Int. d. Mech. Sci. Voh 38, No. 10. pp. 1073 1088, 1996

Copyright F 1996 Elsevier Science Lid Printed in Great Britain. All rights reserved 0020-7403/96 $15.00 + 0.00





Room D310, Wilton Centre, ICI Engineering Technology, Middlesborough, Cleveland, TS90 8JE, U.K.

(Received 20 March 1995, and in revisedform 20 September 1995)
AbstraeI--Dynamic crack propagation in tapered double cantilever beam (TDCB) specimens is analysed via beam theory and the finite element method. Steady state and transient solutions of the energy release rate G are given for various load conditions. Finite element analysis is performed to obtain the dynamic G at given crack speed or the crack history for a given fracture toughness. The stress wave effects on the dynamic G are discussed. The beam solutions are compared with the finite element results and some experimental phenomena are explained. Copyright © 1996 Elsevier Science Ltd.

Keywords: energy release rate, tapered double cantilever beam, steady state crack growth, fracture toughness, dynamic effects.

NOTATION G~ G Gc a a0 d I E p C fl [k] U~ Uk Uox, c static energy release rate dynamic energy release rate fracture toughness crack length crack length at initiation crack speed second moment of area Young's modulus mass density longitudinal wave speed of materials, C = v/(E/p) Rayleigh damping factor stiffness matrix strain energy kinetic energy external work done by the applied load compliance flexural wave speed correction factor on crack length time applied load applied displacement 1. I N T R O D U C T I O N

A t P vo

The double cantilever beam (DCB) specimen is one of the most commonly used test configurations for measuring fracture toughness of composites, polymers and adhesives. Some theoretical analyses of the DCB specimen, via either simple beam or shear beam theories, can be found in previous literature [1-4], and solutions based on the energy balance have been given for rapid crack propagation and arrest. The application of beam theory to dynamic crack propagation is particularly attractive because it is a one-dimensional analysis and the inclusion of the time variable does not introduce further mathematical difficulty. Although the beam analysis cannot predict the details of crack tip stresses or strains, it does provide an accurate account of energy quantities which form the basis of the Griffith fracture criterion. Parallel to the analytical research, the finite element method provides a powerful alternative of analysing most real crack propagation specimen configurations. Finite element analyses have been carried out to simulate dynamic fractures using the node release technique [5-9]. Detailed singular stress or strain fields and energy quantities may be obtained to a great extent of accuracy provided sufficiently fine mesh is used. Efforts have also been made to validate the numerical results by

From . a clamped boundary condition is assumed which moves at a certain crack speed. This problem becomes more pronounced in composite tests where specimens are usually very slender. it was found that the energy release rate G in the DCB specimen is intrinsically prone to the dynamic effects due to its low flexural stiffness of the arm. very high crack speeds can be achieved by applying relatively slow loading rates. hence less the stress wave effect. T H E O R E T I C A L Owing to symmetry. Instead Berry's method [1] is used to obtain an approximate solution where the dynamic energy terms are calculated on the basis of equilibrium field. The solution for fixed displacement rate loading is particularly useful because experiments are usually carried out under this condition. G. 2.0 (1) where I = Bh3/12. attention is focused on the analysis of dynamic crack growth in the TDCB specimen. At the crack tip. Also because of its large stiffness.1074 Y. the relationship between the fracture toughness and crack speeds cannot be found easily. The close form solution to Eqn (1) with a moving boundary is somewhat intractable. Wang and J. Simple beam solutions for steady-state and transient crack propagations are given for fixed load. 1). as demonstrated by Wang and Williams [9]. In this paper. Uk the kinetic energy and Uoxt the work done by the applied load. 2.000 m s. only half of the specimen is analysed. the measured load will be constant. For example. However the crack velocities they tested were very low compared to the wave speed of the material ( ~ 10. The facts revealed in the paper provide a useful explanation to some experimental observations [15]. The equation of motion for the beam under bending is Ox 2 E1 8xZj + p A ~ 2 -. This is proved later in this paper by a two-dimensional finite element analysis. As a result. Williams. i.l . In fact. If the crack begins to grow from its initial length ao = 0 at a constant speed & a steady state crack growth can be assumed on the basis that the crack has all the dynamic information available during propagation.e. this test configuration has been adopted by the ASTM standard [10] for testing adhesive strength under static conditions (Fig. 12]. a concentrated load P is applied (Fig. examining some simple geometries such as finite strip or parallel strip where the exact solutions are available.1) where the dynamic effects were negligible. Steady-state crack propagation First. E is the Young's modulus. At the loading point. Some analyses of TDCB tests can be found in previous literature [11. For a material of a constant fracture toughness. 13] which is applicable for both isotropic materials and composites. B the uniform width of the arm and h the thickness of the arm. The intention is to improve the understanding of the specimen behaviour at crack speeds of a fraction of the characteristic wave speed of the material.1. h 3 = 3xZ/m where m is a constant. the compliance becomes linear to crack length and the energy release rate G can be expressed as a function of the measured load only. A = Bh. It is therefore suggested that the tapered double cantilever beam (TDCB) should be used for its large stiffness of the arm. 1). The effects of transverse shear and rotation of the beam are considered by using the beam-on-elastic foundation model [-3. Through these researches. and assumed a logarithmic polynomial expression to fit the experimental results to correlate fracture toughness and crack velocity.~Uext G =~\ 8a c~Us OUk~ 8a ~a / (2) where Us is the strain energy. The dynamic energy release rate G is then determined through the total energy balance 1 (. fixed displacement and fixed displacement rate loading. By making a contour profile on the arm. p the density of the material. dynamic crack growth in the TDCB specimen under a fixed displacement Vo is examined. A finite element analysis is also performed for the purpose of comparison. Yaniv and Daniel [12] used TDCB specimens to investigate the rate sensitivity of delamination in composites at crack speeds up to 26 m s .

