Electronic Journal of Structural Engineering (8) 2008

Computer Simulation of Concrete Structures subjected to Blast and Impact

Tuan Ngo Priyan Mendis

Dept. of Civil & Environmental Engineering, The University of Melbourne, Parkville, VIC 3010, Australia ABSTRACT: Computer simulation of the real behavior using the finite element method for nonlinear transient dynamic analysis of reinforced concrete structures under impulsive loading is presented. As a result of the impulsive loads reaching their peak intensity in extremely short durations of time, material nonlinearities and strain rate effects play an important role in the dynamic analysis. In this study, a numerical approach based on an explicit finite element formulation is proposed for the analysis of two-dimensional planar structural problems under impulsive loads. A two-dimensional behavioral rate dependent lattice model (RDLM) capable of analyzing reinforced concrete members subjected to impulsive loading is developed and implemented in the computer program, named RC-IMPULSIVE, for the finite element nonlinear dynamic analysis combined with a tensile crack monitoring algorithm. The opening of new cracks as well as the closing and reopening of existing ones can be traced and is activated after total and elastic evaluation for each sampling point at each time step. A parametric study is carried out on reinforced concrete structures subject to impulsive loading to validate the explicit code and is shown to be in good agreement. Keywords: Concrete, simulation, impulsive loading, lattice model, finite element, high strain-rate 1 INTRODUCTION Despite significant progress having been made in the development of sophisticated computational methods, the accurate prediction of failure modes of reinforced concrete structures under severe impulsive loads is very difficult, due to the complexity of concrete behavior in the high strain-rate domain. As a result of the impulsive loads reaching their peak intensity in extremely short durations of time, material nonlinearities and strain-rate effects play an important role in the analysis. Most current design and analysis methods are based on simplified equivalent static forces or single degree of freedom systems. Although this approach may be suitable for preliminary analysis, it contains inherent inconsistencies. The finite element method provides an effective numerical approach for the modelling of structural problems involving severe impulsive loading. It is recognized that numerical simulations play an increasingly important role, replacing more costly experimental studies. Computer analysis program can be very useful in this endeavour as long as the accuracy and the limitations of the program chosen with respect to areas such as constitutive and failure models are well understood. Thus, one of the main aims of this study is the development of a finite element computer code to predict the nonlinear structural response of two-dimensional concrete structures subject to impulsive loading. The rate dependent lattice model (RDLM) for concrete and steel, developed in this study, is based on a strain rate sensitive concrete and reinforcement material model in tension and compression and is suitable for the two-dimensional analysis of RC structures. Major sources of nonlinearities in reinforced concrete structures has been taken into consideration, such as the progressive cracking of concrete in tension, the inelastic response in compression, the yielding of reinforcing steel, and strain-rate sensitivity of concrete and steel reinforcement. Material nonlinearities including crack propagation, concrete crushing, post-failure residual strength, as well as the strain rate effects on the response of concrete and steel can be modelled by the proposed explicit code RC-IMPULSIVE developed in this study. The explicit code is validated using experimental test results from the Woomera blast trial and other test data available from literature. Numerical analysis and a parametric study of reinforced concrete structures subjected to impact and blast loading were also performed using the computer code. The capability of program RC-IMPLUSIVE in predicting accurately the nonlinear dynamic response of concrete structures under impulsive loading has been proven. 2 FINITE ELEMENT FORMULATION

90

Electronic Journal of Structural Engineering (8) 2008 A comprehensive computer program, RCIMPULSIVE, for the finite element analysis of two dimensional plane stress reinforced concrete structures subjected to impulsive loads has been developed in this study using the explicit central difference integration technique and the proposed strain rate dependent lattice model (RDLM). The program is written in FORTRAN in modular form and possesses sufficient flexibility to add new options resulting from further research. Detailed description of the program can be found in Ngo (2005). The following sections will provide a brief description of the implementation procedure for the finite element equations and constitutive relationships of concrete and steel. 2.1 FE discretization Using the Lagrangian approach and the principle of virtual displacements the equation of motion has the general form as follows:
+ Cu + P(u ) = F Mu

