galetakis 2016

8/17/2019 galetakis 2016

1/8

Experimental investigation of the utilization of quarry dust for the

production of microcement-based building elements by self-flowing

molding casting

Michael Galetakis ⇑, Christina Piperidi, Anthoula Vasiliou, George Alevizos, Emmanouel Steiakakis,Konstantinos Komnitsas, Athanasia Soultana

School of Mineral Resources Engineering, Technical University of Crete, 73100 Chania, Greece

h i g h l i g h t s

Use of limestone dust for the production of self-flowing castable building elements.

Cement/filler ratio affects compressive strength and water absorption of specimens.

Microsilica/cement ratio plays an important role in water absorption values.

All specimens fulfill technical specifications for load bearing building elements.

Developed mix design methodology was proven reliable and practical.

a r t i c l e i n f o

Article history:

Received 30 November 2015

Received in revised form 7 January 2016

Accepted 8 January 2016

Available online 13 January 2016

Keywords:

Quarry dust

Microcement

Self-flowing

Mix design

Andreassen model

Box–Behnken factorial design

Response surface

a b s t r a c t

The management and disposal of fine by-products produced by the aggregate industry, ready-mix con-

crete and asphalt concrete installations (also known as filler or quarry dust), emerges as a major environ-

mental problem of quarrying and construction sector. Even though considerable research has been

undertaken for the utilization of this fine by-product in several applications, it still remains under-utilized, while its disposal and stabilization is also problematic due to resulting emissions of airborne

particle pollutants. In this study the production of self-flowing castable cement-based building elements

incorporating high amounts of quarry dust was investigated in laboratory scale. Quarry dust, microce-

ment, water and concrete admixtures were mixed and casted in steel molds for the production of the

specimens. The initial study of mixtures composition was based on the Andreassen particle packing

model, while the final mix design was determined via Box–Behnken fractional factorial design of exper-

iments, in combination with the response surface method. The compressive strength and water absorp-

tion values of hardened specimens exceed the relevant technical requirements currently in force,

regarding load-bearing as well as decorative building elements, thus opening a new promising field for

the utilization of this by-product.

2016 Elsevier Ltd. All rights reserved.

1. Introduction

The accumulation of fine quarry by-products (

8/17/2019 galetakis 2016

2/8

These fine quarry by-products remain largely unexploited,

while they could be an important source of raw material for the

construction industry in times of economic recession and strict

environmental regulations pressing the construction industry.

They could be recycled as they are, or after minor processing for

the manufacture of innovative products, prolonging significantly

the life of limestone reserves and existing quarries. Moreover, con-

sumption of large quantities of such a fine material will have eco-nomic and environmental benefits due to reduced costs for

disposal or discharge and additional revenues from the sale of

new products.

Based on the above, an experimental investigation of the uti-

lization of limestone fine quarry by-product (filler), produced still

in significant quantities from quarries of crushed aggregates in

the island of Crete (Greece), used as raw material for the produc-

tion of building elements, was carried out. In order to avoid com-

plex or energy intensive processes, the production of building

elements from castable (self-flowing) filler-cement mixtures with

the addition of chemical additives, used in concrete production,

was investigated. This process emerges as technically and econom-

ically attractive, having small energy requirements, being direct

applicable in the workplace of the quarry or concrete and/or

asphalt ready-mix plant, while at the same time facilitating the

consumption of the produced building elements by the construc-

tion industry. Direct implementation of such a recycling method

would supply the aggregate and building material industry sector

with a significant technological and competitive advantage, while

improving indicators related to environmental performance and

sustainability. The latter is particularly important for the quarrying

industry which in recent years has been experiencing strong social

pressure to improve its environmental performance [5].

However, the high volume use of quarry dust in cement-based

building products poses several challenges. The behavior of fresh

and hardened cement-based products depends highly on the

intrinsic properties of fines notably the so-called ‘filler effect’ thus

the use of these by-products requires a thorough characterization

in terms of the specifications which must be fulfilled (e.g. compo-sition and grading). Moreover, the addition of high quantities in

cement-based products results in high water demand, which

causes high drying shrinkage and inferior mechanical and physical

properties [6,7]. Therefore the development of sophisticated mix-

ture design methods ensuring optimal particle packing and the

use of suitable binders and chemical additives to reduce water

demand is of great interest.

This study aims to develop a mix design methodology for the

production of self-flowing microconcrete, consisting of quarry

dust, cement, water and concrete admixtures (additives), which

are suitable for the production of building elements. The mix

design of such type of concrete, known also as powder concrete,

has recently gained considerable attention due to its potential as

material for 3D printing inspired construction techniques that haverecently been developed at laboratory scale for cement-based

materials [8].

The structure of the paper is as below: In Section 2 the quarry

dust quality characteristics which are crucial in mixture design

are presented. The used microcement and admixtures are also

described. Moreover, the mixture design model as well as the

specimens’ preparation, curing and laboratory testing are

described. Emphasis is given on the experimental design by utiliz-

ing the Box–Behnken fractional factorial design. In Section 3 the

obtained results are discussed and analyzed according to the

standard procedure proposed by the factorial design of experi-

ments. The optimal mixture composition is also determined by

employing the response surface methodology. Finally the conclu-

sions and the suggestion for further investigation are given inSection 4.

