In-vehicle network level fault diagnostics using fuzzy inference systems

This paper presents an application of an Adaptive-Network-based Fuzzy Inference System (ANFIS) for pre-diagnosing incipient underlying in-vehicle network problems which possibly could cause further failures. An experiment on ANFIS-based pre-diagnosis of network level faults on Controller Area Network (CAN) by utilising available network protocol signals, such as error frames, is reported. The experimental results show that the pre-diagnostic system can efficiently classify causes of error frames transmitted on a CAN bus, and identify “network health” which indicates healthiness of the network when being used for message communication. The potential causes of the faults can be narrowed down, and further network diagnostics and prognostics can be performed.