A Crashworthiness Study of a

Boeing 737 Fuselage Section

A Thesis

Submitted to the Faculty

of

Drexel University

by

Alan Byar

in partial fulfillment of the

requirements for the degree

of

Doctor of Philosophy

May 2003
 Copyright 2004
Alan D. Byar. All Rights Reserved.
 ii

Dedications

This is dedicated to Ann, for her persistent support, her willingness to sacrifice, and

her extreme patience, and to Christopher, who provided encouragement in his own

manner, and who helped to place this effort in its proper perspective.
 iii

Acknowledgements

This work was supported by a grant from the Federal Aviation Administration

William J. Hughes Technical Center through the FAA-Drexel Fellowship Research

Program. Mr. Gary Frings is the FAA Crashworthiness Program Manager, and John

Zvanya is the FAA Project Manager. Additional financial support was provided by

the Department of Mechanical Engineering and Mechanics at Drexel University.

The author would also like to acknowledge the guidance and support of his

graduate advisors, Professors Tein-Min Tan, Jonathan Awerbuch, and Alan Lau.
 iv

Table of Contents

List of Tables .............................................................................................................. vi

List of Figures ........................................................................................................... vii
Abstract ...................................................................................................................... xv
1. Introduction..............................................................................................................1
2. Review of Current Research ....................................................................................7
2.1 Purpose of Crashworthiness Studies.................................................................7
2.2 Two Numerical Methods for Impact Analysis of Air Vehicles........................9
2.3 Comparison of Automotive and Aircraft Crashworthiness Studies................10
2.4 Effect of Rivets on Impact Response of Simulated Airframe Structures .......14
2.5 Analysis of Energy Dissipation ......................................................................16
2.6 Analysis of Seats and Occupant Injury...........................................................18
2.7 Summary.........................................................................................................20
3. Description of the Test Article...............................................................................22
4. Finite Element Model and Simulation ...................................................................32
4.1 Overview of Model .........................................................................................32
4.2 Overhead Stowage Bins and Supports............................................................37
4.3 Luggage Model ...............................................................................................39
4.4 Simulation .......................................................................................................43
5. Experimental and Simulation Impact Response ..................................................46
5.1 Energy Dissipation During the Impact Event .................................................46
5.2 Deformation History of Fuselage Frames in Simulation ................................49
5.3 Comparison of Deformation Between Simulation and Experiment ...............69
5.4 Comparison of Impact Response Between Simulation and Experiment ........75
5.4.1 Filtering Method used with Experimental and Simulation Data ......75
5.4.2 Acceleration Histories of Heath Tecna Bin ......................................79
5.4.3 Load Histories for Heath Tecna Overhead Bin Supporting
Structures ..........................................................................................81
5.4.4 Acceleration Histories of Hitco Bin..................................................87
 v

5.4.5 Load Histories for Hitco Overhead Bin Supporting Structures ........89
5.4.6 Acceleration Time Histories of Seat Tracks .....................................93
5.4.7 Acceleration Histories of Frames......................................................98
5.4.8 Acceleration Time Histories of Seat Cushions ...............................104
6. Study of Simulation Parameters...........................................................................107
6.1 Effect of Friction Between Airframe and Platform .....................................107
6.2 Effect of Reduced Yield Strength ................................................................113
6.3 Effect of Material Failure Criterion on Impact Response............................117
6.4 Study of Luggage Properties........................................................................124
7. Study of Impact Conditions .................................................................................127
7.1 Effect of Roll Angle to Left at Impact .........................................................127
7.2 Effect of Luggage ........................................................................................133
7.3 Effect of Pitch Angle ...................................................................................138
8. Discussion and Summary.....................................................................................146
List of References ......................................................................................................156
Vita.............................................................................................................................160
 vi

List of Tables

1. Weights of Key Items in Test Article and Simulation........................................37

2. Luggage Material Properties...............................................................................43

3. Summary of Significant Events During Simulated Impact.................................69

4. Comparison of Vertical Displacement ...............................................................74

 vii

List of Figures

1. Vertical Drop Tower Test Facility at the FAA William J. Hughes

Technical Center .................................................................................................22

2. Right-Hand Side of Test Article Showing Outline of Frames and Cargo

Door ....................................................................................................................23

3. Front View of the B737 Fuselage Section Before Drop Test .............................24

4. End View of One of the Four Seat Tracks..........................................................25

5. Internal View of the Cargo Door ........................................................................25

6. Hitco Bin Installed on the Left-Hand Side of the Fuselage ................................27

7. Heath Tecna Bin Installed on the Right-Hand Side of the Fuselage ..................27

8. Supporting Structure of Hitco Bin with Tie Rod and Links ...............................28

9. Supporting Structure of Heath Tecna Bin with L Bracket and Strut ..................29

10. Wood Blocks Stowed in Hitco Bin.....................................................................30

11. Wood Blocks Stowed in Heath Tecna Bin .........................................................30

12. Finite Element Model of the Test Section ..........................................................32

13. Finite Element Model with Skin and Floor Removed ........................................33

14. Front View of Model ..........................................................................................35

15. Material Properties for Aluminum Airframe ......................................................36

16. Heath Tecna Bin and Supporting Structure – View Looking Outboard.............37

17. Heath Tecna Bin and Supporting Structure – View Looking Inboard................38

18. Hitco Bin and Supporting Structure – View Looking Outboard ........................38

19. Hitco Bin and Supporting Structure – View Looking Inboard ...........................39

20. Luggage Stiffness as a Function of Volumetric Strain .......................................40

21. View of Luggage in the Under-Floor Compartment...........................................41

 viii

22. Dissipation of Kinetic Energy During Impact ....................................................48

23. Two Views of Effective Plastic Strain in Frames, Time = 3 ms

to Time = 140 ms ................................................................................................50

24. Typical Seat Track Vertical Velocity Change – FAA Data................................68

25. Front View of Comparison of the Deformed Configuration ..............................70

26. Aft View of Comparison of the Deformed Configuration..................................71

27. Left View of Comparison of Deformed Configuration ......................................71

28. Right View of Comparison of Deformed Configuration ....................................71

29. Front Right Side Angled View, Test and Simulation .........................................72

30. Front Left Side Angled View, Test and Simulation ...........................................72

31. Typical Vertical Displacement of Floor During Impact .....................................73

32. Effect of Filter Frequency on Typical Experimental Acceleration Data ............77

33. Effect of Filter Frequency on Typical Simulation Acceleration Data ................78

34. Heath Tecna Bin Acceleration, Forward Location .............................................80

35. Heath Tecna Bin Acceleration, Aft Location .....................................................80

36. Heath Tecna Bin Acceleration, Center ...............................................................81

37. Heath Tecna Bin Acceleration, Average ............................................................81

38. Heath Tecna Bin Load, Forward Strut...............................................................82

39. Heath Tecna Bin Load, Aft Strut ........................................................................82

40. Heath Tecna Bin Load FS 400 – A Leg of Bracket............................................83

41. Heath Tecna Bin Load FS 400 – B Leg of Bracket ............................................83

42. Heath Tecna Bin Load FS 420 – A Leg of Bracket............................................83

43. Heath Tecna Bin Load FS 420 – B Leg of Bracket ............................................83

44. Heath Tecna Bin Load FS 440 – A Leg of Bracket............................................84

 ix

45. Heath Tecna Bin Load FS 440 – B Leg of Bracket ............................................84

46. Heath Tecna Bin Load FS 460 – A Leg of Bracket............................................84

47. Heath Tecna Bin Load FS 460 – B Leg of Bracket ............................................84

48. Heath Tecna Bin Load FS 480 – A Leg of Bracket............................................84

49. Heath Tecna Bin Load FS 480 – B Leg of Bracket ............................................84

50. Influence Coefficient for Vertical Struts of Heath Tecna Bin ............................86

51. Hitco Bin Acceleration, Forward Location.........................................................87

52. Hitco Bin Acceleration, Aft Location.................................................................87

53. Hitco Bin Acceleration, Center Location............................................................87

54. Hitco Bin Acceleration, Average........................................................................87

55. Hitco Bin Load, Forward Tie Rod ......................................................................89

56. Hitco Bin Load, Aft Tie Rod ..............................................................................89

57. Hitco Bin Load, FS 400 Vertical Link................................................................90

58. Hitco Bin Load, FS 400 Horizontal Link............................................................90

59. Hitco Bin Load, FS 420 Vertical Link................................................................90

60. Hitco Bin Load, FS 420 Horizontal Link............................................................90

61. Hitco Bin Load, FS 460 Vertical Link................................................................90

62. Hitco Bin Load, FS 460 Horizontal Link............................................................90

63. Hitco Bin Load, FS 480 Vertical Link................................................................91

64. Hitco Bin Load, FS 480 Horizontal Link............................................................91

65. Influence Coefficient for Combined Vertical Tie Rods on Hitco Bin ................92

66. Locations of Acceleration Results on Seat Tracks .............................................94

67. FS 380 Left Inside Seat Track ............................................................................95

 x

68. FS 380 Right Inside Seat Track ..........................................................................95

69. FS 418 Left Outside Seat Track..........................................................................95

70. FS 418 Right Outside Seat Track........................................................................95

71. FS 418 Left Inside Seat Track ............................................................................96

72. FS 418 Right Inside Seat Track ..........................................................................96

73. FS 452 Left Outside Seat Track..........................................................................97

74. FS 452 Right Outside Seat Track........................................................................97

75. FS 452 Left Inside Seat Track ............................................................................97

76. FS 452 Right Inside Seat Track ..........................................................................97

77. FS 484 Left Inside Seat Track ............................................................................97

78. FS 484 Right Inside Seat Track ..........................................................................97

79. FS 484 Left Inside Seat Track ............................................................................98

80. FS 484 Right Inside Seat Track ..........................................................................98

81. Locations for Upper Sidewall Acceleration Results...........................................99

82. Locations for Lower Sidewall Acceleration Results ..........................................99

83. FS 400 Left Upper Side Wall............................................................................102

84. FS 400 Right Upper Side Wall .........................................................................102

85. FS 440 Left Upper Side Wall............................................................................102

86. FS 440 Right Upper Side Wall .........................................................................102

87. FS 480 Left Upper Side Wall............................................................................102

88. FS 480 Right Upper Side Wall .........................................................................102

89. FS 400 Left Lower Side Wall ...........................................................................103

90. FS 400 Right Lower Side Wall.........................................................................103

 xi

91. FS 440 Left Lower Side Wall ...........................................................................103

92. FS 440 Right Lower Side Wall.........................................................................103

93. FS 480 Left Lower Side Wall ...........................................................................103

94. FS 480 Right Lower Side Wall.........................................................................103

95. Location of Acceleration Response for Comparison with ATDs .....................104

96. Left Side Response – ATD and Simulation at FS 408......................................106

97. Right Side Response – ATD and Simulation at FS 408

Note – Right side seats collapsed......................................................................106

98. Friction Coefficient = 0.0 (Left) and 1.0 (Right) at t=200 ms ..........................108

99. Average Acceleration Response of Heath Tecna Bin.......................................109

100. Average Acceleration Response of Hitco Bin ..................................................110

101. Acceleration Response of Left Outside Seat Track, FS 418.............................111

102. Acceleration Response of Right Outside Seat Track, FS 418 ..........................111

103. Acceleration Response of Left Outside Seat Track, FS 452.............................111

104. Acceleration Response of Right Outside Seat Track, FS 452 ..........................111

105. Acceleration Response of Upper Left Side Wall, FS 400 ................................112

106. Acceleration Response of Upper Right Side Wall, FS 400 ..............................112

107. Acceleration Response of Lower Left Side Wall, FS 400 ................................112

108. Acceleration Response of Lower Right Side Wall, FS 400..............................112

109. Deformation with 20 Percent Reduction in Yield Strength (Left)

and Baseline (Right) at t=100ms ......................................................................114

110. Average Acceleration Response of Hitco Bin ..................................................114

111. Load in Forward Tie Rod, Hitco Bin ................................................................115

112. Load in Aft Tie Rod, Hitco Bin ........................................................................115

 xii

113. Acceleration Response of Left Outside Seat Track, FS 418.............................115

114. Acceleration Response of Right Outside Seat Track, FS 418 ..........................115

115. Acceleration Response of – Left Inside Seat Track, FS 418 ............................116

116. Acceleration Response of Right Inside Seat Track, FS 418 .............................116

117. Acceleration Response of Left Outside Seat Track, FS 452.............................116

118. Acceleration Response of Right Outside Seat Track, FS 452 ..........................116

119. Effective Plastic Strain Failure Criterion Used in Study ..................................118

120. Post-test Failure of Cargo Door Frame.............................................................120

121. Failure of Elements in Door Frame on Right Hand Side..................................120

122. Fractures in Upper Frames on Left Hand Side .................................................121

123. Close-up View of Partial Failure in Frame at FS 380.......................................122

124. Failure (Left) and No Failure (Right) at t=100 ms ...........................................123

125. Failure (Left) and No Failure (Right) at t=200 ms ...........................................123

126. Effect of Luggage Stiffness, Eo = 1.2 (Left side) and

Eo = 0.6 (Right Side) at Time = 140 ms ...........................................................125

127. Effect of Luggage Stiffness, Eo = 1.2 (Left side) and

Eo = 0.6 (Right side) at Time = 250 ms............................................................127

128. 10 Degree Roll Angle at t=60 ms .....................................................................128

129. 10 Degree Roll Angle at t=140 ms ...................................................................128

130. Average Acceleration Response of Heath Tecna Bin.......................................129

131. Load in Forward Strut of Heath Tecna Bin ......................................................129

132. Load in Aft Strut of Heath Tecna Bin...............................................................129

133. Average Acceleration Response of Hitco Bin ..................................................130

134. Load in Forward Tie Rod of Hitco Bin.............................................................130

 xiii

135. Load in Aft Tie Rod of Hitco Bin.....................................................................130

136. Acceleration Response of Left Inside Seat Track, FS 380 ...............................132

137. Acceleration Response of Right Inside Seat Track, FS 380 .............................132

138. Acceleration Response of Left Outside Seat Track, FS 418.............................132

139. Acceleration Response of Right Outside Seat Track, FS 418 ..........................132

140. Acceleration Response of Left Outside Seat Track, FS 452.............................132

141. Acceleration Response of Right Outside Seat Track, FS 452 ..........................132

142. No Luggage at t=140 ms...................................................................................134

143. No Luggage at t=200 ms...................................................................................134

144. Average Acceleration Response of Heath Tecna Bin.......................................135

145. Load in Forward Strut of Heath Tecna Bin ......................................................135

146. Load in Aft Strut of Heath Tecna Bin...............................................................135

147. Average Acceleration Response of Hitco Bin ..................................................136

148. Load in Forward Tie Rod of Hitco Bin.............................................................136

149. Load in Aft Tie Rod of Hitco Bin.....................................................................136

150. Acceleration Response of Left Outside Seat Track, FS 418.............................137

151. Acceleration Response of Right Outside Seat Track, FS 418 ..........................137

152. Acceleration Response of Left Outside Seat Track, FS 452.............................138

153. Acceleration Response of Right Outside Seat Track, FS 452 ..........................138

154. Acceleration Pulse for Longitudinal Impact Condition ....................................140

155. Velocity Change Resulting From Longitudinal Acceleration Pulse.................140

156. Front and Left Views, Effect of Pitch Angle at t=100 ms ................................141

157. Front and Left Views, Effect of Pitch Angle at t=140 ms ................................141
 xiv

158. Acceleration Response of Heath Tecna Bin, Forward End ..............................142

159. Acceleration Response of Heath Tecna Bin, Aft End.......................................142

160. Load in Forward Strut of Heath Tecna Bin ......................................................142

161. Load in Aft Strut of Heath Tecna Bin...............................................................142

162. Acceleration Response of Hitco Bin, Forward End..........................................143

163. Acceleration Response of Hitco Bin, Aft End ..................................................143

164. Average Acceleration Response of Hitco Bin ..................................................143

165. Load in Forward Tie Rod of Hitco Bin.............................................................144

166. Load in Aft Tie Rod of Hitco Bin.....................................................................144

167. Acceleration Response of Left Outside Seat Track, FS 418.............................144

168. Acceleration Response of Right Outside Seat Track, FS 418 ..........................144

169. Acceleration Response of Left Inside Seat Track, FS 418 ...............................145

170. Acceleration Response of Right Inside Seat Track, FS 418 .............................145

 xv

Abstract
A Crashworthiness Study of a Boeing 737 Fuselage Section
Alan Byar
Tein-Min Tan, Ph.D.
Jonathan Awerbuch, Ph.D.
Alan Lau, Ph.D.

The Federal Aviation Administration (FAA) has been conducting drop tests

since the late 1980’s to determine the impact responses of aircraft structures under

severe but survivable crash conditions. The knowledge learned in each drop test,

however, is limited to one specific test condition. The prohibitively high cost of

conducting actual impact tests makes it necessary to develop a modeling and

simulation capability. Simulations may be used to study numerous issues in detail, to

support the design of improved crashworthiness airframe structures. Crashworthiness

simulations of aircraft, unlike those that are routinely performed in the automotive

industry, are much more difficult to perform due to the complexity of the airframe

structure and the unavailability of proprietary information. In this study, a

structurally realistic finite element model was developed to numerically simulate the

drop test of a Boeing 737 fuselage section that was conducted at the FAA William J.

Hughes Technical Center in November 2000. The model incorporated both

geometric and material non-linearity, and used an explicit-integration algorithm to

solve for the dynamic responses. Emphasis has been placed on predicting the

dynamic response of the structure, including the overall deformation of the fuselage,

the acceleration-time histories, and the load-time histories of key structural

components. It was determined that the fuselage frame and the under-floor luggage
 xvi

play the most important roles in energy dissipation during impact, and that the friction

between impacting surfaces markedly affects the deformation of the entire structure.

The effect of luggage stiffness, friction between fuselage and platform, and material

degradation due to fatigue and corrosion were also examined. Simulation results

compared well with those recorded during the drop test, indicating that the optimized

finite element model that was developed is suitable for use in crashworthiness studies

of aircraft. The model was then extended to study various other impact conditions.
 1

Chapter 1: Introduction

While the history of aviation crashworthiness dates back to the earliest days of

powered flight, modern systematic studies of survivability arose in large part from the

experience of the U.S. Army in the 1960’s. It was observed that numerous army

helicopter crashes had resulted in serious injuries or fatalities, even in cases where the

velocity at impact was relatively low. A series of studies were initiated to determine such

basic issues as the causes of these fatalities, and subsequently the range of impact

conditions for which fatalities could be prevented. It was determined that the vertical

force experienced by the occupant during impact was a key factor in many fatalities [1].

The military has since been deeply involved in developing crashworthiness standards for

military aircraft.

Commercial aircraft have also adopted crashworthiness standards for seats and

interior cabin items. Crashworthiness features have been incorporated into many new

designs, primarily in helicopters, and certification requirements continue to evolve to

enhance survivability; however, there are still areas where improvements could be made.

Crashworthiness research in the aviation industry has focused experimentally on

issues such as component testing of seats and energy absorbing sub-floor structures, and

on assessing occupant injury with test dummies. However, impact testing of complete

instrumented air vehicles is prohibitively expensive and extremely time consuming.

Therefore, such experimental investigations are not frequently conducted.

In the automotive industry, however, crashworthiness testing is an important and

highly visible component in the development of any new design. Testing of complete

vehicles is routine and is coupled with numerical simulations. These simulations are used
 2

to study impact energy absorbing designs under a range of impact conditions. Use of

simulations permits a significant reduction in the number of required experimental tests,

as the simulation may be used to study a variety of designs and impact variables.

