The Dosage of Superplasticizer in Cemented Coal Waste Backfill Material Based on Response Surface Methodology

Cemented coal waste backfill material (CCWBM) is developed for backfilling the goaf in coal mines. As fresh CCWBM slurry is generally transported into underground openings through a pipeline, the fluidity of fresh slurry becomes one of the most important properties. Adding superplasticizer is considered to improve the flow performance of the slurry without alerting the mechanical performance of filling body. ,e dosage of superplasticizer (SP) is related to filling cost, thus response surface methodology (RSM) is adopted to study the influence of material composition on SP when target slump is 250mm.,e effects of fly ash content, fine gangue ratio, and mass concentration on SP are analyzed using the software of Design-Expert and central composite design (CCD), and models are established for SP. Results show that the SP model coincides greatly with the test results and can be applied to analyze and predict SP in CCWBM. Mass concentration, fly ash content, and fine gangue ratio influence SP from high to low.,e interaction of fine gangue ratio and mass concentration between SP is the most significant. ,e fact that the improved aggregate space model can be applied to analyze the fluidity of CCWBM is proved too. ,e research results provide guidance for the design and preparation of CCWBM with favourable performance and low cost in practical production.


Introduction
In China, underground coal mines produce about 95% of total coal [1].Underground coal mining indeed contributes a large amount of coal, but it also creates a tremendous amount of mined-out void and discharges a great deal of solid waste [2,3].Most of the waste is coal gangue, and it is reported that approximately 4.5 billion tons of coal gangue is stocked in China at present [4].e waste is simply disposed aboveground and stockpiled in waste dumps, which causes serious environmental problems.On one hand, the waste dumps occupy much land and the poisonous substance contained in the waste contaminates soil and under water [1,5].On the other hand, the spontaneous combustion of the waste leads to air pollution or even personal injury.Meanwhile, underground void results in land surface subsidence [6,7].Cemented coal waste backfill material (CCWBM), which is an engineered mixture of crushed coal gangue, fly ash, cement, and water, is developed for mine backfilling [8][9][10][11].All the problems aforesaid can be solved by backfilling CCWBM into the goaf [5,9,12,13].
Generally, filling slurry is prepared in a filling plant and then transported to underground openings, thereby necessitating the use of sufficient water to gain favourable fluidity [14].But this could lead to a high water/cement ratio, which will lower the strength and durability of filling body [15,16].Considering this challenge, the additive of superplasticizer into CCWBM is considered to reduce mixing water to enhance mechanical properties.Several studies have been conducted on the influences of superplasticizer on the properties of backfill materials.Ercikdi et al. discussed the influences of various superplasticizers, lignosulfonate (EUCO-FILL 30), naphthalene sulfonate (IKSAMENT NS), and polycarboxylate-based superplasticizer (POLYCAR-100), on the properties of CPB [17].e amount of superplasticizer for lignosulfonate, naphthalene sulfonate, and polycarboxylate was 7%, 6%, and 5.4% (by dry binder mass), respectively, to reach the designed slump (7 inches).
e authors reported that superplasticizers reduced water content (∼6.6%), enhanced mechanical properties (20-50%), and improved durability.Simon et al. studied the effects of polycarboxylate-based superplasticizer on flow characteristics and mechanical performance of CPB [18].It was found that the addition of 4% (by dry binder mass) polycarboxylate-based superplasticizer lowered the yield stress of fresh slurry from 1000 Pa to 3 Pa, but fluidity loss remains to occur over time.In addition, the compressive strength increased from 450 kPa to 1000 kPa.Mangane et al. investigated the mechanical performance and the workability of CPB containing various superplasticizers [14].e results showed that polycarboxylate had the best performance and achieved the target consistency at lower water content (between 6 and 10%) while still conserving the mechanical strength of CPB.Using polycarboxylate-based superplasticizer could also achieve a reduction in binder content (from 5 to 3%) without alerting the mechanical strength of CPB.Huynh et al. found that naphthalene sulfonate formaldehyde condensate (NSF) and polyphosphate-based superplasticizer affected the rheological properties of CPB with Portland cement by influencing the zeta potential of tailings and cement particles [19].Moreover, the flow performance of CPB mixed with blended Portland cement slag could be significantly enhanced by adding polycarboxylate-based superplasticizer as reported by Ouattara et al. [20].Although the above research on CPB has achieved important results, these results cannot be directly applied to CCWBM because CPB is substantially different from CCWBM in material composition [5].So, it is essential to acquire more information about the dosage of superplasticizer (SP), which relates to filling cost and the influence of superplasticizer on the flow characteristic and mechanical performance of CCWBM.
Dong et al. demonstrated that for high density filling slurry, pressure head loss was minimal when the slump of fresh slurry was 250 mm [21].erefore, the slump of all mixtures in the present study is set as 250 mm, and the dosage of superplasticizer is adjusted to reach target slump for all mixtures.e variation of superplasticizer dosage is studied when the mix proportion of CCWBM is changed.Fly ash content, fine gangue ratio, and mass concentration are the three key parameters which decide the mix composition of CCWBM, thus this paper studies SP through analyzing the influences of these three factors.
ough the method of single variable and orthogonal design is often used to analyze the effects of various factors on targets, test quantity is large and the interaction among various factors cannot be analyzed [22].Response surface methodology (RSM) is usually adopted to reflect the influence of the factors on the targets and their interaction with three-dimensional image, which is an integration method of mathematics and statistics.Since it came into being in 1951, RSM has been broadly put into use among agriculture, biology, chemistry, and other fields, yet rarely in backfill materials [23,24].In accordance with this, RSM is adopted to analyze the effects and their interaction with the three factors aforesaid.
In the present study, Design-Expert software is applied to discuss the influences of fly ash content, fine gangue ratio, and mass concentration on SP when target slump is 250 mm.An improved aggregate space model is introduced to explain the variation of SP. e research results provide a reference for subsequent research and practical production of CCWBM.

