You are on page 1of 20



Finite Elements in Analysis and Design 24 (1997) 113-132

Finite element simulation of response of buried shelters to blast loadings

Zhengwen Yang*

Department of Mechanical and Production Engineering, National University of Singapore, 10 Kent Ridge Crescent, S 0511, Republic of Singapore

Abstract In the present paper, the response of buried shelters to blast loadings due to conventional weapon detonation has been investigated using the finite element method. The finite element analysis was carried out using a commercial FEA software package, ABAQUS. The validity of finite element model parameters adopted was established by comparisons with existing empirical formulae available for free-field conditions in a soil half-space. The shock response spectra of mountings in shelters have been obtained and compared with those calculated from a SDOF analysis. The effects of the soil damping, the stiffness and dimensions of the structure, the source depth of the weapon detonation, and the stand-off distance on shock response have been thoroughly discussed.

Keywords: FEA; Buried shelter; Blast loading; Free-field pressure; Shock spectrum

1. Introduction Civil defence shelters are typically built to provide protection to personnel and equipment against the effects o f conventional weapon detonation. Apart from the basic objective o f preventing failure o f the structure itself, a major concern is the dynamic response o f the structure, which can have detrimental effects on its occupants and other contents. A rapid movement o f the shelter may cause injury to its human occupants and cause damage to built-in equipment such as generators and electrical fittings. The overall shape o f the shelter normally takes the form o f a monolithic box made o f reinforced concrete walls and slabs. When designing such underground shelters, structural configurations need to be carefully evaluated by considering the dynamic response under design scenario. Continuous research and development has enabled designers to make significant progress. However, the relevant information appears to be scarce because o f the confidential nature o f the subject. * Corresponding address: Blk. 315, Clementi Ave. 4, # 09-145, S 120315, Republic of Singapore. 0168-874X/97/$17.00 (~) 1997 Elsevier Science B.V. All rights reserved PH S0 1 6 8 - 8 7 4 X ( 9 6 ) 0 0 0 3 3-9


Z. )'anw/FiHite Element,s m .4mt/rM.~ am/,qn 24 (1997) 113 132

The problem of underground protective structures subjected to dynamic loadings has long been a topic of interest in defence engineering. The issue has been addressed in terms of theoretical models, numerical methods, and empirical approaches. Theoretical models, which are mostly based on the philosophy of a single-degree-of-freedom (SDOF) analysis [1 3], have been recommended in recent design codes. Relevant design parameters such as the pressure loadings and their resultant acceleration, velocity and displacement are normally presented in either graphical forms or senti-empirical formulae according to simplified assumptions [3]. In most of the models, the overall soil behavior and the soil-structure interaction mechanisms appear to be oversimplified. The finite element method, among other numerical approaches, is a powerful tool which can be used to analyzed the soil structure interaction problem [4 7]. Complex structural configurations can be modeled using finite elements and the response al any desired point of the structure can easily be determined. However, material modeling is a major concern in this approach. The linear elastic model is widely used, but selection of suitable values of this model is a critical consideration if significant errors of prediction are to be avoided. In the present study, a commercial FEA software package, ABAQUS, was used to carry out the finite element analysis about shock response to blast loadings due to conventional weapon detonation. Viscoelasticity was chosen to model the behavior of the soil material. The validity of finite element model parameters adopted was established by comparisons with existing empirical formulae available for free-field conditions in a soil half-space. The shock response spectra of mountings in shelters have been obtained and compared with those calculated from a SDOF analysis. The effects of the soil damping, the stiffness and dimensions of the structure, the source depth of the weapon detonation, and the stand-off distance on shock response have been discussed in detail.
2. Material model 2. 1. Soil mode/

In civil engineering, soil is defined as any uncementcd or weakly cemented accumulation of mineral particles formed by the weathering of rocks. The void spaces between the particles contain water and air. Due to its water content a soil may exist in the liquid, semi-solid or solid state. The last two states are considered herein. Typically, adopted constitutive relations of soils are elastic, elastoplastic, or viscoelastic, though some other more complex relations have also been proposed. The models employed for the elastic behavior often assume the soil to behave as an isotropic material [8, 9]. It is generally accepted that using these models the response of the soils subjected to, say, static loads, can to some extent be evaluated tbr very simple problems. A more complex elastic model may either take the soil hall-space to be made of different elastic layers or take into account the dependence of the soil modulus on the stress state [10]. When excited by fairly high loadings soils may behave strongly nonlinearly. True nonlinear analyses are, however, not yet possible, although promising nonlinear constitutive models for soils do exist. The nonlinearity is usually described in terms of plasticity [l l, 12]. Viscoelastic soils [13-15] exhibit elastic behavior upon loading tbllowed by a slow and continuous increase of strain at a decreasing rate. Such soils are significantly influenced by the time rate of strain.

