Jtam-A4.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 55, 2, pp. 607-619, Warsaw 2017 DOI: 10.15632/jtam-pl.55.2.607 THE FRACTURE MECHANICS FORMULAS FOR SPLIT-TENSION STRIPS Ragip Ince Firat University, Engineering Faculty, Civil Engineering Department, Elazig, Turkey e-mail: rince@firat.edu.tr The notched beams have been commonly used in concrete fracture. In this study, the splitting-strip specimens, which have some advantages – compactness and lightness – com- pared to beams, are analyzed for effective crack models. Using the Fourier integrals and Fourier series, a formula for the maximum tensile strength of concrete is first derived for an un-notched splitting-strip in the plane of loading. Subsequently, the linear elastic fracture mechanics formulas of the splitting-strip specimens, namely the stress intensity factor KI, the crackmouth opening displacementCMOD, and the crack opening displacement profile COD, are determined for different load-distributed widths via the weight function. Keywords: concrete, Fourier integrals, fracture mechanics, splitting test, weight functions 1. Introduction Thesplit-tension specimensare frequentlyusedtodeterminetensile strengthofmaterials, suchas concrete and rock (Neville, 2011). The split-cylinder test, which is also called the Brazilian split test, was proposed byCarneiro andBarcellos (1949). This test was successfully applied to cubes by Nilsson (1961). However, split-tension test specimens, namely, cylinders, cubes and diagonal cubes, have also been successfully used in concrete fractures over the last decade (Modeer, 1979; Tang, 1994; Rocco et al., 1995; Tang et al., 1996; Ince, 2010, 2012a,b). The splitting concrete specimens exhibit several advantages, e.g., compactness and lightness, and the weight of the specimen can be disregarded in calculation of fracture parameters. The experimental investigations on fracturemechanics of cement-basedmaterials conducted until the 1970s indicated that classical linear elastic fracture mechanics (LEFM) was no longer valid for quasi-brittle materials such as concrete. This inapplicability of LEFM was due to the existence of a relatively large inelastic zone in the front and around the tip of themain cracks in concrete.This so-called fractureprocess zone (FPZ)was ignoredbyLEFM.Consequently, several investigators have developed deterministic fracture-mechanics approaches to describe fracture- -dominated failure of concrete structures. These models could be classified as cohesive crack models and effective crackmodels. Contrary to LEFM, inwhich a single fracture parameter was used such as the critical stress intensity factor, thosemodels needed at least two experimentally determined fractureparameters to characterize failure of concrete structures.Thesemodels could be classified as cohesive crack models and effective crack models: namely the two-parameter model (Jenq and Shah, 1985), the effective crackmodel (Nallathambi andKarihaloo, 1986), the size effectmodel (Bazant andKazemi, 1990) and the double-Kmodel (Xu andReinhardt, 1999). Analytical and numerical studies on split-tension specimens (Modeer, 1979; Tang, 1994; Rocco et al., 1995; Tang et al., 1996; Ince, 2010, 2012a,b; Ince et al., 2015, 2016) have revealed that nominal strength is highly affected by the width of the distributed load and the specimen size. The existing design codes have not considered these effects in the determination of the split-tensile strength of concrete. On the other hand, theoretical and experimental studies with 608 R. Ince splitting strip specimens (Davies and Bose, 1968; Filon, 1903; Schleeh, 1978) are limited when compared to those of other splitting specimens. For this purpose, LEFM formulas for the stress intensity factor KI, the crackmouth opening displacement (CMOD), and the crack opening displacement profile (COD) of splitting strip specimens have been evaluated for different load-distributed widths and initial crack lengths utilizing the weight function in this study. In addition, the maximum tensile stresses in the un-notched split-tension strips have also been determined for different load-distributed widths. 