Deformation Response and Load Transfer Mechanism of Collar Monopile Foundations in Saturated Cohesive Soils

Abstract

Collar monopile foundation is a new type of offshore wind power foundation. This paper explores the horizontal bearing performance of collar monopile foundation in saturated cohesive soil through a combination of physical experiments and numerical simulations. After analyzing the deformation characteristics of the pile–soil system under horizontal load through static load tests, horizontal cyclic loading tests were conducted at different cycles to study the cumulative deformation law of the collar monopile. Based on a stiffness degradation model for soft clay, a USDFLD subroutine was developed in Fortran and embedded in ABAQUS. Coupled with the Mohr–Coulomb criterion, it was used to simulate the deformation behavior of the collar monopile under horizontal cyclic loading. The numerical model employed the same geometric dimensions and boundary conditions as the physical test, and the simulated cumulative pile–head displacement under 4000 load cycles showed good agreement with the experimental results, thereby verifying the rationality and reliability of the proposed simulation method. Through numerical simulation, the distribution characteristics of bending moment and the shear force of collar monopile foundation were studied, and the influence of pile shaft and collar on the horizontal bearing capacity of collar monopile foundation at different loading stages was analyzed. The results show that as the horizontal load increases, cracks gradually appear at the bottom of the collar and in the surrounding soil. The soil disturbance caused by the sliding and rotation of the collar will gradually increase, leading to plastic failure of the surrounding soil and reducing the bearing capacity. The excess pore water pressure in shallow soil increases rapidly in the early cycle and then gradually decreases with the formation of drainage channels. Deep soil may experience negative pore pressure, indicating the presence of a suction effect. This paper can provide theoretical support for the design optimization and performance evaluation of collar monopile foundations in offshore wind power engineering applications.

1. Introduction

In response to global and national calls for sustainable development in civil and structural engineering, a wide range of studies have been devoted to enhancing material performance, soil–structure interaction, energy efficiency, and structural resilience under complex environmental conditions [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]. Within this context, improving foundation design—particularly for offshore wind structures—is of paramount importance. The collar monopile foundation investigated in this study aligns with these sustainability goals by enhancing lateral resistance and reducing material consumption through structural synergy. The collar single-pile foundation is a novel form of offshore wind power foundation, functionally similar to a pile cap with an embedded retaining wall [17,18]. It typically consists of a single pile combined with a gravity-based foundation referred to as a collar. At present, there is limited research on collar single-pile foundations both domestically and internationally. Compared with conventional monopile foundations, the distinguishing feature of the collar single-pile lies in the addition of a circular frictional collar fixed at the pile head near the mudline, which provides additional bearing capacity to the typical monopile system [19]. Research has shown that the improvement in load-bearing capacity due to the collar is reflected in several respects [20,21,22]: first, the contact between the underside of the collar and the surrounding soil not only increases the contact area, thereby resisting vertical deformation of the foundation, but also generates Coulomb friction to counteract lateral deformation. Moreover, the restoring moment provided by the gravity-based collar resists foundation rotation. In addition, the vertical stress induced by the self-weight of the collar strengthens the soil around the pile, thus enhancing the lateral bearing capacity of the foundation.

During the normal service period of offshore wind turbines, horizontal loads are the main control loads [23,24]. The existing calculation methods for the lateral resistance of single piles mainly include the extreme foundation reaction force method, the elastic foundation reaction force method (m method), the composite foundation reaction force method (p-y curve method), and the numerical analysis method based on elastoplastic theory [25,26,27,28]. Among them, the p-y curve method has been widely used in offshore projects such as offshore wind power, and has been adopted by domestic and foreign standards such as the “Code for Design of Wind Turbine Foundations for Offshore Wind Farm Engineering”, the American Petroleum Institute (API), and the Norwegian Classification Society (DNV).

In 1956, Mcclelland [29] observed the stress–strain curve of soil in a clay consolidation undrained test and first proposed the p-y curve method. On this basis, Reese et al. [30] carried out a horizontal loading test of steel pipe piles in sandy soil sites. Based on the test results of the flexible pile model (with a slenderness ratio of 24), a p-y curve model was proposed. This method is simple to apply and can better reflect the nonlinearity of the soil.

With the continuous development of pile foundation forms, researchers have begun to focus on the possibility of introducing structural components into traditional single-pile foundations to improve the bearing performance. Stone et al. [31,32] demonstrated through 1 g small-scale laboratory model tests and centrifuge tests that the lateral and vertical bearing capacities of winged monopile foundations in sand are significantly greater than those of conventional monopile foundations. Stone and Arshi [33,34,35,36] conducted 1 g model tests, centrifuge tests, and numerical simulations in sandy soils, and their research showed that adding a circular wing to a monopile increases its lateral bearing capacity by 50%.

Lehane et al. [37,38], based on centrifuge tests and finite element analysis, investigated the effects of wings in clayey soils. The results indicated that due to the soft nature of the selected clay, the ultimate horizontal bearing capacity of the winged pile was governed by the strength of the surrounding soil, with the wing contributing only about 10% of the total horizontal bearing capacity of the winged monopile. In this case, the presence of the wing did not significantly enhance the pile’s horizontal bearing capacity.

