ANALYSIS OF BLAST LOADS ON BUILDINGS

Naury K. Birnbaum* Richard A. Clegg Greg E. Fairlie

Colin J. Hayhurst Nigel J. Francis
*Century Dynamics Incorporated
7700 Edgewater Drive, Suite 626
Oakland, CA 94621
Century Dynamics Limited
Dynamics House, Hurst Road, Horsham,
West Sussex RH12 2DT, England

ABSTRACT
Analytic techniques and computational fluid dynamics (CFD)
software programs are currently in use for assessing the blast loads on
military structures for a variety of situations. These approaches, of
necessity, are now being extended to the calculation of explosionstructure interaction problems where the loadings are not upon
hardened military targets but rather on civilian structures. This has
become of great importance as high explosive bomb attacks have been
increasingly directed against many commercial, financial and civic
centers consisting of buildings of conventional, soft construction.
One aspect in the protection of such structures is the accurate
prediction of the blast loadings on structural components using
analytic or advanced numerical tools taking into account the
complexity of the building, its geometry and the surrounding
environment. Such an understanding of the loads can help define
building protection options such as selection of materials, relocation of
building services, siting, and construction techniques.
Various different numerical and analytic techniques are discussed
and illustrated by example. Applications discussed include:

Idealized one and two dimensional problems

Blast propagation over multiple structures in two dimensions

Blast in a three dimensional urban setting due to a terrorist bomb

including multiple interactions with buildings and roads.

Explosion within a three dimensional deformable building

Three dimensional blast study of the Oklahoma City Federal

Building
Important effects such as multiple blast wave reflections,
rarefactions, and the negative phase of the blast wave can be readily
modeled in CFD codes. Simplified analytic and semi-empirical
techniques many times ignore such phenomena. Thus, modeling
modern buildings set in congested city centers usually requires the use
of sophisticated CFD numerical calculations. The use of both analytic
techniques and CFD calculations can provide a cost-efficient and
accurate approach to determining blast loads. Such information is
critical in any effort to effectively harden commercial structures.

1. INTRODUCTION
Commercial buildings are constructed quite differently than
hardened military structures and as such are generally quite vulnerable
to blast and ballistic threats. In order to design structures which are
able to withstand explosions it is necessary to first quantify the effects
of such explosions. Such quantification should lead to better and more
cost effective design. Typically, it takes a combination of specialist
expertise, experimental tests, and analysis tools to properly quantify
the effects.
This paper concentrates on various analysis methods available to
predict the loads from a high explosive blast on commercial buildings.
Different analysis methods are appropriate to different situations. The
loading types and structural design requirements are discussed in the
context of what the analyses should provide. A number of example
analyses are shown.

2. SOFTWARE TOOLS
A variety of software tools are available from various sources to
quantify the effects of explosives on structures. The table below
summarizes those in use by the authors but does not purport to
encompass the wide variety of software available to the community.
The various blast load analyses presented in this paper are discussed in
terms of the generic methods that can be used.

AUTODYN-2D

Numerical

Table 1. Software Tools

Century Dynamics

AUTODYN-3D

Numerical

Century Dynamics

AutoReaGas (3D)
MADER codes
US Army codes
BLAPAN
SPLIT-X

Numerical
Numerical & Analytical
Analytical
Analytical
Analytical

Century Dynamics/TNO
Charles Mader
US DOD
Century Dynamics
CONDAT/Century Dynamics

Impact/penetration, high explosive and

blast simulation
Impact/penetration, high explosive and
blast simulation
Gas explosions and blast simulation
Characterization of explosive properties
Weapons effects calculations
Calculation of blast loads on panels
Fragmentation code

3. TYPICAL STRUCTURAL RESPONSE TO EXPLOSIVE LOADINGS

The nature of the explosive loading on a structure can be
device and the type of response to be quantified different software
summarized as shown in the table below. Depending on the size of the
tools can be employed.