2. Materials and methods

2.1. Filler characterization

Filler used in this study was collected from a dry mortar plant located in the

island of Crete (Greece). Filler is accumulated during drying and dedusting of mar-

ble sand, which is the main aggregate of dry mortars. The ultrafine fraction of mar-

ble sand (filler) is removed using an air stream, while bag filters are used for its

collection. Collected dust is then carried and disposed in large silos. Removal of fil-

ler from marble sand is imposed by the specifications regarding the grain size gra-dation of aggregates used in dry mortars. The amount of excess filler removed sums

up to a significant percentage of total marble sand (25%). From filler storage silos

representative samples were collected and their particle size distribution, their

specific surface area and their mineralogical–chemical composition were calcu-

lated. Particle size analysis was determined by a laser particle size analyzer (Mal-

vern Instruments, Mastersize-S), specific surface was estimated according to

Blaine method, while mineralogical analysis was carried out by using an X-ray

diffractometer (Siemens D500).

The grain size analysis of examined filler samples, shown in Fig. 1, indicated

that the removed excess fine material from the marble sand during drying and

dedusting processes is coarser than the conventionally defined filler, as described

by relevant European standards (

8/17/2019 galetakis 2016

3/8

2.3. Mixture design methodology

Mixtures design (calculation of the proportion of the ingredients) was per-

formed in two stages: the initial design stage and the final one. In the initial design

the proportions of the ingredients in the mixtures were estimated according to

Andreassen particle-packing optimization model. In the final stage the determina-

tion of the optimal composition was carried out by using the Box–Behnken frac-

tional factorial design of experiments, in combination with the response surfaces

method. The levels of experimentation in Box–Behnken factorial design were deter-

mined based on results obtained from the first stage.Particle packing optimization in concrete mixture design refers to the selectionof

therightgrainsizesand amounts (proportions) of theusedingredients.The particles

should be selected to fill up the voids between large particles with smaller particles

and so on, in order to increase the particle packing density. The role of packing of

aggregate particles in concrete performance was recognized many years ago [10–

12], butespeciallythe lastdecades,particlesize optimization hasgained newinterest

with the introduction of new types of concrete, such as high performance concrete,

self-compacting concrete, fiber reinforced concrete and ecological concrete [13].

The Andreassen particle-packing model as modified in 1980 by Funk and Dinger

[14] (it uses not only the maximum particle diameter, as the previous models do,

butalsothe minimum one) is considered themostsuitablefor thedesign ofmixtures

containing very fine materials, such as filler, microcement and silica fume.

The determination of mixture composition according to modified Andreassen

model, is based on the observed relationship between the packing density (packing

density is the solid volume of particles in a unit volume) of the grains of a cement-

bonded mixture, its rheology and the physical–mechanical properties of the hard-

ened product. Packing models assume that the grains of aggregates are arranged

upon each another, but due to their different size and to their irregular polyhedral

shape, inter-granular voids are formed. These voids are filled with a paste consisting

of a cement and water thus connecting the aggregate grains into a solid material.

For this reason, the amount of the paste should be at least equal to the inter-

granular spaces. Since the paste exhibits lower mechanical strength when hard-

ened, in comparison to stone aggregates, the fewer the voids to be filled by paste,

the greater the strength and density of the final cured product [15]. However,

low paste amounts affect negatively the rheology of the fresh mixture. In contrast,

high paste amounts result in lower mechanical and physical properties, however

the flowability–workability of the mixture is considerable increased. According to

the modified Andreassen particle-packing model the ideal particle size distribution

is given by Eq. (1):

P ðdÞ ¼ d

q d

qmin

dqmax

dqmin

ð1Þ

where P (d) Percentage of particles with diameter less than d (% passing), d particle

diameter, dmax maximum particle diameter in the mixture, dmin minimum particle

diameter in the mixture, q size distribution parameter.

Funk and Dinger [14] indicated that the value of q = 0.37 leads to the highest

packing density. In high packing density the grain size distribution of aggregates

used leads to a reduction of voids and consequently to lower requirements in

cement paste and thus to maximization of mechanical strength of the cured final

product. However, when high flowability of the fresh mixture is the dominant

design criterion, looser packing density of the final mixture is required. Thus, the

selection of the most appropriate value of q is a compromise between required

mechanical properties and flowability.

Adjusting mixture composition to fit a given Andreassen curve (i.e. size distri-

bution parameter q is predefined) is carried out by using commercial existing com-

puter programs. In this study we used the EMMA (Elkem Materials Mixture

Analyzer) software, developed by Elkem Materials (www.elkem.com/en/Concrete/

). EMMA uses the Andreassen model and allows us to design mixture of

optimally-packed materials. In combination with extensive laboratory testing

EMMA provides the basis for the accurate design of self-flowing castable, self-

consolidating and ultra-high strength concrete. The constraints considered duringthe estimation of the mixture proportions were:

The maximum microcement-to filler and microsilica-to-cement ratios was set

to 0.20 and to 0.15, respectively. The addition of higher amounts of microce-

ment and microsilica was considered uneconomic.