In the aviation industry, there remains a need for basic experimental data on the

impact response of transport aircraft under survivable crash conditions. This has been

addressed by the Federal Aviation Administration (FAA), which has initiated a series of

crashworthiness tests at the William J. Hughes Technical Center, in Atlantic City, New

Jersey. These tests involve vertical impacts of commuter airframes, regional transports,

and fuselage sections of transport aircraft. The primary objective of these tests is to

characterize the impact response of the primary airframe structure, along with the

response of cabin components such as overhead stowage bins and seats.

Parallel to the test program at the FAA is an effort to develop a finite element

based crashworthiness modeling and analysis procedure for the pre- and post-test

analyses. Such a modeling and analysis procedure should provide additional insight into

the detailed acceleration time histories of the fuselage structures. It should also show the

overall pattern of deformation as a function of time, predict potential component failures,

and provide a detailed understanding of the dynamic response of large and complex

airframe structures throughout the entire impact event.

Crashworthiness modeling is intended both to complement and to support

experimental impact testing. An experimental database is essential for gaining

confidence in the simulation results, and for understanding the strengths and limitations

of the numerical model. The simulation, in turn, may be used to study issues that cannot
 3

be readily addressed through testing. The simulation may also be used to provide

guidance for future testing conditions or configurations.

Crashworthiness models, once validated, may be used to study other impact

conditions, thus reducing the need for extensive experimental testing. Furthermore, a

validated modeling methodology may be used as a basis for simulating other airframes.

The finite element simulation approach is also well suited for studying a variety of

crashworthiness related issues, such as oblique impact, and to study a wide range of

design options. Simulations can be used to study the design of improved impact

absorbing airframe structures, and greatly reduce the necessity for impact testing.

Numerical simulations may be used to study the effects of individual components on the

dynamic structural responses of the fuselage. Specific design features such as reinforcing

structures, the content of the cargo, the type of overhead bins used, seat configurations,

gross weight, or structural material can be studied as separate variables. Such studies are

currently not practical to perform on an experimental basis.

The objectives of this study are four-fold. First is to develop a finite element

model to simulate a vertical drop test of a 10-foot long B737 fuselage section containing

two different overhead stowage bins. The particular drop test analyzed herein was

conducted in October 2000 at the vertical drop test facility located at the FAA William J.

Hughes Technical Center. The model is validated through detailed comparisons of

simulation and experimental impact response data. The test section was instrumented to

record acceleration time histories from locations on the floor, frames, and overhead

stowage bins, as well as other data. These results are compared with the results from the

simulation at corresponding locations.

 4

Further validation will be found based on the ability of the simulation to provide a

comprehensive interpretation of the dynamic response of fuselage components

throughout the entire impact event. The ability to understand and explain the resulting

acceleration responses in terms of structural deformation during impact also provides a

basis for improving crashworthiness of aircraft components.

Validation of the simulation is therefore a major objective in this study. There are

still a relatively small number of aircraft crashworthiness simulations that have been

studied in detail, and compared with experimental results. Validation of the current study

will not only increase confidence in aviation crashworthiness simulations, but may also

assist in guiding future numerical studies.

The second objective is to address the issue of finite element mesh density, or

overall level of modeling detail, for the crashworthiness studies. This issue is of great

importance in the aviation industry, in light of Kindervater’s statement [2] that it is not

practical at present to conduct a detailed simulation of a complete aircraft. This

statement is true, using the mesh density currently employed by the automotive industry.

While guidelines do exist for modeling automotive structures for impact analysis, no such

guidelines exist with respect to aircraft structures. A strict use of automotive guidelines

in aircraft will lead to unrealistic model sizes.

An effort was made in the present study to adopt the level of modeling detail

needed to capture the basic buckling and crushing behavior of the fuselage structure.

This should permit the simulation to accurately calculate the basic acceleration pulse that

is experienced by critical cabin items and occupants. The mesh density in this study can

realistically be extended to a simulation of a complete air vehicle, based on current

 5

computing capabilities. This issue is significant, because simulations which focus on

attaining fully convergent results will be restricted to modeling only components or sub-

sections of airframes. This is due to the high mesh density needed for such similitude.

The present study addresses the question as to whether a more reasonable mesh density

might still capture the basic fuselage impact response.

The third objective is to study a selected set of key simulation parameters to

determine basic guidelines for performing future crashworthiness simulations of air

vehicles. Issues studied include friction between the fuselage and the impact surface,

modeling of the luggage, possible degradation of material yield strength due to corrosion

and fatigue, and the use of an element failure criterion.

The fourth objective is to use the validated simulation to study a range of impact

conditions. Experimental testing provides valuable data for a given impact condition and

test configuration, but is not practical for studying a variety of conditions. Numerical

methods, however, are ideally suited for analyzing such issues as changing roll angles,

presence of luggage, and combining longitudinal and vertical impact conditions.

The importance of studying multiple impact conditions is illustrated by the

experience of the automotive industry. It was initially believed that frontal impacts

represented the most significant crash scenario, and such impacts were therefore used to

study worst case collisions. It was later found that offset impacts can be equally

dangerous, if not properly addressed, and further testing conditions were added to the

basic frontal impact requirements. This issue would not have been discovered based on

component level testing. Given the very limited experimental data on aircraft impacts, it

is possible that similar discoveries await the aviation industry.

 6

A future goal would be to employ the present methodology to perform an air

vehicle impact simulation of a complete airframe, including not only the fuselage and

cabin items, but also the occupants. With such a simulation, the methodology would then

be in place to perform air vehicle crashworthiness design studies, to study multiple

impact conditions, and to analyze designs for optimizing occupant survivability.

 7

Chapter 2: Review of Current Research

This section provides an overview of aviation crashworthiness studies, and issues

that are currently of interest in the simulation of aircraft impact events.

2.1. Purpose of Crashworthiness Studies

The interest in aviation crashworthiness extends to the earliest period of flight.

This interest was initially motivated by the simple observation that pilots occasionally

survive impacts which have proven fatal to others, and that the velocity at impact is not

the only factor to explain this difference in outcomes. An excellent review of

crashworthiness history, beginning with the pioneering days of aviation, is provided by

Waldcock [3]. He notes that the first systematic statement of crashworthiness principles

is attributed to Hugh DeHaven, who may be considered the father of aircraft

crashworthiness. DeHaven’s own interest in survivability arose from his crash following

a midair collision, an event in which the other pilot was killed.

Waldcock cites DeHaven’s four basic principles, which are modeled on

observations of the freight and packing industry. In the analogy with packaging, the

container in this case is the fuselage, the interior packaging is the seat and restraint

system, and the objects inside the container are the occupants. DeHaven’s four principles

are listed below:

1) The package should not open up and spill its contents, and should not

collapse under the expected loading conditions.

2) The packaging structures which shield the inner container must not be

made of either brittle or weak materials; they should resist impact force by

yielding and absorbing energy.

 8

3) Articles contained in the package should be held securely and

immobilized. This interior packaging is an extremely important part of the

overall design, for it prevents movement inside the package and the

resultant damage from impact within the package itself.

4) The method of holding an object inside a shipping container must transmit

the forces applied to the container to the strongest parts of the contained

objects.

Modern crashworthiness studies seek to implement these same basic principles.

Early studies relied on experimental methods, while later studies frequently employed

numerical methods in conjunction with testing. A summary of recent analytical studies in

crashworthiness is provided by Kindervater [2]. He offers a clear evaluation of the state

of the art in numerical modeling for aviation crashworthiness, focusing on the design and

analysis of composite fuselage structures. Kindervater makes several noteworthy

observations concerning finite element (FE) studies of aircraft crashworthiness, some of

which are listed below.

• While explicit FE codes are used to simulate complete automotive

vehicle impacts, including occupants and airbags, much less attention

had been paid to simulating aircraft structures under dynamic loading

conditions. In fact, no equally detailed simulation has yet been

performed of an air vehicle with occupants.

 9

• The ideal load curve during impact is one where stiffness increases

gradually while crushing, as this maximizes the amount of impact

energy being dissipated. This is termed the controlled load concept.

• Further research is needed to characterize appropriate failure criteria

for modeling of structures during impact.

2.2. Two Numerical Methods for Impact Analysis of Air Vehicles

Two distinct numerical approaches have been used in modern aircraft

crashworthiness studies. The first is termed a hybrid finite element code which is based

on modeling an airframe with a lumped-mass spring damper system. The term hybrid

arises from the fact that this code requires experimental or otherwise derived data to

describe the impact response of the fuselage structure when it contacts the ground.

Spring elements, having prescribed load deflection curve, are used to simulate crushing.

The fuselage is generally modeled very simply with several hundred elements at most.

An example of such a hybrid code is KRASH [4], which was developed by Lockheed

under sponsorship from the U.S. Army. KRASH has been implemented by the Cranfield

Impact Center into an interactive Air Accident Investigation Tool (AAIT), with a library

of commuter aircraft models [5]. The advantage of codes such as AAIT is that they are

relatively easy to use. Simulations require little computational time, typically a matter of

minutes. The disadvantage of such codes is that they require either substantial effort to

accurately determine the spring parameters, and/or extensive experience and judgement

to reasonably estimate these parameters.

The second approach is that of explicit finite element codes, such as Pam-Crash

[6], MSC Dytran [7], and LS-DYNA [8]. These codes are based on fundamental
 10

engineering principles and, given correct material properties and geometry, require no

prior experimental testing. In theory, such codes are fully predictive. In practice,

however, it is frequently noted that experimental verification remains as an important

requirement [9]. This is due in part to the inevitable simplifications needed to model a

complex structure, which may lead to a slightly different pattern of deformation than

would be observed in testing. Also, as previously noted, characterization of material

properties under impact loading conditions is still in its infancy. Thus, validation of

numerical analyses continues to be a critical concern, particularly for transient dynamic

simulations.

2.3. Comparison of Automotive and Aircraft Crashworthiness Studies

While crashworthiness research has evolved a substantial body of knowledge in

recent years, most of the work has been focused in the automotive industry. The

automotive industry now includes numerical simulation of impacts as a routine and

integrated tool in the design process. In fact, the fabrication of even preliminary

structures does not begin until crashworthiness simulations have been conducted.

Volkswagen, for example, discusses their 10 years of experience with crash impact

analysis, in a paper published in 1994 [10]. Even at that time, with less capable

computational abilities, Volkswagen was routinely employing finite element methods to

simulate a variety of impacts with existing and proposed designs. Other manufacturers

now also engage in numerical simulations of vehicle impacts on a regular basis [11].

This integrated experimental and analytical approach reduces development time and it

also permits the study of a much wider variety of designs and impact conditions, as

compared with conducting experimental testing alone.

 11

The situation in the aircraft industry is very different in that no detailed

simulations are generally conducted of transport air vehicles. This is not to suggest that

no crashworthiness studies are underway in aviation. Rather, such studies are typically

directed toward understanding specific issues, or toward modeling of relatively small

structures. Examples of these studies include a detailed simulation of crushing of small

riveted structures [12], impact response of small sub-floor assemblies [13], dynamic

buckling of beams and plates [14, 15, 16], use of energy absorbing box or foam filled

structures [17], rate sensitive material properties [18, 19], and design of energy absorbing

composite structures for use in general aviation aircraft [20]. Also, the helicopter

industry is heavily involved in crashworthiness research, both analytically and

experimentally [21]. Finally, extensive experimental studies and dynamic certification

tests are performed with seats, as will be discussed below.

While complete vehicle crash simulations for automobiles are now standard

practice, Kindervater, among others, has noted that “rather surprisingly, much less

attention has been paid to the development of FE simulation of aircraft structures under

dynamic crash conditions, despite the extremely high cost of aircraft crash tests” [2].

Several reasons can be given for this situation.

First, a better understanding of the details of impact behavior of aircraft

components is required prior to conducting predictive simulations. As noted above,

numerous studies have been devoted to testing of components or small assemblies

representing sub-floor structures.

Second, modeling full airframes, or even large sections of transport aircraft, is

difficult because of the resulting large model size and because of a general lack of access
 12

to detailed geometric information. In the early 1990’s, a full automotive vehicle

simulation would involve 30,000 to 40,000 elements [10], due to limitations in hardware

performance. Ten years later, full simulations typically involve from 300,000 to 400,000

elements and more. Even with this more increased capabilities, however, it is not

considered feasible to model a full airframe to a level of detail similar to that used in the

automotive industry. Thus, simplifications are required such as not explicitly modeling

rivets or substituting beam elements in place of frames or stringers that should be

modeled with shell elements. Access to geometry is also crucial. For in-house studies,

the automotive industry can access their own geometry and material database for use in

their finite element models. Outside automotive research groups can and do fully

disassemble an automobile, scan each part, and use the digitized data in constructing a

model. No one, however, has attempted to do this with a commercial transport aircraft,

due to the time and expense involved. While the required database for a complete air

vehicle model surely resides in aircraft companies, such finite element models for use in

impact studies have not as yet been constructed.

A third possible reason for the lack of aircraft simulations of complete transport

airframes is the more urgent need to study smaller general aviation and small regional

transports, as these are typically lacking in crashworthy abilities. This poor crashworthy

performance is due both to the lack of available crushing stroke and, in many general

aviation aircraft, to the limited ability of basic composite structures to absorb impact

energy. The crushing stroke refers to the available distance for crushing before intruding

upon occupant space, and this dimension is physically more limited in smaller airframes.
 13

A final difficulty in performing crashworthiness simulations of large airframes is

the lack of experimental data with which to validate a model. The validation step is

essential in gaining confidence in the predictive capabilities of the simulation. An

experimental database is essential in developing dynamic crashworthiness simulations,

both in the automotive and aviation industries.

Some of the above issues are currently being addressed. The issue of

crashworthiness of small composite general aviation airframes has been addressed by

Jackson and Fasanella [22], among others. Their studies have combined testing and

simulation of composite structures, to develop and verify new crashworthy design

concepts. The combination of simulation and testing is an important part of the process.

Simulations can assist in selecting, or possibly ruling out proposed designs and testing

provides a point for validation of the modeling approach. The proposed design concept

in this case was a partially foam filled sub-floor. Jackson and Fasanella concluded that

the exact geometry of the foam filled section was a crucial variable in determining the

pattern of deformation and degree of energy dissipation during impact.

The issue of compiling an experimental database of air vehicle impacts is being

addressed by the Crashworthiness Test Program which is conducted at the FAA William

J. Hughes Technical Center outside of Atlantic City, New Jersey. This program has

conducted fully instrumented tests of general aviation aircraft, regional transports, and

sections of narrow body transport aircraft. This provides an invaluable database for

assessing crashworthiness simulations.

This database has been taken advantage of by Jackson and Fasanella [23, 24], in

cooperation with the FAA. A vertical drop test was conducted of a ten foot long section
 14

of a B737 fuselage section, with a conformal fuel tank attached under-floor. This drop

test was simulated in detail and results were compared with experimental test data. The

overall comparison of simulation and test results was generally good with the simulation

capturing many of the buckling details seen in the actual test. Some features of the test

event were difficult to capture, however, such as the failure of the under-floor beams in

contact with the fuel tank and the failure of the supports for the fuel tank early in the

impact event. This test and simulation represented a significant milestone in the path to a

comprehensive and integrated study of crashworthiness of large transport aircraft.

Other models of transport aircraft have been constructed, but for the different

purpose of assessing the interaction of the aircraft and various ground structures at

impact. One such study was performed by Lawver [25] using beam and shell elements in

complete models of a C-130 and C-141. This study focused on impacts into reinforced

concrete and steel structures. The wings and engines were represented by adding weights

in the wing box area, and the level of detail in the model was not sufficient for evaluating

crashworthiness features in the fuselage structure. Although the effect of impact on the

aircraft cargo was evaluated, the primary focus of this study was to assess the force-time

history on the ground structure and to determine the degree of damage to be expected

under different impact conditions.

2.4. Effect of Rivets on Impact Response of Simulated Airframe Structures

Several studies have addressed the issue of simulating rivets in airframes for

crashworthiness analyses. The primary question is whether or not the inclusion of

simulated rivets will significantly alter the buckling response of the airframe, and hence

the acceleration pulse transmitted to occupants and cabin items. Matzenmiller [26]
 15

studied a number of options that are available for modeling riveted joints, such as discrete

beam elements that may be used to simulate rivets or bolts. The beam elements are used

to connect two adjacent structures such as an angle and frame, or frame and skin. Beam

elements, representing rivets, may be given nonlinear plastic properties, along with any

of several failure criteria. After discussing these options, Matzenmiller constructed a

model to simulate a small lap joint test. The simulation was used to illustrate the effects

of different modeling techniques. Although his results show significant differences in

load transfer using the different joint modeling options, his study was conducted with a

model of a small test coupon. This coupon is not representative of an airframe structure

buckling under impact, and cannot be used to draw any conclusions concerning the

effects of rivets on impact studies. Thus the study by Matzenmiller is not conclusive in

determining whether this level of modeling detail is needed for impact simulations.

Bisagni [27] used the finite element method to simulate the impact response of an

AGUSTA helicopter sub-floor structure. Beam elements with a failure criterion were

used to model the rivets. It was found that the simulation results, in terms of buckling

pattern and load time history during crushing, compared well with experimental data

regardless of whether the rivets were included in the model or not. More significantly,

his study indicated that failure of rivets has very little effect on the peak and average

crushing loads, and hence is not so critical for the prediction of total crash response of the

structure.

Vignjevic and Meo [28] simulated an under-floor structure of a helicopter in a

water impact. The structure is a slightly simplified version of that found in the Lynx

helicopter. In this simulation rivets were ignored and beams and bulkheads were joined
 16

directly. The experimental pattern of deformation corresponded closely with that

obtained through simulation. In the conclusions from this work, Vignjevic and Meo

proposed adding rivets to a more detailed version of the model to assess the effect of

including rivets with a failure criterion.

While the rivet modeling issue is still under investigation, it is important to note

that the effect of simulating rivets may vary with the size of the structure being modeled.

During a crash test rivets tend to bear out of the rivet holes prior to the buckling or

fracture of the structure, and hence dissipate a portion of the impact energy. This portion

of energy may be important for smaller structures, but is too small, in the case of

transport air vehicles, to have any significant effect on the pattern of deformation or

buckling of the under-floor frames. For much smaller structures, use of simulated rivets

may have also have a more significant impact on the resulting buckling patterns, because

they will alter the local load paths. Thus if the primary focus is on determining the

acceleration pulse experienced by occupants and by components in the upper fuselage

section of a transport aircraft, then the detailed modeling of rivets may not be necessary.

A definitive study on the effect of rivets in crashworthiness simulations has yet to be

performed.

2.5. Analysis of Energy Dissipation

In crashworthiness studies, it is very difficult, and in many cases impossible to

determine experimentally the conversion of kinetic energy of the impacting bodies to

internal energy dissipated by the individual components of a structure during impact.

Finite element impact simulations, on the other hand, provide a convenient way for
 17

identifying and optimizing energy absorption of individual components within a

structure.

In his study of an energy absorbing highway guardrail Reid [29] employed an

explicit finite element approach to determine precisely where and how the impact energy

is absorbed and dissipated. Subcomponents of the guardrail system were modeled first in

order to perform a trial simulation with a moving deformable barrier and a simplified

vehicle model. The guardrail structure was intended to buckle, or kink, in a predictable

pattern, and it was therefore termed the sequentially kinking terminal (SKT). After the

trial simulation the complete sequentially kinking terminal was modeled, and an impact

simulation was conducted with an existing pickup truck model obtained from the

National Crash Analysis Center (NCAC). A small car model was also used in a

simulated impact. After validating the simulation through comparisons with

experimental data, the results were analyzed to determine which components were most

responsible for energy absorption.

This study determined the percentage of energy dissipated by the guardrail, the

SKT head, wooden posts, the impacting vehicle, and friction. By carefully assessing the

energy dissipated by each component, it was possible to identify critical design features

and to focus efficiently on optimizing the most important design components. This

illustrates a key advantage of impact simulations over experimental testing.