For explicit schemes the Rayleigh damping is used as C=b0Mn. Using the central differential approximation with some mathematical manipulations, Eq. (5) can be rewritten as:

n +1 = a0 M 1 ( Fn +1 Pn +1 ) a1 (u n +1 u n ) u (6) in which a0 and a1 are integration constants expressed as: b 2 ; a1 = 0 a 0 a0 = t 2 + b0 t

2.3 Numerical procedure of the explicit algorithm At each time step in the solution procedure, with the known displacements from Eq. (2) and the element relations in conjunction with the constitutive relationships are used to compute the nodal internal forces for that time. Then, these forces are employed in the equation of motion to compute the accelerations which are integrated subsequently to find the new displacements. The time step must be small enough to be consistent with the numerical stability limits. The energy balance check is used at each time step to monitor the solution stability. 3 FE MATERIAL MODELLING The rate dependent lattice model (RDLM) for concrete and steel, developed in this study, is implemented in the computer program RC-IMPULSIVE. For modelling concrete, a crack monitoring algorithm to control the cracking process, and a crushing control routine have been included to cover the brittle nature of concrete. This algorithm is activated after strain evaluation in each time step and for each cracked integration point. The RC element is numerically simulated using four-node quadrilateral plane stress elements with 2 2 gauss points of integration. The elements are based on the quadratic interpolation where the basic variables in the nodes are the translations in local element directions. The reinforcement is modelled as one-dimensional 2node discrete truss element in order to account for the bond slip. The constitutive relationships of concrete and steel reinforcement are described in the following sections.

(1)

in which M is the global mass matrix, C denotes the global damping matrix, P(u) is the global internal resisting force vector and F is the externally applied impulsive load vector. 2.2 Numerical integration in time

As a time integration scheme, direct integration method is applied where the equilibrium equations are satisfied at discrete time intervals. The explicit central difference method based on the Newmark- is applied, in which the solution at time t + t is based on equilibrium conditions at time t. The displacement and velocity at time tn+1 are given, respectively by: 1 n + 2t u n u n +1 = u n + t u 2 (2)

n +1 = u n + u

1 n + u n +1 ) t (u 2

(3)

The equilibrium Eq. (1) is written for time tn+1 as: n +1 + Cu n +1 + Pn +1 = Fn +1 Mu (4) where the acceleration vector is assumed as:
n +1 = 2(2 M + t C ) 1 ( Fn +1 Pn +1 ) 2C (2 M + t C ) 1 u n + t u n u 2

(5)

91

Electronic Journal of Structural Engineering (8) 2008 3.1 Rate dependent lattice model (RDLM) In lattice model, a RC structure can be modelled as parallel layers of concrete and reinforcement lattices, as shown in Fig. 1. This model can be expressed using the finite element (FE) formulation, by smearing out concrete and reinforcement lattices into a continuum.The constitutive equation for the RDLM is comprised of six lattice components. These include compressive and tensile concrete lattices, longitudinal and transverse reinforcement lattices, and two shear lattices, to evaluate the shear transfer on the crack surfaces. The main lattice [Dmain] includes the concrete and reinforcement lattices, which are then combined with the shear lattices [Dshear] to obtain the total stiffness matrix [Dtotal] of the constitutive equation for the model; i.e. [Dtotal] = [Dmain]+ [Dshear] (7) Eq. (7) is the general form of the constitutive equation for the RDLM. The details of the model can be found elsewhere (Ngo 2005). The stressstrain relationship for the lattices is described below. 3.1.1 Main lattice The stress-strain matrix of a cracked reinforced concrete element is expressed by assuming a uniformly strained 2D continuum. The local strains in each lattice component are calculated from which the stress vectors in the local coordinate are evaluated using the uniaxial stress-strain relationship of each lattice component. The equivalent stress-strain matrix in the global coordinate system, which is the general form of the constitutive equation for main lattice, is expressed as: { g } = [L ] [Rn ][L ]{ g } = [Dmain ]{ g }
T

(8)

In Eq. (8), the stiffness matrix [Dmain] for the main lattice in the RDLM is written as:
n 4 Ei t i cos i 1 = symm.

