This action might not be possible to undo. Are you sure you want to continue?
Optimisation of steel ﬁbre reinforced concretes by means of statistical response surface method
F. Bayramov, C. Tas demir *, M.A. Tas demir ß ß
Department of Civil Engineering, Istanbul Technical University, Maslak, 80626 Istanbul, Turkey Received 29 October 2002; accepted 25 June 2003
Abstract The objective of this research is to optimise the fracture parameters of steel ﬁbre reinforced concretes to obtain a more ductile behaviour than that of plain concrete. The eﬀects of the aspect ratio (L=d) and volume fraction of steel ﬁbre (Vf ) on fracture properties of concrete in bending were investigated by measuring the fracture energy (GF ) and characteristic length (lch ). For optimisation, three-level full factorial experimental design and response surface method were used. The results show that the eﬀects of ﬁbre volume fraction and aspect ratio on fracture energy and characteristic length are very signiﬁcant. Ó 2003 Elsevier Ltd. All rights reserved.
Keywords: Fibre reinforced concrete; Steel ﬁbre; Fracture energy; Characteristic length; Response surface method; Mix optimisation
1. Introduction In this study, steel ﬁbre reinforced concretes (SFRCs) are used as cement-based composite materials. These materials are reinforced with steel ﬁbres. The ﬁbres used are short, and their aspect ratio (length/diameter), volume fraction, and distribution as well as their matrix properties mainly inﬂuence the performance of SFRC. As a result of the rapid developments in concrete technology, concretes of strength over 100 MPa can be easily produced using ordinary materials and applying conventional mix design methods . The main concern about high strength concrete is its increasing brittleness while its strength is being increased. The higher the strength of concrete, the lower is its ductility. This inverse relation between strength and ductility is a serious drawback when using high strength concrete is of concern. A compromise between these two conﬂicting properties of concrete can be obtained by adding short ﬁbres . Therefore, it becomes a more signiﬁcant problem to improve the ductility of concrete. Steel ﬁbre reinforcement greatly increases the energy absorption and ductility of concrete. For a better understanding of
* Corresponding author. Tel.: +90-212-285-3771; fax: +90-212-2856587. E-mail address: firstname.lastname@example.org (C. Tas demir). ß
the fracture behaviour of concrete structures, knowledge of the post-cracking behaviour of concrete material is essential. The main contribution of steel ﬁbres to concrete can mainly be observed after matrix cracking. If a proper design is made, after the matrix cracking, randomly distributed, short ﬁbres in the matrix arrest microcracks, bridge these cracks, undergo a pull-out process and limit crack propagation [3,4]. Debonding and pulling out the ﬁbres require more energy, therefore, a substantial increase in toughness, and resistance to cyclic and dynamic loading occurs . ASTM C 1018 proposes a method for evaluating the toughness indices and JSCE-SF4 recommends determining the ﬂexural toughness factor [6,7]. These two toughness parameters are not suﬃcient to deﬁne the tensile strain-softening law , which simulates the behaviour of the damaged region ahead of a continuous crack, which is referred to as the fracture process zone . To deﬁne the softening behaviour, Hillerborg et al.  introduced the concept of fracture energy GF . The standard test for the evaluation of this property was established by RILEM 50-FMC Technical Committee . Until recently, the optimisation of SFRCs has been generally made by maximising the fracture energy (GF ). Li et al.  solved what was essentially an optimisation problem and also obtained approximate optimal values of ﬁbre aspect ratio (L=d) and of the frictional bond
0958-9465/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0958-9465(03)00161-6
and characteristic length (lch ). regression analysis and optimisation to provide useful approaches to the problem development. E. compressive strength (fc0 ). it is necessary to have some experimental data. a multiobjective simultaneous optimisation technique is used to optimise SFRC with special emphasis on ductility. GF . In one study. and ﬁbre volume fraction (x2 ¼ Vf ). but also splitting tensile strength (fst ). RSM combines statistical and mathematical methods of experiment design. two variable full factorial experimental design. E. statistically based procedure for planning. modulus of elasticity (E). and fracture energy. Bayramov et al. splitting tensile strength (fst ). Lange-Kornbak and Karihaloo were the ﬁrst to use mathematical optimisation techniques for the maximisation of GF and fc0 (either separately or simultaneously) of SFRC [13. optimisation does not lead to a general solution of the problem. fc0 . such as the characteristic length (lch ). The use of RSM to optimise SFRC mix design variables can signiﬁcantly increase not only ductility (lch ). Without a model. choosing a proper model that ﬁts that data. For this purpose. These two responses are of a conﬂicting nature.14]. 1. They developed rigorous micromechanical relations between tensile strength and fracture energy. In order to build a model. Brandt  addressed to the issues about the simultaneous maximisation of GF . 2k ¼ 2 Â 2 ¼ 4 axial points and 1 centre point as plotted in Fig. / Cement & Concrete Composites 26 (2004) 665–675 strength between the ﬁbre-matrix by ﬁrst relating the objective function GF to these ﬁbre parameters. 