Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Contents lists available at ScienceDirect

Computer Methods in Applied Mechanics

and Engineering
journal homepage: www.elsevier.com/locate/cma

A hydro-mechanical coupled contact method for two-phase

geotechnical large deformation problems within the
SNS-PFEM framework
Huangcheng Fang a, Zhen-Yu Yin a, *, Dingli Zhang b, Liqiang Cao c
a
Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China
b
Key Laboratory for Urban Underground Engineering of the Education Ministry, Beijing Jiaotong University, Beijing 100044, China
c
Institute of Geotechnical and Underground Engineering, Beijing University of Technology, Beijing, 100124, China

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

Keywords: This paper proposes a novel numerical algorithm to solve frictional contact problems in two-
Hydro-mechanical coupling phase porous media, accounting for the large deformation and interface fluid flow, within the
Large deformation framework of the Stable Node-based Smoothed Particle Finite Element Method (SNS-PFEM). The
Finite sliding contact
proposed approach builds on our previous work that was only designed for single-phase large
Dual Lagrange multiplier
Interface fluid flow
deformation contact problems. Within this work, the hydro-mechanical coupling system is dis­
Geotechnical engineering cretized using equal-order interpolation, with a polynomial pressure projection method to
eliminate the pore pressure instability. Surface-to-surface discretization is then applied to
formulate the contact constraint of the two-phase porous medium. Using the dual Lagrange
multiplier approach, a coordinate transformation is introduced to enforce the contact constraint
of the solid phase. This transformation is subsequently utilized to develop a new interface fluid
element, which serves as a tool to model the fluid exchange behaviour between porous media and
contact gaps. Different from existing interface fluid elements, the proposed element is constructed
based on the contact algorithm, thereby enabling it to handle non-conforming interfaces and
tackle challenges associated with large sliding problems. Furthermore, the proposed method can
maintain a well-conditioned system matrix throughout the computation process. These distinctive
attributes significantly enhance the flexibility of numerical modelling. Finally, the accuracy and
performance of our method are verified through four numerical examples. The results show that
the proposed method can effectively simulate various geotechnical problems, including multi-
layer soil consolidation, fluid-soil-structure interaction, and geological interface-controlled
slope failure.

1. Introduction

Geomaterials are usually characterized by microscopic pores and macroscopic interfaces. The fluid flow within these pores and
interfaces significantly influences the deformation and strength of the system. In addition to the fluid flow, geotechnical engineering
also involves large deformation and multi-body contact, which further complicates the problem. In the past decades, numerical

* Corresponding author.
E-mail addresses: huangcheng.fang@polyu.edu.hk (H. Fang), zhenyu.yin@polyu.edu.hk (Z.-Y. Yin), dlzhang@bjtu.edu.cn (D. Zhang), lqcao@
bjut.edu.cn (L. Cao).

https://doi.org/10.1016/j.cma.2023.116743
Received 1 September 2023; Received in revised form 24 December 2023; Accepted 30 December 2023
Available online 14 January 2024
0045-7825/© 2023 Elsevier B.V. All rights reserved.
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

solution strategies [1–4] based on the finite element method (FEM) have been extensively developed to solve such problems. However,
geotechnical large deformation analysis often suffers from the problem of mesh distortion, which may reduce the accuracy and
convergence of FEM-based approaches. The meshfree method [5–9] can eliminate the impact of mesh quality, but it introduces new
challenges in imposing boundary conditions and contact constraints, along with incurring high computational costs for computing
derivatives of physical fields [10].
Recently, the particle finite element method (PFEM) has become increasingly popular in geotechnical large deformation analysis
[11–14]. This method employs a standard finite element procedure to solve the governing equations. The mesh distortion is avoided by
using the Delaunay triangulation technique, where the element connectivity is frequently rebuilt over the existing set of nodes [15,16].
These features enable the PFEM to inherit both the flexibility of the meshless particle method in dealing with large deformation
problems and the rigorous mathematical theory of conventional FEM [17,18]. Since the PFEM stores the historical variables, such as
stresses and strains, at the Gaussian points, it is necessary to transfer these variables from the old mesh to the newly generated one.
Unfortunately, this process will inevitably introduce interpolation errors into the numerical results [15,18].
To overcome the above obstacle, a node-based gradient smoothing approach [10,19] was introduced into the PFEM, leading to the
development of NS-PFEM [20–22]. This method stores all historical variables in the nodes, thus avoiding the need for frequent variable
mapping. However, it still suffers from the issues of over-soft and temporal instability [18,23]. Therefore, a stabilization term is usually
required to increase the stiffness of the solid phase and eliminate the instability. The resulting framework could be referred to as the
stable NS-PFEM (SNS-PFEM) [18,24–26]. On the other hand, since the Delaunay triangulation would result in the loss of curvature
information for high-order elements, PFEM-based methods usually use the three-node triangular element [15]. In such cases, the fluid
pressure and solid displacement are discretized using equal-order interpolation, which may cause non-physical oscillations of pore
pressure in cases of low permeability and fast loading [27]. Thus, a stabilization technique is also required for the fluid phase.
In early geotechnical numerical methods, interface contact elements, such as the Goodman element [28] and thin-layer element
[29] were often used to model multi-body interactions. These methods usually require conforming interface meshes and are insuffi­
cient for modelling large sliding problems[30]. In the PFEM, a similar technique called the continuous constraint method [31] was
commonly utilized for contact analysis. This method uses an auxiliary mesh, generated in the remeshing process of PFEM, to perform
contact detection and impose contact constraints. Despite not requiring conforming meshes, the continuous constraint method
essentially belongs to the category of interface element method. Therefore, it also inherits the limitations of interface elements. On the
other hand, the node-to-surface (NTS) discretization [32,33], developed from the computational contact mechanics [34], has been
shown to outperform the interface contact elements in modelling large sliding. The advantage of this approach lies in its ability to
update the contact geometry relationship during nonlinear iterations. However, since the contact constraint is satisfied only in a
collocation manner, the NTS method is incapable of passing contact patch tests. To overcome this difficulty, a surface-to-surface (STS)
method [3,35–38] was proposed, where the contact constraint was discretized in a weak integral form. The NTS and STS methods
usually employ penalty [39–42] or Lagrange multiplier [43–47] techniques to enforce contact constraints. The resulting system is
either ill-conditioned or requires a time-consuming solver [30]. Fortunately, the recently developed dual Lagrange multiplier method
[48–51] provides a feasible tool to address the aforementioned issues. This method utilizes a static condensation technique to eliminate
the additional multiplier degrees of freedom [30]. The resulting system is well-conditioned and without an increase in system size.
When modelling two-phase geomaterials, it is also necessary to establish constraints for the fluid flow across the interface. This is a
challenging task, and only a few studies have proposed viable solution strategies [52]. In general, the contact constraints of the fluid
phase can be divided into two parts: the continuity of pore pressure at the closed interfaces, and the fluid exchange behaviour at the
opening interfaces. The former can be achieved directly using penalty or Lagrange multiplier methods with NTS/STS formulations [53,
54], in a form analogous to the implementation of solid contact constraint. However, neither penalty nor Lagrange multiplier methods
can directly describe the fluid flow in the contact gap.
In the existing literature, the fluid exchange behaviour was usually modelled by the interface fluid elements [55–57]. There existed
two commonly used elements: two-node and three-node interface elements [55]. The former uses the nodes, which coincide with the
boundary nodes of porous media, to represent the potentials on each side of the interface, while the latter additionally incorporates a
node to represent the potential of the fluid in contact gaps. Many numerical experiences have demonstrated that both of these elements
can effectively model the fluid exchange behaviour. However, the performance of these interface fluid elements is still limited by two
defects: first, they fall under the category of the element-type method, which is inadequate for large sliding problems, as previously
discussed; second, when the interface is closed, the fluid field on both sides of the interface should satisfy the continuity condition.
However, achieving this requires the transversal conductivity to approach infinity (similar to the penalty method), which can lead to
an ill-conditioned system. In addition to the interface element method, an alternative approach employing the extended finite element
method has gained popularity for modelling interface fluid flow [58–60]. This method utilizes a weakly discontinuous enhanced
interpolation [60] to discretize the pore pressure field, endowing it with the capability to capture the discontinuity of pore pressure
gradient across porous media and discontinuities. As a result, the utilization of dedicated elements to simulate the interface fluid
becomes unnecessary, thereby greatly enhancing the flexibility of numerical modelling. However, like the standard finite element
method, the challenge of element distortion induced by large deformation persists in the extended finite element method.
To address the aforementioned issues, a novel numerical algorithm is developed in this work to model the geotechnical large
deformation problems with hydro-mechanical interfaces. In our previous works, both explicit and implicit frameworks of SNS-PFEM
[61,62] were developed to solve the large deformation hydro-mechanical coupled problems. Furthermore, an improved SNS-PFEM
was also proposed in [26] to enhance the efficiency of computing the smooth strain field, accompanied by a single-phase contact
algorithm for addressing the solid contact problem. However, the aforementioned works primarily concentrated on the numerical
implementation of SNS-PFEM, without delving into two-phase contact algorithms. The current work builds upon the advancements

2
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 1. Hydro-mechanical coupling system with a discontinuity.

achieved in [26], directing its attention towards the numerical implementation of the hydro-mechanical coupled contact method.
Specifically, the contact constraint for the solid phase is discretized in a surface-to-surface form, and the discretization of the interface
fluid flow is accomplished through a novel interface fluid element. This novel element is formulated based on the coordinate trans­
formation facilitated by the dual Lagrange multiplier method. The resulting method can handle both the continuity condition at the
closed interface and the fluid exchange behaviour at the opening interface.
Compared to the existing methods, our approach will offer the following advantages: (1) the proposed interface fluid element is free
from the matrix singularity caused by the interface closure, thus avoiding potential numerical instability; (2) our contact algorithm
excels in handling non-conforming meshes, establishing itself as a flexible modelling tool; (3) with the aid of the SNS-PFEM framework,
our method circumvents challenges associated with mesh distortion and time-consuming variable mapping, enabling effective
modelling of large deformation contact problems.
This paper is organized as follows: in Section 2, we present the mathematical description of two-phase large deformation contact
problems, encompassing the hydro-mechanical coupling formulation, the weak form of frictional contact, and the governing equations
of interface fluid flow. Section 3 elaborates on spatial discretization using the SNS-PFEM framework, along with the pore pressure
stabilization technique and remeshing strategy. In Section 4, we propose a hydro-mechanical coupled contact algorithm based on the
dual Lagrange multiplier method, providing detailed insights into the implementation of solid contact constraints and the construction
of interface fluid elements. Section 5 showcases four numerical examples designed to validate the accuracy, effectiveness, and
robustness of the proposed method. Finally, Section 6 concludes the paper with some final remarks.

