Investigation On Fault Feature Analysis And Fault Prediction Methods For LOX/Methane Rocket Engines

Aiming at liquid propellant rocket engine,the research on fault characteristics analysis and fault prediction method is carried out,which is of great significance to diagnose rapidly and detect fault as soon as possible.The paper took a LOX/Methane engine as the research object.A fault simulation platform was set up,and the typical fault characteristics are simulated and analyzed.The selection of engine monitoring parameters and sensor data validation were studied.The prediction method based on ARMA and support vector machine was used to predict the fault of liquid rocket engine.Firstly,the engine structure was decomposed hierarchically,using modular modeling ideas and lumped parameter method modeling,and a fault simulation model of liquid oxygen methane engine was established based on Simulink in MATLAB.Typical faults such as pump leakage,cavitation,pipeline blockage leakage,thrust chamber ablation,etc.were simulated,and fault characteristics were analyzed and extracted,and its fault mode library was built.Then combined with qualitative analysis and quantitative analysis,the monitoring parameters of key components of engine were selected,and the data validation of engine sensor was studied based on principal component analysis method.The results show that in addition to slow drift faults,timely and effective detection can be achieved for many other types of sensor faults.Finally,the prediction method based on ARMA and support vector machine was used to study the fault prediction of liquid rocket engine.The research shows that the prediction method based on ARMA has better short-term prediction on steady state data of the engine,but the calculation efficiency of real-time model updating is low.The calculation efficiency of regression support vector machine based on least squares is high,and it can be used for engine steady state fault prediction and startup state parameter prediction.