1. analyse how certain factors inﬂuence the measurement system. chemical and other industries. These graphs oﬀer an opportunity to visually The additional limitations on the ﬁbre properties given in the set of Eq. or responses are compressive strength (fc0 ). or can be assumed. and the ﬁrst cracking strength of FRC. To optimise a process or to ﬁnd the best-ﬁtting function of a number of experimental points. Thus. For the use of a three-level. Research signiﬁcance For improving the mechanical behaviour and fracture properties of SFRC with respect to aspect ratio and volume fraction of steel ﬁbre. These objective functions were related empirically to the mix. Simon et al. optimal values for the design parameters (L=d and Vf ) were found while the brittleness of concrete is minimised.1. fflex . Response surface modelling Analysis of response surface over the simplex region involves the collection of experimental data. It provides a comprehensive. improvement. fst . as can be seen from the following expression proposed by Hillerborg et al. RSM. optimum design is needed. For 65 6 L=d 6 80. Two factors (or independent xk design variables) that can be selected to describe this system are ﬁbre aspect ratio (x1 ¼ L=d). RSMs are tools for the modelling and analysis of problems where one or more measured responses are inﬂuenced by several factors and the objective is to optimise the response. a statistical RSM has been used. In this study. the dependent variables. The full factorial design for k ¼ 2 independent variables consists of 32 ¼ 9 points. A response surface can simultaneously represent two or more independent and one dependent variable when the mathematical relationship between the variables is known. splitting tensile strength (fst ) and ﬂexural strength at the minimum steel ﬁbre volume fraction.  optimised High Performance Concrete mixtures using this method. after this. but it has had limited use in the concrete industry. Using the experimental results. a model has to be found ﬁrst. splitting tensile strength. executing. respectively. A response surface is the graph of system response as a function of one or more variables. . The eﬀects of aspect ratio and volume fraction of steel ﬁbre on the fracture energy and ductility were determined. and testing the adequacy of the ﬁtted model. modulus of elasticity. and the mix and ﬁbre parameters. respectively. (2) are as follows: For 55 6 L=d 6 65. L ¼ 60 mm and 0:75 6 d 6 0:92 mm are speciﬁed.666 F. ﬂexural strength (fflex ).e. In another work. The main aim of this research is to provide the optimisation by maximising the characteristic length (lch ). respectively. Reasonable ranges can be given as follows: 55 6 L=d 6 80 0:26% 6 Vf 6 0:64% ð2Þ 2. the optimisation procedure is performed using the response surface of the model as the basis for ﬁnding the best solution. fracture energy (GF ). and GF are direct tensile strength. has been widely used to optimise products and processes in manufacturing. and the ﬁbre variables. L ¼ 3ðL=dÞ À 135 and the ranges of L and d are 30 6 L 6 60 mm and 0:55 6 d 6 0:92 mm. In this study. or optimisation. : lch ¼ GF E ft02 ð1Þ where ft0 . A common response surface experimental plan which can be used to ﬁnd optimal settings is a three-level. and evaluating batch . characteristic length and ﬂexural strength were maximised simultaneously at the minimum steel ﬁbre volume fraction. in which response surface method (RSM) is incorporated. that is 2k ¼ 22 ¼ 4 factorial points. the two variable full factorial experimental design allows estimation of a full quadratic model for each response (i.
which is often present when a response is at a maximum or minimum in the selected region .26 667 were given in Table 1.16 g/cm3 and the cement content of concrete was 400 kg/m3 . the volume fractions of cement. b3 . and lch ). mm 0. (3)) account for curvature in the response surface. crushed limestone.92 0. / Cement & Concrete Composites 26 (2004) 665–675 80 Fibre aspect ratio (L/d) 75 70 65 60 55 0. and 80. and then at 1.3. The beams prepared for the fracture energy tests were 500 mm in length and 100 · 100 mm in cross-section. their densities were 2.% 0.85 7. and water were kept constant. g/cm3 7. SFRC beam specimens were tested at the loading rate of 0. mm 60 60 30 Diameter (d). siliceous sand. The density of coarse aggregate was 2. The specimens were cast in steel moulds and compacted on a vibration table.85 Tensile strength.65 g/cm3 . The volume fractions of steel ﬁbres were 0. 3.3 mm/min up to a deﬂection of 2 mm. Steel ﬁbres were scattered on the mixture and carefully mixed to achieve a uniform distribution of the ﬁbres in the concrete.5 mm/min up to a 5 mm deﬂection. and 0. The eﬀective cross-section. Water to cement ratio was 0. the tests for the determination of the fracture energy (GF ) were performed according to the recommendation of RILEM 50-FMC Technical Committee . Test procedure Standard strength tests were conducted in accordance with European Standards (EN 206 and EN 12390).F. siliceous sand. Three cylinders were used for compressive tests. Siliceous sand (0–0. Experimental details 3.75 0. 0. 3. N/mm2 1050 1000 1100 . Orientation of the ﬁbres is generally random. and for the splitting test. and the aspect ratios of ﬁbres (L=d) were 55. 