Abstract: Cigarette ash is an important material evidence at the scene of the case. In order to achieve the purpose of rapid non-destructive detection of cigarette ash, 83 collected cigarette ash samples were theoretically analyzed and then tested by energy dispersive X-ray fluorescence spectrometer. A classification model of cigarette ash based on chemometrics was established. First of all, cluster analysis was used to distinguish samples, and the accuracy of clustering results was tested by regression analysis. Then, the material elements were used as variables and the discriminant classification model was established by discriminant analysis. The results show that the clustering results are good, and the regression analysis shows that the cluster analysis categories and elements can establish a good fitting relationship. The accuracy of the classification model obtained by the discriminant analysis is 100%. If you want to classify the unknown cigarette ash samples, you only need to input the relevant variables of the discriminant model, which will show its location in the discriminant distribution map, and then you can classify the samples. This method is simple, fast, accurate and reliable. This study provides a reference for the actual handling of cases at the grassroots of public security.