Journal of Building Engineering 99 (2025) 111530

Contents lists available at ScienceDirect

Journal of Building Engineering

journal homepage: www.elsevier.com/locate/jobe

A reduced three-dimensional rotational Iwan model for nonlinear

behavior of semi-rigid joints
Jinxin Kuang a , Zhiming Hao a,b,*, Hao Chen a , Xiao Liu a,b, Naixian Li a , Jicheng Li a
a
Institute of Systems Engineering, China Academy of Engineering Physics, Mianyang, 621999, China
b
Shock and Vibration of Engineering Materials and Structures Key Laboratory of Sichuan Province, Mianyang, 621999, China

A R T I C L E I N F O A B S T R A C T

Keywords: The investigation into the rotational behavior of semi-rigid joints under loading is of paramount
Semi-rigid joints importance in the realm of structure design and assessment, given their extensive utilization in
3D nonlinear rotation connecting structural components. In this study, we proposed a novel 3D rotation Iwan model
3D rotational Iwan model
capable of capturing the complex behavior of semi-rigid joints subjected to multi-directional
Nonuniform stiffness degradation
Discretization
loads, and developed ellipsoidal failure criteria and a Weibull distribution density function to
derive specific theoretical expressions governing the behavior of these joints. To facilitate prac­
tical implementation, we devised two distinct methodologies. Firstly, a parameter identification
approach based on fine finite element numerical testing was established to calibrate the theo­
retical model parameters. Secondly, a novel method leveraging equal strength discretization was
developed to facilitate finite element simulations of joints equipped withed the 3D rotation Iwan
model based on the stiffness equations of the theoretical model. The findings demonstrate the
efficacy of the proposed model in accurately characterizing the rotational behavior of joints under
multi-directional loads. Moreover, a comparative analysis with fine finite element models
demonstrated that our approach effectively addresses the challenges associated with difficult
convergence due to an excessive number of contact pairs, and significantly improves computa­
tional efficiency, with an 87.7 % increase.

1. Introduction

Connection structures play a pivotal role in various industries such as energy transmission and architecture. These structures often
rely on semi-rigid joints to facilitate the transmission of forces and moments between structural members, ensuring overall structural
integrity. Common examples of semi-rigid joints include web angle connections, top-and-seat angle connections, and end-plate con­
nections, etc. [1]. Throughout their service life, these joints display multidirectional nonlinear behavior in response to complex loads,
exhibiting interactions between these behaviours, including material yielding and significant deformations [2–6]. However, Existing
research has primarily focused on the nonlinear behavior of steel frame joints in planar configurations [7–16]. It is important to also
consider the impact of complex loads present in actual space frame structures, as the beams, columns, and joint regions may experience
out-of-plane rotations and torsion. Traditional full-scale testing methods face limitations in replicating multidirectional bending loads
on large joints. Additionally, the computational cost associated with detailed finite element analyses is prohibitively high. This
challenge is particularly pronounced in large steel structures with numerous joints and intricate connection arrangements, where

* Corresponding author. Institute of Systems Engineering, China Academy of Engineering Physics, Mianyang, 621999, China.
E-mail address: [email protected] (Z. Hao).

https://doi.org/10.1016/j.jobe.2024.111530
Received 15 April 2024; Received in revised form 26 November 2024; Accepted 6 December 2024
Available online 9 December 2024
2352-7102/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

achieving convergence in simulations involving a large number of contact pairs between members poses significant difficulties. Hence,
there is a pressing need to develop high-precision reduced-order 3D rotational models capable of accurately capturing the nonlinear
rotational behavior of joints under multidirectional loads.
The past decades have been carried out on semi-rigid joints for steel structures, and numerous achievements have been obtained. In
the early days, the rotational behavior of semi-rigid joints was usually characterized by a simple mathematical function fitting to
represent the relationship between bending moment and rotation angle, known as an empirical model. Examples of mathematical
models include the linear, bilinear, multilinear, and non-linear models with power, polynomial, and exponential forms [17–22].
Nevertheless, many empirical equation parameters lack clear physical meaning and are only applicable to specific connection forms,
limiting their universality. Yee et al. [23] proposed an analytical method based on elastoplastic theory to predict the rotational
characteristics of joints and analyze the failure modes of connections. Subsequently, this method underwent gradual development,
evolved into a systematic modeling approach for semi-rigid joints [24–27], and was incorporated into standards such as EC3, AISC, and
GB50017-2017 [28–30], commonly known as the component method model. The joints have been divided into tension regions,
compression regions, and shear regions according to the force characteristics of the joints, and each of the regions is composed of a
number of components [31]. The component method model can directly predict the initial rotational stiffness and ultimate bending
moment of the joints, but it overestimated the stiffness. Additionally, various simplified models for semi-rigid joints have been
developed, including the scissors model [32], which characterizes the relative rotation of beams and columns, the centerline model
[33], which allows for independent rotation of beams and columns at both ends of a joint, and the improved panel zone model [34],
which considers both the deformation of the panel zone and the relative rotation of the beams and columns. However, the usefulness of
these plane rotation models is still limited when applied to complex spatial structures [35].
Initially, Iwan [36] characterized the elastoplastic behavior of metallic materials using a series of ideally elastoplastic parallel
Jenkins elements (spring-slider), i.e., commonly referred to as the Iwan model. This model has been widely employed to describe
nonlinear phenomena such as interface slip in bolted joints. Subsequent advancements have led to various adaptations of the original
model to suit specific applications. For example, Segalman et al. [37] established a four-parameter Iwan model, while Brake [38]

Fig. 1. Rotational Iwan model.