65. and b5 . Details of the tests and dimensions of the specimens are given in Table 3.65 g/cm3 . crushed limestone was used. at 20 °C until 28 days old.45 Fibre volume fraction (Vf). is shown below: y ¼ b0 þ b1 x1 þ b2 x2 þ b3 x2 þ b4 x2 þ b5 x1 x2 1 2 ð3Þ The quadratic terms in (Eq. Details of the mixtures are given in Table 2.63 and 2.2. As schematically Fig. For example. six disc specimens were prepared. As coarse aggregate. cement. and V and the last two digits show the ﬁbre volume fraction.45%.75% and 1. All the specimens were demoulded after about 24 h. The concrete mixes were designated with the following codes: A and ﬁrst two digits show the ﬁbre aspect ratio. In all the mixes.55 Aspect ratio (L=d) 80 65 55 Density. Experimental design points according to three-level full factorial design.36. 0. however. The notched beam specimen tested is shown in the inset of Fig. respectively.85 7.5% by weight of cement for diﬀerent concrete mixtures to maintain suﬃcient workability. b4 . The properties of the steel ﬁbres Table 1 The properties of the hooked-end steel ﬁbres Length (L). 1. limestone ﬁnes. This table also shows some properties of fresh concretes.64%. A80V64 indicates a mix with a ﬁbre aspect ratio of 80 and a ﬁbre volume fraction of 0. At least three specimens of each concrete mix were tested under each type of loading condition at the 28th day.25 mm) and limestone ﬁnes (0–4 mm) were used as ﬁne aggregates. stored in water saturated with lime. NC denotes the control mix.26%. The maximum particle size of aggregate was 16 mm. 2. Cement used was ordinary Portland cement with a density of 3. Since the ratio of compressive strength to tensile strength of the SFRC tested in this work is in the range of 5 to 10.1. Specimen preparation In mixing. highrange water reducing admixture and water were added to the mix. which has 6 coeﬃcients that are b0 . Bayramov et al. Materials and mix proportions A total of 10 concrete mixes were cast for this investigation.64%. was reduced to 60 · 100 mm by sawing in order to accommodate both large aggregates and steel ﬁbres used. b2 . b1 . The full quadratic model for two independent variables. the method suggested by Hillerborg  and also pointed out by Barros and Figueiras  has been used. The amount of high-range water reducing admixture varied between 0. limestone ﬁnes and crushed limestone were blended ﬁrst and then. For all the beams.64 3.
the deﬂections were measured simultaneously by using two linear variable displacement transducers (LVDTs).668 F. height and diameter of the disc specimen.71 395 280 625 946 143 1.36 0. the addition of steel ﬁbres into concrete may have an eﬀect of increasing the ductility in the compressive failure rather than the compressive strength.47 A65V45 396 281 627 949 143 0.3) 0.5 2447 0. respectively.45 55 9 2446 0. it can be concluded that more Table 2 The details of mix concrete proportions Mix code 3 . 397 281 627 950 143 1. for the aspect ratio of 55 and 80.08 A80V45 seen in Fig.26 80 8 2430 0. kg/m3 Air content.64 65 6 2450 0. kg/m3 Limestone ﬁnes (0–4 mm).2) 0. % 396 281 627 948 143 0.36 0. % Fibre aspect ratio (L=d) Slump. There is no signiﬁcant eﬀect of ﬁbre volume fraction on the modulus of elasticity. this increase was 42% and 24%.75 (3) 0. respectively. An increase of the ﬁbre volume fraction (Vf ) from 0. The eﬀect of the volume fraction of steel ﬁbre seems to be more signiﬁcant in concretes with the aspect ratio of 65.4) 0.26 55 5 2415 1.79 A65V64 396 281 626 947 143 0. Splitting tensile strength Splitting tensile strength of a disc specimen is calculated using the following expression: fst ¼ 2P phd ð4Þ A55V64 A55V45 NC Cement (C). cm Unit weight.45 80 5 2430 1.64% has resulted in an increase of 30% in the compressive strength. kg/m3 Crushed limestone (4–16 mm).36 0.26 65 7 2430 0.04 (4.36 – – 7 2398 1. the diameter of the steel ﬁbre and possibly the orientation may have played an important role in compression.36 0.57 398 283 630 953 144 0.97 (3.85 (2. 3. The load was applied by an MTS of 250 kN maximum capacity. Tensile strength values evaluated from splitting tests on disc specimens are included in Table 4 and plotted in Fig. As seen in Fig.9) 0. an increase of the ﬁbre volume fraction from 0 (i. However. 3 and Table 4.36 0.64% has resulted in an increase of 23% in corresponding splitting tensile strength. no signiﬁcant change occurs when the ﬁbre volume fraction is increased.26% to 0. Bayramov et al.5 2454 0. Thus.9) 0.2) 0.45 65 5.78 398 282 629 953 144 1.23 (4.36 0.8) 0. Experimental results The tests were carried out to investigate the eﬀectiveness of steel ﬁbre in the improvement of fracture energy and ductility with regard to the aspect ratio and ﬁbre volume fraction. The load versus displacement curve for each specimen was obtained by recording the average of two measurements taken at the mid span.e.51 A80V64 A80V26 4. 4. / Cement & Concrete Composites 26 (2004) 665–675 398 282 629 952 144 1. kg/m3 High-range water reducing admixture.1. On the other hand. and d are the ultimate load.0 (4) 0.65 396 281 627 949 143 1.2. kg/m3 Water (W). 2. Compressive strength 398 283 630 953 144 1.2) 0. % (kg/m3 ) W/C Fibre volume fraction (Vf ). It is seen that the splitting tensile strength increases with increasing steel ﬁbre volume fraction. normal concrete) to 0.09 (4. h.45 (5.64 55 8 2450 0.8 (3.64 80 6.33 (5.36 0.74 A65V26 where P . kg/m Siliceous sand. thus the eﬀect of the volume fraction of steel ﬁbre on compressive strength is not consistent. 4.36 0. for the ﬁbre aspect ratio of 55. However.36 0. For the aspect ratios of 65 and 80.41 A55V26 Compressive strength test results obtained from cylinder specimens are given in Table 4.
