Sie sind auf Seite 1von 124

BLAST RESISTANT DESIGN FOR ROOF SYSTEMS

A Thesis presented to the Faculty of the Graduate School of the University of Missouri Columbia

In Partial Fulfillment of the Requirements for the Degree Master of Science

by

MARK ANDREW MCCLENDON


Dr. Hani Salim, Thesis Supervisor

DECEMBER 2007

The undersigned, as appointed by the Dean of the Graduate School, have entitled the thesis entitled

BLAST RESISTANT DESIGN FOR ROOF SYSTEMS

Presented by Mark A McClendon a candidate for the degree of Master of Science in Civil Engineering and hereby certify that in their opinion it is worthy of acceptance.

________________________________________________ Dr. Hani Salim

________________________________________________ Dr. Sam Kiger

________________________________________________ Dr. Craig Kluever

ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to my advisor, Dr. Hani Salim, Assistant Professor of Civil Engineering, University of Missouri-Columbia. His guidance has been invaluable during the course of this study. His good nature, patience, frankness, and technical expertise had a profound impact on my academic experience and personal goals. I would also like to thank Dr. Sam Kiger, C.W. La Pierre Distinguished Professor and Director of the Center for Explosion Resistant Design, University of MissouriColumbia. His enthusiasm and technical knowledge helped guide this investigation. A special thanks is given to Dr. Perry Green, Technical Director, Steel Joist Institute and especially Tim Holtermann from Canam Steel for donating testing materials and continuous design input. Gratitude is extended to my fellow students, Aaron Saucier, Rhett Johnson, John Hoemann, Tyler Oesch, and others whose efforts were in my favor. Without their hard work and dedication, testing would have been overwhelming. Furthermore, I would like to thank all the members at the University of Missouri who have provided assistance. Finally, I extend my sincerest thanks to my friends who have kept me sane throughout my academic experience, and my family, especially my mother Verna and my aunt Wilhelmina.

ii

TABLE OF CONTENTS
ACKNOWLEDGEMENTS..............................................................................................ii LIST OF ILLUSTRATIONS...........................................................................................vi LIST OF TABLES...........................................................................................................xii ABSTRACT....................................................................................................................xiii

Introduction 1.1 Problem Statement............................................................................1 1.2 Thesis Objective.................................................................................2 1.3 Thesis Overview.................................................................................3

Review of Literature 2.1 Introduction.......................................................................................4 2.2 Types of Blast Loads.........................................................................4 2.3 Effects on Structures.........................................................................8 2.4 Blast Resistant Design.......................................................................9 2.4.1 2.4.2 Single-Degree of Freedom Models..................................10 CONWEP..........................................................................14

2.5 Roof Systems....................................................................................14 2.5.1 2.5.2 Equivalent Blast Load.....................................................14 Resistance Functions........................................................16

iii

Verification of Equivalent Blast Load Procedure 3.1 Introduction.....................................................................................19 3.2 Review of ANSYS LS-DYNA.........................................................20 3.3 Static Load Simulations..................................................................21 3.3.1 3.3.2 3.3.3 3.3.4 Preprocessing....................................................................22 Loading and Solution.......................................................24 Post Processing.................................................................24 Results...............................................................................24

3.4 Dynamic Load Simulations.............................................................31 3.4.1 3.4.2 3.4.3 3.4.4 Preprocessing....................................................................32 Loading and Solution.......................................................34 Post Processing.................................................................35 Results...............................................................................35

3.5 Blast Load Simulations...................................................................43 3.5.1 3.5.2 3.5.3 3.5.4 Application of Blast Loads..............................................43 Results (Positive Phase Only) .........................................45 Results (Positive and Negative Phase) ...........................51 Equivalent Blast Loading................................................55

3.6 Field Test..........................................................................................60 3.7 Summary..........................................................................................72 4 Static Resistance Function of Open Web Steel Joists 4.1 Introduction.....................................................................................73 4.2 Analytical Resistance Function......................................................74

iv

4.3 Experimental Verification..............................................................78 4.3.1 4.3.2 4.3.3 4.3.4 Testing Samples................................................................78 Test Set-Up.......................................................................80 Testing Apparatus............................................................85 Results...............................................................................86 4.3.4.1 16K5 Joist Test.......................................................86 4.3.4.2 26K5 Joist Test.......................................................93 4.3.4.3 32LH06 Joist Test..................................................99 4.4 Summary........................................................................................104 5 Conclusion and Recommendations 5.1 Conclusions....................................................................................105 5.2 Recommendations..........................................................................107 REFERENCES...............................................................................................................108

LIST OF ILLUSTRATIONS
Figure 2-1: Generalized Blast Pressure History.............................................................6 Figure 2-2: Blast Loading on Structure...........................................................................8 Figure 2-3: Damped, Single-Degree-of-Freedom System.............................................10 Figure 2-4: Idealized Dynamic Response Curves for Triangular Loading................12 Figure 2-5: Equivalent SDOF System............................................................................13 Figure 2-6: Equivalent Load Factor and Blast Wave Location Ratio........................15 Figure 2-7: DAHS Equivalent Loading Technique......................................................16 Figure 2-8: DAHS Equivalent Load...............................................................................16 Figure 3-1: BEAM 4 ANSYS element.........................................................................23 Figure 3-2: Static Loading #1; Maximum loading = 375 lb/ft.....................................25 Figure 3-3: Static Load #2; Maximum Load = 375 lb/ft..............................................26 Figure 3-4: Static Load #3; Maximum Loading = 375 lbs/ft........................................27 Figure 3-5: Static Load #4...............................................................................................28 Figure 3-6: Static Load #5...............................................................................................29 Figure 3-7: Percentage Error of Maximum Deflection................................................30 Figure 3-8: Percentage of Error of Max Deflection Location......................................31 Figure 3-9: BEAM 161 ANSYS LS-DYNA element..................................................33 Figure 3-10: Dynamic Pulse Load #1.............................................................................37 Figure 3-11: Dynamic Load #2.......................................................................................38 Figure 3-12: Dynamic Load #3.......................................................................................39 vi

Figure 3-13: Percentage Error in Maximum Response................................................40 Figure 3-14: Percentage Error in Time of Maximum Response.................................41 Figure 3-15: Percentage Error in Response Period......................................................42 Figure 3-16: Locations of CONWEP measurements....................................................44 Figure 3-17: Blast Pressure Distribution.......................................................................44 Figure 3-18: Actual Blast Loading.................................................................................46 Figure 3-19: Loading Scenario #1..................................................................................46 Figure 3-20: Impulse Comparison.................................................................................47 Figure 3-21: Loading Scenario #2..................................................................................47 Figure 3-22: Loading Scenario #3..................................................................................48 Figure 3-23: Impulse Comparison..................................................................................49 Figure 3-24a: Response Comparison of Loading Scenarios using LS-DYNA...........49 Figure 3-24b: Response Comparison of Loading Scenarios using LS-DYNA (without Scen. #1).............................................................................................................50 Figure 3-25: Dynamic Response for Loading Scenario #2...........................................51 Figure 3-26: Actual Blast Load Positive and Negative Phase...................................52 Figure 3-27: Loading Scenario #1 Positive and Negative Phase...............................52 Figure 3-28: Loading Scenario #2 Positive and Negative Phase...............................53 Figure 3-29: Negative Impulse Comparison..................................................................53 Figure 3-30: Response of Scenario #2 Positive and Negative Phase.........................54 Figure 3-31: Response Comparison of Positive Only and Pos. & Negative Data for Loading Scenario #2.........................................................................................55 Figure 3-32: TM-855 Equivalent Blast Load................................................................57

vii

Figure 3-33: Verification of Equivalent Load Response..............................................58 Figure 3-34: Comparison of Dynamic Response...........................................................59 Figure 3-35: FRP Panel Test...........................................................................................60 Figure 3-36: Static Resistance Function for FRP Panels.............................................61 Figure 3-37: Roof Panel Schematic................................................................................63 Figure 3-38: Pressure-time history at 35 ft....................................................................64 Figure 3-39: Impulse-time history at 35 ft.....................................................................65 Figure 3-40: Pressure-time history at 45 ft....................................................................66 Figure 3-41: Impulse-time history at 45 ft.....................................................................66 Figure 3-42: Pressure-time history at 55 ft....................................................................67 Figure 3-43: Impulse-time history at 55 ft.....................................................................67 Figure 3-44: Deflection of R1 panel at near quarter-point..........................................68 Figure 3-45: Deflection of R1 panel at far quarter-point.............................................69 Figure 3-46: Deflection of R1 panel at midpoint...........................................................70 Figure 3-47: Deflection of R2 Panel at near quarter-point..........................................71 Figure 3-48: Deflection of R2 Panel at far quarter-point.............................................71 Figure 3-49: Deflection of R2 Panel at midpoint...........................................................72 Figure 4-1: Resistance Function for 16K5 Joist............................................................75 Figure 4-2: Resistance Function for 26K5 Joist............................................................75 Figure 4-3: Resistance Function for 32LH06 Joist.......................................................76 Figure 4-4: Maximum deflection of elasto-plastic, one-degree-of-freedom system for triangular pulse load.........................................................................................77 Figure 4-5: Bearing seat plates for 16K5 and 26K5 Joists...........................................81

viii

Figure 4-6: Bearing Seat Plates for 32LH06 Joist.........................................................82 Figure 4-7: Steel Joist Institute Specifications for horizontal bridging......................82 Figure 4-8: Lateral bracing welded to strong floor...................................................83 Figure 4-9: Lateral bracing placement.......................................................................84 Figure 4-10: String Potentiometer..................................................................................84 Figure 4-11: 16-Point Loading Tree...............................................................................86 Figure 4-12: 16K5 trusses prior to testing.....................................................................87 Figure 4-13: 16K5 Joist Failure Sequence 1 of 5 Initial bending..........................88 Figure 4-14: 16K5 Joist Failure Sequence 2a of 5 Failure of lateral bracing.......88 Figure 4-15: 16K5 Joist Failure Sequence 2b of 5 Failure of lateral bracing......89 Figure 4-16: 16K5 Joist Failure Sequence 3 of 5 Failure of horizontal bridging..............................................................................................................89 Figure 4-17: 16K5 Joist Failure Sequence 4 of 5 Continued out-of-plane bending...............................................................................................................90 Figure 4-18: 16K5 Joist Failure Sequence 5 of 5 Failure of bearing seat weld.....................................................................................................................90 Figure 4-19: Static Response for 26K5 Joist System....................................................91 Figure 4-20: Midpoint Static Response for an individual 16K5 joist compared to existing methods................................................................................................92 Figure 4-21: 26K5 Joist System prior to test.................................................................94 Figure 4-22: 26K5 Joist Failure Sequence 1 of 5 Deformation of Tension Chord.................................................................................................................94

ix

Figure 4-23: 26K5 Joist Failure Sequence 2a of 5 Failure of End Tension Member..............................................................................................................95 Figure 4-24: 26K5 Joist Failure Sequence 2b of 5 Failure of End Tension Member..............................................................................................................95 Figure 4-25: 26K5 Joist Failure Sequence 3 of 5 Failure of Secondary Web Member..............................................................................................................96 Figure 4-26: 26K5 Joist Failure Sequence 4a of 5 Failure of End Tension Member..............................................................................................................96 Figure 4-27: 26K5 Joist Failure Sequence 4b of 5 Failure of End Tension Member..............................................................................................................97 Figure 4-28: 26K5 Joist Failure Sequence 5a of 5 Connection Plate Failure.......97 Figure 4-29: 26K5 Joist Failure Sequence 5b of 5 Connection Plate Failure.......98 Figure 4-30: Static Response for 26K5 Joist System....................................................98 Figure 4-31: Midpoint Static Response for an individual 26K5 joist compared to existing methods................................................................................................99 Figure 4-32: Initial Deformation of 32LH06 Joist......................................................100 Figure 4-33: 32LH06 Joist System prior to test..........................................................101 Figure 4-34: 32LH06 Joist Failure Sequence 1a of 1 Compression Web Member Buckling...........................................................................................................101 Figure 4-35: 32LH06 Joist Failure Sequence 1b of 1 Buckling of Secondary Web Member............................................................................................................102 Figure 4-36: 32LH06 Joist Failure Sequence 1c of 1 Continued Buckling and Bending of Tension Chord.............................................................................102

Figure 4-37: Static Response of 32LH06 Joist System...............................................103 Figure 4-38: Midpoint Static Response of individual 32LH06 Joist comparing existing methods..............................................................................................103

xi

LIST OF TABLES
Table 3-1: Minimum Time Steps by Element Size.......................................................36 Table 3-2: Aspect Ratio...................................................................................................36 Table 3-3: Peak Pressure vs. Range...............................................................................44 Table 3-4: CONWEP Blast Data....................................................................................56 Table 4-1: Resistance Data from Engineering Calculations........................................77 Table 4-2: Resistance Data from SBEDS Calculations................................................78

xii

ABSTRACT
The design of structures to resist explosive loads has become more of a concern to the engineering community. Different structural members, such as walls, have been thoroughly evaluated under blast loads. This research focuses on the design techniques for the loading on roof structures and the resistance of open web steel joists, a common roof component. Blast loads are dynamic, impulsive and non-simultaneous over the length of a roof. To design against explosions, a procedure has been developed to devise a uniform dynamic load on a roof that matches the response from blast loads. The objective of this research is to test this procedure and compare its results to the deflections from blast loads. This research uses finite element analysis to evaluate the responses from numerically calculated blast loads and compares them to the equivalent loading response. The numerical pressures are calculated using the Conventional Weapons Effects Program (CONWEP) (Hyde, 1992) and the Single-degree-of-freedom Blast Effects Design Spreadsheet (SBEDS) from the Army Corps of Engineers Protection Design Center. Also, the response of experimentally measured roof blast pressures is compared to the equivalent loading response. While the responses from finite element modeling matched the experimental responses, the equivalent loading procedure did not adequately predict the initial peak deflection or the maximum deflection.

