Subclassification Determination and Weathering Composition Prediction of Ancient Glass Artifact Compositions Based on Clustering Algorithms

Abstract

Ancient glass is susceptible to environmental weathering, and the content of chemical elements varies with the degree of weathering. Through the correlation analysis of weathering degree and color, pattern and composition by Chi-square test, it is found that the weathering degree of glass surface changes based on composition and color, and pattern has little correlation. The principal component analysis of the glass is based on different chemical compositions and whether the glass is weathered or not. Through the supervised learning k-means algorithm, four types are obtained: unweathered high potassium glass, weathered high potassium glass, unweathered lead barium glass, weathered lead barium glass; Further, the glass is classified into 9 subclasses based on the content of other elements to facilitate the study. The cloud model was used to predict the composition of cultural relics before weathering, and the relationship between the main chemical composition of high-potassium glass and lead-barium glass before weathering was obtained, which was convenient for subsequent research and restoration of cultural relics.