14th International LS-DYNA Users Conference Session: Connections

Development and Validation of Bolted Connection Modeling

in LS-DYNA for Large Vehicle Models

Michalis Hadjioannou, David Stevens, Matt Barsotti

Protection Engineering Consultants LLC, Austin, TX, USA

Abstract
As part of the United States Marine Corps (USMC) Mitigation of Blast Injuries through Modeling and Simulation
project, Protection Engineering Consultants performed numerical and experimental investigations to develop
modeling approaches for bolted connections. Vehicle models require efficient yet accurate methods to represent
bolted connections, especially under extreme loading situations where connection behavior may have crucial impact
on the accurate prediction of the response. Efficient connection models require relatively coarse mesh sizes and
computationally cheap element types that allow modeling large numbers of connections in vehicle models. This
paper describes the development and validation of reduced bolted connection models that utilize a combination of
beam and shell elements. The models were developed and validated with data from bolted connections that were
tested under static and dynamic loading conditions. The tests provided valuable data for the refinement of the
models, which are shown capable of simulating connection behavior up to and including rupture.
Important aspects of the modeling procedure are highlighted including contact definitions and bolt preloading, as
well as inherent limitations that exist in such models. The study also demonstrates the importance of material failure
parameters such as triaxiality-dependent and strain-rate-dependent fracture. These parameters influence not only
the connection capacity but also the absorbed energy before the connection fails. Considerations of the absorbed
energy are crucial when assessing the safety of occupants in vehicles under extreme loading conditions.

Introduction
As part of the United States Marine Corps (USMC) Mitigation of Blast Injuries through
Modeling and Simulation project, a simplified yet accurate methodology was developed to
simulate the behavior of steel bolted connections including rupture. This methodology is useful
when there is a need to model numerous bolted connections in vehicle models. The design of
commercial and military vehicles is an iterative procedure and requires the evaluation of
different design options to meet performance criteria. Therefore, such models need to be
computationally cheap and easy to construct but they also have to provide accurate results.
Accurate representation of bolted connections in vehicle models is crucial especially when the
vehicle design focuses on occupant safety using components that are designed to absorb energy
under extreme loading conditions, such as collisions, explosions, etc.
Previous efforts in connection modeling provided a basis for the development of the proposed
simplified modeling methodology presented in this paper. A number of research studies focused
on simplified modeling approaches on bolted connections [1,2]. Depending on the application of
the finite element model and the purpose of the analysis, bolts can be represented with merged
nodes of separate components at the physical locations of the bolts or can be as detailed as
including a finely meshed bolt with the nut tightened by representing the bolt threads.

June 12-14, 2016 1-1

Session: Connections 14th International LS-DYNA Users Conference

Between these two extremes, there are other simplified (reduced) approaches that can potentially
represent the connection behavior with good accuracy and at the same time have significant
gains in terms of computational expense. Sonnenschein [3] has reviewed a number of alternative
simplified modeling procedures of bolted connections using LS-DYNA. A commonly used
approach utilizes beam elements to simulate the bolt shank; these are attached to beam elements
arranged radially around the bolt hole, also known as a spider mesh. The spider mesh is intended
to represent the bolt head and nut. This approach eliminates the need of contact definitions which
can result in inaccurate representation of the bearing stresses that are developed between the bolt
shank and the bolt hole. For that reason, the typical failure modes associated with bolted
connections such as plate tear-out rupture and bolt shear failure cannot be captured.
Sonnenschein [3] suggested an intermediate approach which can account for the bearing stresses
of the bolt in the bolt-hole by avoiding the need to use solid elements. In this case the bolt head
is modelled with shell elements which are connected with beam elements that represent the bolt
shank. The bolt shank interacts with the bolt hole through null beam elements that are placed
around the hole. An alternative to that approach was proposed by Narkhede et al [2] that uses
discrete spring elements to account for the interaction of the bolt shank with the bolt hole. These
two approaches show good agreement with experimental results but neither was capable of
explicitly capturing the connection response up to failure or the failure mode.
The proposed simplified modeling methodology not only captures the connection behavior at the
early loading stages, but it is also able to capture connection rupture with good accuracy either
under static or dynamic loading conditions. The method was rigorously validated against
connections that were physically tested under dynamic and static loading conditions until
complete failure. The data gathered from these tests provided a basis for the development and
validation of simplified bolted connection models. These models use beam and shells elements to
explicitly represent important characteristics of bolted connections such as bolt preloading,
interaction of the separate connection components through contact definitions, and physical
fracture of the connecting parts. Certain limitations of the modeling procedure were also
identified and are discussed in this paper.