F.1 6. However.7 1851 2650 A55V64 45.83 49.7 957 1224 A65V45 69.3 7. 4. mm (fc0 ).4 8. MPa .5 1939 2056 A65V64 74. MPa Modulus of elasticity (E). NC 60. N/m Characteristic length (lch ). (In NC.4 7. the mechanical mismatch between steel ﬁbres and concrete may have also played a role in this behaviour.19].4 6.4 9. the ﬁbres with the aspect ratio of 80 ðL=d ¼ 80Þ were broken into two parts.3. N/m Flexural strength (fflex ). h ¼ 300 £150. Flexural strength The ﬂexural strength of a notched beam subjected to three-point bending test can be evaluated using the following equation: signiﬁcant results were obtained in SFRCs with the ﬁbre aspect ratio of 65. MPa Splitting tensile strength (fst ).04 5. % 0. Bayramov et al.45 Fibre volume fraction (Vf).36 51.0 5.71 46. GPa Fracture energy (GF ). During the crack propagation some ﬁbres are broken but some of them are pulled-out of the matrix. h ¼ 60 100 · 100 · 500 Parameters Compressive strength (fc0 ). the opening and propagation of the crack are controlled by the steel ﬁbres.1 4371 4845 8 L/d=80 L/d=65 L/d=55 7 6 5 0. fst ¼ 5:3 MPa. In most cases. MPa 669 Fig.4 1793 2352 A80V64 55.26 0. 2. Similar results were obtained by Eren and Celik . Splitting tensile strength versus ﬁbre volume fraction with diﬀerent aspect ratio. / Cement & Concrete Composites 26 (2004) 665–675 Table 3 Test methods and specimen size Test type Compression Splitting Three-point bending Specimen Cylinder Disc Beam Dimensions (mm) £150.2 91 169 A55V26 46. The cylinder compressive strength of ß plain concrete used in this study is about 60 MPa.9 6.3 6. and test setup (b). the ﬁbres with the aspect ratio of 65 ðL=d ¼ 65Þ did not break but were pulled out of the matrix.6 49.64 Fig.3 52. 3. Along the fracture plane.52 44.55 49.92 45.4 1024 1327 A80V45 54. MPa Fracture energy (GF ).5 7. MPa Modulus of elasticity (E).5 6.7 6.3 5.7 1011 1599 A55V45 48. GPa Splitting tensile strength (fst ). After Splitting tensile strength (fst).95 46. high strength steel ﬁbre with a tensile strength of 2000 MPa is suggested [18.58 48. The results obtained for the ﬁbres with L=d ¼ 65 might be due to their larger cross-sections compared to that of ﬁbres with L=d ¼ 80. Schematic representation of the evaluation of fracture energy (a).1 6.3 6. For high strength concrete.1 3724 3207 A80V26 51.) completion of the splitting tests the fracture surfaces were examined.1 6.4 12.1 5.4 5.6 3368 3537 A65V26 57. Table 4 Strength and fracture properties of concrete tested Mix code MPa Compressive strength Flexural strength (fflex ). so for the steel ﬁbres of L=d ¼ 80.
this property was determined using the following expression : GF ¼ W0 mð1 À a2 Þgd ðaÞ ðbÞ ¼ GF þ GF þ bðh À aÞ bðh À aÞ ð6Þ where b. however is not totally dissipated.64 0.26 0. ﬁbres carry the load subjected to concrete until the cracking of interfacial bond between ﬁbres and matrix occurs. Here. For a notched beam in three-point bending. depending on their aspect ratios and volume fractions of ﬁbres used. Flexural strength versus ﬁbre volume fraction with diﬀerent aspect ratio. the procedures recommended by RILEM TC 50-FMC Technical Committee were applied to measure GF . In order to obtain the . fflex ¼ 6:1 MPa.5% and 100%. As shown in Fig.) Fig. 5. weight of the beam between the supports. GF of the mixes are shown in Fig. and a are span. the advantage of using ﬁbres for increasing ﬂexural strength of concrete seems to be more signiﬁcant . fflex ¼ 3PS 2bðh À aÞ 2 ð5Þ where P is the ultimate load. during which slow crack propagation occurs.5%. GF and GF are the fracture energies supplied by actuator and the beam weight. The cut-oﬀ point was chosen as 5 mm deﬂection. % 0. The area under the load versus deﬂection at mid span curve (W0 ) was described as a measure of the fracture energy of the material. / Cement & Concrete Composites 26 (2004) 665–675 L/d=65 L/d=55 4500 Fracture energy (GF).e. respectively. Characteristic length Characteristic length ðlch Þ should be taken into consideration in the design of concrete mixes. depth. an increase in the ﬁbre volume fraction from 0 (i. The effects of the aspect ratio and volume fraction of steel ﬁbre on ﬂexural strength are plotted in Fig. MPa 13 11 9 7 5 0. for the ﬁbre aspect ratio of 55. thickness. a=h ¼ 0:4 and S=h ¼ 4. and also in Fig. The reason for the increase in fracture energy with increasing ﬁbre volume fraction and its aspect ratio stems from a great number of ﬁbres forming a bridge in the crack forming tortuous crack propagation. h. The load–deﬂection curves were used for evaluating the fracture energy. and S are the width. because it controls the nominal strength. As seen in Fig. The fracture process of steel ﬁbre reinforced concrete consists of progressive debonding of ﬁbre. The results obtained here are based on the area under the complete load–deﬂection curve up to a speciﬁed deﬂection.) Note that GF for normal concrete was calculated from the mesomechanical relationships . b. (In NC. Bayramov et al. the energy at this deﬂection (i.64 Fig. failure mode. As seen in Fig. respectively.5.26 L/d=80 L/d=65 L/d=55 Flexural strength (fflex).4. h. after matrix cracking. Fracture energy (GF ) is the energy needed to develop one crack completely. For the ﬁbre aspect ratios of 65 and 80. (In NC. 4 and Table 4.45 Fibre volume fraction (Vf). 