2. Mathematical description of two-phase geotechnical large deformation problems

2.1. Hydro-mechanical coupling formulation

Fig. 1 illustrates a representative hydro-mechanical contact problem. The bodies are assumed to be saturated porous media, and the
discontinuity is considered to be filled with single-phase fluid flow. In Fig. 1, Ωm and Ωs denote the master and slave bodies in the
current configuration. Their counterparts in the initial configuration are distinguished with a superscript 0. The surfaces of porous
bodies are classified into five distinct types: displacement boundary Γiu , pore pressure boundary Γip , traction boundary Γiσ , flux
boundary Γiq , and contact boundary Γic , where i takes the values of s for the slave body and m for the master body. In the following
derivation, the symbols Ω and Γ without superscripts s or m represent the domain and boundary of the entire porous medium.
In this section, the Einstein summation convention will be employed for mathematical derivations. Considering the effective stress
principle and assuming an incompressible fluid that follows the linear Darcy’s law, the mechanical equilibrium equations for the solid
phase and the mass conservation equation for the fluid phase read
σij,i + bi = 0 (1)

ε̇v + qi,i = 0 (2)

with
kf ( )
σ ij,i = σ′ij,i − p,i δij , and qi = − p,i + γf h,i ; i, j ∈ {x, y} (3)
γf

where σ ij denotes the total stress, σ′ij the effective stress, p the pore pressure, b the body force, q the fluid velocity, ε̇v the volumetric
strain rate, kf the permeability coefficient, γf the fluid weight per unit volume, δij the Kronecker delta function, and h the elevation. The
boundary conditions of hydro-mechanical coupling problems are given by

3
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

ui = ui , on Γu ; p = p, on Γp ; σ ij nj = ti , on Γσ ; − qi ni = q, on Γq ; i, j ∈ {x, y} (4)

where ui is the displacement, nj the outward normal vector, ui the prescribed displacement, t i the prescribed traction, and q the
prescribed pore pressure.
The present work employs the PFEM framework to develop the large deformation solver, where an updated Lagrangian formulation
is utilized. By introducing the test functions δu and δp and imposing the boundary conditions, the weak form of the equilibrium
equation can be written as
∫ ∫ ∫ ∏ ∏
δεij σij t+Δt dΩ − δui bi t+Δt dΩ − δui ti t+Δt dΓ + δ c + δ p = 0; i, j ∈ {x, y} (5)
Ω Ω Γσ

and the mass conservation equation is as follows

∫ ∫ ∫ ∏
δpε̇tv dΩ − δp,i qti dΩ + δpqt dΓ + δ exc = 0; i, j ∈ {x, y} (6)
Ω Ω Γq

where the superscripts t and Δt are time labels; δΠc and δΠp represent virtual works arising from the effective contact stress and
t
interface fluid pressure, respectively; δΠexc denotes the fluid exchange between the porous medium and discontinuity; σ ij t+Δt = σ′ij +
δij pt + Δσ′ij + δij Δp is the total stress tensor at time t + Δt; εij is the Green strain tensor, defined as

1( ) 1
εij = eij + ηij , eij = ui,j + uj,i , ηij = uk,i uk,j ; i, j, k ∈ {x, y} (7)
2 2

where eij and ηij denote the linear and nonlinear terms, respectively. Besides, the volumetric strain is expressed as
εv = ev + ηv , ev = ux,x + uy,y , ηv = ux,x uy,y − ux,y uy,x (8)

where ev and ηv denote the linear and nonlinear terms, respectively. The nonlinear term is typically neglected in small deformation
analysis as it is a second-order small quantity.
In continuum mechanics, the material constitutive should not rely on the chosen frame of reference. Therefore, an objective stress
J
rate that accounts for the rigid body rotation effect is required [63]. In this study, the Jaumann stress rate σ̇′ij is utilized, leading to the
following stress update scheme [64]:
( ) ( )
(9)
′
σ̇′ij = σ̇ ijJ + σ′ik ⋅ ω̇kj + σ′jk ⋅ ω̇ki = Dijkl ⋅ ėkl + σ′ik ⋅ ω̇kj + σ′jk ⋅ ω̇ki ; i, j, k, l ∈ {x, y}

with a spin tensor:

1( )
ω̇ij = uj,i − ui,j (10)
2

where Dijkl refers to the constitutive matrix.

By substituting the incremental forms of Eqs. (7) ~ (9) into Eq. (5) and linearizing the formulation, we can obtain the final
equilibrium equation as follows [65]
∫ ∫ ∫ ( ) ∫
t t t
δeij Dijkl ⋅ Δekl dΩ + δηij σ′ij dΩ + δeij σ′ik Δωkj + σ′jk Δωki dΩ − δηij δij pt dΩ
(11)
Ω Ω Ω Ω
∫ ∏ ∏ ∫ ∫ ∫ ∫
t+Δt t
− δev ΔpdΩ + δ c + δ p = δui bt+Δt i dΩ + δu i t i dΓ − δeij σ ′ij dΩ + δev pt dΩ
Ω Ω Γσ Ω Ω

and the temporal discretization of the mass conservation equation as follows

∫ ∫ ∫ ∏ ∫ ∫ ∫
δpΔev dΩ + δpΔηv dΩ − θΔt δp,i Δqi dΩ + δ exc = − (1 − θ)Δt δpqt dΓ − θΔt δpqt+Δt dΓ + Δt δp,i qi t dΩ (12)
Ω Ω Ω Γq Γq Ω

where the use of 0.5 < θ ≤ 1 can achieve unconditional temporal stability. In this work, θ is taken as 1, leading to a backward Euler
method.

2.2. Frictional sliding contact

At the contact interface, the solid phase prohibits both normal penetration and tension, which can be mathematically expressed by
the Hertz-Signorini-Moreau condition [34]:

σ′n ≤ 0, gn ≥ 0, σ′n gn = 0 (13)

4
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

where the first inequality describes the non-tensile constraint, the second denotes the non-penetration condition, and the third rep­
resents the complementary relationship between the first two.
The tangential contact behaviour is governed by the Mohr-Coulomb friction law in this work, which can be described by the
Karush-Kuhn-Tucker condition [66] as follows
ψ ≤ 0, β ≥ 0, ψ β = 0 (14)

with

ψ = |τ| − (μ|σ′n | + c), β = − vτ /τ (15)

where τ is the frictional force, μ the friction coefficient, c the interface cohesion, and vτ the sliding velocity. The first inequality in Eq.
(14) sets an upper limit for the frictional force, the second specifies its direction, and the third is a complementary condition.
This work employs the Lagrange multiplier method to impose contact constraints. By introducing a Lagrange multiplier λ to
represent the negative effective contact stress, the weak forms of the normal non-penetration and tangential sticking conditions can
thus be expressed as
∫ ∫
δλn n ⋅ (xs − xm )dΓ ≥ 0, and δλτ τ ⋅ (us − um )dΓ = 0 (16)
Γc Γc

where n and τ are the unit normal and tangential vectors; xs and xm denote the current positions of a material point on the slave and
master surfaces; us and um represent the displacements of a material point on slave and master surfaces. Furthermore, the contact
virtual work can be written as
∏ ∫
δ c= (δus − δum )λdΓ (17)
Γc

2.3. Interface fluid flow

Research on interface fluid flow is less common than that on continuous porous media. As shown in Fig. 1, the interface fluid flow
occurs in both the longitudinal (L-axis) and transverse (T-axis) directions. For longitudinal flow, the generalized Darcy’s law is used to
describe the relationship between the flow velocity and fluid pressure gradient:
( ) kL
qL = − cL p,L + γf h,L , with cT = (18)
γf

where cL is the longitudinal conductivity, kL the longitudinal permeability, p the average pore pressure defined on the mid-plane, and h
the elevation of the mid-plane. The value of kL depends on the interface properties, such as the hydraulic aperture, surface roughness,
and infill material. This paper is primarily concerned with a clean and flat interface. Consequently, any contribution of infill materials
and surface roughness will be omitted. In this case, the longitudinal permeability can be determined by the cubic law [67–71]:
/
γf 3 γf 2
kL = w w= w (19)
12μf 12μf

where μf is the dynamic viscosity of the fluid, and w is the hydraulic aperture. This work assumes that the hydraulic aperture w is
equivalent to the contact gap gn. For the sake of simplicity and conciseness, we use gn to represent both the hydraulic aperture and
contact gap in the subsequent derivations.
For transverse flow, the leak-off model is adopted here to relate the pressure drop and flow velocity:
( ) ( )
qT = − cT p̆ + γf h̆ = − cT ps + γ f hs − pm − γ f hm (20)

where cT is the transverse conductivity, p̆ the pressure drop, h̆ the elevation drop, ps and pm the fluid pressures of slave and master
surfaces, hs and hm the elevations of slave and master surfaces. The foregoing Eq. (20) was originally proposed for two-node interface
fluid elements [56] and its accuracy was shown to be close to the three-node interface elements in previous studies [55].
The transverse conductivity cT in Eq. (20) is usually treated as a constant in the literature (e.g., [72] and [57]) to avoid the sin­
gularity of the permeability matrix when the hydraulic aperture approaches zero. However, this can lead to a discontinuous fluid
pressure field at the closed interface. To avoid this issue, this work employs an aperture-dependent transverse conductivity. As the
hydraulic aperture approaches zero, the transverse conductivity tends toward infinity, thereby ensuring the continuity of pore pressure
at the closed interface. For simplicity, a transverse conductivity inversely proportional to the aperture [73] is used here. By doing so,
Eq. (20) can be restated as follows
( )
( ) kT ps + γf hs − pm − γ f hm kT
qT = − cT p̆ + γf h̆ = − , with cT = (21)
γf gn γf gn

5
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

where kT is the transverse permeability coefficient. The above equation indicates that the fluid leakage rate is proportional to the
pressure gradient across the two contact surfaces. This implies the transverse flow also takes the form of Darcy’s law.
Analogous to Eq. (2), the mass balance equation within the discontinuity can be formulated as
χ̇ + qi,i = 0; i ∈ {T, L} (22)

where χ̇ is the change in storage capacity, estimated to be

( )
χ̇ = ėv + η̇v = (ε̇T + ε̇L ) + ε̇T ε̇L − uT,L uL,T = ε̇T + ε̇L + ε̇T ε̇L (23)

where ε̇T and ε̇L are the transverse and longitudinal strain rates. The above storage capacity rate is calculated by the volumetric strain
defined in (8). Since the subsequently proposed interface element is always parallel to the l-axis, the shear term uT,L is perpetually equal
to zero.
By multiplying Eq. (22) with a test function δp and applying the divergence theorem, the weak form of the mass conservation for the
interface fluid flow can be derived as follows [74]
∫ ∫ ∫
δpχ̇ t dΩ + δpqtn dΓ − δp,i qti dΩ = 0; i ∈ {T, L} (24)
Ωf Γf Ωf

where qtn represents the velocity of the fluid flowing from the porous medium into the discontinuity. Ωf, and Γf are the computational
domain and boundary of the discontinuity. The second term in Eq. (24) corresponds to the fluid exchange term in Eq. (6):
∏ ∫
δ exc = − δpqtn dΓ (25)
Γf

Applying Darcy’s law to Eq. (24) yields

∫ ∫ ∫
kL ( t ) kT ( t ) ∏
δpχ̇ t dΩ + δp,L p,L + γ f ht,L dΩ + δp,T p,T + γf ht,T dΩ − δ exc = 0 (26)
Ωf Ωf γf Ωf γf

