关键词: 液体推进剂火箭发动机;故障检测;人工神经元网络;实时算法

Abstract: he starting process and cutting off processe of liquid propellant rocket engines are inherent nonlinear stochastic ones. It is a very difficult task to carry out detecting faults in such processes. In this paper, based on system identification theory, Authors employ artificial neural networks technique to complete the nonlinear system identification for the starting process of the engine with trubopump system. The fault detection method based on checking identification error is proposed and implemented. The results of detecting faults, which are obtained from testing with a number of practical engine’s fire-test data, show that the fault detection algorithm proposed in the paper is very effective. Because the algorithm needs less computation cost, and the measured parameters that are required for the fault detection algorithm are consistent with ones for the current monitoring system on the ground test, the algorithm studied and implemented may directly be applied into real monitoring system of liquid propellant rocket engines.

Key words: Liquid propellant rocket engine;Fault detection;Artificial neural network;Real time algorithm