Permeability Tests and Multi-Factor Analysis of Sand-Based Cemented Backll Five-Component Samples With Different Mix Ratios

: 11 Extensive coal mining involves the challenge of liberating coal resources under buildings, railways, 12 and water bodies. Sand-based cemented backfill (SCB) mining is considered an effective method to 13 solve this problem while utilizing solid wastes in large quantities. Since the groundwater seeping 14 into SCB pores in a complex mining environment deteriorates the SCB strength and stability, the 15 permeability optimization of SCB samples under multi-factor conditions by adjusting their mix 16 ratios is very topical. Therefore, in this study, a large number of SCB samples were prepared using 17 aeolian sand as aggregates, cement and fly ash as cementing materials, and quicklime and water 18 reducing agent as additives. The mass of the aeolian sand was a fixed value, while other ingredients' 19 content ratios were expressed as a percentage of aeolian sand mass. With all other factors being 20 constant, the level of one factor was changed at a time for univariate analysis. Four levels were set 21 up for each of the four factors, and 16 tests were performed for a total of 13 mix ratios with an axial 22 pressure of 1 MPa and confining pressure of 3 MPa. The effects of mix ratios, pore size, porosity, 23 and surface structure on SCB's permeability were analyzed in detail. Experimental results show 24 variations of quicklime an additive. The results showed that as the adhesive content and the degree of hydration increased, the cemented backfill's unsaturated hydraulic conductivity dropped (Abdul- al. Fall and Adrien explored the effects of curing temperature and time on the cemented backfill's permeability composed of tailings and cement. They reported that the


94
As shown in Fig. 1 (a), the primary pore sizes of aeolian sand varied between 188 and 638μm, 95 with an average of 348μm. A small range of particle size distribution indicated that aeolian particles 96 were refined and uniform, which was conducive to particles' adhesion and reduced the permeability 97 (Zhou N et al. 2019). As shown in Fig. 1 (b), cement's primary particle sizes varied between 9 and 98 50μm, with an average of 21μm. The particle size was distributed within a small range, and the 99 gradation curve showed a small dispersion. A smaller particle size resulted in a larger specific 100 surface area, which implied that more free water would be consumed to moisten the material (Bu 101 YH et al. 2021;Zhang BL et al. 2018). According to Fig. 1 (c), fly ash's primary pore sizes varied 102 between 14μm and 133μm, with an average of 44μm. Thus, the particle size was distributed over a 103 larger range, and the respective curve featured a larger dispersion. According to Fig. 1 (d), 104 quicklime's primary pore sizes varied between 6 and 158μm, with an average of 33μm. Thus, the 105 particle size was distributed over a larger range, resulting in non-uniform compactness of SCB 106 (Zhang TS et al. 2011). 107 (2) Analysis of the surface structure of raw materials via SEM 108 A Zeiss Sigma HV field emission scanning electron microscope (FE-SEM) was used to study 109 the surface structure of raw materials and evaluate its possible effect on the SCB's permeability. As shown in Fig.2 (a), the aeolian sand particles were dense and generally had rough surfaces 116 without pores or fissures. Such surface structure was conducive to the adherence and bonding of 117 cementing materials and reduced the SCB's permeability. According to Fig.2 (b), many scraps 118 adhered to the surfaces of cement particles. Some of them were rod-like hydration products. Many 119 fibrous or reticulate hydration products linked the aeolian sand particles (Benyamina S