E t
1 n 1

i i

cos 2 i sin 2 i
i i

[D ]
main

E t

sin 4 i

cos 3 i sin i 1 n Ei t i cos i sin 3 i 1 n Ei t i cos 2 i sin 2 i 1

E t

i i

(9) The values of ti are the thickness ratios of each lattice component that are smeared out and the subscript i denotes the number of elements in the Z direction (element width direction). i is the angle of inclination of each lattice to global coordinate. 3.1.2 Shear lattice The shear transfer mechanism along the crack surface is shown in Fig. 2. To model the effect aggregate interlock in the RDLM, two shear lattices S1 and S2 are provided. The effect of aggregate interlock is dependent on the relative movement of concrete on two sides of the crack. Two elements of shear lattices S1 and S2 are allocated in perpendicular direction to the surface, which will carry the
Y

Y
w

2 1

Local coordinate

X A RC element with Reinforcement Reinforcement discrete lattices Y

1 2 1
2

Smeared

X Reinforcement lattices Y

Cracked concrete

X Concrete Concrete Smeared main lattices discrete lattices

Fig. 1. Lattice model

92

Electronic Journal of Structural Engineering (8) 2008 based on a series of impact tests on the Split Hopkinson Pressure bar (SHPB) set-up. This applies to concrete under a dynamic load, and takes into account the strain-rate effect by incorporating multiplying factors for increases in the peak stress and strain at peak strength. This model is applicable to concrete strengths varying from 32 MPa to 160 MPa with a strain-rate up to 300 s-1. The rate dependent dynamic peak stress (f'cd) can be expressed as:

shear forces along a crack. The dowel effect can be accounted for by changing the inclination angles of the reinforcement lattice components in proportion to the progress of fracture. The rigidity of the shear lattices in the local coordinate system is denoted as Dshear, . To transform this into global coordinate system, the shear controlling matrix , and global coordinate transformation matrices T1, T2 are used to define the global shear stiffness matrix Dshear as follows:

K cd

[Dshear ] = [T1 ][][Dshear , ][T2 ]

3.2 Strain rate effect of concrete

f' = cd = ' s f cs

1.026

1 for

(10)

K cd =

' f cd > 1 ) A2 for = A1 ln( ' f cs

To obtain a reliable prediction of concrete behavior under impulsive loading, a new strain-rate dependent constitutive model is proposed by the authors

' where f cs = static compressive strength (MPa) s = strain-rate ; = 310-5 s-1 (quasistatic strain-rate)

Fig. 2. Shear transfer model

Kcd

Dynamic

Uniaxial Biaxial

Kcdfc
fcs Dynamic

Sta Kcd fcs

cd

Fig. 3. Rate dependent compressive concrete model

93

Electronic Journal of Structural Engineering (8) 2008

' = 1 /(20 + f cs / 2) ;

= principle compressive strain of concrete cd = dynamic strain at peak stress = cs (-0.00002 f'cs
' cs

1 = 0.0022 f

' 2 cs

- 0.1989 f

+ 46.137 and

+ 0.0057)

' + 0.9866 A1 = -0.0044 f cs

cs

A2

' + 2.1396 = -0.0128 f cs

' 4.26 f cs = static strain at peak stress = 4 f ' E cs cs

= 3.3 Rate dependent compressive concrete model The dynamic concrete stress-strain relationship in compression is shown in Fig. 3 and is expressed as follows: 2 2 ' for cd f cd = K cd f cs cd cd ' f cd = K cd f cs Z d ( cd ) for > cd

1.0 1.0 0.8 0.34( t / cd )

