Data processing method of multi-wavelength pyrometer based on neural network

Multi-wavelength pyrometry is a hopeful method for the true temperature measurement.. The problem for this research area is how to transform the collection of the brightness temperatures into the single true temperature. A linear relationship between logarithm of spectral emissivity and wavelength was assumed in usual data processing of multi-wavelength pyrometer, which is only fit for certain particular materials. A new data processing method based on neural network is presented without the need of "guessing" the emissivity function. The neural network can provide a general solution of reasonably good accuracy. The results of the simulated experiments show that the difference between the calculated true temperature and the real true temperature is within +/-40 K for both trained emissivity samples and untrained emissivity samples. And the true temperature errors of trained emissivity samples are generally less than that of untrained emissivity samples. The new data processing method based on neural network is an effective method for the true temperature measurement of most engineering materials.