S U \ ~ / d x EBv~ 8ma (8) In the case of the fixed displacement loading. Ue~t = O.] dx (3) 3 = H /d'~2/3~113 EBv2[-C)[maa) (7) where C = x/(E/p) the longitudinal wave speed of the material. the displacement profile of the arm can be expressed as v = v o l X l n ( X ) + (1 . simple beam theory with the effect of shear deformation ignored.x ) l where Vo is the end deflection under the concentrated load P Vo = ~ (3) 4raP a (4) and the compliance of the arm c = vo/P is linearly proportional to the crack length a.Dynamic crack growth h 3 3x z = m 1075 . the dynamic G is determined G--~= 1 + mh(a) (9) . 1. From Eqn (2). If only the motion of crack advance is considered. m=2 P h B P V lx Fig. Tapered double cantileverbeam specimen. The strain energy U~ in the arm is ( aEBh3(~2y~2 c~ = jo . the velocity distribution along the arm is ~tand the bending curvature of the arm is 7 / (5) ( ? 2 v Vo (6) c~x2 The kinetic energy Uk in the arm is Uk= ax" o 2 \~t.

The contribution of the kinetic energy is characterized by the second term in Eqn (9) which can be rewritten into the usual form used in dynamic fracture as G = 1 + G~ where Cv is the flexural wave speed 11312C (12) Cv . which can be determined from static equilibrium Tm and at x = 0 ~V (18) V ~a d (19) . If d is constant. Wang and J.. and the displacement profile along the arm is v=vtlXln(X)+(l_X)].67 h(a) for isotropic materials. the solution for a steady-state crack growth can be obtained in a similar manner.1076 Y. then the velocity distribution is Ot and the curvature of the arm is V 1 (14) (13) 02V V OX xa" 2 The kinetic energy Uk and the strain energy Us in the arm are. Equation (10) can be rewritten as E ( Vo ~2 (11) where A = 0. where h(a) is the beam thickness at the crack tip. Williams. respectively 27 (15) (16) Us= 8m\d/" (17) There is the work done by the applied load. G~ is the static energy release rate E~V0~ 2 Gs = 8 m \ a ) " (10) The effects of shear deformation and root rotation at the crack tip may be included by applying a correction factor A on crack length [3.. 13]. G. Here the end deflection Vo = Vt./(72 mh(a))" For the beam under a constant displacement rate V.