The resulting value of the q (Andreassen distribution parameter) of the exam-

ined mixtures should be as close as possible to optimal value of 0.37. Values

between 0.32 and 0.42 were considered acceptable.

Considering the above, several mixtures designs were created and theoretically

examined for optimal packing by using EMMA software. Typical grain size distribu-

tions of the resulting mixtures are shown in Fig. 2. The comparison of the obtained

size distributions with the optimal curve (as defined by the Andreassen model) was

done graphically, while the proportions of the mixture ingredients were estimated

by the trial and error method. After an extensive number of trials a series of mix-

tures, which fulfill the packing optimality criterion, were determined. The resulting

levels of cement-to-filler and microsilica-to-cement ratios in these mixtures werefound to range from 0.1 to 0.20 and 0.05 to 0.15, respectively.

Due to the assumptions and simplifications introduced in the Andreassen

model (e.g. particle characteristics like shape are not taken into account), its estima-

tions were considered approximate and therefore they were used in the initialdesign stage of mixture in this study. Furthermore, packing models cannot incorpo-

rate the effect of admixtures, such as the superplasticizers. However, it is known

that the addition of polycarboxylate-type superplasticizers influences the volume

of hydration phases and their spatial distribution. This effect results in changes of

the particle size distribution, specific surface area and number of particles and thus,

changes the rheological behavior of fresh cement pastes [16]. Thus an extensive

experimentation must follow after the theoretical adjusting of mix proportion by

the use of Andreassen model.

Consecutively, based on these initial results, the levels of control variables

(ingredients proportions and admixtures dosages) were selected for the determina-

tion of the optimum composition based on factorial design of experiments and

response surface methodology.

Factorial design of experiments is a scientific approach that allows an experi-

menter to make intentional changes to the inputs (control variables) of a process

or system to identify and observe the changes that occur to the response parame-

ters. Process is defined as a combination of materials, methods, people, environ-

ment, and measurements, which used together form a service, produce a product,or complete a task [17]. For the optimization of the experimental procedure the

fractional factorial design of experiments, by Box–Behnken, was used. This experi-

mental design requires three levels of experimentation for each control variable

(factor) and is widely used for the development of second-power models, when

using response surface methodology in optimization [18,19].

Response surface methodology (RSM) consists of a number of statistical and

mathematical tools used in the development of an appropriate functional relation-

ship between several inputs x i (referred as control parameters of a process) and a

response of interest y (referred as response variable). RSM is used to determine

the optimum settings of xi that result in the maximum or minimum response y over

a region of interest [19,20]. In RSM the relationship between the response y and the

control parameters xi is generally unknown but it can be modeled empirically. The

most commonly used models are the first- and second-degree polynomial models.

In the multiple response case, the determination of the optimum control

parameters that simultaneously maximize or minimize all the responses is a more

complicated problem. One of the most popular methods used in multi-response

optimization is the method based on the desirability function [19]. The method

finds control parameter values that provide the ‘‘most desirable” response values.

The basic idea of desirability function approach is to transform a multi-response

problem into a single response problem by means of mathematical transformations.

For each response yi, a desirability function di( yi) assigns numbers between 0 and 1

to the possible values of y i, with d i( yi) = 0 representing a completely undesirable

value of yi and di( yi) = 1 representing a completely desirable or ideal response value.

The overall objective function, referred to as total desirability D , is defined as the

geometric means of the n individual desirabilities di( yi). If some responses are con-

sidered to be more important than others, an impact in coefficient wi can be defined

for each response d i. In this case the total desirability is [21]:

D ¼ ðdw11 d

w22

. . .dwii . . .d

wnn Þ

1=

Xn

i¼1

wi

ð2Þ

When the number of response variables is small, instead of using the desirability

function, contour plots for each response variable may be overlaid to show the opti-

mum point graphically. For more details about desirability function and response

surface methodology the interested reader is referred to Myers and Montgomery[19].

1

10

100

0.1 1 10 100 1000

% P a

s s i n g

Particle size (μm)

q=0.32

q=0.42

No 4

No 15

No 7

Fig. 2. Size distributions of three selected mixtures (Nos. 4, 7 and 15). All created

mixtures have q values ranging from 0.32 to 0.42 (modified Andreassen model with

dmax = 200 lm and dmin = 0.1 lm).

M. Galetakis et al. / Construction and Building Materials 107 (2016) 247–254 249

http://www.elkem.com/en/Concrete/http://www.elkem.com/en/Concrete/http://-/?-http://-/?-http://-/?-

8/17/2019 galetakis 2016

4/8

In this study cement-to-filler ratio ( A), superplasticizer-to-filler and cement

ratio (B) and microsilica-to-cement ratio (C ), were considered as the control vari-

ables, while the water-to-cement ratio (W/C ) of the fresh mixture, the compressive

strength (Cs) and the water absorption (Wa) of hardened specimens (after curing for

28 days), were selected as the response variables. Each control variable was tested

in low, medium and high level, coded as 1, 0, 1 respectively. According to standard

Box–Behnken fractional factorial design this results in 12 combinations as shown

schematically in Fig. 3.

