Experimental and Numerical Simulation Studies on the Interface Characteristics Model of Loess and Bamboo Geogrid

Abstract

The widespread loess in western China poses significant challenges to transportation infrastructure construction due to its water sensitivity and collapsibility. This study investigates the interface mechanical properties of bamboo geogrid-reinforced loess under static loading through large-scale indoor pull-out tests and DEM–FDM coupled numerical simulations. The effects of vertical stress, the pull-out rate, the number of transverse ribs, burial depth, and reinforcement material on interface behavior were systematically evaluated. Results show that peak pull-out force increases with vertical stress, the number of transverse ribs, and burial depth, with all curves exhibiting pronounced strain hardening followed by softening characteristics. The pull-out rate exhibits a non-monotonic effect, with peak resistance higher at both lower and higher rates compared to intermediate rates. Bamboo geogrids demonstrate substantially superior performance over geogrids, with approximately four times higher peak pull-out resistance and greater initial stiffness. Numerical analysis reveals increased porosity and decreased coordination number in the grid vicinity, the horizontal stratification of the slip rate along the reinforcement, and concentration of strong force chains ahead of transverse ribs, elucidating the model-derived mechanisms underlying the macroscopic reinforcement effects. The findings confirm that bamboo geogrids provide effective and sustainable reinforcement for loess subgrades, offering a scientific basis for environmentally friendly engineering applications in loess regions. Although potential long-term durability under field environmental conditions requires further verification, the superior mechanical interface performance demonstrated here positions treated bamboo geogrids as a promising sustainable reinforcement option.

1. Introduction

With the ongoing development of western China and associated infrastructure expansion, transportation networks in the Loess Plateau region have expanded significantly [1]. However, the widespread occurrence of loess in this region presents major challenges for road construction. Loess is characterized by high porosity, well-developed vertical joints, pronounced water sensitivity, and collapsibility [2,3,4,5]. The construction of high-grade highways in such terrain frequently requires extensive high embankments and deep cuttings, rendering the subgrade highly susceptible to differential settlement, longitudinal and transverse cracking, and, in severe cases, sliding failure. These issues can substantially compromise project safety and long-term serviceability [6,7,8]. Geogrids, as an established soil reinforcement technique, have been shown to effectively restrict lateral soil displacement and improve overall subgrade stability [9]. Although traditional synthetic materials have high costs and lack environmental advantages, in response to the engineering concept of green and sustainable development, using fast-growing, high-strength and corrosion-resistant treated bamboo to make “bamboo geogrids” has become an economical and environmentally friendly alternative [10]. However, the interface interaction mechanism between bamboo geogrids and loess, as two materials with significantly different elastic–plastic properties, is complex [11,12]. Moreover, roads are subjected to vehicle cyclic dynamic loads during their service life, and the cumulative damage to the loess pore structure may lead to macroscopic roadbed subsidence. Therefore, in–depth exploration of the interface mechanism of bamboo geogrids reinforcing loess and the long-term stability under cyclic loading is of great theoretical and engineering significance for green road construction in the Loess region.

Bamboo geogrids were selected as a sustainable alternative to conventional plastic geogrids due to their rapid renewability, lower cost, and reduced carbon footprint, consistent with green infrastructure development in loess regions [10]. Although untreated bamboo is susceptible to environmental degradation, the 4–5-year-old culms used in this study were treated with boric acid-based preservatives and embedded at the fill–soil interface, which effectively mitigates exposure to UV radiation, moisture fluctuations, and biological attack. While long-term field durability remains to be fully validated, this study focuses on short-to-medium-term mechanical interface behavior under static loading as a necessary step toward engineering application. In the relatively dry loess environment of western China, reduced moisture and UV exposure further enhance durability. For projects requiring extended service life (20–60 years), more complex reinforcement systems are typically adopted; however, for moderate design–life subgrades and eco-friendly applications, treated bamboo geogrids remain a competitive and sustainable option. Further long-term cyclic loading and field exposure tests will be conducted to comprehensively evaluate durability. The stability of reinforced soil structures mainly depends on the interaction between the reinforcement and the soil interface, covering friction, adhesion and interlocking effects [13,14,15]. Currently, the academic community mainly studies this characteristic through direct shear and pull-out tests [16]. Existing studies have extensively analyzed the influence of normal stress, the pull-out rate, fill material compaction degree and grid geometric shape on the interface shear strength [17,18,19,20]. In terms of numerical simulation, compared with the limitation of the finite element method (FEM), in which it is difficult to simulate the discrete behavior of particles, the discrete element method (DEM) shows significant advantages in revealing the interlocking effect between particles, the evolution of force chains and the formation of microscopic shear bands [21,22,23,24,25].

In view of this, the present study aims to (1) experimentally quantify the interface shear behavior of bamboo geogrids in loess under monotonic static pull-out loading through large-scale indoor tests, (2) evaluate the effects of vertical stress, the pull-out rate, the number of transverse ribs, burial depth, and reinforcement material type on peak pull-out resistance and interface strength parameters, and (3) reveal the underlying microscopic reinforcement mechanisms via DEM–FDM coupled numerical simulations. By comparing bamboo geogrids directly with conventional plastic geogrids under identical conditions, this work clarifies the macro–micro linkages governing reinforcement efficiency and provides a scientific basis for the application of sustainable bamboo reinforcement in loess subgrades. The study is preliminary in nature, focusing on static mechanical performance; limitations and future research directions are discussed in the conclusions.

2. Materials and Methods

2.1. Materials

This experiment utilized the loess from the ongoing National Highway 310 southward extension project site, as shown in Figure 1a. In previous research by our team, the basic physical properties of the loess in this project were studied in the laboratory [26]. The basic physical properties of the loess in the section studied in this project are as follows: the natural moisture content decreases with increasing depth, with the average moisture content ranging from 12% to 22%; the liquid limit is 25.3%, the plastic limit is 17.2%, and the plasticity index is 8.085; the particle analysis shows the gradation of the loess, as shown in Figure 1b; the compaction test yielded a maximum dry density of 1.945 g/cm³, and the optimum moisture content was 11.965%; the bearing ratio (CBR) was 11.7% to 15.4% when the compaction degree was 93% to 96%, meeting the minimum CBR requirements for various grades of roadbed fill materials as stipulated in the “Highway Roadbed Design Specifications”. In summary, the physical properties of this loess filler are good and are suitable for use as a roadbed filling material. All physical properties reported above (including CBR values) are mean values determined from multiple laboratory tests on samples collected from the project site at different depths. Detailed statistical data are available in the team’s previous study [26].

