Root Cause Analysis For Intermittent Connection Faults In ControllerArea Networks

In networked manufacturing system, communication plays a magnificent role in information interaction. However, duing to data transmission and data association, network fault diagnosis faces difficulties in guaranteeing efficiency and accuracy. It is known that CAN (controller area networks) access large numbers of nodes with complicated network structure. In some cases, unit fault in the local system will affect other systems directly or indirectly, leading to more complicated network faults. Therefore, analysis of CAN faults has much fundamental magnificence.In this paper, cause analysis is firstly applied to analysis of intermittent connection faults in CAN, based on quasic-periodic event and Bayesian networks model. According to the common feature of connection faults, the method of root analysis is proposed and related experiments validated it through the semi-physical simulation test platform.The method of root analysis can be divided into qualitative and quantitative analysis. Firstly, considering the upper layer system status and bottom layer error frame, qualitative causal analysis is put forward based on the quasic-periodic model, by which all thestate events and error frame events those are correlated can be found.Then further quantitative analysis is conducted based on causal Bayesian network model, which assumes causal Markov hypothesis. Given evidence,we can get posterior probability of interested nodes.At last, connection faults are injected and systematic states are simulated through the semi-physical simulation test platform. Therefore, the experimental results further validate the effectiveness of root causal analysis for connection faults in CAN.