2. A historical overview of splitting tests As indicated inFig. 1a, a split-tension specimen is placedbetween theplatens of the testmachine and the load is subsequently applied until failure, which is caused by splitting along the vertical diameter due to the lateral tensile stress thathas occurred (Neville, 2011).According to elasticity theory (Timoshenko and Godier, 1970), the nominal tensile strength of split-tension specimens is defined as σNc = 2Pc πbh (2.1) wherePc is the ultimate load, b is the specimenwidth andh is the specimendepth.However, Eq. (2.1) is onlyvalid for the concentrated loading condition shown inFig. 1a. Inpractice, theapplied load is distributed on the specimens over a finite width (2t) using soft materials, such as hard cardboard and plywood, as indicated in Fig. 1b (Neville, 2011; Davies and Bose, 1960). Tang et al. (1992) investigated the effect of distributed load in three-point bending beams and split- -tension cylinders. The nominal strength decreased with the increasing width of the distributed load in the split-tension cylinder,whereas this effectwasnot significant in thebendingspecimens. According toTang(1994), themaximumtensile stress valueof theun-notchedcylinder specimens at the plane of loading could be calculated as σmax = 2P πbh √ (1−β2)3 (2.2) where P is the total compressive load and β = 2t/h = t/d is the ratio of the distributed-load width to the specimen depth (d is the characteristic specimen size), as depicted in Fig. 1. Rocco et al. (1995) examined the cylinder and cube specimens and proposed that themaximum tensile stress could be calculated for the un-notched cube specimens at the plane of loading as follows σmax = 2P πbh [ 3 √ (1−β2)5−0.0115 ] β ¬ 0.20 (2.3) Using the boundary element method, a formula for the maximum tensile strength of concrete was similarly derived for un-notched diagonal cubes in the plane of loading by Ince (2012a) as follows σmax = 2P πbh 1 0.931+38.931β4.778 β ¬ 0.25 (2.4) Oneof theadvantages ofdiagonal splitting-cube specimens isnearly const = 1/0.931=1.074 maximumstress in theplaneof loading forβ ¬ 0.15,whichdiffers fromother splitting specimens. Approachesbasedon fracturemechanics andusingnotched split cylinder specimenswerefirst performed by Tweed et al. (1972). They developed a closed-form expression for the geometry factor of the notched split-tension cylinder. Tang (1994) also used the finite element method The fracture mechanics formulas for split-tension strips 609 Fig. 1. Splitting tension test: (a) geometry and stress distribution, (b) notched specimen and distributed loading case to study notched cylinder specimens under the opposite distributed loading and the developed LEFM formulas (Fig. 1b). Cubical and diagonal cubic specimens with a central notch were investigated using effective crack models by Ince (2010, 2012a) who evaluated LEFM formulas with those split-tension specimens for various load-distributed widths and initial crack lengths using the finite element and boundary element method. A series of experimental studies with split-tension and beam specimens were also performed. The results were discussed based on twomost popular fracturemodels: the two-parameter and the size effect models. The results of the tests revealed that the notched cube and diagonal cube tests can be utilized successfully for determining the fractureparameters of concrete. Subsequently, Ince (2012b) derivedan improved version ofLEFMformulas for splitting cylinder and cube specimens and then computed the four- termuniversalweight functionsof the split-tension specimens suchas cylinder, cubeanddiagonal cube by using the boundary element method to simulate the double-K concrete fracturemodel. The numbers of theoretical and experimental studies with splitting strip specimens (Davies and Bose, 1968; Filon, 1903; Schleeh, 1978) were limited when compared to studies conducted with other splitting specimens. For instance, a formula similar to equations (2.2)-(2.4) was not developed. Therefore, in the present research, both un-notched and notched splitting strip specimens have been analyzed using analytical and numerical methods. 