In a similar study Altenhof [30] investigated the energy absorption characteristics

of a steering wheel armature during impact. Altenhof also noted that in the case of

impact involving deformable bodies, “experimental methods do not provide a method of

determining the energy absorbed by any single entity.” Finite element methods, however,
 18

are well suited to quantifying energy absorption of individual components involved in an

impact event. Use of such methods permits crashworthy evaluation and improvements in

steering wheels early in the design phase. This study simulated the impact between a

section of a deformable test dummy and a steering wheel. The model was validated by

comparing simulation results with experimental data. Simulation results were then

analyzed to determine the critical components in dissipating energy during impact.

Again, this permits designers to focus on these critical components, and to efficiently

optimize their designs for energy absorption.

There do not appear to be any similar studies of energy dissipation concerning

transport aircraft. While the under-floor structure in general is known to be critical in

determining the impact response of the fuselage, studies have not yet specifically tracked

the energy dissipation for individual components, such as luggage, frames, bulkheads,

stringers, or skin panels. This issue is addressed in the current study.

2.6. Analysis of Seats and Occupant Injury

The passenger seat is a critical component of any crashworthy design.

Accordingly, extensive dynamic seat testing has been conducted by the FAA as well as

by other groups [31, 32, 33]. Tests are conducted both to certify seats for use in

commercial aircraft and to study proposed new designs for energy absorbing seats and

their supporting structures.

Hooper and Ellis [34] summarize the crashworthiness issues and regulatory

requirements for passenger seats in both small and large aircraft. These requirements

include two dynamic test conditions for seats having dummy occupants. The first test

condition employs combined vertical and longitudinal loads and the second test condition
 19

employs combined longitudinal and lateral loads. The peak acceleration level and pulse

duration is specified for each test, depending on the aircraft type, and maximum

permissible injury criteria are specified.

While seat testing is currently the only available method of certifying seat

performance during impact, numerical methods are also used to study occupant response

in order to predict injuries. Among many available analytical tools, MADYMO is one of

the most commonly used tools for occupant modeling [35]. MADYMO uses rigid body

ellipsoids to model test dummies and currently has a library of validated models to

represent numerous test dummies. Interaction between the rigid bodies and the

surrounding finite element structure is given by experimentally determined contact

parameters. MADYMO excels in duplicating the kinematic behavior of test dummies, by

accurately representing the joints and relative motions of body segments. MADYMO is

also used in both the automotive and aviation industries to predict potential injuries by

calculating, for example, lumbar or femur loads during impact. MADYMO has also been

used to calculate head injuries sustained when contacting an object in front of the

occupant during an impact. The object could be a steering wheel, dashboard, or in air

vehicles, another seat.

The primary drawback of MADYMO is that the program is based on rigid, non-

deformable bodies, and therefore cannot accurately simulate the interaction between

deformable bodies. While MADYMO can utilize finite element codes as a subroutine,

the interaction between the finite elements and the rigid occupant model must be

specified artificially with parameters, rather than employ basic material properties with

deformable structures. These parameters are frequently difficult to determine prior to

 20

testing. To alleviate this problem, work was initiated on a fully finite element based

dummy model and a new series of such models are now available [36]. There are two

primary drawbacks in the adoption of fully finite element occupant models. First, the

joint descriptions and resulting kinematic motion are more difficult to capture with a

finite element model. Second, the use of a complete and detailed finite element dummy

model is computationally intensive. It is likely that both of these difficulties with finite

element dummy models will be overcome with continuing advances in computing

capabilities.

2.7. Summary

One of the clear advantages of numerical studies is noted by Bisagni, who writes

that “finite element analysis can be used to aid the designers in evaluating the

crashworthiness of different structural concepts, and can therefore be an important means

of reducing development costs” [27]. This statement was made in 2002 illustrating the

degree to which the aviation industry lags behind in adopting a more widespread use of

numerical simulations for crashworthiness studies. Kindervater noted earlier, in 1999,

that finite element codes, “though well accepted in the automotive industry, are not yet

established in the aircraft industry” for crashworthiness studies [2].

Although an integrated systems approach to crashworthiness would be preferred,

most tests are still conducted on a component basis. Only a limited effort has been made

to evaluate complete integrated designs, from airframe to occupants. Static testing or

simple analysis is used to satisfy the crashworthiness requirement that individual cabin

items remain adequately fixed during impact. Current crashworthiness certification of

seats is done exclusively with experimental testing, though efforts are underway to
 21

employ numerical methods in conjunction with testing. A complete and predictive air

vehicle impact simulation would include accurate modeling of the airframe, cabin

components, luggage, seats, and occupants, and would capture the interaction of all such

items. The current work does not include occupant models but represents all other

entities in an integrated model. This model may therefore be used to simulate the

response of all structural components to impact loading conditions.

 22

Chapter 3: Description of Test Article

This section describes briefly the drop test that was conducted at the FAA

William J. Hughes Technical Center near Atlantic City, New Jersey. As previously

noted, the test preparation, testing, and collection of experimental data were performed

by the FAA. A detailed description of the test can be found in the FAA report [37].

The test article was a 10-foot long Boeing 737 (B737) fuselage section and is

shown in Figure 1 as it is suspended by a hook at the drop test facility. The section was

dropped from a 14-foot height, resulting in a 30-ft/sec initial impact velocity. This

impact velocity was selected to represent a severe but survivable impact.

Figure 1. Vertical Drop Tower Test Facility at the William J. Hughes Technical Center.
 23

The test section includes seven frames that are spaced 20 inches apart. The

location of each frame on the actual fuselage is identified by a Fuselage Station (FS)

number, which indicates the distance in inches from reference point located near the

forward end of the fuselage. A higher station number indicates a position further away

from the reference point, meaning that higher station numbers are further aft. The frames

on the test section range from FS 380 to FS 500, as shown in Figure 2. Right and left

sides of the airframe are indicated with respect to the pilot.

FS 500 Outline of Cargo Door FS 380

Figure 2. Right-Hand Side of Test Article Showing

Outline of Frames and Cargo Door
 24

The test article includes a floor in the passenger compartment supported by seven

under-floor beams that are attached to the seven frames. When preparing the test section,

the adjacent fuselage sections that existed in front of FS 380 and aft of FS 500 were

removed, resulting in a loss of structural stiffness of the test section. In order to

compensate for the reduced stiffness and support, a second under-floor beam was added

under the beam at each end of the fuselage section, at FS 380 and FS 500, respectively.

The added beam at FS 380 is seen in Figure 3. The under-floor compartment is fully

loaded with luggage, which played a very significant role in the impact response of the

fuselage section.

Added
Floor
Beam

Figure 3. Front View of the B737 Fuselage Section Before Drop Test.
 25

Three rows of triple seats are installed on each side of the test section in the

passenger compartment, with twelve mannequins and six anthropomorphic test dummies

occupying the seats. The seats are anchored to four longitudinal seat tracks, two on each

side, that are rigidly connected to the floor, as seen in Figure 4.

Figure 4. End View of One of the Four Seat Tracks.

Figure 5. Internal View of the Cargo Door.

 26

A significant feature of the fuselage section is a cargo door located on the lower

right side. An interior view of the cargo door, Figure 5, shows that the cargo doorframe

is heavily reinforced. This makes the structure on the lower right side of the fuselage

much stiffer than that of the lower left side. The increased stiffness caused by the

presence of the cargo doorframe has a significant effect on the deformation and

acceleration of the fuselage structure and its components, as was observed during the

actual drop test. This will be discussed in further detail in Chapter 6, when presenting

acceleration response data for the test and simulation.

Two different overhead stowage bins were installed in the fuselage section. They

are a Hitco bin on the upper left side and a Heath Tecna bin on the upper right side, as

seen in Figures 6 and 7, respectively. These two bin designs are currently approved and

certified by the FAA. One of the primary goals of the FAA drop test was to determine

the reaction of these two bins to a severe but survivable impact condition. It is noted that

the impact condition implemented in this particular test resulted in peak acceleration

pulse of approximately 12 to 14 g for the overhead stowage bins. This acceleration pulse

is well in excess of the bin certification requirement, which is based on an equivalent

static load of 5 g.
 27

Figure 6. Hitco Bin Installed on the Left-Hand Side of the Fuselage.

Figure 7. Heath Tecna Bin Installed on the Right-Hand Side of the Fuselage.

Both bins are approximately 60 inches long, extending from FS 410 to FS 470,

but each bin employs a different support structure. For the Hitco bin, the primary vertical

support is provided by a vertical tie rod located at each end of the bin, Figure 8. The top

end of each tie rod is connected to a horizontal cylindrical rod, which is then mounted to
 28

the frames. Outboard support of the Hitco bin consists of a pair of short beams that are

connected to the forward and aft outboard edges of the bin, and vertical and horizontal

links at each end of the beam that are connected to the frames. The short vertical links

provide additional vertical support, while the horizontal links provide the

inboard/outboard supports for the bin. A drag strut, which is not shown in the figure,

provides longitudinal support for the bin.

Vertical and Horizontal

Links
Tie
Rod

Figure 8. Supporting Structure of Hitco Bin with Tie Rod and Links.

The primary vertical support for the Heath Tecna bin is provided by two vertical

struts located at the forward and aft ends of the bin, Figure 9. Each vertical strut is

attached to an overhead C-channel that is fixed to the frames. Outboard support of the

bin is provided by a longitudinal channel that is fastened to the outboard edge of the bin.
 29

The channel extends from FS 400 to FS 480 and is attached to the frames with five L

brackets, each consists of an inclined A leg and a horizontal B leg. The B legs provide

inboard/outboard support while the A legs provide both inboard/outboard and additional

vertical support for the bin.

A Leg of
L Bracket C Channel

B Leg of
L Bracket Vertical Strut

Figure 9. Supporting Structure of Heath Tecna Bin with L Bracket and Strut.

Each bin is loaded with wooden blocks to simulate luggage. The Hitco bin

weighs 57 lb and is loaded with 200 lb of wood blocks, as shown in Figure 10. The

Heath Tecna bin weighs 56 lb and is loaded with 120 lb of wood blocks, as shown in

Figure 11.
 30

Figure 10. Wood Blocks Stowed in Htico Bin.

Figure 11. Wood Blocks Stowed in Heath Tecna Bin.

Two cameras were installed on two overhead camera mounts. One was attached

to the frames at FS 380 and FS 400 and the other to frames at FS 480 and FS 500. The

cameras were used to record the responses of fuselage interior during the drop test. After

the drop test, it was noticed that the weight of the cameras and camera mounts caused

considerable plastic deformation in the frames. It is therefore important to include the

cameras and the mounts in the finite element model in order to account for their effect on
 31

the responses of the neighboring structural components, such as the overhead stowage

bins and frames.

The test article was instrumented at various locations with strain gages and

accelerometers to record the strain and acceleration histories during the drop test. In the

simulation, accelerations and forces are calculated at the same locations for comparisons

with the recorded data.

 32

Chapter 4: Finite Element Model and Simulation

4.1. Overview of Model

A finite element model of the fuselage test section was developed using the

PATRAN computer software code. Following is a description of the components of this

model along with a discussion of significant simulation parameters.

The finite element model represents all features of the test article, with the

primary exception that masses for the test dummies and mannequins were distributed on

seat surfaces. Key components of the model include frames, under-floor beams,

overhead stowage bins, camera mounts, and a cargo door, Figure 12. The model was

constructed based on detailed hand measurements of the actual test article.

The finite element model is designed to produce a high fidelity replica of the

actual test article. The masses of the high speed cameras used in the drop test were

included on the mounting platforms. The right-left orientation of the airframe is given

with respect to the pilot.

Figure 12. Finite Element Model of the Test Section.

 33

An important area of the model is the cutout for the cargo door, shown in the

lower right-hand side of Figure 13. The forward and aft ends of the cutout frames are

reinforced with short beams, linking the door frames with the closest fuselage frame. The

forward edge of the cargo door frame aligns with the frame at FS 420, and the aft edge of

the door is located between frames at FS 460 and FS 480. The short frame sections

between the upper edge of the cargo door and the floor beams are also reinforced.

Figure 13. Finite Element Model with Skin and Floor Removed

This added stiffness from the cargo door reinforcements on the right-hand side

has a major effect on the overall response of the fuselage structure to the impact loading,

as will be shown in Chapter 5. Furthermore, as previously noted, the front edge of the
 34

cargo door coincides with the frame at FS 420, while the aft edge of the cargo door ends

between the frames at FS 460 and FS 480. These two different door edge support

structures result in two different load paths, causing a considerable influence on how the

impact energy is transmitted from the lower frames to the upper frames. The difference

in these two load paths for the front and aft cargo door edges is combined with the further

asymmetry of the cargo door being offset to the rear of the fuselage test section. It would

therefore be expected that the test section would deform differently at its forward and aft

ends. Thus the cargo door causes the fuselage structure to be asymmetrical in both the

right to left and forward to aft directions. This effect of the cargo door cutout is evident

in both the experimental and simulated impact results.

The fuselage section is modeled primarily with shell elements. This includes the

fuselage skin, the frames, the floor and its supporting beams, the cargo door, and the

camera mounts. Using shell elements has the advantage of being able to more accurately

simulate buckling and crippling during impact, as well as more accurately calculating

internal energy absorption.

The seat frames are modeled with beam elements, and the seat surfaces are

modeled with shell elements. The masses of the mannequins and dummies are

distributed on the lower seat surfaces, to approximate the loading of occupants on seat

cushions. This was done to approximate the delay that occurs in transferring inertial

forces from the occupants to the seat tracks. The effect of damping due to the seat

cushions was not included in the simulation. Seats and luggage are indicated in Figure 14.
 35

Seats

Luggage

Empty
Space

Figure 14. Front View of Model.

The overall finite element model consists of 68,007 nodes, 53,407 shell elements,

13,824 solid elements, and 691 beam elements. Solid elements were used exclusively in

modeling the luggage, as will be discussed below. Shell elements employ a reduced-

integration scheme in calculating element stiffness. This is done to enhance solution

efficiency. As a routine procedure in this type of impact analysis, the viscous hourglass

damping option is activated to prevent any potential spurious deformation modes. To

accurately capture plastic deformation, five integration points through the element

thickness were used.

The material properties used in the finite element model are based on those of

typical materials used in the construction of the Boeing 737 fuselage. Aluminum 2024-

T3, which possesses superior fatigue characteristics, is used for the skin panels.

Aluminum 7075-T6, which is a high strength alloy, is used for all other structural

members, including frames, under-floor beams, and stringers. Material properties were
 36

taken from a standard aviation textbook. A point-by-point curve with 100 data points

was used to represent the actual stress strain curves of both aluminum alloys, as shown

below in Figure 15.

Figure 15. Material Properties for Aluminum Airframe.

The total weight of the finite element model is 8,974 lbs, and total weight of the

test article was 8,870 lbs. Weights of significant mass items are listed below in Table 1.

The small difference in weights is due primarily to a slight increase in the simulated

luggage weight. This difference in weight should have little or no effect on the impact

response of the fuselage.

 37

Table 1. Weights of Key Items in Test Article and Simulation.

Test Article Simulation

Item Weight in lbs Weight in lbs

Fuselage Section 1360 1355

Stowed Luggage 3229 3300
Passengers, Seats, and Misc. 3550 3596
Overhead Bins and Luggage 433 433
Cameras and Mounts 298 290
Total Weight 8870 8974

4.2. Overhead Stowage Bins and Supports

For the Heath Tecna stowage bin, the struts, the longitudinal channel, and the two

legs of each of the five L Brackets are modeled with beam elements, seen in Figures 16

and 17. The remainder of the Heath Tecna stowage bin is modeled with shell elements.

Figure 16. Heath Tecna Bin and Supporting Structure – View Looking Outboard.
 38

Figure 17. Heath Tecna Bin and Supporting Structure – View Looking Inboard.

For the Hitco bin, beam elements are used for the tie rods, the cylindrical rods, the

short beams, the vertical links, and the horizontal links. Shell elements are used for the

bin itself. Modeling details for the Hitco bin and its supporting structure are shown in

Figures 18 and 19.

Figure 18. Hitco Bin and Supporting Structure – View Looking Outboard.
 39

Figure 19. Hitco Bin and Supporting Structure – View Looking Inboard.

4.3. Luggage Model

The under-floor luggage is a highly significant component of the model because it

directly affects the extent of crushing during impact. However, limited data is available

on the material properties for luggage, leading to difficulties in accurately modeling its

crushing characteristics.

The only known data on the load deflection characteristics of loaded luggage was

obtained by Jackson and Fasanella [24]. Using generic randomly loaded luggage, they

experimentally obtained data on the modulus of luggage under compressive loading.

This data was then translated into a representation of stiffness as a function of volumetric

strain, and is shown as the solid line in Figure 20. The remaining curves in this figure

represent luggage properties that were studied in the current simulation.

 40

Figure 20. Luggage Stiffness as a Function of Volumetric Strain.

This experimental curve was not adopted for modeling the luggage in this study.

Unfortunately, the experimental data ends at a volumetric strain of 0.35, while both

simulation and experimental results indicate that luggage experiences a volumetric strain

of 0.4 to 0.6, with some elements in the simulation reaching strain values of up to 0.7.

Also, it must be noted that the experimental results drop off significantly at the final

plotted value of 0.35, showing highly non-linear behavior near that point.

A further reason for not adopting the experimental curve is that the luggage was

arranged differently in the experimental drop test and the experimental load deflection

test. While the luggage was packed in an effort to fill the under-floor luggage

compartment of the drop test article, significant gaps still appear. The luggage used in the

load deflection test was packed together without gaps. The degree to which luggage was
 41

packed in the under-floor compartment is indicated in Figure 21. There is evidence of

space between the upper pieces of luggage and the floor section. In the finite element

model the luggage elements extend completely to the underside of the floor.

Figure 21. View of Luggage in the Under-Floor Compartment.

The issue of potential gaps between pieces of luggage is significant. If, for

example, 20% of the under-floor volume is indeed empty due to the manner in which

luggage is packed, then the experimental curve in Figure 13 would be shifted over to the

right by approximately 20%. This shift would place the experimental data well within

the range of values studied in the simulation.

The under-floor luggage in the present study was modeled as viscous foam. This

model includes nonlinear elastic stiffness coupled with a viscous damper. This model
 42

was used to represent the stiffness of the luggage as it increased during crushing, and to

simulate energy dissipation.

The viscous foam equations that were used are [8]:

Et = E0(V-n1)

vt = v0(1-V)n2 where

Et is the current value of E,

vt is the current value of the Viscosity Coefficient,
E0 is the initial value for E (Young’s Modulus),
v0 is the initial Viscosity Coefficient,
V is relative volume, defined as ratio of current to initial volume,
n1 is the exponent in power law for Young’s Modulus, and
n2 is the exponent in power law for Viscosity.

The initial elastic stiffness, E0, influences total deformation while the viscosity, v0,

serves to absorb energy. The user specifies E0 and v0, the initial values for E and v. The

user also specifies values for n1 and n2, which determine the rate at which E and v will

change. Et and vt are calculated based on the their specified initial values and the given

exponents, and on the current relative volume, V. This calculation is performed at each

time step during the solution. As V decreases as a result of crushing both Et and vt

increase. Thus both stiffness and viscosity increase as the relative volume decreases

during impact.

Several different initial values and exponents were tested through simulation. A

comprehensive study of a range of values was conducted to compare the final vertical

crushing distance with that observed experimentally. The parameters used in modeling

the luggage were determined from this study and are listed below in Table 2.
 43

Table 2. Luggage Material Properties.

Variable Value in Simulation

E0 Initial Modulus 0.6 psi
v0 Initial Viscosity Coefficient 0.218
n1 Exponent for Modulus 4.0
n2 Exponent for Viscosity 0.2

Poisson’s Ratio was selected as 0.001 in order to approximate the expected

behavior of luggage during crushing. No data is available on this value for luggage and it

was assumed that no lateral expansion would occur during crushing. Testing would be

required to determine if this assumption is accurate.