By combining Eq. (26) with Eq. (6), the fluid exchange term δΠexc can be eliminated, leading to the attainment of global mass con­
servation [74].
When the transverse conductivity tends toward infinity, the conventional interface element method may result in a singular system
matrix. To address this issue, our approach employs the average pressure and pressure drop as degrees of freedom and discretizes the
interface flow on the mid-plane of the discontinuity. By doing so, we should reformulate the storage capacity rate in Eq. (26) as
∫ ∫ ∫ gtn /2 ∫
( ) ( )
δpχ̇ t dΩ = δp ε̇tT + ε̇tL + ε̇tT ε̇tL dgn dΓ ≈ δp ġtn + ε̇tL gtn + gtn ε̇tL dΓ (27)
Ωf Γd − gtn /2 Γd

the longitudinal flow as

∫ ∫ ∫ ∫
kL ( t ) gtn /2
kL ( t ) kL ( t )
(28)
t
δp,L p + γ f ht,L dΩ = δpt,L p + γ f ht,L dgn dΓ ≈ δp,L p + γ f h,L gtn dΓ
Ωf γf ,L Γd − gtn /2 γf ,L Γd γf ,L

and the transverse flow as

∫ ∫ ∫ gtn /2 ∫ ∫ gtn /2 ( ) ( )
kT ( t ) kT ( t ) ps − pm kT pts + γf hts − ptm − γ f hts
δp,T p,T + γf ht,T dΩ = δp,T p,T + γ f ht,T dgn dΓ ≈ δ t
dgn d
Ωf γf Γd − gtn /2 γf Γd − gtn /2 gn γf gn
∫
kT ( t)
Γ= δp̆ t p̆t + γf h̆ dΓ (29)
Γd γ f gn

where Γd denotes the mid-plane of the discontinuity. The discretization schemes for average pore pressure p and pressure drop p̆ will be
addressed in Section 4.2.
Furthermore, the virtual work of the interface fluid pressure acting on the porous medium can be expressed as follows
∏ ∫
δ p=− n ⋅ (δus − δum )pdΓ (30)
Γc

3. Hydro-mechanical coupling SNS-PFEM discretization

3.1. Enhanced node-based smoothing scheme

In order to avoid the frequent variable mapping in the PFEM, this work employs the SNS-PFEM framework to discretize the hydro-

6
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 2. Schematic of the node-based smoothing technique [26].

mechanical system. For the sake of completeness, our derivation will begin from the NS-PFEM formulation. The core of NS-PFEM is to
reconstruct the gradient field by using a node-based smoothing technique [17], as shown in Fig. 2a. Since the gradient smoothing
process for displacement and pressure is identical, we only present the derivation for the former. To prevent any potential confusion,
the Einstein summation convention is no longer employed, and the time label t is omitted in subsequent derivations.
For a given particle c, the smoothing cell Ωc is constructed by connecting the midpoints of edges to the centres of all neighbouring
elements. The smoothed displacement gradient at Ωc can subsequently be computed as follows
ec ∫
1 ∑ ∑ec
uci,j = uIi,j dΩ = αI uIi,j ; i, j ∈ {x, y} (31)
Ac I=1 Ωc I=1

where ec denotes the number of elements associated with particle c, uIi,j represents the standard FEM displacement gradient of element
I, and Ac is the smoothing cell area. For one-order triangular elements, the smoothing coefficient reads

AIe
αI = (32)
3Ac

where AIe is the element area.

As depicted in Eq. (31), the original NS-PFEM employs a constant gradient within the smoothing cell. However, it has been
observed in numerous applications that this constant gradient can result in over-softening and temporal instability. To address these
issues, various stabilization methods [26,75–78] have been proposed, culminating in the development of SNS-PFEM. In this work, the
stabilization technique with an enhanced gradient smoothing scheme [26] is adopted. As illustrated in Fig. 2b, the smoothing cell Ωc is
further approximated as a circle domain, which is divided into four sub-domains with four virtual integration points located on the two
axes. The circle radius is calculated as
√̅̅̅̅̅
Ac
rc = (33)
π
The displacement gradient is assumed to be linearly distributed in the smoothing circle, expressed as follows
uci,j = uci,j + uci,jx (x − xc ) + uci,jy (y − yc );
̃ i, j ∈ {x, y} (34)

where xc and yc represent the spatial coordinates of particle c. uci,jx and uci,jy are the second-order gradient of the displacement, which can
be approximated as
c
2 ∑ ec
AIe uIi,j − ui,j ∑ec
uci,jx ≈
̃ cosθIe = ΔαIx uIi,j
Ac I=1 3 lIe I=1
(35)
c
2 ∑ ec
AIe uIi,j − ui,j ∑ec
uci,jy ≈
̃ I
sinθIe = ΔαIy uIi,j
Ac I=1 3 le I=1

with
( ) ( ) ∑
ec
̃ x , ϕI = 2bI lI − 2 , ϕ
ΔαIx = αI ϕIx − ϕ ̃x = αI ϕIx
x e e

(36)
I=1
( ) ( ) ∑
ec
̃ y , ϕI = 2hI lI − 2 , ϕ
Δα = αI ϕIy − ϕ
I ̃y = αI ϕIy
y y e e
I=1

7
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

where bIe and lIe the length parameters of element I, as shown in Fig. 2c.
Substituting Eqs. (31) and (35) into Eq. (34) yields the enhanced displacement gradient as follows
ec [
∑ ] ∑
ec
uci,j =
̃ αI + ΔαIx (x − xc ) + ΔαIy (y − yc ) uIi,j = αIenh uIi,j ; i, j ∈ {x, y} (37)
I=1 I=1

The enhanced smoothing coefficient αIenh at the four virtual integration points can be expressed as
( I ) [ ( )] ( I ) [ ( )]
̃ x αI = ηI αI ,
αenh 1 = 1 − rc ϕIx − ϕ αenh 2 = 1 + rc ϕIx − ϕ̃ x αI = ηI αI
1 2
( I ) [ ( )] ( I ) [ ( )] (38)
I ̃
αenh 3 = 1 − rc ϕy − ϕy α = ηI3 αI ,
I I ̃
αenh 4 = 1 + rc ϕy − ϕy α = ηI4 αI
I

where ηI∗ is the correction factor. The above formula indicates that the current SNS-PFEM framework can be directly achieved by
multiplying the smoothing coefficient with a correction factor.
Following a similar procedure, the smoothed gradient of pore pressure can be calculated as
ec [
∑ ] ∑
ec
pc,j =
̃ αI + ΔαIx (x − xc ) + ΔαIy (y − yc ) pI,j = αIenh pI,j ; j ∈ {x, y} (39)
I=1 I=1

where pI,j represents the pore pressure gradient of element I in the standard FEM.
εcij for each particle domain Ωc. Subsequently, performing
By substituting Eq. (37) into Eq. (7), we can obtain the smoothed strain ̃
constitutive integration on the four virtual integration points yields the effective stress as follows
( )
σcij′ = f ̃εcij , ⋯
̃ (40)

where f (*) is the material constitutive relationship.

3.2. Spatial discretization

Temporarily ignoring the contributions of contact and interface fluid flow, and substituting the smoothed gradient fields from Eqs.
(37) and (39) into the equilibrium Eq. (11) and mass conservation Eq. (12), the final system matrix can be written in the following
iterative form:
[ ][ ] [ ] [ ]
K uu − K up Δu Fext − Fint ru
= = (41)
− (K up )T − K ηv − ΔtK pp Δp − (Qext − Qint ) − rp

where Kηv arises from the nonlinear term of the volumetric strain. To ensure matrix symmetry, this nonlinear term is neglected,
resulting in an approximate tangent matrix. Note that the nonlinear term of the volumetric strain is retained in the flux vector Qint. The
resulting system can be expressed as
[ ][ ] [ ]
Kuu − K up Δu ru
up T pp = (42)
− (K ) − ΔtK Δp − rp

where Kuu is the stiffness of the solid phase, expressed as

∫ ∫ ∫
K uu = K uu uu uu ̃ T DB
̃ LI dΩ + B̃ T σ̆ B
̃ NL dΩ + B̃T ̂ ′̃
(43)
LI + K NL + K RO = B LI NL LI σ BRO dΩ
Ω Ω Ω

in which the subscripts LI, NL, and RO represent the linear, nonlinear, and rotation terms, respectively; np is the number of the
smoothing cell; B
̃ is the enhanced smoothing gradient matrix; σ̆ and ̂
σ′ denote the total and rotation stress matrices, respectively. To
ensure symmetry, the rotation term K uuRO is often neglected in the existing literature [43].
For plane strain problems, the enhanced smoothing strain–displacement matrices of node I are expressed as
⎡ ⎤
⎡ ⎤ NI,x 0 ⎡ ⎤
NI,x 0 ⎢ 0 0
̃ I = αI ⎢ NI,y 0 ⎥
B̃ I = αI ⎣ 0 NI,y ⎦, B
enh ⎣ 0
̃ I = αI ⎣ 0
⎥, B 0 ⎦ (44)
LI enh NL NI,x ⎦ RO enh
NI,y NI,x 0.5NI,y − 0.5NI,x
0 NI,y

The stress matrices σ̆ and ̂σ are defined as

′

⎡ ⎤
σxx σ xy 0 0 ⎡ ′ ⎤
⎢ σyx σ yy 0 ⎥ ′ 2σ xx 0 2σxy
0
σ̆ = ⎢
⎣ 0
⎥, ̂
⎦ σ =⎣ 0 2σ′yy − 2σxy ⎦ (45)
0 σ xx σ xy
0 0 σ yx σ yy σ xy σ xy σ′yy − σ′xx

8
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Moreover, the hydro-mechanical coupling matrix Kup and permeability matrix Kpp are given by
∫
K up = B ̃ T mNp dΩ
LI (46)
Ω

∫
K pp = ̃ T kf B
γ−f 1 B p
̃ p dΩ (47)
Ω

with m and B
̃ p defined as
[ ]
T ̃I NI,x
m = [1 1 0] , B I
p = αenh N (48)
I,y

The residual traction and flux vectors ru and rp are presented as follows
∫ ∫ ∫ ∫
ru = NTu bdΩ + NTu tdΓ − ̃T ̃ ′ ̃ T mpdΩ (49)
B LI σ dΩ + B LI
Ω Γr Ω Ω

∫ ∫ ∫
rp = − Δt NTp qdΓ + Δt ̃ T qdΩ −
B Np Δεv dΩ (50)
p
Γv Ω Ω

where Δεv is the volumetric strain increment in the current iteration, calculated according to Eq. (8).

3.3. Polynomial pressure projection

This work employs equal-order interpolation to discretize the displacement and pore pressure. When the permeability is low or the
loading is fast, the term ΔtKpp in Eq. (42) can approach zero, leading to the oscillation of pore pressure. In such cases, the pressure
stabilization technique is required. Many stabilization techniques have been developed, such as the polynomial pressure projection
[79], enhanced strain formulation [80], and fractional step method [81]. This work selects the polynomial pressure projection (PPP)
method to stabilize the pore pressure, due to its simple implementation and high efficiency. The detailed derivation can be found in
[79]. Using the PPP stabilization, the system matrix in Eq. (42) can be further rewritten as
([ ] [ ])[ ] [ ] [ ]
Kuu − K up 0 0 Δu ru 0
up T pp − = − (51)
− (K ) − ΔtK 0 S Δp − rp C

with two stabilization terms defined as

∫ ∐ )( ∐ )
τ (
S= Np − Np Np − Np dΩ (52)
Ω 2G

∫ ∐ ) ∐
τ (
C= Np − Np (Δp − Δp)dΩ (53)
Ω 2G

where G denotes the shear modulus, τ the stabilization parameter, and Δp the pore pressure increment. The selection of the parameter τ
∐
has been extensively discussed in [79] and [82]. Besides, is a projection operator, which takes the form:
∐ ∫
1
p= p(x)dΩ (54)
A e Ωe

For convenience, it is assumed that the terms of ΔtKpp and rp in Eq. (51) already include the stabilization term, and the symbols S
and C will be omitted in the subsequent derivations.