3. Deriving LEFM formulas for split-tension strip In this study, split-tension strip specimens have been simulated for the use in effective crack fracture models based on earlier studies of split-tension tests for concrete fracture. The weight function approach was used to determine the LEFM formulas for split-tension strip specimens with the central initial notch. The weight function method was originally suggested by Bueckner (1970) and Rice (1972) to evaluate the stress intensity factor using a simple integration. When a weight function is determined for a notched body, the stress intensity factors could be calculated for arbitrary loadings on the same body. For the case of mode I, the stress intensity factors can be expressed as follows KI = a ∫ 0 σ(y)m(y,a) dy (3.1) 610 R. Ince where a is the crack length, σ(y) is the normal stress along the crack line in the un-cracked body, and m(y,a) is the weight function (or Green’s functions which are essentially equivalent to the weight functions) (Gdoutos, 1990). For the central notch, an infinite strip subjected to four equal point loads on the notch, as shown in Fig. 2, the Green function was derived by Tada et al. (2000) as follows KI = 2F√ 2d [ +0.297 √ 1− (y a )2( 1− cos πa 2d ) ] √ tan πa 2d [ 1− ( cos πa 2d / cos πy 2d )2 ]− 1 2 (3.2) Fig. 2. A center-cracked infinite strip subjected to four equal point loads The accuracy of Eq. (3.2) is better than 1% for all a/d and y/a (Tada et al., 2000). The normal stresses along the crack line in the un-cracked bodymust be first computed to determine the stress intensity factor according to the weight function method. 3.1. Determination of normal stresses along the crack line in the un-cracked body In this study,a rectangularplate−l < x < land−d < y < dwithunitwidth isonly loadedby a uniformcompressive traction p0 in the central region−t < x < t of bothboundaries y =±d. It is modeled by considering earlier studies on split-tension tests (Ince, 2010, 2012a,b), as depicted in Fig. 3a. In the plane elasticity problems not including body forces, the stress components can be expressed via the following equations (Barber, 2004; Timoshenko and Godier, 1970) σx = ∂2Φ ∂y2 σy = ∂2Φ ∂x2 τxy =− ∂2Φ ∂x∂y (3.3) in which Φ is the so-called Airy’s stress function of x and y. This function also satisfies the following bi-harmonic equation ∂4Φ ∂x4 +2 ∂4Φ ∂x2∂y2 + ∂4Φ ∂y4 =0 (3.4) A stress function based on a polynomial can be utilized for some simple cases such as a continuously distributed loading on the boundaries of the elastic body. However, a more useful solution for discontinuous loading, as shown in Fig. 3a, may be derived by using Fourier series (Barber, 2004;Timoshenko andGodier, 1970;Mirsalimov andHasanov, 2015;Basu andMandal, 2016). The following equation has commonly been used as the stress function in such cases Φ =sin(λx)f(y) λ = nπ l (3.5) where n is an integer and f(y) is only a function of y. When Eq. (3.5) is substituted into Eq. (3.4) in order to determine f(y), the following ordinary differential equation is obtained f(4)(y)−2λ2f ′′(y)+λ4f(y)= 0 (3.6) The fracture mechanics formulas for split-tension strips 611 Fig. 3. (a) Base problem for split-tension strip, (b) case of periodic loading The general solution to the above equation can be written as f(y)= C1cosh(λy)+C2 sinh(λy)+C3ycosh(λy)+C4y sinh(λy) (3.7) Consequently, the stress function of the given problem is Φ =sin(λx)[C1cosh(λy)+C2 sinh(λy)+C3ycosh(λy)+C4y sinh(λy)] (3.8) in which C1, C2, C3 and C4 are arbitrary constants. Since the loading and geometry of the plate in Fig. 3a are symmetric in both axes, and sinh(λy) and ycosh(λy) are odd functions, the constants C2 and C3 are equal to zero (Barber, 2004). Therefore, the stress components can be determined from Eq. (3.3) as follows