Considering that sandy soils generally offer better bearing performance than clayey soils, a series of supplementary centrifuge tests were conducted in sand. The tests revealed that in medium-dense to dense sand, the rotational stiffness and horizontal ultimate bearing capacity of winged monopile foundations were significantly improved due to the presence of the wing. Yang et al. [39], through centrifuge testing and 3D finite element analysis, studied the lateral bending moment response of winged monopile foundations in saturated sand. Their results confirmed that the addition of wings significantly enhanced the lateral bearing capacity and stiffness of the pile, with improvements increasing alongside wing diameter or embedment depth. They also proposed a gravity-based foundation design method for offshore wind turbines based on DNV recommendations, which can predict the bearing capacity of gravity-type winged foundations. Wang et al. [40,41,42] conducted centrifuge tests in both sand and clay and compared the results with finite element simulations. They proposed the superposition method and the equivalent bending moment method. The superposition method equates the ultimate horizontal bearing capacity of a winged monopile foundation to the sum of the monopile capacity calculated using the p-y curve method and the wing foundation capacity estimated using the DNV formula.

In recent years, with the increase in pile diameter and the decrease in the aspect ratio of monopile foundations for offshore wind turbines, the traditional p-y curve method has gradually revealed its limitations in applicability. For instance, most existing p-y curves are established based on tests conducted in stiff clay and sandy soils, making them less suitable for saturated soft clay foundations [43]. Moreover, finite element analysis by Wiemann [44] indicated that for rigid large-diameter piles with low aspect ratios, the p-y curve method may overestimate the stiffness of the pile-soil system at deeper embedment depths.

Although the traditional p–y curve method is widely used for analyzing the lateral bearing behavior of monopiles, recent studies have shown that it has clear limitations when applied to large-diameter rigid piles and saturated soft clay. While some researchers have proposed improvements through experiments and numerical methods, existing studies mainly focus on conventional monopiles. Research on collar monopiles in saturated soft clay remains limited, especially regarding the deformation characteristics and pore pressure evolution mechanisms under cyclic loading. Therefore, investigating the deformation response and load-transfer mechanisms of collar monopiles in saturated soft clay under monotonic and cyclic lateral loading is of significant theoretical and practical importance.

In response to these issues, many researchers in recent years have revised and extended the traditional p-y curve method in an effort to enhance its applicability to various soil conditions and pile types. Choo et al. [45], based on model tests, observed that the traditional p-y curve method specified in the API standard underestimates the influence of vertical loads on the lateral response of monopiles. Zhang [46], through finite element simulations using the Mohr–Coulomb model—simplified to the Tresca model by assigning a friction angle of 0°, and simulating undrained conditions by setting the Poisson’s ratio υ to 0.48—investigated the wedge failure mechanism of piles in clay under undrained conditions. He plotted the corresponding lateral load–displacement (p–y) and pile tip shear–displacement (s–u) curves, revealing that the p–y curve method severely underestimates system stiffness.

Shao et al. [47] introduced three-dimensional frictional effects along the pile into the ultimate soil resistance formula and the differential equation for pile deflection, thereby developing an analytical method for the lateral response of piles at the crest of cohesionless soil slopes. Their results were compared with centrifuge and small-scale model tests. Based on a modified sand wedge failure model, a new p–y curve was subsequently proposed. Achmus et al. [48], using finite element software, developed a stiffness degradation model and conducted parametric studies on the influence of pile geometry, loading conditions, and relative sand density on the performance of piles under cyclic loading.

Shao et al. [49] used numerical methods to investigate the degradation of lateral bearing capacity of monopiles in soft clay under lateral cyclic loading, based on a modified kinematic hardening constitutive model. The simulation results were compared with centrifuge model test data, validating the accuracy of the modified model. Vicen et al. [50], through 1 g laboratory model tests, studied the accumulated rotation and unloading stiffness of bucket foundations in saturated loose sand and proposed an empirical equation for cumulative rotation. Cheng et al. [51,52], based on interface theory, developed a constitutive model for saturated clay that accounts for reverse plastic flow, hysteretic cyclic evolution, plastic strain accumulation, and soil stiffness degradation. The model was implemented into ABAQUS through a user-defined material subroutine (UMAT), successfully predicting the nonlinear hysteretic response of monopiles in saturated clay under cyclic loading, as well as the evolution of bending moment and lateral deflection profiles over loading cycles. Cheng et al. [53], based on finite element numerical analysis, investigated the influence of different cyclic loading patterns on the lateral response of large-diameter monopiles using a simplified clay constitutive model.

2. Methodology

2.1. Experimental Apparatus

2.1.1. Model Pile

The model pile employed in this study was scaled from the large-diameter monopile foundation of an offshore wind-turbine project along the Chinese coast [54]. The prototype pile has a diameter of 5 m and an embedment depth of 70 m. According to the second similarity theorem and equivalent parameter conversion, a geometric scale factor of 1:100 (prototype: model) was adopted. To minimize boundary effects in the laboratory, the outer diameter of the circumferential collar (D₂) was set to three times the monopile diameter (D₁).