Z. Yang/Finite Elements in Analysis and Design 24 (1997) 113-132


The choice of a material model is, however, to be based primarily on the purpose of the analysis. Under blast loading, the initial response is most important. The initial response would normally involve some plastic deformation and it is therefore rational to take the soil as an elastoplastic material. However, plastic deformation would be restricted to the vicinity of the explosive. Beyond a certain distance from the detonation, the response will not involve plastic deformation. Furthermore, design stand-off distances are not short enough to cause plastic deformation very near the shelter. Therefore, if our concern is not the field near the detonation, it is sufficient to take soil as an elastic material for the region around the shelter. With the addition of damping, which takes the water content and other effects into consideration, a linear hysteretic viscoelastic model was adopted for the present FEA. In the linear hysteretic viscoelastic model [15], the linear hysteretic damping is independent of frequency and can be introduced into the solution directly, when working in the frequency domain. The elastic solution can be modified to its corresponding viscoelastic solution by replacing the Lam6 constants, 2 and G, with the following expressions, respectively, z = 2(1 + 2 ~ i ) ,

(1) (2)

G* = G(1 + 2~i),
where ~ is the damping ratio.
2.2. Concrete m o d e l

In most cases, the elastoplastic response of protective shelters to explosive effects should be considered. However, nonlinear effects can be neglected if the detonation is far enough away from the structure. This leads to the validity of a linear analysis. Since the concrete material of the structure is much harder than the soil medium, the pure elastic model without damping was used in the present study.

3. Finite element simulation When analyzing dynamic soil-structure interaction problems, another point, i.e. accounting for the effects of practically unbounded domain of soil, needs to be addressed, in addition to modeling of soil as a material. In the present study, the soil is considered as a semi-infinite medium with an unbounded domain. For static loading, a fictitious boundary can be introduced at a distance from the structure, where the load effects can be expected to be negligible practically. This leads to a finite domain for the soil which can be modeled by finite elements similarly to the structure. However, for dynamic loading, this procedure generally cannot be used because a fictitious boundary would reflect waves originating from the vibrating structure back into the discretized soil region instead of letting them pass through and propagate toward infinity. Thus, radiation condition of an unbounded domain usually has to be taken into account in a dynamic model. Nevertheless, in the case of blast loading whose acting duration is very short, the initial stage of response is of great interest and a fictitious boundary could be set at a sufficient distance away where either the reflective waves are not produced within the duration of interest or the effect of reflection on the response is so small that it can be neglected.


Z. Yanq/Finite E/emettts in Anah'sls am/Desiqn 24 (1997) 113 132

t 7 0

Soil x


Blast Buried Shelter

Fig. I, Schematic of buried shelter to blast loading.

Gap formation at the soil-structure interthce may occur when buried structures are subjected to blast loadings [1, l l] (Fig. 1 ). This phenomenon of separation, however, is not taken into consideration herein and the soil is assumed to adhere to the structure surface. For many soil-structure interaction problems with a buried shelter, to obtain meaningful results a 3-D analysis has to be carried out. The need for such an analysis in the present study was investigated by studying an axisymmetric case and comparing the tree-field pressures in FEA at several locations with those calculated from the empirical formulae given in TM 5-855-1 [3]. It became apparent that the empirical formulae are more agreeable with a 2-D analysis in which elastic or viscoelastic soil models are used. A possible reason is that this kind of analyses does not include lateral extension effects, which would be critical when the tensile strength is very small. Thus, in this study, the plane strain FEA was implemented using the commercial FEA software package ABAQUS developed by Hibbitt, Karlsson & Sorensen, Inc. Another code, I/FEM, was used to generate the necessary finite element meshes. The node and element data were then transformed into the ABAQUS format.

3. 1. Free-field pressure
Detonation of an explosive charge produces a radially expanding compressive wave which propagates through the soil medium. When there are no buried structures in the soil, the pressure generated by the compressive wave is referred to as the free-field pressure. When the compressive wave comes into contact with a buried structure, a change in pressure occurs. In the traditional design approach, pressure loadings on the front wall of the buried shelter are usually obtained by multiplying the free-field pressure with an empirical factor.