σx =sin(λx){C1λ2cosh(λy)+C4λ[2cosh(λy)+λy sinh(λy)]} σy =−λ2 sin(λx){C1 cosh(λy)+C4y sinh(λy)} τxy =−λcos(λx){C1λsinh(λy)+C4[sinh(λy)+λycosh(λy)]} (3.9) When the loading given in Fig. 3a is expanded into the Fourier series, the following equation can be obtained p(x)= p0t l + 2p0 l ∞ ∑ n=1 1 λ sin(λt)cos(λx) (3.10) where the first term gives the mean load. The constants C1 and C4 in Eqs. (3.8) and (3.9) can be determined from the following boundary conditions y =±d ⇒ τxy =0 y =±d ⇒ σy = p(x) (3.11) Substituting the first boundary condition in Eq. (3.9)3, the following equation can bewritten as C4 =−C1 λsinh(λd) sinh(λd)+λdcosh(λd) (3.12) A similar procedure with Eq. (3.9)2 and the second boundary conditions can be carried out for the corresponding constants as C1 = p(x) λ2 sin(λx) sinh(λd)+λdcosh(λd) cosh(λd)sinh(λd)+λd C4 =− p(x) λ2 sin(λx) λsinh(λd) cosh(λd)sinh(λd)+λd (3.13) 612 R. Ince Consequently, the normal stress σx can be derived from Eq. (3.9)1 as σx = 2p0 π ∞ ∑ n=1 ψn(λ,y) n sin(λt)cos(λx) ψn(λ,y)= cosh(λy)[λdcosh(λd)− sinh(λd)]−λy sinh(λd)sinh(λy) cosh(λd)sinh(λd)+λd (3.14) For the case of the concentrated load, the above equation can be converted into the following form since 2p0t = P = const and limλ→0 sin(λ)/λ =1 in which λ = nπ/l σx = P l ∞ ∑ n=1 ψn(λ,y)cos(λx) (3.15) Nevertheless, Equations (3.14)1 and (3.15) are indeed valid for the periodic loadings as indicated inFig. 3b. For this reason, the residual stressesmay naturally occur on the boundaries x = ±l. On the other hand, these series transform the Fourier integrals in the case of l → ∞ in λ = nπ/l and then they can be used for the non-periodic loadings (Girkmann, 1959). By considering ∆λ =(π/l)∆n, Eqs. (3.14)1 and (3.15) can be written respectively, as σx = lim l→∞ [ 2p0 π ∞ ∑ λ=π/l ψ(λ,y) λ sin(λt)cos(λx)∆λ ] = 2p0 π ∞ ∫ 0 ψ(λ,y) λ sin(λt)cos(λx) dλ σx = lim l→∞ [ P π ∞ ∑ λ=π/l ψ(λ,y)cos(λx)∆λ ] = P π ∞ ∫ 0 ψ(λ,y)cos(λx) dλ (3.16) in which, the function ψ(λ,y) in Eqs. (3.16) is the same as ψn(λ,y) in Eq. (3.14)2. From the above discussion, the following data would be the inputs in the analysis: h = 2d = 1mm and P = 1N in the subsequent analysis. The stress analyses have been per- formed for the load-distributed width-to-specimen depth ratios β = 2t/h = t/d = 0, 0.05, 0.1, 0.15, 0.20, 0.25, and 0.29. Figure 4 indicates the elastic stress distribution of σx at x/d = 0, 1, 2 and 3 for β = 0 and 0.25 obtained from Eqs. (3.16) based on the Fourier integrals (the case of the infinitely long strip). In the analysis, the Gauss-Lagurre method with 100 points has been used for the integration procedure since the analytical solution of Eqs. (3.16) might not be available. It is seen from this figure that σx decreases very rapidly with the increasing x and it becomes approximately zero at x/d = 3 for any β value. A similar result for the elastic stress distribution of σy in the middle plane y =0was obtained using a different method based on the Fourier integrals by Filon (1903) again for an infinite beam subjected to two equal and opposite concentrated loads (Girkmann, 1959). In addition, Davies and Bose (1968) simulated the splitting beams with the length/depth= L/h = 3 by using the finite element method for β =0 and 1/12. On the other hand, in the splitting strip specimens, themaximum tensile stress in the plane of loading does not occur at the midpoint (x = y =0), unlike in other splitting specimens. The relativemaximumtensile strength values of the splitting strip specimens and other splitting spe- cimens are given for differentβ values inFig. 