2. 5. Fracture energy versus ﬁbre volume fraction with diﬀerent aspect ratio. 4. At higher aspect ratios. height. The ductility is more than about 40 times greater than that of normal concrete.45 Fibre volume fraction (Vf). GF ¼ 91 N/m. and span. respectively.64%.670 L/d=80 F. m. 4. 9. The reason for the increase in ﬂexural strength is that. normal concrete) to 0. the increases were 56. respectively. Final failure occurs due to unstable crack propagation when the ﬁbres are pulled out and the interfacial shear stress reaches the ultimate strength. and S. 4. g is the gravity acceleration. 5 and Table 4.e. these SFRCs allow obtaining high values of fracture energies and as a result a high ductility. expressed as a function of the aspect ratio and volume fraction of steel ﬁbre. and notch depth. d is S ðaÞ ðbÞ the deﬂection of the beam (5 mm) and.N/m 3500 2500 1500 500 0. It can clearly be seen that fracture energy increases as the ﬁbre volume fraction increases. notch depth. a ¼ L À 1. The ﬁgure shows that. Fracture energy One of the major roles of ﬁbre in concrete is to provide an increase in the energy required for fracture by the resultant crack arresting process. 4. 5 mm). ﬂexural strength increases as the ﬁbre volume fraction increases. and crack growth (crack pattern) . It is seen from the schematic curve that. % 0. Flexural strengths of notched beams subjected to three-point bending tests are shown in Table 4. has resulted in an increase of 33. The increase in the fracture energy is because of the high energy of ﬁbre pull-out and ﬁbre debonding in the fracture process. a.
mm 4000 3000 2000 1000 0 0.22 13468. as both aspect ratio and volume fraction of the steel ﬁbre increase. Bayramov et al.89 16.82 0. 6. Perform the regression again with a partial model containing only those terms that are statistically signiﬁcant (i.91 )74.01 )0. / Cement & Concrete Composites 26 (2004) 665–675 L/d=80 5000 Characteristic length (lch). lch ¼ 169 mm.67 3542. In this study the direct tensile strength (ft0 ) term in this relation was replaced by splitting tensile strength (fst ). % 0.12 )21. nine experimental data were ﬁtted to a polynomial type of mathematical model by adjusting parameters until calculated values were in close agreement with the experimental values.42 )44. in this study a ¼ 0:05).1 21687.40 103.12 )33. The variation of lch with L=d and Vf is shown in Fig. the ﬁtted regression model was GF ¼ 3542:1 À 33:64ðL=dÞ À 11514Vf þ 13468:16Vf2 þ 103:41ðL=dÞVf ð7Þ Fig. and repeat this procedure until the partial model contains only signiﬁcant terms.F. but some of the terms may not be signiﬁcant.16 – b5 – – 0. validating the model by examining the residuals for trends and outliers.41 173. Regression analysis It is very important to choose an appropriate model for describing the shape of the surface for each response.) ductility of the mixes. (In NC. and interpreting the model graphically.32 b3 )0. splitting tensile strength (fst ). Characteristic length versus ﬁbre volume fraction with diﬀerent aspect ratio.6 b1 17. The procedure to specify an appropriate model is presented below: 1.02 – – 3. In the analyses of each response. modulus of elasticity (E). As seen in Fig. lch was calculated using the measured GF . Using the same procedure described above for GF . The full quadratic model was simpliﬁed by using a backward stepwise technique.26 0. A full quadratic model can be estimated with a three-level full factorial design. E and ft0 according to (Eq. ﬁtting a full quadratic model using ANOVA. ﬂexural strength (fflex ) and characteristic length (lch ): fc0 ¼ À530:12 þ 17:26ðL=dÞ þ 17:75Vf À 0:13ðL=dÞ fst ¼ À21:68 þ 0:84ðL=dÞ À 2:91Vf À 0:01ðL=dÞ þ 5:98Vf2 E ¼ À15:89 þ 2:48ðL=dÞ À 74:42Vf À 0:02ðL=dÞ þ 17:15Vf2 þ 0:82ðL=dÞVf 2 2 2 ð8Þ ð9Þ ð10Þ Table 5 Model ﬁtting results for each response Responses fc0 fst E fflex GF lch Coeﬃcients b0 )530. the characteristic length which is a measure of ductility of the mix.48 )0. the following models (Eqs.26 0. Model ﬁtting results for each response are shown in Table 5. (8)–(12)) were used for compressive strength (fc0 ). 5.64 L/d=65 L/d=55 671 accept or reject a certain explanation term is to test its signiﬁcance in order to ensure that the additional term really contributes to the explanation of the response.e.68 )15.43 b2 17.53 .45 Fibre volume fraction (Vf). and ðL=dÞVf are the signiﬁcant terms. 6. (1)) introduced in the ﬁctitious crack model by Hillerborg et al.64 )11514 )5080.64 )600.95 b4 – 5.15 30. only signiﬁcant terms (a < 0:05) were included in the model.84 2. For fracture energy (GF ). it can be concluded that the results obtained give a clear picture of how a quasibrittle concrete transforms into a ductile composite with the addition of steel ﬁbres. Fit the full quadratic model and for each coeﬃcient. increases signiﬁcantly. . those terms that had a t-statistic greater than that of the chosen signiﬁcance level. Each mechanical response of SFRC is analysed individually by examining summary plots of the data. Vf . Thus. 2. Vf2 .13 )0. Hence. The only way to In this case ðL=dÞ. calculate the t-statistic for the null hypothesis that the coeﬃcient is equal to zero.98 17. Calculate t-statistics again and eliminate any insigniﬁcant terms.75 )2. 6 and Table 4.