2
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

proposed an Iwan model considering the collision between overlapping plates and screws, which was applied to the analysis of
Brake-Reuß beams. Further enhancements include the work of Li and Hao [39], who proposed a six-parameter Iwan model featuring a
two-pulse non-uniform density function, which can well describe the residual stiffness of joint and the power relationship of energy
dissipation effectively. Song et al. [40] extended the applicability of the Iwan model by developing 2D and 3D beam element models,
which facilitated the analysis of nonlinear dynamic responses in overlapping bolted connection structures. A significant advancement
was made by Liu and Hao [41], who for the first time introduced a 1D rotation Iwan model in the steel frame joint which contains a
series of parallel rotational Jenkins elements and describes the moment-rotation relationship of semi-rigid joints under in-plane
bending conditions, thus broadening the scope of the Iwan model’s applicability in bolted joints. However, despite these advance­
ments, there remains a need for a 3D rotation Iwan model that considers the nonlinear behavior of the joints under multi-directional
loads to provide a more comprehensive description of joint behavior.
This paper presents the establishment of a 3D rotation Iwan model, which aims to capture the nonlinear rotational behavior of semi-
rigid joints subjected to multi-axial loads. The model is utilized for numerical simulation through discretization. In Section 2, we
introduce the 3D rotation Iwan model for semi-rigid joints, and derive the general analytical expression of the moment-rotation
relationship. Section 3 outlines the methodology for determining model parameters through numerical simulation tests, while also
analyzing the influence of loading conditions on the bending behavior of the joints and conducting comparisons with the existing joint
model. To facilitate finite element simulation, Section 4 proposes an equal strength discretization method. The discretized 3D rotation
Iwan model is then compared with the theoretical results. Section 5 elaborates the conclusions of this paper.

2. The 3D rotational Iwan model

According to the 1D rotation Iwan model, the moment-rotation relationship of beam-column joints is expressed by a series of
parallel rotational Jenkins elements [41], all of which are assumed to be ideal elastoplastic elements, as shown in Fig. 1 (a). With the
increase of bending moment, the rotation Jenkins elements yields gradually, and the bending stiffness of the joints decreases subse­
quently. As shown in Fig. 1 (c), for a joint under multidirectional moment, parallel 3D Jenkins elements are used to characterize the
spatial rotation behavior of the joint region, and all 3D Jenkins elements satisfy the following assumptions.

(1) The rotational deformation in a specific direction for each 3D Jenkins element is equal and corresponds to the rotation angle at
the joint in that particular direction;
(2) The 3D Jenkins element represents an ideal elastic-plastic element, with all elements sharing the same yield function
formulation;
∑
(3) The rotational stiffness of a joint is equal to the sum of the stiffness of all 3D Jenkins elements, i.e., S = nj=1 sj , where S refers to
the rotational stiffness of the joint and sj stands for the rotational stiffness of the jth 3D Jenkins element;
∑n j
(4) The total failure moment of a joint is equal to the sum of the yield moments of all 3D Jenkins elements, i.e., Mu = j=1 mp ,
j
where Mu refers to the ultimate moment of the joint and mp stands for the yield moment of the jth 3D Jenkins element.

The cumulative bending moment at a joint is obtained by summing the bending moments of all three-dimensional Jenkins ele

Fig. 2. Yield surface of 3D Jenkins element.

3
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

ments. As the number of Jenkins elements tends to infinity, the fundamental expression of the 3D rotational Iwan model can be
represented as follows:
∫∞
( ) ( )
M(θ) = m θ, mp ϕ mp dmp (1)
0

where θ refers to the rotation angle of the joint, M stands for the total bending moment of the mode, m represents the bending moment
( )
of the 3D Jenkins element, mp is the bending moment at yield of the 3D Jenkins element and ϕ mp is to the density function of the
bending moment at yield mp .

2.1. The bone line equation of the 3D rotational Iwan model

The yield surface of 3D Jenkins element in moment space is written as several concentric ellipsoidal surfaces, in accordance with
the strength criteria defined for beam-column structures under multi-directional loads [42,43], as shown in Fig. 2. The yield criterion
for 3D Jenkins subjected to 3D bending moments can be written as:
⎛ ⎛ ⎞αi ⎞1/β
3 j
⎝ mi ⎠ ⎠ − 1 = 0(j = 1, 2, …, n)
( ) ∑
j
fp mi = ⎝ j
(2)
i=1 mj
pi

j j
where mi refers to the bending moment of the jth 3D Jenkins element rotating along i direction, m j stands for the yield bending
pi
moment of the jth 3D Jenkins element rotating along i direction, αi and β are the shape parameters of the yield surface.
If 3D Jenkins elements are assumed identical in terms of yield strength, then:

mjpj mjpj
δjr,s = r
, δjr,t = r
(j = 1, 2, …, n) (3)
mjpj mjpj
s t

j j j j
where m j , m j and m j are the plastic moments of the jth 3D Jenkins element along r, s and t directions, respectively, δjr,s and δr,t stand
pr ps pt
for the scaling coefficients, r ∕ =s∕ = t and r, s and t ∈ [1, 3].
Substitute Eq. (3) into Eq. (2) to obtain the relationship between the bending moment of a 3D Jenkins element and its yield moment
in each direction:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎛ ⎞αs ⎛ ⎞αt̅
√ j
j j
√
√ r δ r,s m δ j
r,t m t
mjr = mjpj √1 − ⎝ j s ⎠ − ⎝ j ⎠ = ωjr mjpj 0 < ωjr ≤ 1 (4)
α

r mj mj i
pr pr

j
where ωr refers to the coupling coefficient of the jth 3D Jenkins element.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎛ ⎞αs ⎛ ⎞αt̅
√ j j
√
√ δ j
r,s m δ mt
j
ωjr = √1 − ⎝ j s ⎠ − ⎝ r,t j ⎠
αr
(5)
mj mj
pr pr

i.e.
⎡ ⎤
⎡ ⎤
j mj
⎢ pj1 ⎥
⎢ m1 ⎥ [ ]⎢ ⎥
⎢ j ⎥
ωj3 ⎢ mpj2 ⎥ = ωmp
⎢ j ⎥
m = ⎢ m2 ⎥ = ωj1 ωj2 (6)
⎣ ⎦ ⎢ ⎥
mj3 ⎣ j ⎦
mj
p3