The selected levels of control variables are given in Table 1. The selection of

these levels was based on the results of the initial mixture design based on theAndreassen model and on economic criteria regarding the cost of the used materi-

als. For the calculation of the measurement error of the response variables (W/C , Cs

and Wa) the composition with the design parameters in the middle level (0) was

repeated three times (Mixture Nos. 6, 10 and 11). Thus a total of 15 mixtures were

prepared and tested. The resulting combinations (in randomized sequence) are

shown in Table 2.

The mixtures were prepared according to the proportions of Table 2 in a labo-

ratory mixer. The addition of water was done gradually until the mixture exhibited

the predefined flowability measured by the micro-cone spread test. During this

stage the water-to-cement was calculated. To avoid cement hardening during this

procedure a suitable amount (0.5% w/w of the cement) of a concrete retarder (tar-

taric acid) was added. The micro-cone spread test is a modified version of the flow

cone test EN11041, which is suitable for measuring the rheological ability of fresh

microconcrete [22]. The dimensions of the used cone as well as the spread test pro-

cedure are shown in Fig. 4.

The diameter of the resulting spread cone was set to 11.5 cm. This value was

found to ensure the required flowability for efficient molding casting (self-

flowing castable). Then fresh mixtures were poured in cubic molds

(50 50 50 mm3) and fabricated specimens were cured in a climatic chamber

with a constant temperature of 20 ± 1 C and humidity of at least 95%, for 28 days.

Six cubic specimens were prepared from each mixture. For each hardened specimen

the uniaxial compressive strength (according to ASTM C109) and the water absorp-

tion (according to EN1235/99) were measured. The results of laboratory tests for

each mixture (average of 6 measurements) are given in Table 2. The measurement

error was calculated for uniaxial compressive strength to 2.3 MPa, for water absorp-

tion to 0.22% w/w and for water-to-cement ratio to 0.015.

3. Analysis of results and discussion

To analyze factorial design experiments results the typical sug-

gested methodology was used [18,19]. The analysis was carried out

by using the STATGRAPHICS 5 Plus software package (StatPoint

Technologies, Inc., Warrenton, Virginia), [23]. The analysis

included:

– The calculation of main effects and interactions of factors on the

response variables in order to identify those which have a sig-

nificant effect on the response. The three-level Box–Behnken

fractional design was used to model possible curvature in the

response function.

– The modeling of the response variables by a second power poly-

nomial (non-linear regression).

– The identification of the optimum control parameters of the

process by using the response surface methodology.

3.1. Estimation of main effects and interactions

The estimated standardized values of main effects and interac-

tions of the examined factors on the compressive strength Cs are

given in the Pareto diagram shown in Fig. 5. Standardized values

were calculated by dividing the original value by the measurement

error. Standardized values, whose absolute value exceeded 1.96,

were regarded as statistically significant. This value (1.96) results

from the normal distribution at confidence level 95%. Results indi-

cate clearly the strong positive effect of cement-to-filler ratio (fac-

tor A) on Cs. In contrast, superplasticizer-to-microcement and filler

ratio (factor B) has a smaller effect on the compressive strength.

Finally, the effect of microsilica-to-cement ratio (factor C ) on Cs

is statistically insignificant. Concerning the estimated interactions

only AC and BC are statistically significant and have negative effect

on Cs.

As far as comparison of Cs values with the existing standards is

concerned, according to BS 3921, which defines the technical

requirements for the properties of load-bearing bricks, all tested

specimens exceed the requirements for compressive strength for

Class 2, while some of them meet the requirements of demanding

categories, such as class 7 (Table 3). All specimens also exceed the

requirements of EN 771-3 and ASTM C-90 regarding the compres-

sive strength for load-bearing bricks (>12 MPa).

Regarding the water absorption Wa (Fig. 5), the factor A(microcement-to-filler ratio) was proven to have the greatest

impact on Wa. Then follow factor C (microsilica-to-microcement

ratio), interaction AC and interaction BC holding a slightly lower

value. Interaction AB and factor B appear to have no statistically

significant impact, as they are both below the threshold value of

1.96 (vertical line). Concerning the comparison of Wa with the

existing standards all values are below to 10% which is considered

as the maximum allowable water absorption for load-bearing

bricks (EN 771-3 and ASTM C-90).

Finally, water-to-cement ratio W/C was found to be mainly

affected by the cement-to-filler ratio (factor A). Interactions AC ,

AB, BC as well as factor B have significantly less impact on W/C .

The effect of factor C is insignificant since it is below the threshold

value.The variations of Cs, Wa and W/C (original values) when factors

A, B and C change from the low to the high level are shown in Fig. 6.