Typical Responses
Air blast

Primary fragments
Secondary fragments

TABLE 2. Structural Response to Explosions

Small Devices
Large Devices
Component damage/ failure
Component damage / failure
Structural collapse
Small secondary missiles
Secondary fragments & damage escalation
Local in-structure shock
In-structure shock
Localized damage
Localized damage
Localized damage
Localized damage
Tertiary fragments & damage escalation

4. STRUCTURAL DESIGNERS REQUIREMENTS

The structural designer, in order to design a new structure or
retrofit an existing structure for explosion resistance, generally
requires certain information as summarized below:
Blast Loads:
Peak pressures
Impulses
Load time-histories on structural components
Impact Loads:
Velocities / Trajectories of projectiles
Size / Shape / Materials
In this paper, impact type loads will be ignored with the focus on
blast. However, a number of the software tools available are quite
proficient at handling both types of loads.

Summing the impulse for each sub-section then provides the total
impulse applied to the surface.

The total load-time history is then defined to have an exponential

form with a peak calculated assuming an average peak pressure
applied over all the surfaces.
While this simplified method can supply very useful information
it exhibits the following limitations:

Typical real situations are three-dimensional and the resultant

multiple blast reflections and rarefactions are not treated.

The assumption that the load-time history is applied to the all

parts of the surface at the same time with an exponential form is a
poor approximation for near field effects.

The negative phase of the blast pressure is ignored.

Other analytic methods have been developed to overcome some
of the above limitations. An example of one of these is codified in the
BLAPAN program.2 In this case the total load on a surface at a
particular time is computed by summing the load on each surface at
that time. Thus, the calculation then predicts a load time history which
has the same impulse as calculated by TM5-855 but with a different
time history.
The scope of this paper does not allow for a wide review of
analytic methods but it can be stated that generally these various

5. ANALYTIC METHODS
The US Army manual TM5-8551 provides a useful screening tool
for assessing blast loads when many combinations of explosive device
and location must be considered. Load time histories for buildings and
structural members may be calculated according to the method
outlined in TM5-855. The principal steps of the method are:

Divide a surface into sub-sections and calculate a pressure time

history and impulse for each small area.

Zintilis et al, Protection of Commercial Buildings against

Blast and Ballistic Threats, 7th International Symposium on
Interaction of the Effects of Munitions with Structures,
Mannheim (1995).

TM5-855-1 Fundamentals of Protective Design for

Conventional Weapons, Department of the Army, Technical
Manual (1986)

methods, of necessity, neglect some portion of the true physics of the

phenomena.
Numerical methods such as 2- and 3-dimensional CFD
(Computational Fluids Dynamics) codes can provide a more accurate
approach than the analytical methods. The true physics of the
problem, including 3D geometries, can be handled with such codes.
The detonation of the explosive, rapid expansion and formation of the
blast wave, multiple blast reflections, rarefactions and the negative
phase of the blast can all be modeled.

The differing capabilities of the analytic and numerical methods

outlined above can be illustrated in a sample case of a explosive
device located near an office building as shown in Figure 1. This
shows a 100kg TNT device located in a street 15m from an isolated
office building. One typical component of this building to be assessed
could be a double glazed unit 1.5 m by 2m high with its center 12m
above the ground.
Figure 2 indicates the predicted load-time histories on this panel
using TM5-855 (analytic), BLAPAN (modified analytic) , and
AUTODYN-2D (numerical). The positive phase impulse predicted
using TM5-855 and BLAPAN are identical and the AUTODYN-2D
impulse is within 3% of this value. The differences are in the time
phasing and peak magnitude of the loads. The TM5-855 curve shows a
peak load 57% higher than the other two predictions. The exponential
form of the load time history is evidenced in all cases. The BLAPAN
and AUTODYN results both show similar peak loads and timing to
each other. After 40ms a major difference appears when the blast
wave enters the negative phase. A numerical code is able to capture
the correct physics while the analytic results do not account for this
phenomenon.

6. NUMERICAL METHODS
The numerical methods most appropriate to the simulation of the
blast problem are typically based upon a finite difference, finite
volume, or finite element method utilizing explicit time integration.
By way of illustration, the AUTODYN3 and AutoReaGas 4 system of
programs will be used to illustrate some of the generic capabilities of
these types of codes for predicting the blast loads on structures.
Programs available to the community incorporate a number of
different widely used numerical techniques. Each of these techniques
has certain advantages and disadvantages. No single method can
handle all regimes of a typical problem. The challenge is to apply
these techniques in the most efficient and accurate fashion for an
optimum solution.
The various techniques can be generally characterized as below:

Lagrange : the numerical mesh moves and distorts with the

material motion

Euler: the numerical mesh is fixed with the material flowing

through it. First and second order accurate schemes may be used.
Also, specialized high speed second order Euler methods suitable only
for blast simulations are sometimes available.

