BR-FRL: A Belief Rule-Based Fault Recognition and Location Model for Bus Network Systems

Abstract

For a control system with bus network topology, it is significant while difficult to accurately recognize and locate the fault of transmission medium. The reason is that the occurrence of one fault can lead to the abnormality of multiple terminal nodes. How to use these abnormal signals to realize effective fault recognition and fault location is the key to solving the problem. Inspired by the manual inference in engineering practice, this paper proposes a belief rule-based fault recognition and location model (BR-FRL) for transmission medium in bus network systems. The core idea is fault location based on the results of fault recognition. To adapt to the flexible expansion of bus topology, the designed BR-FRL model has self-organization characteristics. The highlights of this paper include three aspects: 1) The progressive reasoning of fault recognition and location based on belief rules is realized in one model; 2) A new model architecture is proposed, which integrates the belief rules, the group method of data handling, and the cautious conjunctive rule (CCR); 3) The parameter learning based on stochastic gradient descent is applied to BR-FRL, and the sensitivity analysis of evidence weight in CCR of BR-FRL is performed. The effectiveness of BR-FRL is verified by an experiment of gigabit-capable passive optical network.