Study On Diagnosis Of Plasma Parameters Of Hall Thruster By Image Method

Hall thruster is a type of electric propulsion device which is mainly developed in the field of aerospace propulsion in our country.The plume plasma parameter diagnosis method has developed very well.As a plume image diagnosis method,the image method based on deep learning technology has the advantage of rapid calculation and reliable accuracy.Firstly,the data processing method of Hall thruster steady-state plasma diagnosis is studied by combining experiment and simulation.The structure and characteristics of the neural network used in the experiment are described,and the processing method of plume probe diagnostic data is given.Then,we explain the adjustments that need to be made when applying deep learning techniques on plasma diagnosis.A suitable plume image preprocessing method for plasma diagnosis tasks is given.The use of random cropping and random scaling increases the image data has set.The method of increasing random noise increases the probe data set.An image-plasma parameter data set for neural network training was established.Through experimental and simulation methods,various influencing factors of the fitting eff ect of plasma diagnosis model based on deep learning technology are studied.Use the built deep convolutional neural network for Hall thruster steady-state plasma diagnosis.The training speed and performance of different networks are analyzed.The changin g rule of plume cloud diagram of Hall thruster under variable working conditions is studied,and the results are basically consistent with the previous experimental results.That proves the reliability of the model.Then,through experiments,the errors caused by different shooting angles and shooting distances on the calculation results of the network model are studied.The results show that the change in shooting distance has the same effect on the electron temperature and ion density.When there is an offset angle,the ion density calculated by the model increases significantly,the deviation is the largest in the radial position of ±4mm,the deviation is small at the exit of the annular channel,and the electron temperature is different at different shooting angles The decrease is more obvious in the radial area of ±4mm,and it does not change much at the exit of the annular channel.The calculated electron temperature and ion density increase as the shooting distance increases,and decrease as the shooting distance decreases.Through experimental comparison,the variation of the radial and axial plasma parameter errors of the thruster under different interference light intensities is analyzed.It was found that the plasma parameters calculated by the model would increase with the increase of the interference light.The variation within the radial position within ±4mm is small,and the variation within the channel exit is large.The calculated value of the axial electron temperature will fluctuate,and the calculated value of the ion density will be too large.Finally,a method for studying the discharge characteristics of Hall thrusters based on images is proposed.The image method was used to analyze the change of rotating spokes in the thruster during flameout.Experiments have found that rotating spokes appear in the discharge channel of the Hall thruster during flameout,and the entire flameout process lasts about 2.5ms.When the flame is turned off,the rotation speed of the spoke is also linearly attenuated,which is consistent with the change in the discharge voltage of the thruster during the flameout.The thrust measurement of Hall thruster was carried out using image method.The results show that the use of neural network to estimate the plume thrust has certain accuracy and reliability.