xiii

The response of several structural members used in roof construction, such as hot-rolled steel beams and reinforced concrete slabs, are well documented and understood. Open web steel joists (OWSJ) are other types of common roof components. Their responses under loading are not clearly defined, and current methods extrapolate techniques used in the design and analysis of hot-rolled steel beams and reinforced concrete. The resistance function currently used for these members are linear elastic and perfectly plastic after the elastic deflection limit. It is believed that the failure mechanisms of OWSJ significantly are not accurately being taken into account. Three tests consisting of different steel joist pairs are performed. The resistance function is computed from these results and compared to current methodologies. The current resistance methods calculate larger maximum loads than the experimental values and the assumption of a perfect plastic post-peak response ignores the buckling failure of web members. It is recommended that additional research is to be done on the prediction of blast pressures on roofs and on the development of an equivalent uniform dynamic load. It is also recommended that an analytical resistance function for OWSJ be clearly defined, which includes all failure limit states.

xiv

Chapter

Introduction

1.1

Problem Statement
Many research has been performed on the design of structures under explosive

threats. The design and response for a wall system subjected to blast loading and its components are very well known. However, there is a gap in the knowledge for the roof system component under explosive loads. The analysis of any system with dynamic loads, such as blast pressure due to a bomb, can be generalized with an equation of motion as shown in Eqn. 1.1. This equation can be solved by various numerical procedures to calculate the dynamic response (Biggs, 1967). The parameters in this equation are the mass, the resistance, and the load.

&& + R = F (t ) My
where M = mass R = resistance F(t) = applied load

(1.1)

While the mass of any system can be easily calculated, the resistance and load are more complex. Explosive loads are highly impulsive, non-simultaneous, and nonuniform. For walls, the blast pressures are mostly uniform on the entire face of the wall, and they can be accurately estimated using available codes such as CONWEP (Hyde, 1992). In roof systems, these loading pressures change with respect to time and with respect to distance. There is currently a loading procedure developed by the Army Corps of Engineers that equate the response from a blast load to the response from an equivalent uniform dynamic load that is more suited to design (UFC, 2002). However, this procedure has not been verified using experimental or numerical data, as far as the author is aware. Therefore, there is a need to explore and validate the methods and results this technique uses. The resistance of many structural systems used in construction of roof slabs is known. In addition to concrete and hot-rolled steel members, open web steel joists are also common roof components, and they are in a high demand of use due to their low weight and relatively high resistance. Current techniques used in design assume a linear elastic, pure plastic load-deflection curve. It is believed that such truss systems exhibit different failure modes than those currently used in design. Due to their high use in military and commercial buildings, current design practices should be researched.

1.2

Thesis Objective
The overall objective of this research is to develop an analysis and design

procedure for open web steel joist roof system under blast loading. To achieve this goal, the following two specific tasks are realized:

Analyze the roof loading procedure specified by the United Facilities Criteria (UFC 3-340-01, 2002) and compare its validity using numerical simulation and field test data.

Analyze and compare current methods of developing the resistance function for open web steel joists with experimental data using static tests.

1.3

Thesis Overview
Chapter 2 covers a literature review of the methods and techniques used in the

thesis. This chapter includes an explanation detailing the current knowledge of explosives and blast waves. It also contains numerical integration techniques for dynamic response calculation. The equivalent blast load procedure is discussed, as well as methods for determining the resistance function for open web steel joists. In Chapter 3 the ANSYS LS-DYNA program is verified for static and dynamic loads before using the program to develop responses for roof models subjected to blast loads. These responses are compared to the response from the equivalent blast loading. Also, the equivalent blast loading procedure is compared to field test data of a roof subjected to explosive loads. In Chapter 4, the static testing procedure for open web steel joist samples is presented. The results from the static tests are compared to current resistance function techniques used by design engineers. The failure mechanisms present in open web steel joists are also discussed. Finally, Chapter 5 summarizes the analysis of the research and presents recommendations for future work.

Chapter

Literature Review

2.1

Introduction
Understanding and solving the problem statement requires basic knowledge of

blast effects and the responses for roof structures. In this chapter current techniques used to describe blast effects will be discussed. In addition, several design methods for roof structures will be introduced.

2.2

Blast Waves
Explosions can be caused by physical, nuclear, or chemical events. A common

example of a physical explosion occurs when a pressure vessel fails. Pressure is released, along with fragments of the vessel. Changing the structure of atomic nuclei produces nuclear explosions. Vast amounts of energy can be released in short periods by breaking the bonds between protons and neutrons. Nuclear fission divides the nuclei of heavy atoms. Nuclear fusion combines nuclei from light atoms. Nuclear explosions generate kinetic energy, internal energy, and thermal energy. 4

The rapid oxidation of fuel elements develops chemical explosions. This reaction releases heat and produces gas, which expands. Low-end explosives create quasi-static loads. High explosives (chemical and nuclear) in a surrounding medium, such as air or water, cause shock waves in the medium. The blast releases high-pressure gases at high temperatures. These gases naturally expand, and the surrounding medium is consequently compressed (Smith and Hetherington, 1994). The compressed medium, or for the specific case of air, forms a shock front. The shock front travels in a radial direction. As the explosive gases cool and slow their movement, the amount of overpressure the shock front carries decreases. The gases release energy to reach equilibrium towards the atmospheric pressure. However, due to the high pressure and mass of the gases, more expansion is necessary to actually reach equilibrium. This causes the pressure in the shock wave to drop below the atmospheric pressure. After sufficient underpressure is expended, the state returns to the atmospheric pressure. The air behind the shock front also places a load, a drag force, on objects encountered (Smith and Hetherington, 1994). The general shape of a pulse shape is shown in Figure 2-1. Important factors pertinent to burst pressures include the peak pressure, the duration, the air density behind the shock front, the velocity of the shock front, and the impulse of the blast pressure. There are several derived equations that calculate the shock front velocity Us, peak dynamic pressure qs, and air density behind the shock front s based on the peak overpressure, ambient air pressure, and the speed of sound in air at the ambient pressure. Examples of some of these equations are shown in Eqns. 3.1, 3.2, 3.3 (Smith and Hetherington, 1994).

Figure 2-1: Generalized Blast Pressure History (TM 5-1300, 1990).

Us =

6 p s + 7 p0 a0 7 p0

(2.1)

s =

6 p s + 7 p0 0 p s + 7 p0
2

(2.2)

5 ps qs = 2( p s + 7 p0 )
where ps = peak static overpressure

(2.3)

p0 = ambient air pressure in front of shock wave 0 = air density in front of shock wave a0 = speed of sound in air at air pressure

The peak overpressure is related to a factor called the scaled distance, Z (Eqn. 2.4). This is proportional to the distance from the charge and the cubed root of the charge mass. Typically the charge mass is measured in terms of TNT, and other types of explosives are converted to this mass type. As the distance increases, the maximum pressure of the shock wave decreases. The total duration of the shock burst actually increases. It should also be noted that at any particular range, the peak overpressure of the blast wave decays exponentially to the atmospheric pressure (Biggs, 1967).

Z=

R W 1/ 3

(2.4)

where R = distance from blast source W = mass of charge in terms of TNT

When blast waves strike a surface, the overpressure increases. The pressure from the expanding gases build up since there is no medium to compress and displace. Therefore, the burst pressures from a surface explosion are larger than an explosion occurring in the air. The overpressure of a surface burst is approximately twice that of a free-air burst (Smith and Hetherington, 1994). Next, the discussion of blast loads continues to how they affect buildings and objects.

2.3

Effects on Structures
Three main loading conditions are available, as explained by Smith and

Hetherington, 1994. In the first type a relatively large shock wave reaches a structure relatively small enough that the blast wave encloses the entire structure. The shock wave effectively acts on the entire structure simultaneously. Additionally, there is a drag force from the rapidly moving wind behind the blast wave. The structure is, however, massive enough to resist translation. The second condition also involves a relatively large shock wave and a target much smaller than the previous case. The same phenomena happen during this case, but the target is sufficiently small enough to be moved by the dynamic, drag pressure. In the final case, the shock burst is too small to surround the structure simultaneously and the structure is too large to be shifted. Instead of simultaneous loading, each component is affected in succession. For a typical building, the front face is loaded with a reflected overpressure.

Figure 2-2: Blast Loading on Structure (Forbes, 1999). 8

As stated in the previous section, reflected pressure accumulated when the blast wave meets material denser than the medium in which it is traveling. Air molecules are stopped by this material and compressed further by the shock front behind them. This reflected pressure decays over time. Blast pressures simultaneously affect walls with the same pressure-time history on the entire wall area. This means that while the blast pressures change with respect to time, these pressures are evenly distributed on every portion of the wall. Also, due to the perpendicular orientation of the front wall to the shock wave, the blast pressure is reflected and magnified (Forbes, 1999). The roof and sides of the building react instead to incident or side-on pressures. Similar to how the peak pressure decays and load duration grows as the range from the charge increases, the incident pressure attenuates as it transverses the roof and sides, as shown in Fig 2-2. In the presence of a sloped roof, the pressure magnifies due to reflection. After the roof and sides are surpassed, the pressures converge on the back of the building. Again, a reflecting effect amplifies the overpressure. In addition to this blast wave diffraction, drag forces load the structure (UFC, 2002).

2.4

Blast Resistant Design


Engineers employ several different methods for the structural design of structures

resisting blast loads. This can range from simplified hand calculations to utilizing computer programs programmed to deal with complex loading and resistances. Any method used takes into account the knowledge of blast loads described in the previous section and the component resistances that make up the structure (Morison, 2006).

2.4.1 Single-Degree-of-Freedom Model


The first explosive design methods are based on the single-degree-of-freedom (SDOF) model. This model idealizes an entire structure or structural component as one point in the structure. The resistance at this point is also taken as the resistance for the entire structure. The equation of motion for a linear elastic, damped single-degree-offreedom system follows (Biggs, 1967).

&& + ky + cy & = F1 [ f (t )] My
where M = mass of structure k = structural stiffness c = damping coefficient F1 = constant force value f(t) = nondimensional time value

(2.5)

M F1

Figure 2-3: Damped, Single-Degree-of-Freedom System.

10

The dot superscripts represent partial derivatives with respect to time. On dot signifies velocity; two dots mean acceleration. For simplicity, the damping coefficient is assumed to be zero. The equation simulates the response of lumped mass-spring system (Biggs, 1967). Solution of this equation (without damping) leads to Duhammads, or Convolution, Integral (Eqn. 2.6). This gives the deflection-time history.

y = y 0 cost +

& y

sin t + y st f ( ) sin (t )d
0

(2.6)

where y0 = initial displacement & = initial velocity y = circular frequency


k/M

yst = static deflection F1/k f() = non-dimensional load-time function

The military has used two main SDOF methods to deal with explosive threats. The first technique, the Modal SDOF method, turns up in a 1946 manual Fundamentals of Protective Design (Non-Nuclear) EM 1110-345-405, reissued in 1965 as TM 5-855-1 and not superseded until 1986. In this method, the response period is taken as the natural period of the first mode under free vibration. Normalized curves were created to aid in calculation of maximum deflection for various dynamic loads and linear elastic, pure plastic resistance curves. The most notable dynamic load used was a triangular load with zero rise time. An example of this non-dimensional chart is shown. This method only analyzes simplified systems and was not ideal for more general loading histories and resistance functions. It also undercalculates the deflection response and reaction response (Morison, 2006).

11

Figure 2-4: Idealized Dynamic Response Curves for Triangular Loading (Morison, 2006).

The equivalent SDOF method was widely published in 1957 in parts of the USACE manual Design of Structures to Resist the Effects of Atomic Weapons, EM 1110-345-415 Principles of Dynamic Analysis and Design, and EM 1110-345-416 Structural Elements Subjected to Dynamic Loads. While the Modal SDOF method uses the actual mass and loading in the system, this alternative method determines equivalent values of mass, resistance, and loading for the lumped mass-spring system based on the distribution of the structures mass and the systems loading. Using the conservation of kinetic energy, internal strain energy and external work, transformation factors can be found to change the systems mass, resistance and loading into their lumped mass counterparts. These transformation factors are also based on the deflected shape function of the system as a whole from a particular reference point, usually the middle point of a structure (Biggs, 1967).