ALE: Arbitrary Lagrange Euler. Provides a type of automatic

rezoning overcoming some of the limitations of both Euler and
Lagrange.

Structural elements: Typically Finite Element formulation. Thin

shells and other structural members are treated.
In addition to the ability to use different numerical methods for a
problem, it is also possible within some software to apply multiple
methods within a given simulation. For example, this can be very
important for studying the effects of fluid-structure interaction, e.g.
blast loads on a deformable structure. In such a case the structure
might be modeled with a Lagrange or structural representation while
the explosive-air domain would be modeled with an Eulerian
approach. The two domains are then dynamically coupled together in
space and time.
The following sections review specific applications of the above
generic techniques. Details of each simulation are available in the
cited references. Finally, a more detailed case study as applied to the
recent Oklahoma City Federal Building bombing is presented.

7. COMPARISON OF ANALYTIC
RESULTS FOR A VALIDATING CASE

AND

Office Block

15
15 m
m
100 Kg TNT

Figure 1. Explosive Device near an office block, analysis

geometry

NUMERICAL

AUTODYN software for non-linear dynamics, proprietary

to Century Dynamics (1985-1996).
4
AutoReaGas software for reactive gas dynamics and blast
analysis, proprietary to Century Dynamics / TNO (1994-1996).

P2
P1

P5
P3

P4

P6

Figure 2. Analytic/numerical results: panel load history

8. BLAST OVER STRUCTURE (2D EULER)

In this example, the far-field blast loading on two buildings is
computed utilizing a two dimensional Eulerian higher order Flux
Corrected Transport (FCT) approach. 5 Figure 3 shows the interaction
with two idealized buildings of 10 meters width and 15 meters high at
two different times. A plane blast wave of 10 kPa and 0.1 secs
impinges on the structures from the left. The blast wave first reflects at
the front of the first building. The wave then passes over the first
building and diffracts down the back side hitting the ground surface.
Gauges monitor the pressure on the various surfaces of the structures.
Typical pressure traces are shown in Figure 4. The wave phenomena
and interaction with the structures is clearly evidenced. The front of
the second building experiences a substantial reduction in loading due
to the obstruction of the first building.
TNO has carried a number of numerical simulations and
corroborating experiments to validate the method.6 The correctness of
the underlying assumption that the blast loading on objects of large
dimension (i.e. buildings) can be modeled as inviscid compressible
flow has been verified. Moreover, the numerical Eulerian FCT
technique has been shown to accurately handle steep gradients and
shocks with a minimum of artificial diffusion.

Figure 3. Blast Wave Interaction with Two Buildings.

Figure 4. Overpressure-time Traces Gauges P1 - P6

9. TERRORIST BOMB IN CITY CENTER STREET (3D

EULER).
In order to study the effects of partial confinement of a blast
wave in a street, a three dimensional Eulerian FCT technique was used
to model a 1000kg TNT charge detonating at ground level in the
center of a 10m wide street as shown in Figure 5. Because of
symmetry only the left half of the geometry was simulated. The size
of the numerical mesh is 300,000 cells. The office building is 80m
wide by 40m high and is centered directly opposite the street
containing the charge. Two buildings are located on either side of the
street creating a channel. Building 1 is 20m high and 36m long.
Building 2 is 14m high and 40m long. There is an 8m wide and 25m
long blind alley between buildings 1 and 2, and a 12m wide street in

W.P.M. Mercx et al, Simulation of Vapour Cloud Explosion

and Subsequent Blast Propagation in Built-up Areas,
International Gas Research Conference, Cannes, France
(1995).
6
L.H.J. Absil et al, Blast Interaction with Multiple Obstacles,
13th International Symposium on Military Applications of
Blast and Shock (MABS), Vol. 1 pp 177-186 (1993).

front of the office building. All of the buildings are assumed to be

rigid. Figure 6 shows the pressure contour at a time 0.15 secs after
charge detonation. At this time the blast wave has expanded
spherically over the top of buildings 1 and 2 and is about to reach the
office block. The channeling effect along the street is evidenced by the
high pressure at the end of the street near the ground.
The computed pressure history at the base of the office block
directly at the end of the street is shown in Figure 7. The free field
blast derived analytically is also depicted in this figure for
comparison. The free field curve assumes that the blast expands
hemispherically over a flat surface before reflecting from the office
building. The channeling of the blast is shown to increase peak
overpressure by 153% and maximum impulse by 340% compared
with the simplified analytic result. The distribution of maximum
impulse applied to the front surface of the office building is shown in
Figure 8 where the high impulse at the end of the street is clearly
shown.

