Research On Fault Detection Of Rocket Pressurization Feed Systems Components And Fluid Velocity In Pipeline

Pressurization feed system is the core system of rocket thrust,and its reliability is directly related to the safety of launch vehicle.At present,the mainstream fault detection algorithms of rocket include signal analysis,mathematical model and neural network.Combined with the advantages of the above methods,the thesis uses the telemetry data of the core components of a certain type of liquid rocket Pressurization feed system to monitor the operation of each component of the Pressurization feed system.The Pressurization feed system is composed of the pressurization component and the feed pipe.Taking the third stage of a launch vehicle as the fault detection object,the telemetry data returned by the sensor is sampled and preprocessed to reduce the data scale and improve the efficiency of fault detection.Then,the common faults and occasional faults are simulated to expand the fault data scale,so that the fault data and normal data are balanced.Finally,the normalized time series is added to distinguish the rocket operation conditions.The processed telemetry data are used to train BP neural network.The feed system mainly includes propellant pipeline and pressurization gas pipeline of different types and different pipe diameter,which are used to transport propellant,coolant and pressurization gas.The normal operation of pipeline flow is the guarantee of rocket launch successfully,and the velocity of pipeline flow is the most important index.In order to monitor the flow velocity of different pipe diameters,the high frequency ultrasonic is used to flow through the pipeline in the direction of the downstream and the reverse flow.The ultrasonic time delay is calculated by using the correlation method of ultrasonic signal.Then the flow velocity can be obtained by simple calculation.For the difference between the downstream flow and the reverse flow time of the pipeline with high velocity and small diameter,that is,the time delay is in the order of μs to ns,which requires ultra-high frequency ultrasonic to measure.The general transducer can not generate such high frequency ultrasonic signal.Therefore,the cost of velocity monitoring can be reduced by interpolation and then the correlation method is used to calculate the delay.The simulation results of COMSOL multiphysics show that the ultrasonic correlation algorithm is suitable for the velocity monitoring of all sizes of pipelines,and can meet the needs of real-time pipeline velocity monitoring.The experimental results show that BP neural network and ultrasonic correlation algorithm can effectively detect the fault and flow rate of the pressurization feed system.The false alarm rate of normal data and the failure data miss rate of the pressurization system are 2.5% and 3.6%respectively.The results of fault detection can be returned every 0.24 s,which can meet the realtime detection requirements.The propellant pipeline and the pressurization gas pipeline flow of the conveying system can meet the real-time detection requirements The velocity monitoring error is less than 4%,which meets the requirement of fault velocity detection of pipeline.