In the strain-softening region, it is assumed that the stress reduces linearly to Kdf'cf'res. In this study the residual strength f'res is taken as 0.2. 3.4 Rate dependent tensile concrete model For concrete in tension, a bi-linear tension softening model, as shown in Fig. 4, is adopted with the dynamic increase factor, Ktd, proposed by Malvar and Ross (1998). Thus the dynamic tensile strength of concrete (ftd) is defined as:

where,

fcd = compressive dynamic strength under

high strain rate (MPa)

ftd Ktd

Ktd fts /4

Fig. 4. Rate dependent concrete tension softening model

crd

ftd Ktd

C=0.4

crd
Fig. 5. Rate dependent concrete tension stiffening model

94

Electronic Journal of Structural Engineering (8) 2008

where crete (MPa)

1.016 f td = K ts f td 1/ 3 1s 1 K td = for > 1s 1 K td = s for s tensile strength of confts = static

The bond effect coefficient, c, is assumed to be 0.4 in this study. 3.5 Rate dependent reinforcement model The rate effects for reinforcement are considered to be equal in both tension and compression (Fig. 6). The dynamic increase factor (DIF) for yield stress (fyd) recommended by Malvar (1998) is adopted in this study as follows: f yd = f ys 4 10 f = 0.074 0.04 ys 414

and s = strain rate up to 160 s-1 and quasi static strain rate respectively log = 6 - 2; = 1/(8 + 8 f'cs/fco) ; and fco = 10 MPa

The tension stiffening effect in the cracked reinforced concrete, as shown in Fig. 5, due to bond effect is also considered to be the rate dependent and is given by:

f td = K td f ts crd

fyd Eh

Es

yd

Fig. 6. Rate dependent reinforcement model

160 140 Midspan Deflection (mm) 120 100 80 60 40 20 0 0 10 20 Time (ms) 30 40 50 Experiment RC-IMPULSIVE

Fig. 7. Comparison of midspan deflections s

95

Electronic Journal of Structural Engineering (8) 2008

Table 1. Material properties of concrete and steel

Agardh Material properties et al. Ductal panels

Compressive Concrete

strength,

fc

112 6.7 45.3 0.3 156 207 586 207 586

164.2 MPa 22.8 44.1 0.3 325 198 1680 (prestressed)

Tensile strength, ft (MPa) Elastic modulus, Ec (GPa) Poissons ratio, Mode I fracture energy, Gf Longitudinal reinforcement Transverse stirrup
t=0

Elastic modulus, Es (GPa) Yield stress, fy (MPa) Elastic modulus, Es (GPa) Yield stress, fy (MPa)

t = 5ms

t = 50ms

Fig 8. Crack patterns of the HSC beam predicted by RC-IMPULSIVE

Fig 9. Crack pattern of the HSC beam after impact

96

Electronic Journal of Structural Engineering (8) 2008 where fys = static yield strength of steel (MPa) = strain rate of steel 4 NUMERICAL VERIFICATION In order to validate the computer code RCIMPULSIVE, analytical results are compared with the experimental results from the impact test conducted by Agardh et al. (1999) and with the testing on Ductal Panels during the Woomera blast trial in May 2004. The results from the blast trial are presented in detail in Ngo (2005). 4.1 Reinforced concrete beams under impact loading by Agardh et al. (1999) Five high strength concrete beams of 4200mm 340mm 170mm dimensions were subjected to a
60

drop weight of 718 kg striking the middle section of the beams with a velocity of 6.7m/s. The main material properties of concrete and steel employed in the analysis are listed in Table 1. The dynamic analysis is performed using the program RC-IMPULSIVE with a time step of 5 micro seconds and no viscous damping has been considered. The analytical model consists of 595 nodes, 504 four-node elements and 202 truss elements. Fig. 7 shows the time history of the mid-span deflection predicted by RCIMPULSIVE and the experimental results and is found to be in excellent agreement with the experimental results. The crack pattern predicted by this program also follows the crack pattern observed in the impact test (Fig.8 and Fig. 9).