4.4. Simulation

The simulation was performed with LS-DYNA and results were viewed in LS-

POST. LS-DYNA is an explicit non-linear finite element code developed specifically for

modeling impact and contact events. This code has been used extensively in the

automotive industry for crashworthiness studies resulting in a substantial body of

knowledge of modeling techniques for crash and impact events.

One of the key parameters in performing the explicit time integration for the

transient responses of a dynamic equation is the value of the integration time step, ∆t .

Based on the Courant-Friedrichs-Lewy condition [8], convergence of the solution can be

achieved if ∆t is set to be smaller than the time required for an acoustic wave to travel

through the smallest element in the mesh, i.e.,

∆t ≤ Lc C
 44

where Lc is the shortest dimension of the smallest element in the mesh and C = E ρ .

C is the speed of the acoustic wave and ρ is the mass density.

The initial value of ∆t can be specified based on the length of the smallest side of

any element. As the simulation proceeds, the value of ∆t will be adjusted and updated

based on the dimensions of the deformed elements. For this simulation, the initial ∆t

used was 4 microseconds (µs), which was adjusted by the program throughout the

analysis. The final time step was 1 µs as the result of extensive element distortion. The

simulations were carried out up to 500 milliseconds (ms), approximately the same

duration as the impact event during the actual drop test.

The initial velocity used in the simulation was 30 ft/sec, which is the velocity

measured during the impact test. The fuselage section was placed 0.1 inches above the

impact surface at time zero because the contact algorithm in the finite element code does

not allow for the impacting surfaces to be in contact at time zero. Consequently, the

actual impact between the fuselage section and the impact surface occurred at

approximately 0.28 ms. The effect of this 0.28 ms time delay is quite small in

comparison with the duration of the impact event, and it was therefore assumed that

contact in the simulation begins with t = 0.

In the experimental drop test the test article was dropped onto a wooden platform.

In the simulation, the model was dropped onto a rigid floor to simplify the solution

process. While small errors may occur in local stress distributions as a result of this

approximation, its effect on the overall response and the acceleration time histories of

primary structure should be negligible. This issue was studied by Jackson and Fasanella
 45

[24], who concluded that use of a rigid impact surface did not noticeably alter the

resulting impact response of the test article.

A single sliding surface was used between the outer skin of fuselage model and

the rigid impact surface. A boxed area was also selected around the lower fuselage where

all included nodes were prevented from penetrating the impact surface. It was found that

no observable frame to frame or frame to skin penetrations occurred, so that no interior

slidelines were needed. The structures in the model did not move through each other

during buckling and collapse.

With the use of a single sliding surface, the final version of the simulation ran in

12 hours on a Silicon Graphics Octane with dual processors. The final version included

luggage. Without luggage, the simulation ran in approximately 2 – 3 hours. If interior

sliding surfaces were required, the simulation would have taken much longer to perform.
 46

Chapter 5: Experimental and Simulation Impact Response

The impact response of the test article is assessed through examining

accelerations, loads, and patterns of deformation though the duration of the impact event.

Simulation results are first presented showing the duration of the impact event, and the

manner in which impact energy is dissipated. A detailed examination of the frames as

they deform under impact is then presented. This is done to explain the acceleration and

load histories in terms of the underlying pattern of buckling and crushing that occurs in

the lower fuselage. The results for acceleration response and loading in the overhead

bins and supports are then shown, comparing experimental and simulation results. These

results are presented for each overhead stowage bin separately. Finally, acceleration

response data is presented for other sensor locations including seat tracks, side walls, and

seats/ATDs, again showing both the experimental and simulation results.

The overall comparison of experimental and simulation results will serve to

establish confidence in the fidelity and predictive capability of the simulation, while also

addressing potential limitations.

5.1. Energy Dissipation During the Impact Event

The fuselage test section consists of several substructures, including the fuselage

skin, frames, floor, under-floor beams, overhead stowage bins, and seats, as well as the

luggage stowed in the cargo compartment. During the drop test each component absorbs

a portion of the kinetic energy and converts it to internal energy of deformation. While

this is commonly referred to as energy absorption it is more accurate to note that these

structures dissipate the majority of kinetic energy through heat and sound during plastic

deformation. Only a small portion of this energy is stored internally.

 47

By analyzing the percentage of kinetic energy that is dissipated by individual

structural components it is possible to determine which components play the most

significant role in absorbing and dissipating energy during the impact event. This is a

key advantage of numerical methods, in that it is possible to compare energy dissipation

of different structures. This permits the designer to focus quickly on the most significant

structures for energy absorption and to assess the effectiveness of multiple design options.

The conversion of kinetic energy into internal energy, as calculated by the

simulation, is shown in Figure 22. It can be seen that during the first two milliseconds

(ms) no energy conversion takes place, indicating that the structure is still deforming

elastically. In fact, the total kinetic energy increases slightly, as the structure continues to

accelerate under the effect of gravity. In other words, during the first two milliseconds

the downward velocity of the test section is not affected by impact. Once the energy

conversion begins to take place at approximately 3.0 ms, the total kinetic energy begins

to decrease as the result of energy dissipation through structural damping and plastic

deformation but plateaus to a value near zero at approximately 125 ms. At this moment

more than 95% of the impact energy has been converted into internal energy. Thus, all

major deformation and energy dissipation is completed within the first 125 ms of the

impact event. Therefore, even though the finite element simulation has been carried out

up to 500 ms, approximately the same duration as the actual drop test, most of the results

to be presented herein will be up to 250 ms only. In the period between 125 and 250 ms

the fuselage does continue to respond elastically.

 48

Figure 22. Dissipation of Kinetic Energy During Impact.

The amount of internal energy absorbed by the frames and luggage as compared

to the total internal energy is also shown in Figure 22. The results indicate that the

frames dissipate approximately 33% of the total internal energy. The frames are the

single most important structural component in terms of energy dissipation. Other

structures, such as underfloor beams and skin panels, make only minor contributions to

energy dissipation. The frames, therefore, have a dominant effect on the overall dynamic

response of the fuselage section, and a careful examination of the deformation histories of

the frames during the impact event can provide vital information for a better

understanding of how other structural components respond. The simulation shows that

an additional 35% of the impact energy is absorbed by the luggage. That is, the luggage

and frames together dissipate 68% of the total energy during impact. This indicates that
 49

the luggage is also an extremely important factor in the energy absorption process. It will

be shown later that an accurate modeling of the behavior of luggage during impact is

vitally important in obtaining reasonable simulation results.

5.2. Deformation History of Fuselage Frames in Simulation

The recorded acceleration responses are a result primarily of the sequence of

structural loading, buckling, and crushing that occurs in the lower fuselage. It is

therefore necessary to study the sequence of deformation in detail in order to understand

the dynamic response of the structural components that will be presented in the following

sections. This sequence of deformation also indicates how structural collapse and

buckling could be better controlled to mitigate high acceleration peaks in the seats and

overhead stowage bins.

The deformation histories of the overall frames and the cargo doorframe region

that were obtained from the finite element simulation are shown at selected time steps up

to 140 ms in Figure 23. This figure shows the overall deformed test article and the

contour of effective plastic strain. The chosen time steps were selected to coincide with

key events that occur to the frames during impact. It should be noted that the skin and

other structural components, as well as luggage, have been removed from the figure for

clarity.

No plastic deformation was observed during the first 2.5 ms of impact, which

agrees with the results for energy conversion shown in the previous section. At t = 3 ms,

the bottom of the frame, which comes in contact with the rigid floor at t = 0, begins to

deform plastically, as can be seen in Figure 23a.

 50

Right-Hand Side

Figure 23a. Two Views of Effective Plastic Strain in Frames, Time = 3 ms

 51

Right-Hand Side

Figure 23b. Two Views of Effective Plastic Strain in Frames, Time = 7 ms

 52

Right-Hand Side

Figure 23c. Two Views of Effective Plastic Strain in Frames, Time = 13 ms

 53

Right-Hand Side

Figure 23d. Two Views of Effective Plastic Strain in Frames, Time = 20 ms

 54

Right-Hand Side

Figure 23e. Two Views of Effective Plastic Strain in Frames, Time = 30 ms

 55

Right-Hand Side

Figure 23f. Two Views of Effective Plastic Strain in Frames, Time = 40 ms

 56

Right-Hand Side

Figure 23g. Two Views of Effective Plastic Strain in Frames, Time = 50 ms

 57

Right-Hand Side

Figure 23h. Two Views of Effective Plastic Strain in Frames, Time = 60 ms

 58

Right-Hand Side

Figure 23i. Two Views of Effective Plastic Strain in Frames, Time = 70 ms

 59

Right-Hand Side

Figure 23j. Two Views of Effective Plastic Strain in Frames, Time = 80 ms

 60

Right-Hand Side

Figure 23k. Two Views of Effective Plastic Strain in Frames, Time = 90 ms

 61

Right-Hand Side

Figure 23l. Two Views of Effective Plastic Strain in Frames, Time = 100 ms
 62

Right-Hand Side

Figure 23m. Two Views of Effective Plastic Strain in Frames, Time = 120 ms
 63

Right-Hand Side

Figure 23n. Two Views of Effective Plastic Strain in Frames, Time = 140 ms
 64

The plastic zone propagates upward along the frame as the impact event

progresses. At t = 7 ms, the flanges of the lower frame section begin to yield due to the

excessive deformation of the frames, as shown in Figure 23b.

At t = 13 ms, the lower left and right corners (at approximately 7 o’clock and 5

o’clock locations, respectively) of all frames, except the cargo doorframe, begin to yield.

This is seen in Figure 23c. As the impact event continues, the bottom of the frame

gradually flattens. At t = 20 ms, Figure 23d, the flanges in areas with high plastic

deformation begin to exhibit local buckling. Buckling begins, for example, in the

forward and aft frames on the right side. Throughout this segment of the impact event the

cargo doorframe still exhibits very little deformation.

At t = 30 ms, Figure 23e, the lower left corners of all the frames form plastic

hinges. Plastic deformation also occurs on the left-hand side just below the window

openings, and in the frames at FS 380 and FS 500 where the reinforcing under-floor

beams are attached. On the right-hand side, because of the stiff cargo doorframe, only

the frames at FS 380 and FS 500 are forming plastic hinges. At this moment (t = 30 ms),

even though the entire fuselage continues to move downward, impact energy is absorbed

primarily by the plastic hinges which formed in the lower left-hand corner of each frame.

Significantly less energy is dissipated by the plastic deformation and local buckling in the

lower right-hand corners. As a result, the entire fuselage section is seen to slightly tilt to

the left. The upper portion of the structure experiences limited plastic deformation, as

compared to the lower portion.

At t = 40 ms, kinks have begun to form in the lower left corner of each frame, as

seen in Figure 23f. These kinks appear at those sites where plastic hinges have
 65

previously formed. The plastic hinge on the right-hand side of the frame at FS 380

contains significantly less plastic deformation in comparison to that on the left-hand side.

This is because the right-hand side cargo door frames limits buckling on that side.

Meanwhile, plastic deformation has developed quite extensively in the lower right-hand

side frame at FS 500, yet the stiff aft doorframe continues to show very little deformation.

The short reinforcement beams connecting the aft doorframe to the frame at FS 480

provide a strong lateral support for the aft doorframe making it the stiffest section of the

entire lower fuselage structure. In fact, the aft doorframe deforms relatively little

throughout the entire impact event. The aft doorframe provides a stiff load path for the

transmission of impulse loading during impact. This leads to significant shear force

being exerted on the upper doorframe between FS 460 and FS 480. This in turn causes

plastic deformation to develop, as will be shown at later times. Plastic deformation is

also observed at the upper right-hand corners of the frames where outboard supports of

the Heath Tecna bin are attached. The upper left-hand side frames show plastic

deformation much later in the impact event.

From t = 50 to t = 70 ms, Figures 23g to 23i, the lower frames on the left-hand

side continue to crush as the plastic hinges collapse. By t = 60 ms, however, the lower

right-hand frames have reached the point of largest buckling. Further buckling is limited

by the relatively stiff cargo door frame, particularly the frame surrounding the aft cargo

door edge.

Around t = 80 ms, Figure 23j, the plastic hinges on the left-hand side begin to hit

the rigid floor, crushing the frames and setting off a second left side impact. Plastic

deformation is also observed on the left-hand side of each frame where the under-floor
 66

beams are joined. This deformation is not nearly as significant on the right-hand side,

due to the cargo doorframe reinforcement described previously. Plastic deformation of

the two frames at FS 380 and FS 500 is largely contained in the areas where the

reinforcing under-floor beams meet the frames. In addition, plastic deformation is

observed in upper frames at FS 480 and FS 500, where supports of the aft camera mount

are attached.

Between t = 90 ms and t = 120 ms, Figures 23k through 23m, the lower left-hand

frames continue to crush. Additional impact energy is dissipated as these frames deform

plastically. The lower right-hand frames are no longer deforming to any significant

degree, so that the test section continues to tilt downward to the left side. The upper

portions of the frames show less deformation, though the areas where plastic deformation

has already developed still continues to grow.

Deformation at t = 140 ms is shown in Figure 23n. By this point in time, the

maximum dynamic deformation has been reached and the left-hand side of the test

section begins to rebound from the impact surface. This rebound is more noticeable on

the left side and it causes the structure to rotate slightly from left to right. The rebound is

also more pronounced at the front of the test section than at the aft end. This also agrees

with the high speed record of the drop test taken during the experimental impact. As

previously noted, the cargo door is located closer to the aft end of the fuselage test

section, resulting in an asymmetrical impact response between the forward and aft ends

of the fuselage section.

Simulation results show that the initial impact event is over at 130 ms, at which

time the floor reaches zero velocity. The timing of the impact duration is also shown in
 67

the history of energy dissipation where energy dissipation has been largely completed by

t = 130 ms (Figure 21). This point in time indicates the end of the plastic response phase

of the impact event. After this time, however, the upper frames continue to deform

elastically. They will reverberate before coming to rest in the final deformed shape.

The effect of the stiff cargo doorframe can be seen again in Figure 23n, as the

lower right-hand side of the frame shows less deformation in comparison to the lower left

side, where plastic hinges formed. It is noted that the upper cargo doorframe, between FS

460 and FS 480, has buckled due to the shear force exerted on it by the stiff aft doorframe.

The buckling of this aft upper doorframe acts as a buffer that slows and dissipates the

force transmitted to the upper frames. In comparison, the forward doorframe is directly

aligned with the frame at FS 420, allowing a direct transmission of the impact force to the

upper frames. The difference in force transmission is important because it accounts for

the difference in force and acceleration time histories of the forward and aft end of the

Heath Tecna bin, which is located on the right-hand side.

Timing of the simulated impact event agrees with experimental data, which shows

that the floor reaches zero velocity around t = 130 ms, Figure 24. The agreement

between simulation and experiment on the timing of this initial phase of the impact

response is a strong indication that the simulation has captured the essential features of

the impact event.

 68

10
100Hz
20Hz
5

-5
Velocity (ft/sec)

-10

-15

-20

-25

-30

-35
-50 0 50 100 150 200 250
Time (msec)

Figure 24. Typical Seat Track Vertical Velocity Change – FAA Data

Table 3 provides a summary of the significant events during the

simulation. Most noteworthy is that the duration of the crushing phase (plastic

deformation) in the simulation coincides with that observed during the experimental

impact event.
 69

Table 3. Summary of Significant Events During Simulated Impact

Time Event in Simulation

3 ms Plastic deformation initiates.
7 ms Flanges of lower frames begin to yield.
13 ms Corners of lower frames begin to yield.
20 ms Plastic hinge develops on left side.
This results from initial impact, which sets off first
acceleration pulse.
30 ms Lower corners of all left side frames form plastic hinges.
Only corners of forward and aft frames on right side form
plastic hinges.
60 ms Point of maximum buckling reached on right side, setting off
second right side acceleration pulse.
Further right side buckling is limited by cargo door frame.
Fuselage now begins to tilt to left.
80 ms Plastic hinge on left side impacts ground, setting of second
left side acceleration pulse.
90 to 130 ms Fuselage continues crushing, primarily on left side.
130 to Point of maximum dynamic deformation reached.
140 ms Seat tracks reach zero velocity.
Point of maximum buckling of left side frames is reached,
setting off third, smaller, left side acceleration pulse.
Slight rebound begins.
280 ms Point of maximum rebound.
400 ms Section settles down again on simulated platform.
Some reverberation still continues in upper fuselage.

5.3. Comparison of Deformation between Simulation and Experiment

Comparisons are presented between the deformed configurations of the fuselage

section recorded after the experimental drop test and deformations obtained from the

simulation at t = 150 ms. Six different views of the comparison are shown in Figures 25

through 30. It is noted from the drop test results that all seats on the right-hand side have

failed, while all those on the left-hand side have survived. Both of the overhead stowage

bins survived the impact and remained attached to their supports. The under-floor
 70

luggage compartment and luggage has been crushed extensively indicating that the

luggage must have absorbed a considerable amount of impact energy. The asymmetrical

deformation resulting from the existence of the very stiff cargo doorframe on the right

side is clearly seen from both the front and back views. The front view shows more

extensive crushing on the left-hand side and the fuselage tilts down on that side as a result.

The side view, Figure 27, shows that the aft end of the fuselage structure crushes more

than the forward end. This pattern is found in both the simulation and experimental test

results, and is due to the asymmetrical position of the cargo door and cargo door

reinforcing structures.

Figure 25. Front View of Comparison of the Deformed Configuration

 71

Figure 26. Aft View of Comparison of the Deformed Configuration

Figure 27. Left View of Comparison of Deformed Configuration

Figure 28. Right View of Comparison of Deformed Configuration

 72

Figure 29. Front Right Side Angled View, Test and Simulation

Figure 30. Front Left Side Angled View, Test and Simulation
 73

Figure 31. Typical Vertical Displacement of Floor During Impact

Qualitatively, the deformation predicted by the simulation compares very well

with that recorded during the actual drop test. The quantitative difference in right-hand

and left-hand side deformation is illustrated in Figure 31, which shows typical

deformation results from the simulation. Points on the floor on the far left and right side

of the airframe were chosen to illustrate the difference in vertical displacement that

occurs.

The point of maximum deformation on the left side is reached around t = 140 ms

with a slight rebound beginning near t = 160 ms, as seen in Figure 31. As previously

noted, crushing on the right-hand side is limited by the reinforced cargo door frame,

though the right-hand side continues to deform slightly up to 200 ms.

Table 4 compares simulated values for vertical displacement with experimental

values measured by the FAA [37]. The dynamic value is taken at the point of maximum

deformation during impact, while the static value represents a post-test measurement. All

measurements represent the vertical distance traveled by a reference point on the right-
 74

hand and left-hand side of the floor of the fuselage section. The experimental static

measurement was performed physically after the impact event was over. The

experimental dynamic measurements were scaled from photographs taken during impact.

The simulation results are for vertical displacement of a node on the floor, selected to

represent the same point used in making experimental measurements.

Table 4. Comparison of Vertical Displacement

Measurement Type Front Left Front Right Aft Left Aft Right
Experimental –
Static Deformation 23 15.8 27.5 18.4
In Inches
Experimental–
Dynamic Deformation 28 21 30.3 24
In Inches
Simulation –
Dynamic Deformation 27.5 16.1 29.2 18
In Inches

The comparison of simulation and experimental displacement results in Table 4

shows very good agreement on the left-hand side, and only moderate agreement on the

right-hand side. This may be due to the cargo door structure in the simulation being

somewhat stiffer than that in the experimental test article. This would limit crushing and

lead to less overall vertical displacement on that side in the simulation. Note that the

simulation does capture the slight but distinct trend in which vertical crushing is greater

at the aft locations, in comparison to the forward locations.