Figure 6. Street channeled blast pressures 0.15 seconds

after charge detonation

80m

Office Block

Building 2

Building 1

Figure 7. Street channeled blast pressure time history at

base of office block

1000 Kg TNT

Figure 5. Street channeled blast analysis geometry

Figure 8. Street channeled blast maximum impulse on the

office block

10. EXPLOSION AND BLAST WITHIN A 3-D BUILDING

(EULER/ALE)
The calculation of a confined explosion within a simplified 3story building is used to illustrate a coupled 3D Euler/ALE (Arbitrary
Lagrange Euler) technique. In this type of approach the building
structure is no longer assumed to be rigid. This necessitates a much
more complex fluid-structure interaction computation. The basic
geometry of the simulation is shown in Figure 9. In the simulation, all
of the internal walls, floors and roof are treated as deformable
Lagrangian steel shell structures while the air and explosive are
computed using an Eulerian/ALE formulation. In the simulation, a 50
kg TNT charge is located in one of the rooms as shown.

Explosive located at 3 m from

walls and 2 m from floor

9m

Figure 10b. Detail location of TNT charge inside room

10 m
30 meters

Figure 9. 3D building with internal explosion

Figure 10a shows the model of the steel structure. Figure 10b
shows the detail of the initial explosive region (a portion of the
numerical mesh is peeled away to expose the inside). An efficient
approach, which a number of researchers employ, is to establish the
initial state for the 3D analysis from a separate detailed 1D spherical
explosion analysis. The 1D profile of energy, density, velocity,
pressure etc. is then mapped into the 3D grid as a starting point for the
3D calculation. The plot in Figure 10c shows the numerical mesh,
material locations, and the deformed state of the internal walls and
roof at 24 msecs after detonation of the explosive device. In this
Euler/ALE simulation, Euler is used for the explosive-air regions and
Lagrange/Finite Element for the deformable structures. Thus, fluidstructure interaction may readily be taken into account. Further details
of this type of coupled technique are provided in reference 7.

Figure 10c. Deformation of walls and roof (explosive/air not

depicted)

11. OKLAHOMA CITY FEDERAL BUILDING BOMBING CASE

STUDY(2D and 3D EULER)
This is a detailed case study which illustrates the methodology
one may employ to effectively quantify the blast loads for a real
case. The above numerical techniques are employed in one, two, and
three-dimensions.
Figure 10a. Explosion inside building, model setup

11.1 Objectives
On April 19, 1995 a terrorist explosive device was detonated
outside the Alfred P. Murrah Federal Building in Oklahoma City
causing severe damage to 75 buildings. Nearly half of the 200ft x 70ft
federal building tower was brought down along with a major loss of
life. To gain further information relating to the loads which the

N. Birnbaum et al, Calculation of Blast Loading in the High

Performance Magazine with AUTODYN-3D, 26th
Department of Defense Explosive Safety Seminar, Miami
(1994).

explosive device imposed on the building, a numerical analysis of the

event using 2D and 3D analysis software was carried out.
The objectives of this work were to:

Model the detonation and expansion of a 4800lb ANFO

(Ammonium Nitrate Fuel Oil) threat.

Assess the loads that a 4,800lb ANFO explosion would impose

on the Murrah Building assuming the facades to be rigid. This
approach allows the overall load levels around the building to
be initially established without the added expense and
complexity of a coupled analysis wherein the building
structure responds to the blast loading.