It can be readily observed that all three factors exhibit nonlinear

correlation with the response variables Cs, Wa and W/C . Factor A

has the strongest impact on all responses. From the slope of the

curve it is evident that the greater improvement (increase of Cs

and reduction of Wa) occurs during this factor’s shift from low to

medium level. Factor C has a negligible effect on Cs and W/C but

a significant one on Wa, while factor B seems to affect negatively

Cs and having minor effect on Wa and W/C .

The above result leads us to recognize that the microcement-

to-filler ratio is the most important factor with the regard to

compressive strength as well as to water absorption. By

increasing this ratio Cs is increased, while at the same time Wais also decreased, which is highly desirable. The increase of

Fig. 3. Schematic representation of Box–Behnken factorial design for three controlvariables ( A, B, C ).

Table 1

Experimental design parameters and selected levels.

Control variable Symbol Level

Low

(1)

Middle

(0)

High

(1)

Cement-to-filler ratio A 0.10 0.15 0.20

Superplasticizer-to-cement and

filler ratio

B 0.004 0.006 0.008

Microsilica-to-cement ratio C 0.05 0.10 0.15

250 M. Galetakis et al. / Construction and Building Materials 107 (2016) 247–254

8/17/2019 galetakis 2016

5/8

microsilica-to-cement ratio contributes to the decrease of water

absorption; however its effect on Cs is slightly negative due to

the observed strong interactions with factors A and B. Finally,

the increase of superplasticizer-to-microcement and filler ratio

(factor B) results in decrease of Cs, while its effect on Wa is

insignificant.

Table 2

Factorial design by Box–Behnken and test results for the hardened specimens.

Mixture

no

Control variables Response parameters

Cement-to-filler

ratio A

Supeplasticizer-to-filler and

cement ratio B

Microsilica-to-cement

ratio C

Water-to-cement

ratio, W /C

Compressive strength

Cs, MPa

Water absorption

Wa, %

1 0.15 0.008 0.05 1.03 29.7 6.45

2 0.20 0.006 0.05 0.79 49.2 7.31

3 0.10 0.006 0.05 1.65 15.7 9.11

4 0.15 0.004 0.15 1.07 50.5 4.55

5 0.20 0.008 0.10 0.90 41.7 4.17

6 0.15 0.006 0.10 1.03 31.5 5.85

7 0.10 0.004 0.10 1.70 25.6 9.09

8 0.10 0.008 0.10 1.60 17.3 9.19

9 0.20 0.006 0.15 0.95 22.3 2.10

10 0.15 0.006 0.10 1.05 27.2 6.34

11 0.15 0.006 0.10 1.02 32.5 5.94

12 0.15 0.008 0.15 1.02 31.9 5.88

13 0.20 0.004 0.10 0.85 46.3 4.93

14 0.15 0.004 0.05 1.03 35.6 7.02

15 0.10 0.006 0.15 1.50 22.7 7.54

21

40

1mm

0mm

60mmm

Fig. 4. Left: dimensions of used cone, middle: cone filled with fresh mixture, right: spread of mixture and measurement of its diameter.

Effect on Cs (standardized values)

0 2 4 6 8 10 12

C

AB

BC

B

AC

A

+-

Effect on Wa (standardized values)

0 4 8 12 16 20

B

AB

BC

AC

C

A

+-

24

Effect on W/C (standardized values)0 20 40 60 80

C

BC

B

AB

AC

A

+

-

Fig. 5. Pareto diagram showing the main effects of design factors A, B and C , as well as their interactions, on the standardized values of compressive strength Cs , water

absorption Wa and water-to-cement ratio W/C ( A = cement-to-filler ratio, B = superplasticizer-to-cement and filler ratio and C = microsilica-to-cement ratio).

Table 3

Minimum required compressive strength for load-bearing bricks according to BS 3921 and specimens’ classification.

Class of load-bearing bricks 1 2 3 4 5 7 10 15

Min. required compressive strength, MPa 7.0 14.0 20.5 27.5 34.5 48.5 69.0 103.5

Number of specimens within class 0 2 4 4 3 2 0 0

Number of specimens exceeding class 15 15 13 9 5 2 0 0


http://-/?-http://-/?-http://-/?-http://-/?-

8/17/2019 galetakis 2016

6/8

3.2. Modeling and optimization

Subsequently the coefficients of a second degree multiple

regression model, were calculated for each response variable.

The estimated second degree polynomial models are given in

Eqs. (3)–(5).

Cs ¼ 22:3þ 906:5 A 1821Bþ 641C 1425 A2 þ 9250 AB

3390 AC þ 1472000B

2

31750BC þ 255:0C

2

ð3Þ

Wa ¼ 16:49 72:25 A 538:8Bþ 17:48C þ 268:3 A2

2150 AB 364:0 AC þ 32710B2 þ 4750BC 79:67C 2 ð4Þ

W =C ¼ 4:7 36:6 A 91:25B 1:7C þ 79:5 A2 þ 375:0 AB

þ 31:0 AC þ 3437:5B2 125:0BC 10:5C 2 ð5Þ

where: A 9 [ 0.1 0.2], B 9 [0.004 0.008] and C 9 [0.05 0.15].