Raw Materials.
e investigated mixtures were systematically proportioned using coal gangue obtained from Xinyang Colliery, a mine in Shanxi Province of China.e samples were mechanically crushed and then categorized into two groups based on particle size, fine gangue aggregate with MSA of 5 mm and coarse gangue aggregate with MSA of 15 mm.e total gangue in each cubic filling slurry was 950 kg. e fineness modulus μ of the fine gangue aggregate was 3.02.A class F fly ash combined with ordinary Portland cement (CEM I 42.5N) (binary system) was used, and the integral dose of binder was 570 kg/m 3 .e main chemical components and physical properties of coal gangue, cement, and fly ash are summarized in Table 1 [7].A naphthalene sulphonate-based superplasticizer in powder was incorporated in all mixtures to reach target slump.

Test Methods.
e required workability of CCWBM was high fluidity and segregation resistance, which was similar to the workability requirement of self-consolidating concrete (SCC) [25].
e material composition of CCWBM in binders and aggregates was analogous to SCC too [26].
erefore, slump flow test was carried out to evaluate the fluidity of CCWBM according to ASTM C1611-M14 "Standard Test Method for Slump Flow of Self-Consolidating Concrete" [27].No superplasticizer was added in the first mixing of all mixtures.If slump was satisfied, SP of the mixture would be zero, or superplasticizer would be added until the slump was satisfied.Coarse adjustment is 10 g once, about 0.1% of binder, while fine adjustment is 5 g once, about 0.05% of binder.e adding method is selected according to the flow characteristics of fresh slurry.