3.1.1. Empirical formula On the basis of experimental data, the design manual TM 5-855-1 [3] proposed the following empirical formula to describe the decay of the free-field pressure with distance and time: P(R,t) = P , e ''~ = f[]pc ( R ) "c ' '~'



E l e m e n t s in A n a l y s i s a n d D e s i g n 2 4 ( 1 9 9 7 )



1.00E+6 -

~ = 2 . 5 . 1 0 -4 ~ - _ - _ ~ ~ = 2 . 5 . 1 0 -3 ~ = 2 . 5 " 1 0 -z
. . . . . .



~o - 5 . 0 0 E + 5 g.



~ ~ i , i

0.00 ,+0

s o0e-z

l ooe




Fig. 2. Effect of soil damping on free-field pressure (R = 9.81 m).

where fl = 0.47 when SI units are used, f is the coupling factor of the explosive energy, pc the acoustic impedance of the soil, n the attenuation coefficient usually taken to be 2.75, R the radial distance measured from the center of the charge, W the charge weight, t the time measured from the instant of the arrival of the pressure wave, and ta = R/c the arrival time of the pressure wave with c being the soil seismic velocity.

3.1.2. Finite element analysis

In the 2-D analysis adopted herein, the detonation is represented by a pressure load applied on the circumference of a circle with radius R0, whose center coincides with that of the explosive charge. The blast loading on the circumference was calculated using Eq. (3) at R = R0. A typical finite element mesh is made up of 3-node triangular and 4-node quadrilateral elements. Boundary conditions for displacements are either fixed or free. As such, the range of the finite element area for soil was set to be quite far away from the charge so as to achieve good convergence for numerical results. In all calculations, zero initial conditions were also used. The material and charge data used were: Es = 3.45 x 10 7 P a , #s = 0.3, Ps = 1.7 x 10 3 kg/m 3,

f = 1 p c = 5 10 S P a s / m , W = 6 0 N ,


Z. YanH/ t;iHite E/emeHIs in Analr,si,~ and Design 24 (1997) 113 132

k'L'M c pm.pir4.c(zZ(c

.5 00F,'~ 4


300m/s) 165m/s)


" ~ ,'700E ~4
~d .k.





O. OOE~ 0 . . 0.00~'~0 ta)

. . o.OOlJ 2

~ ~ l O01z 1

4 1.50E-I

t:ig. 3. Comparison of tiee-lield pressure: tat IR : 7.33 m); and (b) (R

9.81 m).

where E~,ll~, and p~ are the Young's modulus, the Poisson's ratio, and the mass density of the soil, respectively. The charge input data lead to a pressure of P, = 107 Pa at R, = Ira.

3. 1.3. Results and comparLs'ons Sensitiuitr to damping ratio. Fig. 2 shows how the damping ratio of the viscoelastic soil model affects the free-field pressure at a distance of R 9.81m from the detonation center. If there is no damping and the soil is purely elastic, the pressure response oscillates violently with the first peak nearing 1MPa. Since the medium is assumed to be able to take tension in this analysis, immediately following the first positive peak, a negative peak exceeding 1 MPa occurs. The response that follows is more subdued. If the tensile capacity of the medium is negligible, the response following the first peak will be far more subdued due to energy losses during tensile separation. Introduction of damping reduces the intensity of the first peak itself as well as the consequent deformation, as expected. Fig. 3 compares the FEA results with ~ 2.5% and the corresponding empirical results by Eq. (3) at R = 7.33m and R - 9.8l m, respectively. The empirical curves are plotted for two different seismic velocities, one set to the commonly used value for soils, 300 m/s, and the other, 165 m/s, calculated directly from the elastic constants of the soil. The amount of damping, = 2.5%, used for FEA is within the damping interval [2%, 20%] for most soils. It is observed that with this value of damping the pressure (maximum 16.5 KPa in Fig. 3(b) at the location far away from the detonation is relatively low, compared to the yield capacity of the soil. The FEA results are observed to be in good agreement with the empirical formula with c - 300 m/s when t > 30 ms. This result can

Z. Y a n g l F i n i t e Elements in Analysis and Design 24 (1997) 113-132


2 , 0 0 E + 4 --

FEM(xx-stress) * ~ . . . . FEM(principal stress) empirical(c=3OOm//s) empirical(c = f 6 5 m / s )

. . . . .

