Study on the composition of glass artifacts based on PSO particle swarm neural network

Abstract

The analysis of the composition of glass artifacts has always been the focus of research in related fields. By analyzing and identifying the different compositions of glass artifacts from different eras and fields can support the understanding of the history of glass artifacts, their achievements and thus their exit from human civilization. In this paper, we use multiple correlation analysis, kernel principal component analysis, BP neural network model optimized by PSO particle swarm algorithm to study the composition of ancient glass artifacts.To address the shortcomings of the conventional BP neural network model, this paper adopts the BP neural network model under the optimization of PSO particle swarm algorithm to identify and study the unknown artifact types. And different subjective assignments are made to glass artifacts with different weathering degrees to discover the level of model sensitivity.Then the S-W normal distribution test was used and it was found that most of the chemical components did not obey normal distribution. Therefore, nuclear principal component analysis was chosen to find out the correlation between different types of components. On the basis of the identified two correlation coefficients and the paired data were normally distributed, by paired sample t-test, it was found that the magnitude of the differences between the chemical components of different categories was large. This was supplemented by correlation analysis and kernel principal component analysis. Finally, the composition identification study of glass artifacts was realized, which provided a new methodological path for the composition study of glass artifacts.