Study on strength prediction and strength change of Phosphogypsum-based composite cementitious backfill based on BP neural network

Abstract

The uniaxial compressive strength (UCS) of the backfill is crucial for ensuring the stability and safety of the goaf. It can lead to the collapse of the goaf and subsidence changes in the ground surface. To accurately assess and predict the strength of phosphogypsum-based composite cementitious backfill, the effects of four cementitious material factors (quicklime, phosphogypsum, NaOH, and cement) on the strength of the backfill were investigated. A BP neural network strength prediction model for phosphogypsum-based composite cementitious backfill based on orthogonal tests was established. The changes in the strength of the internal structure of the backfill at different curing periods (7 days, 14 days, and 28 days) were analyzed using SEM-EDS tests. The results showed that the established orthogonal BP neural network prediction model had fast convergence speed and high prediction accuracy. The prediction model's correlation coefficient (R) was as high as 0.99903, and the relative error was within 5 %. The results of the grey correlation analysis indicate that the phosphogypsum content is the primary factor influencing the strength of the backfill. Furthermore, the change rule of backfill strength in response to varying cementitious material content has been elucidated. The internal strength of the backfill at different curing periods is primarily influenced by the formation of 3CaO • AlO • 3CaSO • 31 HO (AFt). The findings of these studies are of great significance for the evaluation and prediction of the strength of phosphogypsum-based cementitious composite backfill and provide strong support for further research in this field.