 75

5.4. Comparison of Impact Response between Experiment and Simulation

The impact response of the test article is determined primarily by the acceleration

time histories as measured through accelerometers, and by load time histories recorded

through calibrated strain gages. Validation of the simulation requires a close comparison

of experimental and simulation time histories for all available data. Accordingly,

acceleration time histories are presented below for selected points on the overhead

stowage bins, seat tracks, and frames, where all selected points correspond to

experimental sensor locations. For the overhead stowage bins, load time histories of the

primary supporting structures are also presented. Before presenting these data, the

filtering method used in obtaining the results will be discussed.

5.4.1. Filtering Method used with Experimental and Simulation Data

Unfiltered data from accelerometers is difficult to interpret, as it contains

numerous frequency responses superimposed upon a fundamental pulse. The overall

structure has several components, and each will have its own mode of oscillation along

with multiple harmonics. In crash impact analysis, the primary concern is with the

fundamental mode, which must be extracted from data containing high frequency ringing.

Methods of filtering the basic, or raw, data have been developed for use in

analyzing automotive crash tests. However, the appropriate filter depends upon the

structural stiffness and duration of the impact event. A very short duration impact event

would require a higher frequency filter, while events of longer pulse duration will require

a lower frequency filter. For example, if a low frequency filter is applied to a very short

duration event, it will essentially average out the impulse over too long a period, and

therefore fail to capture the peak acceleration value. The opposite problem occurs with
 76

the current study. When a high frequency filter is applied to a longer duration event it

will leave high frequency ringing in the results, even after filtering. Such results will

therefore overstate the magnitude of the peak value, and understate the pulse duration.

A standard algorithm for filtering raw acceleration data is presented in SAE

J211/1 [24]. The lowest frequency filter in this standard is a 100 Hz (Channel Frequency

Class 60) filter. However, analysis of the experimental data revealed that the 100 Hz

filter still employs too high a filtering frequency to extract the correct peak and pulse

duration of the fundamental acceleration response. An investigation was therefore

undertaken by the FAA [24, 37] in which a range of filter frequencies between 5 Hz and

100 Hz were evaluated. The FAA determined that the 20 Hz filter leads to results that

accurately show the duration and magnitude of the fundamental acceleration pulse, and

that the 20 Hz filter does not distort the resulting data. One method of checking the

validity of filtered results is to integrate the acceleration results to find the velocity time

history. The resulting plot of velocity should show the correct change in velocity over

time, as long as the filter has not altered the underlying acceleration pulse.
 77

50
100Hz
20Hz

30

10
Acceleration (g)

-10

-30

-50

-70
-50 0 50 100 150 200 250
Time (msec)

Figure 32. Effect of Filter Frequency on Typical Experimental Acceleration Data

The effect of different filtering frequencies on acceleration data is shown in

Figure 32, where results for a typical sidewall sensor have been filtered both with a 100

Hz (SAE Class 60) filter and with a 20 Hz filter. The results filtered with the 100 Hz

filter show significant oscillations, making it difficult to determine the pulse duration. In

contrast, the 20 Hz filtered results show a clear pulse from 65 ms to 145 ms. The

magnitude of the acceleration pulse is also more accurately shown by the results using the

20 Hz filter. The 100 Hz results show a peak of around 26 g, while the 20 Hz results

correctly show the magnitude of the fundamental pulse as approximately 18g.

Furthermore, results with the 20 Hz filter correlate well with observable impact events,

such as buckling and collapse of frames.

Simulation results were also analyzed with 20 Hz and 100 Hz filters, to verify that

the 20 Hz filter should be used for simulation data. Figure 33 shows the effect of using
 78

these two different filtering frequencies on the simulation results. As with the

experimental data, this comparison indicates that the 20 Hz filter is better able to capture

the peak acceleration of the fundamental mode, and also shows a more accurate

description of pulse duration. Results with the 100 Hz filter, in contrast, do not show any

clear pulse duration, because there are still high frequency oscillations. For this study, all

experimental and simulation acceleration time histories are filtered with a 20 Hz filter

following the principles of SAE J211/1.

Figure 33. Effect of Filter Frequency on Typical Simulation Acceleration Data

When performing transient dynamic analysis, it is important to use an appropriate

sampling interval to avoid aliasing. Aliasing occurs when the sampling interval is larger

than the frequency response, resulting in energy being transferred incorrectly to other
 79

frequencies. Sampling intervals ranging from 0.25 to 1.0 ms, corresponding to a

frequency range of 4,000 to 1,000 Hz, were checked in calculating accelerations. Results

show very little difference in either the peak accelerations or in the shape of the

acceleration time history curves. In this study, a sampling interval of 0.5 ms,

corresponding to a sampling frequency of 1,000 Hz, was used.

5.4.2. Acceleration Histories of Heath Tecna Bin

A comparison of the experimental and simulation impact response of the

overhead stowage bins will serve to further validate the simulation. If there is a

reasonable correspondence in results it will indicate that the model accurately captures

the essential buckling behavior of the test article during impact.

Simulation results indicate that dissipation of kinetic energy is largely completed

within the first 150 ms of the impact event. This indicates that no additional plastic

deformation occurs beyond this point. However, simulation results also show that the

upper frames continue to deform elastically and reverberate. Since the overhead bins are

attached to the upper portion of the frames, their responses are influenced not only by the

plastic deformation that occurs, but also by the elastic responses of the upper frames. For

this reason, the acceleration and load time histories discussed in this section are all

presented for a duration of 250 ms.

Acceleration time histories for the Heath Tecna bin, installed on the right-hand

side of the fuselage, are presented in Figures 34 through 37. The peak acceleration value

is approximately 13 g’s for the simulation, and approximately 12 g’s for the experimental

data. For the simulation, the timing of the peak acceleration occurs near 50 to 55 ms,

while the experimental data shows the peak occurring between 75 and 110 ms.
 80

The close correspondence of peak acceleration values indicates that the simulation

has accurately captured the fundamental acceleration pulse experienced by the bin during

impact. The mismatch in timing of the peak values indicates that the simulation is

somewhat too stiff on the right-hand side.

Study of the high-speed films that recorded the experimental impact event

indicates that buckling occurs on the right-hand side beginning at approximately 20 ms.

After this point, the structure continues buckling, or crushing, for another 20 to 30 ms,

after which it reaches its point of maximum deformation, effectively increasing its

stiffness. This leads to a second acceleration pulse on the right-hand side, which is

reflected in the experimental data.

Figure 34. Heath Tecna Bin Acceleration, Figure 35. Heath Tecna Bin Acceleration,
Forward Location Aft Location
 81

Figure 36. Heath Tecna Bin Acceleration, Figure 37. Heath Tecna Bin Acceleration,
Center Average

The most likely explanation for this slight shift in phase for the peak acceleration

value concerns the modeling of the cargo door. In the test article, the door is pressed into

place with latches and bolts. In the simulation, the cargo door is fixed to the surrounding

doorframe. Thus the simulation will more quickly transmit impulses from impact up

through the frames. The fixed connection in the simulation could also influence the

timing of buckling in the cargo door reinforcements.

5.4.3. Load Histories for Heath Tecna Overhead Bin Supporting Structures

Load time histories for the supporting structural members of each bin were also

analyzed. As previously noted the Heath Tecna bin is supported by a pair of primary

vertical struts attached to the forward and aft ends of the bin. There is also an outboard

supporting structure that supports inboard-outboard loads and provides secondary vertical

support for the bin. Load time history plots are significant because they can be used to

predict the conditions under which the supporting structure may fail, assuming that a

failure load is known experimentally.

 82

Vertical struts provide the primary vertical support for the Heath Tecna bin. The

load time histories for these supports are shown in Figures 38 and 39. The magnitude of

the simulation loads match very well with the experimental values, though the exact

timing of the peak values shows a slight discrepancy with the experiment.

Figure 38. Heath Tecna Bin Load, Figure 39. Heath Tecna Bin Load,
Forward Strut Aft Strut

The A leg of the L Bracket provides a portion of the vertical support during

impact. These loads are less significant for two reasons. First, as previously noted, the

majority of loading is supported by the vertical struts, so that the magnitude of loading in

the bracket is significantly smaller. Second, testing by the FAA [37] has shown that any

failure in the bin, in which the bin detaches from its supports, will occur first in the struts,

and not in the outboard bracket.

The load histories in the brackets are presented in Figures 40 through 49. The A

leg of the bracket is angled upward, while the B leg only reacts inboard-outboard loads.

In both cases, the simulation load time histories compare reasonably well with those of

the test article. While there are some discrepancies in the distribution of loads to

individual brackets, the overall load carried by the combined brackets is comparable in
 83

both simulation and experimental results. The simulation was not sufficiently detailed in

modeling bracket attachments to accurately capture the distribution of secondary loads

through each bracket.

Figure 40. Heath Tecna Bin Load, Figure 41. Heath Tecna Bin Load,
FS 400 – A Leg of Bracket FS 400 – B Leg of Bracket

Figure 42. Heath Tecna Bin Load, Figure 43. Heath Tecna Bin Load,
FS 420 – A Leg of Bracket FS 420 – B Leg of Bracket
 84

Figure 44. Heath Tecna Bin Load, Figure 45. Heath Tecna Bin Load,
FS 440 – A Leg of Bracket FS 440 – B Leg of Bracket

Figure 46. Heath Tecna Bin Load, Figure 47. Heath Tecna Bin Load,
FS 460 – A Leg of Bracket FS 460 – B Leg of Bracket

Figure 48. Heath Tecna Bin Load, Figure 49. Heath Tecna Bin Load,
FS 480 – A Leg of Bracket FS 480 – B Leg of Bracket
 85

The FAA drop test was intended in part to address the issue as to whether static

testing is comparable to dynamic testing for overhead stowage bins. This issue can be

answered by determining the percentage of total vertical load carried by the struts at the

point in time when peak loading occurs. This calculation was performed by the FAA

using experimental data, and the result was termed the dynamic influence coefficient for

the vertical struts. The influence coefficient is therefore the percentage of total vertical

load from the overhead bin that is supported by the two struts.

The influence coefficient is significant because static testing of each overhead

stowage bin results in a load distribution that is constant between all supporting structures,

so that the struts always carry the same percentage of the total load. Under conditions of

dynamic impact loading, however, the percentage of vertical load carried by each

structural support will vary.

Under conditions of static testing, the influence coefficient is 66%, meaning that

66% of the vertical load is supported by the two struts. The influence coefficient under

dynamic loading conditions is most significant at times when peak vertical loading occurs.

Two pulses were observed in the experimental data with peaks occurring at 13 and 102

ms. The first peak load, at 13 ms, is very early in the impact event, and the magnitude of

this peak is relatively small. The second peak load, at 102 ms, represents the point at

which the vertical load reaches its maximum value. At 102 ms the experimental

influence coefficient is 93%, as shown in Figure 50.

 86

Figure 50. Influence Coefficient for Vertical Struts of Heath Tecna Bin

The dynamic influence coefficient was calculated using simulation results and is

also shown in Figure 50. At the most critical point in time, the point at which peak

loading occurs, the simulation calculates an influence coefficient of 0.88, which is very

close to the experimental value. In general, the dynamic influence coefficient based on

simulation results fluctuates around 75 percent. A comparison of dynamic and static

influence coefficients shows that impact loading condition leads to a noticeable change in

the load distribution throughout the supporting structures of this overhead bin, and that

this load distribution is not accurately represented with static testing. More generally,

these results show that the simulation can be used to track the dynamic influence

coefficient with reasonable accuracy.

 87

5.4.4. Acceleration Histories of Hitco Bin

Acceleration time histories for the Hitco overhead stowage bin, installed on the

left-hand side of the fuselage, are presented in Figures 51 through 54. The peak

acceleration value is approximately 10 g’s for the simulation, and approximately 11 g’s

for the experimental data. For both the simulation and experiment, the timing of the peak

acceleration occurs near 120 ms. The close correspondence of peak acceleration values

and pulse duration indicates that the simulation has accurately captured the fundamental

acceleration pulse experienced by the bin during impact.

Figure 51. Hitco Bin Acceleration, Figure 52. Hitco Bin Acceleration,
Forward Location Aft Location

Figure 53. Hitco Bin Acceleration, Figure 54. Hitco Bin Acceleration,
Center Location Average
 88

A comparison of the time at which the peak acceleration pulse occurs indicates

that the simulation closely matches the experimental results. In both cases, the peak

acceleration pulse occurs around 100 to 150 ms. This is the primary acceleration pulse,

and is the most important portion of the impact response. The magnitude of the smaller

initial acceleration pulse in the simulation shows a stronger than expected pulse near 15

ms, prior to the initiation of buckling on the left side. Because the later acceleration time

history matches very well with experimental data, it is likely that the overall structural

modeling possesses the correct stiffness. Possible reasons for this slight initial

discrepancy, such as the omission of rivets in the current model, will be examined in

Chapter 8.

These time histories also show that the aft end of the stowage bin experiences a

higher peak acceleration pulse than the forward end. This response is captured in both

the experimental and simulated results. The difference in peak acceleration between the

two ends of the overhead stowage bin is due to asymmetry in the test article introduced

by the location and design of the cargo door, as previously noted. The cargo door is

offset toward the aft end of the fuselage, and is responsible for this difference in response

in the forward and aft ends of the bin. Thus even the acceleration responses on the left-

hand side of the airframe are affected by the right-hand side cargo door reinforcement.

The timing of the acceleration pulse is due to the pattern of deformation and

buckling that occurs in the left-hand side of the airframe. The history of deformation

shown previously indicates the formation of a plastic hinge in the lower left-hand side

frames. The impact leading to the formation of the hinge results in the first, lower
 89

acceleration pulse. The period during which crushing occurs is reflected in the decrease

in acceleration prior to the primary pulse. When this plastic hinge impacts the platform,

it results in the longer primary acceleration pulse, followed by additional buckling, and

finally stabilization.

The simulation therefore provides a clear means of understanding, as well as

calculating, the acceleration pulses which are experienced in the overhead stowage bins.

Such an understanding is essential for future studies of methods to control the dissipation

of impact energy.

5.4.5. Load Histories for Hitco Overhead Bin Supporting Structures

Load time histories of the primary vertical supporting members of the Hitco bin

are presented in Figures 55 and 56. The secondary supports are vertical and horizontal

links, Figures 57 through 64. Peak load values match very closely with the experimental

data, again with the exception of a higher than expected initial peak near 20 ms. A

comparison of pulse duration and magnitude between simulation and experiment shows

that the simulation has captured the basic structural response of the Hitco bin.

Figure 55. Hitco Bin Load, Figure 56. Hitco Bin Load,
Forward Tie Rod Aft Tie Rod
 90

Figure 57. Hitco Bin Load, Figure 58. Hitco Bin Load,
FS 400 Vertical Link FS 400 Horizontal Link

Figure 59. Hitco Bin Load, Figure 60. Hitco Bin Load,
FS 420 Vertical Link FS 420 Horizontal Link

Figure 61. Hitco Bin Load, Figure 62. Hitco Bin Load,
FS 460 Vertical Link FS 460 Horizontal Link
 91

Figure 63. Hitco Bin Load, Figure 64. Hitco Bin Load,
FS 480 Vertical Link FS 480 Horizontal Link

As with the struts of the Heath Tecna bin, it is possible to calculate an influence

coefficient for the tie rods of the Hitco bin. Again, the influence coefficient is the

percentage of total vertical load that is carried by the two primary vertical supports of the

Hitco bin. These supports are the tie rods that attach toward the inboard side of the Hitco

bin. Figure 65 shows a plot of the dynamic influence coefficient based on simulation

results. The experimental dynamic influence coefficient is shown at two different points

in time, 14 ms and 114 ms. These two experimental points were selected because they

are the times at which peak experimental loading occurs in the Hitco bin, as seen

previously in Figures 55 and 56. The point at 114 ms represents the time at which the

primary loading pulse occurs, and is therefore the more important of the two points. The

earlier time, 14 ms, represents the time of peak loading of a much smaller pulse.
 92

Figure 65. Influence Coefficient for Combined Vertical Tie Rods of Hitco Bin

Static testing shows that the vertical tie rods carry 66.5% of the total vertical load

in the Hitco bin, which is shown as the static influence coefficient. The remainder of the

load is carried by the vertical links attached to the outboard edge of the bin.

Dynamic testing shows that the influence coefficient for the tie rods is 56% at the

most critical point in time, 114 ms. A comparison of the experimental influence

coefficient at 114 ms with that obtained through the simulation shows that the simulation

accurately represents the load distribution between the tie rods and links during the most

critical period of impact. Peak loading in the simulation occurs between 100 and 120 ms,

similar to the experimental results, so the dynamic influence coefficient is of most

significance during this period. In general, the dynamic influence coefficient based on

simulation results fluctuates around the experimental static value. The only exception to
 93

this is the period from 50 to 70 ms, which occurs when the overhead stowage bin is very

lightly loaded, and the values during this period are therefore not significant.

Thus the results seen in Figure 64 show first that the static and dynamic influence

coefficients are similar for this bin configuration, and second that the simulation has

accurately modeled the influence coefficient of the Hitco overhead stowage bin.

5.4.6. Acceleration Time Histories of Seat Tracks

The acceleration response of the seat tracks provides an indication of the force

which will be transmitted to the seats and occupants during impact. The acceleration

response results from a complex interaction between the lower frames, the under-floor

beams, the tracks attached to those beams, the seats, and the occupants.

In the test article, the seat cushions absorb energy, and also serve to delay the time

at which inertial forces are transferred from the occupants to the seat frames. Further, the

seat frames undergo plastic deformation, also affecting the forces transferred to the seat

tracks. It was observed that seats on the right-hand side of the test article failed during

impact, and this event significantly alters both the magnitude and timing of the forces

transferred to the seat track.

In the simulation, the effect of the seat cushions is not included, as it was beyond

the scope of the current study to address energy absorption in the seat structures.

Although seats are included in the simulation, masses for the dummies are distributed on

the seat surface. Structural failure of the seats was not permitted in the simulation, due to

a lack of appropriate data on the failure loads. Thus, the seats respond elastically, even

though permanent plastic deformation would be expected in several locations. This can

be seen in the simulation, as outboard (window) seats deflect significantly. The test
 94

results also show that these seats experience permanent deformation, even on the left-

hand side where the basic seat structure remains intact.

Fourteen locations were selected for calculating acceleration time histories, as

shown in Figure 66. These locations correspond to experimental sensor locations.

Acceleration time history results for these locations are shown in Figures 67 through 77.

Figure 66. Locations of Acceleration Results on Seat Tracks

The most important features of the acceleration time history are illustrated best in

Figures 69 and 70. First, and most significantly, it is seen that the magnitude of the

experimental peak acceleration response is closely matched by the simulation results.

Slight differences in timing of the pulse are due to the above noted simplifications in the

simulated seats and occupant masses.

 95

Figure 67. FS 380 Left Inside Seat Track. Figure 68. FS 380 Right Inside Seat Track.

Figure 69. FS 418 Left Outside Seat Track. Figure 70. FS 418 Right Outside Seat Track.

Second, Figures 69 and 79 show that the magnitude of the right-hand side

acceleration pulse is greater than the left-hand side by 7 or 8 g, as seen in both

experimental and simulation results. This difference is due to the fact that the right-hand

side response is influenced by the relatively stiff cargo door reinforcing structure, which

restricts deformation on that side.

Third, it can be seen from results in Figure 67 through 70 that the right-hand side

acceleration pulse typically shows a sharper, shorter duration impact. This is because the

frames and cargo door on the right-hand side buckle uniformly at one location only, and
 96

do not exhibit the more widespread and gradual crushing seen on the left-hand side. This

causes the right-hand side acceleration pulse to be somewhat shorter than the left.

It should be noted that the outside seat tracks are located closer to the frames, and

therefore more accurately illustrate the effect of the impact being transmitted from the

frames to the under-floor beams. The inside seat tracks are closer to the center of the

under-floor beam, and hence are more influenced by the oscillations of the beam itself, as

well as its interaction with the seats.