The bamboo specimens were 4–5 years old, with culm diameters of at least 50 mm and wall thicknesses ranging from 5 to 15 mm. The material was fabricated into two structural forms: flat bamboo strips and hollow cylindrical tubes. Mechanical testing was conducted in three categories: tensile and flexural properties were determined in accordance with established methods for evaluating the physical and mechanical characteristics of bamboo used in construction; shear properties were assessed following standardized procedures for determining the shear strength of bamboo; and compressive behavior was evaluated based on the testing protocol adopted in previous studies on bamboo-reinforced composite anchor rods.

In this experiment, this material was used. The bamboo fiber materials were crosswise and lengthwise connected, and then fixed with rivets and iron wires to form a strong bamboo fiber grid material as shown in Figure 2a. The longitudinal and transverse bamboo strips (width 30 mm, thickness 8 mm) were spaced at 30 mm × 30 mm grid openings. Nodes were connected using stainless steel rivets (diameter 4 mm) and galvanized iron wires (diameter 2 mm). Prior to assembly, the bamboo was treated with a boric acid-based preservative to enhance corrosion resistance against moisture and acidity, as mentioned in the Introduction. The relevant parameters of the bamboo fiber grid are presented in

Table 1. The basic performance indicators of bamboo geogrids.

Material Dimensions/mm		Compressive Strength/MPa	Tensile Strength/MPa	Shear Strength/MPa	Bending Strength/MPa
Width	Thickness	69.49	12.62	189.07	152.80
4	8

Note: Values are means of at least three replicate tests [27].

. The bamboo fiber grid has high strength, multi-directional load-bearing capacity and good engineering applicability [27]. It can be used as an effective natural reinforcement material for roadbed reinforcement. The mechanical properties of the bamboo fiber grid are shown in Figure 2b.

The geogrid is made of bidirectional geogrid, which is made of polypropylene. The mesh size is 30 × 30 mm, as shown in

Table 2. Basic performance indicators of geogrids.

Mesh Size/mm	The Dimensions of the Transverse Ribs/mm		Ultimate Tensile Strength/MPa
	Horizontal	Vertical	Horizontal	Vertical
30 × 30	30	30	20	20

Note: Values are means of at least three replicate tests.

. The basic performance indicators are presented in Table 2. Before the test, the geogrid was cut to a size of 90 × 90 mm, and finally approximately formed into a 400 × 400 mm geogrid, with the mesh size and overall size being the same as those of the bamboo-reinforced geogrid.

2.2. Test Plan

2.2.1. Static Loading Pull-Out Test Plan

This study systematically designed a pull-out test scheme for bamboo geogrids in compacted loess using the control variable method, aiming to explore the influence of four key factors—vertical stress, pull-out rate, the number of cross ribs of the grid, and burial depth—on the interface interaction.

As shown in

Table 3. Test conditions for static load pull-out test.

Research Variables and Levels	Fixed Parameters	Experimental Group
(a) Vertical load test	Pull-out rate: 10 mm/min Number of transverse ribs: 3 Burial depth: 250 mm	10 kPa
		20 kPa
		30 kPa
		40 kPa
		50 kPa
(b) Pulling rate test	Vertical stress: 20 kPa Number of transverse ribs: 3 Burying depth: 250 mm	5 mm/min
		20 mm/min
(c) Number of transverse ribs test	Vertical stress: 20 kPa Pulling speed: 10 mm/min Burying depth: 250 mm	0 rib
		1 rib
		2 ribs
(d) Burial depth test	Vertical stress: 20 kPa Pulling speed: 10 mm/min Number of transverse ribs: 3	125 mm
		375 mm
(e) Grid material test	Number of transverse ribs: 3 Burying depth: 250 mm Vertical stress: 10–50 kPa Pulling rate: 5–20 mm/min	Geogrid
		Bamboo grid

, the static load test was divided into four series:

(a)

Vertical load test: In actual engineering, the loads received by the loess high-fill roadbed surface are generally not high. Considering the strength issue of the bamboo geogrid, this test used loads within the range of 50 kPa. Under the conditions of a pull-out rate of 10 mm/min, 3 transverse ribs, and a burial depth of 250 mm, this test applied five levels of vertical stress ranging from 10 to 50 kPa to establish the relationship between interface shear stress and normal stress and solve the interface strength parameters.

(b)

Pull-out rate test: This test used the pre-test method to measure the curve change pattern of the bamboo geogrid within the range of 5–20 mm/min. The rate value was selected as it was relatively suitable. Then, the influence of the rate on the pull-out of the bamboo geogrid was analyzed. Therefore, under the conditions of a vertical load of 20 kPa, 2 cross ribs, and a burial depth of 250 mm, this test set three pull-out rates of 5, 10, and 20 mm/min to evaluate the rate effect during the loading process.

(c)

Cross rib number test: Keeping the previous baseline conditions unchanged, four grids with cross rib numbers of 0, 1, 2, and 3 were used to reveal the passive impedance and locking mechanism of the cross ribs.

(d)

Burial depth test: Keeping other conditions unchanged, the burial depth of the grid was adjusted to 125, 250, and 375 mm to analyze the distribution law of interface forces at different burial depths.

(e)

Grid material test: Keeping the number of cross ribs of the geogrid and bamboo geogrid intact, separate tests were conducted on vertical load and pull-out rate. The vertical load test was carried out under vertical stress of 10, 20, 30, 40, and 50 kPa, and the pull-out rate test was conducted at a rate of 5, 10, and 20 mm/min to obtain the force laws of the geogrid and bamboo geogrid in the soil.

2.2.2. Pull-Out Test Procedure

Pull-out tests were performed using a custom-designed loading apparatus. Compacted loess was placed in layers into the test box to achieve the target dry density, with the bamboo geogrid reinforcement positioned horizontally at the predetermined embedment depth and aligned parallel to the pull-out direction. A predetermined normal stress was then applied via the vertical loading system and maintained constant throughout the test. Pull-out loading was subsequently initiated using a servo-controlled motor that displaced the gripping fixture (and attached reinforcement) at a constant rate. Throughout the test, a data acquisition system continuously recorded time-series measurements of normal stress, pull-out force, and fixture displacement at a fixed sampling frequency. These data were used to compute interface shear stress and to characterize the complete pull-out response. All pull-out tests were repeated at least three times under identical conditions.

2.3. Interface Strength Analysis Method

During the pulling test, the interface frictional resistance varies with the different strains at each point of the reinforcement material, showing a non-uniform distribution [28]. In this study, with a high-stiffness bamboo grid and high-hardness geogrid, the interface frictional resistance can be regarded as uniformly distributed and balanced with the pull-out force. This value is the shear strength of the grid–soil interface (${\tau }_{\mathrm{gz}}$), which can be expressed by the following Formula (1):

$${\tau }_{\mathrm{gz}}={\displaystyle \frac{F_{\max }}{2L_{\mathrm{e}}W}}$$

(1)

where F_max represents the peak pull-out force of the interface (kN); L_e represents the length of the geogrid buried in the loess fill material (mm); W represents the width of the geogrid buried in the loess fill material (mm).