As the applied load escalates, some 3D Jenkins components begin to yield, resulting in a loss of bending stiffness at the joint and a
consequent progressive decline in its overall bending stiffness. The total bending moment at the joint, denoted by Eq. (1), is partitioned
into the summation of elastic and yield components.
∫ sθ ∫∞
( ) ( )
M(θ) = ωmp ϕ mp dmp + Sθϕ mp dmp (7)
0 sθ
⏟̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏟ ⏟̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏟
yielding elastic

Let’s consider the coupling effect of the multidirectional moment on joint failure only and ignore the effect of the loading state on
( )
the bending stiffness of the joint, that is, eliminate the bending stiffness in Eq. (7). Let φ = mp /s, ρ(φ) = − s2 ϕ mp , and make all 3D
Jenkins elements have the same coupling coefficient ω. Substitute it into Eq. (7) to obtain:

4
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

∫θ ∫∞
M(θ) = ω φρ(φ)dφ + θρ(φ)dφ (8)
0 θ
⏟̅̅̅̅̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅̅̅̅̅⏟ ⏟̅̅̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅̅̅⏟
yielding elastic

Eq. (8) shows the moment-rotation equation of the joint under 3D bending. To be specific, the coupling coefficient ω can be ob­
tained from the failure function of the joint. In 3D moment space, the limit bending moment of the beam-column joint can be written as
a multi-directional moment function, and the limit failure criterion of the joint can be written as:
( )А )1/В
3 (
∑ Mi i
Fp (Mi ) = − 1 = 0(j = 1, 2, …, n) (9)
i=1
Mui

where Mi refers to the bending moment of the joint rotating around i direction. Mui stands for the ultimate bending moment of joint
rotation around i direction. Аi and В are the shape parameters.
In order to obtain the relationship between the joint bending moment and the limit bending moment, Eq. (9) is transformed into the
following equation:
⎡ ⎤ ⎡ ⎤
M1 Mu1
M = ⎣ M2 ⎦ = [ ω1 ω2 ω3 ]⎣ Mu2 ⎦ = ωMu (10)
M3 Mu3
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )Аs ( )Аt
where, ωr = 1 − MMus − MMut ,r∕
=s∕
= t and r, s and t ∈ [1, 3].
Аr

s t

2.2. Density function of 3D rotational Iwan model

Eq. (8) represents the 3D rotational Iwan model, wherein the density function ρ(φ) denotes the failure distribution of 3D Jenkins
elements. The density function ρ(φ), in essence, delineates the rate of stiffness degradation during joint rotation, which is equivalent to
the second derivative of the moment-angle relationship curve with respect to the rotation [32]. The degradation trend of the joint’s
bending stiffness is non-uniform from the onset of rotation. Throughout this process, the 3D Jenkins elements gradually undergo
yielding. Upon the failure of the joint, all 3D Jenkins elements yield simultaneously. The power and exponential functions are com­
bined into a density function of a form that is similar to that of a Weibull distribution. This distribution is used to accurately represent
the yield rotation distribution of the 3D Jenkins elements, which provide an effective description of the failure behavior of the joint
under complex loading. Mathematically, this can be denoted as follows:
⎧
⎪
⎪ ΔSi k
⎨ Mdi k φ ki − 1 e− Mpi φi i , φ ≥ 0
di i i i
ρ(φi ) = pi (11)
⎪
⎪
⎩ 0 ,φ < 0 i

where i = 1, 2, and 3, indicating that the joint rotates along the i direction; Mpi refers to the yield moment of the joint rotating along the
i direction; di stands for the amplitude parameter of yield moment, representing the amplitude range of plastic moment; ΔSi represents
the range of bending stiffness declining of the joint, ΔSi = Sei − Spi ; ki refers to the shape parameter, which is used to characterize the
declining speed of bending stiffness of joints; Spi stands for the residual stiffness when the joint is judged to be a failure.

Fig. 3. Density function of Weibull distribution.

5
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

According to the basic characteristics of Weibull distribution density function, the variations in the density function curves under
diverse combinations of the amplitude parameter di and the shape parameter ki are minor, as shown in Fig. 3. The φ1i refers to the
initial angle of the joint yielding, and φ∞i is the limit angle of the joint judged as a failure. Since θ = φ1i , the joint starts yielding, the
bending stiffness of the joint begins to decline, and the declining rate drops gradually. The declining rate decreases, and approaches
zero infinitely with the increase of the angle of rotation, as exemplified by Curve 1 in Fig. 3. Variations in the loading state lead to
alterations in the moment-rotation correlation of the joints, reflected in modifications in the amplitude parameter di and shape
parameter ki of the curves. While some moment-angle curves exhibit linearly elasticity at the initial stage of loading, the rate of decline
in bending stiffness markedly accelerates post-yielding, eventually tapering off as the angle progresses, as depicted by Curve 2 in Fig. 3.
When the yield angle φ1i of the joint approaches zero, the joint manifests nonlinearity throughout the entire rotation process, as shown
by Curve 3 in Fig. 3.
By substituting Equation (11) into Equation (8), the expression for the moment-rotation relationship of semi-rigid joints can be
derived as follows:
( )
1
di +1+
ki [ ( ) ( )]
ωi Mpi 1 1 ΔSi ki Mp (di +1) − ΔSi k
θ i
Mpi i
Mi (θi ) = ( ) Γ − Γ , θi + (1 − ωi ) i θi e (12)
ki 1 ki ki Mpi ΔS
1+
ki
ΔSi

where, Γ(., .) represents an incomplete gamma function.