3.4. Remeshing strategy

In large deformation analysis, elements may undergo squash, extension, or distortion. Consequently, the implementation of
remeshing becomes necessary. In the PFEM, the computational domain is defined as a collection of particles (i.e., mesh nodes). The
remeshing process only requires the reconstruction of element connectivity over the existing set of particles. Moreover, when adjacent
particles are either too close or too far apart, the insertion and removal of particles can also be executed. Many node adjustment
algorithms have been designed for the PFEM, interested readers can refer to [15,26,83,84]. This work employs the remeshing strategy
proposed in [26]. Since the SNS-PFEM stores all historical variables at the node/particle, the operation of node adjustment is very
simple and computationally cheap. Meanwhile, the variables mapping from the old mesh to the new one is unnecessary, thus elim­
inating the interpolation errors.

9
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

4. Hydro-mechanical contact method for non-conforming interfaces

4.1. Contact discretization for solid phase

To arrive at the discretization of contact virtual work and constraints, the Lagrange multiplier is approximated along the slave
surface using dual basis functions:
∑ns
λ= ϕ zi
i=1 i
(55)

where zi is the discrete nodal multiplier and ns is the slave node number. The Lagrange multiplier defined above represents the effective
contact stress. The dual basis ϕ is constructed by a biorthogonality condition at the element level [85], which is given by
∫ ∫
ϕi Nj dΓ = δij Nj dΓ; i, j = 1, 2, ⋯, nes (56)
Γec Γec

where Nj is the standard FEM shape function and nes is the node number of the surface element e. For 2D problems, the dual basis
function can be explicitly expressed as follows
1 1
φ1 = (1 − 3ζ), and φ2 = (1 + 3ζ) (57)
2 2

where ζ denotes the elemental coordinate.

By inserting Eq. (55) into Eq. (16), the discrete forms of contact constraints can be written as
∫ ∫ [∑
ns
(
∑
ns ∑
nm
)]
δλn n ⋅ (xs − xm )dΓ = δzni ϕi ni ⋅ Nj xsj − Nk xmk dΓ = 0 (58)
Γc Γc i=1 i=1 k=1

∫ ∫ [∑
ns
(
∑
ns ∑
nm
)]
δλτ τ ⋅ (us − um )dΓ = δzτi ϕi τ i ⋅ Nj usj − Nk umk dΓ = 0 (59)
Γc Γc i=1 j=1 k=1

Following a similar procedure, the discrete contact virtual work in Eq. (17) is derived as

∏ ∫ ∫ [( ∑ ns ∑
nm
)
∑
ns
]
δ c= (δus − δum )λdΓ = Nj δusj − Nk δumk ⋅ ϕi zi dΓ (60)
Γc Γc j=1 k=1 i=1

The integrations in Eqs. (58)~(60) are performed on the projection of the master surface onto the slave surface.
Inserting the contact constraint and virtual work from Eqs. (58)~(60) into the system Eqs. (42), we can obtain a system with the
fully tied condition as follows
⎡ ⎤⎡ ⎤ ⎡ ⎤
K RR K RM K RS − LRP 0 ΔuR rR
⎢ K MR K MM T T ⎥⎢ ⎥
K MS − LMP − M R ⎥⎢ ΔuM ⎥ ⎢ rM ⎥ ⎢
⎢ ⎥
⎢ K SR
⎢ K SM K SS − LSP D T RT ⎥ ⎢ ⎥ ⎢
⎥⎢ ΔuS ⎥ = ⎢ rS ⎥
⎥ (61)
⎣ − LPR − LPM − LPS − ΔtΦ 0 ⎦⎣ Δp ⎦ ⎣ − rp ⎦
0 − RM RD 0 0 z 0

where the subscripts S, M, and R denote the slave, master, and remaining nodes, respectively. R, D, and M are the rotation matrix and
two mortar matrices. The rotation matrix of node i is given by
Ri = [ ni τi ] (62)

in which ni is calculated by averaging the normal vectors of all edges connected to node i. τ i is computed as the orthogonal vector of ni.
Besides, D and M have the following nodal block form:
∫ ∫
Dij = ϕi Nj dΓ ⋅ I2 = δij Nj dΓ ⋅ I2 ; i, j = 1, 2, ⋯, ns
∫Γc Γc
(63)
Mij = ϕi Nk dΓ ⋅ I2 ; i = 1, 2, ⋯, ns , k = 1, 2, ⋯, nm
Γc

with a 2 × 2 identity matrix I2. Due to the dual basis functions, D is a diagonal matrix.
It can be seen from Eq. (61) that the Lagrange multipliers introduce zero diagonal terms and increase the system size, thereby
escalating the computational overhead associated with solving linear equations. Fortunately, the dual Lagrange method provides a
low-cost treatment to eliminate the additional multipliers, which is achieved by the following coordinate transformation [30,48,86]:
( )
α = R us − D− 1 Mum = R(us − Hum ) (64)

10
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

with a single-node form:

( )⎫
∑nm
⎪
αni = ni ⋅ usi − Hij umj ⎪
⎪
⎪
⎪
j=1
⎬
( ) ; i = 1, 2, ⋯, ns (65)
∑
n m ⎪
⎪
⎪
⎪
ατi = τ i ⋅ usi − Hij umj ⎪
⎭
j=1

where α is the relative displacement in the local Cartesian coordinate system.

Using the transformation defined in Eq. (64), the system equation can be condensed into
⎡ ⎤⎡ ⎤⎡ ⎤⎡ ⎤ ⎡ ⎤
I 0 0 0 K RR K RM K RS − LRP I 0 0 0 ΔuR rR
⎢ 0 I HT 0 ⎥⎢ K MR K MM ⎥⎢
K MS − LMP ⎥⎢ 0 I 0 0 ⎥⎢ ΔuM ⎥ ⎢ rM + H rS ⎥
⎥ ⎢ ⎥ ⎢ T
⎢
⎣ 0 0 R 0 ⎦⎣ K SR
⎥⎢ = ⎥ (66)
K SM K SS − LSP ⎦⎣ 0 H RT 0 ⎦⎣ Δα ⎦ ⎣ RrS − Dz ⎦
0 0 0 I − LPR − LPM − LPS − ΔtΦ 0 0 0 I Δp − rp

abbreviated as
[ ][ ] [ ]
̂
K − L Δ̂
u ̂r u
T = (67)
− L − ΔtΦ Δp − rp

Note that the main computational cost of matrix condensation is attributed to the inversion of D. However, this cost can now be
considered negligible due to the strict diagonalization of D.
For the convenience of imposing contact constraints, the slave nodes S are categorized into four subsets: active nodes А, inactive
nodes I, sticking nodes T, and sliding nodes L. These subsets satisfy the following complementary conditions:
A ∪ I = S, A ∩ I = ∅; T ∪ L = S, T ∩ L = ∅ (68)
In this way, the contact constraints for Eq. (66) can be imposed as follows
( )tr
⌢n
Δαni = − g i , ∀i ∈ A

Δατi = 0, ∀i ∈ I
(69)
zni = 0, ∀i ∈ T
/ ( )tr
⃒ ⃒ ⌢n
zτi = − vτi ⃒vτi ⃒ ⋅ μ z i , ∀i ∈ L

⌢n tr
where the trial contact gap ( g i ) and Lagrange multiplier ( z i )tr are calculated based on the configuration at the beginning of each
⌢

iteration:
( )tr ( )
⌢n ∑
nm
gi = ni ⋅ xsi − Hij xmj (70)
j=1

( )tr [ ]tr
(71)
⌢ ⌢n ⌢τ
zi = zi zi = D−ii 1 Ri ri

4.2. Interface flow discretization for fluid phase

In the existing literature, the interface fluid elements are typically constructed based on the underlying meshes of contact surfaces.
These methods are insufficient in cases of finite sliding where the discontinuous displacement is significant. This section aims to
develop a new hydro-mechanical interface element that is suitable for non-conforming interfaces and capable of handling finite sliding.
As delineated in Section 2.3, two key physical quantities associated with interface fluid flow require discretization: the average pore
pressure p and pressure drop p̆. Drawing an analogy with the relative displacement defined in Eq. (64), the nodal average pore pressure
can be expressed as
( )
1 1 s ∑ nm
s
pi = pi − p̆i = p + m
Hij pj (72)
2 2 i j=1

and the pressure drop can be written as

∑
nm ∑
nm
p̆i = psi − D−ii 1 Mij pmj = psi − Hij pmj (73)
j=1 j=1

11
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Subsequently, a standard FEM discretization is employed to reconstruct the pore pressure fields:
∑ns
p̆(x) = i=1
Ni (x)p̆i (74)
∑ns
p(x) = i=1
Ni (x)pi (75)

where Ni denotes the 1D linear interpolation. The interface element is positioned at the mid-plane of the discontinuity and uses the
same element connectivity as the slave surface.
This work posits an equivalence between the hydraulic aperture and the contact gap. Therefore, we use the normal relative
displacement of the contacting surfaces to characterize the change in the hydraulic aperture, which can be discretized as
∑ns
Δgn (x) = Δαn (x) = i=1
Ni (x)Δαni (76)

where the normal relative displacement αni is defined in Eq. (65). Moreover, the longitudinal strain of the interface element is
calculated as
∑ns
εL = i
u si
Ni,L ̂ (77)

s
where ̂
u i is the longitudinal displacement of slave node i, which is given by

̂ si = τ i ⋅ usi
u (78)

By substituting the discrete fields of Eqs. (74) ~ (77) and the interface flow formulations of Eqs. (27) ~ (29) into the mass con­
servation Eq. (26), we can obtain the following incremental form
( ) ( ) ( )
δpT LTf Δαn + δpT ΔtΦLf Δp + δp̆T ΔtΦTf Δp̆ − δΠexc = δpT rf (79)

with
∫
LTf = NT NdΓ (80)
Γd

∫
ΦLf = NT,L cL N,L gn dΓ (81)
Γd

∫ ∫
kT
ΦTf = NT NdΓ = NT cT NdΓ (82)
Γd γ f gn Γd

∫ ∫ ∫
δpT rf = − δpT NT χ gn dΓ + ΔtδpT NT,L qL gn dΓ + Δtδp̆T NT qT dΓ (83)
Γd Γd Γd

where Lf denotes the hydro-mechanical coupling matrix, ΦLf the longitudinal conductivity matrix, ΦTf the transverse conductivity
matrices, and rf the residual vector. The hydraulic aperture gn (i.e., contact gap) can be calculated using Eq. (70). Similar to Eq. (41),
the nonlinear term of storage capacity change is neglected on the left-hand side but retained on the right-hand side.
Inserting the interpolation of average pore pressure from Eq. (75) and the change in the hydraulic aperture from Eq. (76) into the
virtual work of the interface fluid pressure defined in Eq. (30), we can obtain
∏
δ p = − δαTn Lf p (84)

Before assembling the interface fluid flow into the system, it is necessary to unify the degrees of freedom. According to Eqs. (72) and
(73), we can obtain the following transformation

ps = p̆ + D− 1 Mpm = p̆ + Hpm (85)

1 1
p = p̆ + D− 1 Mpm = p̆ + Hpm (86)
2 2

where D and M are consistent with those of the solid contact as defined in Eq. (63), except that the identity matrix I2 should be reduced
to I1.
Applying the transformation of Eq. (85) to the system matrix of Eq. (67) yields

