Prediction by regression for strength of chemically active composite materials

Abstract

The work is devoted to the statistical modeling for short- and long-term predictions of the chemically active properties of the materials by the data on their physical and chemical characteristics measured by non-destructive testing. These materials primarily include products from cement and concrete. Their resistance to compression and bending are the most important quantitative characteristics, which is the basis for the design of various buildings and structures. Information about their strength at a given time, as well as its possible value in the future has the importance paramount in different technological solution. These include problems of establishing optimal mineralogical composition, the definition of moments of time removing formwork, condition monitoring of bridges, dams, power plants and major construction projects, assess their stability, including of natural disasters etc. Statistical analysis of experimental data is the basis for decision-making process, allowing minimization of economic losses. Linear and nonlinear regression modeling of hardening process of such materials is implemented, and an adaptive algorithm is proposed and tested for the prediction of strength and comparative analysis of experimental data.