Investigation of the strength of concrete-like material with waste rock and aeolian sand as aggregate by machine learning

Abstract

Abstract Solid waste filling is an important development direction for filling mining technology. This paper proposes to use waste rock and aeolian sand as aggregates to fill the underground extraction area in order to reduce the environmental damage of related waste solid. The experiments are optimized by response surface methodology and multivariate nonlinear response models are constructed to investigate the effects of different factors on uniaxial compressive strength (UCS) of concrete-like material (CLM). The performance of different swarm intelligence optimization algorithms is analyzed and combined with support vector regression model (SVR) to construct an intelligent prediction model for UCS. The results show that the packing density has a maximum value of 0.74 when the proportion of waste rock is around 0.6. The response model constructed in this paper has a P-value < 0.01 and R2 > 0.8, which indicates its high significance and goodness of fit. The UCS of CLM increases with the increase of cement content and slurry mass fraction, while it also increases and then decreases with the increase of proportion of waste rock. The ratio of waste rock and aeolian sand will affect the compactness of cemented structure. The better the ratio, the higher the average grey value of the cemented structure, the more compactness the cemented structure. The whale optimization algorithm-SVR model constructed in this paper has a prediction accuracy of more than 99% for UCS of CLM, which achieves high accuracy and fast prediction of UCS under multifactor conditions.