Both the pile shaft and the collar were fabricated from an aluminum alloy with a Young’s modulus of 71 GPa. The collar was welded to the pile shaft, with the weld bead centered 300 mm below the pile head, as shown in Figure 1. The principal dimensions of the model monopile are a 50 mm outer diameter, a 3 mm wall thickness, and a 1000 mm total length, whereas the collar measures 150 mm in outer diameter, 50 mm in inner diameter, and 3 mm in thickness.

2.1.2. Model Soil

Kaolin powder derived from primary ore was employed to prepare the cohesive model soil. The particle-size distribution was determined in accordance with the Chinese Standard Methods of Soil Testing for Civil Engineering (GB/T 50123-2019) [55]; the pertinent physical properties are listed in

Table 1. Soil properties.

Parameter	Value
Median particle diameter ($d_{50}$/mm)	0.0046
Relative density ($G_s$)	2.57
Poisson’s ratio (${\mu }_s$)	0.47
Liquid limit ($W_I$)	43.07%
Plastic limit ($W_P$)	24.70%
plastic index ($I_P$)	18.37%

. The preparation and placement procedures were as follows.

The preparation and placement procedures were conducted in accordance with standard practices. Kaolin slurry was saturated while maintaining a standing water layer above the soil surface (Figure 2a). After placing and compacting a bedding layer of fine sand, saturated kaolin was backfilled in controlled 10 cm lifts. The model pile was installed vertically at the center of the tank. To accelerate consolidation, a preload was applied under a constant water head (Figure 2b). Following preload removal and the drainage of surface water, the model was left to stabilize prior to testing.

2.1.3. Model Tank and Instrumentation Layout

The model container was fabricated from steel plates, tempered-glass panels and reinforced concrete, with internal dimensions of 1.20 m × 1.35 m × 0.90 m. The minimum clearance between the model pile and each container wall exceeds 10 D₂ (where D₂ denotes the pile diameter), so boundary effects can be neglected. Throughout the test, the container remained watertight, meeting all accuracy requirements.

To monitor the evolution of pore-water pressure in the soil and the load–deformation behavior of the pile during loading, all sensors were linked to a Donghua 5922 multifunctional dynamic data-acquisition system. The instrumentation comprised one displacement transducer, four pore-pressure transducers and one tension–compression load cell. As illustrated in Figure 3a, the displacement transducer was mounted at the pile head; the load cell was positioned 2 D₂ above the ground surface; pore-pressure sensor P1 was embedded 1 D₂ below the mudline, while the remaining three pore-pressure sensors were installed successively downward at vertical intervals of 4 D₂.

2.1.4. Loading Apparatus

Two sets of apparatus were employed to apply horizontal loads: one for monotonic (static) loading and the other for cyclic loading.

The stable application of load and the precise control of the force transmitted to the model pile are essential for the static test. A fixed pulley system was utilized (Figure 3b), and the slow maintained-load method specified in the Chinese specification Technical Code for Testing of Building Pile Foundations (JGJ 106-2014) was followed, applying increments of 20 N (±0.5 N). A steel strand connected the loading tray at one end and the model pile at the predetermined loading point at the other; the pulley converted the tray’s vertical weight into a horizontal force acting on the pile. The horizontal load was adjusted by successively adding or removing sandbags from the tray until it was fully loaded, at which point the horizontal force at the pile head was approximately 200 N. The in situ arrangement of the monotonic loading system is presented in Figure 3b.

A horizontal cyclic loading system must combine operational stability with precise control of load amplitude and frequency. Conventional solutions based on hydraulic servo actuators or vibrators, although widely used, entail high capital and installation costs and typically deliver forces in the kilonewton range—well above the tens to hundreds of newtons required for the present small-scale study. Balancing cost and ease of operation, the experiment therefore employs an electric-motor-driven eccentric wheel coupled to the pile via a flexible coupling (Figure 4a), with a tension–compression load cell mounted on the coupling to monitor the cyclic force in real time. Laboratory calibration shows that an eccentricity of 5 mm at a frequency of 1 Hz yields a slow decay in load amplitude (the motor and eccentric wheel are shown in Figure 4b,c). Accordingly, an eccentricity of 5 mm and a loading frequency of 1 Hz are adopted here. During installation, the eccentric wheel is finely adjusted so that the collar monopile experiences different load amplitudes on its two flanks, allowing the influence of amplitude variation on cumulative pile–head displacement to be assessed. The proposed method offers simple operation, stable and controllable loading frequency, and low cost, although the cyclic load amplitude diminishes gradually with the number of cycles.

2.2. Numerical Simulation

To verify the accuracy of the numerical model, a finite element model consistent with the specifications of the laboratory test was first established and validated through comparison with the experimental results. On this basis, in order to better reflect actual engineering conditions and conduct further analysis, a finite element model of a gravity-based collar monopile foundation and the surrounding soil was developed based on parameters from a real-world project. The total length of the pile is 30 m, with an outer diameter of 3 m, a wall thickness of 0.06 m, and an embedment depth of 20 m. The collar has a height of 1.5 m, an outer diameter of 12 m, and an inner diameter of 3 m.

