基于机器学习的风化硅酸盐玻璃原成分预测及亚分类方法

Prediction of Original Ingredients of Portland Glass and Research into Subclassification Methods Based on Machine Learning

玻璃在中国史料中早有记载,但是由于长期存在名称和质地的混淆,且近现代有关中国古代玻璃的研究起步较晚,关于古代硅酸盐玻璃的风化和成分研究比较缺乏。以往研究古代玻璃器的著作,多是从王朝更替的角度,对文化交流、化学分析等方面研究玻璃的文化艺术形态及其自身的运行发展的规律,较少有学者系统建立数学模型并使用智能算法定性定量开展风化硅酸盐玻璃原成分预测及亚分类方法研究。本工作以多组风化和未风化硅酸盐玻璃为研究对象,提取其化学成分含量、纹饰和颜色等数据,利用Spearman系数分析了纹饰、颜色和玻璃大类之间的相关性并研究影响表面风化的因素;利用决策树进行大致分类,然后用神经网络预测玻璃风化前主要化学成分的含量,并总结硅酸盐玻璃的分类依据。此后通过K-means聚类建立分类模型:确定最佳类别数,进行亚类划分,寻找铅钡玻璃和高钾玻璃的最优分类数量。研究结果表明,只有玻璃类型对表面风化具有显著影响;风化过程中参与度较高的化学成分为二氧化硅、氧化铝、氧化铅、氧化钡、氧化铅和五氧化二磷;风化后,铅钡玻璃二氧化硅含量明显下降,氧化铅含量明显上升,而高钾玻璃二氧化硅含量明显上升,氧化钾氧化钙和氧化铝含量明显下降;高钾玻璃分为3个亚类,铅钡玻璃分为4个亚类。为后续利用机器学习研究古代硅酸盐玻璃的风化和成分提供了参考。

Glass as a material has existed in China for a long time,but the related studies on ancient glass in China started relatively late due to the long-term confusion of name and texture,leading to a lack of research on the weathering and composition of ancient silicate glass.Some previous studies on ancient glass mainly discussed the artistic character and development laws of glass with respect to cultural exchange and chemical analysis from the perspective of dynastic succession.A few work established the related mathematical model and used the intelligent algorithm for qualitative quantification of weathering silicate glass original composition prediction and subclassification method.This paper was to use multiple groups of weathered and unweathered silicate glasses and collect/extract the data on their chemical composition content,ornamentation and color.The relations among the patterns,color,types of glass and surface weathering were analyzed by the Spearman coefficient.The decision tree for a rough classification and neural network to predict the main chemical composition of glass before its weathering was given,and the classification basis of silicate glass was summarized.Besides,the subcategorization at the optimal quantity of categories to conduct subclass classification was established,and a reasonable amount of barium glass and high potassium glass was selected.The results show that the type of glass has an influence on the surface weathering,and there are silicon dioxide,aluminum oxide,lead oxide,barium oxide and phosphorus pentoxide involved in the weathering process.Moreover,the amount of silicon dioxide decreases and lead oxide increases sharply in lead barium glass,while vice versa in high potassium glass after weathering.