The first derivative of Eq. (12) to θi can be used to obtain the corresponding stiffness equation:
( )
θi i [ ]
ΔSi k
M(d
p
i +1) −
Mp ΔSi
Si (θi ) = i e 1 − (1 − ωi ) ki θi ki (13)
i

ΔSi Mpi

Eq. (12) represents the analytical expression of the 3D rotational Iwan model that considers the loading state, and characterizes the
moment-rotation relationship of the semi-rigid joint. The mechanical response of semi-rigid joints can be divided into three stages: (1)
Elastic stage (OA), when joints perform linear elastic rotation and no 3D Jenkins element yields; (2) Yield stage (AD), when joints get
involved in nonlinear elastoplastic rotation and 3D Jenkins elements start to yield gradually; (3) Failure stage (after Point D), which
occurs when the joint reaches the specified rotation limit. At this point, the joint’s contribution to structural stability loss is considered
as a failure, and all 3D Jenkins elements yield simultaneously. It should be noted that even after exceeding the specified limit rotation
angle, the joint may still possess the ability to resist rotational deformation, which is accounted for by the residual bending stiffness
parameter Spi (Spi ≥0). The model expression contains six parameters with definite physical meaning, such as ωi , Mpi , Sei , Spi , di and ki .
The characteristic curve parameters of the whole rotation process can be identified by the tangent BD of the secant OB and point D, as
shown in Fig. 4. When OB exactly happens to be the tangent of the starting point, φ1i = 0, that is, the whole rotation process of the joint
is nonlinear. In the equation, ωi represents the influence of the loading state on the mechanical response of joints. Mpi , Sei and Spi can be
obtained directly from the intersection coordinates of the secant OB and BD, respectively, and they represent the plastic moment and
the flexural stiffness degradation level of the joint. di refers to the amplitude parameter characterizing the amplitude moment level of
the curve, and ki indicates the flexural stiffness degradation rate of the joint.

Fig. 4. Moment-rotation relationship of semi-rigid joints.

6
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

3. Parameter identification of 3D rotational Iwan model

First of all, M-θ curves of joints in different working conditions were obtained through a series of simulations, and the characteristic
point data of the curves were extracted. The model expression was developed using the nonlinear regression method. The specific
parameter identification steps are as follows.

1. The corresponding yield coupling coefficient ωi was calculated from Eq. (10) based on the loading state of the nodal region.
2. The M-θ curves of joints were obtained experimentally, and initial bending stiffness Sei , residual stiffness Spi and yield moment Mpi
were identified by drawing secant lines using the method shown in Fig. 4.
3. The substitution of the coordinate values pertaining to the transition point between the elastic and plastic stages (referred to as
Point A in Fig. 3), the transition point between the plastic and failure stages (referred to as Point C in Fig. 3), and the characteristic
point associated with the plastic bending moment (referred to as Point D in Fig. 3) into Eq. (12), along with the parameters derived
from step 1 and step 2, yields a set of nonlinear equations involving the parameters di and ki . Thus, the nonlinear equations are
solved to determine the values of the parameters di and ki .

3.1. Experimental verification of finite elements

Firstly, a finite element model was established against SC3 specimens in Literature [11] to analyze the mechanical response of
beam-column joints without stiffener overhanging endplates under unidirectional static load, and the effectiveness of the model was
also verified by the test results. Fig. 5 shows the SC3 specimens, with both beams and columns made of Q345 H-shaped steel and the
beam and column connected by 10.9 M20 high-strength bolts.
Fig. 6 shows the solid element model of the beam-column joint established. The bolt was modeled using an elastoplastic triple-
broken line approach, while the beam, column, and end plate were represented using an ideal elastoplastic model. The material
parameters can be found in Ref. [11]. The surface-to-surface contact form with the penalty function method and a friction coefficient of
0.44 was used between the end-plate and column flange plates, as well as between the bolt, column flange, and end-plate. Fig. 7 shows
the comparison between the simulation results and the test results shown in the literature [11]. It can be observed that the moment
response curve obtained from the numerical model aligns well with the experimental curve, demonstrating the model’s accuracy.

Fig. 5. The SC3 specimens.

7
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Fig. 6. The fine-FEM Model.

Fig. 7. Comparison FEM with experiment.

3.2. Finite element numerical experiment and parameter identification

A series of 12 simulation tests were performed to investigate the mechanical response of joints under different load combinations,
taking into consideration the influence of the loading state. The influence of different M3 values on the rotation behavior of joints along

Table 1
The finite element numerical experimentation results.
Num. Bending Moment in X-Direction (M1/ Bending Moment in Y-Direction (M2/ Bending Moment in Z-Direction (M3/ Loading
kN⋅m) kN⋅m) kN⋅m) State

M0T0 300 0 0 1D
M1T0 278 0 24 3D
M2T0 240 0 48 3D
M3T0 199 0 72 3D
M4T0 138 0 96 3D
M5T0 0 0 120 2D
M0T1 291 15 0 1D
M1T1 257 15 24 3D
M2T1 217 15 48 3D
M3T1 169 15 72 3D
M4T1 91 15 96 3D
M5T1 0 15 104 2D
M0T2 0 30 0 1D