64 80 65 0.45 Fibre volume fraction (Vf).26 55 0.0 5.8 5. Splitting tensile strength (fst). The variation of fst with L=d and Vf . (7)–(12).1 5. fst . the ﬁtted values that were obtained for each of the mechanical properties of SFRC are shown in Table 6. GF .2 Splitting tensile strength (fst ). % Fibre aspect ratio (L/d) Fig.3 6.1 46. 8.9 6.8 7.6 Flexural strength (fflex ). GPa 48.6 Fracture energy (GF).5 46. Fig. MPa 11.1 0.MPa 7.45 0. / Cement & Concrete Composites 26 (2004) 665–675 Table 6 Fitted values for each mechanical properties of SFRC Fibre aspect ratio (L=d) Fibre volume fraction (Vf ). and lch as a function of the aspect ratio and volume fraction of steel ﬁbre are plotted in Figs.9 50. Bayramov et al.6 49.45 0.5 5.26 0.5 6.26 fflex ¼ 16:67 À 0:12ðL=dÞ À 44:64Vf þ 30:22Vf2 þ 0:4ðL=dÞVf lch ¼ 21687:6 À 600:43ðL=dÞ À 5080:32Vf þ 3:95ðL=dÞ2 þ 173:53ðL=dÞVf ð12Þ ð11Þ Flexural strength.3 44. all independent 0.0 By using the Eqs. The variation of fflex with L=d and Vf .6 43.672 F.7 68. 7–10.7 6. % Fitted values Compressive strength (fc0 ).0 46.3 71.3 64.26 0.6 49.7 50. N/m 4251 3413 2575 1738 900 0. MPa 11.1 49. (fflex). N/m 4251 2084 900 3767 1896 1009 3445 1772 1082 Characteristic length (lch ).6 51. (7)–(12).6 6.8 6. fflex .9 6. % 80 65 0.26 55 65 Fibre aspect ratio (L/d) Fig. 9. Response surfaces of the mechanical properties. MPa 57.4 9. The variation of GF with L=d and Vf .3 6. 7. respectively. mm 4523 2841 1159 3286 2101 917 3448 2595 1742 80 80 80 65 65 65 55 55 55 0.1 6.45 Fibre volume fraction (Vf).1 53.5 7. .64 0. 7.64 80 6.64 0.9 Fracture energy (GF ).8 10. % 0.26 55 Fibre aspect ratio (L/d) 0. Optimisation After building the regression model and establishing relationships between mix design variables and the responses expressed in Eqs.7 6.2 9.45 0.2 45.4 8.6 6.45 Fibre volume fraction (Vf).64 0.5 Modulus of elasticity (E).5 6.64 0. Mpa 6.8 7.