It should also be noted that the sensor location at FS 380 is very close to the

forward end of the fuselage section, and may be influenced by the second reinforcing

under-floor beam at that location. This means that both the sensor and simulation results

for that location may not be typical of the fuselage as a whole.

Figure 71. FS 418 Left Inside Seat Track Figure 72. FS 418 Right Inside Seat Track
 97

Figure 73. FS 452 Left Outside Seat Track Figure 74. FS 452 Right Outside Seat Track

Figure 75. FS 452 Left Inside Seat Track Figure 76. FS 452 Right Inside Seat Track

Figure 77. FS 484 Left Outside Seat Track Figure 78. FS 484 Right Outside Seat Track
 98

Figure 79. FS 484 Left Inside Seat Track Figure 80. FS 484 Right Inside Seat Track

5.4.7. Acceleration Histories of Frames

Acceleration time histories of the frames provide an indication of the acceleration

pulse as it moves from the point of contact toward the overhead stowage bins. The

acceleration response was recorded experimentally at six locations on the upper frames,

three on each side, as shown in Figure 81. Similarly, results were obtained at six

locations on the lower side walls, as shown in Figure 82. The lower side wall nodes are

located 12 inches above the floor, in the center of the frame.

 99

Figure 81. Locations for Upper Sidewall Acceleration Results

Figure 82. Locations for Lower Sidewall Acceleration Results

 100

A comparison of experimental and simulation acceleration time histories for the

upper side walls are shown in Figures 83 through 88. Peak acceleration values for the

upper side walls show a close correlation between experimental and simulation results,

though there is a slight shift in phase. Results for the lower side wall show a moderate

correlation between experimental and simulation acceleration time histories. The

discrepancies may be due to local effects in the model, as the experimental sensor is

located very close to the frame and under-floor beam joint. This joint is not modeled in

detail in the finite element simulation. Results for the upper frame locations show a

better comparison between simulation and experimental results than the lower frame

locations. The upper frame experimental sensors are far away from the frame and under-

floor beam joint, and the upper frame simulation results are less influenced by local

effects near the floor level.

It should be noted that the right-hand side peak acceleration values are nearly

identical for both the lower and upper side walls. This indicates very little plastic

deformation or buckling occurs in the right side frames, as the acceleration pulse is

transmitted directly through the frames. This observation is supported by a post-test

examination of the test article, which indicates only minor buckling of the inside portion

of the frames on the right-hand side.

In contrast, the left-hand side results for both test and simulation show that the

peak acceleration value in the upper frame is approximately 25 to 30% lower than that in

the lower frame. This would indicate that substantial plastic deformation, or failure,

occurred in the left side frames. Again, this is supported by a post-test examination,

which indicates substantial failures in each of the frames on the left side. Although the
 101

magnitude of this effect is overestimated in the simulation, it nevertheless agrees with the

basic trend, showing more significant plastic deformation on the left side.

The fact that the upper left-hand side frames show a greater degree of plastic

deformation than those on the right-hand side is an illustration of the general complexity

of the impact event. The right side experiences a more significant acceleration pulse, as

has been noted, due to the reinforced cargo door structure. This leads, for example, to the

failure of the right side seats, while the left side seats remain upright. On the upper

frames of the test article, however, the left side shows larger fractures, as measured by the

degree to which the failure extends through the frame.

This greater deformation seen in the left-hand side upper frames can be explained

as resulting not only from the acceleration pulse, but also from the downward tilt on the

left side, which changes the magnitude and location of peak bending stresses on the upper

frames. As has been noted, this effect is also captured in the simulation, which shows

higher plastic strains on the upper left-hand side frames, as compared to the upper right-

hand side frames.

 102

Figure 83. FS 400 Left Upper Side Wall Figure 84. FS 400 Right Upper Side Wall

Figure 85. FS 440 Left Upper Side Wall Figure 86. FS 440 Right Upper Side Wall

Figure 87. FS 480 Left Upper Side Wall Figure 88. FS 480 Right Upper Side Wall
 103

Figure 89. FS 400 Left Lower Side Wall Figure 90. FS 400 Right Lower Side Wall

Figure 91. FS 440 Left Lower Side Wall Figure 92. FS 440 Right Lower Side Wall

Figure 93. FS 480 Left Lower Side Wall Figure 94. FS 480 Right Lower Side Wall
 104

5.4.8. Acceleration Time Histories of Seat Cushions

Although the six anthropomorphic test dummies (ATDs) were not included in the

simulation, it is still possible to compare acceleration response results from the seats with

those of the instrumented ATDs. Each ATD had an accelerometer mounted near the

pelvis, which is close to the seat surface. These experimental results from the ATD were

compared with simulation results from the seat cushion itself. Thus while the comparison

does not use identical points, the proximity of the seat cushion to the accelerometer on

the ATD provides a reasonable point of comparison. Figure 95 shows the location of two

seat cushions that were used to compare with experimental results from the ATDs.

Location of Acceleration Response on Seats at FS 408

Figure 95. Location of Acceleration Response for Comparison with ATD’s.

 105

The simulation correctly identifies the peak acceleration value for the left-hand

side seat, though the simulated cushion shows a slightly higher frequency rebound than is

found on the ATD. This comparison is shown in Figure 96. Figure 97 shows the right

hand side response, again comparing the simulated seat response to that of the ATD. It

must be noted that the right hand side seats collapse in the test event, which explains the

lower peak acceleration value for the experimental sensor. In the simulation, where no

collapse occurs, the peak acceleration response it higher than that on the left hand side, as

expected. This agrees with the trends seen in test results, as well as the results for seat

track response in both test and simulation, as presented earlier. The right-hand side seats

consistently experience an acceleration pulse which is approximately 5 g’s higher that

those seen on the left hand side. The acceleration pulse shape and duration for the

simulated seat agrees very well with that of the experimental ATD.

This comparison shows that the simulation may be used to assess the acceleration

pulse experienced by the occupants, to study occupant injury. This represents an

important step in using the current simulation to study occupant injuries resulting from

impact. There are three approaches which could be taken in studying occupant response.

First, the acceleration pulse from the seat could be used as input to an occupant model, to

determine the loads experienced by the occupant, and to assess potential injuries. A

second approach would be to include an occupant model in the existing simulation.

Several choices currently exist for fully finite element occupant models, one of which

could be placed in the existing seat structure. A third approach is a hybrid model,

coupling MADYMO with LS-Dyna. Use of an existing MADYMO dummy model

would be less difficult to implement than a full finite element occupant model.
 106

MADYMO has a set of validated models, and has been used successfully with LS-Dyna

in previous simulations.

Figure 96. Left Side Response – ATD and Simulation at FS 408

Figure 97. Right Side Response ATD and Simulation at FS 408

NOTE – Right side seats collapsed
 107

Chapter 6: Study of Simulation Parameters

In developing the finite element model of the fuselage section, a series of hand

measurements of the actual test article were carefully conducted to ensure that the

geometrical dimensions of the model were as accurate as possible. There are, however,

other factors that may influence the outcome of the simulation, and must be examined as

well. Parametric studies have been conducted to investigate four of these factors,

luggage stiffness, friction between the fuselage and platform, degradation of material

properties due to fatigue and corrosion, and failure of the structure during impact.

6.1. Effect of Friction Between Airframe and Platform

It was initially assumed in the simulation that the fuselage would not slide on the

impact surface during the impact event. This assumption was based on detailed

observation of the experimental impact sequence. Close study of the high-speed film of

the impact sequence shows no discernable slipping between the skin and platform. The

airframe does not appear to rotate or slide, even when vertical rebound occurs after the

initial impact.

The coefficient of friction was set to 1.0 for the simulation results presented in

Section 5.0, which will be referred to as the baseline simulation. A friction coefficient of

1.0 means that no sliding is permitted after the fuselage skin and the impact surface come

into contact. A friction coefficient of 0.0 means that no friction exists between the two

surfaces after impact, and surfaces are free to slide.

Friction coefficients of 0.0, 0.5, and 0.8 were selected for study. It was found that

the simulation impact response with a friction coefficient of 0.5 was almost identical to

the response with a coefficient of 0.0. Further, it was found that a 0.8 coefficient of
 108

friction led to only slight changes in simulated impact response of the fuselage. Thus the

simulation impact response does not appear to be affected by friction until the friction

coefficient is 0.8 or higher. For this reason, the comparisons in this section are presented

for friction coefficients of 0.0 and 1.0, covering respectively conditions of sliding and

fixed contact after impact.

Simulation impact results using a friction coefficient of 0.0 show that the absence

of friction noticeably alters the pattern of deformation, and that sliding does indeed occur

when friction is omitted. Without friction the center section of the airframe crushes

upward, and this affects the buckling pattern of the left-hand side frames, as seen in

Figure 98. This upward deflection at the center is clearly evident in comparison to the

fixed case, and also is in contrast with the previously shown experimental results. For

both cases the fuselage section begins to rotate clockwise, though this rotation is more

extreme in the absence of friction. Thus the use of a friction coefficient of 0.0 leads to

results that are qualitatively less accurate in comparison with the baseline simulation.

Figure 98. Friction Coefficient = 0.0 (Left) and 1.0 (Right) at t=200ms
 109

The effect of friction on the acceleration results is less pronounced.

Acceleration response results from the Heath Tecna bin show that the overall magnitude

of the peak acceleration response is similar for both cases, though the sliding condition

leads to additional oscillations in the bin response. This is seen in Figure 99.

Figure 99. Average Acceleration Response of Heath Tecna Bin

The acceleration response results for the Hitco bin are shown in Figure 100, again

comparing results with frictional coefficients of 0.0 and 1.0. The peak acceleration of the

initial pulse is lower for the sliding case, and also occurs later. The delay in the initial

pulse is due sliding that occurs before the impact load is transferred up through the

frames. The first acceleration pulse with fixed contact occurs around 20 ms, while the

first pulse occurs around 45 ms when sliding is permitted. The primary acceleration

pulse around 120 ms, however, is identical in magnitude and timing for both the fixed

contact and sliding conditions.

 110

Figure 100. Average Acceleration Response of Hitco Bin

Similar minor differences in are observed in the acceleration response results for

seat track locations. Typical results are shown in Figures 101 through 104. The sliding

condition generally affects the timing of the acceleration pulse, though the peak

magnitude changes only marginally. In comparing left and right side acceleration

responses, it is seen that the left side response of the fuselage is influenced to a greater

degree by the absence of friction. Changes in the right-hand side acceleration response

show minor differences in peak magnitude, but the change does not follow a predictable

pattern. At FS 418 the sliding condition leads to a lower peak acceleration response for

the outside seat track, while at FS 452 the sliding condition leads to a higher peak

acceleration response. These results are shown in Figures 102 and 104.
 111

Figure 101. Acceleration Response of Figure 102. Acceleration Response of

Left Outside Seat Track, FS 418 Right Outside Seat Track, FS 418

Figure 103. Acceleration Response of Figure 104. Acceleration Response of

Left Outside Seat Track, FS 452 Right Outside Seat Track, FS 452

This same trend continues when examining acceleration response results from the

side wall locations, with little change evident in the peak values, as shown in Figures 105

through 108. One exception is the lower left side wall at FS 400, in which sliding prior to

buckling clearly lowers the peak acceleration response.

 112

Figure 105. Acceleration Response of Figure 106. Acceleration Response of

Upper Left Side Wall, FS 400 Upper Right Side Wall, FS 400

Figure 107. Acceleration Response of Figure 108. Acceleration Response of

Lower Left Side Wall, FS 400 Lower Right Side Wall, FS 400

In summary, it was found that the omission of friction leads to a qualitative

change in the impact response of the simulation, but does not have a significant effect on

the acceleration response of the bins or seat tracks. The most significant changes in

acceleration response tended to be on the left-hand side, because of the different buckling

pattern of the left-side frames when using a 0.0 friction coefficient. The absence of

friction permits these frames to slide, unlike the behavior of left-side frames in the test

article. In both the test and the baseline simulation the lower left-side frames buckle in
 113

an S pattern, while use of the sliding condition in the simulation permits the left-side

frames to fold over.

6.2. Effect of Reduced Yield Strength

The frame material is 7075-T6 aluminum, an alloy that is known to be susceptible

to corrosion. After 20 to 25 years of service, it is possible that exposure to moisture and

fatigue may have degraded the original material properties of the test article, resulting a

lower yield stress than would be found with the original material.

To determine if a reduction in yield strength was needed to account for the effect

of fatigue and corrosion, a 20 percent decrease in yield strength was implemented and

results were compared with the baseline simulation results. Overall, it was found that the

simulation is not highly sensitive to changes in the material yield strength. While the

reduction in yield strength leads to an increase in total vertical crushing, the 20 percent

reduction in yield strength leads to only moderate changes in the acceleration response

during impact.

The effect on deformation of reducing the material yield strength is seen in Figure

109. A 20 percent reduction in yield strength represents a significant change in material

properties, thought the impact response shows only minor differences in the two cases.

Although the change in yield strength does result in additional crushing, the overall

pattern of deformation remains the same for both conditions. One reason for the minor

difference in deformation in the two cases is that vertical displacement during impact is

also limited by luggage. In the absence of luggage, the change in yield strength might

have a more significant effect on the resulting deformation.

 114

Figure 109. Deformation with 20 Percent Reduction in Yield Strength (Left)

and Baseline (Right) at t=100ms

The acceleration response results for the Hitco overhead stowage bin shows only

minor variations with the reduction in allowable yield stress, as seen in Figure 110. The

load time histories of the support structures in the Hitco bin also indicate the effect of

reducing the yield stress is very slight. This is seen in Figures 110 and 111.

Figure 110. Average Acceleration Response of Hitco Bin

 115

Figure 111. Load in Forward Tie Rod, Figure 112. Load in Aft Tie Rod,
Hitco Bin Hitco Bin

Acceleration results from typical seat track locations are presented in Figures 113

through 118. All of the acceleration results indicate a reduction in the magnitude of the

peak acceleration pulse with a reduction in yield stress. These results taken together

indicate a moderate but noticeable effect on the impact response of the fuselage.

Figure 113. Acceleration Response of Figure 114. Acceleration Response of

Left Outside Seat Track, FS 418 Right Outside Seat Track, FS 418
 116

Figure 115. Acceleration Response of Figure 116. Acceleration Response of

Left Inside Seat Track, FS 418 Right Inside Seat Track, FS 418

Figure 117. Acceleration Response of Figure 118. Acceleration Response of

Left Outside Seat Track, FS 452 Right Outside Seat Track, FS 452

It was determined that the reduction in yield strength was not needed in the

simulation. This is based on the additional degree of crushing seen in the simulation

using the reduced yield strength, which does not compare as well with experimental data

as the baseline simulation results.

 117

6.3. Effect of Material Failure Criterion on Impact Response

Failures were observed in both the upper and lower frames of the test article

during impact. The baseline simulation does not include failure, and assumes that the

material will yield indefinitely, following the stress-strain relationship given in Figure

119.

Simulation of element failure is potentially important, however, as fractures will

result in load paths changing while the impact proceeds. Fractures could also affect the

resulting pattern of buckling, and could alter the shape of the collapsing frames, leading

to changes in the acceleration pulses experienced by the occupants or by cabin items.

An element failure criterion was implemented in the simulation to determine if the

resulting impact response would more accurately represent the experimental impact

response of the test article. There are several available options for use in simulating

element failure. The failure criterion option that was selected for study is effective

plastic strain, with a failure limit set at 12 percent. The point at which element failure

occurs is shown in Figure 119. Element failures did not occur in the skin section, so the

only material affected in this study was the 7075-T6 aluminum.

 118

Point of Element Failure

Figure 119. Effective Plastic Strain Failure Criterion Used in Study

A review of the literature on dynamic inelastic failure by Jones [38] notes that

there is still great debate on the appropriate failure criterion for beams under impact

conditions. Plastic work per unit volume, according to Jones, appears to be a good

candidate for widespread use as a failure criterion, though there are numerous other

candidates, depending on whether failure is tensile or shear.

Even with a selected failure criterion, the choice of failure limit was also

uncertain. Vignjevic and Cavalcanti [13] studied the effect of failure limits with effective

plastic strains of 12 and 18 percent in aluminum alloys, as well as the use of Tresca stress

for predicting failure. Their study was performed with the main frame of the Lynx

helicopter, which was modeled with a very fine mesh. They determined that all three

simulations, with three different failure criteria, produced very similar results, though the
 119

Tresca criterion led to a slightly different buckling pattern in the outside flange. They

concluded that the choice of a failure criterion did not significantly affect either the

collapse mechanism or the force-displacement results. All three simulations

corresponded well with the observed experimental buckling. Based on their study, no

clear recommendation could be made for an optimal failure criterion.

The test article in the present study does show post-impact fractures in several

places, including the upper frames on both sides, multiple fractures and crushing in the

lower frame sections, and fractures near the door. Figure 120 shows a failure in the

reinforced frame around the forward edge of the door. The failure, however, does not

involve the complete fracture of the section, despite the substantial buckling and

permanent deformation that is observed. A similar pattern of failure in the supporting

frame around the cargo door is seen around the aft edge.

It is significant to note that the failure in the finite element simulation leads to

removal of the element, effectively resulting in a fracture with the width of the failed

element. In the experimental drop test, fractures in the frame sections were frequently

partial, so that the frame was still able to transmit load, and still able to resist crushing.

With the complete removal of elements in a section of the frame, as shown in Figure 121,

the stiffness of the section drops drastically, resulting in a complete redistribution of

loads. This is because a frame with a row of failed elements will no longer carry any

load, and will therefore not transmit forces to the upper frames. Also, the failed elements

lead immediately to the formation of a plastic hinge on the right side, permitting the

frames to buckle. Element failure therefore affects both the timing and magnitude of the

impact response.
 120

Figure 120. Post-test Failure of Cargo Door Frame

Figure 121. Failure of Elements in Door Frame on Right Hand Side

 121

A comparison of the failures highlighted in Figures 120 and 121 shows

that while the simulation correctly identifies the location of failure, it does not correctly

represent the extent of the failure, which is substantially greater in the simulation.

FS 380

Figure 122. Fractures in Upper Frames on Left Hand Side

 122

Figure 123. Close-up View of Partial Failure in Frame at FS 380

The simulation results for the upper frames do not show any deleted or failed

elements. This corresponds reasonably well with the observation that very small

fractures are typically found in the upper frames, rather than complete failure of the

frame section, as seen in Figures 122 and 123. Observed fractures in upper frames on the

right-hand side are less extensive than fractures on the left-hand side.

Figures 124 and 125 show the pattern of deformation during impact, illustrating

the effect of employing the failure criterion. The right-hand side shows evidence of early

buckling, resulting from the failure of elements around the reinforced cargo door frame.

The failed elements, shown previously in Figure 121, lead to a greater degree of

deformation on the right side than is seen in either the baseline simulation or the actual
 123

drop test. Thus, in this case, the use of element failure has resulted in a less accurate

simulation of the impact event.

By 100 ms, it is clear that the use of the failure criterion has significantly changed

the pattern of deformation, in that crushing on both right and left sides is essentially

equal. Again, this is due to the failure of elements around the reinforced cargo door on

the on the right-hand side. This behavior, with equal crushing on both right and left

sides, does not correspond with either the experimental or baseline simulation results. At

200 ms the trend continues, with the fuselage section leaning slightly to the right with use

of element failure, seen in Figure 125.

Figure 124. Failure (Left) and No Failure (Right) at t=100ms

Figure 125. Failure (Left) and No Failure (Right) at t=200ms

 124

Thus the observed pattern of deformation resulting from the inclusion of failure in

the present simulation leads to a less accurate description of the impact response. This is

because a much finer mesh is needed to capture the partial and gradual failures evidenced

in the experimental test article. However, the significant increase in mesh refinement that

would be required is clearly not practical for the current study.