12

Accurate shape functions result in accurate solutions. When the shape function is not as well known, the solution becomes more approximate and more errors are entered. The static deflection shape under the same loading distribution as the dynamic load is relatively easy to calculate and results in very accurate answers. The equation of motion for this equivalent SDOF system is given in Eqns. 2.6a and 2.6b. This can be solved in the same manner as the Modal SDOF models. Analysis of the equation allows for the use of only one transformation factor KLM, the ratio of the mass factor to the load factor. It should also be noted that the transformation factors for load and resistance are the same (Biggs, 1967).

& = Fe (t ) M e && y + ke y
or

(2.6a)

&& + K L ky & = K L F (t ) K M My

(2.6b)

where KM = mass transformation factor KL = load/resistance transformation factor M = total mass k = stiffness F = total load

p k Figure 2-5: Equivalent SDOF System. 13 y

2.4.2 CONWEP
The Conventional Weapons Effect Program (CONWEP) (Hyde, 1992) uses blast test data and empirical equations to calculate load data for different types of explosives with varying loading sequences. This program outputs parameters of the pressure-time history including side-on and reflected pressure, incident and reflected impulse, time of arrival, positive phase duration, shock front velocity, and peak dynamic pressure. This program only computes data corresponding to the positive pulse of a blast load.

2.5

Roof Systems
This section discusses design techniques for roof systems subjected to blast loads.

Looking at the general equation of motion, the mass of a structure is a well-known value. The loading function and, more importantly, the resistance are the less realized components. Different methods for determining these quantities are presented.

2.5.1 Equivalent Blast Load


The Unified Facilities Criteria 3-340-01 (which supersedes Army Manual TM 5855-1/Airforce AFPAM 32 1147/NAVYFAC P-1080/DSWA DAHSCWEMAN-97) formulates an equivalent uniform blast loading for the roof and sidewalls of a structure. The shape of this load is a triangular pulse with an unequal rise and decay times. The equation is shown in Eqn. 2.7. The load factor is a function of the Lwb/L, which is the ratio of the length of the blast wave at the back of the roof to the length of the roof. The drag coefficient is a function of the dynamic pressure. This equivalent load is calculated using values from the CONWEP program.

14

Por = C E Psob + Cd q ob
where Por = peak pressure Psob = peak incident pressure at back of roof CE = equivalent load factor qob = dynamic pressure at back of roof Cd = side-on element dynamic drag coefficient

(2.7)

CONWEP data is taken at a range equal to the radial distance from the blast source to the back of the roof. The length of the shock wave is calculated from the shock front velocity and the positive phase duration of the shock front. Using Fig. 2-6, the load factor and the ration of D/L can be computed, where D is the point on the roof where the peak pressure is said to occur. The shape of the loading function, as shown in Fig. 2-8, demonstrates that the maximum pressure on the roof takes time to build. As the shock front transverses the roof, the changing pressure-time history accumulates to a peak value. In Fig. 2-8, tf is the time the shock wave hits the roof. The time to reach the peak pressure point is defined as td. The loading ends after the blast wave reaches the end of the roof and the duration of the loading expires, tb and tofb, respectively.

Figure 2-6: Equivalent Load Factor and Blast Wave Location Ratio (DAHS, 2002). 15

Figure 2-7: DAHS Equivalent Loading Technique.

Figure 2-8: DAHS Equivalent Load (TM 5-1300, 1990).

2.5.2 Resistance Function


There are several different types of roof structures. Some types include concrete slab and I-beam composites, open web steel joists (OWSJ) and metal decking, steel joists, purlins and decking.

16

For these systems, the I-beam and the steel truss are the significant components with respect to design. The behavior of I-beams is well-known and documented. However, the response of open web steel joists to failure is not well documented. Conventional engineering designs assume a simple bilinear linearly elastic, perfect plastic resistance function. The Steel Joist Institute (SJI) has developed charts to ascertain the maximum allowable load per length for every truss as a function of joist type and length (SJI, 2005). The allowable deflection is L/360, where L is the effective joist length. In addition, the Steel Joist Institute provides an equation for the approximate joist moment of inertia, based on the maximum live load and the effective length of the truss. This equation is described in Eqn. 2.8. The derivation for this moment of inertia originates from the stiffness of a simply supported beam under a uniform load and includes a 15% increase in deflection due to elongation in the truss web. Along with the equivalent SDOF method, current engineering practices define a resistance function shape for use in design.
I j = 26.767( w LL )( Lspan ) 3 (10 6 )

(2.8)

where wLL = maximum allowable live load of joist (lb/ft) Lspan = effective length (ft)

A current engineering method of deriving the resistance formula takes into account the properties of the joist members and the Steel Joist Institute load specifications. The truss can be designed as a simply supported beam with a uniformly distributed load. Using the design tensile strength of the web members as a limit, the elastic section modulus of the joist cross-section can be computed.

17

Using this value and the joist material characteristics, an effective moment capacity can be computed. The maximum moment for a simply supported beam under uniform load can be determined from the load and the length of the beam. Included into this maximum moment calculation are strength increase factors for the joist materials. Using this relationship, the ultimate resistance can be computed. Similarly, using the deflection equation for the same system, the stiffness and elastic deflection limit can be calculated. The maximum response is calculated using numerical techniques and/or idealized charts based on the ratios of load duration to the joist period and resistance magnitude to the forced load magnitude. SBEDS (2004) uses a slightly different approach. For one, the moment of inertia is calculated from the actual cross-sections of the joist, including the top chord, the bottom chord, and some factor for the web. These calculations prove to result in higher moments of inertia for a given steel truss compared to the Steel Joist Institutes approximate calculation. The program defines a value Lshear, which is the maximum joist length designed according to SJI with the maximum total allowable load. For K-Series joists this maximum total load is 550 lb/ft. For the deeper joists LH and DLH, this maximum shear load is based on individual truss and is detailed in the SJI Load Tables document. The effective moment capacity is calculated from the maximum shear load and the Lshear quantity, assuming a simply-supported beam with a distributed load. SBEDS program back-calculates a resistance value from the joist length and effective moment capacity. This resistance value is also factored using strength increase factors. The elastic deflection limit and stiffness are calculated using the simply supported beam assumption mentioned earlier.

18

Chapter

Verification of Equivalent Blast Load Procedure

3.1

Introduction
Blast loads are dynamic, impulsive, and non-uniform. An experimental blast load

has been formulated to emulate the response of a blast load acting on a roof system. This loading resembles a triangular pulse with a rise time and decay time. This loading shape has been in use for over fifty years. With the advent of more advanced computer technology and finite element analysis programs, current techniques for expressing blast loads on the roofs of structures can be evaluated. This chapter explains the programs and procedures used in blast and dynamic loading estimation.

19

3.2

Review of ANSYS Program


Finite element analysis is a numerical method of deconstructing a structure or

system into smaller, simpler discrete regions. These regions, also called finite elements, are linked and controlled by several governing equations, including equations of equilibrium, compatibility equations and constitutive relations. The computer solves these equations simultaneously using different numerical schemes to express the behavior of the structure as a whole. In general the number, size, and shape of the elements control the program results. ANSYS was created in 1970 by Dr. John Swanson (Swanson Analysis Systems, Inc.). The program ANSYS ED Release 9.0 is used in this research. This is a student level version meant for academic use. It supports many of the capabilities present in the full version of ANSYS Multiphysics with some limitations on types and size of elements, material types, and degrees of freedom. ANSYS is able to construct static and dynamic structural analysis (both linear and non-linear), heat transfer problems, fluid flow problems, acoustics, and electromagnetic analyses. The product ANSYS LS-DYNA is a result of the partnership between ANSYS, Inc. and Livermore Software Technology Corporation and was first introduced in 1996. It is an explicit non-linear structural simulation used for dynamic analyses. This is applicable for structures experiencing large deformations and short time durations, including impact test simulations, drop test problems, explosive simulations, and metal forming.

20

For ANSYS use, three main steps must be performed preprocessing, completing a solution, and post processing. First, a two- or three-dimensional model of the structure or system has to be formed. The entire structure can be modeled or parts of the structure can be modeled. Symmetry can also be used to simplify the construction of the model. Next, the different material properties that make up the model have to be described and applied to the models geometry. The system can now be discretized into elements. The number, size and type of element is based on the analysis type and the users judgment. Once the system is idealized into finite elements with their own properties, physical constraints and different loading types can be positioned on the model. The user directs the program to solve the problem and output results. These results can be illustrated using tables, graphs, or contour plots. The ANSYS program includes a graphical user interface (GUI) to aid in executing these steps. To gain confidence in ANSYS, static and dynamic analyses were performed using simple models. These analyses are also beneficial in validating the use of ANSYS in analytical experimentation. The next sections depict the methods used for static and dynamic loads.

3.3

Static Load Simulations


The ANSYS program was used to compare deflection curves derived from

kinematic boundary conditions with the results obtained from ANSYS finite element modeling. A linearly elastic, isotropic BEAM-type element was used in the modeling. Its dimensions were 4 x 18 x 20. The beam was modeled as being simply supported. The material of the beam is modeled as steel.

21

3.3.1 Preprocessing
The element type used in this analysis is called BEAM4. It is described as a uniaxial element with tension, compression, torsion, and bending capabilities. The element contains two nodes at either end. Each node (shown as I and J in Fig. 3-1) has six degrees of freedom from translations in the local x, y, and z directions and rotations in the local x-, y-, and z-axes. This element is used in static analyses to simulate the reactions of an elastic, untapered section in 3D. Figure 3-1 shows the rotational aspects of the beam element. The numbers depict the six different surfaces on the element. This technique is commonly used during the loading stage to determine on which face to apply the load. For different elements, the ANSYS program supplies different ways of detailing each section. The GUI offers an automated section development that allows users to choose the shape of the element cross-section and its dimensions. For BEAM4 elements, a set of real constants is inputted. These constants include the cross-sectional area, the weak and strong moments of inertia, the cross-sectional dimensions, and the mass density per unit length. Different material models for several structural materials are available in the ANSYS program. For this model, a linear elastic, isotropic material model was used. The modulus of elasticity (and also the mass density) is chosen for this particular material type.

22

Figure 3-1: BEAM 4 ANSYS element (ANSYS, 2004).

For the creation of a simple beam, keypoints are chosen as the locations of the beams ends. A straight line is drawn between these keypoints. A third keypoint is also drawn to define the orientation of the element. In the Figure 3-1 this point is denoted as node K. The next step is to mesh the line. This will apply elements along the lines geometry. The user defines the number and size of the finite elements. The ANSYS ED version limits the number of possible nodes, lines, and elements. The user can apply the real constants and material attributes to the element mesh and use some node K to orient the elements. The line is then meshed into the number of prescribed elements.

23

3.3.2 Loading and Solution


The structure under consideration is a simply supported beam. To that effect, the node at one end of the beam is constrained from movement in the x, y, and z directions; the node at the other end of the beam is constrained in the x and y direction. Next, loads are placed on one surface of the beam elements to simulate bending about the weak axis of the beams cross-section. This model is then solved.

3.3.3 Post-Processing
ANSYS provides several post processing options. A list of the nodes and positions in the global directions can be provided, or a graphical representation of the original and deflected position of the model can be drafted. A contour of the stresses in the beam model can be generated.

3.3.4 Results
Five different static load configurations were tested. Each configuration contained only distributed loads. For each static loading, three test beams were modeled. Each beam was divided into a different number of segments. The tests were designed to determine how the finite element model deflection compared to expected results and how discretization of the model affects results. Each static load was tested on a beam model consisting of 1 segments, 6 segments, or 1 segments, or conversely 20 beam elements, 40 beam elements, or 240 beam elements, respectively.

24

The first static load is a uniform static load. The loading amount is designed to result in a discernable deflection curve, while keeping the stress of the beam in the elastic range. As can be seen in Figure 3-2, the expected deflection curve as well as the different curves from the three test beams are fairly accurate. The displacement of the beam reaches maximum in the center of the beam, as expected. This loading also results in the greatest amount of deflection from all five curves. The shape of the results is consistent with known kinematics. From the graph, there appears to be little difference between the calculated values and each ANSYS test.

6 5 4 3 2 1 0 0 5 10 Distance from Support (ft) 15 20


400 350 300 250 200 150 100 50 0 0 5 10 15 20 Distance from Support (ft)

Defl. Eqn. 1' Beam Elements 6" Beam Elements 1" Beam Elements

Deflection (in)

Figure 3-2: Static Loading #1; Maximum loading = 375 lb/ft.

Load (plf)

25

The next static load is a triangular loading that decays over the length of the beam. The displacements for the three ANSYS tests also appear fairly close to expected answers. Due to the centroid of the loading, the maximum deflection is off the center of the beams length. The total load on the beam is half that of the uniform loading, and the range of the displacement reflects this.
3 2.5 Deflection (in) 2 1.5 1 0.5 0 0 5 10 Distance from Support (ft) 15 20
400 Load (plf) 300 200 100 0 0 5 10 15 20 Distance from Support (ft)
Defl.Eqn 1' Beam Element 6" Beam Element 1" Beam Element

Figure 3-3: Static Load #2; Maximum Load = 375 lb/ft.