12
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

⎡ ⎤
⎡ ⎤ ̂ ⎡ ⎤⎡ ⎤ ⎡ ⎤
I 0 0 0 ⎢ K − LR − LM − LS ⎥ I 0 0 0 Δ̂u (̂r u )
⎢0 I ⎥ ⎢ − LT − ΔtΦRR − ΔtΦRM ⎥
− ΔtΦRS ⎥⎢ ⎥⎢ ⎥ ⎢ − r ⎥
0 0 ⎥⎢ 0 I 0 0 ⎥⎢ ΔpR ⎥ ⎢ ( )
(87)
p R
⎢ ⎢ R
⎥⎢ = T
( ) ⎥
⎣0 0 I H T ⎦ ⎢ − LT − ΔtΦMR − ΔtΦMM − ΔtΦMS ⎥ ⎣0 0 I 0 ⎦⎣ ΔpM ⎦ ⎣ − rp M (− H rp S ⎦
⎣ M ⎦ )
0 0 0 I 0 0 H I Δp̆ − rp S
− LTS − ΔtΦSR − ΔtΦSM − ΔtΦSS

abbreviated as
[ ][ ] [ ]
̂
K ̂
− L Δ̂
u ̂r u
̂ T
̂ = (88)
− L − Δt Φ Δ̂
p − ̂r p

Note that the above system equations temporarily neglect the fluid exchange term δΠexc. By applying the transformation of Eq. (86)
to the interface fluid flow of Eq. (79), and subsequently inserting Eq. (79) into the system equations, the fluid exchange term can be
completely eliminated. In this way, the final system equations can be expressed as
[ ( ) ][ ] [ ]
K̂ − L ̂+L ̂f Δ̂
u ̂r u
( )T ( ) = (89)
− L ̂+L ̂f − Δt Φ ̂ +Φ̂f Δ̂
p − ̂r p − ̂r f

with

L 2
̂ f = W T ΦLf W 2 + W T ΦTf W 1
̂ f = W T Lf , Φ
2 1
(90)
∫ ( ∫ ) ( ∫ )
̂r f = − W T2 NT χ gn dΓ + W T2 Δt NT,L qL gn dΓ + W T1 Δt NT qT dΓ (91)
Γd Γd Γd

[ ]
1
Δ̂
u = [ ΔuR ΔuM Δα ]T , Δp = [ ΔpR ΔpM Δp̆ ]T , W 1 = [ 0 0 I ], W 2 = 0 H I (92)
2
Remark. Our method is also applicable to the hydro-mechanical contact between the two-phase porous media and single-phase
solid. In such cases, it is only necessary to define the fluid pressure degree of freedom on the solid surface, without the need to dis­
cretize the fluid field inside the solid.

4.3. Discussion on the singularity of transverse permeability

For cases involving a closed interface, the transverse permeability matrix ΦTf should tend towards infinity to ensure the pore
pressure continuity. However, this may lead to the system matrix being singular. Fortunately, the proposed interface element employs
the pressure drop as the degree of freedom, enabling the direct elimination of singularity by normalizing the diagonal of the system
matrix. It is worth noting that this normalization is not applicable to the conventional interface element. A simple proof is provided
below:
In order to facilitate analysis, we reorganize the system matrix in the following form
⎡ ⎤
K L1 L2
⎢ T ⎥
S=⎢ ⎣ L1 Φ11 Φ12 ⎥
⎦ (93)
LT2 ΦT12 Φ22 + ΦTf

Applying diagonal normalization to the above matrix yields

⎡ ⎤
K̃ ̃1
L ̃2
L
˘ ⎢ ⎥
̃
S = GT SG = ⎢ ̃T ̃
⎣ L1 Φ11
̃ 12
Φ ⎥
⎦ (94)
T T
̃ ̃ ̃
L2 Φ12 Φ22 + ΦTf ̃

with a scaling operator G defined as

(√̅̅̅̅̅)− 1
Gij = Sii δij (95)

When ΦTf approaches infinity, the matrices L

̃ 2 and Φ
̃ 12 can be treated as zero, resulting in the degeneration of Eq. (94) to
⎡ ⎤
̃
K ̃1
L 0
⎢ T ⎥
̃ T ⎢
S = G SG ≈ ⎣ L̃
1
̃ 11
Φ 0 ⎥⎦ (96)
0 0 ̃ Tf
Φ

From the above equation, it becomes apparent that if Φ

̃ Tf is a full-rank matrix, the normalized system matrix will no longer be

13
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 3. Terzaghi consolidation: (a) problem definition; (b) non-conforming meshes.

singular.
According to Eq. (82), the elemental transverse permeability matrix of the proposed method can be explicitly calculated as:
∫ [ ] [ ]
kT kT L 2 1 L 2 1
ΦTf = NT NdΓ = = cT (97)
Γd γf gn γ f gn 6 1 2 6 1 2

where L is the element length. It can be seen that the resulting equations are linearly independent. Therefore, the singularity caused by
zero aperture can be eliminated through the diagonal normalization.
In conventional interface element methods, the pore pressure drop is also used to discretize transverse flow. However, these
methods do not directly treat the pressure drop as the degree of freedom; instead, they consider it as an intermediate variable [87]:
n+
∑ n−
∑
p̆ = Ni+ q+
i − Nj+ q+ + + − −
j = N q − N q (98)
i=1 j=1

where N+ and N− represent the shape functions for the top and bottom edges of the interface element; n+ and n− are the corresponding
node numbers; q+ and q− denote nodal pore pressure vectors. Then, the transverse permeability matrix can be expressed as:
∫
ΦTf = [N+ , − N− ]T cT [N+ , − N− ]dΓ (99)
Γd

Taking an interface element with length L as an example, Eq. (99) can be calculated as follows:
⎡ ⎤
2 1 − 2 − 1
L⎢ 1 2 − 1 − 2⎥
ΦTf = cT ⎢ ⎥ (100)
6⎣− 2 − 1 2 1 ⎦
− 1 − 2 1 2

The above matrix is rank deficient. When cT approaches infinity, both matrices in Eqs. (93) and (96) will become ill-conditioned.
Although assigning a finite value to cT can prevent the system matrix from becoming singular, it may lead to a deviation from the
precise fulfilment of pore pressure continuity at the closed interface. A specific numerical example showcasing the aforementioned
matrix singularity will be presented in Section 5.2.

5. Numerical examples

In this section, four numerical experiments will be conducted to validate the accuracy and performance of the proposed hydro-
mechanical contact algorithm. The numerical experiments include a benchmark test (i.e., one-dimensional Terzaghi consolidation)
and three typical geotechnical problems (i.e., strip footing penetration, shallowly buried foundation uplift, and delayed failure of slope
excavation). Except for the Terzaghi consolidation, all other examples employ the Mohr-Coulomb constitutive model and are simu­
lated under conditions of large deformation and finite sliding. The gravity effect is only considered in the last example.

5.1. One-dimensional Terzaghi consolidation

The Terzaghi consolidation problem is a commonly used benchmark for evaluating the newly developed hydro-mechanical
coupling method [88,89]. Here, we focus on validating the effectiveness of the proposed method in solving the contact problem of

14
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 4. Comparison of pore pressures: (a) without PPP stabilization; (b) with PPP stabilization.

Fig. 5. Comparison of consolidation degrees: (a) without PPP stabilization; (b) with PPP stabilization.

saturated porous media. To achieve this goal, two independent blocks with non-conforming meshes are employed for the numerical
simulation, as illustrated in Fig. 3. The left and right sides of the two blocks are fixed in the horizontal direction, while the bottom
surface of the lower block is fixed in the vertical direction. All three aforementioned surfaces are impermeable. Besides, the top surface
of the upper block is subjected to a uniform pressure of p0 = 10 kPa, with the pore pressure constrained to zero. In this example, a linear
elastic material with an elastic modulus of E = 100 MPa and Poisson’s ratio of ν = 0 is used for the solid phase, while the fluid phase is
modelled using the linear Darcy’s law with permeability coefficients of kx=1 m/s and ky = 10− 9 m/s. The unit weight of the fluid is
taken as γf = 10 kN/m3. The simulation is performed under the small deformation condition. The two blocks are tied together using the
proposed algorithm, where the interface aperture is considered to be 10− 16 m as a surrogate for zero. The analytical solutions [90] for
the pore pressure and degree of consolidation in the present problem can be expressed as
[ [ ]
4p0 ∑∞ (− 1)j− 1 π y] 2
2 π cv
p= cos (2j − 1) exp − (2j − 1) t (101)
π j=1 2j − 1 2H 4 H2
[ ]
u(H, t) − u(H, 0) 8 ∑∞ 1 2
2 π cv
D= =1− 2 2
exp − (2j − 1) t (102)
u(H, ∞) − u(H, 0) π j=1
(2j − 1) 4 H 2

with a consolidation coefficient

15
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 6. Pore pressure distribution at t = 10000s: (a) contour; (b) results on the contact surface.

Fig. 7. Footing penetration: (a) model setup; (b) finite element mesh.

kf E(1 − v)
cv = (103)
γ f (1 − 2v)(1 + v)

where H is the height of the sample, and u(H, t) denotes the displacement of the top surface at time t.
Fig. 4 presents the computed results of pore pressures, encompassing scenarios with and without the implementation of PPP
stabilization. In instances where PPP stabilization is not employed, significant oscillations in the pore pressure occur at the start of the
simulation. Conversely, with the implementation of PPP stabilization, precise results are achieved, aligning consistently with the
analytical solutions across all time steps. These results indicate that PPP stabilization can be seamlessly incorporated into our
framework without necessitating any adjustments. Fig. 5 depicts the evolution of the consolidation degree over time. It can be seen that
the use of PPP stabilization leads to a marginal reduction in the accuracy of the consolidation degree during the initial loading stage.
This is primarily attributed to the additional stabilization term disturbing the undrained condition of the initial loading stage.
Compared to pore pressure oscillations, this slight error in the consolidation degree can be considered negligible. Fig. 6a plots the
contour of pore pressure at t = 10000s, while Fig. 6b presents the pressure distribution on the contact interface. The pore pressure
distribution across the entire computational domain is observed to be smooth, without any non-physical oscillations near the contact
interface. Furthermore, the pore pressure values on the two sides of the contact interface are identical. These findings demonstrate that
our contact method has the ability to accurately transmit stresses across non-conforming interfaces.

16
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 8. Comparison of condition numbers between the two methods.

Fig. 9. Pore pressure distributions: (a) the conventional interface element method; (b) the proposed method.

17
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 10. Shear stress distributions: (a) the conventional interface element method; (b) the proposed method.