In the actual pulling test, the vertical stress transferred to the interface between the reinforcement and the soil is not entirely from the vertical stress of the external load, but also includes the force of the soil and the test device [29]. Therefore, a real stress value needs to be obtained through conversion, which is called the effective normal stress (σ_v), and its calculation formula is as follows:

$${\sigma }_{\mathrm{v}}={\sigma }_{\mathrm{n}}+\gamma h+G_{\mathrm{b}}$$

(2)

where σ_n represents the normal stress (in kPa) generated at the interface between the reinforcement and the soil due to additional external loads; γ represents the density of the loess fill material (kN/m³), with a value of 17.9 kN/m³; h represents the laying height of the reinforcement material (m); and G_b represents the normal stress (in kPa) generated by the loading plate, which was measured and calculated to be 0.78 kPa by dividing the self-weight of the loading plate (determined using a calibrated electronic scale) by the contact area of the plate with the soil surface.

$$f^{\ast }={\displaystyle \frac{{\tau }_{\mathrm{gz}}}{{\sigma }_{\mathrm{v}}}}$$

(3)

The above coefficient is analogous to the friction coefficient. In pull-out tests, this coefficient can be adopted to characterize the interfacial pull-out resistance. We need to use the above formula in combination with the data obtained from previous tests, and finally calculate the strength coefficient of the bamboo grid and geogrid interface under different vertical stresses and different pull-out rates. From these coefficients, we can obtain the force distribution pattern between the reinforcement and the soil.

3. Test Results and Analysis

Pull-out tests revealed the interface shear behavior between loess and bamboo geogrid reinforcement, compared against conventional geogrids. The key variables assessed were normal stress, the pull-out rate, the number of transverse ribs, and embedment depth. Force–displacement curves generally exhibited an initial steep rise (elastic mobilization), peak strength, post-peak softening, and approach to residual resistance, attributable to progressive frictional and bearing resistance, shear band development, and interface rearrangement. Peak pull-out force, critical displacement, apparent shear strength parameters, and energy dissipation varied systematically with the tested factors. The subsections below present the experimental observations and quantitative comparisons, emphasizing mechanistic distinctions arising from the rigid nodal structure and surface characteristics of bamboo geogrids relative to plastic alternatives.

3.1. Analysis of the Influence of Vertical Load and Pulling Rate in the Pull-Out Test

3.1.1. Analysis of the Vertical Load Effect of Bamboo Geogrid Versus Conventional Geogrid

The indoor pull-out test results of the interface between loess and bamboo-reinforced grid are shown in Figure 3a. The curves in Figure 3a illustrate the trend of pull-out force changing with displacement. From the pull-out force–displacement curves under different vertical stress conditions, it can be seen that the pull-out evolution process of the loess–bamboo-reinforced grid has obvious hardening–softening characteristics. In the initial stage of pull-out, all curves show a steep upward trend, indicating a significant elastic stage, and the pull-out force increases rapidly with the increase in displacement; as the displacement continues to increase, the curves enter a nonlinear growth stage and successively reach the peak; after reaching the peak, they enter the stress softening stage, and the pull-out force begins to decrease with the displacement.

The curve patterns in Figure 3a show that vertical stress has a significant strengthening effect on the pull-out performance. As the vertical stress increases from 10 kPa to 50 kPa, the corresponding peak pull-out force steadily increases from 5.2 kN to approximately 9.98 kN, and the displacement required to reach the peak, that is, the critical displacement, also shows a slightly lagging trend with the increase in pressure. From the deep evolution law perspective, this phenomenon reflects that the increase in vertical stress significantly enhances the orthogonal stress at the surface of the grid and the interface with loess, thereby strengthening the friction resistance between the interface and the bamboo-reinforced nodes and the lateral impedance effect of the bamboo-reinforced nodes [22]. After reaching the peak, the curves show varying degrees of strain softening, and the pull-out force gradually decays and eventually approaches the stable residual strength, which is caused by the formation of shear bands at the interface and local damage of the soil structure [30]. The higher the vertical stress, the larger the area enclosed by the curve, indicating that the interface has a stronger ability to absorb energy before failure; that is, the bamboo-reinforced grid has more excellent reinforcement and anchoring performance under higher compaction or deep burial conditions.

3.1.2. Analysis of the Pull-Out Rate Effect on Bamboo Geogrid Versus Conventional Geogrid

In the pull-out test of the loess bamboo geogrid, by analyzing the pull-out force–displacement curves under different pull-out rates, it can be observed that the pull-out rate has a significant impact on the performance of the grid.

As shown in Figure 4a, at a pull-out rate of 10 mm/min, the force between the grid and the loess is the smallest. When the pull-out speed is reduced to 5 mm/min or increased to 20 mm/min, the force between the grid and the loess increases to a certain extent. From the relationship curves of the bamboo geogrid under different pull-out rates, it can be seen that within the test rate range, the pull-out performance under the 5 mm/min condition is the most excellent: its peak pull-out force reaches approximately 8.4 kN, and the ascending slope of the curve in the first half is the largest, reflecting extremely high initial interface stiffness; as the pull-out rate increases to 10 mm/min and 20 mm/min, the peak pull-out force respectively decreases to approximately 6.4 kN and 7.7 kN. Among them, the peak under the 10 mm/min rate shows a significant drop compared to the 5 mm/min rate, and at the 20 mm/min rate, the force between the bamboo geogrid and the loess increases significantly, indicating that after a certain critical speed, increasing the rate can increase the force between the bamboo geogrid and the loess, while the 20 mm/min rate has a peak higher than the 10 mm/min rate, but its arrival at peak displacement is significantly delayed, indicating that the formation process of the interface shear zone under a high flow rate is more complex. From a deep physical law perspective, the pull-out force of the bamboo geogrid mainly depends on its natural bamboo nodes and grid nodes’ strong squeezing and end-bearing effect on the loess [26]. At low-speed pull-out, the mechanical transmission between the soil particles and the grid nodes is more sufficient, and the particles can form a stable “force chain” structure through displacement rearrangement to resist the pull-out force; as the pull-out rate increases, the soil in the interface shear zone undergoes instantaneous structural disruption due to intense disturbance, and the rearrangement of soil particles lags behind the movement of the grid, resulting in a reduction in the effective passive soil pressure in front of the nodes. Moreover, the curves at different rates after the displacement exceeds 150 mm all trend toward a consistent residual strength, indicating that the rate effect mainly affects the dynamic interlocking stage before interface failure, and has a weak influence on the sliding friction characteristics of the interface after complete failure. Figure 4b presents the indoor pull-out test force–displacement curves of the geogrid under three loading rates. From the test results, it can be seen that the peak load of the geogrid under different loading rates has obvious differences. Among them, the peak load under the 5 mm/min condition is the largest, followed by 20 mm/min, and the peak under the 10 mm/min is the smallest. This indicates that the pull-out bearing performance of the geogrid is sensitive to the loading rate but does not monotonically increase with the increase in the rate. The influence of the rate on the pull-out force is similar to that of the bamboo geogrid, and this will not be elaborated here. All three curves show that the load first rises rapidly and reaches the peak under the increase in displacement, then enters a slow decline or basically stable stage, reflecting that the geogrid undergoes a stage evolution from elastic deformation to peak strength and then to grid structure adjustment and interface sliding dominance during the pull-out process.