6.4. Study of Luggage Properties

A description of the viscous foam luggage model was previously presented in

Section 4. This material model includes several variables that influence the change in

luggage stiffness during crushing. Numerous studies were therefore undertaken to

identify the effect of these variables on the simulated impact response, and it was

determined that a key variable was that of initial luggage stiffness, E0.

The effect of initial luggage stiffness, E0, on the pattern of deformation was

studied in detail, using values of E0 ranging from 0.3 psi to 1.8 psi. The value that was

selected for E0 was 0.6 psi. This was based on comparisons with simulation and

experimental data on the overall extent of crushing, the duration of the impact event, and

the resulting acceleration time histories.

Typical results for deformation during impact are shown for two cases, E0 = 0.6

psi and E0 = 1.2 psi. A front view of simulation results is shown in Figure 126,

comparing the deformed shapes at 140 ms. The left-hand side of Figure 126 shows

results with an initial stiffness of E = 1.2 psi, while the right side shows results with an

initial stiffness of E = 0.6 psi. In comparing the right and left sides, it is evident that

luggage stiffness significantly effects the vertical displacement experienced by the

 125

simulated test section. This in turn affects the time duration of the impact event, which in

turn affects the acceleration pulses that are experienced in fuselage structure and

overhead stowage bins. Higher initial luggage stiffness leads to a significant decrease in

deformation, while lower initial luggage stiffness leads to greater deformation.

While Figure 126 shows only qualitative effects of luggage stiffness, it is clear

that luggage modeling is an important variable in determining overall crushing behavior.

Maximum dynamic deformation occurs at around t = 100 ms for the stiffer luggage, and

at around t = 140 ms for the softer luggage model. Thus the softer luggage not only

permits additional deformation, but also changes the duration of impact. By t = 250 ms,

in both cases the upper fuselage has rebounded after initial flexing, and there is also a

slight rebound up from the platform of the entire fuselage, particularly on the left-hand

side. This is seen in Figure 127, which also shows the significant effect of luggage

stiffness on post-impact rebound and residual deformation.

Figure 126. Effect of Luggage Stiffness, E0 = 1.2 (Left side)

and E0 = 0.6 (Right side) at Time = 140 ms
 126

Figure 127. Effect of Luggage Stiffness, E0 = 1.2 (Left side)

and E0 = 0.6 (Right side) at Time = 250 ms

The above luggage study also illustrates a major issue in aircraft crashworthiness,

in that the type and quantity of underfloor luggage greatly influences the impact response

of the airframe and seat tracks. Since this issue has not yet been addressed

experimentally, a validated luggage model could permit the computational study of the

effect of luggage on aircraft crashworthiness.

 127

Chapter 7: Study of Impact Conditions

An important advantage of crashworthiness simulations is the ability to study a

range of impact conditions, and to examine the effect of variations in the configuration of

the simulated test article without conducting expensive tests. This chapter presents the

impact response results from three studies; roll angle, luggage, and pitch angle. For each

study, results are compared with the baseline simulation.

7.1. Effect of Roll Angle to Left at Impact

One of the issues addressed in the drop test study of the fuselage section with

overhead stowage bins was the effect on bin response of dynamic loading in comparison

with equivalent static loading. It was shown in Chapter 5 that dynamic impact had a

significant effect on loading for the Heath Tecna bin on the right-hand side, and only a

minor effect on loading of the Hitco bin. However, only one dynamic impact condition

was addressed. It is possible that a roll angle at impact, as could occur in actual crash

conditions, might significantly alter the impact response of the overhead bins. A similar

issue may be raised concerning the effect of roll angle on seat and occupant response, an

issue that has not previously been studied. This section addresses the issue of whether a

roll angle at impact could have a significant effect on the loads in the overhead stowage

bin supports, and on the acceleration pulse experienced at the seat tracks.

A 10 and 15 degree roll angle at impact was introduced in the simulation, and

results were compared with previous simulation results having a 0 degree roll. Typical

deformation results are shown for the 10 degree roll angle condition at t = 60 ms and t =

140 ms, in Figures 128 and 129. Total time duration of the impact is very similar to that
 128

of the baseline case, with vertical velocity reaching zero near 140 ms. It is evident that

substantial crushing occurs on the left hand side.

Left Side

Figure 128. 10 Degree Roll Angle at t=60ms

Left Side

Figure 129. 10 Degree Roll Angle at t=140ms

The average acceleration response of the right-hand side Heath Tecna bin is

shown in Figure 130. The peak acceleration response of this bin increases slightly for the

roll angle conditions, with the average peak acceleration increasing from 13 g to 15 g.

These results are reasonable, as the Heath Tecna bin is on the high side of the fuselage,

and therefore somewhat cushioned from the impact. The timing of the acceleration

pulses, however, changes noticeably, and an additional pulse is found between 100 and
 129

150 ms. This change in timing and duration of the acceleration pulses leads to a

noticeable increase in loading of the struts around 120 ms with the 10 degree roll

condition. This effect is more evident in the aft strut, which shows a significant increase

in peak load in comparison to the zero degree roll angle condition. The load histories of

the forward and aft struts are shown in Figures 131 and 132.

Figure 130. Average Acceleration Response of Heath Tecna Bin

Figure 131. Load in Forward Strut of Figure 132. Load in Forward Strut of
Heath Tecna Bin Heath Tecna Bin
 130

The Hitco bin is located on the left-hand side, closest to the impact, and shows

more significant changes in impact response. The effect of the roll angle is clearly seen in

the acceleration response results presented in Figure 133, which shows that the resulting

acceleration pulse arrives earlier, is slightly higher, and is significantly longer than with

the zero degree role case. Peak acceleration increases from 11 g to 13 g, and pulse

duration is close to 100 ms, up from 70 ms with no roll angle.

Figure 133. Average Acceleration Response of Hitco Bin

Figure 134. Load in Forward Tie Rod of Figure 135. Load in Aft Tie Rod of
Hitco Bin. Hitco Bin
 131

A significant increase in peak load results from the introduction of a roll angle, as

seen in Figures 134 and 135. The baseline peak axial load is 1300 lbs, while the 10

degree roll condition shows a peak load of over 2400 lbs. Thus the 10 degree roll

condition almost doubles the load response seen in the primary support structure of the

Hitco bin.

The results presented in Figures 136 through 141 illustrate the effect of roll angle

on acceleration pulses for the left and right seat tracks. In general, the left roll angle

leads to an increase in the magnitude of the peak acceleration response, and a decrease in

the right side response. This change in peak acceleration would also effect the behavior

of the seats, and would likely lead to the collapse of seats on the left-hand side. Timing

of the pulse also changes, with the left-hand side pulses occurring earlier with the roll

angle, and the right-hand side pulses being delayed.

The roll angle study shows first that a relatively moderate roll angle can

significantly alter the impact response of both seats and overhead bins. This could be

important for certification testing, as current test conditions do not consider the effect of

roll angle. Second, this study indicates there is a worst-case roll condition in terms of the

resulting loads on structural supports in the overhead stowage bins. As the fuselage

section continues to tilt past 10 degrees, a greater percentage of the load is taken by the

secondary supporting structure. This means that the influence coefficient of the overhead

bin struts varies with roll angle. If the influence coefficient were plotted against roll

angle, the peak value would occur near the 10 degree roll condition. A more complete set

of data is needed, however, to precisely determine the most critical roll angle.
 132

Figure 136. Acceleration Response of Figure 137. Acceleration Response of

Left Inside Seat Track, FS 380 Right Inside Seat Track, FS 380

Figure 138. Acceleration Response of Figure 139. Acceleration Response of

Left Outside Seat Track, FS 418 Right Outside Seat Track, FS 418

Figure 140. Acceleration Response of Figure 141. Acceleration Response of

Left Outside Seat Track, FS 452 Right Outside Seat Track, FS 452
 133

7.2. Effect of Luggage

The baseline simulation shows that the luggage plays a significant role in

absorbing impact energy. To more clearly show the effect of luggage on the impact

response of the fuselage section, a simulation was conducted with the luggage removed.

This illustrates the response of an air vehicle that has little or no luggage loaded in the

under-floor compartment. The weight of the luggage was also therefore deleted, and the

new total weight was 3024 lbs lighter than the previous fuselage model with luggage.

The absence of luggage leads to a very different pattern of deformation during

impact, as seen in Figure 142. It should be noted that no contact surface was defined

between the lower frames and the floor. As a result, the lower frames move through the

floor, although this clearly could not occur in an experimental setting.

Deformation in the case without luggage continues until the frame sides that are

just below the floor beams are able to resist the crushing load. The greater degree of

vertical crushing observed without luggage shows that the luggage adds a significant

degree of stiffness to the fuselage section. The lack of energy absorption without luggage

is clearly illustrated in Figure 143, which shows more extensive downward deflection in

the upper frames. This deflection occurs because of the secondary impact of the right and

left frames after buckling, and because of the resulting impact force that is transferred

through the frames.

 134

Figure 142. No Luggage at t=140ms

Figure 143. No Luggage at t=200ms

The average acceleration response of the Heath Tecna bin is shown in Figure 144.

The initial peak occurs earlier for the case without luggage, although the value is similar

to that of the baseline case. There is also a noticeable pulse between approximately 130

and 180 ms, and another pulse cresting near 250 ms, at a time when the baseline case is

clearly moving toward zero. The duration of the impact event is therefore longer for the

case without luggage, and significant acceleration pulses are still experienced after the

frames have reached their final crushing stroke.

 135

Figure 144. Average Acceleration Response of Heath Tecna Bin

The load time histories of the Heath Tecna bin supports illustrate the high initial

peaks and continued oscillations that occur without luggage. This is attributable to the

lack of damping in the absence of luggage. These continued oscillations contrast with the

damped response of the baseline case and the experimental test, both of which show that

the bin response approaches zero before 250 ms. This is significant, as it supports the use

of the viscous foam model in characterizing the luggage. These results are seen in

Figures 145 and 146.

Figure 145. Load in Forward Strut of Figure 146. Load in Forward Strut of
Heath Tecna Bin Heath Tecna Bin
 136

The average acceleration response results for the Hitco bin is shown in Figure

147, again comparing the case without luggage with the baseline simulation. The

average acceleration results show an increase of approximately 50 percent in peak

acceleration value for the case with luggage removed.

Figure 147. Average Acceleration Response of Hitco Bin

The results presented in Figures 148 and 149 provide clear evidence of the effect
of omitting luggage. These force histories indicate severe oscillations that result in a
compressive load of approximately 2200 lbs for the aft tie rod. Peak tension loads have
also increased to between 1600 and 1900 lbs without luggage.

Figure 148. Load in Forward Tie Rod of Figure 149. Load in Aft Tie Rod of
Hitco Bin Hitco Bin
 137

The effect of luggage on acceleration response is significant, though the effect is

different on the right and left side of the airframe. The left side impact response shows

an increase in the magnitude of peak acceleration of typically between 50 and 100

percent without luggage, as seen in the seat track responses shown in Figures 150 through

153. The right side response is mixed, and frequently shows a decrease in acceleration

without luggage. However, the lower frames were permitted to penetrate the floor beams

and floor section, though this could not occur in practice. Thus the values for

acceleration response without luggage cannot be taken as conclusive.

Figure 150. Acceleration Response of Figure 151. Acceleration Response of

Left Outside Seat Track, FS 418 Right Outside Seat Track, FS 418
 138

Figure 152. Acceleration Response of Figure 153. Acceleration Response of

Left Outside Seat Track, FS 452 Right Outside Seat Track, FS 452

This study demonstrates that luggage has a significant effect on the resulting fuselage

impact response. If luggage is omitted from the under-floor compartment, the resulting

acceleration impact response for seats and overhead bins increases substantially.

7.3. Effect of Pitch Angle

Aviation crashworthiness tests are typically conducted with separate tests to study

vertical and longitudinal impacts, with each component being independently evaluated.

Even in cases where swing tests are conducted, the time and expense involved generally

restricts testing to obtaining a single data point, which may be of limited value in

assessing other angles of impact or impact on surfaces other than concrete. While

experimental tests provide valuable data, their results are valid only for the single test

condition under consideration.

Vertical drop tests, such as the current study, employ an initial impact velocity,

and then assess the resulting acceleration response of the fuselage section and cabin

items. Longitudinal impact tests, however, cannot follow this procedure. Unlike

automobile impacts, an aircraft impact will cover an extended area, and will vary
 139

significantly depending on the terrain, pitch angle, and other complex factors, such as

separation of wings or fractures in the fuselage section. For longitudinal testing, a typical

acceleration pulse is therefore provided. The selected pulse is based on an analysis of

actual aircraft accidents.

Fortunately, a triangular acceleration pulse with a peak at 100 ms provides a very

good approximation of an actual impact event, according to a recent FAA study [39].

The FAA determined that peak acceleration response values may be in the range of 6 to

16 g’s. This acceleration pulse represents the initial, and most critical, moment of

impact. After this initial pulse, friction associated with sliding slowly reduces the

longitudinal velocity to zero.

The experimental procedure followed by the FAA was to accelerate the fuselage

test section in the opposite direction, that is, in the aft direction. The acceleration pulse is

controlled by a pneumatically operated piston inside a closed cylinder. The magnitude of

acceleration is then controlled by a metering pin. The FAA recorded the instrumented

response of the overhead stowage bins and the under-floor conformal. The actual tests

were conducted at the Transportation Research Center Inc.’s Impact Simulator Facility in

East Liberty, Ohio, in 1997.

Three peak acceleration values were studied experimentally by the FAA, 6 g, 9 g,

and 16 g. The 16 g impact resulted in substantial damage to the frames, and failures

where the overhead stowage bins were supported. The 6 g case resulted in no observable

damage to the bins, but resulted in failure of the fuel cell supporting brackets. After

removal of the fuel cell, the 9 g test case resulted in no significant damage to the

overhead bins or frames. The 9 g acceleration case was selected for simulation in the
 140

current study, in combination with the previous 30 ft/sec vertical impact. The

longitudinal acceleration profile that was used in the simulation is given in Figure 154.

Figure 154. Acceleration Pulse for Longitudinal Impact Condition

This acceleration pulse leads to an experimentally determined velocity change of

approximately 32 ft/sec, again with the applied acceleration being in the aft (positive X)
direction, with respect to the occupants.
Figure 155 shows the change in velocity during the simulation, for a point located
on the cabin floor. This velocity profile accurately matches the profile resulting from the
experimental 9 g acceleration pulse.

Figure 155. Velocity Change Resulting From Longitudinal Acceleration Pulse

 141

Results are presented for the condition with a 9 g longitudinal acceleration pulse
combined with a 30 ft/s vertical impact. The impact deformation response is seen at two
selected times in Figures 156 and 157. For the side view in this sequence, the upper skin
panels have been removed to more clearly present the seat response during impact.

Figure 156. Front and Left Views, Effect of Pitch Angle at t=100ms

Figure 157. Front and Left Views, Effect of Pitch Angle at t=140ms

The acceleration responses of the Heath Tecna bins are shown in figures 158 and

159. Both the forward and aft locations for acceleration results indicate only slight

changes in the vertical impact response of the Heath Tecna bin. This shows that the

addition of a longitudinal acceleration pulse has a limited effect on the vertical impact

response. Similar results are seen for loading in the supporting structure of the Heath

Tecna bin, Figures 160 and 161.

 142

Figure 158. Acceleration Response of Figure 159. Acceleration Response of

Heath Tecna Bin, Forward End Heath Tecna Bin, Aft End

Figure 160. Load in Forward Strut of Figure 161. Load in Forward Strut of
Heath Tecna Bin Heath Tecna Bin

Results for the Hitco Bin acceleration response are presented in Figures 162

through 164. Forward and aft locations are shown separately, followed by the average

acceleration response. Peak acceleration results are similar for the primary pulse around

100 to 150 ms, showing the inclusion of longitudinal acceleration has a limited influence

on the vertical response.

 143

Figure 162. Acceleration Response of Figure 163. Acceleration Response of

Hitco Bin, Forward End Hitco Bin, Aft End

Figure 164. Average Acceleration Response of Hitco Bin

The load response in the struts that support the Hitco overhead stowage bin again

shows that the combination of longitudinal and vertical acceleration pulses are largely

independent, although the results do tend to diverge at later times. This is seen in Figures

165 and 166. The differences that occur are generally more in terms of timing than pulse

magnitude. This is understandable for the overhead bins, which see peak acceleration

pulses later in the impact event.

 144

Figure 165. Load in Forward Tie Rod of Figure 166. Load in Aft Tie Rod of
Hitco Bin. Hitco Bin

The seat track responses also show the limited effect of imposing a longitudinal

acceleration component in conjunction with a vertical impact, as seen in results presented

in Figures 167 through 170. The right-hand side in particular is largely unaffected. The

effect on the left-hand side is more notable, though the difference again is more of timing

than magnitude.

Figure 167. Acceleration Response of Figure 168. Acceleration Response of

Left Outside Seat Track, FS 418 Right Outside Seat Track, FS 418
 145

Figure 169. Acceleration Response of Figure 170. Acceleration Response of

Left Inside Seat Track, FS 418 Right Inside Seat Track, FS 418

Overall, the acceleration response results indicate that there is a limited

interaction between vertical and longitudinal acceleration pulses during impact. The

changes in vertical impact response resulting from the addition of a longitudinal pulse are

sufficiently minor to conclude that these two impact conditions may be tested separately

without inducing serious errors. This conclusion is noteworthy, as current experimental

testing relies on the assumption that vertical and longitudinal impacts are in fact

independent, with little non-linear interaction between the two cases.

This conclusion may be limited to narrow range of impact angles, as the

kinematics differ for the two cases. At some point, it is likely that the nonlinear

interaction between longitudinal and vertical impulses will become evident, and that a

simple linear combination of impact responses will no longer lead to valid results.
 146

Chapter 8: Discussion and Summary

In validating a finite element impact simulation, attention should not be focused

entirely on a detailed comparison of acceleration results between simulation and

experiment [27]. The objective instead should be to replicate the basic behavior of the

test article during impact, and to obtain simulation results that are meaningful and

physically reasonable. While a comparison with experimental data is important, and

validation of the dynamic simulation is vital, the experimental data itself is subject to a

certain degree of variability. This is observed in automobile impact studies, where

repeated tests yield, in some cases, noticeably different acceleration pulses. This may be

due to slight variations in manufacturing, slight differences in placement of the

accelerometers, or noise and other errors involved in the data acquisition process. Also,

the exact location of the nodes used in extracting simulation results may affect the

acceleration response data. While comparisons of acceleration data are important, the

most critical validation is a comparison of the observed pattern and timing of deformation

during impact. In this regard, the simulation shows a high level of fidelity to the actual

impact event.

The simulation results also compare well with experimental acceleration response

data, in that they correctly show the magnitude of the fundamental acceleration pulse,

while generally capturing the timing and duration of the pulse. The simulation also

captured the correct magnitude of loading in the struts and tie rods that support the

overhead stowage bins. The accuracy of the simulation in showing the impact response

of the overhead stowage bins is a clear indication that the overall fuselage model is a
 147

reasonable representation of the actual test article. This is because the overhead bin

response arises from the pattern of buckling and crushing that occurs in the lower

sections of the fuselage.

Further validation is found in the ability of the simulation to clearly explain the

acceleration time histories in terms of the deformation of the lower frames. One example

of this is the acceleration response of the Hitco bin, as was previously noted in Section

5.3. The initial impact leads quickly to buckling of the frames and the formation of a

plastic hinge on the lower left-hand side. The initial impact is reflected in a small

acceleration pulse that decreases after buckling is initiated. The plastic hinge then

impacts the platform, resulting in a second acceleration pulse, followed by additional

buckling. In general terms, it is possible to explain the acceleration pulses as a sequence

of impact loading resulting in buckling and crushing, followed again by impact loading of

the deformed and therefore stiffer structure. For crashworthiness applications, the ideal

response would be to have continuous structural crushing, meaning that the structure

gradually increases in stiffness.

