Research On Multi-spectral Radiometric Thermometry Algorithm Based On Optimization Theory

Temperature measurement technology can be divided into non-contact temperature measurement technology and contact temperature measurement technology.With the rapid development of the society,the requirements of people for the quality of life are gradually improved.Non-contact temperature measurement technology is becoming more and more popular because the non-contact temperature measurement is fast and will not cause any impact on the measured object.Multi-wavelength radiation thermometry can obtain the true temperature and spectral emissivity of the object at the same time,which is a unique advantage in the field of radiation thermometry.The difficulty and emphasis of multi-wavelength radiation thermometry is the changeable spectral emissivity.The spectral emissivity change with materials,as well as temperature,wavelength,roughness,oxidation degree and so on.After the continuous exploration of scholars in recent decades,the solutions to the problem of variable spectral emissivity can be divided into four types.The first assumes that the spectral emissivity is a constant or that its variation is negligible,thus it is greatly simplified during the processing of multiwavelength data.This method assumes a gray-body model.The second assumes that the relationship between spectral emissivity and wavelength is a polynomial function,and then use mathematical methods such as least square method to compute each coefficient of the polynomial,and bring them into the equations to get the true temperature.This method assumes a wavelength model.The third assumes that the spectral emissivity and true temperature have a function relation to calculate the true temperature.This method assumes a true temperature model.The fourth is the neural network model method.Through training,neural network can invert the true temperature without assuming the spectral emissivity model.In this paper,the unknown and changeable spectral emissivity problem is solved by transforming the multi-spectral temperature data processing problem into the optimization problem without assuming emissivity model based on the theory of iterative method,and the problem of unestimable error caused by assuming spectral emissivity model is avoided.This method can automatically identify different emissivity trends.The theoretical basis of radiation thermometry is summarized in this paper,and three mathematical models of multi-wavelength radiation thermometry are introduced.For the unknown spectral emissivity problem,an optimization method is proposed to process multi-wavelength radiation thermometry data.In the third chapter,BFGS algorithm,DFP algorithm and Broyden algorithm in the quasi-Newton method are used separately.These three algorithms can automatically identify different emissivity variation trends,and all of them can invert the true temperature and spectral emissivity.The maximum relative error of inversion temperature is less than 2 %.The efficiency of this algorithm is high,and the calculation time is less than 0.2 s.In the fourth chapter,the objective function and equality constraints of SQP algorithm are constructed.The same data in the third chapter are used for simulation experiments.The results show that the algorithm can automatically identify different emissivity.In the fifth chapter,we use the experimental data of Fourier spectrometer(FT-IR)to verify the practicability and effectiveness of the above four algorithms.