8
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Direction 1 was compared using two sets of specimens: M0T0-M5T0 and M0T1-M5T1, respectively, when M2 = 0 kN m and M2 = 15
kN m. All joints were subjected to force loading, and the direction coordinates of the bending moment acting on joints are consistent
with those in Fig. 1 (c). Conditional loads M2 and M3 were applied first, followed by the load M1. In particular, specimens M0T0, M1T0,
and M0T2 are the working conditions when the joint is subjected to unidirectional load. From these specimens, the failure moments for
the joint rotation along directions 1, 2, and 3 under unidirectional bending can be derived: Mu1 = 300 kN m, Mu2 = 30 kN m, and Mu3 =
120 kN m. The primary mode of failure of a joint under torque is due to excessive rotation of the members, even though the joint may
not have fully yielded. In such cases, the beam has lost its stability. Consequently, the torsional failure moment is determined to be 30
kN m.
(( )1.68 ( )1.55 ( )2.48 )2
|M1 | |M2 | |M3 |
Fp (Mi ) = + + − 1=0 (14)
300 30 120

Table 1 presents the results obtained from a series of simulation tests. The failure moment value along Direction 1 exhibits a
decreasing trend as the bending moment in other directions increases. The joints in Table 1 have reached the limit state under all
loading states. Substituting the resultant bending moment into Eq. (9) enables the determination of the joint’s failure function
expression through regression analysis. According to the symmetry of geometric characteristics and mechanical characteristics of
joints, it is assumed that the failure surface of each quadrant symmetrically aligns with the coordinate plane. After extending the failure
surface symmetry to the other quadrants, the failure surface shown in Eq. (14) can be obtained. Fig. 8 shows the failure surfaces of
joints in moment space. The corresponding yield coupling coefficients ωi can be obtained based on the Step 1.
Tables 2–4 show the parameters of M-θ curves obtained using the method in Fig. 3. Tangent stiffness was employed for the initial
rotational stiffness of joints, i.e., φ1i = 0. With the increase of the lateral out-of-plane moment M3, the yield moment Mp1 and initial
bending stiffness Se1 decrease in in-plane vertical direction 1 under the same torque. For the same lateral out-of-plane moment M3, the
characteristic parameters of these M-θ curves will decrease with the increase in torque M2. The situation above implies that multi-
directional bending exerts a significant influence on the mechanical response of the joint. This behavior can be attributed to the
stress superposition effect on the joint section under multi-directional loads, which causes the bolt, end plate, or flange plate to reach
failure conditions prematurely. As a result, the joint bends more readily under multi-directional bending compared to unidirectional
bending. Consequently, beam-column joints designed using the conventional unidirectional loading method may experience damage
at loads lower than those anticipated during actual application.
Through the substitution of curve parameters and each group of test data presented in Table 2 into Equation (11) using the
methodology outlined in Step 3, we can derive nonlinear equations pertaining to the parameters di and ki . By employing nonlinear
optimization theory, the optimal combination of parameters di and ki can be obtained, subsequently yielding the theoretical curve.
Table 2, Tables 3, and Table 4 shows the parameters di and ki of each working condition model solved by a genetic algorithm. Fig. 9
shows the comparison between the theoretical curves obtained via regression analysis of each data group and the simulation test
results, indicating that the theoretical model can characterize the moment-rotation relationship of joints under different loading
conditions. Meanwhile, the out-of-plane moment exerts a significant influence on the ductility of joints rotating in the vertical plane.
With an increase in the magnitude of the out-of-plane moment M3, the ductility of joints rotating along Direction 2 experiences a
discernible decrease.
Fig. 10 presents a comparative analysis of the 3D rotational Iwan model with three alternative models, the three-parameter power
model, the three-parameter exponential model, the four-parameter exponential model, and the component model, with reference to
the FEM results of M0T0, M0T5, and M0T2 specimens. Specifically, it illustrates the moment-rotation curves of SC3 specimens

Fig. 8. The failure surface of joints in M space.

9
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Table 2
The parameters of joint M1-θ1 curves.
Num. Se1 /kN⋅m/rad Sp1 /kN⋅m/rad Mp1 /kN⋅m φ∞1 /rad ω1 d1 k1

M0T0 18115 510 271 0.065 1.000 2.544 0.951

M1T0 17638 239 266 0.059 0.989 2.564 0.923
M2T0 16766 239 232 0.043 0.937 2.632 0.944
M3T0 15471 232 193 0.036 0.821 2.706 1.004
M4T0 12150 143 136 0.026 0.601 2.836 1.121
M0T1 16652 196 281 0.064 0.780 2.477 1.022
M1T1 16553 178 250 0.053 0.767 2.554 1.019
M2T1 15841 111 213 0.042 0.705 2.636 1.051
M3T1 13408 95 166 0.035 0.559 2.731 1.121
M4T1 9165 88 89 0.021 0.228 3.078 1.201

Table 3
The parameters of joint M2-θ2 curves.
Num. Se2 /kN⋅m/rad Sp2 /kN⋅m/rad Mp2 /kN⋅m φ∞2 /rad ω2 d2 k2

M0T1 276 229 10 0.055 0.147 3.123 1.646

M1T1 276 161 12 0.057 0.364 3.184 1.586
M2T1 281 146 11 0.066 0.477 3.419 1.300
M3T1 285 61 12 0.087 0.497 3.474 1.277
M4T1 284 16 13 0.148 0.451 3.371 1.322
M5T1 272 9 13 0.169 0.459 2.580 1.008
M0T2 281 29 23 0.293 1.000 2.621 1.013

Table 4
The parameters of joint M3-θ3 curves.
Num. Se3 /kN⋅m/rad Sp3 /kN⋅m/rad Mp3 /kN⋅m φ∞3 /rad ω3 d3 k3