3.1.3. Comparison of Interface Performance Between Bamboo Geogrid and Geogrid

Comparative pull-out tests under identical conditions (vertical stresses of 10–50 kPa and pull-out rates of 5–20 mm/min) revealed substantial differences in interface behavior between bamboo geogrids and traditional geogrids. Bamboo geogrids consistently exhibited peak pull-out forces approximately four times higher than those of geogrids across all tested conditions (

Table 4. Strength parameters of different grid interfaces under different vertical stresses.

Material Type	Normal Load ${\mathit{\sigma }}_{\mathbf{n}}$ (kPa)	Effective Normal Stress ${\mathit{\sigma }}_{\mathbf{v}}$ (kPa)	Section Shear Strength ${\mathit{\tau }}_{\mathbf{gz}}$ (kPa)	Frictional Coefficient f*
Bamboo geogrid	10	15.26	13	0.85
	20	25.26	16	0.63
	30	35.26	19.25	0.54
	40	45.26	23.5	0.52
	50	55.26	25	0.45
Geogrid	10	15.26	3.63	0.24
	20	25.26	4.13	0.16
	30	35.26	5	0.14
	40	45.26	5.75	0.12
	50	55.26	6.5	0.12

and

Table 5. Strength parameters of different grid interfaces under different pull-out rates.

Material Type	Pulling Rate v (mm/min)	Effective Normal Stress ${\mathit{\sigma }}_{\mathbf{v}}$ (kPa)	Section Shear Strength ${\mathit{\tau }}_{\mathbf{gz}}$ (kPa)	Frictional Coefficient f*
Bamboo geogrid	5	25.26	21.02	0.83
	10	25.26	16.09	0.64
	20	25.26	19.15	0.76
Geogrid	5	25.26	4.87	0.19
	10	25.26	3.63	0.14
	20	25.26	4.3	0.17

). This superior resistance is accompanied by markedly greater initial stiffness and more pronounced strain-hardening followed by softening in the pull-out force–displacement curves.

In contrast, geogrids displayed greater flexibility, resulting in an extended initial deformation phase and higher overall ductility. The delayed stress transfer in geogrids reflects gradual tensioning and progressive mobilization of frictional and bearing resistance, whereas the high rigidity of bamboo geogrids enables rapid mobilization of interfacial friction and passive bearing at the transverse rib nodes.

Interface shear strength parameters further highlight these differences: apparent friction coefficients for bamboo geogrids ranged from 0.45 to 0.85 under varying normal stresses and 0.64 to 0.83 under varying rates, significantly exceeding those of geogrids. The unique surface texture and nodal structure of bamboo geogrids provide enhanced embedment and interlocking with loess particles, leading to superior uplift resistance and better constraint of shear-induced settlement. These attributes make bamboo geogrids particularly effective for controlling early deformation in loess subgrades compared to conventional alternatives.

3.2. Analysis of the Influence of the Number of Transverse Ribs in the Pull-Out Test

For different numbers of transverse ribs, this study selected the vertical stress condition of 20 kPa, and respectively took the conditions of no transverse ribs, single transverse ribs, double transverse ribs and full transverse ribs for the bamboo grid, as shown in Figure 5a. The development law of the pull-out force–displacement curve was explored to more clearly analyze the mechanical action law of the number of transverse ribs on the bamboo grid in reinforcement. The test data showed that the number of transverse ribs had a significant impact on the improvement of interface mechanical properties, as shown in Figure 5b. Without transverse ribs, there was no obvious pull-out force peak. In the single transverse ribs condition of the grid, a clear pull-out force peak appeared after the transverse ribs began to form. From the single transverse ribs to the full transverse ribs, the increase in the pull-out force peak was not much. This indicates that the formation of the transverse and longitudinal rib structure of the grid is the core mechanism of the reinforcement effect of the bamboo grid, and the number of transverse ribs only enhances the grid strength within a certain range. Comprehensive study of the entire number of transverse ribs shows that the increase in the number of transverse ribs has a strong influence on improving the initial stiffness and peak strength of the interface [22]. The full transverse rib structure forms continuous end-bearing units in the pull-out direction, and its pull-out force peak and shear stiffness are the largest. This multi-transverse rib structure enhances the pull-out stability of the grid in the loess and has better mechanical stability than the single transverse rib structure.

The test data indicate that the number of transverse ribs has a significant impact on the improvement of the interface mechanical properties. Moreover, the data in

Table 6. Similar friction coefficients for different numbers of cross ribs.

Number of Ribs	Effective Normal Stress ${\mathit{\sigma }}_{\mathbf{v}}$ (kPa)	Section Shear Strength ${\mathit{\tau }}_{\mathbf{gz}}$ (kPa)	Frictional Coefficient f*
0	25.26	5.99	0.24
1	25.26	14.52	0.57
2	25.26	15.64	0.62
3	25.26	16.09	0.64

shows good logical consistency. From the perspective of the interface shear strength parameters, the addition of transverse ribs transforms the interface from smooth friction to mechanical adhesion, resulting in a sudden increase in cohesion, and the friction angle performance continuously improves with the increase in the number of transverse ribs. This improvement in microscopic parameters is directly reflected in the macroscopic shear strength: as the number of transverse ribs increases from zero to some quantity and from few to many, the cross-sectional shear strength steadily rises, driving the friction coefficient to increase synchronously. Thus, it can be seen that increasing the transverse ribs can effectively strengthen the interaction between the reinforcement and the soil, by enhancing the interface’s adhesion efficiency, ultimately achieving an overall improvement in the comprehensive shear resistance of the interface.