Static and Dynamic Pendulum Testing of Bolted Connections

Experiments
Protection Engineering Consultants (PEC) designed a series of tests on bolted connections to
provide data for the validation of simplified modeling approaches for bolted connections. The
tests applied direct shear to bolted connections as shown in Fig. 1; different bolt diameters, plate
thicknesses, and bolt-hole to edge distances were evaluated. By adjusting these parameters,
typical failure modes were achieved including plate tear-out failure and direct shear fracture of
the bolt as shown in Fig. 2.

1-2 June 12-14, 2016

14th International LS-DYNA Users Conference Session: Connections

bearing load
shear

bearing
load

Fig. 1: Bolted connection loaded in direct shear

(a) plate tear-out failure (b) Bolt shear fracture

Fig. 2: Typical failure modes of physical tests; (a) plate tear-out failure, (b) bolt shear fracture
The tests were performed by Southwest Research Institute (SwRI) using a universal loading
machine for the static tests and a 2250-lbf pendulum for the dynamic tests; see Fig. 3. By
adjusting the drop-height of the pendulum, different load rates were achieved. The connection
specimens were loaded through a specially designed frame that accommodated the connection
specimens shown in Fig. 3(b). Similar test specimens to those that were tested dynamically were
also tested statically until total failure. A total of twenty-four pendulum and seven static tests
were performed during the testing program. All specimens consisted of rolled homogeneous
armor (RHA) steel plates and Grade 8 bolts. Similar tests were performed with welded
connection specimens but they are outside the scope of this paper. The data gathered from these
tests were used to validate the simplified bolted connection models.
The strains in all specimens were measured using digital image correlation (DIC). In particular,
for the dynamic tests, the peak axial strain (thus peak axial force), the pendulum impact velocity
and failure mode were used as the basis for comparison with the numerical models. In addition,
the static tests were instrumented with conventional load cells and string potentiometers since
these test configurations allowed their usage.

June 12-14, 2016 1-3

Session: Connections 14th International LS-DYNA Users Conference

moving
end

loading
frame
2250-lbf
pendulum

connection
specimen

loading
frame

2250-lbf
pendulum
fixed
end connection
specimen
FE model
(a) static testing (b) dynamic testing

Fig. 3: Testing apparatus used for the connection tests; (a) static testing, (b) dynamic testing

Proposed Simplified Modeling Procedure

Generating the Finite Element Mesh

The bolt-head and nut are represented with fully integrated shell elements (elform=16) that are
placed in the surface defined from the bolt-head and nut mid-thickness. The thickness of the shell
elements is equal to the thicknesses of the bolt-head and nut, as shown in Fig. 4. The bolt shaft in
actual bolts typically consists of two portions; (a) the bolt shank, i.e. the unthreaded part and (b)
the threaded part. Both portions are represented with Hughes-Liu two-node beam elements
(elform=1) with a diameter equal to the nominal diameter of the bolt. The junction between the
beam elements of the bolt shaft and the shell elements of the bolt head is realized with constraint
equations using the *CONSTRAINT_NODAL_RIGID_BODY (CNRB) keyword. The
constraint nodes of the shell elements in the bolt-head and nut extend to an area that is equal to
the area covered from the bolt shaft, as shown in Fig. 4(c). The CNRB definitions also include
nodes at the ends of the beam elements to account for the actual deformable part of the bolt shaft
as shown in Fig. 4(b). That is needed because the clear distance between the bolt head and nut,
also known as grip length, is defined by the total thickness of the connecting parts and the
thickness of the washers. The washers are modelled as separate parts that interact with the other
components of the connection, i.e. bolt and connected plates. Fig. 5 shows the numerical model
of the bolt/nut including the washers. The thinning option in the *CONTROL_SHELL card
(istupd=4) was used since the elements at the vicinity of the bolt hole experience high plastic
strains and Poissons ratio effect become influential.