M1T0 1690 1514 13 0.014 0.425 3.928 1.976

M2T0 1733 1415 34 0.028 0.625 2.753 1.560
M3T0 1743 1159 57 0.044 0.754 2.430 1.320
M4T0 1752 490 79 0.078 0.880 2.381 1.091
M5T0 1810 81 104 0.243 1.000 2.283 0.926
M1T1 1765 1221 16 0.015 0.125 3.972 1.372
M2T1 1745 1115 36 0.031 0.359 2.892 1.301
M3T1 1743 792 56 0.052 0.598 2.577 1.225
M4T1 1749 235 84 0.096 0.771 2.379 1.090
M5T1 1647 27 102 0.257 0.845 2.256 0.984

subjected to vertical moment M1, torque M2, and out-of-plane moment M3 individually. The moment-rotation relationship of the 3D
rotating Iwan model in all directions can be matched with the FEM results，why is attributed to its adaptive density function, which is
adept at accurately reflecting the stiffness degradation within the joint. Conversely, the three-parameter power model approximates
the FEM curve effectively up to joint yield, yet exhibits a rapid decline in stiffness post-yield, resulting in a curve approaching a
horizontal asymptote swiftly. The residual stiffness of the joints considered in the four-parameter power model allows the asymptotes
of the curve to always have a certain slope, but it also results in the ultimate bending moment being greater than the FEM results.
Concurrently, the component model can predict the rotational performance of the joint, but the stiffness may be somewhat over­
estimated. Additionally, Eurocode 3 has not yet provided the component mode of the joint regarding torque M2 and out-of-plane
bending moment M3. Furthermore, the three-parameter exponential model overestimates the ultimate bending moment of the
joint. In contrast, while each model exhibits distinct strengths and weaknesses, the 3D rotational Iwan model emerges as the most
reliable in capturing the nuanced behavior of the joints under various loading conditions.

4. Discretization and application of 3D rotational Iwan model

Since the density function ρ(φ) in the theoretical model is continuous, the number n of 3D Jenkins elements approaches ∞, but a
limited number of 3D Jenkins elements are required in practical numerical calculation to obtain the rotational characteristics of
connected joints. The present study introduces a discrete methodology premised on the principle of uniform strength, entailing that
every discrete element possesses an identical yield moment. As illustrated in Fig. 10, the area encompassed by the stiffness curve and
the coordinate axis, as delineated by Eq. (13), is partitioned into n equi-areal sub-regions. Each sub-region corresponds to the yield
j
moment m j (i = 1, 2, 3) characteristic of the respective Jenkins element.
pi
The yield moment of the jth Jenkins element is equal to the area of the region Jj in Fig. 11. If the region is small enough, its area is

10
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Fig. 9. Comparison of theoretical curves and numerical results.

approximately trapezoidal. That is,

∫ j
/
Si
1( j )( )
mj j = Si − 1 (θi )dS = θi + θj−i 1 Sji − Sj−i 1 = Mui n (15)
pi
Si
j− 1 2

where j ∈ [1, n]; Si − 1 (θi ) is the inverse function of Si (θi ).

j
The yield moment m j of the 3D Jenkins element in each direction under different loading conditions can be calculated using Eqs.
pi
j
(14) and (15). In other words, using the equal-strength discretization method, m j can be determined through the limit bending
pi
moment of the joint and the number of Jenkins elements.
j
The bending stiffness si of the jth Jenkins element can be written as:

sji = Sji − Sj−i 1

(16)

Fig. 12 shows the FE model constructed based on a 3D rotational Iwan model. In the joint connection region, the contact surface of
the end plate and the column flange plate were coupled with reference points 1 and 2, respectively. Between these two points, a series
of n parallel 3D Jenkins elements were incorporated. The beams and columns were constructed with linear elastic materials with a
Young’s modulus of 200 GPa and a Poisson’s ratio of 0.3. Fig. 13 (a) shows the finite element calculation results of specimen M0T0 with
different numbers of elements. With the increase in the number of 3D Jenkins elements, the results of the FE model and the theoretical
model coincide with each other gradually. For n = 1 and n = 2, the simplified constitutive model curves of double and triple broken
lines were obtained, respectively. When n = 20, the results of the FE model basically coincide with the theoretical curves.
In order to establish the connection joints of the 3D rotating Iwan model taking into consideration the influence of loading con­
ditions, the corresponding stiffness curves presented in Tables 2–4 were discretized utilizing the equal strength method, with the
discretization number of the 3D Jenkins element set at 20 (n = 20). The initial stiffness parameters of 3D Jenkins elements were
discretely obtained from stiffness curves corresponding to M0T0, M5T0, and M0T2 under unidirectional loading conditions, and then
the constitutive parameters of 3D Jenkins elements under different loading conditions were matched by predefined field variables. The
simulation analysis was divided into two steps. In the first step, the conditional load Mr or Ms, or Mr & Ms was applied to the beam ends,

11
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Fig. 10. Comparison of results between rotation models.

Fig. 11. The discretization of stiffness curve.

j j
the bending moments of 3D Jenkins elements were read, and the corresponding bending stiffness si and yield moment m j were
pi
matched in order. Moving on the second step, the original conditional load Mr, Ms, or Mr & Ms was kept constant and the Mt was added
until the joints failed, here r ∕
=s∕= t and r, s and t ∈ [1, 3]. The model calculation steps are shown in Fig. 14.
As a result of the constraints imposed by the limited experimental conditions, the bending stiffness of the 3D Jenkins element under

12
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Fig. 12. The FE model of Beam-Column Joint Using 3D Rotational Iwan Model.

intermediate conditions was obtained using bicubic interpolation method [44]. Subsequently, the interpolated outcomes were
modernized across the predesignated field variables. In Fig. 13 (b), a comparative analysis is presented pertaining to the 3D rotational
Iwan model through four group verification scenarios. The discrete findings of M0T0, M2T0, M4T0, M0T1, M2T1 and M4T10 were
employed as interpolation data, juxtaposed against the theoretical results of M1T0, M3T0, M1T1, and M3T1 as the benchmark
verification data. Notably, a high degree of conformity was observed between the aforementioned sets of data, thereby validating the
model’s efficacy in characterizing the rotational behavior of joints subjected to 1D, 2D, and 3D bending loading conditions. The
computational analyses were executed on a formidable computational infrastructure comprising a 14-core processor system incor­
porating the 12th Generation Intel (R) Core (TM) i7-12700H processor with a 2.30 GHz base frequency and 16 GB of memory. The
processing time required by the CPU for the refined Finite Element model amounted to 11,785 s, whereas the 3D rotational Iwan model
executed the computations in 1447 s, signifying a notable enhancement in computational efficiency by 87.7 %.