in this circumstance. giving a bound to the behaviour. in which case the static energy release rate can be written as 2raP2 GsEB 2 . given a constant Gc = 1 kJ m . For example. Therefore.1. and . it is noticed that given a constant crack speed and a constant displacement rate. (26) 2. If the crack initiates at a finite length. Hence we have = "x/~Oc) which is the static solution Eqn (21). In reality. the crack speed will initiate at 132ms -1 and gradually reduce to 126ms -1. Therefore the solution also applies to the beam under a fixed load. To achieve a steady-state crack growth with a constant G.Dynamic crack growth From Eqn (2). Transient crack propagations The steady-state solutions given above are valid only where the crack initiates from zero length. for tests of materials of low Gc-values and high crack speeds. the dynamic G becomes transient as a result of the delay of the stress wave reflection from the loading end.z.+ l l C : ]" It shows that given a fixed loading rate V. for a 300 mm long aluminium bonded joint under a fixed loading rate of V = 2 m s. the crack speed depends solely on the fracture toughness Gc of the material at that crack speed.~ . the applied load is constant. in which case the crack tip is constantly informed about the change of the boundary condition by the continuous waves reflected from the loading end. The dynamic G will drop from its initial value due to the increase of the kinetic energy in the beam. the dynamic effects cannot be ignored and Eqn (24) must be used to correlate crack speed and fracture toughness. From Eqn (18).3~/(mh{a)) {22) This velocity of flexural waves is crack length dependent. The limiting case is d --+CF. The relation between crack length and time can be estimated via Eqn (20) which can be rewritten as 9 a 2 G=~m m For G = Gc where Gc is constant. the second term in the square root in Eqn (24) can be ignored and a constant crack velocity is a good approximation provided that the material's rate dependency on such small crack speed variation is negligible. crack speed must slow down gradually. However. we have /{8mG¢ 9mh(a}~ (24) t = a 4 t . Equation (20) indicates that the dynamic G decreases at a constant crack speed as the beam thickness h(a) increases with crack length.E .2. it is difficult to detect and measure such a small change and a constant crack velocity can be reasonably assumed. the dynamic G is 1077 G-~ G=l_9 -~ mh(a} (d) 2 (20) where Gs is the static energy release rate G~ = ~ and the characteristic fiexural wave speed is {21) ~/(11)c c v .

The solution can be expressed in terms of a series. a "holding back force" F is postulated at the node immediately adjacent to the crack tip node. The crack growth is simulated by releasing the nodal constraints along the pre-defined crack path at a given crack speed.J'.]". and Cv is determined from Eqn (22) at a = ao. For the next step. The interval for the first wave arrival at the crack tip is given by a. ~ = a/Cv and Cv is determined from Eqn (12) at a = ao.IX. transient behaviour is again observed when the crack initiates at a finite length..J'M'M N. In this node release technique.. G. In the finite element analysis.~<1 ~ x.a ". ~2 in Eqn (27) is negligible.N". the static solution is always obtained prior to the crack initiation.~)) G=Go(1 +az) 1 -~) (27) where Go = E/8m(vo/ao) 2. A s c h e m a t i c finite element mesh of T D C B specimen. . Equation (27) becomes analogous to the transient solution of a parallel strip under axial loading [7].. = ao ~ _ ~) (28) where ~ = d/Cv. Some modifications were made for the present calculations.d"a "4~__M_"4" . 3. giving a step-wise decrease of the dynamic G. A two-dimensional plane stress FE model for the TDCB specimen (Fig. the equation of motion in the two-dimensional domain is solved by an iterative time integration scheme. and decays linearly from the initial value to zero as the crack tip propagates from the released node to the crack tip node. Repeating this procedure gives G-values at various crack lengths.I'MM'M'M ". W a n g a n d J. 2) was set-up with a total of 3208 d.a~".4 ". 2. Under fixed displacement rate control.f. as expected since both the geometries have a linear compliance with respect to crack length.s'4 ~'. However the transient solution cannot be obtained in this case apart from that the dynamic G will oscillate about the steady-state solution as a result of wave reflections. Given a fixed displacement Vo.4"4". Williams. . [-7]. when ao < a < al.[/]x.J'Q". When ~ is small.r'. Linear triangular elements with equally distributed lump masses were used with in a nominal element size of 2 mm. N U M E R I C A L A finite element analysis of high speed TDCB tests was carried out using a F O R T R A N programme written by Keegstra [5]. ~ . the transient energy release rate at each step can be estimated by averaging the steady-state solution during the wave reflection period. ao takes the value of al and a new G-value can be determined. then remain more or less constant until the first reflected wave arrives. The dynamic G is then computed through the total energy balance and the work done by the "holding back force". Following the method used in Ref. Once the crack starts to propagate. The static G-values were also calculated via the virtual crack closure method [14] and the results were compared with that of the beam analysis.s'. The dynamic G-values were computed for both steady-state and transient crack propagations discussed in the early section.]M"OM'MM'M'M "xlM'NI'M"xlN N N NI'M N I ' x l N f M _ ~ N N N N ~ r'.. a l = a0 ((1 + ~)/(1 . ! ~ _ " 4 _ " 4 _ " 4 ~ ~ L=300 mm Fig. Both the schemes should give the same G-values. Small viscous damping was applied in the region of the beam ahead of the e j v0 A l . Another drop of the G-value will take place to accommodate the new boundary condition brought back by the stress wave. And this behaviour is repeated for further crack extension.1078 Y.

