Neural Networks Based Real-Time Fault Detection for A Liquid Rocket Propulsion System

摘要

The real-time fault detection is very crucial in developing the online health monitoring techniques for rocket propulsion systems, particularly when the manned space missions are accompanied. The neural networks based approach provides an alternative solution to the design of model-based fault detection methods for detecting the potential failures of propulsion systems. In this paper, a general framework for neural networks based fault detection is developed for a class of liquid rocket propulsion systems. The design approach consists of system modeling, residual generation, and fault detection. First, feed-forward neural networks are used to model the complicated dynamics of propulsion system for simplifying the modeling process and improving the real-time performance of model-based fault detection. Second, a real-time fault detection architecture using the established neural networks approximator is designed. By using the real measurements from ground firing test, an example is provided for demonstrating the effectiveness of the proposed approach to the real-time fault detection of a liquid rocket propulsion system.