Предиктивное моделирование эксплуатационных свойств пеностекла с использованием моделей машинного обучения на основе линейной регрессии

Abstract

Building construction requires the use of efficient thermal insulation materials such as foam glass in view of energy conservation. The paper considers predictive modelling of the performance properties of foam glass using machine learning models. The paper presents a mathematical description of the additives impact in the charge on the properties of foam glass. Nine charge compositions for foam glass synthesis were developed and their main microstructure parameters were determined. The authors tested the regression models using the JupyterNotebook software environment and the SciKit-Learn library in the Python programming language. The paper analyses the regression equation coefficients and estimates the modelling error. The obtained results confirm the effectiveness of predictive modelling of foam glass performance properties on the basis of linear regression.