5.2. Footing penetration

The second example pertains to a common geotechnical problem of footing penetration, which has been extensively studied in
previous literature [18,89]. However, most of the existing investigations focused on homogeneous and continuous soil media. In
contrast, this example considers a soil mass with a cavity. The computational model is presented in Fig. 7, where the upper part is an
impermeable footing (i.e., a single-phase solid) and the lower part is a saturated soil mass. The cavity is located at a depth of 2 m,
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅
featuring an aperture of h = 52 − x2 − 2 6 m within the range of − 1 m < x < 1 m.
The penetration is achieved by imposing a vertical displacement of 1.0 m to the top surface of the footing in 20 incremental steps
over 3000 s. The soil is modelled using the Tresca constitutive model with an elastic modulus of E = 1 MPa, Poisson’s ratio of ν = 0.3,
and cohesion of c = 10 kPa; the footing is assumed to be an elastic material with E = 30 GPa and ν = 0.3; the friction coefficient and
cohesion on the contact interfaces are taken as 0.8 and 2 kPa; the Darcy permeability of the soil and the transverse permeability of the
interface fluid flow are both set to kf = 10− 5 m/s; the unit weight and dynamic viscosity of water are taken as γf = 103 kN/m3 and μf =
10 Pa•s, respectively. For comparison purposes, the present model will be calculated using both the proposed method and the con­
ventional interface element method. The transverse conductivity coefficient cT utilized in both methods is inversely proportional to the
hydraulic aperture, as expressed in Eq. (21). The finite element mesh is depicted in Fig. 7b, where matching meshes are utilized on the
upper and lower surfaces of the cavity to facilitate the generation of elements for the conventional interface method.
In Section 4.3, we have briefly discussed the matrix singularity in cases of zero aperture. Here, we further demonstrate the
singularity-free nature of the proposed method. Fig. 8 provides the condition number of the diagonally normalized system matrix
across all time steps. As the loading progresses, the hydraulic aperture gradually decreases, eventually leading to the interface closure.
In this example, the interface closure commences from the 11th time step. Therefore, the condition number of the conventional
interface element method undergoes a sharp escalation after this time step, reaching a notably high level of 1015. In contrast, the
condition number of our method remains at a low level throughout the simulation. This observation indicates that the proposed
method is free from the singularity, thereby preventing potential numerical instability.
Fig. 9 and Fig. 10 depict the distributions of pore pressure and shear stress obtained through the two methods. When no closure
occurs at the soil interface, both methods yield satisfactory results, displaying a symmetrical pattern in pore pressure and a smooth

18
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 11. Footing uplift: (a) model setup; (b) finite element mesh.

Fig. 12. Computed uplift resistance at the top surface of the footing.

distribution of shear stress. However, as the loading progresses, relative sliding occurs along the interface, leading to non-matching
meshes on the two contacting surfaces. Since the interface elements are constructed based on the initial conforming meshes of the
contact surfaces, their constraint accuracy decreases with the increasing extent of interface sliding. Additionally, the matrix singularity
induced by interface closure may also reduce the accuracy of numerical operations. As a result, we can observe that when the
penetration depth of the footing reaches 1 m, the result obtained from the conventional interface element method is unsatisfactory,
characterized by the asymmetric pattern in pore pressure and the non-smooth distribution of shear stress near the interface. On the
contrary, our approach consistently produces high-quality results throughout the simulation. This excellent performance primarily
stems from three aspects: firstly, our approach allows for the construction of interface fluid elements on non-matching meshes; sec­
ondly, the contact geometry relationship can be continuously updated during the computation; and thirdly, the system matrix can
remain in a well-conditioned state even in the event of interface closure.

5.3. Uplift of a shallowly buried footing

The third example is a challenging geotechnical problem of footing uplift, which involves the complex fluid-soil interaction, as well
as the nonlinearities arising from material behaviour, large deformation, and multi-body contact. Fig. 11 gives the geometric
configuration and numerical mesh of the computational model. In this example, both the soil and footing are modelled as two-phase

19
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 13. Comparison of results between high and low uplift velocities: (a) V = 200 mm/h; (b) V = 0.25 mm/h.

Fig. 14. Delayed failure of a cut slope: (a) model setup; (b) and (c) finite element meshes of two models.

porous media, with permeability coefficients of kf = 10− 6 m/s for the soil and kf = 10− 11 m/s for the concrete footing; the transverse
permeability of the interface is set to k = 5 × 10− 7 m/s; the unit weight and dynamic viscosity of water are taken as γf = 103 kN/m3 and
μf = 10 Pa•s. Furthermore, the soil behaviour is described by the Tresca constitutive model using parameters of E = 100 kPa, ν = 0.3,
and c = 1 kPa, while footing is assumed as an elastic material using parameters of E = 30 GPa and ν = 0.3. The friction coefficient and
cohesion on the soil-structure interface are set to 0.3 and 0, respectively. Other model information can be found in Fig. 11.
As is well-known, the uplift can generate negative excess pore water pressure at the footing bottom, resulting in a suction that
counteracts the movement of the structure. To showcase the capability of our method in tackling such problems, nine different uplift
velocities, i.e., V = 200 mm/h, 2 mm/h, 1 mm/h, 0.75 mm/h, 0.5 mm/h, 0.25 mm/h, 0.1 mm/h, 0.01 mm/h and 0.001 mm/h, are
simulated in this example. The uplift is attained by applying a vertical displacement of 0.24 mm to the top surface of the footing in 24
incremental steps.
In Fig. 12, the computed resistances, i.e., the reactions at the top surface of the footing, are plotted against the uplift displacement.
As the uplift velocity increases, the uplift resistance increases, but their relationship is not a simple linear proportionality. In theory,
there exist upper and lower limits for the uplift resistance. The former refers to the undrained strength, while the latter is the fully
drained strength. Fig. 13 displays the final distributions of pore pressure on the deformed models subjected to uplift velocities of 200
mm/h and 0.25 mm/h. It is seen that the high uplift velocity can cause a significant negative excess pore pressure, resulting in a large
suction at the footing bottom, which, in turn, causes the soil to bulge and adhere to the footing. In contrast, at low uplift velocities, the
soil separates from the footing without experiencing significant heave at the bottom, primarily due to the decrease in suction. Besides,

20
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 15. Evolution of total plastic strains over time.

Fig. 16. Comparison of the final collapse status: (a) with a geological interface; (b) without any geological interface.

the deformation of the upper soil layer under the low uplift velocity is considerably greater than that under the high uplift velocity.
Therefore, the contact nonlinearity of the former is more severe than that of the latter.

5.4. Delayed failure of a newly excavated slope

The delayed failure of the newly excavated slope is simulated in this example. As illustrated in Fig. 14, an inclined geological
interface is present within the stratum. The bottom surface is fixed in both directions, while the left and right surfaces are fixed
horizontally. An associated Mohr–Coulomb model is adopted to describe the material behaviour of soil, with properties of E = 10 MPa,
ν = 0.35, c = 5 kPa, and ϕ = φ =25◦ , where ϕ and φ are friction and dilation angles. The Darcy permeability of the soil and the
transverse permeability of the interface are both set to kf = 10− 5 m/s. Additionally, the friction coefficient and cohesion of the soil
interface are taken as 0.3 and 2 kPa, respectively. The gravity field is considered in this example, where the unit weights of the soil
skeleton and water are 20 kN/m3 and 10 kN/m3, respectively.
The entire numerical simulation consists of two distinct stages. In the first stage, the gravity load is applied to generate the initial
stress field. The period of this stage is 100 years, completed in one step. Before excavation, the displacement generated in the first stage
is reset to zero. In the second stage, the excavation is completed all at once in the first step, then only the time load is applied until the
slope fails. The period of this stage is 20 days, with a time step size of 0.2 days for the first five days, and 0.5 days for the following 15
days. Furthermore, a stratum without any interface is also simulated for comparison. The finite element meshes for both cases are given
in Fig. 14. To quantify the failure of the slope, a total plastic strain is employed in this example:
∫
εtotal = εep dΩ (104)
Ω

where εep is the equivalent plastic strain.

Fig. 15 illustrates the normalized total plastic strain of two models plotted against the time step. If the threshold for complete failure
is defined as reaching 98% of the maximum total plastic strain, the first model (with the interface) fails completely on the 7.5th day
after excavation, while the second model (without the interface) fails on the 12th day. The maximum total plastic strain of the first
model is found to be 854.74, while that of the second model is 1243.69. It can be concluded that the presence of the geological interface
can result in a rapid slope failure, but the total plastic strain is reduced.

21
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

Fig. 17. Comparison of the final pore pressure: (a) with a geological interface; (b) without any geological interface.

The final collapse status of the slopes is presented in Fig. 16, along with the distributions of equivalent plastic strain (PEEQ). In the
first model, the slope slides directly along the geological interface without generating any new sliding surface. No obvious shear failure
is observed on the left side of the interface. This can be attributed to the use of a low friction coefficient and interface cohesion, which
greatly reduces the shear traction acting on the left part of the soil. In the second model, a new sliding surface is generated, and the
failure range is significantly larger than that in the first model. The final distribution of pore pressure is presented in Fig. 17. Although
the two models have significant differences in the failure modes, their pore pressure distributions exhibit remarkable similarities. This
is mainly because the pore pressure is a transient variable, which is only related to the current physical state. It should be emphasized
that, even in the presence of significant interfacial sliding, the pore pressure field of the first model remains smooth and continuous,
which once again proves the high accuracy of our method.

6. Conclusions

In this work, a hydro-mechanical contact method is proposed to model the geotechnical multi-body interaction problems. Using the
SNS-PFEM framework, all field variables are stored at nodes, thus avoiding frequent variable mapping. Then, a polynomial pressure
projection is used to eliminate the potential instability of pore pressure that arises from equal-order interpolation. The contact
constraint of the solid phase is imposed by the dual Lagrange multiplier approach. This method introduces a coordinate transformation
to condense the multipliers. The system matrix can thus maintain an unchanged size. Using the same coordinate transformation, a
novel interface element is constructed for modelling the fluid flow in contact gaps. This element can be established on the non-
conforming meshes and allows for large sliding. When the interface aperture approaches zero, the proposed element degrades into
a continuity constraint of the pore pressure, and the system matrix can remain well-conditioned without any singularity.
Finally, four numerical examples are performed to demonstrate the superior performance of the proposed method. The results of
the first example agree well with analytical solutions, indicating the high accuracy of our method. The results of the remaining three
examples show the effectiveness, flexibility, and robustness of our method in tackling diverse geotechnical problems.

CRediT authorship contribution statement

Huangcheng Fang: Conceptualization, Formal analysis, Investigation, Methodology, Software, Validation, Writing – original
draft. Zhen-Yu Yin: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing –
review & editing. Dingli Zhang: Formal analysis, Investigation, Writing – review & editing. Liqiang Cao: Investigation, Visualization,
Writing – review & editing.

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.

Data availability

Data will be made available on request.

Acknowledgements

This research was financially supported by the Research Grants Council (RGC) of Hong Kong Special Administrative Region
Government (HKSARG) of China (Grant No.: 15217220, N_PolyU534/20, R5037-18F), and the Hong Kong Polytechnic University
Strategic Importance Fund (ZE2T) and Project of Research Institute of Land and Space (CD78).

