| As an important part of the infrared camera load and low temperature optical system,the satellite-borne Stirling pulse tube cryocooler provides a low temperature cold source for the infrared focal plane devices to improve the imaging quality of the infrared camera.Since the 12 th Five Year Plan,the rapid development of infrared focal plane remote sensing technology in China has promoted the level of satellite-borne pulse tube cryocooler with independent intellectual property rights.With the development of infrared remote sensing load from pre-research to business application,it also puts forward higher requirements for the performance consistency of satellite-borne pulse tube cryocooler products.It is necessary to get rid of the original R&D production mode and establish shelf type aerospace product system with long life,high reliability,high consistency and efficient production,so as to meet the increasingly heavy demand of aerospace research and development tasks.In order to achieve spectrum products,the performance of satellite-borne pulse tube cryocooler must have good consistency.Therefore,the task and goal of this project is to determine the main factors affecting the performance of the cryocooler and reduce the performance fluctuation of the cryocooler,that is,to improve the performance consistency of the cryocooler,which is also a complicated and difficult systematic project and require a large amount of resources and intelligence.This paper takes a model of 12W@85K coaxial pulse tube cryocooler as the main research object,a 6W@95K linear pulse tube cryocooler was designed for auxiliary analysis,and specially designed a detachable compressor and cold finger to carry out the special experiment of variable paremeters.Firstly,numerical simulation analysis and experimental research are carried out on the influence of various factors on the performance of the cryocooler within the process control range(the variation range of each influencing factor is small).Then,the performance prediction model of the cryocooler is obtained by Logistic ordinal regression,Random Forest and XGBoost regression based on machine learning language method,and the feature importance of each factor affecting the performance fluctuation is obtained.Finally,the regression model results and the experimental results were compared and analyzed.The main research work is as follows:1)Study on the thermodynamics and dynamics of pulse tube cryocooler(1)According to the basic thermodynamics theory of pulse tube cryocooler,the following researches are carried out: Firstly,based on the enthalpy flow phase modulation theory,the phase diagram of the whole pulse tube cryocooler is established,the phase condition of the internal components of the pulse tube cryocooler is analyzed,and the optimal phase area of the whole pulse tube cryocooler is analyzed;Secondly,the energy losses of main parts is studied,and the loss of regenerator is mainly analyzed,including the influence of incomplete heat exchange loss,pressure drop loss and heat conduction loss on the efficiency of cryocooler.(2)Based on the basic control equation of compressor piston,the following research is carried out: Firstly,through the vector analysis method,the mechanical analysis of compressor piston is carried out.Secondly,Euler equation is introduced to convert the control equation and voltage balance equation of compressor into plural form for solution,and then according to the definition of resistance,the relationship between motor efficiency and PV power conversion efficiency of compressor is obtained,and the factors affecting compressor efficiency are analyzed.(3)Coaxial and in-line pulse tube cryocooler at 90 K temperature range are designed and manufactured.Enthalpy flows,energy losses,and cooling performance of the two PTCs are compared.The research shows that the linear pulse tube cryocooler has less energy losses and better cooling performance,and the thermal paremeter results are in good agreement with the basic theory in this chapter.By analysing the thermodynamics and dynamics of pulse tube cryocooler,which provides support for the one-dimensional numerical modeling of cryocooler,and provides ideas for improving the consistency of the cooling performance.2)Numerical simulation and experimental study on the factors influencing the performance fluctuation of pulse tube cryocoolerFirstly,the thermodynamic and dynamic analysis of the pulse tube cryocooler is used to establish the one dimensional numerical model of 12W@85K coaxial pulse tube cryocooler.Then,the paper studies the influencing factors which will fluctuate the performance of the cryocooler theoretically by numerical simulation,experimental analysis and correlation analysis.The influence results of each influencing factor on the performance of the cryocooler are obtained in the process control range(the range of each influencing factor is small).The experimental results show that the mesh filling rate of the regenerator(The actual input power change at the 12W@85K refrigeration capacity is 6.11wac)and the magnetic induction intensity of the compressor(The actual input power change of the 12W@85K refrigerating capacity is3.52wac)is the first and second effect on the performance of the cryocooler;the sealing clearance of the compressor piston,the resistance of the compressor motor,the compressor and the cold connecting pipe of different shapes,the filling depth of regenerator,the diameter of wire mesh,the volume of gas reservoir,the length of inertance tube,charged pressure and the temperature of the hot end of the regenerator have little influence on the performance of the cryocooler.Therefore,it can be obtained that to reduce the performance fluctuation of the cryocooler,the consistency of the mesh filling rate of the regenerator and the magnetic induction intensity of the compressor should be controlled.3)Performance regression and feature importance of influencing factors of pulse tube cryocoolerOn the basis of the research in Chapter 4 of this paper,the parameters which have a certain influence on the performance of the cryocooler or fluctuation range(max-min/avg)> 1% are selected as independent variables,including: compressor piston clearance sealing value,magnetic induction intensity of compressor,compressor motor resistance value,wire mesh filling rate of regenerator,diameter of wire mesh and thickness of wire mesh as the independent variable x and the cryocooler performance is the dependent variabley(The actual input power of12W@85K is used to characterize the performance of the cryocooler).The six influencing factors of cryocooler performance fluctuation are analyzed by three methods: Logistic ordered regression,Random Forest regression and XGBoost regression based on machine learning language.The regression results show that the significant variable influencing cryocooler performance is the wire mesh filling rate of regenerator;Both Random Forest regression and XGBoost regression reveal the first and second important characteristics influencing cryocooler performance are the mesh filling rate of the regenerator and the magnetic induction intensity of compressor,which are consistent with the results in Chapter 4.The feature importance of the other four variables is relatively low,and the order in each model is slightly different,which has relationship with the objective function of the model and the amount of sample data used for regression.Through Random forest regression and XGBoost regression,the prediction models of cryocooler performance and independent variables are established,and the average of mean relative errors are 5.62% and 4.59%,and the average of certainty coefficients are 0.805 and 0.906,respectively.The XGBoost regression has higher accuracy.Through the Random Forest and XGBoost method,the performance of the cryocooler is regressed,and the feature importance of the influencing factors are determined.Through the improvement and control,the performance fluctuation can be reduced and the performance consistency can be improved.Combined with the statistical analysis of big data and theoretical experimental research,it is also the first time in the field of space cryocooler.With the increase of sample data of satellite-borne cryocooler in the future,it will also provide support for the further improvement of the accuracy of regression model. |
PDF Full Text Request