Use only the information available in public domain literature

Achieve the above objectives using generally available

numerical techniques

When the blast wave reaches the end of the 1D mesh, the solution
profile of density, energy, velocity, and pressure is saved. This is then
mapped into the 3D mesh as an initial condition.
200ft

Parking Lot
(20ft tall)
70ft

45ft

70ft

330ft x 200ft high

Figure 12. Geometry and Extent of Numerical Model used

for Full 3D Numerical Analysis (Large Model)
11.2.2 3D Analysis of Blast Wave Expansion/
Interaction with Building (Large Model)
A high order 3D FCT approach is used to calculate the overall
loads on the entire building structure. The method used in this analysis
was optimized to be applicable only to ideal gas behavior coupled
with rigid, rectilinear structures. This approach allows the efficient use
of very large computational domains. In a full Euler treatment,
allowing such features as viscosity, material strengths, and multiple
materials, many more variables must be carried at each computational
element. The result is that very large models may not be practical.
However, by reducing the complexity of the material behavior very
large problems may be simulated.
Figure 12 shows a plan view of the overall dimensions and
location of the explosive device. The structures are assumed rigid and
the initial state of the explosion is mapped from the 1D spherical
model. The rest of the computational domain is initialized with air at
ambient atmospheric conditions.
A number of pressure gauges are placed around the structure to
monitor pressures and impulse time histories.
11.2.3 Large Model Results
The analysis required 12 hours of CPU time running on a DEC
Alpha/3000-400 workstation. The final results are shown in Figures 13
to 16 and take the form of:

Sequence of overpressure contour plots - front elevation of

building

Contour plot of impulse on front elevation

Contour plot of overpressure - rear elevation of building

Pressure time history plots at selected gauge points on the

front and rear of the building
Other contour and time history plots of variables such as impulse,
dynamic pressure, and velocities are also of interest but not shown
here.

J=2
I = 451

I=1
r = 726.5 mm

Explosive Device

420,000 numerical cells: 3ft x 3ft x 3ft

11.2.1 Detonation/Initial Expansion of device

The explosive device used at Oklahoma City was estimated to
consist of 4800 lbs ANFO + 1 cartridge of TNT as a booster. The
explosive performance and in particular the numerical modeling of
ANFO are generally not well characterized. Therefore, it was
assumed to use a TNT equivalency of 0.6lb of TNT per lb of ANFO.
The explosive detonation and expansion was then modeled as
equivalent TNT of 2880 lbs using the well established JWL equation
of state data. For this analysis, the assumption was made that the
device was resting on rigid ground. To set the initial conditions for the
3D analysis a 1D spherically symmetric simulation was performed of
a sphere of twice the actual charge weight (5760 lbs). Figure 11
illustrates the 1D model setup.

Air (101.33 kPa)

230ft

15ft

11.2 Numerical Analysis

Two analyses were performed: A Large model which
encompassed the entire building and parking structure and a Small
model with higher resolution which included only the north facade and
east corner of the building.
Each numerical analysis was performed in two steps:

Modeling of the 1D detonation and initial expansion of the

4,800lb threat before interaction with the building takes
place using a 2D code

Remap the results of the 1D analysis into a 3D numerical

grid and analyze using a 3D code
The loads on the building can then be assessed from the results of
the analysis by means of contour plots of pressure and impulse over
the surface of the building and recordings of pressure variations with
time at designated gauge points on the structure.

2619 kg Sphere TNT

Federal Building
(130ft tall)

J=1
r = 4700 mm

Figure 11. 1D Spherically Symmetric Model Setup

Figure 13a. Full building analysis, blast overpressure, front

elevation, t = 0

Figure 13c: Full building analysis, blast overpressure, front

elevation, t = 30msecs

Figure 13b. Full building analysis, blast overpressure, front

elevation, t = 5 msecs

Figure 13d: Full building analysis, blast overpressure, front

elevation, t = 70msecs

Figure 15. Full Building analysis, blast overpressure at t =

160 msecs, rear elevation

Figure 13e: Full building analysis, blast overpressure, front

elevation, t = 120 msecs

Gauge #3

Gauge #7

Gauge #9

Gauge #8
Gauge #14

Figure 16. Typical pressure records front and rear locations

`
11.2.4 Large Model Conclusions
Typically, it is not possible to accurately resolve the peak
pressures and impulses at all locations of the building using a single
analysis because of the large range in scaled distances involved in a
particular problem. For example, to accurately resolve the peak
pressures at all length scales it is estimated that approximately
250,000,000 cells of size 0.125m x 0.125m x 0.125m would be
required. While it would not be impossible to attempt such a
simulation with that resolution, such a problem would tax the
resources of even the largest supercomputers currently available. By
way of illustration, a recent state-of-the-art calculation of the World