22
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

References

[1] H. Sabetamal, M. Nazem, J.P. Carter, S.W. Sloan, Large deformation dynamic analysis of saturated porous media with applications to penetration problems,
Comput. Geotech. 55 (2014) 117–131, https://doi.org/10.1016/j.compgeo.2013.08.005.
[2] D. Sheng, P. Wriggers, S.W. Sloan, Improved numerical algorithms for frictional contact in pile penetration analysis, Comput. Geotech. 33 (2006) 341–354,
https://doi.org/10.1016/j.compgeo.2006.06.001.
[3] M.A. Puso, T.A. Laursen, A mortar segment-to-segment frictional contact method for large deformations, Comput. Methods Appl. Mech. Eng. 193 (2004)
4891–4913, https://doi.org/10.1016/j.cma.2004.06.001.
[4] J. Yang, Z.Y. Yin, F. Laouafa, P.Y. Hicher, Internal erosion in dike-on-foundation modeled by a coupled hydromechanical approach, Int. J. Numer. Anal. Methods
Geomech. 43 (2019) 663–683, https://doi.org/10.1002/nag.2877.
[5] Z. Wiȩckowski, The material point method in large strain engineering problems, Comput. Methods Appl. Mech. Eng. 193 (2004) 4417–4438, https://doi.org/
10.1016/j.cma.2004.01.035.
[6] W. Liang, J. Zhao, H. Wu, K. Soga, Multiscale, multiphysics modeling of saturated granular materials in large deformation, Comput. Methods Appl. Mech. Eng.
405 (2023), https://doi.org/10.1016/j.cma.2022.115871.
[7] W. Liang, H. Wu, S. Zhao, W. Zhou, J. Zhao, Scalable three-dimensional hybrid continuum-discrete multiscale modeling of granular media, Int. J. Numer.
Methods Eng. 123 (2022) 2872–2893, https://doi.org/10.1002/nme.6963.
[8] T. Belytschko, S. Xiao, Stability analysis of particle methods with corrected derivatives, Comput. Math. with Appl. 43 (2002) 329–350, https://doi.org/10.1016/
S0898-1221(01)00290-5.
[9] J.J. Monaghan, Smoothed particle hydrodynamics and its diverse applications, Annu. Rev. Fluid Mech. 44 (2011) 323–346, https://doi.org/10.1146/annurev-
fluid-120710-101220.
[10] W. Zeng, G.R. Liu, Smoothed Finite Element Methods (S-FEM): an overview and recent developments, Arch. Comput. Methods Eng. 25 (2018) 397–435, https://
doi.org/10.1007/s11831-016-9202-3.
[11] E. Oñate, S.R. Idelsohn, F. Del Pin, R. Aubry, The particle finite element method. an overview, Int. J. Comput. Methods. 1 (2004) 267–307, https://doi.org/
10.1142/S0219876204000204.
[12] E. Oñate, S.R. Idelsohn, M.A. Celigueta, R. Rossi, Advances in the particle finite element method for the analysis of fluid-multibody interaction and bed erosion
in free surface flows, Comput. Methods Appl. Mech. Eng. 197 (2008) 1777–1800, https://doi.org/10.1016/j.cma.2007.06.005.
[13] X. Zhang, K. Krabbenhoft, D. Sheng, W. Li, Numerical simulation of a flow-like landslide using the particle finite element method, Comput. Mech. 55 (2015)
167–177, https://doi.org/10.1007/s00466-014-1088-z.
[14] L. Monforte, M. Arroyo, J.M. Carbonell, A. Gens, Numerical simulation of undrained insertion problems in geotechnical engineering with the Particle Finite
Element Method (PFEM), Comput. Geotech. 82 (2017) 144–156, https://doi.org/10.1016/j.compgeo.2016.08.013.
[15] M. Cremonesi, A. Franci, S. Idelsohn, E. Oñate, A state of the art review of the Particle Finite Element Method (PFEM), Arch. Comput. Methods Eng. 27 (2020)
1709–1735, https://doi.org/10.1007/s11831-020-09468-4.
[16] F. Sengani, F. Mulenga, A review on the application of particle finite element methods (PFEM) to cases of landslides, Int. J. Geotech. Eng. 16 (2022) 367–381,
https://doi.org/10.1080/19386362.2020.1814027.
[17] W. Zhang, W. Yuan, B. Dai, Smoothed particle finite-element method for large-deformation problems in geomechanics, Int. J. Geomech. 18 (2018), https://doi.
org/10.1061/(asce)gm.1943-5622.0001079.
[18] Y.F. Jin, Z.Y. Yin, X.W. Zhou, F.T. Liu, A stable node-based smoothed PFEM for solving geotechnical large deformation 2D problems, Comput. Methods Appl.
Mech. Eng. 387 (2021), https://doi.org/10.1016/j.cma.2021.114179.
[19] G.R. Liu, T. Nguyen-Thoi, H. Nguyen-Xuan, K.Y. Lam, A node-based smoothed finite element method (NS-FEM) for upper bound solutions to solid mechanics
problems, Comput. Struct. 87 (2009) 14–26, https://doi.org/10.1016/j.compstruc.2008.09.003.
[20] W.H. Yuan, B. Wang, W. Zhang, Q. Jiang, X.T. Feng, Development of an explicit smoothed particle finite element method for geotechnical applications, Comput.
Geotech. 106 (2019) 42–51, https://doi.org/10.1016/j.compgeo.2018.10.010.
[21] W. Zhang, Z. hao Zhong, C. Peng, W. hai Yuan, W. Wu, GPU-accelerated smoothed particle finite element method for large deformation analysis in
geomechanics, Comput. Geotech. 129 (2021), https://doi.org/10.1016/j.compgeo.2020.103856.
[22] W. Zeng, G.R. Liu, D. Li, X.W. Dong, A smoothing technique based beta finite element method (βFEM) for crystal plasticity modeling, Comput. Struct. 162
(2016) 48–67, https://doi.org/10.1016/j.compstruc.2015.09.007.
[23] H. Feng, X.Y. Cui, G.Y. Li, A stable nodal integration method with strain gradient for static and dynamic analysis of solid mechanics, Eng. Anal. Bound. Elem. 62
(2016) 78–92, https://doi.org/10.1016/j.enganabound.2015.10.001.
[24] H. Yang, X.Y. Cui, S. Li, Y.H. Bie, A stable node-based smoothed finite element method for metal forming analysis, Comput. Mech. 63 (2019) 1147–1164,
https://doi.org/10.1007/s00466-018-1641-2.
[25] Y. Li, G.R. Liu, A novel node-based smoothed finite element method with linear strain fields for static, free and forced vibration analyses of solids, Appl. Math.
Comput. 352 (2019) 30–58, https://doi.org/10.1016/j.amc.2019.01.043.
[26] H. Fang, Z.-Y. Yin, M. Peng, D. Zhang, Improved SNS-PFEM framework with dual mortar method to model geotechnical large deformation contact problems,
Comput. Methods Appl. Mech. Eng. 412 (2023), https://doi.org/10.1016/j.cma.2023.116091.
[27] J.B. Haga, H. Osnes, H.P. Langtangen, On the causes of pressure oscillations in low-permeable and low-compressible porous media, Int. J. Numer. Anal. Methods
Geomech. 36 (2012) 1507–1522, https://doi.org/10.1002/nag.1062.
[28] R.E. Goodman, R.L. Taylor, T.L. Brekke, A model for the mechanics of jointed rock, J. Soil Mech. Found. Div. 94 (1968) 637–659, https://doi.org/10.1061/
jsfeaq.0001133.
[29] C.S. Desai, M.M. Zaman, J.G. Lightner, H.J. Siriwardane, Thin-layer element for interfaces and joints, Int. J. Numer. Anal. Methods Geomech. 8 (1984) 19–43,
https://doi.org/10.1002/nag.1610080103.
[30] M. Zhou, B. Zhang, C. Peng, W. Wu, Three-dimensional numerical analysis of concrete-faced rockfill dam using dual-mortar finite element method with mixed
tangential contact constraints, Int. J. Numer. Anal. Methods Geomech. 40 (2016) 2100–2122, https://doi.org/10.1002/nag.2524.
[31] J.M. Carbonell, E. Oñate, B. Suárez, Modelling of tunnelling processes and rock cutting tool wear with the particle finite element method, Comput. Mech. 52
(2013) 607–629, https://doi.org/10.1007/s00466-013-0835-x.
[32] T.J.R. Hughes, R.L. Taylor, J.L. Sackman, A. Curnier, W. Kanoknukulchai, Finite element method for a class of contact-impact problems, Comput. Methods Appl.
Mech. Eng. 8 (1976) 249–276, https://doi.org/10.1016/0045-7825(76)90018-9.
[33] J.O. Hallquist, G.L. Goudreau, D.J. Benson, Sliding interfaces with contact-impact in large-scale Lagrangian computations, Comput. Methods Appl. Mech. Eng.
51 (1985) 107–137, https://doi.org/10.1016/0045-7825(85)90030-1.
[34] P. Wriggers, Computational Contact Mechanics, Springer, Berlin, 2006.
[35] T.W. McDevitt, T.A. Laursen, A mortar-finite element formulation for frictional contact problems, Int. J. Numer. Methods Eng. 48 (2000) 1525–1547, https://
doi.org/10.1002/1097-0207(20000810)48:10<1525::AID-NME953>3.0.CO;2-Y.
[36] M. Tur, F.J. Fuenmayor, P. Wriggers, A mortar-based frictional contact formulation for large deformations using Lagrange multipliers, Comput. Methods Appl.
Mech. Eng. 198 (2009) 2860–2873, https://doi.org/10.1016/j.cma.2009.04.007.
[37] T. Doca, F.M. Andrade Pires, J.M.A. Cesar De Sa, A frictional mortar contact approach for the analysis of large inelastic deformation problems, Int. J. Solids
Struct. 51 (2014) 1697–1715, https://doi.org/10.1016/j.ijsolstr.2014.01.013.
[38] M. Hirmand, M. Vahab, A.R. Khoei, An augmented Lagrangian contact formulation for frictional discontinuities with the extended finite element method, Finite
Elem. Anal. Des. 107 (2015) 28–43, https://doi.org/10.1016/j.finel.2015.08.003.
[39] A. Pantano, R.C. Averill, A mesh-independent interface technology for simulation of mixed-mode delamination growth, Int. J. Solids Struct. 41 (2004)
3809–3831, https://doi.org/10.1016/j.ijsolstr.2004.02.018.

