CTCS-3 Communication Failure Diagnosis Expert System

ABSTRACT:Recently, the construction of high-speed railway is booming in China. With the rapid increase of train speed, the security requirements of train control system are also becoming more and more demanding. Being one key technology in the Chinese Train Control System-level 3 (CTCS-3), the reliability of the GSM-R system directly affects the normal CTCS-3 operation. However, due to the complexity of current communication network, the actual communication failure in CTCS-3 is obviously uncertain and unexpected. Therefore, how to locate C3 communication failures quickly and accurately is a crucial index which can be measured by the reliability and maintainability of train control system. Currently, communication failures are mainly diagnosed by experts who have rich experiences and related data, making the diagnosis process inefficient, long time consuming and unable to pass down to the successors.This thesis combines the artificial intelligence and diagnosis experiences together, designs a communication fault diagnosis expert system based on signaling data. Firstly, the article clarifies the definition of communication failure, and then proposes a fault diagnosis method based on signaling data. Through representative failure cases in actual engineering tests, the article verifies the feasibility of this method. Then the fault diagnosis process based on long-time field experiences is summarized as the experimental basis of failure diagnosis expert system. As a result, this paper establishes a fault diagnosis model and the relate criterion of communication failure based on interface signaling messages. A method is proposed to abstract sampling data from original failure data with designated attributes. This paper also introduces a class selector as a novel way in generating knowledge database, and chooses ID3 algorithm of decision tree as a core construct method in class selector through the comparison of three intelligent algorithms. At last, the gross structure and function model of expert system can be designed through the stages mentioned above.