The collar monopile foundation consists of three components: a hollow steel pipe pile, a gravity-based collar made of reinforced concrete, and a connection section made of high-strength grouting material, which serves to link the collar with the steel pipe pile. For model simplification, the grouting material is treated as reinforced concrete. Given that the stiffness and strength of these three materials are significantly higher than those of the surrounding soil, all components of the collar monopile foundation are modeled as linear elastic materials in the finite element analysis. Specifically, the reinforced concrete is assigned a density of 2500 kg/m³, an elastic modulus of 30 GPa, and a Poisson’s ratio of 0.3; the steel is assigned a density of 7850 kg/m³, an elastic modulus of 206 GPa, and a Poisson’s ratio of 0.3. The hollow steel pipe pile and the gravity-based collar are connected using shared nodes. The foundation soil is assumed to be homogeneous clay, modeled using the Mohr–Coulomb elastoplastic model. The soil is assigned a density of 960 kg/m³, an elastic modulus of 10 MPa, a Poisson’s ratio of 0.37, an internal friction angle of 15.7°, and a cohesion of 15 kPa.

2.2.1. Soil Stiffness Attenuation Model Considering Cyclic Cumulative Pore Pressure

Under horizontal cyclic loading, soft cohesive soils accumulate plastic strain, leading to stiffness degradation. To capture this behavior, this study adopts the stiffness–degradation model proposed by Hu [56], as given in Equation (1). A user subroutine (USDFLD) was developed in Fortran to implement the model by describing the material’s stress–strain response, and the corresponding field variables were defined and embedded into ABAQUS for numerical simulation.

$$\delta ={\displaystyle \frac{E_N}{E_1}}={\displaystyle \frac{\left(q_d-u_N\right)/{\epsilon }_N}{\left(q_d-u_1\right)/{\epsilon }_1}}={\displaystyle \frac{{\epsilon }_N}{{\epsilon }_1}}{\displaystyle \frac{\left(q_d-u_N\right)}{\left(q_d-u_1\right)}}=N^{-S}({\displaystyle \frac{a+b·N^{c·D_d}}{a+b}})$$

(1)

where $\delta$ is the softening index; N is the number of loading cycles; $E_N$ is the secant modulus at the N-th loading cycle; $u_N$ is the pore water pressure of the specimen at the N-th cycle; ${\epsilon }_N$ is the axial strain at the N-th cycle; $D_d$ is the dynamic deviator-stress level, defined as $D_d=q_d/q_{ult}$; $q_d$ denotes the dynamic deviator stress; $q_{ult}$ denotes the undrained shear strength; and s, a, b, and c are empirical constants determined from tests. The model incorporates the influence of excess pore-water pressure on the softening of saturated soft clay and couples the effects of the number of cycles N and the dynamic deviator stress level $D_d$ on pore-pressure evolution. It further establishes an explicit relationship between the secant shear modulus after N cycles and that of the first cycle. With this relationship, the equivalent stiffness for any cycle count can be obtained without iterative computation, which greatly reduces the computational effort and thereby enhances its suitability for engineering applications.

2.2.2. Boundary Conditions

Both the collar monopile foundation and the surrounding soil are modeled as axisymmetric bodies, with the soil domain idealized as a semi-cylindrical region. A symmetry boundary condition is applied along the half-axisymmetric plane perpendicular to the Y direction, constraining translation in Y to reduce computational cost. To minimize boundary effects, the soil domain extends 10 D₂ (where D₂ is the pile diameter) laterally on either side and 1.5 H (H being the pile embedment depth) vertically. Boundary conditions are assigned as follows: the ground surface is free; the lateral boundary on the non-symmetry side is constrained in X and Y; the base is fully fixed in X, Y, and Z. The model with boundary conditions is illustrated in Figure 5a.

A surface-to-surface contact formulation is adopted to model the relative displacement and shear transfer at the pile–soil interface. The normal contact behavior is defined as hard contact, which prevents penetration of the slave surface into the master surface while permitting separation under tension. Tangential behavior is represented by a Coulomb friction penalty model, with the interface friction coefficient set to u = tan(0.75φ), where φ is the soil’s internal friction angle. This treatment ensures that the simulated frictional response matches actual conditions and provides a realistic description of shear transfer, friction, and sliding along the pile–soil interface.

For the selection of element types in mesh generation, both the foundation model and the collar monopile model adopt eight-node linear reduced integration hexahedral solid elements (C3D8Rs), thereby avoiding the occurrence of “hourglass modes” during the computation. Before meshing, the Partition Cell function is used to appropriately partition the geometric models of the foundation and the pile to meet the requirements for structured meshing. A biased mesh technique is applied to refine the mesh near the pile–soil interface, ensuring sufficient mesh density at the contact surface to enable the accurate calculation of the stress distribution in the corresponding region.