1-4 June 12-14, 2016

14th International LS-DYNA Users Conference Session: Connections

shells

nut
CNRB

thread
beam

deformable
grip
length

shank

CNRB
head

CNRB area

(a) actual (b) plan view (c) perspective view (d) extruded view
bolt/nut

Fig. 4: Finite element model of bolt/nut, (a) actual bolt/nut, (b) plan view, (c) perspective view,
(d) extruded view

washer

(a) plan view (b) perspective view (shell

thicknesses enabled)

Fig. 5: Finite element model of the bolt/nut including washers, (a) plan view, (b) perspective
view
The plates that are attached together with the bolts are represented with fully integrated shell
elements. The bolt-hole should always be centered to the bolt-hole shaft to avoid any initial
penetrations between them. The reference plane of the shell elements is always placed at the
mid-thickness of the plates. Fig. 6 shows a side view of a complete single bolted connection
finite element model including the plates. Around the bolt-holes of the plates, very thin null
beam elements are used to account for the contact of the bolt shaft with the bolt-hole as shown in
Fig. 7. The null beam elements do not have any structural contribution in the connection
response and are implemented in LS-DYNA as regular beams that use the *MAT_NULL
material definition. Null beams are also placed at the washers to account for the contact between
the bolt shaft and the washers.

June 12-14, 2016 1-5

Session: Connections 14th International LS-DYNA Users Conference

washer (a) side view bolt reacting plate

loaded plate

(b) side view (shell thicknesses enabled)

Fig. 6: Side view of the complete connection model

null beams

Fig. 7: A view of the plate depicting the null beam elements placed around the bolt-hole
Interaction of the Connecting Parts
The individual connection components interact with each other by defining two separate contact
groups. The first group includes the beam elements of the bolt shaft and the null beam elements
that are placed around the bolt-hole only. This contact group explicitly simulates the bearing
stresses of the bolt shaft against the bolt hole using the *CONTACT_AUTOMATIC_GENERAL
keyword. The second contact group includes the interaction of all remaining parts, i.e. the
connected plates, bolt-head, nut, and washer using the *CONTACT_SINGLE_SURFACE
keyword. In both contact definitions a static and dynamic friction coefficient of 0.5 and 0.4 are
used respectively (fs=0.5, fd=0.4). Applying a viscous damping coefficient of 10-20% for the
contact helps in eliminating high-frequency oscillations [4]. The contact option that ignores
initial penetrations (IGNORE=2) is used because in cases where relatively thick and small
elements are used, self-contact between the shell elements might result in early termination of
the analysis without any meaningful results.
Bolt Pretension
Pretension in the bolt is applied using thermal contraction. The principle is to numerically shrink
the bolt shaft by cooling it enough to result in the desired tension. That is an iterative process and
usually a couple of iterations will result in finding the required temperature drop that results in
the desired pretension in the bolt shaft. LS-DYNA does not currently have non-iterative
preloading types for Hughes-Liu beam elements and therefore the only available option for bolt
pretension is thermal contraction.