3.3. Analysis of the Influence of Embedding Depth in the Pull-Out Test

For different grid burial depths, this study selected the 20 kPa vertical stress condition and respectively took the bamboo geogrids with burial depths of 125 mm, 250 mm and 375 mm to explore the development law of their pull-out force–displacement curves, thereby more clearly analyzing the mechanical action law of the buried depth on the bamboo geogrid in the reinforcement process, as shown in Figure 6a. The test data show that the burial depth also has a significant impact on the improvement of the interface mechanical properties, as shown in Figure 6b. Numerically, the peak value of the pull-out force–displacement curve under the 125 mm burial depth is smaller than that of the 250 mm condition and also smaller than that of the 375 mm condition. This indicates that the deeper the burial depth, the greater the force on the grid. This is because it is subjected to higher soil gravity. However, the pull-out force values between adjacent conditions do not show a very obvious pattern. This indicates that in complex soil conditions, the force on the grid is not only affected by the overburden pressure but also by the position of the grid in the model box, such as the amount of soil on the upper and lower sides of the grid. This study shows that the position of the grid within the soil is an important factor affecting the interface mechanical law. By analyzing these three different burial depth conditions, to some extent, the influence law of burial depth on the force of the grid is analyzed and summarized.

As shown in

Table 7. Friction coefficients of interface at different burial depths.

Buried Depth	Effective Normal Stress ${\mathit{\sigma }}_{\mathbf{v}}$ (kPa)	Section Shear Strength ${\mathit{\tau }}_{\mathbf{gz}}$ (kPa)	Frictional Coefficient f*
125 mm	23.02	14.38	0.62
250 mm	25.26	16.00	0.63
375 mm	27.50	17.95	0.65

, the test data indicate that the interface mechanical properties do not increase linearly with the increase in burial depth; instead, they exhibit certain nonlinear characteristics. From the parameter comparison, as the burial depth increases, although the effective normal stress does rise, the fluctuations in interface cohesion and the non-continuous changes in interface friction angle restrict the growth of the macroscopic shear strength. This suggests that under deep burial conditions, the high stress environment may suppress the shear expansion effect of the soil around the cross ribs, preventing the strengthening of the bond between the reinforcement and the soil with the increase in pressure, resulting in the gradual flattening of the pseudo-friction coefficient at greater depths. In summary, the contribution of burial depth to the interface stability has a significant boundary effect. The improvement of the interface shear resistance is limited by the dynamic balance between the microscopic mechanical parameters and the macroscopic stress environment. The experimental results presented in Section 3 are now complemented by numerical interpretations in Section 4, where clear distinctions are maintained between measured macroscopic behavior and model-derived particle-scale insights.

4. Numerical Simulation Analysis of DEM Pull-Out Tests

To elucidate the model-derived microscopic mechanisms governing the loess–bamboo geogrid interface during pull-out, discrete element method (DEM) simulations were carried out using the coupled framework of PFC3D 7.00 and FLAC3D 7.00 software (Itasca Consulting Group, Inc., Minneapolis, MN, USA) based on the boundary wall method. This approach enables explicit representation of particle-level interactions, force chain evolution, and shear band development while incorporating continuum-scale boundary conditions. The model was calibrated against triaxial compression data and validated against large-scale laboratory pull-out results. The following subsections describe model setup and verification, post-processing techniques for microscopic parameters, and quantitative analysis of displacement fields, force chains, porosity, coordination number, and the slip rate under varying pull-out rates, transverse rib configurations, and embedment depths.

4.1. Model Establishment and Verification

4.1.1. Model Establishment for Pull-Out Test

To complement the experimental findings presented in Section 3, a DEM–FDM coupled numerical simulation was performed. The numerical results presented below are intended to provide mechanistic insights at the particle scale and should be clearly distinguished from the direct experimental measurements. This study employs the DEM–FDM coupling algorithm based on the boundary wall method. Through the wall elements in PFC, the forces and moments between the particles and the solid elements are transmitted, achieving the deformation coordination between the continuous medium and the discrete medium [31]. During the calculation iteration process, the velocity and force are exchanged in real time as the core coupling variables: the node velocity solved by FLAC is transmitted to PFC through the coupling boundary, and the updated interface force response of PFC is sent back to FLAC as the boundary condition, and the cycle iteration continues until the system satisfies the balance criterion [32]. The schematic diagram of the coupling calculation is shown in Figure 7.

The overall size of the model is set to 650 mm × 500 mm × 300 mm. The elastic model is constructed with a width of 400 mm and a rib width of 10 mm for the bamboo fiber grid, which is arranged at the central interface with a height of 150 mm. The loess is simulated using a parallel bonding model, and the particle size is reasonably enlarged to balance the calculation efficiency and macroscopic response accuracy. The particle size was enlarged by a factor of 5 (from the original loess gradation median diameter of approximately 0.03 mm to 0.15 mm) to improve computational efficiency while maintaining macroscopic behavior. A sensitivity analysis was performed by testing enlargement factors of 3, 5, and 7; the selected factor of 5 yielded the best agreement with experimental peak pull-out forces and triaxial stress–strain curves. The model construction is carried out in the order of layering sampling, servo preloading, embedding of reinforcement materials, servo consolidation, and coupling activation. Under the assumption that the initial stress field is only generated by self-weight, the contact force between the grid element and the soil particles is realized through the wall-zone command. The direct shear test simulates by dividing the upper and lower shear boxes through the horizontal layered wall, applying a constant normal stress to the top servo cover plate, and giving the upper box a horizontal velocity to trigger shear; the pull-out test directly controls the grid to move at a stable speed horizontally while maintaining a constant uniformly distributed load at the top. This coupling model aims to precisely reproduce the energy dissipation process, shear band evolution law, and load transfer mechanism at the reinforcement–soil interface at the micro-scale level, providing a reliable numerical basis for loess reinforcement engineering.

In the discrete element software, the numerical modeling of the bamboo geogrid pull-out test follows a rigorous mechanical loading logic. The numerical simulation modeling process for the bamboo geogrid pull-out test is as follows:

(a)

A rectangular model boundary is constructed using the wall element, and a layered compaction sampling method is employed to generate a granular assembly of loess with heterogeneous characteristics.

(b)

After the initial ground stress of the soil is balanced, the grid elements are embedded at the predetermined depth to construct a reinforcement–soil coupling model. During the boundary condition-setting stage, the particles at the ends of the grid are logically grouped and a constant horizontal velocity vector is assigned to simulate the displacement-controlled pull-out loading in the actual test.

(c)

After the model is established, a vertical load is applied until the system’s kinetic energy reaches the stable threshold.

(d)

Then, the horizontal velocity of the grid is activated, and the built-in historical monitoring pointer (History) is activated simultaneously to capture the evolution laws of the pull-out force and displacement in real time. The relevant process for establishing the model is shown in Figure 8.