Simulation of the horizontal cyclic forces induced by waves, wind and operational loads in ABAQUS proceeds by first determining the horizontal force and moment acting on the turbine foundation top according to design codes and project requirements. A reference point is then created above the geometric center of the model top—taken here as 10 m above the mudline, consistent with the typical range of 2.5 D₂ to 4.0 D₂—and coupled to the top surface of the steel tubular pile via a coupling constraint. The computed equivalent loads are finally applied at this reference point, as shown in Figure 5b. Considering the ease of operation and the effectiveness of achieving equilibrium, this study adopts the ODB import method for geostress initialization.

2.2.3. Model Verification

To ensure the reliability of the parametric finite-element analysis, a numerical model was developed using the exact geometry of the collar monopile employed in the laboratory tests (Figure 6). The computed load–displacement response (Figure 7a) agrees well with the experimental data. A slight under-prediction of stiffness is observed for displacements below 0.4 cm, which can be attributed to the concrete preload blocks used during soil surcharging; these blocks increased the initial stiffness of the collar monopile–soil system in the physical test, thereby steepening the measured curve in the small-deformation range. As shown in Figure 7b, it can be found that the top displacement curve of the ring-wing single-pile foundation numerical model under 4000 cyclic loads fits well with the measured results of the indoor model test. Overall, the numerical model demonstrates reliable and accurate performance in reproducing the static load behavior of the collar monopile foundation.

To further validate the numerical approach and assess the applicability of the stiffness degradation model in cohesive soils, additional finite element models were developed based on the configurations reported in References [19,57]. These included both static and cyclic loading conditions (Figure 8). The simulated load–displacement responses showed close agreement with the experimental results, confirming the accuracy of the adopted constitutive model and the reliability of the numerical framework in capturing the deformation behavior of collar monopiles under different loading scenarios.

3. Result

3.1. Analysis of Monotonic Loading Tests

Under horizontal monotonic loading, the external force is shared by both the collar and the pile shaft; however, the manner in which the load is transmitted between these two components has yet to be comprehensively elucidated. To investigate this load-transfer mechanism, the deformation of the pile–soil interface at the mudline was examined under four discrete load levels H. Figure 9 illustrates the deformation patterns when horizontal loads of 20 N, 60 N, 120 N and 200 N were applied, and the observations are summarized below.

At an initial load of about 20 N, no discernible heave or cracking developed at the collar perimeter, and disturbance to the surrounding soil was minimal. This indicates that at low load levels, the lateral deformation of the collar-pile system is limited, the horizontal load is carried mainly by the pile shaft, and only a minor portion is transferred to the collar.

When the load was raised to roughly 60 N, visible cracks appeared on the soil surface in the lower right direction and directly beneath the collar. The arcuate cracking pattern along the collar edge demonstrates that the load share and lateral displacement of the collar had grown. A crack located about 0.1 D 2 below the collar center resulted from rotation of the pile–collar assembly, which lifted the collar edge to its maximum height at this position and destabilized deeper soil strata, confirming that the collar perimeter imposes the most pronounced disturbance.

With a further increase to about 120 N, the initial cracks deepened and propagated laterally along the collar perimeter. Pronounced horizontal sliding of the collar and indentation of the underlying soil became evident, signaling an increase in pile–soil relative displacement. The intensified disturbance enlarged the rotation angle of the collar, producing an additional crack about 0.3 D 2 below and to the left of the collar.

Once the applied load reached around 200, the sliding of the collar at the mudline became prominent, and the surrounding soil virtually lost its horizontal bearing capacity, with lateral displacement approaching 5 mm. Soil collapse occurred around the cracks, spawning numerous secondary cracks of enlarged dimensions. These observations reveal irreversible plastic deformation in the soil adjacent to the collar, a consequent reduction in mudline bearing capacity, and discernible uplift of one side of the collar.

3.2. Analysis of Cyclic-Loading Test Results

In saturated clay, all interparticle voids are completely filled with pore water. When subjected to horizontal cyclic loading, the collar monopile executes a reciprocating motion: the collar continuously rises and falls about its center of rotation, repeatedly compressing the soil skeleton beneath it. This process markedly increases the pore-water pressure in that zone, while the fabric of the surrounding soil is continually reworked, causing pore pressure to accumulate progressively. Because saturated clay possesses a low coefficient of permeability—and because the collar inhibits dissipation of excess pore pressure beneath it—the excess pore-water pressure in the underlying soil rises sharply during the early stages of cyclic loading (i.e., when the number of load cycles, N, is small). As N increases, however, disturbance induced by the collar monopile triggers irreversible plastic failure in the soil, and cracks open at the mudline along the outer edge of the collar. These newly formed cracks serve as drainage paths, enabling a rapid release of the excess pore pressure in the soil surrounding the cracks. Figure 10a shows that when N = 100, an elliptical water blister approximately 5 cm in diameter forms at the collar perimeter, demonstrating that the presence of the collar causes a sharp rise in pore-water pressure beneath it at the onset of cyclic loading, while the appearance of surface cracks allows the excess pressure to dissipate rapidly.