5. Conclusion

A 3D rotation Iwan model featuring coupled failure was developed within this study to address the multidirectional nonlinear
rotational issue in semi-rigid joints. The rotational behaviors of joints under multidirectional loads were characterized using a set of
parallel 3D Jenkins elements complying with the ellipsoidal failure criterion. A Weibull distribution density function, incorporating
clearly defined physical parameters, was introduced and subsequently used to derive the specific expression of the model. A fine-FEM
model of bolted beam-column joints with overhanging endplates was established, and comparisons between the calculated outcomes
and experimental results were conducted to confirm the model’s efficacy. Following this, a series of simulation tests under different
loads were carried out, and a parameter identification method was proposed. Compared to the three-parameter and four-parameter
power models, the three-parameter exponential models, and the component models, the 3D rotation Iwan model was found to
accurately capture the variations in bending stiffness of semi-rigid connection joints during loading. Additionally, an equal-strength
discretization method was proposed, and the theoretical model was applied to finite element numerical simulations (Iwan-FEM). The

Fig. 13. Comparison of calculation results of 3D rotational Iwan model with theoretical curves.

13
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

Fig. 14. The calculation flow of 3D rotational Iwan model.

calculation results of Iwan-FEM are consistent with the results of fine-FEM, which could avoid the convergence problem of contact
nonlinear analysis and significantly improve the calculation efficiency.
The proposed 3D rotation Iwan model is a phenomenological model whose establishment and parameter identification are based on
the M-θ nonlinear relationship curve of the joint. The model can be theoretically applied provided that the form of the M-θ relationship
skeleton curve of the joint is described in Fig. 4, such as the reinforced concrete joints, bolted joints and welded joints that common
joints in the engineering. Moreover, it is evident that for a genuine spatial joint, the impact of coupling under significant spatial
deformation also necessitates examination. Future research will focus on extending the model to consider the coupling effects between
major and minor axis directions in spatial joints, further enhancing its practical applications in structural analysis.

CRediT authorship contribution statement

Jinxin Kuang: Writing – review & editing, Writing – original draft. Zhiming Hao: Writing – review & editing. Hao Chen: Soft­
ware. Xiao Liu: Validation. Naixian Li: Validation. Jicheng Li: Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgments

This work was supported by the Outstanding Young Scientist Foundation of Sichuan Province of China (grant No.
2023NSFSC1913). The support is gratefully acknowledged.

Data availability

Data will be made available on request.

References

[1] W.F. Chen, N. Kishi, M. Komuro, Semi-rigid connections handbook, J. Ross Pub (2011).
[2] N.S. Trahair, Buckling and torsion of steel angle section beams, Res. Rep. - Univ. Sydney, Dep. Civ. Eng. (2003) 1–31.
[3] N.S. Trahair, Biaxial bending of steel angle section beams, Res. Rep. - Univ. Sydney, Dep. Civ. Eng. (2002) 1–25.