The third static load is also triangular; however it rises to a peak halfway across the beams length and descends the rest of the length to zero. The total load of this curve is the same as that of Static Load #2. However, the centroid of the loading is at the center of the beams length and it is balanced about the center of the beam. Little difference can be discerned from the different test beams in this graph. The maximum loading is located in the center of the beam. The final two static loads are neither symmetrical nor are they continuous along the beam.

26

4 3.5 3 Deflection (in) 2.5


Load (plf)
Defl. Eqn. 1' Beam Element 6" Beam Element 1" Beam Element

2 1.5 1 0.5 0 0

400 300 200 100 0 0 5 10 15 Distance from Support (ft) 20

10 Distance from Support (ft)

15

20

Figure 3-4: Static Load #3; Maximum Loading = 375 lbs/ft.

Static Load #4 consists of three varying uniform loads acting on three different equal length sections of the beam. While the deflection curves of the previous three static loads can be easily found in a book or manual (AISC, 2001), the derivation of the deflections for Static Loads #4 and #5 involves calculating the internal moment in the beam and integrating twice. Also, the kinematic boundary conditions of the beam state that the slope and deflection must be continuous along the beam and at the changes in loading. The general shape of the displacement curves is customary to the expected results. The loads center of gravity is about 9 from a support, and the maximum beam dislocation is located about an inch from the center of the beams length. The total load is little more than half of Static Load #1, and the deflection is in the right scale.

27

The curves for the 1 length beam element and for the 6 length beam element are discernable Figure 3-5 around the center of the beam. The test with the smallest element length (conversely the largest number of elements) appears to coincide with that of the deflection equation.
3 2.5 2 1.5 1 0.5 0 0 5 10 Distance from Support (ft) 15 20
Load (plf) 300 200 100 0 0 5 10 15 20 Distance from Support (ft)
Defl. Eqn. 1' Beam Element 6" Beam Element 1" Beam Element

Deflection (in)

Figure 3-5: Static Load #4.

As can be seen by the deflection curves of Static Load #4, the 1 beam segment overestimates and the 6 curve underestimates the expected result. It can also be seen that the difference in expected and calculated values decreases with increased discretization. Static Load #5 has similar results. It is made up of three different uniform loads acting on equal length sections of the beam, along with one extra section that remains unloaded. The curves all underestimate the deflection, but they arrive closer to the expected value as the element size increases. Again, due to the loading schematic, the location of maximum deflection is offset from the center, and the curves reflect this. 28

1.5 1.25 Deflection (in) 1 0.75


Load (plf)
Defl. Eqn. 1' Beam Element 6" Beam Element 1" Beam Element

0.5 0.25 0 0

300 200 100 0 0 5 10 15 20 Distance from Support (ft)

10 Distance from Support (ft)

15

20

Figure 3-6: Static Load #5.

The ANSYS results appear to resemble the expected values, for the most part. To further examine this, the maximum deflection and the locations of maximum deflection are compared to the derived deflection equation. As can be seen from Figures 3-2 through 3-6, the continuous loads have less error than discontinuous loads. In addition, the symmetrical loads (Static Loads #1 and #3) have very small deviations from the expected values. And, as expected, the percentage of error decreases as the element size decreases.

29

Finally, the location along the beam model of the maximum displacement is compared. The number of elements in the model determines the number of nodes in the beam. Each node is spaced the length of an element, and so accuracy of the displacement and location of displacement is limited by the length of an element. For each loading, the location of the maximum deflection is no more than half a foot from the center of the beam. The closest the 1 beam element models can measure is 1, so there is a relatively larger error for every loading. The smaller the element, the more accurate the measurement is. For Static Loads #1 and #3, the maximum deflection is located at the center of the beam, so there is no error for any beam model.

5.00% 4.50% 4.00% 3.50% 3.00% Error 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% 1' 6" Element Size 1"
Static Load 1 Static Load 2 Static Load 3 Static Load 4 Static Load 5

Figure 3-7: Percentage Error of Maximum Deflection.

30

5.00% 4.50% 4.00% 3.50% 3.00% Error 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% 1' 6" Element Size 1"
Static Load 1 Static Load 2 Static Load 3 Static Load 4 Static Load 5

Figure 3-8: Percentage of Error of Max Deflection Location.

From these tests, it can be concluded that ANSYS provides good values for displacement, with less than 5% error. If available, the smaller element results in close to exact answers, but larger element sizes can give reasonable answers. The following section details the process of evaluating the dynamic analysis capabilities of ANSYS.

3.4

Dynamic Load Simulations


The ANSYS program can be also be used to model structures subjected to

dynamic loadings. However, a sub-program called LS-DYNA exists in ANSYS that is more suited towards explicit solutions and dynamic loads that are applied over short durations. LS-DYNA computes solutions faster than ANSYS and can supply more information (i.e. data points). On the other hand, ANSYS has more graphical support than LS-DYNA.

31

The LS-DYNA program is used to verify the maximum deflection of a simply supported beam under different pulse loadings. An equivalent single degree of freedom (SDOF) model was used to determine both the maximum deflection of the beam under a uniformly loaded dynamic load and the shape of the time-history curve of this deflection. This maximum value and the shape of the graph as a whole were compared to the LSDYNA results. The same beam dimensions from the static tests were used. Also, the magnitude and duration of the dynamic load were constant for each pulse loading.

3.4.1 Preprocessing
The beam element used in this analysis, BEAM 161, consists of a node at either end of the element (I and J) and an orientation node (node K) on a different line than that of the beam direction. This element is used for explicit dynamic analyses only, and can be used to model isotropic, linear elastic materials. It can model beams of different crosssection shapes. Each node has six physical degrees of freedom translation and rotation in three directions and also three nodal velocity degrees of freedom and three nodal acceleration degrees of freedom. These latter are not technically degrees of freedom, but are computed and stored as such. The BEAM 161 element has two main methods of calculation. A Hughes-Liu beam element assumes a constant moment along its length and detects stresses in the center. A Belytschko-Schwer element generates a linearly changing moment along the element and measures stresses at either end. The cross-sectional dimensions are defined as real constants. Next, the material properties of the beam are defined. The modulus of elasticity and Poissons ratio can be inputted.

32

Figure 3-9: BEAM 161 ANSYS LS-DYNA element.

For this model, the material was assumed to be steel, with a Youngs modulus of 30 x 106 psi and a Poissons ratio of 0.3. Also, because this is an explicit solution, the weight of the structure is considered a necessary value; therefore, another material property is the mass density. The beam can be modeled very simply with a line drawn between two keypoints. A third keypoint that is not collinear with either endpoint is also necessary for element orientation. The line that makes up the beam can be divided into a certain number of elements or a particular element size. This is necessary from a practical point to have a node in the center of the beam to measure the maximum deflections. So, at the very least the beam model has to be divided into 2 elements. After construction of the line keypoints and line, the beam can be meshed into the desired number of elements with the set material and geometrical properties.

33

3.4.2 Loading and Solution


Three different pulse loadings were used to test the dynamic load capabilities of LS-DYNA. A rectangular pulse loading with instantaneous acceleration was tested because it is a very simple, uniform load. The two other loadings tested are triangular pulse loads. The first triangular pulse load is an instantaneous loading with a linear decay to zero. The second pulse loading has an equal time to rise from zero and time to decay to zero. The first triangular pulse loading allows for comparison to results with the rectangular loading and is similar to the pulse loading from a blast load. The second triangular load is also used for comparison and is similar to the equivalent loading of a roof structure subjected to a blast load. All three dynamic loads have equal magnitudes and durations. These pulse loads were inputted into the model as array parameters. The loads were all linear, and very few points need to be inputted. The program interpolates the points between those given. Two arrays are defined for each load. One array gives the time values and the other the load values. The elements that make up the beam can be defined as components where the loads will be acting. The pulse loads are uniformly distributed across the length of the beam, and are set to act in the center of the beams cross-section. The test beam is designed as a simply supported, and so translational constraints are put on the endpoints. The duration of the analysis can then be specified and the program can be directed to output ANSYS solutions or LS-DYNA solutions. ANSYS solutions have a limit of 100 points and are usually used to examine the deflection of an entire structure at a specific time.

34

The LS-DYNA solutions have a limit of 1000 points, and because of their large size are usually limited to observe the time history of important points or nodes on the beam. The purposes of these tests do not necessitate a very large number of output points. The length of time under investigation should be the length of at least half a response period.

3.4.3 Post-Processing
ANSYS has the capabilities to display the deflection of the entire beam model at a particular time, or graph the displacement time history of a particular node on the beam. For these tests the time history of the center node of the beam is graphed.

3.4.4 Results
For each dynamic load the model divides the beam into sets of 2 elements, 6 elements, 10 elements, 20 elements, 40 elements, and 240 elements; or element lengths of 10, 3 4, 1, 6, and 1 respectively. The LS-DYNA program calculates the time steps used in the numerical integration based on the dimensions and properties of the beam and the element length (ANSYS, 2004).

t = 0.9

L c

(3.1)

where L = length of element c = wave propagation velocity

35

c=

(3.2)

where E = modulus of elasticity = mass density

The wave propagation velocity (Eqn. 3.2) is calculated from the Youngs Modulus and density of the beam, and the smallest time step (Eqn. 3.1) used in the program is the ratio of the smallest element length in the model and the propagation velocity. For all models the element length is uniform. Six models were used with different lengths of beam element. The following table shows the number of beam elements, respective lengths, and used and corresponding time steps.

Table 3-1: Minimum Time Steps.

Number Element Smallest Length Time Step of (sec) (ft) Elements 2 120 0.0005342 6 40 0.0001781 10 24 0.0001068 20 12 5.342E-05 40 6 2.671E-05 240 1 4.451E-06

Table 3-2: Aspect Ratio. Number Length: of Depth Elements Ratio 2 30:1 6 10:1 10 6:1 20 3:1 40 3:2 240 1:4

As can be seen, the more elements in a beam (i.e. the shorter the element length), the smaller the time steps are. It would be assumed that the smaller the time steps used in numerical integration, the more accurate the solution would be. It must be noted that too small of a time step could lead to a numerical instability in which the solution is not bounded and the resulting deflection approaches infinity. This was not encountered in this investigation.

36

The ANSYS test data was compared to a time history calculated using a single degree of freedom (SDOF) model and load and mass factors to correct for the shape of the load and the shape of the beam (refer to Sec. 2.4.1 for details).
1.2 1 0.8 Deflection (in) 0.6 0.4 0.2 0 0 -0.2 Time (msec) 10 20 30 40
2 elem (LS-DYNA) 6 elem (LS-DYNA) 20 elem (LS-DYNA) 240 elem (LS-DYNA) Duhammad's Integral

Load (plf)

400 200 0 0 2 4 Time (msec) 6

50

60

70

80

90

Figure 3-10: Dynamic Pulse Load #1.

Dynamic Load #1 is a rectangular pulse load with an instantaneous rise and decay. Several deflection time history curves are shown in Figure 3-10. The beam model with two elements and only one node between the endpoints drastically underestimates the response. Using six elements results in a displacement within 20% of the theoretical response. Coincidentally, this is the result without including the load and mass factors. Subsequent decreases in element length arrive to a more accurate response. The period response also becomes more and more exact. The beginning curvature of the graphs demonstrates zero initial displacement and zero initial velocity. However, the more elements modeled, the more warped the graphs become.

37

Dynamic Load #2 is a triangular pulse load with an instantaneous rise and a linear decay to zero. The time history curves resemble those for Dynamic Load #1. This displays the precision of the ANSYS program. Again, the graphs show a zero initial deflection and a zero initial velocity (Fig. 3-11). The accuracy of the answer increases with the decrease in element length. The maximum displacement is less than that of the first dynamic loading, which is the expected result. The response period is the same, which is consistent with the period of Dynamic Load #1 and with the natural period of the beam model. Unfortunately, the curves also become more and more wavy with the addition of elements. In addition to the shape, the waviness affects the time of maximum deflection. This causes a difference between the calculated and expected time of maximum response.

0.6 0.5 0.4 Deflection (in) 0.3


Load (plf)

2 elem (LS-DYNA) 6 elem (LS-DYNA) 20 elem (LS-DYNA) 240 elem (LS-DYNA) Duhammel's Integral

0.2 0.1 0 0 10 20

400 300 200 100 0 0 2 4 Time (msec) 6

30

40

50

60

70

80

90

-0.1 Time (msec)

Figure 3-11: Dynamic Load #2.

38

Dynamic Load #3 is also a triangular pulse load with an equal time to rise and time to decay. The time history curves are very similar to that of Dynamic Load #2 in terms of shape and magnitude. The similar shapes are due to the accuracy of the analysis. The close responses result from the equal impulse of the pulse loads. Due to the beam models properties, it is an impulsive sensitive system. The resistance of the beam is less related to the peak load than to the pulses impulse. The shape of the curves relates to the aspect ratio of the beam elements. The depth of the beam, and thus the depth of each element, measures 4 inches. The aspect ratio of the largest beam element (as shown in Table 3-2) is 30:1, and the aspect ratio of the smallest beam element is 1:4. The elements with aspect ratios less than 5:1 are specified as deep beams, and have different kinematic and equilibrium conditions than those of regular beams. The ANSYS code chosen was designed only for regular beams.
0.6 0.5 0.4 Deflection (in) 0.3 0.2 0.1 0 0 -0.1 10 20 30 40 50 60 70 80 90
400 Load (plf) 200 0 0 2 4 Time (msec) 6

2 elem (LS-DYNA) 6 elem (LS-DYNA) 20 elem (LS-DYNA) 240 elem (LS-DYNA) Duhammel's Integral

Time (msec)

Figure 3-12: Dynamic Load #3.