\ \

\ \
\ \ \ \ \

1.00E+4 5q -.~ 5.00E+3

- 5, 0 0 E + 3 --~T--]--r--T--T--r--q--r--T--q--r--T--q--T--]--T--T-----~ O. OOE + O 5, O O E - 2 1. OOE - 1 l, 5 0 E - 1




Fig. 3. Continued.

perhaps be considered as an illustration of the validity of the use of 300 m/s as the seismic velocity in practical applications and of 2.5% as the damping ratio for the visocelastic soil model in FEA. These parameters were adopted throughout the remaining FEAs. It is also observed that the normal stress in the x-direction is nearly equal to the principal stress at the point considered herein and thus can be chosen to represent the actual pressure. Effect of charge depth. Shown in Fig. 4 is a comparison of free-field pressure at a horizontal distance of R = 9.81 m from the detonation for the depths of detonation at H = 10 m and H = 20 m. It is evident that the depth of the explosive charge has no significant effect on the freefield pressure. Further studies have shown that this conclusion is true as long as the charge location is not too shallow. That is to say, when the charge is located deeply enough, when compared with the stand-off distance at which the pressure is measured, the surface effect of the soil half-space can be ignored. 3.2. Response with buried shelter In this section, finite element studies of the soil half-space with a buried shelter are discussed. In all cases the depth of charge is assumed to be at the same horizontal level as the center of gravity



Y a n g / F i n i t e Elements in Analysis and Design 24 ( 1 9 9 7 ) 113 132

-- H=lOrn
2.00E+4 *-~ = * * H = 2 O T r L


L \


'\, \~

5. O O E + 3
/ / i / /



,~ ....

T~ ~

T 2






Fig. 4. Effect of charge depth on free-field pressure (R

9.81 m).

o f the shelter. Considered herein is a box-type shelter with a rectangular cross-section, consisting o f two walls and two slabs as shown in Fig. 1. The shelter was modeled with 2-D beam elements. The effects of different rigidities of different shelters were considered by varying the thickness o f wall and elastic modulus o f wall material as well. The case o f a buried rigid solid block was also considered for comparison purposes. The same soil and charge properties given in Section 3.1.2 above were used. As noted before, the detonation was represented as a pressure pulse o f peak intensity Pa - - 107 Pa at a radius R0 -- 1 m. Adopting typical concrete properties for the shelter wall material, the following were assigned: the Poisson's ratio /~c = 0.2 and the mass density Pc -- 2.3 x 103 kg/m 3. 3.2.1. Effect o f shelter stiffness Two shelters o f wall thickness ts = 0.5 m were considered, with one having a material elastic modulus Ec = 2.6 x 10 ] Pa and the other 1000 stiffer, i.e., Ec = 2.6 x 10 j3 Pa. In addition, a rigid solid block was also considered. The shelter size is 2 0 m x 4 m , the depth o f detonation is H = 10m, and the stand-off distance is L = 10 m. The pressure at point 1 o f these three cases were compared with each other and the free-field pressure at the same location, in Fig. 5. It is observed that an increase in the rigidity o f the shelter causes an increase in the peak pressure. A very stiff box-type shelter behaves just like a rigid solid block in the soil medium. The peak pressure with a buried rigid solid block is about 1.6 times the peak free-field pressure. This value is slightly higher than 1.5

Z. Yang/Finite Elements in Analysis and Design 24 (1997) 113-132


.... free-field(FEM) -- -- b u r i e d s t r u c t u r e , Ec=2.6* 101N/(rnZ b u r i e d s t r u c t u r e , E c = 2 . 6 " 101aN//m z ..... b u r i e d r i g i d s o l i d block

3. OOE+4

"x \

1. OOE+4

, \\ , k ~

// '/


- i. O O E + 4 - ~

o. o o E + o

5. o o E - e

~ ~. o o E - ~

1.5oE- 1

Time (See)
Fig. 5. Effect of buried shelter on pressure field (R = 9.81 m).

suggested in TM 5-855-1 [3] for a conservative evaluation but less than 2 which is the theoretically idealized value at a rigid surface.
3.2.2. Effect o f s h e l t e r - w a l l rigidity

