Fault detection and diagnosis for railway track circuits using neuro-fuzzy systems

Railways are expected to operate with ever increasing levels of availability, reliability, safety and security. One way of ensuring high levels of dependability is through the use of condition monitoring systems. This paper presents the results of research on fault detection and diagnosis methods for railway track circuits. The proposed method uses a hybrid quantitative/qualitative technique known as a neuro-fuzzy system. Such a hybrid fault detection and diagnosis system combines the benefits of both fuzzy logic and neural networks, i.e. the ability to deal with system imprecision and to learn by neural network training processes. It is shown that the proposed method correctly detects and diagnoses the most commonly occurring track circuit failures in a laboratory test rig of one type of audio frequency jointless track circuit.