5, comparatively. In this figure, the relative location values (y/d) of the maximum tensile stresses are also given for the infinite strip specimens. As shown in Fig. 5, the maximum tensile stress of the strip specimens occurs at the midpoint for β ­ 0.29. Similar to themaximum tensile stress value of the un-notched diagonal splitting-cube specimen, Eq. (3.17) is proposed for themaximum tensile stress value of an un-notched splitting strip specimen in the plane of loading σmax = 2P πbd ( 1 0.7+1.685β0.617 ) β ¬ 0.29 (3.17) The accuracy of Eq. (3.17) is 0.5% for the particular data given in Fig. 5. The fracture mechanics formulas for split-tension strips 613 Fig. 4. Stress distribution of σx at x/d =0, 1, 2 and 3 for β =0 and 0.25 in infinitely long strip Fig. 5. Comparison of the nominal strengths of split-tension specimens 3.2. LEFM formulas Using the boundary collocation method, Isida (1971) proved that the central cracked plate could be practically regarded as an infinite strip when the length/depth= L/h ratio of a plate is ­ 3.Nevertheless, itwas indicated in theanalysis forEqs. (3.14)1 and (3.15) basedon theFourier series for the finite stripwith L/h =2 (Fig. 3b) that the stress distribution of σx on the vertical axis at x =0 was exactly the same as in the case of the infinitely long strip for β = 0 to 0.25. Thismay be explained by Saint-Venant’s principle. The sums of the series were continued to be 614 R. Ince taken until the absolute value of the n-th termwas less than 1e-100. Consequently, the number of terms ranged from 192 (small y values) to 340 (large y values). Meanwhile, similar results were found for certain β values (0.05, 0.10 and 0.20) by Schleeh (1978) who studied on a plate with L/h =2using the Fourier series. Additionally, in this study, it has also been observed that for σx stress distribution along the plane x = 0, the maximum difference between the solution with L/h = 2 and that of L/h = 1.5 was 0.1%. Nevertheless, it will be seen in the following Sections of this work that this difference is muchmore significant for a cracked body. In this study, the analysis of the split-tension strip specimens based on fracture mechanics has been performed for all cases including combinations obtained with α values of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7 and 0.8 and load-distributed width-to-specimen depth ratio values of β = 0, 0.05, 0.1, 0.15, 0.2, and 0.25. A parallel analysis was also performed using the boundary element program (BEM). In all analyses, the stress intensity factor KI was computed usingEq. (3.1). In practice, the polynomial approach was commonly utilized to interpret normal stresses along the crack line σ(y) in Eq. (3.1) (Anderson 2005). Therefore, in this study, the function of σ(y) was computed at specific points including 0, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35 and 0.4mm on the infinitely long strip with h = 2d = 1mm and then expressed as the 8th degree polynomial by using the least squares method for each β value. Green’s function in Eq. (3.2) was used as the weight function m(y,a) in Eq. (3.1), since it was essentially equivalent to the weight function, as discussed above. The integration procedureswere achieved bymeans of theGauss-Chebyshev method. In this study, the stress intensity factor is defined as follows for the splitting strip specimens KI = σN √ πaY (β,α) (3.18) where σN is the nominal stress according to Eq. (2.1), a is half length of the notch, and Y (β,α) is the geometry factor, where β is the relative load-distributed width (t/d) and α is the relative crack length (a/d). The Y (β,α) geometry factors of the splitting specimens are obtained as Y (β,α) = KI/(σN √ πa), where KI is calculated from Eq. (3.1). Fig. 6. Geometry factor Y (β,α) values for split-strip specimens Figure 6 shows the individual geometry factors obtained from the analytical solution; these factors are presented as symbols for each α and β value. The geometry factors have been ge- nerated with the least square method for each β value, as indicated by the curved solid line in Fig. 6. The general form of the selected function is Y (β,α) = A0(β)+A1(β)α+A2(β)α 2+A3(β)α 3+A4(β)α 4+A5(β)α 5 (3.19) The fracture mechanics formulas for split-tension strips 615 where the coefficients Ai (i = 0 to 5) are functions of β, as summarized in Table 1. Equation (3.19) fits all the results from the analytical solutionswith an accuracy of 0.1% for 0.1¬ α ¬ 0.8 and any β value. It is emphasized that