Strength investigation of the cemented paste backfill in alpine regions using lab experiments and machine learning

Abstract

Cemented paste backfill (CPB) is increasingly used in mines in alpine areas. The strength of CPB is unstable due to the influence of multiple factors such as low temperature. In order to obtain the uniaxial compressive strength (UCS) of CPB under the action of different factors, a large number of laboratory tests had to be conducted in the past, which is time-consuming and laborious. In this paper, a research on the UCS is conducted by means of a combination of indoor tests and machine learning. The sparrow search algorithm is used to optimize the extreme learning machine, thus solving the uncertainty of the extreme learning machine in determining the initial weights and thresholds. The strength prediction model of CPB based on SSA-ELM model is constructed with the curing temperature, sand-cement ratio, slurry concentration and curing time as input variables and UCS as output variables. Compared with the original ELM, the PSO-ELM and BP-ANN, the correlation coefficients improved from 0.8817, 0.9511, 0.9187 to 0.9979, which indicates that the SSA-ELM model has better prediction accuracy. In addition, correlation analysis shows that the low temperature(<20 °C) environment has the largest effect on the UCS, which affects the development of UCS by influencing the process of hydration reaction.