Figure 14: Full building analysis, impulse at 70 msecs, front

elevation

Trade Center bombing was modeled using ~18,000,000 cells (an order
of magnitude smaller than indicated above). The calculation, as well
as the associated data handling, imposed a very heavy load on a
modern CRAY supercomputer.8
Therefore, as a practical matter, for a full investigation into the
loading applied to the building, regardless of the computer resources
available, one typically needs to run several analyses covering
different scale ranges. For example, it is suggested that at least three
analyses be used to obtain reasonably accurate predictions over all the
building facades:
i) Entire building (1m x 1m x 1m cells) - Large Model
ii) North facade and East corner of building (0.5m x 0.5m x 0.5m
cells) - Small Model
iii) North facade local to device (0.125m x 0.125m x 0.125m cells) extent depends on
particular feature of the building which
is of interest
Figure 18a. Small model overpressure, initial condition
11.2.5 3D Blast Interaction with Building (Small Model)
To illustrate the above point, a local analysis of the North facade
and East corner of the building (see Figure 17) was performed using
double the spatial resolution (relative to the Large Model). This
results in a model of 560,000 cells. Figures 18 illustrate some of the
Small Model results.

10m
Building North Facade

5m
16m

25m
40m

14m

Figure 18b. Small model overpressure, t = 40 msecs

70m x 40m High

560,000 cells (0.5m x 0.5m x 0.5m)

Figure 17. Geometry and Extent for 3D Numerical Analysis

of North Facade and East Corner (Small Model)

Figure 18c. Small model overpressure, t = 82 msecs

This smaller model provides better resolution of the peak
pressures and impulses in the immediate vicinity of the explosive
device, important in defining the response of the structure and its
glazing.

Baum et al, Numerical Simulation of Blast In The World

Trade Center, American Institute of Aeronautics and
Astronautics, Inc. AIAA 95-0085 (1995) .

10

11.3 Oklahoma City case study conclusions

Several numerical analyses of the bombing of the Alfred P. Murrah
Federal Building in Oklahoma City were carried out using currently
available CFD techniques. The main conclusions to be drawn are:
A 1D/2D approach can be used for the analysis of the detonation
and initial expansion of HE assuming a spherically symmetric
expansion.
Computational efficiency is achieved by remapping the results of
the symmetric analysis into a 3D numerical grid as initial
conditions.
An optimized FCT Euler processor can be used efficiently to
simulate the expansion and interaction of a blast wave against rigid
buildings and surfaces. This conservative approach allows the
overall load levels around the building to initially be established
without the added expense and complexity of a coupled analysis
involving a responding building structure.
Typically, even for those with access to supercomputer power, it is
not possible to accurately resolve the peak pressures and impulses
on all building surfaces using a single model of the whole building.
However, a series of localized models can be used to help accurately
resolve the solution in the areas of concern.
The loads on the building can be assessed by means of contour plots
of pressure/impulse over the building surfaces and via recordings of
variable time histories at specified locations in the numerical grid.
12. SUMMARY AND FUTURE WORK
Currently available analytic and numerical techniques can be
used to effectively assess the loadings on buildings resulting from the
detonation of high explosive devices. Numerical techniques
encompassing Eulerian, Lagrangian, ALE, and structural
representations can address different aspects of the loading and
response of buildings. Three dimensional simulations would ideally be
finely resolved using on the order of 106-107 cells or more, which may
require the breaking up of a larger problem into smaller detailed
analyses, even for those with access to a supercomputer.
Further work needs to be done to include additional features in
the numerical techniques used for explosive-structure interaction.
Many of the numerical simulation methods assume only rigid nonresponding structures. A fully coupled fluid-structure method
including such effects as glass breakage, flows into the building, and
structural deformation would allow a better estimate of the physical
blast loads as well as insight into such issues as progressive failure of
the building structure.
Specialized Euler methods, such as the optimized FCT approach
discussed, should be used to extend high resolution blast calculations
to 107 cells or more using supercomputer facilities and massively
parallel processors. Such calculations can provide important
benchmarks and guidance in the use of smaller, less refined, analyses
for broad use by design engineers.
Sophisticated and detailed blast load analyses can and should be
regularly performed for all large public buildings. The accurate
prediction of these loads provides the building designer with the
necessary information to allow for better protection for the building
and its people. Consideration should be given to making blast
resistance, as is presently done for fire, a part of the building code for
all large public buildings.

11