Machine learning based strength prediction method for cement-based grouting material

Abstract

In order to accurately predict the compressive strength of cement-based grouting material by small diameter core sample method, the compressive strength tests of cement-based grouting material standard size test blocks and small diameter core samples were carried out by pressure testing machine, and based on the test data, support vector machine regression (SVR) and random forest regression (RFR) were used to predict the compressive strength of cement-based grouting material. The results show that the standard size test blocks all show the failure pattern of the quadrangular cone with positive and negative continuation, while the small diameter core samples with a high diameter ratio of 0.7 and 1.0 show a cone failure form with positive and negative connections, and the small diameter core samples with a high diameter ratio of 1.2 show an oblique crack shear failure form; The compressive strength values of standard size test blocks and small diameter core samples all follow a normal distribution and have no outliers; As the age increases, the compressive strength of standard size test blocks and small diameter core samples increases, and they have the characteristics of higher early strength; The compressive strength of the core sample with a diameter of 46 mm is less and more susceptible to the influence of machining accuracy; At a given age and diameter, the compressive strength value of the core sample with a high diameter ratio of 0.7 is the largest, and the degree of dispersion of compressive strength is the smallest; The RFR prediction model has a better effect on the compressive strength of cement-based grouting material. The proposed method can accurately predict the compressive strength of cement-based grouting material, which provides some reference for the prediction and research of compressive strength of cement-based grouting material.

Keywords

• non-metallic building material; cement-based grouting material; machine learning; small diameter core sample; compressive strength