the solution will stabilize at a > 50 mm. The damping takes the simple form of/3 [k].1.Dynamic crack growth Table 1.2. where [k] is the stiffness matrix and/3 the Rayleigh damping factor. The details of the material properties and dimensions are given in Table 1. the applied load is constant.1 was appropriate. The force at the loading point computed by the finite element compares very well with that of the beam solution. The dissipated work done by the damping force was also included in the total energy balance for G calculations. Static analysis First a static stress analysis was carried out via the finite element method. It was found that /3 ~ 0. This is clearly due to the change of the boundary condition. As mentioned earlier. L. The boundary conditions of the beam are shown in Fig.5 Cims 1) 5073 L(mm) 300 1079 Note: the constant rn = 2.5a2) l/a B(mm) 12. Again good agreement between the finite element and beam theory was found. the compliance of the beam is 4m c = g k a. In general. 4. On the other hand. Dynamic crack propagations at speeds ranging from 100 to 1000 ms -1 are examined in order to see the dynamic effect. for which is often used in adhesive joint tests. The longer the crack length. According to Eqn (20). 4. However as the crack approached to the free end (a > 250 mm). 2. the initial crack length ao was set to a very small value (one element length) to approximate the steady-state condition (ao = 0). is 300 mm. the better the agreement. the dynamic G vanishes at a ~ 50 mm. 500 and 1000 m s 1). and the thickness h varies along with crack length. It can be seen that the steady-state solution is indeed a good approximation to the behaviour. Three sets of the finite element results for dynamic crack growth under fixed displacement rate control are also presented (Figs 8-10). The compliance of the beam was calculated at various crack lengths with a fixed displacement vo = 0. When the remaining ligament on the crack path gets very short. under fixed displacement rate control. the assumption of the clamped end condition is no longer valid and large shear deformation and rotation will occur. the discrepancy at very short crack length (a < 30 mm) can be attributed to: (1) beam theory is not applicable for short beam as the effect of shear deformation becomes more pronounced. The results are presented in Figs 5-7 in terms of G/Gs vs a. resulting in a sharp increase in the compliance and static G-values (Fig. a is the crack length. which is also clearly predicted by the finite element (Fig. The discrepancy immediately after the onset of crack growth is due to possible spurious stress waves generated by the irregular finite element mesh in the region as well as the reasons mentioned in the previous section. the finite element predicted good compliance values even at short crack lengths (Fig. From beam theory. at fi = 1000 m s. The total length of the specimen. Steady-state analysis The dynamic G-values of steady state crack growth under fixed displacement control were computed at three different crack speeds (200.3 p(kgm z) 2720 h(a)(mm) (1. The material of the TDCB specimen in the present analysis is an aluminium alloy. (2) triangular finite elements tend to be over-stiff and (3) only a small number of elements employed in the region will usually underestimate the beam compliance. R E S U L T S 4. 10). 3). 4). a large discrepancy was seen.1. The beam solution shows that for crack speeds less than 200 m s-1 the behaviour is quasi-static. Material properties and dimensions of the specimen E(GPa) 70 v 0. In the computation. (a9) Compared with Eqn (29). They cover the range of crack speeds usually seen in practical tests. as shown in Fig. crack tip to reduce the effect of stress waves reflected from the free end [9].5 mm applied at the loading end. I1 for .