The displacement of the mid-region of a wall relative to its supported boundaries depends on the wall rigidity. In the case of shelter wall, due to soil-structure interaction, the loading on the wall also changes with wall displacements. Fig. 6 illustrates the influence of the shelter-wall rigidity on the displacement normal to the near-wall interface. The wall rigidity was modified by changing the thickness of the near wall, tw, as well as the material elastic modulus. In all cases input data about the shelter and values of H and L are the same as in Section 3.2.1. It is observed that with low rigidity at the near wall, large deformations occur followed by a rebound. Eventually, all the three curves approach the same displacement level which is the same as the displacement of the shelter as a whole, the vibration having been damped by the soil. The initial displacement of the flexible wall is as high as 2 times that of the final displacement of the shelter. Note that the flexible wall has deflected in tandem with the pressure variation, ahead of the rest of the shelter. The displacement of the wall relative to the shelter is also governed by the natural frequency of the wall, which depends on the wall thickness and material elastic modulus.


Z, Yang/PTnite Elements in Analysis and Design 24 (1997) 113 132 a: ~ 0 . 2 5 ~ t k E ~ = 2 . 6 , 10 fO N / / m 2 6 t~ 0 , 5 0 m E ~ = 2 . 6 * l O l N / m z ' ~ , t3 2 C a s e c: t~: 1.00vrL, f , , ~ = 2 . 6 10 N ~ 7 ~ Case Case


2. OOE 3



i /r ~ , ,i /

" c-:.

j' /


I/ I/

5. OOE 4

J~ ~
t}s /',



] 7 I I "T--I i I
--T I [ [

5. OOE 2



Fig. 6, Effect of structural rigidity on wall displacement ( R - - 10 in).

Depicted in Fig. 7 is the displacement distribution along the near wall at different times for a flexible wall. Displacements at the upper and lower ends of the wall are nearly the same as the displacement of the shelter as a whole. As can be expected, the flexible walls deflect more relative to the edges, when compared with rigid walls. When the ground surface reflection of the pressure pulse reaches the shelter, the shelter tilts by a small angle, manifesting as a differential movement between the top and bottom edges of the wall. These two figures and more results have also shown that the ratio of the maximum deflection to the wall span is less than 5 10 - 4 which is within elastic limits of the materials considered herein. Therefore, the linear analysis is deemed to be rational.
3.2.3. Shock response spectrum

Shock response spectra are used to determine shock isolation requirements for vital equipment mounted in shelters. A response spectrum describes the relative displacement, velocity and acceleration amplitudes as a function of the natural frequency of mountings in shelters using a linear analysis as is in the present study. Each scenario of adverse conditions may provide different response spectra.
3.2.3. 1. S D O F model. A widely adopted approach in deriving a response spectrum is to consider the rigid-body motion of a shelter using a S D O F model. The response spectrum is conventionally

Z. Yang/Finite Elements in Analysis and Design 24 (1997) 113-132

..... t=O. O 2 5 S e c t=O. O 7 5 S e c t=O. 1 2 5 S e c


Case -8.00E+O

* * * * *

. . ~


E - 1. OOE + 1


-1.20E+1 ] i ~ 4.00- 4

8, O O E - 4



2. O O E - 3


Displacement (m)
Fig. 7. Deformation o f the near wall.

represented as a graph of pseudo-velocity versus angular frequency co with the pseudo-velocity being defined as the product of co and the amplitude of the relative displacement of the mounting. When a mounting is modeled as a spring-mass system, the equation of motion of the mounting can be written in the form i; + co-y = -//0, (4)

where y is the relative displacement of the mounting with respect to the shelter, u0 the displacement of the shelter, and co = 2 n f with f indicating the natural frequency of the mounting. A closed-form expression of the response spectrum of a mounting in a shelter was previously obtained by other investigators [16], i.e.
I)p ~ ~acota


~-gV/( 1 + co2ta)(1 2 + cozt~) where Vp denotes the pseudo-velocity, ~aa the peak value of the average normal free-field pressure at the near surface, p-O the average of the near and far side soil impedances, ta the averaged arrival time of the pressure pulse, and tb the characteristic transit time across the structure, which is





in Anulysis

und Design 24 (1997)


Case -8.OOE+0

+-*H* y-M

t-0.025Sec t=O.O75Sec t=O. 125Sec


:1 4.00E-4


1 1 ~--7-T-


1 /

8 OOE --4

1 20Ep3




Fig. 7. Continued.


