In order to explore the difference and contribution of volatile substances in quality levels of strong-flavor Baijiu， sensory evaluation， gas chromatography-ion mobility spectrometry and chemometrics were used to detect and analyze the volatile substances in quality levels of strong-flavor Baijiu. Through GC-IMS detection， 41 variables that can characterize the difference of volatile substances in three grades of Baijiu were screened， including 13 esters， 8 alcohols， 6 ketones， 3 aldehydes and 4 other compounds. Principal component analysis and partial least squares discriminant analysis based on peak volume of 41 variables. The results showed that the cumulative contribution rate of the first two principal components of PCA reached 80.1%， which can effectively distinguish different quality levels of Baijiu samples. PLS-DA can screen 17 characteristic biomarkers based on variable importance projection， draw clustering heat maps for the above substances to distinguish the contribution of aroma substances to different Baijiu samples， and construct a K-nearest neighbor model， with a discrimination accuracy of 100%. This study can provide a theoretical basis for the rapid evaluation of the classification of strong-flavor Baijiu.