Test procedures 127
The multi-factor experiments on SCB samples with different mix ratios consisted of three parts: 128 (i) sample preparation and pretreatment; (ii) permeability tests and analysis of their results; and (iii) 129 microstructural examination and analysis. The main test procedures are schematically presented in 130 Aeolian sand, fly ash, cement, quicklime, and water reducing agent were mixed at different 135 mix ratios. Then water was added. The mixture was prepared into a cemented backfill paste (CBP) 136 using an NJ-160A cement paste mixer. After that, CBP was poured into a cylindrical mold of 50mm 137 in diameter and 100mm in height. CBP was fully shaken via a shaker for 1-2 min to make it compact 138 and repel the air bubbles. The mold was removed after leaving the CBP to stand for 24 h. The 139 cylindrical samples were placed into a curing box to cure for 28 days. The specimens were cured 140 and then cut and polished with a cutting machine and a polishing machine, respectively. It was 141 ensured that the flatness of the two end surfaces was below 0.5mm and their parallelism was below 142 0.02 mm, which satisfied the Standard for Tests Method of Engineering Rock Masses (GB50218-143 94). After the above surface treatment, the samples were soaked in the water to make them water-144 saturated. 145 (2) Permeability test 146 up together with tape into a waterproof and heat-shrinkable sleeve. It was then placed into the groove 148 at the bottom of the seepage meter, and the groove was capped. An axial pressure was applied to the 149 sample. Water was injected into two pressurized vessels of the hydraulic pressure loading system 3 150 in Fig. 3. A specific pressure difference was ensured between the two vessels. The inlet and outlet 151 pipes were connected to the corresponding joints of the seepage meter. Each joint's tightness was 152 checked, and a closed channel filled with water was formed between the seepage meter and the 153 vessels. After the test was started, water in the two pressurized vessels entered the inlet and the 154 outlet pipes simultaneously. Therefore, a difference in hydraulic pressure was created between the 155 inlet and outlet pipes (and, thus, between the upper and lower surfaces of the sample). The hydraulic 156 pressure levels were monitored on the real time scale via the pressure gauge and the pressure 157 transmitter, whose readings were recorded by the data collection and analysis system. 158 The SCB samples' permeability was calculated as follows: 159 where cf is the coefficient of volume compressibility of the fluid, which was taken as 4.85×10 -3 in 161 this study; V is the volume of the voltage stabilizer; H is the sample height, taken as 100mm; µ is 162 the dynamic viscosity of the seeping fluid (water), taken as 0.89×10 -3 Pa.s for 25℃; tf is the end 163 time of equidistant sampling; A is the cross-sectional area of the sample, taken as 1.96×10 3 mm 2 ; P10 164 and P20 are the initial pressures of vessels 1 and 2 in the upstream and the downstream, respectively; 165 P1f and P2f are the final pressures acting on vessels 1 and 2 in the upstream and the downstream at 166 the end time, respectively. 167

(3) Microstructural examination 168
The influencing factors of SCB's permeability were analyzed on a microscopic scale. The 169 Archimedes method, mercury intrusion porosimetry, and SEM scanning were employed to detect 170 the porosity, pore size, and surface structure of SCB. Using the Archimedes method, the sample 171 block mass m1 was weighed after drying for 12 h. The mass m2 was weighed after sealing and gas 172 extraction for 2 h plus immersion in the distilled water for 12 h. The mass m3 was weighed after 173 wiping off the water. Thus, the porosity n was the ratio of the mass of fluid seeping into the pores 174 to that of fluid displaced by the sample: where m1, m2, and m3 are masses of the sample block after the described above procedures. 177 The mercury intrusion porosimetry was implemented by a YG-97A capacitive mercury 178 injection apparatus. The maximum pressure imposed by the apparatus was 60 MPa. The measurable 179 range of pore size of the samples was 0.016 to 63µm, with a precision of 0.5% (Fridjonsson EO et 180 al. 2013). The surface structure was observed by SEM scanning using a Zeiss Sigma-HV FE-SEM. 181 The samples were dried at a constant temperature of 40℃ for 12 h and then crushed. Smaller 182 samples with 10mm×10mm×5mm dimensions were harvested from inside the larger sample blocks. 183 With the natural face formed by crushing facing upwards, the sample was fixed to the base using 184 conductive adhesive or glue, followed by gold spraying. 185