4.1.2. Model Verification for Pull-Out Test

In the numerical simulation construction, the loess component is simulated using the discrete element method (DEM) to accurately capture the microscopic motion and stress characteristics of soil particles; the bamboo grid is modeled based on the linear elastic finite element model (FDM), and its mechanical parameters are set according to the previous experimental results of the research group [27].

Since the microscopic parameters of the particles in the discrete element model and the macroscopic mechanical indicators of the material are not linearly correlated, in order to ensure the representation accuracy of the numerical model, this study has carried out a strict calibration process for the loess parameters. Firstly, macroscopic stress–strain curves were obtained through indoor triaxial compression tests, and equal-sized triaxial numerical specimens were established for comparative simulation [33,34,35].

As shown in Figure 9a, the numerical simulation curves and the indoor test curves are in good agreement in terms of the initial modulus, peak strength, and evolution trend. Through repeated iterative corrections, the microscopic parameters of the loess determined finally are as shown in

Table 8. Parameter calibration results.

Parameter Types	Calibration Results	Parameter Types	Calibration Results
dp_nratio	0.5	pb_ten	2 × 104
porosity	0.3	pb_coh	2 × 105
emod	5 × 107	pb_fa	20
kratio	0.5	fric	0.8

, laying a reliable mechanical foundation for the subsequent simulation of the interface characteristics of the reinforcement–soil system [36].

To verify the reliability of the DEM–FDM coupling model, this study, based on the above-mentioned indoor large-scale pull-out test conditions, constructed a numerical verification model. By comparing the mechanical responses of the numerical simulation and the physical model test, the results are shown in Figure 9b. The peak pull-out force, strain softening trend, and interface displacement evolution law obtained through numerical calculation are highly consistent with the measured data from the test. Comparative analysis indicates that this coupling model can accurately capture the macroscopic mechanical characteristics and internal evolution logic of the reinforcement–soil interface, confirming the scientificity and feasibility of using the DEM–FDM coupling method to analyze the characteristics of the loess–bamboo grid interface, and showing it can be used for further in-depth microscopic mechanism exploration.

4.1.3. DEM Post-Processing Method

In discrete element numerical simulation, the macroscopic mechanical response curve can usually be directly output. However, the microscopic parameters such as porosity, coordination number, and the slip rate, as well as the spatial–temporal evolution characteristics of the cross-sectional stress field, need to be extracted through secondary development and post-processing using the FISH language. Currently, the post-processing methods for microscopic data mainly include two types: one is the spatial information statistics based on the measurement circle method; the other is the classification and aggregation of particles or contacts using the built-in GROUP command. Figure 10a shows the layout scheme of the measurement circles in the model, and Figure 10b is the contact point (Contact) grouping cloud map based on the XZ section.

In this study, the measurement circle method was employed to quantitatively characterize the porosity, coordination number, and stress field. The specific layout scheme was as follows: a 10 × 10 measurement circle array was uniformly constructed within the XZ section. Through the real-time coverage of the monitoring area, the spatiotemporal evolution laws of the micro-scale parameters during the numerical simulation process were dynamically obtained.

In the slip rate analysis, based on the built-in pointer commands provided by the discrete element software, FISH language functions were written to monitor and record the contact slip status during the simulation process in real time. The slip rate is defined as the ratio of the current number of slip contacts to the total number of contacts. To further achieve the spatially refined characterization of the slip rate, this study used the GROUP command to divide the model into regions (as shown in Figure 10), thereby quantitatively extracting the local slip rate evolution data of specific regions or specified layers. It should be noted that the porosity, coordination number, and the slip rate discussed in the following sections are model-derived quantities obtained from the calibrated DEM–FDM simulations rather than direct experimental microscopic observations.

4.2. Numerical Microscopic Analysis of Pull-Out Test Using DEM Numerical Simulation

4.2.1. Displacement and Force Chain Changes

The following analysis presents particle-scale quantities extracted from the DEM–FDM coupled model. These quantities serve as indicators of internal behavior within the numerical framework and should be interpreted in conjunction with the experimental macroscopic observations. In the discrete element numerical simulation of the pull-out test, the evolution of the displacement field at the interface of bamboo-reinforced soil and the reorganization of the force chain structure jointly reveal the mechanical essence of the interface transitioning from local compression to overall mobilization [37]. The displacement field analysis indicates that the particle displacement vectors exhibit a significant spatial stratification effect and progressive diffusion feature: the initial displacement field is highly confined to the first-level meshed fill soil at the pull-out end, presenting a clear localization concentration; as the pull-out displacement increases to the peak stage, the displacement field spreads to the entire reinforcement range, and the moving particles are around the grid area, presenting a horizontal diffusion and longitudinal distribution characteristic towards the pull-out direction, as shown in Figure 11a. This is because of the mechanical interception effect from the cross ribs, forcing the particles at the front edge of the cross ribs to undergo lateral displacement, significantly enhancing the overall anchoring force by increasing the soil bulging work. At the same time, the dynamic reconfiguration of the force chain structure further explains the skeleton characteristics of load transfer.

According to Figure 11b, during the pull-out process, the interface force chain undergoes a drastic transformation from isotropic to anisotropic. The strong force chains rapidly accumulate and coarsen at the front edge of the cross ribs, and form a strong force chain network radiating towards the depth of the soil. The simulation shows that after the pull-out is completed, the density of the force chain around the cross ribs reaches its peak, while the surface force chains of the longitudinal ribs are relatively sparse. This quantitatively confirms that the pull-out resistance is dominated by the passive resistance provided by the cross rib nodes. The coordinated evolution of the displacement and force chain proves that the bamboo grid lattice constructs a solid mechanical conduction framework through the coordinated displacement of the fill soil particles within the mesh and the continuous locking effect of the strong force chains at the microscopic level, thereby achieving efficient displacement restraint and strength enhancement of the loess fill in the macroscopic sense. The interpretations presented above are based on numerical modeling. Due to necessary simplifications (e.g., enlarged particle sizes for computational efficiency and idealized contact laws), the ability of the model to fully represent real-scale particle interactions remains limited.

4.2.2. Analysis of Model-Derived Microscopic Parameters of Interfaces with Different Rates

The influence of the pull-out rate on the model-derived microscopic mechanisms at the reinforcement–soil interface is illustrated through contour plots of porosity, coordination number, and the slip rate in Figure 12. At the lowest rate of 5 mm/min, particles had sufficient time for rearrangement, rolling, and interlocking, leading to the development of a well-defined passive resistance zone ahead of the transverse ribs [38]. This process induced pronounced dilation near the interface, resulting in a distinct high-porosity region centered around the grid as shown in Figure 12a. The coordination number exhibited a continuous, low-value band parallel to the interface in Figure 12b, reflecting substantial disruption of the particle contact network. The slip rate showed moderate localization within this zone in Figure 12c.