39

To further explore the ANSYS results, the maximum deflection, time of maximum response, and period are compared to the expected values. They are compared according to the number of finite elements and the type of loading. Figure 3-13 shows a clear decrease in maximum deflection error with an increase in element number. This trend is common for every pulse load. Figure 3-14 presents the error in calculating the time of maximum deflection. The beam model results in a value within 3% of the SDOF models prediction. Unfortunately, the accumulation of elements past that level also increases the amount of error. The error in the magnitude of maximum displacement for the 20-element beam is less than 2% for all loads.
60% 50%
Dynamic Load #1

40% Error 30% 20% 10% 0% 2 6

Dynamic Load #2 Dynamic Load #3

10 20 Number of Elements

40

240

Figure 3-13: Percentage Error in Maximum Response.

40

35% 30% 25% 20% 15% 10% 5% 0% 2 6 10 20 40 240


Number of Elements

Dynamic Load #1 Dynamic Load #2 Dynamic Load #3

Error

Figure 3-14: Percentage Error in Time of Maximum Response.

Finally, the period of response is examined for the test beams. Figure 3-15 shows a marked decrease in error with the increase in element number. The error for the 20element beam is about 3%, while the error for the 240-element beam is close to zero.

41

20%

16%

Dynamic Load #1 Dynamic Load #2 Dynamic Load #3

12% Error 8% 4% 0% 2 6

10 20 Number of Elements

40

240

Figure 3-15: Percentage Error in Response Period.

In conclusion, the ANSYS/LS-DYNA models result in answers close to the expected values. None are the answers are perfect, and additional programming with the LS-DYNA and other finite element analysis software needs to be performed to determine if these are currently the most accurate outcomes. For the current purposes, the error is small enough to be applicable. In the next section, the response of beams under blast loads is explored. For these analyses, an aspect ratio between 3:1 and 3:2 appears to result in the best solutions in terms of maximum deflection, time of maximum deflection, and response period. An element length of 1 is used for the next section.

42

3.5

Analysis of Blast Load Response Using LS-DYNA


The same steps as before are used to generate a model of a simply-supported

beam. Twenty beam elements make up the beam, corresponding to an element length of 1 each. The responses of the blast loads are compared to the response of the equivalent dynamic blast load using ANSYS. The response of the equivalent loading is also derived using Duhammads integral.

3.5.1 Application of Blast Loads


The pressures on a structure from a blast are non-uniform and highly impulsive. The load pulses on a roof contain a positive downward phase and a negative section phase. The peak pressures decay and the load duration increases as the blast wave traverses the roof length. Using ANSYS, different sections of the beam require different loading characteristics at different times. The loading response is analyzed in stages. First, the positive pressure phase is idealized using a triangular pulse with an instantaneous rise time and a time to decay. For each subsequent section of the beam, the peak load decreases and the time history of the loading changes to represent the accentuation of the blast pulse. Next, the same process is repeated using both positive and negative phase data. Finally data from an actual blast test is used in analysis. To reflect the variable loads, the beam model is divided into five differing loading sections. The loads on each section have a different initial load time, peak load, and final time. The loads are inputted as time and load array parameters similar to the pulse loadings in Sec. 3.3. Instead of one set of time and load arrays, there are five sets of time and load arrays. The test beam is divided into five separate components, where each set of time and load data is applied to each component. 43

30

35

40

45

50

Figure 3-16: Locations of CONWEP measurements.

Figure 3-17: Blast Pressure Distribution.

The application CONWEP (Hyde, 1992) generates time of arrival, incident peak pressure, positive phase duration, and incident impulse output used in the arrays. In CONWEP, a hemispherical surface explosive of 1000 lbs of TNT is analyzed at a range of 30 feet from the simply-supported beam. Pressure information is recorded at ranges of 30 feet (the front end of the beam), 35 feet, 40 feet, 45 feet, and 50 feet (corresponding to the back end of the beam). These ranges are shown in Figure 3-16.

Table 3-3: Peak Pressure vs. Range.


Range of Data Origin (ft) 30.0 35.0 40.0 45.0 50.0

Peak Pressure (psi) 134.2 94.92 69.83 53.23 41.75

Range of Data Distribution 30-32' 32-37' 37-43' 43-48' 48-50'

44

These five loading patterns are distributed along the beam to replicate actual blast characteristics. The pressure from the 30 feet range is distributed (as shown in Fig. 3-17) over the first two feet of the beam, from 30 to 32. The pressure from the 35 feet range is allocated from 32 to 37. The 40, 45, and 50 feet range data is applied over the next 6, 5, and 2 feet, respectively. Table 3-3 displays the peak pressure and the range to which it relates. The beam model supports line loads instead of surface loads. The pressures are multiplied by the width of the cross-section to calculate the associated distributed load.

3.5.2 Results (Positive Phase Only)


The actual blast loads can be seen in Fig. 3-18. In the actual blast pressures, the peak decays in a nonlinear manner to zero. The first trial positive phase pressures decay linearly over the same length of time (see Fig 3-19). An important check in this analysis compares the impulses from the actual data to the approximate data. It can be shown that the first trial impulses are from 2 to 7 times that of the actual pulses (see Fig. 3-20). To correct this, the pulses were modified to a bilinear decay. From the simple dynamic models performed earlier, it can be ascertained that the beam model is impulse sensitive. Therefore, the actual shape should not matter, and the arrangement of the bilinear decay should be irrelevant. This decay model is performed with two separate examples (Figs. 3-21 and 3-22). One shape follows the pattern of decreasing primary and secondary peaks, along with increasing durations. The third scenario keeps the bilinear decay form without following a discernable configuration.

45

140 120 Incident Pressure (psi) 100 80 60 40 20 0


0 5 10 15 Time (msec) 20 25 30
Beam Segment 1 Beam Segment 2 Beam Segment 3 Beam Segment 4 Beam Segment 5

Figure 3-18: Actual Blast Loading.

2500

2000

Load (lbs/in)

1500

Beam Segment 1 Beam Segment 2 Beam Segment 3 Beam Segment 4 Beam Segment 5

1000

500

0 0 0.005 0.01 0.015 Time (sec) 0.02 0.025 0.03

Figure 3-19: Loading Scenario #1.

46

1600 1400 Incident Impulse (psi-msec) 1200 1000 800 600 400 200 0 30.0 35.0 40.0 Distance (ft) 45.0 50.0
Real Data Trial #1 Data

Figure 3-20: Impulse Comparison.

2500

2000

Load (lb/in)

1500

Beam Segment 1 Beam Segment 2 Beam Segment 3 Beam Segment 4 Beam Segment 5

1000

500

0 0 0.005 0.01 0.015 Time (sec) 0.02 0.025 0.0

Figure 3-21: Loading Scenario #2.

47

2500

2000

Beam Segment 1 Beam Segment 2 Beam Segment 3

Load (lbs/in)

1500

Beam Segment 4 Beam Segment 5

1000

500

0 0 0.005 0.01 0.015 Time (sec) 0.02 0.025 0.03

Figure 3-22: Loading Scenario #3.

To further continue validation of the input data, the beam is divided into four equal-sized components instead of five different sized components. Four different pulses load these sections. For Scenario #4, loads from 30, 35, 40, and 45 feet range act on the four segments. This is repeated in Scenario #5 using loads from the 35, 40, 45, and 50 feet range. The dynamic responses from these five loading combinations are plotted together (Fig. 3-23) and the impulses from the three main load sets are compared to the actual positive impulses. The deflection-history for Scenario #1 greatly exceeds the deflection of the other tests. Scenarios #2-5 agree with other more closely. Scenarios #2 and 3 barely differ discernibly. Scenario #4 and Scenario #5 represent the upper and lower bounds, respectively (Figs. 3-24a and b). This occurs because the pulses closer to the blast source are larger than the pulses farther away. From these results, the best loading data to use is Scenario #2. Also, it can be seen that the impulse for Scenario #2 matches the actual impulses.

48

240 230 Incident Impulse (psi-msec) 220 210 200 190 180 170 160 150 140
Exact Data Trial #2 Trial #3

30.0

35.0

40.0 Distance (ft)

45.0

50.0

Figure 3-23: Impulse Comparison.


8 6 4 Deflection (in) 2 0 0 -2 -4 -6 -8 Time (msec) 50 100 150 200 250

Scenario #1 Scenario #2 Scenario #3 Scenario #4 Scenario #5

Figure 3-24a: Response Comparison of Loading Scenarios using LS-DYNA.

49

2.5 2 1.5 1 Deflection (in) 0.5 0 -0.5 -1 -1.5 -2 -2.5 Time (msec) 0 50 100 150 200 250
Scenario #2 Scenario #3 Scenario #4 Scenario #5

Figure 3-24b: Response Comparison of Loading Scenarios using LS-DYNA (without Scen. #1).

Figure 3-25 displays the deflection-history of the beam model at various points. As expected, the midpoint deflection measures the highest of the beam. The maximum deflection happens after all the load pulses have come and gone. This outcome is due to the high natural period of the beam relative to the short duration of the blast. The beam takes longer than the time of the load to fully react. The rightmost quarter-point encounters a maximum before the left quarter-point, although the left has a higher deflection than the right.

50

2 1.5
Left Quarter-point

1 0.5 0 0 -0.5 -1 -1.5 -2 50 100

Midpoint Right Quarter-point

Deflection (in)

150

200

250

Time (msec)

Figure 3-25: Dynamic Response for Loading Scenario #2.

3.5.3 Results (Positive and Negative Phase)


Similarly, the estimation of the negative phase begins with the shape of a triangle. Initially, the negative phase of the blast loading starts at zero. There exist a decay time (rise in the absolute sense) to the maximum negative load and a rise time (decay in the absolute sense) back to zero (Fig. 3-27). Comparison of the actual impulses and the approximate triangle impulse (Fig. 3-29) concludes that the approximate negative impulse is less than the actual by about 11%. To keep the design and input simple, a contour using three lines in implemented. This phase is designed to have the same impulse as the actual load.

51

140 120 Incident Pressure (psi) 100 80 60 40 20 0 -20 0 20 40 60 80 Time (msec) 100 120 140
Beam Segment 1 Beam Segment 2 Beam Segment 3 Beam Segment 4 Beam Segment 5

Figure 3-26: Actual Blast Load Positive and Negative Phase.

2500 2000 1500 1000 500 0 -500

Beam Segment 1 Beam Segment 2 Beam Segment 3 Beam Segment 4 Beam Segment 5

Load (lbs/in)

0.02

0.04

0.06 0.08 Time (sec)

0.1

0.12

0.14

Figure 3-27: Loading Scenario #1 Positive and Negative Phase.

52

2500

2000
Beam Segment 1 Beam Segment 2

1500 Load (lb/in)

Beam Segment 3 Beam Segment 4 Beam Segment 5

1000

500

-500 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Time (sec)

Figure 3-28: Loading Scenario #2 Positive and Negative Phase.

-350

-300 Negative Impulse (psi-msec)

Actual Trial 1

-250

Trial 2

-200

-150

-100

-50

0 30.0 35.0 40.0 Distance (ft) 45.0 50.0

Figure 3-29: Negative Impulse Comparison.

53

The dynamic response of Figure 3-30 is the result of the positive phase and the aforementioned negative phase from Scenario #2. From Figure 3-31, it can be seen that the initial maximum deflection is less than the deflection from only the positive phase. The suction pressure from the negative phase decreases this deflection and enhances the deflection in the rebound. The deflection bounces back to a higher value that the positive phase loading, as can be seen from the midpoint results. The same event happens at the left and right quarter-points. Again, the right point experiences higher loading before the left point (Figs. 3-28 and 3-30).

Left Quarter-point Midpoint Right Quarter-point

Deflection (in)

0 0 -1 50 100 150 200 250

-2

-3 Time (msec)

Figure 3-30: Response of Loading Scenario #2 Positive and Negative Phase.

54

3 2 1 0 -1 -2 -3 0 50 100 150 Time (msec) 200 250

Deflection (in)

Positive Phase Only Pos. & Negative Phase

Figure 3-31: Response Comparison of Positive Only and Pos. & Negative Data for Loading Scenario #2.

The next section covers the equivalent loading defined by TM-855-1. The method is discussed and the response is compared to both the positive phase response and the deflection from the positive and negative blast pressure.

3.5.4 Equivalent Blast Loading


The procedure to define the blast loading on a roof structure is used for the beam model. The range to the target structure is the same range used in the blast data, namely 30 feet. The range to the back of the structure equals the range to the target plus the length of the beam model. In this example, there is no structure height. Blast data from this range of 50 feet is used in the analysis. The CONWEP application can calculate the necessary information from known bomb size and distance. This data is shown in Table 3-4.

55

Table 3-4: CONWEP Blast Data.