Design of the experimental scheme 186
The mass of the aeolian sand was a fixed value. The contents of other ingredients were 187  Table 1. As seen in Table 1, four levels were set up for each of the four factors, 193 and 16 tests were performed under a total of 13 mix ratios, insofar as three samples (B3, C3, and D4) 194 were identical to the standard sample A2, with fly ash content of 70%, cement content of 15%, 195 quicklime content of 15%, and WRA content of 0.4%. 196

Influence of mix ratio on SCB's permeability 200
The axial and confining pressures were fixed at 1 and 3 MPa, respectively. Three identical 201 samples were tested by each test scheme, and their permeability values (K1, K2, and K3) were averaged, 202 as shown in Table 2 203 The averaged results on permeability variation with each ingredient's content are plotted in Fig. 4. As shown in Fig. 4 (a), when the fly ash content was increased from 60 to 90%, the SCB's 211 permeability exhibited a sharp rise, followed by a gradual drop, ranging between 0.773×10 -2 and 212 3.34×10 -2 mD. A more detailed analysis of Fig. 4(a) revealed that within the fly ash content range 213 between 60 and 70%, the SCB's permeability was highly sensitive to the fly ash content variation, 214 reaching its maximum of 3.34×10 -2 mD at 70%. After the fly ash content exceeded 70%, both the 215 permeability and the above sensibility decreased. This trend in concert with findings of (Behera SK 216 et al. 2019), which reported that a water film was formed and adsorbed by the spherical surfaces 217 when the fly ash came into contact with water. As the fly ash content increased, the amount of free 218 water consumed by the formation of water film decreased dramatically. Consequently, the hydration 219 of other cementing materials, such as cement, would be incomplete. Given the above, the SCB's 220 permeability increased significantly as the fly ash content grew from 60 to 70%. As the fly ash 221 content reached a certain level, fly ash would act as a fine aggregate, filling pores, enhancing the 222 backfill's compactness, and resulting in a gradual permeability decline (Ismail I et al. 2013). 223 According to Fig. 4 (b), as the cement content increased from 5 to 10%, SCB's permeability 224 exhibited a slight gradual drop. As the cement content was increased from 10 to 20%, the 225 permeability gradually grew at a decreasing rate, reaching its minimum of 2.30×10 -2 mD at the 226 cement content of 10%. The permeability range was between 2.3×10 -2 and 4.0×10 -2 mD. 227 Noteworthy is that cement is usually added as the main cementing material in SCB. Cement 228 hydration can give rise to ettringite (AFt) and calcium silicate hydrate (C-S-H), facilitating the 229 bonding of adjacent aeolian sand particles. The cement content of 5% is considered a too low dose 230 for cementing materials, which fails to ensure the bonding of aeolian sand particles. As the cement 231 content was increased to 10%, the number of hydration products was increased, reducing the SCB 232 permeability. As the cement content further increased to 15% and 20%, the dry material's specific 233 surface area grew sharply because there were many small-sized particles, which surface moistening 234 consumed a large amount of free water. The fine cement particles could also envelop the water 235 molecules to form a loose flocculent structure, which further led to an increase in permeability 236

(Kwan AKH and Chen JJ 2013; Qureshi T et al. 2018). 237
According to Fig. 4(c), as the quicklime content increased from 5 to 20%, SCB's permeability 238 showed a decreasing trend, and this drop was slowed down gradually. SCB's permeability varied 239 more significantly as the quicklime content increased. The variation range was from 3.0×10 -2 to 240 7.1×10 -2 mD. Noteworthy is that CaO in the quicklime is hygroscopic, and its hydration gives rise 241 to Ca(OH)2, which can facilitate the hydration of glass microspheres in fly ash (Panchal S et al. 242 2018). In this study, increasing the quicklime content reduced the migration of excess free water 243 during the preparation of SCB samples, reducing the backfill's microsphere loss. As a result, the 244 SCB became more compact, and its permeability dropped with the quicklime content. 245 As shown in Fig. 4 (d), as the water reducing agent content increased from 0.1 to 0.4%, the 246 SCB's permeability varied slightly and tended to stabilize. The permeability ranged between 2.9×10 -247 2 and 3.3×10 -2 mD. On the one hand, water reducing agent adhered to the surface of cement particles, 248 causing repulsion between like charges and reducing the enveloping effect offered by cement 249 particles to free water. Therefore, the raw materials came into full contact with free water, facilitating

