Research On Multi-spectral Temperature Measurement Technology Based On Spectrum Recognition

Temperature is an important parameter in the thermodynamics, physics, materials, metallurgy and other subjects. With the development of aviation and aerospace technology, and the improvement of industrial process measurement and control, there are more requirements for temperature measurement method and accuracy. Multi-spectral thermometry measures the object spectral radiance at a certain monment to calculate the object temperature with the advantage of high speed, non-contact measurement, no upper limit of temperature and so on, has the important economic and social significance for the development of aviation, aerospace, national defence and industry.The existing muti-spectral thermometry can not obtain overall emissivity characteristics due to the fewer amount of wavelengths. Especially for the untrained specimens, the application is subject to the large calculated error of temperature. Therefore, the temperature measurement method based on emissivity model recognition in order to improve the true temperature measurement accuracy and expand the application field of multi-spectrum thermometry is an important scientific problem to be solved. In the existing multi-spectral thermometry, a lack of the temperature measurement for special objects with compl ex emissivity characteristics is a scientific problem to be solved. The existing radiation thermometers have disadvantges of high lower limit temperature, fewer wavelength amount and limited universality. The development of the multi-spectral radiation thermometer with middle-infrared band, more wavelength amount, fast response speed and suitable for multi-class objects is critical technical problem to be solved.The subject "Research on multi-spectral temperature measurement technology based on spectrum recognition" is to solve the problems in the emissivity mode recognition, multi-spectral thermometry for special objects and multi-spectral radiation thermometer. The main contents of this paper are as follows:1. Muliti-spectral thermometry based on spectrum recognition is proposed. Aimed at the problem of the large calculated error for the untrained specimens in the existing muliti-spectral thermometry based on learning, multi-class objects spectral radiation characteristics at wide spectrum and multi-wavelength are tested. Five emissivity models are established and the corresponding spectral radiance are generated as research subjects. The feature extraction method based on principal component analysis(PCA) is applied to extract the spectral intensity and shape feature. The spectral training specimens are established. T he objects recognition method based on least squares support vector machine(LSSVM) is studied for emissivity model recognition. The multi-spectral temperature measurement model is established based on wavelength optimizing. The adaptive simulated annealin g(ASA) algorithm is applied to solve the object temperature. The simulation results show that the temperature calculated error is 0.69%. The universality of muliti-spectral thermometry is improved.2. Multi-spectral thermometry based on spectrum recognition for special objects is proposed. Aimed at the lack of the temperature measurement for special objects with complicated emissivity characteristics, the spectral emissivity models with subsectional functions are established on the analysis of the spectral radiation characteristics of the special objects. The wavelet analysis method is applied to detect and eliminate the absorption band and pulse interference. The singular points are detected to preclassify and used to subsect the object spectrum. For each subsectional spectrum, The object recognition method based on LSSVM is applied for emissivity model recognition. The multi-spectral temperature measurement model with subsectional function is established and solved with the ASA.The simulated results show that the temperature calculated error is 0.68% for the special spectrum. The temperature measurement for special object is solved.3. The mid-infrared multi-spectral temperature measurement system is developed. Aimed at the problem of high lower limit temperature, fewer wavelength amount and limited universality in the existing apparatus, the grating beam splitting is adapted to increase the wavelength amount to 256. The Pb Se detector is applied to expand the spectrum from near infrared range to the mid-infrared range 2.5-5μm. The lower limit of temperature is reduced to 500℃ without liquid nitrogen cooling. So the practicalility of the system is improved. The upper computer and lower computer are designed to improve the portability. The zero point correctio n method is studied. The sample heating apparatus in the temperature range of 500-1300℃ is developed for the experiments. The calibration method of monochromator is proposed based on the continuous spectrum source to improve the accuracy of monochromator. The temperature of the real objects is measured by this system. The expanded uncertainty of temperature measurement result is estimated about 2.6 4%.