When the number of loading cycles rises to N = 300, disturbance in the soil surrounding the collar intensifies markedly. Figure 10b indicates that continued cyclic loading causes progressive softening of the soil skeleton outside the collar, leading to a further extension of the pre-existing drainage channel along the collar perimeter. The reciprocating “push-out–pull-in” motion of the collar monopile under the external load greatly amplifies the interaction between the collar edge and the adjacent soil, and the persistent oscillation steadily degrades the strength of the soil outside the collar. Concurrently, the collapse of soil around the drainage channel occurs under cyclic disturbance; the clay skeleton at the collar edge ruptures, liquefies into slurry, and churns under the motion of the collar, demonstrating that this portion of saturated clay has completely lost its load-bearing capacity and confirming that the soil distributed along the collar perimeter is the first to fail under cyclic loading. In addition, a fine crack approximately 5 cm in length is observed from the outer edge of the collar on the right side, extending about 1 cm outward toward the upper-right quadrant.

As the number of load cycles further increases to N = 600, the disturbance induced by the collar continues to accumulate. Figure 10c shows that the crack located at the upper-right quadrant of the collar propagates outward along the mudline, deepening and widening to create a distinct interface with the surrounding soil. Simultaneously, persistent degradation at the collar edge forms a trench approximately 5 mm wide along this interface. Moreover, the clay originally covering the upper side of the collar has transformed into a slurry, owing to the collapse of its soil skeleton under the external loading; loss of structural integrity in this zone renders the pore water expelled by cyclic loading increasingly turbid.

After approximately 4000 loading cycles, the specimen was allowed to rest for 30 min. As shown in Figure 10d, the original trench at the collar perimeter had widened to about 2 cm, and a series of outward-propagating cracks developed along the trench, forming a circumferential fracture zone. During the rest period, the water gradually clarified, revealing a metallic sheen on the upper surface of the collar. This observation indicates that cyclic loading had pulverized the soil overlying the collar into slurry, which mixed with the surrounding water. Upon quiescence, the pore-water pressure within the soil decreased and stabilized; the water previously expelled by excess pore pressure, together with suspended clay particles, flowed back through the cracks into the soil mass, thereby reducing the clay coating on the collar surface and exposing the metal beneath.

3.3. Analysis of Pore-Water Pressure Under Cyclic Loading

The foregoing examination of mudline failure in saturated clay demonstrates that variations in excess pore-water pressure manifest conspicuously during soil degradation—for example, through the emergence of water blisters within drainage channels around the collar. This highlights the pivotal role of excess pore-water pressure in governing the deformation of the collar monopile foundation. Accordingly, the present section investigates the evolution of excess pore-water pressure at monitoring points P1, P2, P3 and P4 as the number of loading cycles N increases.

As illustrated in Figure 11, the absolute magnitude of the measured excess pore-water pressure surrounding the collar monopile initially increases and subsequently decreases with the number of loading cycles. Monitoring points P1, P2 and P3 exhibit nearly identical trajectories—an initial rise followed by a gradual decline—because they are located 1 D₂, 5 D₂ and 9 D₂ below the mudline, respectively. Under horizontal cyclic loading, these points experience simultaneous compression arising from the interaction between the collar-pile shaft and the surrounding soil, which elevates the excess pore pressure. During the early stages of cyclic loading, dissipation is impeded, and the pressure therefore accumulates; once additional loading cycles create effective drainage pathways, the excess pore-water pressure begins to diminish.

Throughout the entire monitoring period, the excess pore-water pressure recorded at point P1 (located 1 D₂ below the mudline) exhibits the greatest intra-cycle fluctuation. Owing to its proximity to the collar in the vertical direction, P1 is most strongly affected by the collar–soil interaction under cyclic loading, and the surrounding soil is subjected to the most intense disturbance. The evolution can be summarized as follows: When the number of loading cycles is below 300, the excess pore-water pressure at P1 rises sharply. As the cycles increase to between 300 and 750, continual rearrangement of clay particles enlarges the drainage pathway at the mudline, so the pressure continues to increase but at a markedly reduced rate. Once the number of cycles exceeds 750, the drainage channel has become sufficiently developed to dissipate excess pore pressure rapidly; the pressure generated during each cycle is then smaller than the amount discharged, and the measured excess pore-water pressure diminishes as N continues to grow.

The trends recorded at points P2 and P3 resemble those at P1, although the cycle counts at which peak pressures occur shift to lower values with increasing depth. For P2, located 5 D₂ below the mudline, the excess pore-water pressure rises sharply when the number of cycles N is below 50; as N increases to between 50 and 400, the rate of increase diminishes; the pressure then tends to stabilize between 400 and 700 cycles, and once N exceeds 700, it gradually declines. At P3, positioned 9 D₂ below the mudline, a rapid pressure rise is observed for N<30; the growth rate slows between 30 and 330 cycles; relative stability follows in the range 330–470 cycles; and, beyond 470 cycles, the pressure begins to decrease. These observations indicate that, with greater embedment depth, the disturbance imparted by the pile shaft and collar to the surrounding soil under cyclic loading becomes progressively weaker, allowing the excess pore-water pressure to reach its peak after fewer cycles. Overall, the amplitude of pore-pressure variation diminishes vertically from P1 to P3 as depth increases.