The correlation coefficients R between the observed and pre-

dicted values of the responses were estimated to 0.93, 0.97 and

0.98 for Cs, Wa and W/C respectively. These values indicated that

there is a reasonable agreement between predicted and observed

values and can be concluded that the fitted polynomial modelsdescribe adequately the influence of the selected factors to the

responses under study. The estimated second degree polynomial

models, shown in Eqs. (3)–(5), are valid only when the design fac-

tors are within the range of experimentation.

After that the individual desirability function for each response

variable was estimated considering the requirements for the phys-

ical and mechanical properties of load-bearing structural elements

according to EN 771-3 and ASTM C-90. The simultaneous optimiza-

tion of response variables (maximize Cs and minimize Wa) is

achieved by maximizing the overall desirability function. Follow-

ing this methodology it is proven, as shown in Fig. 7, that the com-

bination of factorial levels which maximizes compressive strength,

0.1B

0.008 0.15

C s , M P a

17

21

25

29

33

37

41

A0.2 0.004

C0.05 0.1

B0.008 0.15

W a , %

4.6

5.6

6.6

7.6

8.6

9.6

A0.2 0.004

C0.05 0.1

B0.008 0.15

w a t e r - t o - c e m e n t r a t i o

0.87

1.07

1.27

1.47

1.67

A0.2 0.004

C0.05

Fig. 6. Change of compressive strength, Cs, and water absorption, Wa, when the design factors A, B and C shifts from low to high level ( A = cement-to-filler ratio,

B = superplasticizer-to-cement and filler ratio and C = microsilica-to-cement ratio).

Cement-to-filler ratio

Microsilica-to-cement ratio

Superplasticizer-to-cement and filler ratio=0.004

0.1 0.12 0.14 0.16 0.18 0.2 0.05

0.070.09

0.110.13

0.15

0

0.2

0.4

0.6

0.8

1

D e s i r a b i l i t y

Fig. 7. Plot of desirability optimization function (simultaneous maximization of Cs and minimization of Wa).

20

25

3

4

5

6

7

8

9

30

35

40

40

M i c r o s i l i c a - t o - c e m e n t r a t i o

45


Suprplasticizer-to-cement and filler ratio=0.04

Cs

Wa

0.1 0.12 0.14 0.16 0.18 0.2

0.05

0.07

0.09

0.11

0.13

0.15

0.152

0.072

0.004

Fig. 8. Determination of cement-filler and microsilica-to-cement ratios (points

shown as circles) for Cs = 40 MPa and Wa = 7%.

Superplasticizer-to-cement and filler ratio=0.004


M i c r o s i l i c a - t o c e m e n t r a t i o

0.80

0.85

0.90

W/C=0.75

1 . 0

1 . 0

5

1 . 1

0

1 . 2

0

1 . 4

0

0.1 0.12 0.14 0.16 0.18 0.2

0.05

0.07

0.09

0.11

0.13

0.15

1 . 6

0

1.070.072

0.152

Fig. 9. Determination of water-to-cement ratio given that cement-to-filler

ratio = 0.152 and Cs = 40 MPa and Wa = 7%.


8/17/2019 galetakis 2016

7/8

while at the same time minimizes water absorption is attained

when the microcement/filler and the silica fume/microcement

ratios are at their highest levels, while the superplasticizer/(micro-

cement and filler) ratio is at its lower level.

The obtained values for Cs and Wa were 40.46 MPa and 2.18%

respectively, while the reached desirability value is 0.97 (the

impact of Cs and Wa on the overall desirability was considered

identical).

In addition to optimization, Eqs. (3)–(5) allow us to the design

mixtures with predefined values of compressive strength and

water absorption. This is particularly useful since the producedself-flowing castable building elements must fulfill specific stan-

dards’ requirements related to the proposed use. An illustrative

example indicating the way to design a self-flowing castable mix

with resulting Cs = 40 MPa and Wa = 7% is shown in Figs. 8 and 9.

Fig. 8 shows the overlaid contours of Cs and Wa derived from

Eqs. (3) and (4), while in Fig. 9 the contours of water-to-cement

ratio (Eq. (5)) are given. Superplasticizer-to-filler and cement ratio

was kept at the low level (0.004) since it was found to have negli-

gible effect on Cs and Wa.

First the points which fulfills the requirements Cs = 40 MPa

and Wa = 7% were identified. In this case three such points

(shown as small circles in Fig. 8), representing different mix

designs, were found. The selection of the most appropriate mix

design among these is usually based on the resulting materials’cost. In this example the mix design with cement-to-filler

ratio = 0.152 and microsilica-to-cement ratio = 0.072 was arbi-

trary selected. Next the water-to-cement ratio was estimated

(1.07), as shown in Fig. 9.

Forthe validation of the above methodthreeadditional mixtures

(shown in Table 4) with Cs = 45 MPa, Wa = 7%, Cs = 40 MPa, Wa = 7%

and Cs = 35 M, Wa = 8% wereselected. Their ingredients’ proportions

were estimated similarly to the presented example and finally the

properties of the produced specimens(after 28 days of curing) were

determined. Results of the laboratory tests, shown in Table 4, indi-

cate that there is a reasonable agreement between the predicted

(design) and the obtained values from the laboratory tests.

