Research On Multispectral Radiation Temperature Measurement Inversion Algorithm Based On GIM-LSTM

Multi-spectral radiation temperature measurement technology has the advantages of fast response speed,no interference with the measured field and no upper limit of measurement.It is one of the most powerful tools in the field of high temperature measurement.It has been widely used in high temperature measurement fields such as rocket engine plume diagnosis,smelting,petrochemical,glass,welding and semiconductor.It calculates the temperature value through data processing inversion based on the radiation intensity information of multiple spectral channels.Because the spectral emissivity of the target is unknown,the multi-spectral radiation temperature is difficult to solve directly.This paper is funded by the National Natural Science Foundation of China.In view of the above problems,a multi-spectral radiation temperature measurement inversion algorithm based on generalized inverse-long short-term memory network is proposed,and the theoretical and experimental aspects are studied in depth.In terms of numerical simulation,466560 sets of temperature data sets composed of 243 emissivity models were constructed,and the classification of data sets was realized by generalized inverse algorithm.In this paper,the Recurrent Neural Network(RNN),Long Short-Term Memory Neural Network(LSTM)and Generalized Inverse Matrix(GIM)algorithms are combined to propose generalized inverse-recurrent neural network and generalized inverse-long short-term memory network two multispectral radiometric temperature inversion algorithms.Firstly,the generalized inverse algorithm is used to classify the spectral emissivity data set.The first identification of spectral emissivity is realized,and then sent to the corresponding secondary identification sub-network to realize temperature inversion.The simulation results show that the generalized inverse-recurrent neural network and generalized inverse-long short-term memory network inversion algorithms are better than those without generalized inverse participation.The generalized inverse-long short-term memory network algorithm is superior to the generalized inverse-recurrent neural network inversion algorithm.In the experimental aspect,the measured data of the rocket plume temperature measured by the public multispectral pyrometer and the self-built simulated radiation source temperature data based on the blackbody furnace and the filter are used to verify the results.The results show that the multi-spectral radiation temperature measurement data processing algorithm based on generalized inverse-long short-term memory network meets the measured requirements.