this accuracy is valid for the strips with L/h ­ 2, as discussed above. Table 1.Ai and Bi coefficients for split-strip specimen Coefficient β = t/d 0 0.05 0.10 0.15 0.20 0.25 A0 0.7570 0.7656 0.7781 0.7751 0.7534 0.7200 A1 0.5009 0.2618 −0.1872 −0.4591 −0.4724 −0.3435 A2 −2.8063 −1.1558 1.9581 3.8574 3.9428 2.9948 A3 11.4780 6.2913 −3.6245 −9.9595 −10.7060 −8.1581 A4 −16.6660 −9.2409 5.1032 14.5480 16.0060 12.5460 A5 9.9819 5.9303 −2.0490 −7.6288 −8.9671 −7.5169 B0 0.8882 0.8883 0.8851 0.8733 0.8515 0.82158 B1 1.5494 1.4740 1.3054 1.1399 1.0223 0.94892 B2 −6.0402 −5.5487 −4.4507 −3.3906 −2.6901 −2.3399 B3 16.806 15.2700 11.8060 8.4028 6.0978 4.9091 B4 −21.027 −18.8330 −13.8580 −8.9269 −5.5668 −3.8579 B5 11.503 10.2930 7.5156 4.6857 2.6659 1.5512 Fig. 7. Comparison of geometry factor Y values of split-tension specimens for β =0 The variation of the geometry factors is summarized for the splitting specimens, namely: cylinder, cube, diagonal cube and infinite strip for the case of concentrated loading in Fig. 7. TheY functionsof cylindrical andcubical specimenshavebeenobtained fromtheBEMsolutions found in the literature.TheBEMsolutions for splitting strip specimenswithL/h =1.5 and2are also given inFig. 7.Aone-quartermodel has beenused for the plain strain conditions, because of its symmetry in theBEM study. Similar to earlier studies of the author, 250 boundary elements, including 2 crack tip elements, provided reliable results for the crack tip singularity in a one- quarter specimen. The following factors have been the inputs in the analysis: Young’s modulus E =1MPa, Poisson’s ratio ν =0.2, h =2d =200mm, and P =100N. A detailed explanation of theBEMsimulationsmaybe seen elsewhere (Ince, 2012a,b). The accuracies of BEMsolutions 616 R. Ince of the strip specimens are greater than 1.2% for L/h = 2 and better than 7.5% for L/h = 1.5 in comparison with the solution of the infinitely long strip in Fig. 7. On the other hand, the behavior of splitting strips is very different from other splitting specimens, as is clearly shown in Fig. 7. For α =0, the value of Y of strip specimens approach approximately to 0.75, while that of the other specimens approach to 1. Furthermore, the Y function of the strip is approximately parallel to that of the cube. Tada et al. (2000) proposed that Castigliano’s theoremmay be used for calculating opening displacements of the crack surface in a cracked body, as follows COD(y)= 2 E′ a ∫ aF KIP ∂KIF ∂F da (3.20) where KIP is the stress intensity factor due to loading forces, KIF is the stress intensity factor due to virtual forces and E′ = E/(1−ν2) for plane strain and E′ = E for plane stress. In Eq. (3.20), a is length of the notch and aF is location of the virtual force in the vertical distance from the center of the specimen, in which the COD value is computed (Fig. 2). In this study, Equations (3.18) and (3.1) are used for KIP and KIF , respectively. From Eq. (3.20), CMOD values are computed as follows CMOD =COD(y =0)= 2 E′ a ∫ 0 KIP ∂KIF ∂F da (3.21) The CMOD includes not only the elastic constants but also the size of the specimens. The general forms of the specimens are usually described in a polar coordinate system (Tang 1994). Therefore, theCMOD is defined as follows in this study CMOD = πσNa E′ V1(β,α) (3.22) V1(β,α) dimensionless function inEq. (3.22) is calculated bynormalizingCMOD values obtained fromEq. (3.21) with πσNa/E ′ values. Figure 8 indicates individual V1 values for each α and β. Fig. 8. Non-dimensional V1(β,α) values for split-strip specimens The V1(β,α) values are generated by the least squaresmethod for each β, as depicted in Fig. 8. Similar to Eq. (3.19), the following equation is chosen for the general form of the V1(β,α) function The fracture mechanics formulas for split-tension strips 617 V1(β,α) = B0(β)+B1(β)α+B2(β)α 2+B3(β)α 3+B4(β)α 4+B5(β)α 5 (3.23) where the coefficients Bi (i =0,1, . . . ,5) are functions of β that are listed in Table 1. Equation (3.23) fits all the results from the analytical solutionswith an accuracy