基于K-Means聚类的玻璃风化成分分析与鉴别

Glass cultural relics are easily weathered by the environment, resulting in the change of their internal chemical composition, which affects the category judgment. In this paper, by eliminating the wrong data and filling in the missing values, the control variable method is used to observe the relationship with the other three variables, taking whether weathering is a fixed value. Chi-square test was used to analyze the difference of known categorical variables, and the significance P was used to judge whether there was difference between glass color type and surface weathering. Through the known chemical composition data, the representative chemical elements of high potassium and lead-barium before and after weathering were selected to draw the change map before and after weathering, and the fluctuation of data was observed. The weighted average prediction model was established, and the standard normal distribution function was used for weighting. The linear mapping relationship between the chemical composition content of different kinds of glass before and after weathering was calculated, and the chemical composition content before and after weathering was predicted. On this basis, a subclass division model was established based on the standard deviation, and the K-Means algorithm was used to select appropriate indicators for the four categories of whether different types of glass are weathered, and the appropriate indicators were selected according to the significance P-value for subclass division, which was divided into three subclasses. Finally, the sensitivity analysis is carried out and the data is disturbed, which reflects the good robustness and universality of the model in this paper to a certain extent.