1-6 June 12-14, 2016

14th International LS-DYNA Users Conference Session: Connections

Material Constitutive Behavior

A number of material models in LS-DYNA can be used to simulate the behavior of the
connection components. In this study the *MAT_SIMPLIFIED_JOHNSON_COOK was used to
represent the steel plates and *MAT_PIECEWISE_LINEAR_PLASTICITY was used for the
bolts. It is crucial to define the material law using actual experimental data as a benchmark for
replicating physical uniaxial tension coupon tests with numerical tension coupons. The objective
is to identify the correct true stress-strain parameters so that the resulting engineering stress-
strain curve from the analysis is similar to the benchmark curve. This calibration process tends to
be sensitive to the specific geometry and gage length and it is therefore suggested that numerical
coupon geometries match those of the experimental tests.
Material Damage Criteria
An important aspect of the modeling procedure is the implementation of advanced damage
models that are able to simulate the post-yield behavior of steel until rupture and capture
behaviors associated in extreme loading conditions. In this study, damage is modelled using the
GISSMO (Generalized Incremental Stress-State dependent damage Model) damage model. A
detailed description of GISSMO is provided by Effelsberg et al [5]. In the analyses performed
herein two different failure and two different mesh regularization criteria were employed.
The first failure criterion is triaxiality-dependent which associates the plastic strain at fracture of
the shell elements with the level of triaxiality of each element. That is defined as a triaxiality
curve (LCSDG) in GISSMO that associates the effective plastic strain at fracture for different
triaxiality ratios. The dependence of fracture strain of steel with triaxiality has been proven by a
number of researchers such as Johnson and Cook [8] but the experimental data available for
different types of metals are still very limited.
The second failure criterion is the introduction of strain-rate dependent scaling factors for
equivalent plastic strain to failure. A curve (LCSRS) defines scaling factors for the plastic strain
at fracture at the different strain rates. This accounts for the fact that steel fractures at lower
strains as the load rate increases. Data for these damage parameters were based on the
experimental work of Whittington et al. [6].
The last two criteria basically implement mesh regularization for different mesh sizes. Those
require the definition of two additional curves. The first one scales the equivalent plastic strain to
failure with the element size (LCREGD). In general the scaling factors increase as the element
sizes decreases. The second one defines a fading component that depends on the element size
and accounts for the strain localization (FADEXP) at higher strain rates. Values for these two
criteria were adopted from the work of Ozamut et al. [7] that focused on identification of these
parameters for steel.

Remarks and Limitations of the Simplified Modeling Procedure

A number of parameters in the modeling procedure were found to have appreciable influence on
the connection response. Those are discussed in this section and limitations associated with the
simplified modeling procedure are also presented.
Bearing Forces
In an actual bolted connection or a detailed finite element model with brick elements, the bearing
forces of the bolt shaft with the bolt holes are distributed over the contact area between the bolt
June 12-14, 2016 1-7
Session: Connections 14th International LS-DYNA Users Conference

shaft and the bolt holes, as shown in Fig. 8(a). Depending on the thickness of the connected
plates and the diameter of the bolt, the contact area can increase significantly. On the other hand,
when the proposed simplified connection model is used, this contact area is reduced to a contact
edge, which consists of the shell elements around the bolt-hole that are in contact with the bolt
shaft as shown in Fig. 8(b) and Fig. 9. Therefore the distribution of bearing forces due to that
contact is different compared to the actual one and this results in a different stress distribution.
This difference does not typically cause any issues at relatively low loads, but at higher loads,
such that the connection is about to rupture, this can lead to significantly different results. A
good approach that has been proven to provide better results is to always have the reference
planes of the shell elements at their mid-thicknesses (nloc=0).

(a) detailed model with solid elements

(section cut)

extruded view
(b) simplified model

Fig. 8: Bearing forces, (a) detailed model with solid elements, (b) simplified model

bolt shaft

null beams

bolt head

Fig. 9: Interaction of the bolt-shaft with the null beams in the simplified connection models

Triaxiality Damage Criterion

An important parameter is the definition of a triaxiality-dependent failure criterion, which affects
the connection response, especially when plate tear-out fracture is modelled. Bearing of the bolt-
shaft with the bolt-hole results in an uneven triaxiality state at the vicinity of the bolt-hole. An
example of the triaxiality state of a plate that is bearing against a bolt is shown in Fig. 10(a). Fig.
10(b) shows the Von Misses stress distribution at the same state. It is noticeable that although the
stresses at the bearing side of the bolt-hole are fairly uniform, the triaxiality ratios are
considerably different. The definition of the fracture strain as a function of triaxiality ratio

1-8 June 12-14, 2016

14th International LS-DYNA Users Conference Session: Connections

eventually dictates which of those elements fail first and results in a more realistic block shear
failure as shown in Fig. 11(a). On the other hand, if the failure strain is independent of the
triaxiality ratio, the failure mode is not realistic and is shown in Fig. 11(b). In that case elements
fail right at the bearing surface between the bolt-shaft and the bolt-hole which also affects the
obtained capacity of the connection.