By using the single composite response (D) given in Eq. fflex and volume fraction of steel ﬁbre (Vf ) are also considered to be of Composite desirability (D) 0. which are deﬁned for each response. Therefore.64 80 0. simultaneously. (13) takes on the form: D ¼ ðd1 Â d2 Â d3 Þ3 1 ð16Þ Fig. 11. mm 4523 3592 2662 1731 917 673 0.e. In cases of the maximising and minimising of individual responses. the overall function becomes zero. the ﬁtted value. lch and fflex ) are considered to be of equal importance. max fj Àmin fj ð14Þ dj ¼ > > min fj < Yj < max fj > : 1 Yj P max fj 8 Yj 6 min fj >1 >h > max f ÀY iwtj < j j and 0 < dj < 1. thus L ¼ 60 mm and d ¼ 0:785 mm.45 Fibre volume fraction (Vf). 8 Yj 6 min fj >0 >h > Yj Àmin fj iwtj < and 0 < dj < 1.404 0. The ﬁgure shows that. % 0. variables are varied simultaneously and independently in order to optimise the objective functions. which shows that d is between 0. respectively.26 55 65 Fibre aspect ratio (L/d) actual (experimental) values of jth response included in the optimisation. can be used to optimize the responses simultaneously . A desirability function (dj ) varies over the range of 0 6 dj 6 1. and minimum and maximum 0.202 0 ð13Þ where n is the number of responses included in the optimisation. these three responses (fst . The power value wtj is a weighting factor of the jth response. respectively. In this work. The predicted response values and associated uncertainties (at 95% conﬁdence level) are lch ¼ 4068 Æ 415 mm. (14) and (15). lch and fflex . the desirability will be deﬁned by the formulas given in Eqs. (13). min fj and max fj are the desirability function. fst . The objective of optimisation is to ﬁnd the ‘‘best settings’’ that maximise a particular response or responses.45 Fibre volume fraction (Vf). wtj ¼ 1 and maximised simultaneously. the optimal values of design variables are Vf ¼ 0:64% and L=d ¼ 76:44. / Cement & Concrete Composites 26 (2004) 665–675 Characteristic length (lch). Response surface plot of the composite desirability (D) when fst . the multiobjective optimisation problem is solved. The variation of lch with L=d and Vf . d2 and d3 are the desirability functions of fst . E ¼ 49:2 Æ 0:4 GPa and fc0 ¼ 65:7 Æ 4:7 MPa. for optimisation. To attain a less brittle concrete. The cost of the steel ﬁbres used in the production of composites is also important from the application is of concern. the volume fraction of steel ﬁbre must be minimised to get an economical mixture. Thus. So it is necessary to maximise fst . A numerical optimisation technique using desirability functions (dj ). (2) : D ¼ ðd1 Â d2 Â d3 Â Á Á Á Â dn Þn 1 where d1 . i. respectively. Bayramov et al. max fj Àmin fj dj ¼ ð15Þ > min fj < Yj < max fj > > : 0 Yj P max fj where dj .606 0. Since 65 6 L=d 6 80.92. The solution of this multiobjective optimisation is shown in Fig. fst ¼ 7:1 Æ 0:2 MPa. lch and fflex . For n ¼ 3. a concrete with the highest splitting tensile strength (fst ). fflex ¼ 11:3 Æ 0:7 MPa. the highest characteristic length (lch ) and also the highest ﬂexural strength (fflex ) is to be obtained. Yj .26 55 Fig.808 0. lch and fflex are maximised simultaneously.64 80 0. .F. a commercially available (Design-Expert) software package was used. Optimisation usually involves considering several responses simultaneously. If any of the responses or factors falls outside their desirability range. the limitations of L and d are: L ¼ 60 mm and 0:75 6 d 6 0:92 mm. Eq. GF ¼ 4136 Æ 237 N/m.% 65 Fibre aspect ratio (L/d ) 0. lch . 10. 11. D is maximised over the feasible region of the design variables given in Eq. Numerical optimisation can optimise any combination of either factors or responses. Thus.75 and 0. which is the geometric mean of the individual desirability functions.
The second and third authors wish to acknowledge the grant of The British Council (Britain–Turkey Partnership Program for the academic link between ITU and Cardiﬀ University). where d1 . wtj ¼ 1 and optimised simultaneously. the optimal values of design variables are Vf ¼ 0:558% and L=d ¼ 75:87. d ¼ 0:791 mm. ACI Mater J 1992.46 Fibre volume fraction (Vf). i. (13) takes on the form: D ¼ ðd1 Â d2 Â d3 Â d4 Þ4 1 uncertainties (at 95% conﬁdence level) are lch ¼ 3359Æ 310 mm.45 0. L ¼ 60 mm. / Cement & Concrete Composites 26 (2004) 665–675 equal importance. fflex and Vf .076 References  Balaguru P. i. Response surface plot of the composite desirability (D) when fst .  Banthia N.436 0. ACI Mater J 1992. Contour plot of the composite desirability (D) when fst . Trottier JF.229 0. and Vf ¼ 0:558%.44. the optimal values of design variables obtained are as follows: a steel ﬁbre volume fraction of 0.674 F.e.152 0.64 0. Part II: Toughness characterization. Patel M.e.45 Fibre volume fraction (Vf).87. ACI Mater J 1995. lch and fflex ) are concerned. Concrete reinforced with deformed steel ﬁbres.e.% 0. Ashour SA.92(2):146–54. d3 . and the ﬁnancial supports of TUBITAK (Project: ICTAG 1665) and DPT (State Planning Organization. Concluding remarks Response surface method is a promising approach for optimising SFRCs to meet several performance criteria such as minimum brittleness. The experimental design made by using RSM provides a thorough examination of SFRC properties over the selected range of ﬁbre volume fraction and aspect ratio. i.89(5):449–55.  Wafa FF.% 0.64% and an aspect ratio of 76.343 0.558% and an aspect ratio of 75.64 80 0. The predicted response values and associated Composite desirability (D) 0. ﬁtting quadratic models that are usually assumed to represent each concrete property of interest. GF ¼ 3125 Æ 165 N/m. lch and fflex ) and cost optimisation the optimal values of design variables obtained are as follows: a steel ﬁbre volume fraction of 0. fflex ¼ 9:4 Æ 0:5 MPa. d2 .26 55 65 Fibre aspect ratio (L/d ) 80 Fig. The results show that the predictiveness of the polynomial regression model is satisfactory. . Compatibility of the matrix with the steel ﬁbre should be investigated and an optimisation based mixture design is to be used as a future work.457 0.89(6):541–6. lch and fflex are maximised and Vf is minimised.246 0. 0. Flexural toughness of steel ﬁbre reinforced concrete. 12 and 13. ð17Þ 7. lch . lch and fflex are maximised and steel ﬁbre volume fraction (Vf ) is minimised simultaneously. fst . In order to provide an adequate representation of the responses. In this case Eq. When the mechanical properties (fst . The support given by BEKSA is also gratefully acknowledged. These ﬁgures show that.381 0. Narahari R. E ¼ 48:6 Æ 0:3 GPa and fc0 ¼ 65 Æ 3:8 MPa.246 0. 12.26 55 65 Fibre aspect ratio (L/d ) Fig. can be done in identifying optimal mixes. Project: 2002K120340). The solution of this multiobjective optimisation is shown in Figs. The addition of steel ﬁbre improves fracture properties signiﬁcantly. Mechanical properties of high-strength ﬁbre reinforced concrete. respectively.152 0. For both mechanical properties (fst . and d4 are the desirability functions of fst . Bayramov et al. Acknowledgements The ﬁnancial support of TUBITAK (The Scientiﬁc and Technical Research Council of Turkey)––BAYG is greatly appreciated by the ﬁrst author. fst ¼ 6:8 Æ 0:1 MPa. 0.322 0. 13.114 0 0. lch and fflex are maximised and steel ﬁbre volume fraction (Vf ) is minimised simultaneously.
