| Multispectral radiometric thermometry is a powerful tool to measure the high temperature of non-gray body surface.The multi-spectral radiation temperature data processing algorithm based on constraint optimization solves the problem that it is difficult to solve the underdetermined equations due to unknown emittance,but the accuracy and efficiency of the inversion are difficult to meet the measurement requirements due to the blindness of the iterative initial value selection.Funded by the national natural science foundation of China(No.61975028),this project aims to apply generalized inverse theory to the inversion process of multispectral radiation temperature measurement,and study the multispectral radiation temperature measurement technology based on generalized inverse from the theoretical and experimental aspects.In order to solve the above problems,a preliminary exploration is made.Firstly,based on the principle of multi-spectral radiation temperature measurement,the model of multi-spectral radiation temperature inversion underdetermined equations is constructed,which is converted into the form of matrix equation.By using the plus sign inverse matrix in the four m-p generalized inverse matrices,a temperature value and a group of spectral emissivity are directly calculated.Then this set of spectral emissivity of initial value,the two kinds of constrained optimization algorithm for iterative calculation,one kind is the traditional gradient projection method,the other is the outer point penalty function method,through the six kinds of typical materials under 1800 k emissivity model simulation,the results show that based on generalized inverse results as the initial value,the accuracy and efficiency of traditional gradient projection algorithm is improved,the computing efficiency is still low.For this reason,the generalized inverse-outside-point penalty function constraint optimization algorithm is introduced.Compared with the generalized inverse-gradient projection algorithm,the average relative error of temperature inversion of the former six materials is 1.6%without noise,while the latter is 5.0%,and the computational efficiency of the former is about 150 times higher.In the case of adding 5%random noise,the generalized inverse-outer point penalty function constraint optimization algorithm is compared with the generalized inverse-gradient projection algorithm,the relative error of the former six material temperature inversion in the case of no noise was 1.7%,3.0%,while the former is about 93 times higher operation efficiency,show that the generalized inverse-outer point penalty function method has good performance in terms of retrieval precision and efficiency.In order to verify the validity of the generalized inverse-outer point penalty function method,based on the United States ocean optics near infrared optical fiber spectrometer and optical lens,set the multi-spectral radiation temperature measurement experiment device,after black-body furnace calibration,radial and axial temperature of butane flame temperature distribution were measured,and compared with thermocouple measurements.Experimental results show that when the butane flame temperature scope in 500℃～680℃,the average relative error of axial temperature measurement is 2.1%,the average relative error of the radial temperature measurement is 1.9%,the uncertainty of the result of the experiment were analyzed,and the total uncertainty of 3.0%,generalized inverse-outer point penalty function is verified the effectiveness of the algorithm.Outside point method based on generalized inverse-multi-spectral radiation constrained optimization algorithm,to solve the constrained optimization algorithm of the initial value selection problem blindly,so as to realize the cases without assuming emissivity model of gray body material rapid inversion of surface temperature and spectral emissivity,which is spectral radiation temperature measurement technology to achieve more theoretical basis is provided for distributed real-time measurement. |