Peak Pressure (psi) Positive Phase Duration (msec) Time of Arrival (msec) Incident Impulse (psi-msec) Shock Front Velocity (ft/s) Peak Dynamic Pressure (psi) 41.95 15.73 12.83 150 2068 30.41

The results of CONWEP are used to describe the pulse loading in Figure 3-32. The first noticeable characteristics of the equivalent blast loading and the load from blast data are the differences in time of arrival and loading duration. While the equivalent load begins at a relatively much later time than the actual arrival time of the blast pressure, the durations vary by about eight percent for the positive phase only of the blast loading. The equivalent loading contains only a positive phase, while blast loads have both positive and negative phase. The length of time the blast affects a structure greatly increases when including the negative phase. The positive and negative blast pressure has a much larger duration than the equivalent load pressure. The peak loading of the blast pressures are all greater than the peak load for the DAHS equivalent load. Additionally, the sum of the positive impulses from the blast loads equals three times that of the equivalent loading. The total impulse from the positive and negative phase of the blast loading differs 15% from the equivalent impulse. To keep these details in perspective, the design of the equivalent loading applies the pulse to the entire beam, while the actual blast loads have different effects on different portions of the beam model.

56

500 450 400 350 Loading (lb/in) 300 250 200 150 100 50 0 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 Time (sec)

Figure 3-32: TM-855 Equivalent Blast Load.

The response of this load on the beam model can be calculated using ANSYS. Also, due to the shape of the load and how it is applied to the entire beam, the deflection time history can be derived using Duhammads Integral. First, we will evaluate how the ANSYS program reproduces the deflection time history by comparing it to the hand numerical integration methods (Fig. 3-33). As expected, there is a very good correlation between the hand calculation and the ANSYS results. There is a 3% difference between the maximum deflections and also between the response periods.

57

4 3 2 1 0 0 -1
Duhammel's Integral

Deflection (in)

20

40

60

80

100

120

140

160

-2 -3 -4

ANSYS

Time (msec)

Figure 3-33: Verification of Equivalent Load Response.

Since it has been verified that ANSYS calculates the equivalent response correctly and again verifies that ANSYS accurately calculates dynamic responses, the deflection for the equivalent load can be compared to the approximate blast loading data. Figure 3-34 contains the deflection from the positive phase blast data, the positive and negative phase blast data, and the equivalent pulse load.

58

4 3 2 Deflection (in) 1 0 -1 -2 -3 -4 0 50 100 150 Time (msec) 200 250


Positive Phase Only Pos. & Negative Phase Equiv. Load.

Figure 3-34: Comparison of Dynamic Response.

Analyzing the curves, several points can be observed between the three simulations. There is a large variation between the first peak of the equivalent response and the positive phase only response, and the equivalent response first peak is almost twice that of the positive and negative response. While the subsequent peaks match for the positive and negative phase and the equivalent load, the first peak is the most valued by design engineers. The period response for the equivalent loading almost matches the positive phase response data, and the positive and negative phase data varies by only six percent. Finally, the time of the maximum response differs between curves. As noted earlier, the starting time of the equivalent loading is about ten milliseconds after the blast wave is calculated to affect the beam model. It is believed that the data including the positive and negative blast phase is more accurate than the positive phase only. The next step in the research is to apply the equivalent load procedure to field test data.

59

3.6

Field Test

A research project by Hoemann and Salim (2007) was performed to analyze the resistance and dynamic response of roof members consisting of fiber reinforced polymers (FRP). The results of these tests can be used to further evaluate the DAHS load and procedure. The test setup is shown in Figures 3-35 and 3-37. The structure consists of six feet high concrete support blocks with earth berms along the sides to prevent blast pressure from forming underneath the roof. The charge weight is 1,000 lbs of ammonium nitrate and fuel oil (ANFO), and the standoff has a range of 30 feet to the face of the support block and 33 feet to the roof component.

33-ft Standoff to panel face

1,000-lbs ANFO

30-ft Standoff to block face

Figure 3-35: FRP Panel Test. The roof components are structural panels made up of honeycomb fiber reinforced laminations. The cross-section of an FRP panel contains core layers consisting of an FRP material formed into a sinusoidal wave bonded to a flat piece of similar makeup.

60

Two FRP components were used in the explosive field test. Both roof panels were 4 feet wide with a simply supported span of twenty feet. One FRP sample has a crosssection depth of 12 inches. It is denoted as R1. The second sample, R2, is 15 inches deep. The experimental static resistance functions of these structural members are shown in Figure 3-36. The sample roof was designed against blast loads considering only the elastic range of the FRP materials resistance.

Static Tree Testing 12 and 15 Deep Panel Sections


20

16

Equivalent Section of Panel R2

Pressure (psi)

12

Equivalent Section of Panel R1

0 0 2 4 6 8 10 12

- Displacement (in.)

Figure 3-36: Static Resistance Function for FRP Panels.

61

The test setup includes pressure gages on the roof and on the surface at equal radial distances. There are also deflection gages at the midpoint and two quarter-points for both FRP panels. This schematic is shown in Figure 3-37. It should also be noted that the test explosive is placed so that the resulting blast wave travel along roof direction. For this research project, the measured pressures are inputted into an ANSYS LSDYNA model. The response is measured against field deflections. Additionally, the dynamic response from the DAHS equivalent load is compared to the experimental values. Comparing the blast pressures measured on the roof and on the surface, it is noted that the loads on the ground surface are larger than the loads on the roof. The surface pressures do match the pressures and loading characteristics computed from CONWEP, which proves that this program correctly calculates pressures at the ground level but not at higher elevations. Most likely the blast wave loses more energy from striking the roof, or that blast pressures are not equal from radial distances at varying incident angles.

62

Panel R1
FF8 FF7

Panel R2
FF6 D5 D 5-ft FF5 FF4 D3 D4 5-ft FF3 D1 10-ft FF1 5-ft D2 5-ft FF2

10-ft

35-ft 30-ft

Displacement Gauges Free Field Pressure Gauges

Figure 3-37: Roof Panel Schematic (Hoemann and Salim, 2007).

63

Figures 3-38 through 3-43 show blast pressure-time histories from the ground, rooftop, and calculated pressures from CONWEP and SBEDS. The CONWEP data was used to determine the time of arrival of the shock front and SBEDS was used to generate simulated pressure-time history. Impulses are also calculated from the pressure-time histories.

105 90 75
FF1 - pressure FF2 - pressure CONWEP/SBEDS Data

Pressure (psi)

60 45 30 15 0 -15 -30 0 8

16 24 32 40 48 56 64 72 80 Time (msec)

Figure 3-38: Pressure-time history at 35 ft.

For Figures 3-38 through 3-43, the CONWEP/SBEDS simulation has the same peak pressure measured at the ground level. It is also of interest that the blast pressures measured on the ground and at the rooftop have the same positive phase durations. Also, the measured negative phases appear to have the same durations, although they are not well defined.

64

200 160

Impulse (psi-msec)

120 80 40 0 -40 -80 -120 0 8 16 24 32 40 48 56 64 72 80 Time (msec)


FF1 FF2 CONWEP/SBEDS Data

Figure 3-39: Impulse-time history at 35 ft.

Comparable to the pressure history, the impulse-time history shows lower values on the roof than on the ground. The actual impulses for the simulation are greater than those measured on the rooftop or on the ground. The peak incident pressures for the 35 ft range computed by CONWEP is almost twice the impulse measured at the roof (Fig. 337). This same trend occurs in the impulse diagram (Fig. 3-38). The cause of the shock wave reduction when it reaches the roof is expected to affect the pressures less as the blast wave transverses the roof. The CONWEP/SBEDS simulation should closer predict the measured pressure levels. While this does occur, there is still appreciable error between the measured and calculated blast pressures, as shown in Figures 3-40 and 3-42.

65

60 50

Pressure (psi)

40 30 20 10 0 -10 0 8

FF4 FF5 CONWEP/SBEDS Data

16 24 32 40 48 56 64 72 80 Time (msec)
Figure 3-40: Pressure-time history at 45 ft.

180 150

Impulse (psi-msec)

120 90 60 30 0 -30 -60 -90 0 8 16 24 32 40 48 56 64 72 80 Time (msec)


Figure 3-41: Impulse-time history at 45 ft. 66
FF4 FF5 CONWEP/SBEDS Data

40 32
FF7 FF8 CONWEP/SBEDS Data

Pressure (psi)

24 16 8 0 -8 -16 0 8

16 24 32 40 48 56 64 72 80 Time (msec)
Figure 3-42: Pressure-time history at 55 ft.

180 150

Impulse (psi-msec)

120 90 60 30 0 -30 -60 -90 0 8


FF7 FF8 CONWEP/SBEDS Data

16 24 32 40 48 56 64 72 80 Time (msec)
Figure 3-43: Impulse-time history at 55 ft. 67

Figures 3-44 and 3-45 display the response of the FRP roof panels from the field data and the simulated deflections from the measured roof pressures. The deflections at the quarter-points predicted by ANSYS LS-DYNA match closely to those measured in the field. In the model, damping is not considered, whereas the effect occurs in the experiment. These models use the measured roof blast pressures.

Normalized Deflection (delta/EI)

2.4 1.8 1.2 0.6 0 -0.6 -1.2 -1.8 -2.4 0 50 100 150 Time (msec) 200 250
Field Test Response ANSYS LS-DYNA Response

Figure 3-44: Deflection Response of R1 panel at near quarter-point.

68

Normalized Deflection (delta/EI)

2.4 1.8 1.2 0.6 0 -0.6 -1.2 -1.8 -2.4 0 50 100 150 Time (msec) 200 250
Field Test Response ANSYS LS-DYNA Response

Figure 3-45: Deflection Response of R1 panel at far quarter-point.

From Figure 3-46, it can be seen that the ANSYS LS-DYNA model continues to closely match the field response, not including damping effects. The DAHS procedure is calculated using CONWEP data. Due to the differences between measured pressures, a corrected DAHS procedure is also used. The only factor changed in the corrected procedure is the peak pressure. The equivalent load response, however, measures almost three times the initial peak deflection and twice that of the rebound deflection. In design, it has been customary to focus on the first maximum deflection point. The rebound effects are similar to the model in Sec. 3.5, where the subsequent peaks are larger than the first. The corrected DAHS response predicts values closer to measured deflections, but the first peak deflection is still twice of that measured. In that regard, the DAHS procedure is still very conservative.

69

Normalized Deflection (delta/EI)

6 4.5 3 1.5 0 -1.5 -3 -4.5 -6 0


Field Test Response LS-DYNA Response DAHS Response Corrected DAHS

50

100 150 Time (msec)

200

250

Figure 3-46: Deflection Response of R1 panel at midpoint.

Panel R2 displays similar results. The ANSYS LS-DYNA simulations approximately measure the primary peak deflections, but appear to significantly underestimate the rebound peaks (Fig. 3-49). The deflections during rebound are larger because of a tie-down anchor failure during the negative pressure phase duration. The DAHS procedure predicts higher deflections than is measured in the first peak. It does come close to the peak responses due to rebound, but this may be inconclusive due to the anchor failure. The corrected DAHS procedure does more closely approach the measured deflections for this test, but again, that may be inconclusive due to the anchor failure.

70

Normalized Deflection (delta/EI)

2.4 1.6 0.8 0 -0.8 -1.6 -2.4 -3.2 -4 0 50 100 150 Time (msec) 200 250
Field Test Response ANSYS LS-DYNA Response

Figure 3-47: Deflection Response of R2 Panel at near quarter-point.

Normalized Deflection (delta/EI)

3 1.5 0 -1.5 -3 -4.5 -6 0 50 100 150 Time (msec) 200 250


Field Test Response ANSYS LS-DYNA Response

Figure 3-48: Deflection Response of R2 Panel at far quarter-point.

71

Normalized Deflection (delta/EI)

3.2 2.4 1.6 0.8 0 -0.8 -1.6 -2.4 -3.2 0 50 100 150 Time (msec) 200 250
Field Test Response LS-DYNA Response DAHS Response Corrected DAHS

Figure 3-49: Deflection Response of R2 Panel at midpoint.

3.7

Summary
It appears from the ANSYS LS-DYNA curves that the equivalent loading

procedure does not predict the maximum deflection well. The curves define linear elastic materials and do not include any dampening effects. The first deflection curve from the exact blast loading is more than 50% less than the deflection response from the equivalent loading. The DAHS procedure more closely approaches the peak deflections from rebound, but is still conservative. The field test data highlights an interesting phenomenon in blast pressures on the tops of structures. CONWEP accurately predicts the ground surface pressures, but encounters greater error on roofs. More research is needed to verify this and change the program to account for the load reduction.

72

Chapter

Static Resistance Function of Open Web Steel Joists

4.1

Introduction
Current resistance functions for open web steel joists include a linear elastic curve

close to the elastic limit and a fully plastic curve after yield. It is believed that the failure mechanisms in a steel joist contradict this behavior, and static tests are performed to qualify this theory. The steel trusses in this research are tested to failure to evaluate the maximum load, deflection, and amount of energy able to be absorbed of a typical joist system.