At monitoring point P4—located below the pile’s zero-displacement plane—the recorded excess pore-water pressure is negative. When the number of load cycles N is below 60, the absolute magnitude of this negative pressure rises rapidly; between 60 and 300 cycles, the growth rate diminishes; and once N surpasses 300, the absolute value begins to decline. This behavior arises because the collar monopile undergoes rotation beneath the mudline under cyclic loading: points P1, P2 and P3 lie above the zero-displacement plane, whereas P4 lies below it, so the pile segment at P4 moves in the opposite direction, generating suction in the surrounding soil. Similar observations were reported in the laboratory monopile tests of Liao [58]. After approximately 300 cycles, further enlargement of the drainage pathway allows the negative pressure at P4 to dissipate progressively, mirroring the decreasing trends observed at P1, P2 and P3.

To further characterize the cumulative deformation of the collar monopile under horizontal cyclic loading, dynamic load–time histories at various cycle counts were constructed from the force recorded by the tension–compression load cell and the pile–head displacement measured by the displacement transducer (Figure 12). The results demonstrate a progressive degradation of pile–soil stiffness with increasing cycle number. During the first ten cycles, the soil on the side subjected to the larger load amplitude exhibits a pronounced stiffness reduction and fails rapidly; as the cycles continue, the influence of load amplitude on stiffness diminishes, and, once the number of cycles reaches approximately 1000, the incremental effect of additional cycles on stiffness becomes negligible.

4. Discussion

Previous investigations [19,39] indicate that the collar monopile, or hybrid monopile, attains a higher ultimate lateral bearing capacity than a conventional monopile foundation. When embedment of the collar itself is disregarded, its lateral performance is governed by three interacting mechanisms: frictional resistance beneath the collar rim, the restoring moment generated by the collar, and the interaction between the pile shaft and the surrounding soil. At the mudline, these contributions must balance the externally applied horizontal load and the pile–head bending moment. The mudline shear force in the shaft originates from the lateral load at the pile head and is mobilized through compression between the collar and the adjoining soil. Laboratory tests reveal that increasing horizontal load produces lateral displacement of the shaft and simultaneous sliding and rotation of the collar, yet the internal transfer of forces within the foundation cannot be fully resolved experimentally. To elucidate the distribution of shear and bending moments and to highlight the role of the collar, a numerical model of the collar monopile foundation was developed in Abaqus and subjected to detailed simulation.

4.1. Analysis of the Load-Transfer Mechanisms of Conventional Monopiles and Collar Monopiles

The comparative load-transfer analysis focuses on the shear force and bending moment acting at the mudline of the pile shaft, the basal friction developed beneath the collar, the externally applied horizontal load at the pile head, and the theoretical total bending moment at the mudline. The mudline shear force and bending moment are extracted from the ABAQUS 2020 output by means of a Python 3.8.2 script, whereas the basal friction beneath the collar is obtained by subtracting the mudline shear force from the applied horizontal load.

Figure 13a reveals a sequential evolution in the mudline shear force of the pile shaft within the collar monopile. In the initial phase, when the pile–head displacement is less than 0.12 m (approximately 0.1 D₂), both the shaft displacement and collar rotation are small, and the shear force mobilized by the interaction between the pile shaft and the surrounding soil dominates the overall lateral resistance. As the displacement increases further, both the mudline shear force and the horizontal friction beneath the collar rise concurrently. Once the pile–head displacement reaches about 0.23 m (roughly 0.2 D₂), the continually increasing horizontal load intensifies the bearing effect between the collar base and the seabed, causing a portion of the vertical load originally carried by the shaft to be transferred to the collar; consequently, the center of horizontal load transfer at the mudline gradually shifts toward the collar. During this interval, the collar resistance grows almost linearly, whereas the horizontal force borne by the shaft diminishes, so that although the shaft still contributes most of the lateral capacity, the collar’s share becomes appreciable. When the displacement exceeds 0.23 m (0.2 D₂), rotational effects become more pronounced, the friction under the collar is fully mobilized, and the collar assumes the primary role in sustaining the overall horizontal load.

At the mudline, the external bending moment induced by the horizontal load equals the sum of the restoring moment provided by the collar and the pile–head bending moment. Figure 13b shows that the theoretically calculated external moment—obtained by multiplying the pile–head horizontal load by the distance from the load application point to the mudline—greatly exceeds the bending moment extracted from ABAQUS at the mudline of the pile shaft. This disparity indicates that the collar’s restoring moment carries the vast majority of the external moment, clearly demonstrating that the addition of the collar markedly enhances the bending resistance at the pile head.

4.2. Analysis of the Lateral Bearing Capacity and Component Synergy of the Collar Monopile Foundation

Horizontal static loading was applied to three finite element models—namely, the traditional monopile model, the gravity-based collar model, and the collar monopile model—to evaluate the corresponding foundation responses (Figure 14). The geometric parameters of each model are as follows: traditional monopile foundation (e = 0.05 m, D₂ = 3 m, H = 20 m), gravity-based collar foundation (D₁ = 12 m, D₂ = 3 m, H₁ = 1.5 m), and collar monopile foundation (D₁ = 12 m, D₂ = 3 m, H₁ = 1.5 m, H = 20 m), where e denotes the monopile wall thickness, D₁ the collar diameter, D₂ the monopile diameter, H₁ the collar height, and H the monopile height. By applying horizontal loads at the loading point of each model and measuring the corresponding horizontal displacement of the pile shaft at the mudline, three load–displacement (P–Δ) relationships were obtained, as shown in Figure 15a: the P–Δ curve for the collar monopile foundation, the P–Δ curve for the traditional monopile foundation, and the P–Δ curve for the gravity-based collar foundation.