23
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

[40] A. Pantano, R.C. Averill, A penalty-based interface technology for coupling independently modeled 3D finite element meshes, Finite Elem. Anal. Des. 43 (2007)
271–286, https://doi.org/10.1016/j.finel.2006.10.001.
[41] A.R. Khoei, M. Vahab, M. Hirmand, Modeling the interaction between fluid-driven fracture and natural fault using an enriched-FEM technique, Int. J. Fract. 197
(2016) 1–24, https://doi.org/10.1007/s10704-015-0051-0.
[42] W. Han, M. Sofonea, Convergence analysis of penalty based numerical methods for constrained inequality problems, Numer. Math. 142 (2019) 917–940,
https://doi.org/10.1007/s00211-019-01036-8.
[43] F. Ben Belgacem, The Mortar finite element method with Lagrange multipliers, Numer. Math. 84 (1999) 173–197, https://doi.org/10.1007/s002110050468.
[44] K.C. Park, C.A. Felippa, R. Ohayon, Partitioned formulation of internal fluid–structure interaction problems by localized Lagrange multipliers, Comput. Methods
Appl. Mech. Eng. 190 (2001) 2989–3007, https://doi.org/10.1016/S0045-7825(00)00378-9.
[45] P. Farah, W.A. Wall, A. Popp, A mortar finite element approach for point, line, and surface contact, Int. J. Numer. Methods Eng. 114 (2018) 255–291, https://
doi.org/10.1002/nme.5743.
[46] H. Fang, D. Zhang, M. Zhou, Q. Fang, M. Wen, A contact algorithm for cohesive cracks in the extended finite element method, Int. J. Numer. Methods Eng. 121
(2020) 2747–2766, https://doi.org/10.1002/nme.6329.
[47] M.C. Cojocaru, A. Matei, Weak solutions via two-field lagrange multipliers for boundary value problems in mathematical physics, Math. Model. Anal. 27 (2022)
561–572, https://doi.org/10.3846/mma.2022.15827.
[48] A. Popp, M. Gitterle, M.W. Gee, W.A. Wall, A dual mortar approach for 3D finite deformation contact with consistent linearization, Int. J. Numer. Methods Eng.
83 (2010) 1428–1465, https://doi.org/10.1002/nme.2866.
[49] A. Popp, B.I. Wohlmuth, M.W. Gee, W.A. Wall, Dual quadratic mortar finite element methods for 3D finite deformation contact, SIAM J. Sci. Comput. 34 (2012)
B421–B446, https://doi.org/10.1137/110848190.
[50] A. Popp, A. Seitz, M.W. Gee, W.A. Wall, Improved robustness and consistency of 3D contact algorithms based on a dual mortar approach, Comput. Meth. Appl.
Mech. Eng. 264 (2013) 67–80, https://doi.org/10.1016/j.cma.2013.05.008.
[51] H. Fang, D. Zhang, Q. Fang, L. Cao, M. Wen, An efficient patch-to-patch method for coupling independent finite element subdomains with intersecting
interfaces, Comput. Methods Appl. Mech. Eng. 2021 (2021) 114209, https://doi.org/10.1016/j.cma.2021.114209.
[52] H. Sabetamal, M. Nazem, S.W. Sloan, J.P. Carter, Frictionless contact formulation for dynamic analysis of nonlinear saturated porous media based on the mortar
method, Int. J. Numer. Anal. Methods Geomech. 40 (2016) 25–61, https://doi.org/10.1002/nag.2386.
[53] W. Wang, M. Zhou, B. Zhang, C. Peng, A dual mortar contact method for porous media and its application to clay-core rockfill dams, Int. J. Numer. Anal.
Methods Geomech. 43 (2019) 1744–1769, https://doi.org/10.1002/nag.2930.
[54] P.S. Donzelli, R.L. Spilker, A contact finite element formulation for biological soft hydrated tissues, Comput. Methods Appl. Mech. Eng. 153 (1998) 63–79,
https://doi.org/10.1016/S0045-7825(97)00065-0.
[55] J.M. Segura, I. Carol, On zero-thickness interface elements for diffusion problems, Int. J. Numer. Anal. Methods Geomech. 28 (2004) 947–962, https://doi.org/
10.1002/nag.358.
[56] J.M. Segura, I. Carol, Coupled HM analysis using zero-thickness interface elements with double nodes. Part I: theoretical model, Int. J. Numer. Anal. Methods
Geomech. 32 (2008) 2083–2101, https://doi.org/10.1002/nag.735.
[57] M. Peng, Y. Tian, C. Gaudin, L. Zhang, D. Sheng, Application of a coupled hydro-mechanical interface model in simulating uplifting problems, Int. J. Numer.
Anal. Methods Geomech. 46 (2022) 3256–3280, https://doi.org/10.1002/nag.3450.
[58] A.R. Khoei, Extended Finite Element Method: Theory and Applications, John Wiley & Sons, West Sussex, 2015, https://doi.org/10.1002/9781118869673.
[59] A.R. Khoei, M. Hirmand, M. Vahab, M. Bazargan, An enriched FEM technique for modeling hydraulically driven cohesive fracture propagation in impermeable
media with frictional natural faults: numerical and experimental investigations, Int. J. Numer. Methods Eng. 104 (2015) 439–468, https://doi.org/10.1002/
nme.4944.
[60] A.R. Khoei, M. Vahab, M. Hirmand, An enriched–FEM technique for numerical simulation of interacting discontinuities in naturally fractured porous media,
Comput. Methods Appl. Mech. Eng. 331 (2018) 197–231, https://doi.org/10.1016/j.cma.2017.11.016.
[61] Y.F. Jin, Z.Y. Yin, Two-phase PFEM with stable nodal integration for large deformation hydromechanical coupled geotechnical problems, Comput. Methods
Appl. Mech. Eng. 392 (2022), https://doi.org/10.1016/j.cma.2022.114660.
[62] Z.Y. Wang, Y.F. Jin, Z.Y. Yin, Y.Z. Wang, A dynamic SNS-PFEM with generalized-α method for hydro-mechanical coupled geotechnical problems, Comput.
Geotech. 159 (2023), https://doi.org/10.1016/j.compgeo.2023.105466.
[63] M. Nazem, D. Sheng, J.P. Carter, S.W. Sloan, Arbitrary Lagrangian–Eulerian method for large-strain consolidation problems, Int. J. Numer. Anal. Methods
Geomech. 32 (2008) 1023–1050, https://doi.org/10.1002/nag.657.
[64] M. Nazem, D. Sheng, J.P. Carter, Stress integration and mesh refinement for large deformation in geomechanics, Int. J. Numer. Methods Eng. 65 (2006)
1002–1027, https://doi.org/10.1002/nme.1470.
[65] J. Yuan, M.A. Hicks, Large deformation elastic electro-osmosis consolidation of clays, Comput. Geotech. 54 (2013) 60–68, https://doi.org/10.1016/j.
compgeo.2013.05.012.
[66] W. Karush, Minima of functions of several variables with inequalities as side conditions, Traces Emerg. Nonlinear Program. (2014) 217–245, https://doi.org/
10.1007/978-3-0348-0439-4_10.
[67] D.T. Snow, A Parallel Plate Model of Fractured Permeable Media, University of California, Berkeley, 1965.
[68] P.A. Witherspoon, J.S.Y. Wang, K. Iwai, J.E. Gale, Validity of Cubic Law for fluid flow in a deformable rock fracture, Water Resour. Res. 16 (1980) 1016–1024,
https://doi.org/10.1029/WR016i006p01016.
[69] T. YW, T. CF, Channel model of flow through fractured media, Water Resour. Res. 23 (1987) 467–479, https://doi.org/10.1016/0148-9062(87)90977-6.
[70] S.R. Brown, Fluid flow through rock joints: the effect of surface roughness, J. Geophys. Res. 92 (1987) 1337–1347, https://doi.org/10.1029/JB092iB02p01337.
[71] M.E. Thompson, S.R. Brown, The effect of anisotropic surface roughness on flow and transport in fractures, J. Geophys. Res. 96 (1991), https://doi.org/
10.1029/91jb02252.
[72] B. Cerfontaine, A.C. Dieudonné, J.P. Radu, F. Collin, R. Charlier, 3D zero-thickness coupled interface finite element: formulation and application, Comput.
Geotech. 69 (2015) 124–140, https://doi.org/10.1016/j.compgeo.2015.04.016.
[73] A. Lavrov, Coupling in hydraulic fracturing simulation. Porous Rock Fracture Mechanics: With Application to Hydraulic Fracturing, Drilling and Structural
Engineering, 2017, pp. 47–62, https://doi.org/10.1016/B978-0-08-100781-5.00003-8.
[74] A.R. Khoei, M. Vahab, A numerical contact algorithm in saturated porous media with the extended finite element method, Comput. Mech. 54 (2014)
1089–1110, https://doi.org/10.1007/s00466-014-1041-1.
[75] G.R. Liu, T. Nguyen-Thoi, K.Y. Lam, A novel alpha finite element method (αFEM) for exact solution to mechanics problems using triangular and tetrahedral
elements, Comput. Methods Appl. Mech. Eng. 197 (2008) 3883–3897, https://doi.org/10.1016/j.cma.2008.03.011.
[76] Z.Q. Zhang, G.R. Liu, Temporal stabilization of the node-based smoothed finite element method and solution bound of linear elastostatics and vibration
problems, Comput. Mech. 46 (2010) 229–246, https://doi.org/10.1007/s00466-009-0420-5.
[77] M. Hillman, J.S. Chen, An accelerated, convergent, and stable nodal integration in Galerkin meshfree methods for linear and nonlinear mechanics, Int. J. Numer.
Methods Eng. 107 (2016) 603–630, https://doi.org/10.1002/nme.5183.
[78] C.T. Wu, Y. Wu, Z. Liu, D. Wang, A stable and convergent Lagrangian particle method with multiple nodal stress points for large strain and material failure
analyses in manufacturing processes, Finite Elem. Anal. Des. 146 (2018) 96–106, https://doi.org/10.1016/j.finel.2018.04.007.
[79] J.A. White, R.I. Borja, Stabilized low-order finite elements for coupled solid-deformation/fluid-diffusion and their application to fault zone transients, Comput.
Methods Appl. Mech. Eng. 197 (2008) 4353–4366, https://doi.org/10.1016/j.cma.2008.05.015.
[80] P. Mira, M. Pastor, T. Li, X. Liu, A new stabilized enhanced strain element with equal order of interpolation for soil consolidation problems, Comput. Methods
Appl. Mech. Eng. 192 (2003) 4257–4277, https://doi.org/10.1016/S0045-7825(03)00416-X.

24
H. Fang et al. Computer Methods in Applied Mechanics and Engineering 420 (2024) 116743

[81] M. Pastor, T. Li, X. Liu, O.C. Zienkiewicz, M. Quecedo, Fractional step algorithm allowing equal order of interpolation for coupled analysis of saturated soil
problems, Mech. Cohesive-Friction. Mater. 5 (2000) 511–534, https://doi.org/10.1002/1099-1484(200010)5:7<511::AID-CFM87>3.0.CO;2-S.
[82] W.C. Sun, J.T. Ostien, A.G. Salinger, A stabilized assumed deformation gradient finite element formulation for strongly coupled poromechanical simulations at
finite strain, Int. J. Numer. Anal. Methods Geomech. 37 (2013) 2755–2788, https://doi.org/10.1002/nag.2161.
[83] M. Cremonesi, L. Ferrara, A. Frangi, U. Perego, Simulation of the flow of fresh cement suspensions by a Lagrangian finite element approach, J. Nonnewton. Fluid
Mech. 165 (2010) 1555–1563, https://doi.org/10.1016/j.jnnfm.2010.08.003.
[84] J.M. Rodriguez, J.M. Carbonell, J.C. Cante, J. Oliver, The particle finite element method (PFEM) in thermo-mechanical problems, Int. J. Numer. Methods Eng.
107 (2016) 733–785, https://doi.org/10.1002/nme.5186.
[85] B.I. Wohlmuth, A mortar finite element method using dual spaces for the Lagrange multiplier, SIAM J. Numer. Anal. 38 (2001) 989–1012, https://doi.org/
10.1137/S0036142999350929.
[86] H. Fang, D. Zhang, M. Zhou, Q. Fang, M. Wen, X. Hu, A virtual interface-coupled extended finite element method for three-dimensional contact problems, Int. J.
Numer. Methods Eng. 122 (2021) 386–402, https://doi.org/10.1002/nme.6541.
[87] Z.Z. Fu, S.H. Liu, Formulations of a hydromechanical interface element, Acta Mech. Sin. Xuebao. 27 (2011) 697–705, https://doi.org/10.1007/s10409-011-
0482-4.
[88] J.M. Segura, I. Carol, Coupled HM analysis using zero-thickness interface elements with double nodes-Part II: verification and application, Int. J. Numer. Anal.
Methods Geomech. 32 (2008) 2103–2123, https://doi.org/10.1002/nag.730.
[89] Z.Y. Wang, Y.F. Jin, Z.Y. Yin, Y.Z. Wang, A novel coupled NS-PFEM with stable nodal integration and polynomial pressure projection for geotechnical problems,
Int. J. Numer. Anal. Methods Geomech. 46 (2022) 2535–2560, https://doi.org/10.1002/nag.3417.
[90] A. Verruijt, Soil Mechanics, Delft University of Technology, Delft, 2006.

25