Abstract

A central composite design models was carried out to model the influence of key mixture parameter on fresh and hardened properties affecting the performance of Self-Compacting Concrete (SCC). Such responses included slump flow as a filing ability, L-Box as a passing ability and sieve analysis as resistance to segregation. Thirty-one mixtures were prepared to derive the numerical models and evaluate the accuracy. The models are valid for a wide range of mixture proportioning. The study presents, the derived numerical models that can be useful to reduce the test procedures and trials needed for the proportioning of self-compacting concrete. The qualities of these models were evaluated based on several factors such as level prediction, residual error, residual mean square and correlation coefficients.

---

INTRODUCTION

Sonebi (2004) in his study investigated and developed medium strength SCC by using pulverized fuel ash and (SP). In his analysis used factorial design to get mathematical models on five parameters. The parameters were cement, PFA (W/P) and SP. Also, the responses of the statistical models are slump flow, fluidity, loss orimet time, V-funnel time, L = box, JRing combined to the orimet, JRing combined to the cone, rheological parameter, segregation and compressive strength at 7, 28 and 90 days. The developed models were valid with 0.38-0.72 W/P, 60-216 kg m^-3 of cement, 183-317 kg m^-3 of PFA and 0-1% by mass of powder. The results showed that MS-SCC at 28 day compressive strength can achieve to 30-35 MPa by using 211 kg m^-3.

Prasad et al. (2008) were used Artificial Neural Network (ANN) to predict a 28 days compressive strength on SCC and normal concrete and addition of high fly-ash and they found that the proposed ANN are validated by estimation of slump flow and compressive strength of SCC.

Felekoglu et al. (2007) they are focusing on the development of mix design of SCC and making adjustment to W/C ratio and the dosage of SP. The parameters are five mixtures were investigated, the responses are slump flow, V-funnel and L-Box was carried to determine optimum parameters for SCC mixtures. The hardened properties of SCC were studied also.

Domone (2007) in his studied 70 paper on hardened properties of Self Compacting Concrete (SCC) the data were analyzed and relations between the cylinder and cubic compressive and tensile strengths and modulus of elasticity.

The conclusions were addition of limestone powder makes a substantial contribution to strength and the performance of structural elements is largely predicted by properties of materials measured and analysis of the data were enough to give confidence for SCC.

Laboratory test, such as the Slump flow, Sieve Size, L-Box and 28 days compressive strength tests were conducted using indigenous materials Malaysia.

MATERIALS AND METHODS

The materials that implemented in the research are:

Cement: Ordinary Portland cement of available in local market is used in the investigation. The Cement used has been tested for various proportions as per (ASTM C150-85A) (ASTM Standard C 150, 2006) the specific gravity was 3.15 and fineness was 2091cm² gm^-1.

Coarse and fine aggregate: coarse aggregate: Crushed angular granite material of 20 mm max size from a local source was used as course aggregate. The specific gravity of 2.45, absorption value was 1.5%, fineness modulus 6.05 and bulk density of 1480 kg m^-3 confirms to ASTM C 33-86 (ASTM C 33-03, 2006) was used.

The fine aggregates consisted of river sand with maximum size of 4.75 mm, with a modulus of fineness Mx = 4.16; normal grading. Specific gravity was 2.33 and absorption value was 6.4%.

Fly Ashes (FA): Type-II fly ash from Kapar Thermal Power Station, Selangor, Malaysia, was used as cement replacement material. Fly Ash for use as Pozzolana and Admixture. Class F fly ash was obtained had a specific gravity of 2.323 and fineness of 2423 cm² g^-1 determined as confirms to (ASTM C 618) (ASTM C 618-05, 2006).

Superplasticizer (SP): Polycarboxylicether (PCE) based super-plasticizer, which is Brown Color and free flowing liquid and having Relative density 1.15 Super Plasticizer confirms to ASTM C 494-92 (ASTM C494/C494M-05a, 2006). Type A and Type F in aqueous form to enhance workability and water retention. A sulfonated, naphthalene-formaldehyde super plasticizer and a synthetic resin type Air-Entraining Admixture (AEA) were used in all the concrete mixtures.

