Abstract: In the study of the effect of stress rock infrared radiation, although AIRT temporal curve is used widely, due to uneven distribution and anisotropy of rock specimen composition and damage, the AIRT tempo - ral curve is always in a variety of forms with common features hidden and difficulty to be classified clearly. The paper, based on infrared observation experiment of siltstone under uniaxial compression, resamples and normalizes the original AIRT time series, then carries out grey correlation clustering analysis and classifies them with development trend of similar degree between sequences as quantitative indicators. Results show that for the test, when the critical value is 0.7, the AIRT temporal curve can be divided into two categories, the first class curve of which accounts for 60 % of all, with obvious infrared abnormal precursors, and the data pro - cessing method has very important reference significance for the AIRT evolution characteristics and catastro - phe precursor quantitative research.