~ = 200 rn/s ----o-- Finite element Beam theory a~ I . v o = 0. . Comparison of compliance of TDCB specimen. J . 5. Fixed displacement. Comparison of static G in TDCB specimen. G. Williams.1080 6- Y. 32 1 ) f ~ L I 0 0 100 200 300 a (ram) Fig. 5- 4. Comparison of dynamic G-values. a o = O. 4. Wang and J.5 mm. 12 10 ~= 6- 4 j 100 200 :1 0 300 a (mm) Fig. i 100 a (rnm) 200 300 Fig. 3.

5Fixed displacement. h = 1000 m/s 4.5 mm.5 mm. 5Fixed displacement rate.Dynamic crack growth 5Fixed displacement. • . 8. v o = 0. tl = 500 m/s 4. . 6. V = 5 m/s. v o = 0. ----o-Finite element Beam theory 3 O 100 200 300 a (ram) Fig. Comparison of dynamic G-values.i = 200 m/s ----o-Finite element Beam theory 3t5 2O 0 ! . Comparison of dynamic G-values. --~o-Finite element Beam theory 1081 3. a o = O. a o = O. 7. . 0 1 O0 200 300 a (ram) Fig. 0 0 I 1 t 100 a (mm) 200 300 Fig. r3 2. . . Comparison of dynamic G-values. a 0 = O. .

5 J m -2 and a constant displacement rate V = 5 m s-1 is applied. 3Fixed displacement rate. Comparison of dynamic G-values... A different finite element a p p r o a c h was also carried out to c o m p u t e crack speed variation at a given G-value. as shown in Fig.o-Finite element Beam theory L3 o ~ * '!'"L' '" i . The oscillation at the beginning and near the end of the beam is due to the b o u n d a r y effects. V = 5 m/s.. then the crack speed by E q n (24) will decrease from 500 m s . This case can be c o m p a r e d with the previous one with a fixed loading rate V = 5 m s . An example is shown in Fig. fi = 500 m s-1. Figure 12 shows the G-values from beam theory and the finite element for ci = 500 m s ..1082 5- Y. If the fracture toughness is assumed to be a constant Gc = 437.1 and a constant P = 1093 (N)...~.3331331 1 O0 a (mm) 200 300 Fig. Wang and J...1.... In order to achieve a constant G crack growth. This argument is further confirmed by the finite element result with fixed load control. Fixed displacement rate. /= = 500 m/s 4----o-Finite element Beam theory kb 0 100 a (mm) 200 300 Fig.1 and fi = 500 m s .. .. G. the crack speed has to decrease according to E q n (24). 14.5 J m -2 and loading rate ~ . The dynamic G c o m p u t e d by using this varying crack speed..1 to 320 m s . 15 with a fixed Gc = 437. Comparison of dynamic G-values.. Williams. V = 5 m/s.1..5 J m 2. ~ = 1000 m/s -~.. as both have the same initial G-value but one is a constant G crack p r o p a g a t i o n and the other has a constant crack velocity. a o . 10. is indeed approximately equal to the constant value of 437. Figure 13 shows that the displacement rate at the loading point calculated by the finite element is identical to that of the fixed displacement rate case of V = 5 m s .. 9. ao ~ O.-= O.

i 0 100 a (ram) 200 300 Fig. ao = 0.2000 v "1o 1083 Finite element Beam theory q 1000 0 100 a (mm) 200 3OO Fig. 11. 13. 5Fixed load. Comparison of applied load. ~ = 500 mls 4 3 E v 2 '1 0 • - .Dynamic crack growth 3000 Fixed displacement rate. 12. & = 500 m/s ---o-. P = 1093 N. Comparison of dynamic G-values. . End displacement vs crack length. 5Fixed load. P = 1093 N. & = 500 m/s ---e-Finite element Beam theory Lb 2- 0 0 100 a (ram) 20O 300 Fig. V = 5 m/s. a o = 0. ao = 0.