All concrete mixes were prepared in 40 L batches in a rotating planetary mixer. The batching sequence consisted of homogenizing the sand and coarse aggregate for 30 sec, then adding about half of the mixing water into the mixer and continuing to mix for one more minute. The mixer was covered with plastic cover to minimize the evaporation of the mixing water and to let the dry aggregates in the mixer absorb the water. After 5 min, the cement and fly ash were added and mixed for another minute. Finally, the SP and the remaining water were introduced and the concrete was mixed for 3 min.

Slump flow, L-Box, V-funnel were used to test the workability, passing ability of SCC. With the L-Box, the height of concrete in the vertical part, after the flowing of concrete, was considered in the analysis of the results. The resistance to segregation was measured by sieve size, a fresh concrete was poured from 2 kg panel over sieve number 5 to observe the quantity of concrete passing the sieve after 5 min (Sonebi, 2004). Nine 100x100x100 mm cubic were cast and moist for each mix to determine compressive strength after 3, 7 and 28 days.

Development of statistical models: Statistical experimental design of four factors at two levels was used to evaluate the influence of 2 different levels for each variable on the relevant concrete properties. Such 2 level factorial design requires a minimum number of tests for each variable (Montgomery, 2005). The fact that the expected responses do not vary in a linear manner with the selected variable and to enable the quantification of the prediction of the responses, a central composite plan was selected, where the response could be modeled in a quadratic manner.

 

Table 1:	Value of coded variables

 

Since, the error in predicting the responses increases with the distance from the centre of the modeled region, it is advisable to limit the use of the models to an area bound by values corresponding to -α to +α limits.

The parameters were carefully selected to carry out composite factorial design, where the effect of each factor is evaluated at five different levels, in codified values of -α, -1, 0, 1, +α. The value of α value is chosen so that the variance of the response predict by the model would depend only on the distance from the centre of the modeled region. The value of α value is taken here as ±2. Seven replicate central points were prepared to estimate the degree of experimental error for the modeled responses as shown in Table 1. Appropriate MiniTab software was used for statistical analysis of the results (MINITAB Handbook, 2003).

Four key parameters that can have significant influence on the mix characteristics of SCC were selected to derive the mathematical models for evaluating relevant properties. The experimental levels of the variables (maximum and minimum), boundary of cement content, W/P, fly-ash content, Sp dosage are defined. The modeled experimental region consisted of mixes ranging between the coded variables of -2 to +2 and is given in Table 1. The derived statistical models are valid for mixes with W/P ranging from 0.3-0.38 by mass, dosages of SP ranging from 7.2-10.8 kg m^-3 1.8% of total powder content (by mass) (Su et al., 2001), cement content ranging from 400-450 kg m^-3. The mass of coarse aggregate was 25-35% by volume of the mix. The SCC responses modeled were slump flow, L-Box ratio, segregation resistance and 28 day compressive strengths (EFNARC, 2002).

RESULTS AND DISCUSSION

Derived models: The mix proportions and test result of 31 mixes prepared to derive the central composite surface design models are summarized in Table 2 and 3, respectively. The result of the derived models in this research is prepared, along with the correlation coefficients and the relative significance. The estimates for each parameter refer to the coefficients of the model found by a least-square method. The significant of each variable on a given response is evaluated using t-test values based on Student’s distribution. Probabilities <0.05 are often considered as significant evidence that the parameters are not equal to zero; contribution of the proposed parameter has a highly significant influence on the measured response. The R² values of the response surface models for the slump flow, L-Box, resistance to segregation and 28 days f_c are 80.8, 55.9, 60.6 and 82.3% in full quadratic equation with respect to the linear, interaction and pure quadratic. The high correlation coefficient of the response shows good correlations that considered at least 95% of the measured values can be accounted for proposed models.

The accuracy of the proposed models was determined by comparing predicted to measured values. Table 2

shows, the mix proportions and properties of fresh and hardened SCC of all mixes used in the central composite design.

Numerical models for slump flow (mm) y1, compressive strength (Mpa) Y11, L-Box (ratio) y4, segregation resistance (%) y8: The linear, interaction, full and pure quadratic models for slump flow, compressive strength at 28 days, Lox and resistance to segregation are shown in Table 3 with values of R², R²-adj., F and p-values and Lowe and upper values of Residuals. As shown in the Fig. 1, the residuals of slump flow linear model plot versus the run order fluctuate in a random pattern around the center line.

