Influencing Factors on Backfill Strength and a Combined Strength Prediction Model
ZHANG Peng, GAO Qian, WEN Zhen-jiang, ZHANG Tao
2021, 42 (2):
232-241.
DOI: 10.12068/j.issn.1005-3026.2021.02.013
The water-cement ratio, packing density and specific surface area are important factors that influence backfill strength. However, few studies have investigated the significance of these influencing factors. Hence, uniaxial compressive strength tests were carried out on test blocks aged 3d, 7d and 28d using a 3-factor 5-level orthogonal design. A variance analysis was performed on the experimental results. The results from this statistical analysis indicated the following F-value ratios for the three factors(water-cement ratio∶packing density∶specific surface area) at different ages: 3d(698.404∶26.148∶0.910), 7d(862.626∶35.465∶1.286) and 28d(1585.404∶31.695∶1.338). According to the analysis results, the influencing factors had the following order of significance: water-cement ratio > specific surface area > packing density. Due to the small F-values of the specific surface area and packing density, these variables were not considered as factors affecting strength when establishing a strength prediction model. To obtain a backfill with the maximum strength, the optimal mixture had a 1.2 water-cement ratio, a specific surface area of 410m2/kg, and a packing density of 0.6%. To study the relationship between the water-cement ratio and strength, the water-cement ratio in the optimal mixture was varied, and 30 groups of strength tests were performed with samples containing different water-cement ratios after ageing for 3d, 7d and 28d. Based on the experimental results, a combined model was established for predicting strength at 3d, 7d and 28d with a combination of grey theory, fuzzy set theory and Markov theory. The measured values were plotted on the same scatter diagram as the values predicted with the GM(1,2) model, the regression analysis model and the combined model. A subsequent analysis verified that the combined model had higher accuracy and robustness than other models.
References |
Related Articles |
Metrics
|