Aircraft fault diagnosis and decision system based on improved artificial neural networks

The goal of this work is to build an aircraft fault diagnosis and decision system, which uses data-driven methods to automatically detect and isolate faults in the aircraft, while keeping its reliability and safety. As a fundamental specification, this fault diagnosis system should not be a black box, the condition monitoring and the results of comprehensive diagnosis shall be illuminated to engineering consulting services, and it can help engineers to accumulate the knowledge for reengineering purposes (including diagnosis operational rules) and improve the design of new aircraft. In comparison with some methods, Artificial Neural Networks (ANN) has been shown to be more advantageous and is currently used in fault diagnosis system. For example, it hasn't any problem of conflict of new rules, which is a big problem in Expert System while adding new fault. In this work, ANN is improved. Its speed of learning and the iteration times can be self-corrected or mutated. Moreover, neural network can be combined with other optimization methods, like genetic methods, to achieve a better performance. Furthermore, according to the different types of sensors, certain sub-networks are built to assist the principal network or replace it in some anomaly condition. A decision system treats the results of all the networks and comes to a conclusion, which will be sent to pilot, airport command center (ACC), or fault tolerant system.