40 M id-span D eflection (m m )

Woomera Test Results Outward Deflection: 37.0 mm

20

0 0 -20 10 20 30 40 50 60 70 80 90 100

-40

Woomera Test Results Inward Deflection: 50.4 mm

-60 Time (msec)

Fig 10. Mid-span deflection history of 100mm thick Ductal Panel 1 at 30m predicted by RCIMPULSIVE compared with Woomera test results
t=0

t =5 0 ms

t = 100 ms

Fig 11. Crack patterns of Ductal Panel -1 predicted by RC-IMPULSIVE

97

Electronic Journal of Structural Engineering (8) 2008 crack patterns of concrete at various stages in the loading history of the panel are also displayed. These results match the experimentally recorded crack pattern in Fig. 12 quite well on both front and rear faces. It may be further observed that although there is some diffusion in the crack pattern, as in all smeared crack analysis, strain localisation is clearly monitored. Similar to the experimental behavior, only minor cracks are developed and concrete crushing does not occur for this panel. 5 CONCLUSION The computer program, RC-IMPLUSIVE, developed in this study, offers excellent opportunities for linear and nonlinear dynamic analysis of reinforced concrete structures where deformations, stresses, and the progressive fracture in concrete and steel can be traced. The dynamic response of the test panels under blast loading, as predicted by RCIMPULSIVE, shows that an accurate numerical simulation of the experimental observations is possible using the proposed nonlinear numerical method, despite the highly variable nature of the loading conditions. The inclusion of strain rate effects on the proposed rate dependent lattice models for concrete and steel leads to a realistic nonlinear dynamic analysis of reinforced concrete structures under impulsive loading conditions. 6 ACKNOWLEDGMENTS This work was conducted as part of the doctoral studies of the first author. The financial support of the University of Melbourne is gratefully acknowledged.

4.2 Ductal panels subject to blast loading A total of three ductal panels were prepared for the blast trial at Woomera, South Australia, in May 2004. Each panel had the dimension of 1000 x 2000 mm. The panel thicknesses were 75 mm and 100 mm. The one-way ductal panels, supported only on two vertical sides, are modelled as simply supported beams with a span of 2m. The panels were prestressed to 20% of the ultimate strength of the tendons. Concrete is modelled using plane stress elements with four nodes. The prestressed tendon is simulated as perfectly bonded truss elements. The FE model includes 205 nodes, 160 four-node elements and 40 truss elements. The prestressed effects are modelled by setting initial tensile stresses for steel elements, as well as initial compressive stresses for concrete elements. The time step length is chosen to be 2.5 micro-seconds. The result for ductal panel 1 is presented here. The main material properties of concrete and steel employed in the analysis for the panel are listed in Table 1. Panel 1 is a 100 mm thick prestressed Ductal panel located at 30m stand-off distance. The blast results in an average reflected impulse at the panel surface equal to 3771 kPa.msec with an average reflected pressure of 1513 kPa. The computed displacement-time history and experimental measurements at the mid-span of Ductal Panel 1 are plotted for comparison in Fig. 10 which shows very good agreement. The peak inward deflection of 50.4 mm was recorded which agrees closely with the analytical results, for which the computed peak inward displacement of 48.6 mm is reached in 14.4 msec. Again there is good agreement between the computed peak outward displacement of 35.3 mm reached at time t =37.9 msec, compared with the recorded deflection of 37.0 mm. The deformed shape of Ductal Panel-1 at different time intervals is plotted in Fig. 11 in which the

REFERENCES [1] Ngo, T. (2005), Behaviour of High Strength Concrete subjected to Impulsive Loading PhD Thesis, Department of Civil & Environmental Engineering, University of Melbourne, Australia. [2] Malvar, L. J. and Ross, C. A. (1998), Review of Strain-rate Effects for Concrete in Tension, ACI Materials Journal, 95(6): 735-739. [3] Malvar. L. J. (1998), Review of Static and Dynamic Properties of Steel Reinforcing Bars, ACI Materials Journal, 95(5): 609-616. [4] Agardh, L., Magnusson, J. and Hansson, H. (1999), High Strength Concrete Beams subjected to Impact Loading Research Report, Defense Research Establishment FOA, Sweden.

Fig. 12. Crack pattern of Ductal Panel -1

98