4. Conclusions

The experimental investigation of the usability of limestone fil-

ler for the production of building elements by free flow casting

indicates that:

– The main factor that affects the uniaxial compressive strength

(28 days) and water absorption of the prepared specimen is

the cement-to-filler ratio. Microsilica-to-cement seems to play

an important role in water absorption values. The interactions

of cement-to-filler and microsilica-to-cement and filler were

also proven important.

– According to technical specifications regarding load bearing

building elements, specimens of all examined compositions ful-

fill the requirements for compressive strength and waterabsorption.

– The developed mix design methodology, based on Response

Surface Analysis, was proven reliable and practical.

In order to make a complete investigation of the prepared spec-

imens with regard to their suitability for production of building

elements, additional properties should be also considered. Long

term durability tests such us the accelerated weathering tests

and/or the resistance to high or low temperature environment

(freeze–thaw and thermal shock test) should be carried out.

Furthermore, in order to evaluate the proposed technology,

specimens of larger dimensions, resembling them of market prod-ucts of the building industry should be produced, while the use of

high frequency mixing as a method of improving dispersion of

microsilica and ultrafine cement grains among aggregates should

also be investigated.

Acknowledgments

This work has been performed under the framework of the

‘‘Cooperation 2011” project DURECOBEL (11_SYN_8_584) funded

from the Operational Program ‘‘Competitiveness and Entrepreneur-

ship” (co-funded by the European Regional Development Fund

(ERDF)) and managed by the Greek General Secretariat for Research

and Technology.

Additionally, we would like to thank Cementa AB (Heidelberg

Cement Group) and especially Mr. Mikael Dellhammar, for provid-

ing us the ‘‘Ultrafine 12” microcement sample used in this

investigation.

References

[1] R. Hill, R. Dawson, M. Mundy, Utilisation of aggregate materials in roadconstruction and bulk fill, Resour. Conserv. Recycl. 32 (2001) 305–320.

[2] D.J. Harrison, From waste to wealth: developing saleable mineral productsfrom quarry waste, in: R.K. Dhir, M.D. Newlands, J.E. Halliday (Eds.), Recyclingand Reuse of Waste Materials, 2003, pp. 281–286.

[3] M. Galetakis, G. Alevizos, K. Leventakis, Evaluation of fine limestone quarry by-products, for the production of building elements – an experimental approach,Constr. Build. Mater. 26 (2012) 122–130.

[4] M. Galetakis, V. Zourbakis, I. Koinakis, K. Leventakis, G. Alevizos, Aggregateproduction on the island of Crete, Greece: current production conditions andfuture perspective, Bull. Geol. Soc. Greece XLVII (3) (2013) 2030–2039 .

[5] N. Almeida, F. Branco, J. DeBrito, J.R. Santos, High-performance concrete withrecycled stone slurry, Cem. Concr. Res. 37 (2007) 210–220.

[6] L. Courard, F. Michel, J. Pierard, Influence of clay in limestone fillers for self-compacting cement based composites, Constr. Build Mater. 25 (2011) 1356–1361.

[7] S.D. Charkha, Experimental investigation of M30 design mix concrete withpartial replacement of conventional ingredients, Int. J. Res. Civ. Eng. Archit.Des. 1 (2) (2013) 38–45.

[8] A. Perrot, D. Rangeard, A. Pierre, Structural built-up of cement-based materialsused for 3D printing extrusion techniques, Mater. Struct. (2015), http://dx.doi.org/10.1617/s11527-015-0571-0.

[9] A. Goldman, A. Bentur, The influence of microfillers on enhancement of concrete strength, Cem. Concr. Res. 23 (4) (1993) 962–972.

[10] W.B. Fuller, S.E. Thompson, The laws of proportioning concrete, Am. Soc. Civ.Eng. J. Transp. 59 (1907) 67–143.

[11] A.N. Talbot, F.E. Richart, The strength of concrete – its relation to the cement,

aggregate and water, Bulletin No. 137, Engineering Experiment Station,University of Illinois, Urbana, 1923.

Table 4

Design parameters and obtained results from the laboratory tests of the validation mixtures.