given by
fb =



with H,, being the total mass of the shelter and its contents per unit area exposed to the blast. Finite element analysis. The displacement, velocity and acceleration calculated at points in a shelter using FEA can be used to define the base motion, to which the equipment and other objects mounted in the shelter will be subjected. Knowing the base motion, it is possible to derive the corresponding response spectrum for a mounting, using a simple spring-mass system. Since the base motion data are discrete in the time space, a step-by-step time-marching method should be employed to obtain the response of a mounted spring-mass system. Eq. (4) describes the motion of a spring-mass system due to base acceleration. The initial conditions of the relative displacement y are yl,=, = _illtzO= 0. Assuming the acceleration of the shelter is linear between the discrete accelerations at known time points, the relative displacement

Z. Y a n g / F i n i t e Elements in Analysis and Design 24 (1997) 113-132


Case c
-8. OOE+O -

..... t=O.O25Sec ~ * * ~ t=O. O75Sec ~.~ t=O. 125Sec


I I I I I I [ I I I I I I I [ I I I I I I I ] I I I I I I I [



-1.20E+1 I 4.00E-4

8. OOE-4



2. OOE-3


Fig. 7. Continued.


yj+~ at time t = tj+~ can be expressed in terms of the relative displacement yj at t = O, the base acceleration //oj, and the increment in base acceleration &//0j over the time duration fO, as follows:
Yi+l = (Yl)j+l -t- (Y2)j+l + (Y3)j+t,


(Yl)j+l = y/COS

~obtj + J~J sin co(~tj,



(.v2),+, = u ~ ( c o s ~o60 - 1),

(O 2




6//oj (sinco60 fO36tj


(lO) (11)

aO = 0+1 - O,

6iioj =//oj+l -//oj-


Z. Yan q / fqnite Elements m Analysi.~ am/ Design 24 ( 1997 j 113 132

The expressions of the relative velocity and acceleration are readily obtained from the above equations. The calculations proceed step-by-step through the discrete time intervals, starting from t=0. The maximum relative velocity I.f'i...... is known as the pseudo-velocity:




As already noted, the response spectrum is indeed a graph of c t, versus ~,). The response spectrum thus denotes the relative motion of a continuous spectrum of spring-mass systems subjected to the same base excitation. By definition, on the other hand, a simple harmonic motion v has the lbllowing relations among the maximum displacement ]YI,, ..... the maximum velocity J.f'l,~,:, and the maximum acceleration


= J.f'i ...... _- I-i}i"'~"~



Therefore, if any one of the three quantities, i3'1....... JJ'J....... or ].~"l,-~,~ is known, the other two can be calculated, provided ~o is known. In other words, the response spectrum can also be interpreted in terms of I)'lma~ as well as I.i;'l....... It should be noted that the maximum values JY]........ i.f'lm,lx, and J.i;'j ..... can be separately lbund from the time-marching scheme. Given that the computed v need not exactly be a simple harmonic motion, these values may not fit the description in Eq. (14). It is necessary to exercise care when choosing a value of j f'] ...... fbr plotting the response spectrum. Furthermore, it is necessary to ensure that the base acceleration //0, which is available only in discrete values, does not contain excessive noise. Thus, iio needs to be moderated or verified using known ~i0 and u0. R e s u l t s a n d c o m p a r i s o n s .

The response at the center of the near wall was obtained. Five

cases were studied.

Case A B C D E

H(m) 10 10 20 10 10

l,lml I0 l0 I I) 15 i5

Shelter dimensions (m m) 20.0 x 4.0 12.0 x 2.4 20.0 ,, 4.0 20.0 4.0 10.0 ".~ 2.0

The material and charge data were the same as those stated above, with the exception of k,'~. 2.6 1013 Pa and t,, = 1 m; this assumes a rigid shelter and provides the average base motion of the shelter. Two methods were used to obtain the peak values i3'1....... I./'l....... and ].i,I...... In method 1 the peak value was determined based on all the results over the entire time interval adopted for calculation

Z. Yang/ Finite Elements in Analysis and Design 24 (1997) 113-132


.... ..... ~
1 0 ~=~.