Porosity 255
The influence of the contents of fly ash, cement, quicklime, and water reducing agent on SCB's 256 permeability was verified by measuring the corresponding changes in porosity. The porosity (n) of 257 each sample was assessed by the Archimedes method via Eq.(2). The test results are shown in Table  258 3 and Fig. 5. 259  As shown in Fig. 5, the variation curve of the overall porosity of SCB for various contents of 266 raw materials coincided well with the variation curve of permeability. As shown in Fig. 5 (a), SCB's 267 permeability and porosity reached their maxima at the fly ash content of 70%. In Fig. 5 (b), the 268 porosity curve was less steep than the permeability one, though both curves shared a consistent trend. 269 According to Fig. 5 (c), both permeability and porosity dropped with the quicklime content and 270 nearly coincided at the last stage. As seen in Fig. 5 (d), WRA content had a very slight effect on 271 SCB's porosity, which was earlier explained. 272 Thus, SCB's porosity ranged between 1.91 and 6.53% under different contents of raw materials 273 and closely correlated with the permeability variation. 274

Particle size effect on permeability 275
The relationships of SCB's permeability with pore distribution, pore size, and mix ratio were 276 analyzed. The pore size distributions of each sample are shown in Fig. 6. 277 As seen in Fig. 6 (a), the permeability reached its maximum at the fly ash content of 70%. The 278 primary pore sizes (top three pore sizes in terms of occurrence frequency) were 0.25-0.63 µm, 279 accounting for 53.5% of the total pore size. At fly ash contents of 60, 80, and 90%, the primary pore 280 sizes of SCB significantly exceeded those at 70%, ranging between 0. 16  According to Fig. 6 (b), when the cement content was 5, 10, and 15%, the primary pore sizes 290 of SCB ranged between 0.25 and 0.63 µm. The primary pore sizes for cement contents of 5 and 10% 291 accounted for 50.1 and 48.6% of the total pore sizes, respectively. The difference between the two 292 was small, and so was the corresponding permeability change. The pore size of 0.4 um accounted 293 for the highest proportion (about 25.49%) for the cement content of 15%; the proportion of pore 294 size of 0.25 µm dropped from 19.65% at the cement content of 10 to 15.33%, corresponding to that 295 of 15%. The share of large-sized pores at the cement content of 15% exceeded those at 5 and 10%. 296 The permeability also increased correspondingly. At the cement content of 20%, the primary pore 297 sizes were distributed in the range from 0.4 to 1µm, which significantly exceeded those of all other 298 contents. The corresponding permeability was also the highest. The above results indicated that the 299 cement content variation between 5 to 10% had a relatively small impact on the pore size distribution 300 of SCB, in contrast to that in the range from 10 to 20%. Generally speaking, the primary pore sizes 301 of SCB correlated more closely with permeability within the cement content range from 5 to 20%. 302 As seen in Fig. 6 (c), the primary pore sizes of SCB at quicklime contents of 5, 10, and 15% 303 were within the range from 0.25 to 0.63 µm. However, the share of pores with a size of 0.63µm 304 gradually decreased with quicklime content, accounting for 21.60, 17.71, and 12.69% of the total pore size for quicklime contents of 5, 10, and 15%, respectively. At the quicklime content of 20%, 306 the distribution of primary pore sizes differed little from that of 15% quicklime content. The 307 permeability dropped with the quicklime content and exhibited a saturation. Thus, when the 308 quicklime content was in the range from 5 to 20%, SCB's primary pore sizes positively correlated 309 with permeability: the smaller the primary pore sizes, the lower the permeability. If the primary pore 310 sizes remained unchanged, the corresponding permeability tended to stabilize. 311 The analysis of Fig. 6 (d) revealed that when the WRA content ranged from 0.1 to 0.4%, the 312 primary pore sizes of SCB ranged between 0.25 and 0.63 µm. Although the share of pores with a 313 size of 0.4µm fluctuated, this pore size consistently accounted for the highest proportion. Meanwhile, 314 the permeability changed less significantly and tended to stabilize. The above results indicated that 315 within the WRA content range from 0.1 to 0.4%, SCB's primary pore sizes closely correlated with 316 permeability. However, the WRA content had a weaker effect on SCB's pore size distribution. 317