Furthermore, to analyze the lateral bearing advantages of the collar monopile foundation in saturated cohesive soil compared to the traditional monopile, the three sets of data obtained above were further processed and analyzed in this section. First, the shear force at the cross-section of the pile shaft was extracted from the numerical results of the collar monopile model to determine the portion of the applied horizontal load transmitted to the pile shaft at the mudline. This yielded the P–Δ curve for the monopile portion of the collar monopile foundation, where P represents the shear force at the mudline of the monopile. Second, two data sets from Figure 15a were linearly superimposed to derive the P–Δ curve representing the sum of the load at the head of the traditional monopile and the load at the top of the gravity-based collar. Lastly, the P–Δ curve for the collar portion of the collar monopile was obtained by subtracting the monopile shear P–Δ curve from the total applied load P–Δ curve of the collar monopile foundation. These operations produced three distinct P–Δ curves describing the horizontal resistance composition at the mudline of the collar monopile foundation, as shown in Figure 15b: (1) the mudline shear force P–Δ curve of the collar monopile shaft, (2) the P–Δ curve of the combined traditional monopile head load and gravity-based collar top load, and (3) the P–Δ curve representing the frictional resistance at the bottom of the collar in the collar monopile foundation.

The ultimate horizontal bearing capacity of a single pile was determined using the allowable displacement method, where the corresponding horizontal load at a surface displacement of 0.02 D₂ (i.e., 0.06 m) is defined as the ultimate horizontal capacity [59]. When the horizontal displacement reaches 0.06 m, both the traditional monopile and the collar monopile foundations reach their ultimate bearing states. At this point, the ultimate horizontal bearing capacity of the traditional monopile foundation (F_ult1) is 745 kN; the horizontal bearing capacity of the gravity-based collar foundation (F_H1) is 96 kN; and the sum of F_ult1 and F_H1 is 841 kN. The ultimate horizontal bearing capacity of the collar monopile foundation (F_ult3) is 1191 kN. Within the collar monopile foundation, the contribution from the monopile portion is 1037 kN, while the contribution from the collar portion (F_H2) is 154 kN.

Compared with the traditional monopile foundation, the monopile portion of the collar monopile foundation exhibits a capacity increase of 292 kN, representing 39.2% of F_ult1. The collar portion of the collar monopile foundation (F_H2) shows an increase of 58 kN over the gravity-based collar foundation (F_H1), equivalent to a 60% improvement. Overall, the ultimate horizontal bearing capacity of the collar monopile foundation (F_ult3) exceeds the combined capacity of the traditional monopile and gravity-based collar foundations by 350 kN, which corresponds to a 41.6% increase over the sum of F_ult1 and F_H1.

The synergistic interaction among the components of the collar monopile foundation enables its horizontal bearing capacity to exceed the simple sum of a traditional monopile and a gravity-based collar foundation, indicating a significant structural coupling effect. This cooperative mechanism not only enhances the bending resistance at the pile head and the lateral stiffness of the surrounding soil but also improves the bearing capacity of the monopile through the restoring moment and friction provided by the collar. Additionally, the presence of the monopile reinforces the soil beneath the collar, further increasing the overall sliding stability and ultimate bearing capacity of the foundation.

5. Conclusions

On the basis of model tests on collar monopiles embedded in saturated cohesive soils, this study systematically investigates pile–soil deformation under horizontal monotonic loading and cumulative deformation under horizontal cyclic loading. The temporal and spatial evolution of excess pore-water pressure around the pile is examined, and, in conjunction with finite-element simulations, the load-transfer mechanism of the collar monopile is elucidated. The principal conclusions are as follows.

(1)

As the horizontal monotonic load increases, cracks initiate around the collar and propagate outward, while pile–head displacement grows, indicating a progressively larger contribution of the collar to lateral resistance.

(2)

Under horizontal cyclic loading, the soil adjacent to the collar undergoes crack initiation, extension, local collapse, and trench formation and widening; the drainage path enlarges, and expelled pore water and fine particles re-infiltrate during the rest period.

(3)

Excess pore pressure is positive in the shallow zone and negative at greater depth, with the largest fluctuations near the surface; shallow sensors show a rise–fall trend, whereas deeper ones display a fall–rise trend.

(4)

In the early cycles, load amplitude governs stiffness loss; as the number of cycles increases, the amplitude effect diminishes and cycle count becomes the dominant factor controlling the stiffness evolution of the collar monopile–soil system.

It should be noted that this study does not consider combined vertical–horizontal loading, variable cyclic loads due to offshore environmental changes, or seismic effects, which will be addressed in future investigations.