73

4.2

Analytical Resistance Function


Three types of open web steel joists are considered in this research. Their sizes

are chosen to represent three differing aspect ratios. The truss types are 16K5, 26K6, and 32LH06 (SJI, 2005). Before actual experimentation of the joists, other resistance methods can be employed. Figures 4-1 through 4-3 display the response curves for all three sets of trusses according to SBEDS (2004) calculation and an example of current engineering practice. The ultimate resistances for the trusses almost coincide for the SBEDS calculation and the engineering practice. The engineering resistance ranges from 94 to 99% of the SBEDS resistance. However, the stiffness and elastic deflection limit did not match as well. The stiffness ratio of engineering calculation to SBEDS decreases as the aspect ratio decreases. For the smallest truss, the ratio is about 75%; for the largest joist, it is more than 27%. Since the ultimate resistances are almost equal with respect to the procedure and the stiffness ratio changes, the elastic deflection limit also has to change with a similar trend. Stiffness and resistance are related according to Eqn. 4.1 (Biggs, 1967).

ye =

Ru ke

(4.1)

where ye = elastic deflection limit Ru = ultimate resistance ke = effective stiffness

74

900 800 700 600 Load (lb/ft) 500 400 300 200 100 0 0 2 4 6 8 10 Deflection (in) 12 14 16
Engineering Calc SBEDS Max Allowable Load, Defl. Ultimate Load, Defl.

Figure 4-1: Resistance Function for 16K5 Joist.

1200 1000
Engineering Calc

800 Load (lb/ft) 600 400 200 0 0 2 4 6

SBEDS Max Allowable Load, Defl. Ultimate Load, Defl.

8 10 Deflection (in)

12

14

16

Figure 4-2: Resistance Function for 26K5 Joist.

75

1600 1400 1200 Load (lb/ft) 1000 800 600 400 200 0 0 2 4 6 8 10 Deflection (in) 12 14 16
Engineering Calc SBEDS Max. Allowable Load, Defl. Ultimate Load, Defl.

Figure 4-3: Resistance Function for 32LH06 Joist.

The calculation for ultimate resistance in the engineering procedure uses wellknown values. The calculation for stiffness utilizes an approximate formula from the Steel Joist Institute (2004) to calculate joist moment of inertia. The SBEDS calculations use data from actual joist cross-sectional properties to calculate the stiffness. The maximum deflection in the engineering procedure is based on the ratio between the ultimate resistance, the loading, and the ratio of pulse duration to the joist natural period. This relationship is graphed for single-degree-of-freedom models with an elastic-plastic resistance shape subjected to triangular pulse loads in Fig. 4-4. This maximum deflection depends on the burst load. The SBEDS calculation can generate pressure-impulse diagrams that calculate the maximum deflections each truss can resist and the associated pressures and impulses.

76

The maximum load and ultimate load based on allowable stress design from the Steel Joist Institute are also noted in Figures 4-1 through 4-3. For the K-Series joists the maximum resistance is 83% of the engineering calculation and about 78% of the SBEDS calculations. For the LH-Series truss the maximum load is about 77% for both the engineering calculation and the SBEDS calculation. Tables 4-1 and 4-2 summarize the results of the engineering calculation and SBEDS, respectively.

Figure 4-4: Maximum deflection of elasto-plastic, one-degree-of-freedom system for triangular pulse load (Biggs, 1967).

Table 4-1: Resistance Data from Engineering Calculations.

16K5 26K5 32LH06

Stiffness (lb/ft/in) 231.34 499.057 622.7

Ru (lb/ft) Ru (SJI) 774.411 640 1093.03 903.333 1462.72 1128.23

Ye (in) 3.3475 2.1902 2.349

T (msec) 53.07 39.25 38.93

77

Table 4-2: Resistance Data from SBEDS Calculations.

16K5 26K5 32LH06

Stiffness (lb/ft/in) 308.3155217 838.5155286 2302.532127

Ru (lb/ft) Ye (in) T (msec) 812.689 2.64924 44.03 1168.28 1.40354 30.52 1478.05 0.64483 22.01

Additionally, both procedures allow for the calculation of joist natural periods. The natural periods from the engineering calculations are larger than the SBEDS calculations. This is probably due the smaller stiffnesses derived from the engineering method. The natural periods for the engineering calculations are 20-30% more than the SBEDS calculations for the K-Series trusses. For the LH-Series, the natural period is about 77% more than the SBEDS calculations.

4.3

Experimental Verification
This section will discuss the testing of the joists systems to establish the resistance

function. The joists were loaded until ultimate failure. Attention was kept to the Steel Joist Institute specifications for placing and loading OWSJ. The results from the testing are compared to the analytical resistance functions.

4.3.1 Testing Samples


Three pairs of joists are examined in this research. Two joist sets are from the KSeries type joists and one is a LH-Series truss. All trusses are 24 feet long.

78

The 16K5 steel joist has an effective depth of 15.14 in from the top chord centroid to the bottom chord centroid. The compression (top) chord consists of two equal leg angles 1.5 inches wide with a thickness of 0.155 in. The tension (bottom) chord is also made up of two equal leg angles 1.5 inches wide and 0.130 in thick. The end tension members are made up of solid round bars in diameter. The secondary web members are also solid round bars 11/16 in diameter and are designed to react in compression. The rest of the web is made up of tension and compression solid round bar members. The diameters for the round bars in the truss include 11/16 and 19/32. It should also be noted that the diameters of the web members decrease closer to the center of the joist length. The web members delineate 9 panel points in the top chord and 8 panel points in the bottom chord. Also, in construction of the joist a camber of 0.30 inches is required. The 26K5 truss has an effective depth of 25.02 inches. The compression chord contains two equal leg angles 2 wide with thickness of 0.166 inches. The tension chord is made up of two equal leg angles 1.5 wide with thickness of 0.109 inches. The end tension diagonal is a round steel bar 15/16 in diameter. The secondary web members consist of U-shaped members in diagonal near the ends and vertical in the web. They are designed to be in compression These U-shaped members are 1 wide and 1 tall with a thickness of 0.090 inches. The web consists of crimped-end U-shaped members near the outside of the web and regular U-shaped members in the web set. The regular web members have the same dimensions as the secondary web members. The crimped-end members are slightly thicker than the other web members. The web members connect to make four panel points in the top chord and three panel points in the bottom chord. The truss had a camber of 0.30 inches.

79

The 32LH06 steel truss has an effective length of 31.09 inches. The compression chord is made up of two equal leg angles of 1.75 wide with thickness of 0.143 inches. The bottom (tension) chord is made up of two equal leg angles of 1.5 wide with 0.109 inches thickness. The end tension diagonal is a solid round bar with a 1 diameter. The rest of the web members consist of U-shaped members. The secondary (compression) members have 1 width and 1 height and thickness of 0.090 inches. The tension web members have the same dimensions. The compression web members have width of 13/8 and height of 1-3/8 and thickness of 0.118 inches. There are no crimped members. The necessary camber is 0.30 inches. This joist contains three panel points in the top chord and 2 panel points in the bottom chord. Note the top chord has a larger cross-sectional area compared to the bottom chord for all truss types. For the 16K5 steel truss the cross-sectional areas became smaller closer to the center of the joist length. The slenderness ratio of the web members grows larger closer to the center of the joist length. This trend continues for the other two joists.

4.3.2 Test Set-Up


For our testing purposes, the open web steel joists (OWSJ) are designed for resting on steel supports. According to the Steel Joist Institute 42nd Edition Standard Specifications (2005), no less than 2 of the end of a K-Series OWSJ may lay on the steel support. The end of the OWSJ is welded to the steel supports using a minimum of two 1/8 fillet welds one inch long each. For LH and DLH-Series OWSJ, no less than 4 of the joist end may be placed on the support, and a minimum of 2 - fillet welds 2 inches long each must be used.

80

Figure 4-5: Bearing seat plates for 16K5 and 26K5 Joists.

The bearing seats of the open web steel joists are welded to a 3/8 thick plate that is bolted to the support system. The support system itself is bolted to a strong floor. Due to the dimensions of the support, the width of the plate is limited to 13 inches. The length of the plate is also 13 inches. Due to the plate thickness and LRFD standards, the width of the fillet welds for the bearing seats will be and 3 inches long. The bearing seats are laid out on the plate to allow room for the welds on each side and for a joist center to joist center distance of 6.5 inches. Six 1 diameter and two diameter regular-strength (A307) bolts are used to connect the plate to the strong floor support.

81

Figure 4-6: Bearing Seat Plates for 32LH06 Joist. Due to the joist type, length, and specifications approved by the Steel Joist Institute, horizontal bridging consists of continuous, horizontal steel members (equal leg angles) attached to both the top and bottom chord. Each attachment will be made by welding and is to have a slenderness ratio (between attachments) of less than 300. Each weld should resist at least 700 lbs of horizontal force. Diagonal bracing is not necessary for any truss.

Figure 4-7: Steel Joist Institute Specifications for horizontal bridging.

82

Both K-Series joists need a minimum number of two rows of horizontal bracing. The spacing has to be uniform throughout the joist. The equal angle horizontal bracing can be welded to the third points on the truss, i.e. at 8 from either end. The LH-Series joist only requires one row of horizontal bracing. The equal angle horizontal bracing for this pair of trusses can be welded at the middle of the trusses. The horizontal bridging consists of 1.5 equal leg angles. They are welded according to the figure with a 1/8 wide weld one half an inch long. Including the horizontal bridging between trusses, the ends of joist systems require connections at wall or any other structural component capable of resisting lateral loads. These connections can be fixed, hinged, or rollers allowing movement in the vertical direction. The latter connection type was chosen for several reasons: stability of the joists, smaller effect on resistance of truss system, and ease of construction. Five sets of lateral bracing were constructed from 12 inches long angles with channel members welded to either end. Figures show these bracings. The angles were welded to the strong floor and grease was applied to the sides of the channels to simulate a roller-type connection. The lateral bracings were placed in the center and spaced at 4 feet intervals along the length of the steel joist system.

Figure 4-8: Lateral bracing welded to strong floor. 83

Figure 4-9: Lateral bracing placement.

String potentiometers are placed in the center of the joist system length and at the quarter- points of the joist system length. These deflection gages are connected to the plywood boards that run the length of the joists. Figures show the connection of the strain pots.

Figures 4-10: String Potentiometer.

84

4.3.3 Testing Apparatus


Each test will consist of a pair of OWSJ. They will be connected at the supports, by the horizontal bridging, and by boards of plywood the entire length and width of the joist pair. The joist pair will be tested using an 18 loading tree (Fig. 4-11) with 16 loading points acting along the joists. The variation in lengths between the joist structure and the approximate distributed loading arises due to equipment restraints and available trusses. The loading tree consists of a pneumatic pump that pulls a beam made up of two channel sections back-to-back. Two smaller beams each made up of two channel sections back-to-back are connected to the larger beam. Two plates are connected to each of these smaller beams, and two smaller plates are connected to the previous plates. Each of the final plates in the loading tree has connections at the ends to apply load to a testing sample. The net effect is that the load from the pneumatic pump is distributed to two points, the loads from those two points are distributed to four points, and finally the loads from the four points are distributed to sixteen points. A general schematic of the loading tree is shown below. The loading tree has a maximum load of 110 kips. The equipment pulls up to simulate a downward distributed force, and so the test sample must be placed upside-down. Straps were used to connect the truss system to the loading tree due to their relatively high depths.

85

Figure 4-11: 16-Point Loading Tree.

4.3.4 Results
The following section presents the results of the steel joist testing. Each joist system was tested until it could not sustain any more load. This was determined by observing the failure mechanisms of the joist during the test and the decay of load resistance with an increase in deflection. The test samples were loaded using a deflectioncontrolled rate of 0.5 inches per second.

4.3.4.1 16K5 Joist System


Figure 4-12 show transverse and side views of the 16K5 truss system. As stated in Section 4.3.3, the loading tree does not cover the entire length of the joist. Also, it can be seen that the trusses are slightly warped, due to the camber in the truss and their own inherent flexibility out of the plane of loading.

86

Figure 4-12: 16K5 trusses prior to testing.

Initially the truss system displayed very little noticeable displacement. Then, the middle of the truss began to bow out-of-plane (Fig. 4-13). The lateral bracing at the center of the joist length fails due to this deflection (Figs. 4-14 and 4-15). The welds on the opposite side of the transverse bending fail. This bending is mainly due to failure in the compression flange. As the top chord (in figures the bottom is the top) continues to bend, the tension chord begins to twist. Eventually the horizontal bracing in the tension chord fails (Fig. 4-16). Finally, the welding connecting the compression chord to the bearing seats fails (Fig. 4-18). The truss system is now characterized as being unable to support any further load. Figures show this sequence of events.

87

Figure 4-13: 16K5 Joist Failure Sequence 1 of 5 Initial bending.

Figure 4-14: 16K5 Failure Sequence 2a of 5 Failure of lateral bracing.

88

Figure 4-15: 16K5 Joist Failure Sequence 2b of 5 Failure of lateral bracing.

Figure 4-16: 16K5 Joist Failure Sequence 3 of 5 horizontal bridging.

89

Figure 4-17: 16K5 Joist Failure Sequence 4 of 5 Continued out-of-plane bending.

Figure 4-18: 16K5 Joist Failure Sequence 5 of 5 Failure of bearing seat weld.