At an intermediate rate of 10 mm/min, the response displayed transitional behavior. Particle rearrangement lagged behind the imposed shear displacement, yet partial structural recovery persisted. Compared to the 5 mm/min case, the high-porosity zone became narrower and less intense in Figure 12d, while the low-coordination-number band remained present but with reduced continuity and width in Figure 12e. Slip activity was distributed over a broader area with intermediate gradients in Figure 12f, indicating the coexistence of localized damage and limited reorganization.

Under the highest rate of 20 mm/min, rapid shear induced intense relative sliding, with contact network disruption outpacing rearrangement. Localized dilation and coordination number reduction were less pronounced in Figure 12g,h, and the low-coordination band became discontinuous. In contrast, the slip rate field expanded significantly, reaching its highest values and showing the broadest high-slip region in Figure 12i. These observations suggest that elevated rates suppress deep structural adjustment while promoting more widespread particle mobilization and energy dissipation.

The macroscopic pull-out force–displacement curves under different rates, presented in Figure 4a, show the highest peak force of approximately 8.4 kN and the steepest initial stiffness at 5 mm/min, followed by a reduction to about 6.4 kN at 10 mm/min and a partial recovery to 7.7 kN at 20 mm/min with noticeably delayed peak displacement. These trends are mechanistically explained by the microscopic evolutions captured in Figure 12. At 5 mm/min, efficient particle rearrangement and pronounced dilation produce a distinct high-porosity zone centered around the grid, a continuous low-coordination-number band indicating extensive contact network disruption, and localized slip activity. At 10 mm/min, the reduced peak and transitional stiffness correspond to a narrower and less intense high-porosity region, partial continuity in coordination number reduction, and intermediate slip distribution, reflecting delayed reorganization amid ongoing damage. At 20 mm/min, the partial force recovery and broader post-peak response align with suppressed localized dilation, discontinuous low-coordination-number changes, and the broadest high-slip region, suggesting widespread particle mobilization that compensates for restricted deep structural adjustment and contributes to the observed non-monotonic rate dependence.

To further clarify the competing mechanisms behind the non-monotonic rate effect, the evolution of the three model-derived parameters is compared. At a shear rate of 5 mm/min, the pronounced high-porosity zone as illustrated in Figure 12a, the continuous low-coordination-number band shown in Figure 12b, and the localized slip presented in Figure 12c collectively indicate efficient particle rearrangement and strong interlocking, which lead to high energy dissipation through dilation and contact network disruption. At the intermediate rate of 10 mm/min, the narrower high-porosity region depicted in Figure 12d, the reduced continuity of the low-coordination-number band displayed in Figure 12e, and the moderate distributed slip shown in Figure 12f reflect insufficient time for stable force chain formation, resulting in minimum effective passive resistance and the lowest energy dissipation. At 20 mm/min, the broader but less intense high-porosity zone as seen in Figure 12g, the discontinuous low-coordination band illustrated in Figure 12h, and the most extensive high-slip region presented in Figure 12i demonstrate that rapid shear promotes widespread particle mobilization, allowing partial recovery of energy dissipation despite suppressed localized dilation. These differences clearly distinguish the rearrangement-dominated regime at low rates from the disturbance-dominated regime at high rates, with the intermediate rate representing the least efficient transition.

Although these model-derived trends in porosity, coordination number, and the slip rate provide valuable insights into possible particle-scale mechanisms, they are dependent on the calibrated contact model parameters and simplifying assumptions in the DEM simulation. Care should be taken not to overinterpret them as direct physical measurements.

4.2.3. Analysis of Model-Derived Particle-Scale Parameters at Different Numbers of Transverse Ribs

The evolution of microscopic interface parameters (porosity, coordination number, and slip rate) as a function of the transverse rib number is presented in Figure 13. Increasing the number of transverse ribs progressively shifts the interface interaction from predominantly frictional (longitudinal rib surface only) to a composite mechanism dominated by continuous passive end-bearing combined with friction.

In the absence of transverse ribs, pull-out resistance relies primarily on frictional interaction along the longitudinal elements. Particle rearrangement is minimal, leading to relatively uniform distributions of porosity and coordination number across the interface zone, with no pronounced localization or stress concentration evident in Figure 13a and Figure 13b. The slip rate remains low and broadly distributed, as shown in Figure 13c.

With the addition of a single transverse rib, localized compression and dilation develop immediately ahead of the rib, resulting in a distinct high-porosity region at the front edge in Figure 13d. The coordination number exhibits a well-defined low-value zone (minimum) in front of the rib in Figure 13e, indicating localized disruption of the particle contact network. The slip rate begins to show moderate concentration in the vicinity of the rib in Figure 13f.

Under the full transverse rib configuration, the mechanical interlocking and passive resistance effects are maximized. Multiple transverse ribs impose continuous and superimposed compression–shear loading on the soil, producing pronounced dilation that extends not only ahead of individual ribs but also merges between adjacent ribs, forming extended high-porosity regions in Figure 13g. The coordination number develops a continuous, band-like, low-value zone (dark blue) spanning the full reinforced length in Figure 13h, reflecting widespread and connected disruption of force chains. The slip rate reaches its highest values and forms a sharp, high-gradient slip layer, particularly in the central and trailing zones of the grid, as illustrated in Figure 13i. These results demonstrate that a dense transverse rib arrangement substantially enhances soil mobilization, intensifies localized structural disturbance, and maximizes micro-scale energy dissipation through enlarged zones of particle rearrangement and relative sliding.

The force–displacement curves in Figure 5b demonstrate a clear progression in pull-out resistance with an increasing transverse rib number: a negligible peak in the absence of ribs, the emergence of a distinct peak with a single rib, and the highest peak forces and sustained post-peak strength under the full-rib configuration. This macroscopic behavior is directly supported by the microscopic rearrangements shown in Figure 13. Without transverse ribs, the flat, low-resistance curve corresponds to uniform and minimal changes in porosity and coordination number, along with diffuse low-slip activity, consistent with friction-dominated sliding lacking significant interlocking. With a single transverse rib, the distinct peak force aligns with localized high-porosity development ahead of the rib, a concentrated low-coordination-number zone, and emerging slip concentration, indicating initial activation of passive bearing. In the full-rib case, the maximal peak forces and enhanced post-peak performance match extended high-porosity regions that merge between ribs, a continuous band-like low-coordination-number zone spanning the full reinforcement length, and the highest slip gradients in central and trailing areas. These micro-scale features reflect superimposed compression–shear loading and enhanced soil mobilization, which account for the observed transition from simple friction to composite end-bearing friction dominance and the limited additional gain beyond a certain rib density. Although these model-derived trends in porosity, coordination number, and the slip rate provide valuable insights into possible particle-scale mechanisms, they are dependent on the calibrated contact model parameters and simplifying assumptions in the DEM simulation. Care should be taken not to overinterpret them as direct physical measurements.

