Research On Multispectral True Temperature Retrieval Algorithm Based On Optimization Principle

Multispectral thermometry is a process of retrieving the true temperature o f radiators by measuring the information of multispectral radiations and using related theories and algorithms.The solution of Spectral emissivity is still the key and difficult y in multispectral thermometry.Theoretically,it is necessary to know enough spectral information to obtain the true temperature of the radiator.Considering that the spectral emissivities of actual radiators at different spectr um and temperatures are usually inconsistent,and the solution of spectral emissivit y is an unavoidable problem in non-contact radiation temperature measurement,it is o f great scient ific significance and application value to carry out the research on the solution of multispectral emissivit y and the inversion methods of true temperature.After decades of development,the solution spectral emissivity can be generalized into four types of models.One is the grey body hypothesis model,which considers that spectral emissivity is a constant or its change can be neglected in the process of temperature inversion;the other is the wavelength hypothesis model,which considers that there is a certain relationship between spectral emissivit y and wavelength in the process of temperature inversion.Thirdly,the true temperature hypothesis model,which considers that there is a certain relationship between spectral emissivity and true temperature in the inversion process of the true temperature,and establishes a model between spectral emissivity and true temperature and realizes the inversion of true temperature wit h iteration method;Fourthly,the establishment of a neural network model,which achieve true temperature inversion by the learning neural network.Based on the uniqueness of true temperature and the analysis of different hypothetical models,the thesis tries to find a general true temperature inversion method without the hypothesis of spectral emissivit y model,and carries out the research work wit h multispectral true temperature inversion method as the core.The thesis summarizes the characteristics of traditional mult ispectral true temperature inversion theories and methods.In view of the complexity of selecting spectral emissivit y model in the existing multispectral true temperature inversio n process,a true temperature inversion method based on the constrained optimization principle of single objective function minimization is proposed.This method does not need to assume spectral emissivity model and convert the true temperature solution problem into the constrained optimization problem.Compared with the trad itional second measurement method,under the same init ial conditions,the true temperature inversion speed of the new method is increased by more than 98%.The true temperature inversion based on the single objective functio n minimization method has a higher speed,but the inversion accuracy is lower than that of the traditional second measurement method,and the some errors are more than 1%.In view of the low accuracy of the single objective function minimizatio n method in true temperature inversion,anot her multi objective function minimization method based on constrained optimization principle is proposed.The inversion accuracy of this method is approximately the same as that of second measurement method,but the inversion speed of the true temperature is still significantly higher than that of second measurement method.Compared with the single objective function minimization method,the inversion accuracy of the mult i objective function minimization method is better than that of the single objective function minimization method.All the inversion errors are within 1%,and the some errors are zero.Therefore,the multi objective minimization method is more suitable for true temperature inversion with higher accuracy.In view of the fact that Wien’s formula can not be used to replace Planck’s formula in some case in the actual multispectral radiation measurement,a method of true temperature inversion based on Planck’s principle is proposed.I n the method,it is still unnecessary to establish spectral emissivit y model,and convert the problem for solving true temperature into an optimal problem with constraint conditions.The simulation results show that all the inversion errors of this method are less than 1%,which further expands the application scope of multispectral true temperature inversion algorithm based on constrained optimization principle.The relationship between the true temperature points and the reconstructio n accuracy in the construction of two-dimensional true temperature field is explored.The influence of different interpolation methods on the prediction accuracy of temperature field distribution under different sampling numbers is discussed.Based on the relationship between the numbers of points,two hypothesis test methods are used to test the temperature field.In addit ion,the relevant test conclusions are given,which provide the relevant theoretical analysis and objective evaluation methods for the construction of the true temperature field o f rocket exhaust and the structural design of the flat blackbody.