 Hillerborg A. Report to RILEM TC 50-FMC.  Gao J. Tension softening diagrams and evaluation of properties of steel ﬁbre reinforced concrete. vol. Inﬂuence of the yield strength of steel ﬁbres on the toughness of ﬁbre reinforced high strength concrete. 1311– 21. 1985. vol. editor. Adv Cem Based Mater 1998.  Barros JAO. Modeer M. Concrete fracture energy tests performed by 9 laboratories according to a draft RILEM recommendation. ASTM C 1018-97.  Hillerborg A. Sweden. In: Proceedings. Flexural behaviour of SFRC: testing and modelling.18(106):285–90. p. materials. London: E & FN Spon. .  ASTM. Local approach to fatigue of concrete. Bayramov et al. In: K€ nig G et al. Karihaloo BL. 1991. Oslo.  Karihaloo BL. Worldwide Advances in Structural Concrete and Masonry.F.  Myers RH.  Brandt AM. In: Proceedings of the 5th International Symposium on Utilization of High Strength/High Performance Concrete. € ß  Eren O. Optimisation technique for the design of high-performance ﬁbre-reinforced concrete.  Lange-Kornbak D. Adv Cem Based Mater 1996. 2. Rokugo K. the CCMS Symposium. Backer SA. 496– 505. Leipzig. 2002. 3.  Gr€ newald S. J Mech Phys Solids 1991. Kunieda M.  Hordijk DA.6:773–82. Lund.  Derringer G. Struct.  RILEM 50-FMC.. New York: John Wiley & Sons.39:607–25. Wang Y.19:307–13. 675  Simon MJ. Philadelphia. Determination of fracture energy of mortar and concrete by means of three-point bend tests on notched beams. Design of concrete mixes for minimum brittleness.  Lange-Kornbak D. PhD dissertation. Chicago. 1995. Delft University of Technology. Eﬀect of silica fume and steel ﬁbers on some properties of high-strength concrete. Response surface methodology: process and product optimization using designed experiments. Method of tests for ﬂexural strength and ﬂexural toughness of steel ﬁbre reinforced concrete. Tokyo. 1996. mechanical properties and performance.12(4):214–9.11(4):331–9. Committee of fracture mechanics of concrete. Norway: Norwegian Concrete Association. Cem Concr Compos 1997.  Li VC. J Qual Technol 1980. Celik T. Sun W. 58–61. Kamada T. 1997. June 2002. 1984. Suich R. Struct Multidiscip Optimisation 2001. p. Japan Society of Civil Engineers (JSCE). Morino K. Constr Build Mater 1997. Eng Fract Mech 2000. 1999. High strength self-compacting ﬁbre u reinforced concrete: behaviour in the fresh and hardened state. Report TVBM-3015. Mechanical properties of steel ﬁbrereinforced. Figueiras JA. Standard test method for ﬂexural toughness and ﬁrstcrack strength of steel ﬁbre reinforced concrete (using beam with third-point loading). Optimising high-performance concrete mixtures using statistical response surface methods. Walraven JC.  Vandewalle L. 977–89. Analysis of crack formation and crack growths in concrete by means of fracture mechanics and ﬁnite elements. Concrete Library of JSCE. Cem Concr Res 1976.65:235–45. Simultaneous optimization of several response variables. Delft. Montgomery DC. 1983. Mater. Design of ﬁbre-reinforced DSP mixes for minimum brittleness. Lagergren ES. Netherlands. J Mater Civ Eng 1999. Micromechanical model of tension softening and bridging toughening of short random ﬁbre reinforced brittle matrix composites. lightweight concrete.7:89– 101. 6th International Symposium on HSC/ o HPC.3:124– 32. Peterson PE. p.  JSCE-SF4. Lange-Kornbak D. Uchida Y. high-strength. / Cement & Concrete Composites 26 (2004) 665–675  Kurihara N. Wathne LG. Lund Institute of Technology.11(7–8):373–82. Cement-based composites. p. Karihaloo BL.21(1):32–9.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.