A second example illustrating the clear link between the acceleration response

and observed pattern of deformation is found in a comparison of the right-hand and left-

hand side acceleration pulses of the seats and seat tracks. As previously noted, the right-

hand side cargo door reinforcement has a dominant effect on the overall impact response

of the fuselage. The additional stiffness of the reinforced cargo door limits the overall

vertical deflection on the right-hand side, which leads to higher right-hand side

acceleration pulses. This effect is seen in the experimental impact response in which all

right-hand side seats failed. The simulation results also show that the right side is
 148

significantly stiffer that the left, and therefore experiences generally higher acceleration

pulses.

Minor discrepancies were observed in comparing simulation and experimental

results due to the inevitable modifications required in constructing the finite element

model. This is most clearly shown in the finite element representation of the cargo door,

which uses a simplified method of attaching the cargo door to the frame. In the model,

the cargo door is attached directly to the frame. In the test article, the cargo door is not

directly fixed to the frame, but is essentially pressed onto the frame from inside, and held

in place with locking bolts.

The door-to-frame connection in the test article influences the timing of the force

that is transferred through the door, and the degree to which the door itself will buckle.

The simplification in which the door is directly connected to the frame will therefore

have some affect on the timing of acceleration responses in the region of the cargo door.

While this region could be modeled in more detail, the complex geometry of the door and

its attaching mechanisms results in many unknowns, such as the exact degree of contact

pressure exerted by the door on the frame, and the percentage of load transferred through

each of the bolts. Hence a simplified version of the cargo door attachment was adopted.

A second modification in the model was to join all structural members together

directly, rather than simulate the rivets and bolts. As previously discussed in Section 2.4,

it is unlikely that the omission of rivets will have a significant effect on the fundamental

impact response of the fuselage section. However, the early initial acceleration peaks

sometimes seen in the simulation results could be explained in part due to the direct

element-to-element connection of joined sections, leading to slightly stiffer joints. Such

 149

joints occur where sections of the frame are spliced together, and where the frames are

joined to the under-floor beams. In the test article, such joints have some degree of

compliance and it is possible that the time required to load the rivets will lead to some

delay in transferring impact loads through riveted joints. One method of dealing with this

would be to model these joints in greater detail, with explicitly modeled rivets, in order to

assess the affect of including rivets in the simulation. However, this minor improvement

in the fidelity of the simulation would come at a greatly increased cost in time and

complexity.

A third modification was to slightly alter the initial condition of the test article.

Prior to testing, the fuselage section was suspended from a cable attached to the upper

section of the fuselage. This would cause the test article to deform under its own weight

prior to release. This also would cause the test article to vibrate after release as it moves

towards equilibrium during freefall. This could influence the simulation results in a

random manner, depending on what phase of oscillation the frames are in when impact

occurs. Though this effect could be small, it would be most noteworthy in the first few

milliseconds after impact. In the simulation the initial condition prior to impact is that of

the fuselage section when sitting on the ground under 1g of gravity.

A fourth modification was made in modeling the seats and test dummies. The

masses of the six ATDs and twelve mannequins were lumped onto the seat surfaces. This

affects the timing of the impulse transmitted to the seat track. In the test article,

deformation of the seat cushions by the test dummies causes a time delay in transmitting

forces to the seat frame. In the simulation, this effect is not captured, as the dummy

masses are directly attached to the seat. Also, the seats in the simulation were permitted
 150

to deform, but not to fail, as accurate modeling of seat failure was beyond the scope of

the present study. Thus the experimental failure of seats on the right-hand side altered

the timing and magnitude of the force transmitted to the right-hand side seat tracks and

floor beams. The collapse of these seats was not reflected in the simulation.

Even with the above modeling modifications, extensive comparisons of

simulation and experimental results show that the simulation captures the fundamental

acceleration response and deformation resulting from impact. This indicates that the

modeling simplifications were reasonable, and do not adversely affect the utility of the

simulation. This is a key point, as it shows that the level of modeling detail used in the

current simulation is adequate for crashworthiness studies.

Furthermore, the current model can be extended to include the complete air

vehicle, and still remain within current practical computing limits for performing

dynamic impact simulations. Using a network of PCs, it is currently possible to perform

overnight simulations of automobile impact events where models contain 500,000 to

600,000 elements. If the current level of modeling detail is extended to include an entire

airframe, the model would contain approximately 500,000 elements. Thus a detailed

crashworthiness model of a full transport air vehicle, including all primary structure, seats,

and critical cabin items, is therefore practical at the present time.

The validated model was used to study the effects of three selected simulation

parameters. First, friction between the test article and the impact surface was studied. It

was determined that friction has a clear qualitative effect on deformation, though the

effect on simulation peak acceleration results was less evident. Without friction, the test

article tends to slide slightly on impact. This initial sliding causes a slight reduction in
 151

the initial peak in acceleration that is sometimes found in simulation results for frame

locations. The reduction in the early peak acceleration value occurs because the frame

section is allowed to slide inward on initial contact, rather than staying fixed at or near its

impact point. More generally, the fact that friction will affect the response of even a

vertical impact onto a rigid platform indicates that the type of impact surface, such as

concrete, soft soil, or sand, will clearly affect the impact response of the airframe [40].

This also indicates that characterization of the impact surface in cases with a longitudinal

impact velocity is quite important.

The second parameter studied was the effect of material yield stress, which was

reduced by 20% to account for the effects of corrosion and fatigue. However, the

sensitivity of the simulation to material yield strength was found to be moderate, rather

than decisive. A reduction in yield strength of 20% led to small changes in the peak

acceleration responses of the overhead stowage bins, for example. The reduced yield

strength did lead to an increase in vertical displacement, or crushing, of approximately

10% on both right and left sides. The reduced yield strength did improve the correlation

with experimental data at some points, but the overall effect was minor. Based on this

study, it would appear that a reduction in yield strength of approximately 10% would

offer the best description of material properties for impact simulations of aging aircraft.

The third parameter that was studied was the inclusion of a material failure

criterion. Element failure was set at 12% effective plastic strain, meaning that the

element was deleted at this point. Fractures were observed in the test article after impact,

and the failure criterion was an effort to simulate this behavior. However, the use of this

failure criterion did not improve simulation results, and led in fact to buckling behavior
 152

that was contrary to that observed in the experiment. It is possible that current

uncertainties in an appropriate dynamic failure criterion make its implementation

susceptible to error. It is very likely that a significantly finer mesh is needed to correctly

implement a failure criterion, as the complete removal of failed elements does not

accurately describe the observed fractures in the test article.

The simulation was also used to determine the amount of energy dissipated by

each section of the airframe. It was found that the frames and luggage combined are

responsible for approximately 68% of the energy dissipation during the impact. A careful

examination of the patterns of deformation also indicates that the frames and luggage are

the two most significant factors influencing overall structural impact response.

A study was therefore conducted to characterize the behavior of luggage under

impact, and to determine the most appropriate material properties. The luggage was

modeled as viscous foam with properties that were dependent upon its current volumetric

strain during the impact event. This permitted the luggage to gradually increase in

stiffness during crushing and to absorb impact energy, as would be expected.

The effect of luggage was shown more clearly by a simulation in which luggage

was entirely deleted from the under-floor compartment. The absence of luggage led to an

increase in the peak acceleration values at floor level of approximately 25%. Even more

significantly, the simulated impact without luggage resulted in the almost total collapse

of the upper frame section. The collapse led to the intrusion of the bins into the occupant

space, a clear violation of the desired impact response, in which a survivable occupant

space should remain intact. Thus crashworthiness, for a commercial transport aircraft,

depends in large part on the type and amount of luggage in the under-floor compartment.
 153

A study of roll angles at impact found that the loads in the primary vertical bin

supports change significantly with a roll angel of 10 degrees to the left. The 10 degree

roll condition resulted in an increase in peak tie rod loads in the Hitco bin of

approximately 90 percent, and would likely have led the bin to detached under such

conditions. The 15 degree roll angle condition resulted in peak tie rod loads that were 80

percent higher than those of the zero degree roll condition. This shows that as roll angle

increases to 15 degrees and beyond, the secondary links begin to carry more of the load.

Of the conditions studied, the 10 degree roll represents the worst case condition for roll

angle.

The left-hand side seats also experienced a significant increase in loading with the

10 degree roll angle, which again could influence their possible failure, or contribute to

an increased level of occupant injuries. This finding is potentially significant, as it

contrasts with the static qualification testing currently being conducted for overhead bins.

Static tests for overhead stowage bins are currently only performed with a zero degree

roll condition. Dynamic seat testing for certification currently includes pitch and yaw

angles, but not a roll angle.

Another issue that has received attention is that of testing vertical and longitudinal

impacts separately. This issue is significant because current aviation crashworthiness

testing is frequently done with separate tests for vertical and longitudinal impacts. A

combined test case was simulated, with a 9 g longitudinal pulse combined with the 30

ft/sec vertical impact. It was found that these two impact modes are largely independent,

partly because of the differences in timing for the two acceleration pulses and impact

forces. In general terms, the effect of the vertical impact tends to occur relatively early in
 154

the impact event, while the effects of the longitudinal impact occur somewhat later. Thus

it was found that the use of independent vertical and longitudinal impact testing is largely

justified, as limited non-linear effects were found in combining the two impact

conditions. While testing is often employed based on this assumption, the exact effect of

combining the two load cases has not been previously studied. The present study shows

that the independent use of vertical and longitudinal testing yields valid results.

Future work based on the present simulation could include the investigation of a

broader range of alternate impact conditions or configurations. The ability to evaluate the

impact response for other test cases is a major advantage of crashworthiness simulations.

Impact conditions that could be studied are impact into soft soil, water, or sloping terrain.

It would also be possible to study a variety of test configurations, such as changing the

amount of luggage loaded in the under-floor compartment or in bins, altering the number

of passengers, changing overhead bin supports, or loading more rigid cargo in the under-

floor compartment. Finally, since the simulation has accurately captured the acceleration

response of the lower seat cushions under impact loading, it would be possible to employ

the simulation in future studies of occupant injury.

In sum, the current study represents a significant step in conducting air vehicle

crashworthiness simulations. In addition to validating the basic simulation, studies were

conducted addressing the properties and effect of luggage, the effect of a roll angle at

impact, and the effect of combining vertical and longitudinal impacts. With

approximately 60,000 elements, the present simulation could be readily adapted to model

an entire transport air vehicle, including a small number of finite element dummies, and

still remain within the current practical computational limits. As computational

 155

capabilities continue to expand, it is likely that complete crashworthiness simulations,

including fuselage structure, cabin items, seats, and occupants, will become routine for

use in evaluating and improving crashworthy air vehicle designs.

 156

List of References

[1] Laananen, D. H., and Winkelman, K. L., “Analysis of energy-absorbing seat

configurations for aircraft,” International Journal of Crashworthiness, Vol 1 No
4, 1996.

[2] Kindervater, C. M., Kohlgruber, D., and Johnson, A., “Composite vehicle
structural crashworthiness – A status of design methodology and numerical
simulation techniques”, International Journal of Crashworthiness, Vol 4 No 2,
1999.

[3] Waldock, W. D., “A Brief History of Crashworthiness”, SAFE Association

Symposium, Phoenix, Arizona, September 20, 1997.

[4] Gamon, M., Wittlin, G., and LaBarge, B., “KRASH 85 Userr’s Guide –
Input/Output Format”, Final Report DOT /FAA/CT-85/10, May 1985.

[5] Anon., “Air Accident Investigation Tool – User’s Manual”, Cranfield Impact
Centre, Cranfield England.

[6] PAM-CRASH, Engineering Systems International SA, F-94588 Rungis, France.

[7] Anon., “MSC Dytran User’s Manual”, The MacNeal-Schwendler Corporation,

Los Angeles, CA, 1999.

[8] Anon., “LS-DYNA3D User’s Manual”, Livermore Software Technology

Corporation, Livermore, CA, 1997.

[9] Anon., “Computational Methods for Crashworthiness”, NASA Conference

Publication 3223, October 1993.

[10] Hillmann, J., Konig, C., and Wang, X., “10 Years of Crash Simulation at
Volkswagon”, Second International LS-DYNA3D Conference, San Francisco,
California, September 20 and 21, 1994.

[11] Kirkpatrick, S. W., Simons, J. W., and Antoun, T. H., “Development and
Validation of High Fidelity Vehicle Crash Simulation Models”, SRI International,
Menlo Park, CA, 1998.

[12] Webb, D. C., Webster, J., and Kormi, K., “Finite Element Simulation of Energy
Absorption Devices under Axial Static Compressive and Impact Loading”,
International Journal of Crashworthiness, Vol 6 No 3, 2001.
 157

[13] Vignjevic, R., and Cavalcanti, M. J., “Numerical simulation of the Lynx
helicopter main lift-frame component collapse”, International Journal of
Crashworthiness, Vol 2 No 1, 1997.

[14] Alves, M., and Jones, N., “Impact failure of beams using damage mechanics: Part
I – Analytical model”, International Journal of Impact Engineering, Volume 27,
pp. 837-861, 2002.

[15] Kenny, S., Pegg, N., and Taheri, F., “Finite element investigations of the dynamic
plastic buckling of a slender beam subject to axial impact”, International Journal
of Impact Engineering, Volume 27, pp. 179-195, 2002.

[16] Cui, S., Hao, H., Cheong, H. K., “Numerical analysis of dynamic buckling of
rectangular plates subjected to intermediate-velocity impact”, International
Journal of Impact Engineering, Volume 25, pp. 147-167, 2001.

[17] Hanssen, A. G., Hopperstad, O. S., and Langseth, M., “Design of aluminium
foam-filled crash boxes of square and circular cross-sections”, International
Journal of Crashworthiness, Vol 6 No 2, 2001.

[18] Ahmed, T. U., Ramachandra, L. S., and Bhattacharyya, S. K, “Elasto-plastic

response of free-free beams subjected to impact loads”, International Journal of
Impact Engineering, Volume 25, pp. 661-681, 2001.

[19] Jones, N., “Dynamic material properties and inelastic failure in structural
crashworthiness”, International Journal of Crashworthiness, Vol 6 No 1, 2001.

[20] Jackson, K. E., “Impact Testing and Simulation of a Crashworthy Composite

Fuselage Concept”, International Journal of Crashworthiness, Vol 6 No 1, 2001.

[21] Fasanella, E. L., Boitnott, R. L., Lyle, K. H., and Jackson, K. E., “Full-scale crash
test and simulation of a composite helicopter”, International Journal of
Crashworthiness, Vol 6 No 4, 2001.

[22] Jackson, K. E., and Fasanella, E. L., “Crashworthy Evaluation of a 1/5-Scale

Model Composite Fuselage Concept”, U. S. Army Research Laboratory Vehicle
Technology Center, Langley Research Center, Report Number NASA/TM-1999-
209132 ARL-MR-441, 1999.

[23] Fasanella, E. L., and Jackson, K. E., “Crash Simulation of a Vertical Drop Test of
a B737 Fuselage Section With Auxiliary Fuel Tank”, U.S. Army Research
Laboratory, Vehicle Technology Center, Langley Research Center, April 2000.
 158

[24] Fasanella, E. L., and Jackson, K. E., “Crash Simulation of Vertical Drop Tests of
Two Boeing 737 Fuselage Sections”, U.S. Army Research Laboratory, Vehicle
Technology Center, Langley Research Center, March 2002.

[25] Lawver, D., Nikodym, L., Tennant, D., and Levine, H., “Non-linear numerical
modeling of aircraft impact,” International Journal of Crashworthiness, Vol 6
No 4, 2001.

[26] Matzenmiller, A., Schweizerhof, K., and Rust, W., “Joint Failure Modeling in
Crashworthiness Analysis”, Second International LS-DYNA3D Conference, San
Francisco, California, September 20 and 21, 1994.

[27] Bisagni, C., “Crashworthiness of helicopter subfloor structures”, International

Journal of Impact Engineering, Volume 27, pp. 1067-1082, 2002.

[28] Vignjevic, R., and Meo, M., “Simulation of helicopter under-floor structure
impact on water,” International Journal of Crashworthiness, Vol 6 No 3, 2001.

[29] Reid, J. D., “Tracking the energy in an energy absorbing guardrail terminal”,
International Journal of Crashworthiness, Vol 3 No 2, 1998.

[30] Altenhof, W., Saverio, P., Zamani, N., and Gaspar, R., “An experimental and
finite element investigation into the energy absorption characteristics of a steering
wheel armature in an impact”, International Journal of Impact Engineering,
Volume 27, pp. 9197-212, 2002.

[31] Cheng, Z. Q., Pilkey, W. D., Balandin, D. V., Bolotnik, N. N., Crandall, J. R., and
Shaw, C. G., “Optimal control of helicopter seat cushions for the reduction of
spinal injuries”, International Journal of Crashworthiness, Vol 6 No 3, 2001.

[32] Nicholson, C. R., and Turnour, S. R., “Toward Simplified Methodologies for Seat
Design to FAR Part 23.562”, Society of Automotive Engineers, Inc., 2001.

[33] Nicholson, C. R., Turnour, S. R., and Chapman, H., “The Design and Testing of
Buckling Monocoque Seating Structures for Aircraft”, Society of Automotive
Engineers, Inc., Paper Number 1999-01-1599, 1998.

[34] Hooper, S. J., and Ellis, D. R., “Aviation safety and crashworthy seat design”,
International Journal of Crashworthiness, Vol 2 No 1, 1997.

[35] Teulings, A. M. G. L., Gowdy, V., Wismans, J. S. H. M., and Aljundi, B.,
“MADYMO Validation of Side Facing Sofa Sled Tests”, North Atlantic Treaty
Organization, Research and Technology Organization, Paper Number
AC/323(HFM)TP/7, RTO Human Factors and Medicine Panel, Wright-Patterson
Air Force Base, Ohio, October 1998, Published August, 1999.
 159

[36] Fredriksson, L. A., and Nilsson, L., “An Advanced Finite Element Data Base of
the Hybrid III Test Dummy: Phase I, Model Generation and Sled Test
Validation”, Second International LS-Dyna Conference, The Westin St. Francis,
San Francisco, CA, 20th and 21st September, 1994.

[37] Abramowitz, A., Smith, T. G., Vu, T., and Zvanya, J., “Vertical Drop Test of a
Narrow-Body Transport Fuselage Section with Overhead Stowage Bins,” Draft
Report Number DOT/FAA/AR-01/100, March 2002.

[38] Jones, N., “Dynamic material properties and inelastic failure in structural
crashworthiness”, International Journal of Crashworthiness, Vol 6 No 1, 2001.

[39] Ramalingam, V. K., and Lankarani, H. M., “Analysis of impact on soft soil and its
application to aircraft crashworthiness”, International Journal of
Crashworthiness, Vol 7 No 1, 2002.

[40] McGuire, R., “Longitudinal Acceleration Test of Overhead Luggage Bins and
Auxiliary Fuel Tank in a Transport Airplane Airframe Section, Part II”, Report
Number DOT/FAA/AR-99/4,II, October 2000.
 160

Vita

Alan Byar obtained a combined Bachelor of Science and Master of Science

degree in Mechanical Engineering from Drexel University in 1988. During this time he

was working with Dyna East Corporation, designing conventional warheads. He then

moved to Boeing as a structural analyst, employing both classical and finite element

analysis methods to analyze airframe structures. In 1999, he returned to Drexel with an

FAA Fellowship to study aircraft crashworthiness. He has presented papers at several

international crashworthiness conferences, presenting simulations based on numerical

analysis of the impact response of commercial aircraft. He is currently with the Boeing

Company Phantom Works in Seattle, performing impact studies and developing a model

to analyze crashworthiness of the new Boeing 7E7 transport aircraft.