Establishment of RSM Model
where Y was response; X i was independent variable; b i , b ii , and b ij denoted regression coefficient of first degree item, second degree item, and cross degree item, 2 Advances in Materials Science and Engineering respectively; k was the number of factors; and e was the error caused by test and regression [21,28].e dosage of superplasticizer (SP) was selected as response in this study.Fly ash content, fine gangue ratio, and mass concentration were the key parameters in preparation of CCWBM.eir variation affected the mix proportion of CCWBM, thus directly affecting SP.Consequently, fly ash content, fine gangue ratio, and mass concentration were chosen as function factors, and tests of three-factor-threelevel were designed.e levels of test factors are shown in Table 2.

Test Results.
Test plan and results are shown in Table 3.It was found that mass concentration had a significant influence on SP that SP significantly grew with the increase of mass concentration.When fine gangue ratio rose from 25% to 35%, SP hardly changed.SP tended to be larger with the increase of fly ash content.
ere was no remarkable bleeding in slump test for all mixtures because CCWBM contained a large amount of powder including fly ash with small particle size [29].

Fitting the Model.
Setting SP as response, the polynomial regression model expressed as equation ( 2) which relates all the variables is calculated using Design-Expert software according to equation (1): (2) Table 4 and 5 show the results of variances analysis for this model and significance tests for regression coefficients.
e probability p value gained from ANOVA is relatively low (p (model>F) < 0.0001) which means that the model is significant [24].e determination coefficient, R 2 , is 0.9671, indicating that the model can explain response value change of 96.7% and only 3.29% of response variance cannot be explained by this model [24,30].Furthermore, predicted value and actual value are mainly in a straight line in Figure 1 [22].All of above demonstrate that the model with a high precision can be used to SP analysis and prediction.

Analysis of Test Results
Figures 2-4 are contours and 3D images of the effects of fly ash content, fine gangue ratio, and mass concentration and their interactions on SP.   2, SP gradually becomes large with the growth of y ash content when the dosage of y ash is less than 390 kg/m 3 .An aggregate space model is introduced to explain SP change and adjusted to be reliable to some extent [30].As illustrated in Figure 5, water is the only owable ingredient in fresh lling slurry.Water molecules enwrap cement and y ash particles, forming a lubricating layer between them, which is the minimal unit of slurry ow and cement-y ash paste.e larger the space between powder particles (de ned as powder space), the better the uidity of cement-y ash paste [26].Cement-y ash paste surrounds ne gangue particles and forms a lubricating layer between them, which makes up the middle unit of slurry ow and mortar.e owability of mortar depends on the space between ne gangue particles (de ned as ne gangue space) and the uidity of cement-y ash paste.e better the uidity of cement-y ash paste, the lower the friction on ne gangue particles, the more excellent the owability of mortar.e uidity of mortar tends to be better with bigger ne gangue space.When the ne gangue space tends to in nity, the uidity of mortar will be close to the uidity of cement-y ash paste [30].Mortar enwraps coarse gangue particles and forms a lubricating layer between them, which forms the maximal unit of slurry ow. e owability of fresh slurry relies on the space between coarse gangue particles (de ned as coarse gangue space) and the uidity of mortar.e better the uidity of mortar, the lower the friction on coarse gangue particles, the more favourable the owability of fresh slurry.e uidity of fresh slurry tends to be better with larger coarse gangue space.When coarse gangue space tends to be in nite, the uidity of fresh slurry will be close to the uidity of mortar [30].However, when the powder space is too large, there is so much free water in mixtures that the initial setting time of slurry is prolonged and the dry shrinkage of back ll body is high, which damages the strength and durability of back ll body.When the ne gangue space is too big, the viscosity of mortar is small and thus the segregation of fresh slurry is serious.e volume of coarse gangue is small and the strength of back ll body is low with large coarse gangue space.erefore, lling slurry must enjoy moderate powder space, ne aggregate space, and coarse gangue space to gain favourable performance [29,31].
Because the particle size of y ash is smaller than that of cement [28,32], the amount of powder particles in mixtures increases with the growth of y ash content and powder space decreases, which leads to the reduction of the owability of cement-y ash paste.Naphthalene sulphonate-based superplasticizer can be adsorbed on the surface of cement and y ash particles and a ects the zeta potential of powder particles.is prevents cement occulation and reduces constructral water on the surface of powder particles [19,[33][34][35].As a result, free water in mixtures grows and its distribution is optimized.Meantime, a solvent layer is formed on the surface of powder particles, which plays a good lubricating role.e above two e ects increase powder space and therefore, with the growth of y ash content, SP gradually increases to reach the target slump.Nevertheless, the variation of SP is opposite when the dosage of y ash is greater than 390 kg/m 3 .Because the mass of cement decreases, y ash can play a better role in dispersing cement particles [36,37].Moreover, as the surface electrical properties of y ash is different from cement, cement occulation and bound water decrease.ese e ects ensure that mixtures have a moderate initial powder space, so SP becomes small.

