Machine Learning-Based Comprehensive Analysis of Phase, Composition, and Microstructure of Yuan Dynasty Oil-Glazed Porcelain

The City Ruins in Jining in Inner Mongolia were an important transportation and trade hub in the Yuan Dynasty. Archaeological excavations have unearthed a large number of black-glazed oil-spot porcelain pieces, providing significant materials for studying ancient Chinese ceramic firing techniques. Using technologies such as super depth-of-field microscopy, X-Ray Fluorescence (XRF), X-Ray Diffraction (XRD), Scanning Electron Microscopy with Energy Dispersive X-ray Spectroscopy (SEM-EDS), and Raman spectroscopy, this paper conducts a scientific archaeological analysis on representative samples. The results show that the chemical composition of the body of the glazed oil-spot porcelain unearthed from Jining site possesses remarkable characteristics of northern Chinese porcelain, which has rich aluminum and poor silicon. Analysis of the micro-structure and phase of the crystals on the glaze surface reveals an iron oxide crystalline phase transformation in the order of γ-Fe۲O۳→ε-Fe۲O۳→α-Fe۲O۳. The formation mechanism of the glaze surface is identified as anorthite crystallization --intergranular liquid phase separation -iron oxide crystallization. The glaze's chemical composition features rich Fe۲O۳ (۶.۲۶۸%), rich CaO (۵.۵۲۷%), and poor Al۲O۳ (۱۴.۹۴۵%), with an average RO content of ۰.۷۶. Analysis of the micro-structure and phase of the crystals on the glaze surface reveals an iron oxide crystalline phase transformation in the order ofγ-Fe۲O۳→ε-Fe۲O۳→α-Fe۲O۳. The formation mechanism of the glaze surface is identified as anorthite crystallization --intergranular liquid phase separation -iron oxide crystallization. This paper utilized an Artificial Neural Network (ANN) to better understand the conditions and interrelationships among parameters. These practical networks have proven successful in diverse functions and complex problems, even with training data errors. The study showed neural networks to investigate enamel phase transformation, and formation by manipulating chemical compositions, with acceptable error confirmed through linear regression. The ANN model was able to predict the enamel microstructure, phase transformation, and formation mechanism with an error of less than ۱% compared to the target values.

کلیدواژه ها:

Oil-spot ، Black glazed porcelain ، ε-Fe۲O۳ ، Microstructure Phase ، Composition