Williams. . . 0. 14. ao = 0.2 0. ~i varies (equation 24) ----o--Finite element Beam theory 0 ~ + • t 0 100 a (mm) 200 300 Fig.6 0.1084 5- Y. V = 5 m/s. Fixed d~splacement rate. Comparison of crack length vs time.5 ram. vo = 0. .B Fig.0 0. 400 Fixed displacement. Comparison of dynamic G-values. & = 100 m/s ---o-300 i Finite element Beam theory gb o L~ 200 100- 0 0 100 a (mm) 200 300 Fig. 400 ----o-- Equation (24) Finile element 300 2oo 100 0 • n T . 16. ao = 60 ram. Wang and J. G. 15. Comparison of G-values.4 t (ms) 0.

ao = 60 ram. V = 2 m/s.. 6000 Fixed displacement rate. 18. ~i = 100 m/s 5000 ----o-Finite element Beam theory 4000. % = 0. MS 38-I0-D . Comparison of dynamic G-values.Dynamic crack growth 400 Fixed displacement. 19. & = 200 m/s ----o-300. o Finite element Beam theory 1085 N--. E 200 100 0 0 100 a(mm) 200 300 Fig. o 2000 % o 1000 100 200 300 a (mm) Fig. Comparison of dynamic G-values. ao = 60 ram. a o = 60 mm. 3000. 17. Comparison of dynamic G-values.5 mm. &. f = 200 m/s 5000 ----o-4000 N" o Beam theory Finite element Wave intervals • E 3000 2000 1000 1 ~ T 100 a (mm) 200 300 Fig.. 6000 Fixed displacement rate. V = 5 m/s.

4000 o Finite element Beam theory 3000 0 _J 2000 1000 i i i 100 200 300 a (mm) Fig.3 0. 6000 Fixed displacement rate.7 l (ms) Fig. Comparison of the end load. V = 5 m/s. ao = 60 ram. a o = 60 ram. Comparison of dynamic G-values. V = 5 m/s. Williams.6 0. 20. & = 200 m/s 5000 ~ .4 0. G. 2O00 1800 1600 14 O0 1200 ---o-Fixed displacement rate. Comparison of crack speed.2 0.Finite element L~ 300~ 20O 100 0 100 a (mm) 200 300 Fig. . f = 500 m/s Beam theory ~ .0 0. 21. ao = 60 mm Equation (24) Finite element ~1000 800 '°° o 6oo~ 40O 2O0 l i ooo o ooo o l i ~ i i i 0 0. V = 5 m/s.1086 1000 Y. 22.1 0. 90 01 800J Fixed displacement rate. Wang and J.5 0.

Transient crack propagation will occur and its effect becomes more pronounced at high loading rates. Figure 22 shows the finite element prediction of the crack speed at V = 5 m s . as shown in Fig. showing a good agreement. 19). ~ . . The steady state crack speed predicted by Eqn (24) is also plotted in the same figure. the end load exhibits a large oscillation. 21 for V = 5 m s.1 0.e. The wave reflection intervals can be determined by Eqn (22). which appears to be slightly lower than the average crack speed computed by the finite element. the quasi-static solution. To examine this transient effect. a finite initial crack length of 60 mm was used with both fixed displacement and fixed displacement rate loading applied.2 o.. i.0 0.5 mm. A comparison of crack length vs time relation is given in Fig. given in terms of crack length vs time.02C. Two finite element results are given in Figs 16 and 17 at crack speeds of 100 and 200 m s-1 with a fixed displacement vo -./ " ~ . Now if a fixed fracture toughness is given.a 0.. Nevertheless the steady state solution appears to give the mean value of the transient G. crack velocity can be predicted at a given loading rate. Comparison of crack length vs time. (4) Under fixed displacement rate control. C O N C L U S I O N S (1) The analysis provides a simple solution of dynamic crack propagation in TDCB specimens which seems to give a good correlation compared to the finite element results. G decreases at a constant ~i. then ~i will decreases. 200 and 500 ms -~. The relation of crack length vs time is given for a constant G crack propagation by beam theory.4 0. Transient analysis In practical tests. 5. Similar bebaviour was seen for the cases of fixed displacement rate control (Figs 18-20) at fi = 100. It can be seen that the dynamic G oscillates as a result of the wave effects but the trend follows that of the transient beam solution.--o-- / i / / . the dynamic effects are small.0. is compared with Eqn (24).t. (2) The dynamic effects are negligible for ~i < 100 m s.. (5) For ci < 200 m s.5 J m-2.Dynamic crack growth 500 Fixed 400 1087 displacement rate. resulting in unstable crack growth (a > 250 mm) as seen in experiments.3.t. This is typically seen in high rate experiments.1 in aluminium specimens.. (3) Both the static and dynamic G-values exhibit a sharp increase when the crack approaches to the end of the specimen.6 0.t with a fixed Gc = 437. 300 Equation (25) //f///~ ~1 0"0 g 200 E 100 0 0. This can be easily identified by the severe oscillation of the measured load. 4. Analytical solution is not available in this case. ao = 60 mm -. 23.5 0. ti < 0.7 t (ms) Fig. V = 5 m s. If Gc is constant. The result..t and ci = 200 m s. The fracture toughness can be approximately estimated by measuring the average crack speed [Eqn (25)]. 23 which shows that in high rate tests. Eqn (24) gives a better prediction of the crack history than Eqn (25).. V Finite e l e m e n t E q u a t i o n (24) = 5 m/s.t. Or if the fracture toughness is known. the crack speed will accelerate and decelerate as the driving force G changes with time. As expected. we usually initiate the crack at a finite length. which have an obvious effect on the dynamic G-values (Fig.