" "~ ","'~"~"

wy, (method 1) y (method 1) y " / w (method 1 ) wy (method 2) y:, (method 2) y '//w (method 2)

1 0 -~

1 0 -3_


1 0 ~6

1 0 -7

I [111




1 0 -t


10 ~

10 ~



Fig. 8. Shock spectrum from finite element analysis.

of the response of the spring-mass system, while in method 2 the peak value was chosen by considering the oscillations within the final steady response when the system eventually vibrates freely. Fig. 8 shows the pseudo-velocities in Case A calculated from Eq. (13) and Eq. (14) using these two methods. Although the three curves given by method 2 seem to meet the requirements of Eq. (14) for most frequencies within the frequency range from 0.01 Hz to 1000 Hz considered herein, the response at very low frequencies near 0.01 Hz as well as at very high frequencies near 1000 Hz appears to be discontinuous, which implies that this method can not generate correct peak response of the spring-mass system. On the other hand, the curves obtained using method 1 satisfy Eq. (14) only for the frequencies from 10 Hz and 80 Hz. A possible reason is that errors occur in the relative velocity and acceleration when they are calculated from the relative displacement which is believed to be the most accurate. However, it is observed that among all these curves the solid one calculated based on the displacement is the only smooth curve. For this reason, the solid curve with method 1 can be considered to represent the correct pseudo-velocity. This consequence can also be expected, as compared with method 2, method 1 takes into account a much longer span of response.

Comparison with SDOF results. Comparisons were carried out between the response spectra produced using the FEA results of shelter movement and those produced by the SDOF model.


Z. Yan.q I l~Tnite Elements in Analysis amt Design 24 (1997) 113-132

FEM, t,=O. 15, Case A ........ FEM, t,:O. 84, Case A .... SDOF, 1. OH 1.0Q, 1.0pc *~*~*'-*~* SDOF, O. 4t'O.8Q, 1.0pc SDOF, O.4H~o'O.5Q, O.5pe to ~,~,~

-~ ,.4~~- _-~-..

-.".~ 10 2


0 \

]~ I I I -T[TW ....... ~; I I '-~ T 1 J T 'TIT~ I I I I I I I i




10 z

10 3


Fig. 9. Continued.


Fig. 9 shows the comparisons of the FEAs in Cases A and D with their corresponding SDOF results. Here tr is the truncated-response time defined as the time after which the response of the spring-mass system in FEA has been truncated when determining the maximum relative displacement. The maximum relative displacement of the system after t = tr should not exceed that of the system when t < tr. It can be seen that a good evaluation of the response spectra can be achieved within a relative short time tr = 0.15 s truncated for the response. It is also observed that the FEA curves almost lie between two SDOF curves with different values of Hf,, t,, and p~. Effect o f shelter dimensions. From Fig. 10 it is found that the pseudo-velocities of the mountings in the smaller buried shelter (Cases B and E) are always larger than those in the bigger shelter (Cases A and D) because the former has a less total mass and is easier to be displaced. Effect o f charge depth. Fig. 11 indicates that the charge depth has no significant effect on the response spectrum. Further studies have shown that this is true provided the charge is located deeply enough, i.e., provided the depth of detonation is of the same order as the shelter dimensions. Effect o f stand-off distance. The effect of the stand-off distance on the response spectrum is shown in Fig. 12. Increasing the stand-off distance will yield lower pseudo-velocities as expected.

Z. Yang l Finite Elements in Analysis and Design 24 (1997) 113-132 FEM, Case D . . . . . . . . S D O F 1. OH 1.0t~ , 1.0pc .... S D O F ' O. 4I-f~ ' . 7 5 t a .50pc


I0 -1




1 0 -2. co o

t ~3 10 -s




"% -,, ,


1 0 -4



I I fill I


1 0 -~


I0 ~

10 ~


Freqaze~zcy (Hz)
Fig. 9. Comparison of shock spectrum.


10 - 1 _


I0 0





I IIlll


I IIlll


10 z

10 3


Freq~zevtcy (Hz)
Fig. 10. Continued.














' ~






" C.,3 0



" "F

" r ". . . . . . . . .

T " l " - wr~

!g) 10 z


i i r-17 ~q
10 ~


/~'re q'~z~zc',q


Fig, I0. Effect of structure dimensions on shock spectrum.








,// //4"'
2 J" X \ \ \, \




Q 10 @/;--s
F ~ i T-I I ; -T]---




, ,,,,~
10 a



Freq'u, er~cv


10 ~

Fig. II. Effect of charge depth on shock spectrum.

Z. YanglFinite Elements in Analysis and Desiqn 24 (1997) 113-132

- Case A




****~ Case C

1 0 -~

"d Q




1 0 ~'


10 z

10 3



Fig. 12. Effect of stand-off distance on shock spectrum.