90

The load-deflection responses of the 16K5 truss system are shown in Figure 4-19. D1 measures the midpoint, while D2 and D3 are the quarter-points (Fig. 4-19). As can be seen, the system experiences linear displacement initially. The more noticeable bending of the compression chord out of the loading plane releases some of the energy absorbed. Following this, the system is able to receive more energy. The load resistance also increases in a fairly linear manner. The failure of the horizontal bracing releases more built-up energy, and finally the bearing seat weld failure finishes the usability of the truss system. Failures are denoted in the chart as areas where there is a large gap between data points and a downward slope.

35,000 30,000 25,000


Load (lbs)

20,000 15,000 10,000 5,000 0 0 5 10


Displacement (in)
D1 D2 D3

15

20

Figure 4-19: Static response for 16K5 truss system.

91

900 800 700 600 Load (lb/ft) 500 400 300 200 100 0 0 2 4 6 8 10 Deflection (in) 12 14 16
Experiment Engineering Calc SBEDS Max Allowable Load, Defl. Ultimate Load, Defl.

Figure 4-20: Midpoint Static Response for an individual 16K5 joist compared with existing methods.

The midpoint static response for an individual 16K5 truss was determined by dividing the system response by two times the effective joist length (Fig 4-20). It can be seen that the experiment measures a lower yield than existing analytical methods. The analytical methods assume failure mainly by bending. The original stiffness of the 16K5 truss corresponds well to the engineering calculation stiffness. Additionally, the maximum experimental load coincides with the ultimate load defined by SJI (2004) and steel joist manufacturers.

92

4.3.4.2 26K5 Joist System


In this test, the truss system was also slightly askew (Fig. 4-21). The failure mechanisms were also different. The first noticeable deflection was the bending of the ends of the tension flange in both trusses (Fig. 4-22). This phenomenon would be caused by high tension in the end tension member, buckling of a compression web member, or a combination of the two. The source is evident from the next failure mode. The end tension member of one joist fails, and the secondary web member and another web member from the other joist buckles (Figs. 4-23 and 4-24). From Figure 424, it can be seen that the weld for the tension member failed for one angle in the tension chord. A part of the other angle in the tension chord was ripped. This was the same portion that was bending at the start of the test. Up to this failure, the load-deflection behavior was fairly linear (Fig 4-30). When the end of the tension chord begins to bend, the response curves. The results of the end tension member break were a large release of energy and redistribution of forces. This is the most likely cause of the web compression member buckling. Since the end tension member in one truss failed, the next web member in that truss takes its place as a tension member. The truss system continues to absorb energy and the behavior has a linear shape (4-30). Figure 4-24 shows the weld from this new tension member fails. It should be noted that this member was originally designed as a compression member.

93

The end tension member of the other truss fails. From Figure 4-26, this failure seems to be caused by the twisting of the tension chord more than the bending of the member. The load-deflection behavior after this point continued to drop without the anticipation of rebound. It was noticed afterwards that the welds bonding the bearing seats to the connection plates failed (Fig. 4-29).

Figures 4-21: 26K5 Joist System prior to test.

Figure 4-22: 26K5 Joist Failure Sequence 1 of 5 Deformation of Tension Chord.

94

Figure 4-23: 26K5 Joist Sequence 2a of 5 Failure of End Tension Member.

Figure 4-24: 26K5 Joist Failure Sequence 2b of 5 Failure of End Tension Member.

95

Figure 4-25: 26K5 Joist Failure Sequence 3 of 5 Failure of Secondary Web Member.

Figure 4-26: 26K5 Joist Failure Sequence 4a of 5 Failure of End Tension Member.

96

Figure 4-27: 26K5 Joist Failure Sequence 4b of 5 Failure of End Tension Member.

Figure 4-28: 26K5 Joist Failure Sequence 5a of 5 Connection Plate Failure.

97

Figure 4-29: 26K5 Joist Failure Sequence 5b of 5 Connection Plate Failure.

45,000 40,000 35,000 30,000


Load (lbs)
D1 D2 D3

25,000 20,000 15,000 10,000 5,000 0 0 5 10


Displacement (in)

15

20

Figure 4-30: Static Response for 26K5 Joist System.

98

1200 1000 800 Load (lb/ft) 600 400 200 0 0 2 4 6 8 10 Deflection (in) 12 14 16
Experiment Engineering Calc SBEDS Max Allowable Load, Defl. Ultimate Load, Defl.

Figure 4-31: Midpoint Static Response for individual 26K5 Joist compared with existing methods. Again, it can be seen from Figure 4-31 that the current methods calculate more resistance than is experimentally observed. The initial slope of the elastic response curve of the joist, the stiffness, compares well with the engineering calculation value. Again the ultimate load of the test closely matches the ultimate resistance of the joist defined by the SJI and the manufacturer. Figure 4-31 demonstrates the multiple failure modes a steel joist has to resist before ultimate failure.

4.3.4.3 32LH06 Joist System


The test set-up for this truss system is displayed in Figure 4-33. The tension chord in one of the trusses is slightly warped and twisted (Figure 4-32). This was not expected to affect the strength of the joist system.

99

This failure system is similar to that of the 26K5 joists in that it is localized in an area other than the midpoint. Like the previous joist types, there was little noticeable deformation at the beginning of the test. The primary failure mechanism is the buckling of web compression members near the end of the truss (Figs. 4-34 through 4-36). The buckling of these members redistributed the forces and caused the secondary web (compression) members to buckle as well. This resulted in a loss of energy from the system. Again the initial load-deflection behavior changes linearly. After the multiple buckles, the response showed no sign of rebounding. The test was stopped.

Figure 4-32: Initial Deformation of 32LH06 Joist.

100

Figures 4-33: 32LH06 Joist System prior to test.

Figure 4-34: 32LH06 Joist Failure Sequence 1a of 1 Compression Web Member Buckling.

101

Figure 4-35: 32LH06 Failure Sequence 1b of 1 Buckling of Secondary Web Member.

Figure 4-36: 32LH06 Failure Sequence 1c of 1 Continued Buckling and Bending of Tension Chord.

102

30,000 25,000 20,000


Load (lbs)

D1 D2 D3

15,000 10,000 5,000 0 0 2 4 6 8 10


Displacement (in)

Figure 4-37: Static Response of 32LH06 Joist System.

1600 1400 1200 Load (lb/ft) 1000 800 600 400 200 0 0 2 4 6 8 10 Deflection (in) 12 14 16
Engineering Calc Experiment SBEDS Max. Allowable Load, Defl. Ultimate Load, Defl.

Figure 4-38: Midpoint Static Response of individual 32LH06 Joist comparing to existing methods.

103

The steel joist in this test did not reach the expected resistance value as shown in Figure 3-38. The joist came close to reaching the maximum allowable load (SJI, 2005). Since this truss is a longspan steel joist, the joist is designed to span longer lengths of about 30 ft or more (SJI, 2005). It is possible that the experimental resistance was much lower because truss span to depth ratio is too small. This might have resulted in a sheardominated behavior instead of bending. The full resistance of the joist and its members was not utilized.

4.4

Summary
From these static tests, it can be concluded that more research is necessary to

develop the static resistance functions of open web steel joists. The experiments for two out the three tests showed ultimate resistances close to the Steel Joist Institutes value of ultimate load. From the third test it is recommended for LH-Series Joists that longer spans be tested. Current blast design methodologies assume that the resistance function of open web steel joists behave similar to hot-rolled beams and reinforced concrete slabs. The current techniques used to define the resistance function of open web steel joists do not take into account the buckling of web members and the resultant strain softening from the loss of resistance and energy.

104

Chapter

Conclusions & Recommendations

5.1

Conclusion
The equivalent uniform blast procedure defined by DAHS was compared to

numerical blast data simulated using finite element analysis. The results of the equivalent blast did not adequately predict the response of the simulated blast load. The equivalent load procedure was also compared to experimental blast data. The finite element model closely predicted the response from the field test. However, the equivalent load did not adequately predict the deflection measured in the field test. Three different open web steel joists were tested under static uniform loads. The resistance function for each steel joist was calculated from these results and compared to current methods for developing truss resistance function. The experimental resistance function did not match these analytical resistance functions currently used in SBEDS. The following are the conclusions that can be drawn from this research.

105

The DAHS procedure is very conservative and does not compute accurate responses for roof systems.

Current methods for calculating blast pressures such as CONWEP and SBEDS correctly computes pressures at the ground level, but do not reasonably evaluate blast pressures on roofs.

The equivalent load procedure does not adequately predict response, even after adjusting the pressure input calculated from CONWEP or SBEDS.

The current methodology to develop resistance function for open web steel joists measures a higher peak load than experimentally tested.

The SBEDS assumption of perfect plastic behavior post-peak does not take into account the various failure mechanisms observed in open web steel joists.

In summation, current methods to calculate the blast loading on a roof and to design for the resistance of open web steel joist are not accurate. Therefore, the following recommendations are provided to develop an engineering method for blast design of open web steel joist roof system.

106

5.2

Recommendations

The following recommendations are made based on the observations of this study. Current methods to measure blast pressures achieve good results at the ground level. It has been observed that roof systems affect the blast load and effectively reduce the loading. The calculation of blast pressures on rooftops needs to be further researched. The equivalent uniform blast load procedure requires more research to better predict response. This can be accomplished by conducting an array of numerical solutions using various roof and blast parameters. Currently engineers only focus the first peak deflection in dynamic analysis. Higher deflections can occur after this. Engineering analysis and design should consider the first peak deflections and subsequent maximum deflections. The resistance function for open web steel joists should be developed based on more analytical and experimental research. Existing and additional field test in roof systems should be evaluated to verify developed engineering analysis and design.

107

REFERENCES
American Institute of Steel Construction (AISC), (2001). AISC Manual of Steel Construction Load and Resistance Factor Design. Chicago, Illinois. ANSYS ED Code Version 9.0, (2004). SAS IP. Biggs, J., (1967). Introduction to Structural Dynamics. McGraw-Hill Book Co. Chock, J. M. K., Kapania, R. K, (2001).Review of Two Methods for Calculating Explosive Air Blast. The Shock and Vibration Digest. Vol. 33, No. 2, pp. 91-102. Departments of the Army, the Navy, and the Air Force, (1990). Structures to Resist the Effects of Accidental Explosions, Department of the Army Technical Manual TM 51300, Department of the Navy Publication NAVFAC P-397, Department of the Air Force Manual AFM 88-22. Department of Defense, (2002). Design and Analysis of Hardened Structures to Conventional Weapon Effects (DAHS). United Facilities Criteria (UFC) 3-340-01. Forbes, D. J., (1999). Blast Loading on Petrochemical Buildings. Journal of Energy Engineering. Vol. 125, No. 3, pp. 94-102. Galambos, T., Xykis, C., (1991). The Effect of Lateral Bracing on the Stability of Steel Trusses. Journal of Construction Steel Research. Vol. 20, pp. 251-258. Hoemann, J., Salim, H.A., (2007). Experimental Evaluation of Structural Composites for Blast and Fragmentation Resistance Design. Air Force Research Laboratory. AFRLML-TR-2006-4578, Tyndall AFB, Florida.
Hyde, D. (1992). ConWep - Application of TM5-855-1, Structural Mechanics Division, Structures Laboratory, USAE Engineer Research and Development Center (ERDC), Vicksburg, Mississippi, USA.

Li, Q. M., Meng, H., (2002). Pressure-Impulse Diagram for Blast Loads Based on Dimensional Analysis and Single-Degree-of-Freedom Model. Journal of Engineering Mechanics. Vol. 128, No. 1, p. 82. Morison, C. M., (2006). Dynamic Response of Walls and Slabs by Single-Degree-ofFreedom Analysis A Critical Review and Revision. International Journal of Impact Engineering. Vol. 32, pp. 1214-1247. Nebuda, D., Oswald, C.J., (2004). SBEDS (Single-Degree-of-Freedom Blast Effects Design Spreadsheets). Proceedings from the 31st DoD Explosives Safety Seminar, San Antonio, Texas. 108

Smith, P. D., Hetherington, J.G., (1994). Blast and Ballistic Loading of Structures, Butterworth-Heinemann. Steel Joist Institute (SJI), (2005). 42nd Edition Standard Specifications Load Tables and Weight Tables for Steel Joists and Joist Girders. Steel Joist Institute. Sunshine, D., Amini, A., Swanson, M., (2004). Overview of Simplified Methods and Research for Blast Analysis. ASCE. U.S. Army Corps of Engineers, (2002). Lightweight Steel-Deck Roof System BlastResponse Evaluation. U.S. Army Engineering Research and Development Center. Geotechnical and Structures Laboratory (ERDC-GSL) Technical Report, TR-02-14. U.S. Army Corps of Engineers, (2005). Component Explosive Damage Assessment Workbook. U.S. Army Corps of Engineers. Protective Design Center (PDC) Technical Report, TR-05-02. Yost, J. R., Dinehart, D. W., Gross, S. P., Pote, J. J., Gargan, B., (2004). Strength and Design of Open Web Steel Joist with Crimped-End Members. Journal of Structural Engineering. Vol. 130, No. 5, p. 715.

109

Das könnte Ihnen auch gefallen