 

Table 2:	The mix proportions and properties of fresh and hardened SCC of all mixes used in the central composite design

 

 

Table 3:	Statistical models of slump flow, compressive strength at 28 days, L-Box and resistance to segregation summary

 

 

Table 3:	Continued

 

 

Fig. 1:	Residual of workability (slump flow, L-Box and segregation resistance) and compressive strength at 28 days for various statistical model

 

The range of residual varies between -106.4 and 99.9.Also, for interaction model -103.3 to 99.9, full quadratic model -87.5 to 89.17 and pure quadratic -83.7 to 94.9. Linear model show high residual range which indicates a poor fit. The adjusted correlation coefficient a 80.8% and the adjusted R²64.1 that fits the data full quadratic model with a significant p-value approach to zero better than other models.

Moreover, the residuals of compressive strength linear model plot versus the run order fluctuate in a random pattern around the center line. The range of residual varies between -4.043 and 5.097 for linear. Also, for interaction model -4.062 to 6.320, full quadratic model -3.055 to 2.869 and pure quadratic -4.813 to 1.649. Interaction model show high residual range which indicates a poor fit as shown in Fig. 1. The adjusted correlation coefficient and 82.3% and the adjusted correlation coefficient 66.9% that fits the data of model full quadratic with a significance value that is very close to zero, Fig. 2, which is more significant and better than other models.

In addition, the residuals of L-Box linear model plot versus the run order fluctuate in a random pattern around the center line.

The range of residual varies between -0.7935 and 0.6724 for linear. Also, for interaction model -0.7935 to 0.3959, full quadratic model -0.4273 to 0.3515 and pure quadratic -0.7365 to 0.7365. Pure quadratic model show high residual range, which indicates a poor fit as shown in Fig. 1.

The adjusted correlation coefficient a 55.9% and the adjusted correlation coefficient 17.3% that fits the data of model full quadratic with a significance value that is close to zero than other values in Fig. 2, which is more significant and better than other models.

Furthermore, the residuals of segregation resistance linear model plot versus the run order fluctuate in a random pattern around the center line. The range of residual varies between -18.15 and 18.15 for linear. Also, for interaction model -17.11 to 18.15, full quadratic model -15.75 to 14.17 and pure quadratic -17.31 to 14.17. Linear model show high residual range which indicates a poor fit.

 

Fig. 2:	Comparison between measured workability and strength and predicted value from statistical models

 

The adjusted correlation coefficient a 60.6% and the adjusted correlation coefficient 26.2%, that fits the data of model full quadratic with a significance value that is close to zero than other values in Fig. 2, which is more significant and better than other models.

CONCLUSION

The effect of the concrete constituents such as cement, water-powder ratio, fly-ash and super-plasticizer on workability of concrete and compressive strength were investigated based on the result of this research the following conclusions can be drawn:

•	A central composite design is a useful tools to evaluate parameters effects of mixture and the interaction between the parameters on SCC that can reduce the number of trials to achieve balance among mix variables

•	Numerical models established for the SCC mixtures can be useful in design of concrete and selecting constituent materials

•	Central composite was selected where, the response modeled in a quadratic manner, while seven replicate central points were prepared to estimate the degree of experimental error response model

•	Graphical analysis of the residuals shows the deviation between the measured data and the fit one could be effective methods to test the adequacy of the regression model fit

•	Fluctuating of measured residual data in random manner show a satisfactory plot on the band and its clear in full quadratic models for all the sixteen models and its clear in Fig. 2

•

Full quadratic models in all the response (slump flow, L-Box, segregation resistance and 28 days compressive strength) shows a high correlation coefficient (R2) and adjusted correlation coefficient and less level of significant from the four predictions models (linear, interaction, full quadratic and pure quadratic) were developed

ACKNOWLEDGEMENTS

This research described in this study was conducted at the Civil of Engineering (COE) Department at Universiti Tenaga Nasional (Uniten) at Malaysia and was funded by the Uniten Internal Grant and Ministry of Science, Technology and Innovation (MOSTI), E-Science Grant. (03-02-03-SFO140) also, the researcher would like to thank BASF chemical company for their helps in implementing this research.

How to cite this article:

Hashem AL-Mattarneh, Kamal Nasharuddin Mustapha, Arabi N.S. Al Qadi and Qahir N.S. AL-Kadi. Central Composite Design Models for Workability and Strength of Self-Compacting Concrete.
DOI: https://doi.org/10.36478/jeasci.2009.177.183
URL: https://www.makhillpublications.co/view-article/1816-949x/jeasci.2009.177.183