4. Conclusions and recommendations (1) An elaborate and extensive analysis of shelter response was carried out using the ABAQUS FEA software package. A viscoelastic soil model was adopted. A 2-D analysis was selected instead of a 3-D one following comparisons with an existing empirical formula for free-field pressure waves. The soil parameters to be adopted in further studies were also chosen on the basis of comparisons with the empirical formula. A wave velocity of 300 m/s and a damping ratio of 2.5% were found to be suitable. (2) Various cases of buried shelters subjected to detonations at the same level as the shelter were studied using 2-D FEA. (a) Rigid structures experience higher pressure and less displacement during the first half-wave of response, when compared with more flexible counterparts. Due to the ground surface reflection of the pressure wave, the shelter wall deforms slightly asymmetrically as the shelter tilts by a very small angle. The deformation within elastic limits has verified the present linear analysis. (b) The FEA compares well with SDOF model for the response spectrum. (c) Either the lighter the shelter or the shorter the stand-off distance, the larger the peak pseudovelocity. The charge depth has no significant effect on the response spectrum. For further investigations on the same line the following are suggested. (1) Adoption of more complicated soil models with the objective of obtaining a more realistic representation and more accurate analysis;


Z. Yan{t/Finite Elements #~ Anah'sis and Desi#n 24 (1997) 113 132

(2) Consideration of the effects of large detbrmation in the soil near the detonation charge, to simulate a more realistic scenario; and (3) Incorporation of the influence of gap tbrmation at the soil-structure interface on the response of shelters.

Acknowledgements The financial support from the National Science and Technology Board of Singapore and the discussions with Drs. K.Y. Earn and W.A.M. Alwis of the National University of Singapore are gratefully acknowledged. References
[1] E. Hinman, "Shock response of underground structures to explosion", ~th h*t. Co*~ll on lQhration and Shock, Lyons, France, 1988. [2] I.N. Psycharis, "Effect of base uplift on dynamic response of SDOF structures", J. Slrucl. )Lq. 11% pp. 733 754, 1991. [3] Fundamental of Protection Design for Conventional Weapons. TM 5-855-1, US Army Engineers Waterways Experimental Station, Vicksburg, 1986. [4] O.K. Gyebi and G. Dasgupta, "'Finite elemenl analysis of viscoplastic soils with Q-factors", Soil Dyn. Earthquake En~l. 11 pp. 187 192, 1992. [5] Y.K. Cheung and J.X. Zhu, "'Dynamic interaction analysis of a circular cylindrical shell of finite length in a hallspace", Earthquake Eng. Struet. Qrn. 21. pp. 799-809 1992. [6] J.X. Zhu et al., "Dynamic response of underground structures by time domain SBEM and SFEM", Appl. Math. Mech. 11, pp. 1129 1136 1990. [7] S. Valliappan, "'Numerical modeling of tunnel blasting", in: (i. Swoboda (ed.), Numerical Methods in Geomeehanics, Innsbruck, Balkema, Rotterdam, pp. 145 154, 1988. [8] P. Dangla, "A plane strain soil-structure interaction model", Earthquake Enq. Strut. D3'n. 16, pp. 1115-1128, 1988. [9] W. Ruecker et al., "Behavior of three-dimensional structures due to transient loads in the interior of an elastic halfspace", in: A.S. Cakmak and 1. Herrera (eds.), Proc. 4th hit. Cot!F On Soil Dynamics and Earthquake Engineering, Mexico City, 1989. [10] P.V. Lade, Model and parameters tbr the elastic behavior of soils, in: G. Swoboda (ed.) Numerical Methods in Geomechanics, Innsbruck, Balkema, Rotterdam, pp. 359 364. 1988. [11] P. Weidlinger and E. Hinman, "Analysis of underground protective structures". J. Struct. En,q. A S C E 114, pp. 1658 1673, 1988. [12] R. De Boer, "'On plastic delbrmation of soils", Int..I. Pk,sticitv 4 pp. 371-391, 1988. [13] B.L. Kutter and N. Sathialingamn, "Elastic-viscoelastic modeling of the rate-dependent behavior of clays", Geoteehnique 42, pp. 427~441, 1992. [14] J. Dominguez and R. Abascal, "'Seismic response of strip footings on zoned viscoelastic soils", J. Eng. Mech. ASCE 115, 913 934, 1989. [15] W.S. Pi, "Dynamic tire/soil contact surface interaction model for aircraft ground operations", J. AireraJ? 25, pp. t038 1044, 1988. [16] W.A.M. Alwis and K.Y. Lain, "Response spectrum of underground protective structures", l~;inite Elements Anal Des. 18, pp. 203-209, 1994.