Mixture no Design parameters Estimated mixture proportions (w/w) Obtained values

from laboratory

tests

Cs, MPa Wa,% Microcone spread (cm) W/C C/F M/C Cs Wa

16 35 8 11.5 1.58 0.105 0.125 33.7 8.50

17 40 7 11.5 1.07 0.152 0.072 39.5 7.33

18 45 7 11.5 0.79 0.193 0.059 46.1 7.44


http://refhub.elsevier.com/S0950-0618(16)30015-0/h0005http://refhub.elsevier.com/S0950-0618(16)30015-0/h0005http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0025http://refhub.elsevier.com/S0950-0618(16)30015-0/h0025http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://dx.doi.org/10.1617/s11527-015-0571-0http://dx.doi.org/10.1617/s11527-015-0571-0http://refhub.elsevier.com/S0950-0618(16)30015-0/h0045http://refhub.elsevier.com/S0950-0618(16)30015-0/h0045http://refhub.elsevier.com/S0950-0618(16)30015-0/h0045http://refhub.elsevier.com/S0950-0618(16)30015-0/h0050http://refhub.elsevier.com/S0950-0618(16)30015-0/h0050http://-/?-http://refhub.elsevier.com/S0950-0618(16)30015-0/h0050http://refhub.elsevier.com/S0950-0618(16)30015-0/h0050http://refhub.elsevier.com/S0950-0618(16)30015-0/h0045http://refhub.elsevier.com/S0950-0618(16)30015-0/h0045http://dx.doi.org/10.1617/s11527-015-0571-0http://dx.doi.org/10.1617/s11527-015-0571-0http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0035http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0030http://refhub.elsevier.com/S0950-0618(16)30015-0/h0025http://refhub.elsevier.com/S0950-0618(16)30015-0/h0025http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0020http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0015http://refhub.elsevier.com/S0950-0618(16)30015-0/h0005http://refhub.elsevier.com/S0950-0618(16)30015-0/h0005http://-/?-

8/17/2019 galetakis 2016

8/8

[12] A.H.M. Andreasen, J. Andersen, Ueber die Beziehung zwischen Kornabstufungund Zwischenraum in Produkten aus losen Körnern (mit einigenExperimenten), Kolloid-Zeitschrift 50 (1930) 217–228.

[13] S.A.A.M. Fennis, J.C. Walraven, J.A. Uijl, Defined-performance design of ecological concrete, Mater. Struct. (2012), http://dx.doi.org/10.1617/s11527-012-9922-2.

[14] J.E. Funk, D.R. Dinger, J.E. Funk, Jr., Coal grinding and particle size distributionstudies for coal-water slurries at high solids content, Final Report, EmpireState Electric Energy Research Corporation (ESEERCO), New York, 1980.

[15] Fung. Kwan, Packing Density Measurement and Modeling of Fine Aggregate

and Mortar, Department of Civil Engineering, University Hong Kong, China,2009. pp. 349–356.

[16] A. Zingg, L. Holzer, A. Kaech, F. Winnefeld, J. Pakusch, S. Becker, L. Gauckler, Themicrostructure of dispersed and non-dispersed fresh cement pastes – newinsight by cryo-microscopy, Cem. Concr. Res. 38 (2008) 522–529.

[17] S.R. Schmidt, R.G. Launsby, Understanding Industrial Designed Experiments,fourth ed., Air Academy Press, Colorado Springs, 1998.

[18] R.H. Myers, ‘‘Response surface methodology – current status and futuredirections”. (with discussion), J. Qual. Technol. 31 (1999) 30–44.

[19] R.H. Myers, D.C. Montgomery, Response Surface Methodology: Product andProcess Optimization Using Designed Experiments, second ed., John Wiley &Sons, New York, 2002.

[20] G.E.P. Box, N.R. Draper, Empirical Model-Building and Response Surfaces, JohnWiley & Sons, New York, 1987.

[21] G. Derringer, R. Suich, Simultaneous optimization of several response

variables, J. Qual. Technol. 12 (4) (1980) 214–219.[22] B. Felekoglou, Effects of PSD and surface morphology of micro-aggregates on

admixture requirement and mechanical performance of micro-concrete,Cement Concr. Compos. 29 (6) (2007) 481–489.

[23] Manguistics Inc., Statgraphics 5 Plus. User’s Guide, 2005.

http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060http://dx.doi.org/10.1617/s11527-012-9922-2http://dx.doi.org/10.1617/s11527-012-9922-2http://dx.doi.org/10.1617/s11527-012-9922-2http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0085http://refhub.elsevier.com/S0950-0618(16)30015-0/h0085http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0100http://refhub.elsevier.com/S0950-0618(16)30015-0/h0100http://refhub.elsevier.com/S0950-0618(16)30015-0/h0105http://refhub.elsevier.com/S0950-0618(16)30015-0/h0105http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0110http://refhub.elsevier.com/S0950-0618(16)30015-0/h0105http://refhub.elsevier.com/S0950-0618(16)30015-0/h0105http://refhub.elsevier.com/S0950-0618(16)30015-0/h0100http://refhub.elsevier.com/S0950-0618(16)30015-0/h0100http://refhub.elsevier.com/S0950-0618(16)30015-0/h0100http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0095http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0090http://refhub.elsevier.com/S0950-0618(16)30015-0/h0085http://refhub.elsevier.com/S0950-0618(16)30015-0/h0085http://refhub.elsevier.com/S0950-0618(16)30015-0/h0085http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0080http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://refhub.elsevier.com/S0950-0618(16)30015-0/h0075http://dx.doi.org/10.1617/s11527-012-9922-2http://dx.doi.org/10.1617/s11527-012-9922-2http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060http://refhub.elsevier.com/S0950-0618(16)30015-0/h0060

galetakis 2016

Documents