analyzed.Then,the systematic IC fault diagnosis method consisting of three algorithms with different theoretical framework,design and implementation principles,and ap-plicability,which are context free grammar based IC fault diagnosis algorithm,Bayesian based IC fault diagnosis algorithm,and tree based IC fault diagnosis algorithm,is proposed,and its ap-plicability is extended to the situation where IC faults coexist with global interference.Finally,the sensor is physically implemented and an experiment platform is constructed,the effective-ness of the three IC fault diagnosis algorithms for diagnosing various IC fault scenarios and their robustness to global interference are verified by experiments.In chapter 5,the diagnosis of IC faults in the CAN network with complex topology is studied based on the research results in chaper 4.Firstly,the structural characteristics of complex topol-ogy is analyzed,the fusion of distributed sensor measurement results and the representation of error event tuple mode are analyzed.Based on the analysis of the necessity of the initial sensor deployment method,the diagnosability of the network for different types of IC faults is studied.Secondly,in light of the error event tuple mode generated by the existing sensors for each node in the network,the filter search algorithm is developed to diagnose the local IC faults and derive the trunk IC faults.Then,the sub-networks that are not fully diagnosable to the trunk IC faults are determined according to the network topology and fault diagnosis results,and the optimal sensor deployment strategy is proposed to add the minimum number of sensors to the sub-networks to achieve the complete diagnosis of IC faults.Finally,the effectiveness of the filter search algo-rithm in diagnosing multiple IC fault scenarios and the validity of the optimal sensor deployment strategy which deploys the minimal number of sensors to realize the complete diagnosability of IC faults in the CAN network are verified by experiments.In chapter 6,the research work of this dissertation is summarized,the innovation points are highlighted,and the further research work is discussed.