Model-Based Detection and Monitoring of the Intermittent Connections for CAN Networks

Monitoring the reliability of an industrial network has become more and more important due to the complexity of the networked systems and the costs associated with the system downtimes. Among the numerous factors that affect the reliability of networks, intermittent connection (IC), a situation where communication lines suffer from short random disconnections because of various reasons, is a frequent and hard troubleshooting problem. It leads to both quality drawback issues and threat to personal safety. Hence, early detection of ICs is a necessary and challenging task for maintenance of the networked systems. In this paper, a novel model-based IC detection method is proposed based on a generalized zero-inflated Poisson model. A maximum-likelihood-estimation-based model parameter estimation method and optimization of the ranked probability control chart are applied for monitoring the IC events. Characteristics including minimum sampling size and confidence interval of estimated parameters are also studied for model interpretation and IC error detection. A test bed is constructed, and a computer-controlled emulation system is developed. A case study is conducted to demonstrate the modeling and control chart design procedure, as well as the effectiveness of the proposed methodology on IC error detection and monitoring.