There are many kinds of ancient glass in China, each with its own characteristics, which not only has artistic and practical value, but also deserves more attention for its historical and scientific value. The ancient glass is the witness of history and the carrier of civilization. The study of ancient glass is of great significance for excavating the value of cultural relics and inheriting Chinese civilization. This paper will study the chemical composition of these glass relics, analyze the difference between the composition and content of different types of glass, and establish a model to solve the problem of distinguishing and identifying cultural relics:Firstly, the correlation analysis of the relationship between certain variables such as weathering, type, texture and color of glass products is conducted by spearman rank correlation coefficient, and the conclusion is drawn that surface weathering is strongly related to glass type and texture, and is strongly related to glass color. Because the chemical composition will affect some physical (such as color) and chemical properties of glass, on the basis of this conclusion, the correlation analysis is carried out by using Lambda correlation measurement method and Goodman Guluska tau-y coefficient for chi square test, and the conclusions are drawn: ① High potassium glass is easy to be weathered, while lead barium glass is not easy to be weathered;② Class AC is easy to be weathered, while Class B is not easy to be weathered; ③ Light blue and blue-green glass is easily weathered. Firstly, the two types of glass are divided into four categories according to surface weathering, and the statistical rules of the content of each component before and after weathering are obtained by comparing and analyzing the content differences of each component before and after weathering; In the third part, the chemical composition content was predicted by means of mean ratio and regression model, and the results were obtained.In this paper, the classification rule is obtained through discriminant analysis, and the discriminant function is determined by Wilke Lanbda test showed that the classification rule was statistically significant; Secondly, based on K-means clustering algorithm, the two types of glass are divided into subcategories by Elbow Method, interval statistics and Silhouette Coefficients), Canopy algorithm, etc. determine the number of clusters as 3, and obtain specific classification results with mathematical software. Finally, the gray correlation analysis shows that the average values of various correlation degrees are large, that is, the samples within each classification are close to each other, indicating that the classification is reasonable. The classification sensitivity is obtained by comparing the content of main chemical components with the original classification results.