E ect of Mass Concentration.
As can be seen from Figure 3, SP signi cantly grows with the increase of mass concentration.Because of the reduction of mixing water, powder space becomes small.It is necessary to enlarge the dosage of superplasticizer to reach large powder space.Advances in Materials Science and Engineering increase its volume.
e increase of cement-y ash paste volume can enlarge the ne aggregate space and thus improve the uidity of mortar.However, when mass concentration is 81%, the variation of SP is opposite because compared with mass concentration of 79%, the reduction of mixing water is signi cant.e increase of ne gangue volume will cause the increase of mortar volume and the reduction of coarse gangue, which will enlarge coarse gangue space.e variation of SP with ne gangue ratio is a ected by mass concentration, indicating that the interaction between ne gangue ratio and mass concentration is signi cant.

Rank of Signi cance.
e rank of the factors' signi cance for SP is mass concentration > y ash content > ne gangue ratio, and the curves on the 3D response surface change gradually from steep to gentle.Mass concentration a ects SP with steep curve on the 3D response surface, while the in uence of y ash content on it is weak with gentle curve.
is is also demonstrated in Table 4 where the p value of C is less than 0.0001 [22,36].4.5.Interaction.Contour shape can re ect the strength of the interaction e ect.e oval shape shows that the two factors interact signi cantly, while circular indicates the opposite [22].
erefore, the interaction between y ash content and ne gangue ratio is signi cant.However, the contour shape of BC is neither circular nor oval, and the 3D response surface is obviously distorted.In addition, the p value of BC is less than 0.0001 as shown in Table 4, so the interaction between ne gangue ratio and mass concentration is the most signi cant [37].

Conclusions
e SP model of CCWBM build in this paper is tting well and thereby could be applied to analyze and predict the dosage of superplasticizer of CCWBM.As the test results illustrate, SP rises signi cantly with the increase of mass concentration and tends to be larger with bigger y ash content.e e ect of sand ratio on SP is related to mass concentration.e rank of the factors' signi cance for SP is mass concentration > ne gangue ratio > y ash content, and the curves on the 3D response surface change gradually from steep to gentle.e interaction between mass concentration and ne gangue ratio is the most remarkable.Moreover, a reasonable explanation of the variation of SP is made through using the improved aggregate space model and thus it could be used to analyze the uidity of CCWBM.However, the SP model of CCWBM is built based on the results of slump ow test.As the properties of CCWBM have been greatly a ected by raw materials and material composition, the model in the present paper needs to be revised and improved.

Cement paste Cement Mortar
Coarse aggregate Fine aggregate Fly ash Water

3. 1 .
Design of Response Surface.According to the model of Design-Expert, a second-order polynomial model was selected to fit the response:

Figure 1 : 3 )Figure 3 :Figure 2 :Figure 4 :
Figure 1: Predicted vs. actual.e predicted value and actual value are mainly in a straight line.

Table 4 :
ANOVA of response surface quadratic model for SP.

Table 2 :
Levels of factors of RSM.

Table 3 :
Design of tests based on CCD and test results.
Advances in Materials Science and Engineering4.1.E ect of Fly Ash Content.As shown in Figure

Table 5 :
Signi cant test of the regression coe cients.