of 0.2% for 0.1¬ α ¬ 0.8 and any β value. In practice, COD values in a cracked body normalize the CMOD value to determine the profile of the crack surface.Consequently, theCOD/CMOD ratio is independentof the specimen size, but does depend on the specimen geometry and loading type (Tang 1994). In this study, the normalized crack profile of the strip specimen is described by means of regression analysis via a formula similar to the earlier form for the cylindrical and cubical specimens proposed by Ince (2012b) as follows COD(β,y,a) CMOD = √ ( 1− y a )2 + [ 2.367−0.038(1+β)15.9α1.17 (y a )0.961][y a − (y a )2] (3.24) in which y is the vertical distance from the center of the specimen as shown in Fig. 2. The accuracy of the equation is greater than 3.3% for 0.1¬ α ¬ 0.8 and any β value. Equations (3.18) to (3.24) are based on LEFM for split-tension strip specimens. The concre- te fracture parameters for effective crack models (two-parameter model, size effect model and double-Kmodel) could easily be calculated using these equations. The coefficients Ai and Bi for β values, which are not given in Table 1, could be derived by interpolation. 4. Summary and conclusion Recently, split-tension specimens such as cylinders, cubes and diagonal cubes have been com- monly used to determine the tensile strength of cement-basedmaterials. The split-tension strips have been used to determine the fracture parameters of concrete using the effective crackmodels such as the two-parameter model, the size effect model and the double-K in this article. Based on these theoretical and numerical investigations, the following conclusions can be drawn: • The number of theoretical and experimental studies on split-tension strip specimens is limited. Therefore, in this study, a formula for the maximum tensile strength of concrete has been developed for un-notched strip specimens. The results of the analysis reveal that the derived formula is valid for stripswith the ratio of length/depth= L/h ­ 2both for the Fourier integral (the case of the infinite long strip) and the Fourier series (the case of the finite strip). Similarly, it has been indicated from the parallel analysis, which was based on the boundary elementmethod, that the LEFM formulas of cracked strip specimens are valid for strips with L/h ­ 2. • The initial crack of the splitting strip specimens starts at about the location of the loading point, while the initial crack in other splitting specimens starts at the center of the section. However, the initial crack location approximates to the center of the section with the increasing load-distributedwidth-to-specimen depth ratios β, and it is in the center of the section for β ­ 0.29. • In this study, only double symmetrical strips have been analyzed and they have no vertical displacement and no shear stress at the middle line y = 0, as is clearly shown in Fig. 3. Consequently, by considering the upperhalf of the strip (L/d ­ 4), the results of this study may also be utilized in the analysis of a cracked elastic layer resting on a rigid smooth base in soil and rock mechanics. This problem was investigated by Marguerre (1931) for uncracked elastic layers. 618 R. Ince References 1. Anderson T.L., 2005, Fracture Mechanics: Fundamentals and Applications, CRC Press, Boca Raton 2. Barber J.R., 2002,Elasticity, Kluwer Academic Publishers, Dordrecht 3. Basu S., Mandal S.C., 2016, P-wave interaction with a pair of rigid strips embedded in an orthotropic strip, Journal of Theoretical and Applied Mechanics, 54, 2, 579-592 4. Bazant Z.P., Kazemi M.T., 1990, Determination of fracture energy, process zone length, and brittleness number from size effect with application to rock and concrete, International Journal of Fracture, 44, 2, 111-131 5. BuecknerH.F., 1970,Anovel principle for the computation of stress intensity factors,Zeitschrift für Angewandte Mathematik and Mechanik, 50, 9, 529-546 6. Carneiro F.L., Barcellos A., 1949, Tensile strength of concrete (in French),RILEMBulletin, 13, 98-125 7. Davies