14
J. Kuang et al. Journal of Building Engineering 99 (2025) 111530

[4] X. Cheng, Y. Chen, Ultimate strength of H-sections under combined compression and uniaxial bending considering plate interaction, J. Constr. Steel Res. 143
(2018) 196–207.
[5] X. Cheng, D. Duan, A. Ali, Y. Chen, Constitutive model for thin-walled H-sections bent about weak-axis considering local buckling, Structures 25 (2020) 56–71.
[6] X. Cheng, D. Duan, Y. Chen, Sectional constitutive model for the thin-walled steel H-section beam-column element, J. Constr. Steel Res. 156 (2019) 202–214.
[7] L.S. Silva, H. Craveiro, J. Conde, R. Simões, C. Martins, A. Catinaccio, D. Mladenov, J.B. Lopes, D.A. Feito, C. Bault, Experimental behaviour of large-size beam-
to-column steel joints for the DUNE Neutrino Experiment, Thin-Walled Struct. 191 (2023) 110996.
[8] J.D. Gao, H.X. Yuan, H. Qian, X.W. Chen, Structural behaviour and design of stainless steel end-plate beam-to-exterior column minor-axis joints, Thin-Walled
Struct. 181 (2022) 110119.
[9] J.D. Gao, X.X. Du, H.X. Yuan, M. Theofanous, Hysteretic performance of stainless steel double extended end-plate beam-to-column joints subject to cyclic
loading, Thin-Walled Struct. 164 (2021).
[10] Q. Yang, H. Ban, Balanced design philosophy of superior high-performance steel cover-plated beam-to-column joints, Thin-Walled Struct. 184 (2023).
[11] B. Gil, R. Goñi, T-stub behaviour under out-of-plane bending. I: experimental research and finite element modelling, Eng. Struct. 98 (2015) 230–240.
[12] B. Gil, R. Goñi, E. Bayo, Initial stiffness and strength characterization of minor axis T-stub under out-of-plane bending, J. Constr. Steel Res. 140 (2018) 208–221.
[13] B. Gil, R. Goñi, F. Bijlaard, E. Bayo, Major axis steel joint under torsion: stiffness and strength characterization, Eng. Struct. 180 (2019) 586–602.
[14] Y. Shi, G. Shi, Y. Wang, Experimental and theoretical analysis of the moment-rotation behaviour of stiffened extended end-plate connections, J. Constr. Steel
Res. 63 (2007) 1279–1293.
[15] G. Shi, Y. Shi, Y. Wang, Behaviour of end-plate moment connections under earthquake loading, Eng. Struct. 29 (2007) 703–716.
[16] G. Shi, Y. Shi, Y. Wang, M.A. Bradford, Numerical simulation of steel pretensioned bolted end-plate connections of different types and details, Eng. Struct. 30
(2008) 2677–2686.
[17] A.R. de Alvarenga, Plastic-zone advanced analysis – formulation including semi-rigid connection, Eng. Struct. 212 (2020) 110435.
[18] S.M. Maxwell, W.M. Jenkins, J.H. Howlett, A Theoretical Approach to the Analysis of Connection Behaviour, Joints in Structural Steelwork, 1981, pp. 2–49.
[19] P.D. Moncarz, K.H. Gerstle, Steel frames with nonlinear connections, J. Struct. Div. 107 (1981) 1427–1441.
[20] E. Attiogbe, A. Member, G. Morris, Moment - rotation functions for steel connections, J. Struct. Eng. 117 (1991) 1703–1718.
[21] C. Daz, P. Mart, M. Victoria, O.M. Querin, Review on the modelling of joint behaviour in steel frames, J. Constr. Steel Res. 67 (2011) 741–758.
[22] E.M. Lui, W.F. Chen, Analysis and behaviour of flexibly-jointed frames, Eng. Struct. 8 (1986).
[23] Y.L. Yee, R.E. Melchers, Moment-rotation curves for bolted connections, J. Struct. Eng. 112 (1986) 615–635.
[24] H. Yang, W. Liu, X. Ren, A component method for moment-resistant glulam beam-column connections with glued-in steel rods, Eng. Struct. 115 (2016) 42–54.
[25] G. Shi, X. Chen, Moment-rotation curves of ultra-large capacity end-plate joints based on component method, J. Constr. Steel Res. 128 (2017).
[26] S.N. Sadeghi, A. Heidarpour, X.L. Zhao, R. Al-Mahaidi, A component-based model for innovative prefabricated beam-to-hybrid tubular column connections,
Thin-Walled Struct. 132 (2018) 265–275.
[27] Ž. Bučmys, A. Daniūnas, J.P. Jaspart, J.F. Demonceau, A component method for cold-formed steel beam-to-column bolted gusset plate joints, Thin-Walled
Struct. 123 (2018) 265–275.
[28] I. Vayas, J. Ermopoulos, G. Ioannidis, Design of Steel Structures to Eurocodes, Springer Nature Switzerland AG, Cham, Switzerland, 2019.
[29] R. Tartaglia, M. D’Aniello, G.A. Rassati, J.A. Swanson, R. Landolfo, Full strength extended stiffened end-plate joints: AISC vs recent European design criteria,
Eng. Struct. 159 (2018) 155–171.
[30] Chinese Code, Standard for Design of Steel Structures. GB50017-2017, 2017.Beijing, China, (in Chinese).
[31] C. Faella, P. Vincenzo, G. Rizzano, Structural Steel Semirigid Connections: Theory, Design, and Software, CRC press LLC, USA, 2000.
[32] H. Krawinkler, S. Mohasseb, Effects of panel zone deformations on seismic response, J. Constr. Steel Res. 8 (1987) 233–250.
[33] D.A. Foutch, S.Y. Yun, Modeling of steel moment frames for seismic loads, J. Constr. Steel Res. 58 (2002) 529–564.
[34] H. Krawinkler, The State-Of-The-Art Report on System Performance of Moment Resisting Steel Frames Subjected to Earthquake Ground Shaking, Federal
Emergency Management Agency(FEMA), Washington D.C., 2000.
[35] J.M. Cabrero, E. Bayo, The semi-rigid behaviour of three-dimensional steel beam-to-column joints subjected to proportional loading. Part I. Experimental
evaluation, J. Constr. Steel Res. 63 (2007) 1241–1253.
[36] W.D. Iwan, A distributed-element model for hysteresis and its steady-state dynamic response, J. Appl. Mech. Trans. ASME 33 (1964) 893–900.
[37] D.J. Segalman, A four-parameter Iwan model for lap-type joints, J. Appl. Mech. Trans. ASME 72 (2005) 752–760.
[38] M.R.W. Brake, A reduced Iwan model that includes pinning for bolted joint mechanics, Nonlinear Dynam. 87 (2017) 1335–1349.
[39] Y. Li, Z. Hao, A six-parameter Iwan model and its application, Mech. Syst. Signal Process. 68–69 (2016) 354–365.
[40] Y. Song, C.J. Hartwigsen, L.A. Bergman, A.F. Vakakis, Dynamical response analysis of jointed structures by 2-D/3-D adjusted Iwan beam elements, 44th AIAA/
ASME/ASCE/AHS/ASC Struct. Struct. Dyn. Mater. Conf. (2003) 1–11.
[41] X. Liu, Z. Hao, Constitutive model for rotation behaviour of semi-rigid steel beam-to-column joints, J. Constr. Steel Res. 198 (2022) 107563.
[42] R. Verwaerde, P.A. Guidault, P.A. Boucard, A non-linear finite element connector model with friction and plasticity for the simulation of bolted assemblies,
Finite Elem. Anal. Des. 195 (2021) 103586.
[43] N.S. Trahair, Biaxial bending and torsion of steel equal angle section beams, Res. Rep. - Univ. Sydney, Dep. Civ. Eng. 133 (2005) 1–28.
[44] Y. Li, F. Qi, Y. Wan, Improvements on bicubic image interpolation, in: Proc. 2019 IEEE 4th Adv. Inf. Technol. Electron. Autom. Control Conf. IAEAC 2019, 2019.

15