Fault detection in intelligent early failure warning sensors system for train rotary door operator

Summary form only given. The scope of this paper relates to sensor fault detection in intelligent early failure warning sensors system for safety critical single throw mechanical equipment, using Vapor´s train rotary door operator (TRDO) as a first case study. Work carried out at The University of Birmingham includes design and development of state-of-the-art control and data acquisition system for Vapor´s TRDO using Lab Windows/CVI. Fault diagnosis of engineering systems involves identification of components that cause variations from usual healthy behaviour. When observed abnormal behaviour crosses certain threshold a warning is initiated for failure. Neural networks are being used for system identification and modelling of healthy behaviour of equipment using various sensors. Failures are predicted from deviation of any sensor´s data from its healthy model. Faulty component is identified after analysis of information from all sensors and using database available from fault modes and effects criticality analysis (FMECA) of the TRDO. Sensor failure detection and sensor value validation has also been investigated. Different types of sensors are employed for monitoring various parameters used in electro-pneumatic operation of train door. These parameters include air flow, air pressure, angular displacement, lateral displacement, opening and closing time. These parameters relate to one another and this property is exploited to detect sensor failures. Detection of sensor failures using analytical redundancy is also discussed