(a) Triaxiality ratio (b) Von-Misses stresses

Fig. 10: Triaxiality ratio and von-Mises contours on a plate that is bearing against a bolt (bolt
not shown for clarity)

(a) triaxiality-dependent (b) triaxiality-

fracture independent fracture

Fig. 11: Plate tear-out fracture, (a) triaxiality-dependent fracture, (b) triaxiality-independent
fracture

June 12-14, 2016 1-9

Session: Connections 14th International LS-DYNA Users Conference

Comparison with Experimental Tests

The static and dynamic pendulum tests of the connections were simulated to validate the
proposed simplified modeling approach. For the dynamic tests a separate detailed finite element
model of the pendulum testing was created as shown in Fig. 12. The connection specimens were
represented with the simplified modeling approach presented herein and the remaining
components of the test setup were included, such as the loading frame and the pendulum. For
each test, the peak axial force that was developed at the specimen and the corresponding impact
velocity of the pendulum were measured. Fig. 13 shows a comparison of the peak axial force
versus the impact velocity of ten (10) of the dynamic pendulum tests against the response that
was calculated with the finite element models. Each data series represents the same specimen
configuration that was tested under different impact velocities. It is evident that these finite
element models were able to capture the variation of the peak axial force under different impact
velocities within 15%. Fig. 14 shows a comparison of the failure modes that were obtained with
LS-DYNA and the failure modes of the physical tests. Both failure modes, i.e. plate tear-out
fracture and bolt shear were quite similar to the failure modes that were observed during the
physical tests.

loading FE model
frame

2250-lbf
connection pendulum
specimen

Fig. 12: Finite element model used to replicate the pendulum tests

1-10 June 12-14, 2016

14th International LS-DYNA Users Conference Session: Connections

120

100

LC1A
Peak Force (kips)

80
LC1A_dyna
LC2A
60
LC2A_dyna
LC1B
40
LC1B_dyna
LC2B
20
LC2B_dyna

0
0 50 100 150 200 250
Impact Velocity (in/s)

Fig. 13: Peak force versus impact velocity, experimental data and LS-DYNA simplified
simulations

(a) plate tear-out

Physical Test LS-DYNA

(b) bolt shear fracture location

Fig. 14: Comparison of failure modes in the simplified bolted connection specimens modelled
with LS-DYNA and the physical tests, (a) plate tear-out, (b) bolt shear

June 12-14, 2016 1-11

Session: Connections 14th International LS-DYNA Users Conference

Summary
This paper presented a simplified modeling approach that can be used to model bolted
connections under static and dynamic loading conditions. This approach can be used in large
models with multiple connections and has been validated against static and dynamic test data.
This methodology is proven capable of simulating the behavior of bolted connections until
rupture with good accuracy. Crucial aspects of the modeling behavior for accurate simulation of
the connection response and prediction of the failure modes associated with connection failure
were discussed.

Acknowledgment

The authors would like to acknowledge Navistar Defense, LLC, for the contributions of
materials used in testing and technical advice during the execution of this project.

References

[1] K.K. Nakalswamy, Experimental and numerical analysis of structures with bolted joints subjected to impact
load, n.d.
[2] S. Narkhede, N. Lokhande, B. Gangani, G. Gadekar, Bolted joint representation in LS-DYNA to model bolt
pre-stress and bolt failure characreristics in crash simulations, in: 2010.
[3] U. Sonnenschein, Modelling bolts under dynamic loads, in: 2008.
[4] Contact parameters LS-DYNA Support, (n.d.). http://www.dynasupport.com/tutorial/contact-modeling-in-ls-
dyna/contact-parameters (accessed September 18, 2015).
[5] J. Effelsberg, A. Haufe, M. Feucht, F. Neukamm, P. Du Bois, On parameter identification for the GISSMO
damage model, in: Dearborn, MI, USA, n.d.
[6] W.R. Whittington, A.L. Oppedal, S. Turnage, Y. Hammi, H. Rhee, P.G. Allison, et al., Capturing the effect of
temperature, strain rate, and stress state on the plasticity and fracture of rolled homogeneous armor (RHA)
steel, Mater. Sci. Eng. A. 594 (2014) 8288. doi:10.1016/j.msea.2013.11.018.
[7] B. Ozarmut, H. Richter, A. Brosius, Parameter identification of a damage model for the process chain forming
crash, in: 2014.
[8] G.R. Johnson, W.H. Cook, Fracture characteristics of three metals subjected to various strains, strain rates,
temperatures and pressures, Eng. Fract. Mech. 21 (1985) 3148. doi:10.1016/0013-7944(85)90052-9.

1-12 June 12-14, 2016