Fault Localization Of Intermittent Connections For CAN Networks

Supported by National Natural Science Foundation of China, this dissertation presents a data driven approach for fault localization with clear knowledge of CAN network topology, electrical specification, medium access and fault confinement mechanism and the features of intermittent connection (IC). As a common cable transient connection fault, the IC problem is a systemwise phenomenon, not only the node that has the IC problem can be shut down, other nodes can be shut off as well. By interrupting the network communication, the IC problem can deteriorate the performance of the network, leading to system-level failure in severe cases. Hence, it is desirable to detect and locate the IC problems before further damages occur. By passively capturing IC-induced network errors, a triple is used to describe each discrete error event. Based on statistical significance, the discrete error events are mapped to alarms for different nodes. There is causal association between the patterns of alarms with IC fault locations. By proposing two algorithms which can solve the alarm correlation and fault identification problem, the fault localization is figured out by searching the optimal solution with the minimized information cost.The dissertation is organized as follows.In chapter 1, the research background of fault diagnosis and localizing is described in detail, especially emphasizing the existing research techniques for this field. This chapter concludes with a summary of research contents and significance.In chapter 2, after defining the specific meaning of intermittent connection clearly, the re-search objectives are proposed with the particular denotes. Meanwhile, the whole research method and corresponding assumptions and constraint conditions are briefly summarized.In chapter 3, by analyzing how the IC faults interrupt the network communication, a test-bed is designed to monitor the status of CAN bus and capture the physical signal the moment an error frame occurs. By decoding the physical signal according to CAN network electrical specification, useful information can be interpreted. Combined node ID and pattern recognizing, a triple is used to describe each discrete error event induced by IC, including the interrupted packet, the monitoring position for the interested error event, the logical feature of the decoded physical signal. This chapter is aimed at acquiring fundamental data for the development of research method, all important parameters have been received through digital logical signal, which is creative for data driven method, and is more reliable for our goals.In chapter 4, based on statistical significance, the discrete error events are mapped to alarms for different nodes. There is causal association between the patterns of alarms with IC fault lo-cations. A single fault may result in a large number of alarms, and it is often difficult to isolate the true cause of the fault. By proposing PIA algorithm which can solve the alarm correlation and fault identification problem, the fault localization is figured out by searching the optimal solution with the minimized information cost. The effectiveness of proposed algorithm has been verified using different experiment cases.In chapter 5, by analyzing the limitations of PIA algorithm, the optimized algorithm PNIA is proposed. The effectiveness has been verified using developing experiments when the CAN network suffers interference resources other than IC faults.In chapter 6, the research work is summarized, as well as its scientific contributions and innovations. Future work is also presented.