Useful discussions with colleagues are acknowledged. Blackman. U. S. Swansea. Wang. Kanninen. J. ASTM STP 972 (edited by J. J. University of Swansea.. Freund. Blackman and Mr A. 12. Numerical investigations of rapid crack propagation.D.K. 25. Numerical studies in dynamic fracture mechanics.R. Yaniv and I. 8. An analysis of tapered double cantilever beam fracture toughness test for adhesive joints. pp. Technology and Medicine. U. Owen). E. M. Phys. Mech. D3433-75 (1980). Fract. The wave speed can be calculated and the wave effect intervals predicted. Mech. M. 415~431 (1974). R. 25. Phys. J. Equations of motion at constant deformation. P. Taylor. Standard practice for fracture strength in cleavage of adhesives in bonded joints. R. 367 376 (1989). 10. R. 617 631 (1993). B. 660-672. Wang. A. Thesis. American Society for Testing and Materials. REFERENCES 1. Daniel. 77 83. (1995). Solids. ASTM STP 677 (edited by C. Testing and Design. N. Computer studies of dynamic crack propagation in elastic continua. 323 331 (1994). (6) The transient behaviour of crack growth initiating from a finite length is governed by the velocity of flexural wave propagation. J.. Numerical Method in Fracture Mechanics (edited by A. A dynamic analysis of unstable crack propagation and arrest in the DCB test specimen. P. Rybicki and M. G. Imperial College of Science.F. Y. Fract. Keegstra. F. Wang and J. Smith). J. on Structural Adhesives in En~lineering. 5. J. Berry. American Society for Testing and Materials. Rydholm. 207-216 (1960). Mech. K. G. Composite Materials. Berry. 8. Philadelphia (1979). Kinloch and Y.. Williams. Y. . Fredriksson and F. 35.. particularly those with Dr B. G. ASTM. 27. Mech. Williams. 13. Mech. Engn9 Fraet. 11. Taylor. The higher the crack speed. Wang and J. Composites. 7. Dynamic G calculations for an axially loaded parallel strip. 6. Int. Int. Williams. B. 245 261 (1985). the larger the transient effect on the G-values. L. Technol. J. 2. Engng Fract. Solids. Some kinetic considerations of the Griffith criterion for fracture--l. Some kinetic considerations of the Griffith criterion for fracture--II. Philadelphia (1988). pp. G. B. Bristol. S. (1978). 46. F. 10. A finite element calculation of stress intensity factors by a modified crack closure integral. 15. J. G. Nilsson. Williams. Composites Sci.S. Fracture Mechanics. pp. W. D. C. A. A simple model of the double cantilever beam crack propagation specimen. Height tapered double cantilever beam specimen for study of rate effects on fracture toughness of composites.1088 Y. 194~206 (1960). Phys. 931-938 (1977). 9. J. Equations of motion at constant force. End corrections for orthotropic DCB specimens. Luxmore and D. Acknowledgements~he authors wish to thank EPSRC for providing the financial support for the project. 4.N. London (1977). Proceedings of the 4th Conf. G. 69-79 (1977). The impact testing of adhesive joints. Solids. Kanninen. Whitcomb). A numerical study of dynamic crack growth in isotropic DCB specimens. P. J. Atluri and T. 9. Nishioka.K. 651-667. 8. 241 258. pp. Wang and J. Ph. 14. 3..