Deep neural network-based fault diagnosis approach for the rocket propelling nozzle in the glide

This paper presents an intelligent approach for fault diagnosis about rocket nozzle in glide phase. In contrast to the commonly used observer or any other technologies over the complicated rocket, the proposed approach is based on a deep neural network for processing fault problems without rocket mathematical model accurately. The intelligent method, based on a Dynamic neural network with cross-entropy lose function, automatically identifies the faults of rocket nozzle, effectively acting as a fault classifier. This approach provides a simple solution for modeling difficult problems and allows multi-classes faults to be recognized in a straightforward way.