Surface structure 318
The evolution law of SCB's microstructure and its effect on permeability were studied under 319 different mix ratios. SCB samples with different contents of fly ash, cement, quicklime, and WRA 320 were examined by SEM. The observations with the magnification power of 5000X are shown in Fig.  321 7. According to Fig. 7 (a), with fly ash contents of 60, 70, 80, and 90%, SCB's permeability first 322 increased and then decreased as the fly ash content increased. The permeability reached its 323 maximum at the fly ash content of 70%. According to the SEM images, at the fly ash content of 324 60%, the smallest pore sizes were observed on the SCB surface. These pores were mostly quasi-325 circular or polygonal in shape. The hydration of cementing materials was more significant, leading 326 to higher overall compactness and lower permeability of SCB. When the fly ash content increased 327 from 60 to 70%, the pore sizes on the SCB surface increased dramatically and cementing materials' 328 hydration was incomplete. As a result, the aeolian sand particles could not be sufficiently bound to 329 each other. The pores were mostly polygonal and strip-like, and SCB had a loose overall structure 330 and larger permeability (Zhao HT et al. 2020). At fly ash contents of 80 and 90%, the pore size 331 gradually dropped, according to the SEM images in Fig 7(a). A large number of scraps filled the 332 pores, reducing the SCB permeability.

338
In Fig. 7 (b), under the cement content of 5, 10, 15, and 20%, SCB's permeability first dropped 339 and then gradually rose as the cement content increased. It can be seen from the SEM images that 340 at the cement content of 10%, there were more cemented materials on the surface than at 5%. Pores 341 and fissures were smaller, and the overall structure was more compact. Correspondingly, the 342 permeability was lower (Alfonso I et al. 2019). When the cement content grew to 15 and 20%, there 343 were many cement flocs due to incomplete hydration. A larger amount of free water was enveloped 344 at higher cement contents, which inhibited the hydration. Fewer cemented materials were produced, 345 pores were larger, and the overall structure was looser. Therefore, the permeability gradually 346 As shown in Fig. 7 (c), the permeability gradually dropped with quicklime content at quicklime 348 contents of 5, 10, 15, and 20%. The pore sizes had a wider distribution wider range in the respective 349 SEM images, decreasing gradually during this process. Moreover, the amount of cemented materials 350 produced gradually increased, indicating a higher degree of hydration and a looser SCB structure. 351 Such variations were consistent with the permeability changes. 352 According to Fig. 7 (d), SCB's permeability varied little at WRA contents of 0.1, 0.2, 0.3, and 0.4% and tended to stabilize. In the respective SEM images, pore sizes of the SCB surface were 354 small at any WRA content. The pore shape and pore size distribution density were also similar at 355 different WRA contents. Thus, changes in the WRA content had little impact on SCB's surface 356 structure and permeability. 357

Sensitivity analysis of multiple influencing factors 358
The so-called range analysis was adopted in this study due to its simplicity, rapidity, 359 conciseness, and efficiency (Chen JJ et al. 2020). This approach implies that the overall response 360 ij K under different levels of each factor (where i is the level and j is the factor) is calculated. The  Table 4. 366 Next, the factors were ranged in terms of their influence degree. Each factor was plotted along 368 the x-axis and the range along the y-axis. The influence degree of each factor on SCB's permeability 369 was thus visualized in Fig. 8. 370 aquiclude strata in a typical mining area of China. J Cleaner Prod 267. https://doi.org/ 10.1016/j.jclepro.2020.122109.