Experimental and machine learning-based optimization of mix design and compressive strength prediction for a novel dual-waste backfill material

To achieve in-situ utilization of coal mining wastes and address environmental and engineering challenges, this study investigates a novel dual-waste backfill material (NDWB) prepared by combining coal gangue (CG) and high-salinity concentrated mine water (HSCM water). The effects of water-binder ratio (w/b), sulfate ion concentration, CG powder content, and Ordinary Portland Cement (OPC) content on the uniaxial compressive strength (UCS) and fluidity of NDWB were examined through orthogonal and single-factor experiments. The findings indicate that increasing the w/b consistently reduces UCS across all curing ages. Sulfate ion concentration initially enhances UCS but decreases at higher levels, whereas at 14 and 28 days, higher sulfate concentrations further improve strength. The UCS is maximized at 50 % CG powder content, while excessive amounts negatively impact strength. UCS first decreases and then increases with OPC content, reaching its lowest at 10 %, while 30 % OPC content yields the highest strength (4.126 MPa at 28 d). Moreover, orthogonal experiments indicated the w/b as the primary factor influencing both strength and fluidity. with the optimal mix determined as w/b= 3.0, sulfate ion concentration= 20,000 mg/L, CG powder content= 50 %, and OPC content= 30 %. Additionally, strength prediction models (decision tree, random forest, gradient boosting decision tree, light gradient boosting machine, and extremely randomized trees) were developed, with GBDT showing the highest accuracy on both training and test datasets. This study establishes a theoretical basis for the application of NDWB in mine backfilling and offers valuable guidance for its practical engineering implementation.