4.2.4. Analysis of Model-Derived Particle-Scale Parameters at Different Embedment Depths

The evolution of microscopic interface parameters (porosity, coordination number, and slip rate) with increasing embedment depth is illustrated in Figure 14. Increasing burial depth nonlinearly augments the effective normal stress and confinement at the reinforcement–soil interface, thereby influencing the spatial extent, intensity, and localization of microstructural changes during pull-out.

At the shallowest embedment depth of 125 mm, overburden pressure remains relatively low, resulting in loose particle constraints and limited pull-out disturbance. Changes in porosity and coordination number appear dispersed across the interface zone, with no strong localization evident in Figure 14a,b. The slip rate shows moderate, broadly distributed activity in Figure 14c.

At an intermediate depth of 250 mm, increased confining pressure suppresses free dilation and concentrates shear deformation within a narrower band adjacent to the grid. The high-porosity region becomes more pronounced and closer to the grid surface in Figure 14d, while the low-coordination-number zone exhibits greater concentration and improved continuity in Figure 14e. The slip rate displays a slightly broader but still localized distribution in Figure 14f, reflecting a transitional response between shallow and deep conditions.

Under the deepest embedment of 375 mm, high confining pressure compacts the soil tightly around the grid, requiring the pull-out process to overcome substantial passive resistance. This leads to intense localized dilation and structural reorganization within the shear zone. The high-porosity region reaches its greatest extent and intensity (darkest red) around the grid in Figure 14g, accompanied by the most pronounced and continuous low-coordination-number band (deepest blue) in Figure 14h. The slip rate increases markedly with depth, showing the highest values and the widest high-slip region in Figure 14i. These results indicate that, despite tighter macroscopic confinement at greater depths, the need to accommodate large grid displacement induces more extensive coordinated particle sliding and relative motion at the micro-scale.

The force–displacement curves in Figure 6b illustrate a nonlinear increase in pull-out capacity with embedment depth: the lowest peak at 125 mm, an intermediate rise at 250 mm, and the highest peak combined with a more sustained post-peak response at 375 mm. These macroscopic trends are mechanistically linked to the confinement-dependent microstructural changes presented in Figure 14. At the shallowest depth of 125 mm, the lower peak and broader softening correspond to dispersed variations in porosity and coordination number, accompanied by moderate and diffuse slip activity, reflecting loose particle constraints and limited localized dilation. At 250 mm, the transitional increase in resistance aligns with a more concentrated high-porosity zone near the grid surface, improved continuity in the low-coordination-number reduction, and slightly broader slip distribution, indicating progressive narrowing of the shear zone under rising confinement. At 375 mm, the highest peak and enhanced resistance match the most intense localized dilation, with the widest and deepest high-porosity region around the grid, the most pronounced continuous low-coordination-number band, and the maximal extent and gradients of slip activity. These observations demonstrate that greater overburden pressure intensifies shear zone activity and coordinated particle sliding at the micro-scale, thereby supporting superior macroscopic performance under deep burial conditions, albeit with boundary-limited gains due to suppressed dilation. Although these model-derived trends in porosity, coordination number, and the slip rate provide valuable insights into possible particle-scale mechanisms, they are dependent on the calibrated contact model parameters and simplifying assumptions in the DEM simulation. Care should be taken not to overinterpret them as direct physical measurements.

5. Conclusions

This study systematically investigated the interface mechanical characteristics between bamboo geogrids and loess through large-scale pull-out tests and DEM–FDM coupled numerical simulations. The key findings are summarized and discussed as follows:

(1)

The pull-out tests and coupled DEM–continuum simulations demonstrate that bamboo geogrids significantly outperform conventional plastic geogrids at the loess interface. Under identical conditions, bamboo geogrids consistently achieve peak pull-out forces approximately four times higher, accompanied by markedly greater initial stiffness and distinct strain-hardening followed by softening. This enhanced performance stems from the material’s high rigidity, which facilitates rapid mobilization of interfacial friction and passive bearing at transverse rib nodes, thereby providing superior anchorage and early deformation control in loess fills.

(2)

Macroscopic pull-out resistance increases systematically with normal stress, transverse rib number, and embedment depth, exhibiting characteristic post-peak softening attributable to shear-band formation and localized soil damage. Transverse ribs transform the interaction from predominantly frictional to a composite mechanism incorporating adhesion and continuous passive end-bearing, resulting in substantial gains in apparent interface shear strength and friction coefficient. Embedment depth exerts a nonlinear strengthening effect through increased confinement, although gains diminish at greater depths due to suppressed dilation. The pull-out rate exerts a non-monotonic influence: peak resistance is highest at the extremes (5 mm/min and 20 mm/min) and lowest at the intermediate rate (10 mm/min). This pattern reflects rate-dependent, particle dynamics-efficient rearrangement and stable force chain formation at low rates versus intense disturbance, lagged reorganization, and widespread mobilization at high rates.

(3)

Numerical results from the DEM–FDM simulations reveal that pull-out induces localized dilation (increased porosity), contact network disruption (decreased coordination number), and stratified slip along the reinforcement, with strong force chains concentrating ahead of transverse ribs. Systematic variations in the rate, rib configuration, and depth alter these model-derived evolutions. These numerical findings directly support the observed macroscopic strength enhancement and softening behavior obtained from the laboratory pull-out tests.

(4)

Limitations and future perspectives: This study provides valuable insights into the static interface behavior of bamboo geogrids in loess; however, several limitations should be acknowledged. The experiments were conducted under monotonic static loading only, without considering cyclic traffic loads or long-term environmental exposure (UV radiation, moisture fluctuations, acidity). The DEM–FDM, while well-calibrated, is a simplified 3D representation that does not fully capture particle shape variability or biodegradation effects. Consequently, the present work should be regarded as a preliminary mechanical investigation. Future research will focus on cyclic pull-out and direct shear tests under simulated traffic loading, accelerated environmental aging tests to quantify fatigue life, and full-scale field trials combined with multi-physics numerical modeling that integrates mechanical, hydraulic, and chemical degradation processes. Such a complex approach is essential for the safe, long-term application of bamboo geogrids in real engineering projects.

In conclusion, bamboo geogrids offer a robust, sustainable reinforcement alternative for loess subgrades, particularly in high-fill applications susceptible to differential settlement. Their superior interfacial performance and environmental advantages support wider adoption in collapsible-soil regions. The established macro–micro linkages provide a mechanistic foundation for optimized design and highlight the need for future field-scale studies under combined loading and environmental conditions.