J.D., Bose D.K., 1968, Stress distribution in splitting tests,ACI Journal, 65, 662-669 8. FilonL.N.G., 1903,On the approximate solution of the bending of a beamof rectangular section, Transaction of the Royal Society of London Series (A), 201, 63-155 9. GdoutosE.E., 1990,FractureMechanics Criteria andApplications, KluwerAcademicPublishers, London 10. Girkmann K., 1959,Plate and Shell Structures (in German), Springer-VerlagWien 11. Ince R., 2010, Determination of concrete fracture parameters based on two-parameter and size effect models using split-tension cubes,Engineering Fracture Mechanics, 77, 2233-2250 12. Ince R., 2012a, Determination of concrete fracture parameters based on peak-load method with diagonal split-tension cubes,Engineering Fracture Mechanics, 82, 100-114 13. Ince R., 2012b, Determination of the fracture parameters of the double-K model using weight functions of split-tension specimens,Engineering Fracture Mechanics, 96, 416-432 14. InceR., GörM., ErenM.E., Alyamaç K.E., 2016, The effect of size on the splitting strength of cubic concretemembers, Strain, 51, 2, 135-146 15. InceR., GörM., Alyamaç K.E., ErenM.E., 2016,Multi-fractal scaling law for split strength of concrete cubes,Magazine of Concrete Research, 68, 3, 141-150 16. Isida M., 1971, Effect of width and length on stress intensity factors of internally cracked plates under various boundary conditions, International Journal of Fracture, 7, 3, 301-316 17. Jenq Y.S., Shah S.P., 1985, Two-parameter fracturemodel for concrete, Journal of Engineering Mechanics, ASCE, 111, 10, 1227-1241 18. MarguerreK., 1931,Pressuredistribution byan elastic layer ona rigid rough layer (inGerman), Ingenieur-Archiv, 2, 108-117 19. Mirsalimov V.M., Hasanov S.G., 2015, Effect of damages on crack development in coating on elastic foundation, Journal of Theoretical and Applied Mechanics, 53, 4, 823-836 20. ModeerM., 1979,A fracturemechanics approach to failure analyses of concretematerials,Report TVBM-1001, Division of BuildingMaterials, University of Lund, Sweden 21. Nallathambi P., Karihaloo B.L., 1986, Determination of the specimen size independent frac- ture toughness of plain concrete,Magazine of Concrete Research, 38, 135, 67-76 22. Neville A.M., 2011,Properties of Concrete, Pearson Education Limited, Essex, UK. 23. Nilsson S., 1961, The tensile strength of concrete determined by splitting tests on cubes,RILEM Bulletin, 11, 63-67 The fracture mechanics formulas for split-tension strips 619 24. Rice J.R., 1972, Some remarks on elastic crack-tip stress fields, International Journal of Solid Structures, 8, 751-758 25. RoccoC.,GuineaG.V.,Planas J., ElicesM., 1995,The effect of the boundary conditions on the cylinder splitting strength,Proceedings of FRAMCOS-2:Aedificatio Publishers, 75-84,Freiburg, Germany 26. Schleeh W.S., 1978, Determination of the tensile strength of concrete (in German), Beton, 28, 2, 57-62 27. TadaH., Paris P.C., IrwinG.R., 2000,Stress Analysis of Cracks Handbook, ASMEPress,New York 28. Tang T., 1994, Effect of load-distributed width on split tension of unnotched and notched cylin- drical specimens, Journal of Testing and Evaluation, ASTM, 22, 401-409 29. Tang T., Ouyang C., Shah S.P., 1996, A simple method for determining material fracture parameters from peak loads,ACI Materials Journal, 93, 2, 147-157 30. TangT., Shah S.P.,OuyangC., 1992,Fracturemechanics and size effect of concrete in tension, Journal of Structural Engineering, ASCE, 118, 3169-3185 31. Timoshenko S.P., Godier J.N., 1970,Theory of Elasticity, Mc-GrawHill, NewYork, USA 32. Tweed J., Das S.C., Rooke D.P., 1972, The stress intensity factor of a radial crack in a finite elastic disc, International Journal of Engineering Science, 10, 323-335 33. Xu S., Reinhardt H.W., 1999, Determination of double-K criterion for crack propagation in quasi-brittle fracture, Part I: Experimental investigation of crack propagation, International Jour- nal of Fracture, 98, 2, 111-149 Manuscript received November 9, 2016; accepted for print December 12, 2016