The contact model selection is based on the nature of the materials. Rock, an anisotropic and non-homogeneous material, breaks and fractures under cutting forces. To simulate particle bonding, this study adopts the Hertz–Mindlin contact model, which is the default in EDEM due to its accuracy and computational efficiency. In this model, the normal force Fn is a function of the normal overlap δn, defined as 3Fn=43E∗R∗δn32. Equivalent Young's modulus E∗ and equivalent radius R∗ are defined as 41E∗=1-vi2Ei+1-vj2Ej,51R∗=1Ri+1Rj, where Ei, vi, and Ri and Ej, vj, and Rj are the Young's modulus, Poisson's ratio, and radius of each contact sphere.

The damping force Fnd is 6Fnd=-256βStm∗vnrel,7β=-ln⁡eln⁡2e+π2,8Sn=2E∗R∗δn,9m∗=1mj+1mi-1, where e is the recovery coefficient, and vnrel is the relative normal velocity.

The Hertz–Mindlin with bonding model describes interactions between bonded particles. A bond is created upon contact and resists both normal and tangential motions until the stress exceeds a critical threshold, at which point the bond breaks. This effectively simulates fracture and crushing in rock and concrete. Once bonding is established, forces and torques are initialized to zero and adjusted incrementally at each time step as follows: 10δFn=-νnSnAδt,11δFt=-νtStAδt,12δMn=-ωnStJδt,13δMt=-ωtSnJ2δt, where A=πRB2 and J=12πRB4.

The bond is broken when the normal and tangential shear stresses exceed predefined values: 14σmax⁡<-FnA+2MtJRB,15τmax⁡<-FtA+MnJRB. This study uses a soft-rock material with a compressive strength of 2.40 MPa and tensile strength of 0.5 MPa. The key bonding parameters are listed in Table 2 (Tian et al., 2019).

Virtual uniaxial compression and Brazilian splitting tests were conducted to determine the compressive and tensile strengths of the bonded-particle model. Strength is calculated from the force at which the bonded particles fail macroscopically.

For the Brazilian test, the particle bed was modeled as a 1200 mm diameter, 600 mm thick disk. At t = 0, the upper platen applied pressure at 0.1 m s⁻¹ until the sample split. The schematic and force diagrams are shown in Figs. 2 and 3, respectively.

The uniaxial compression test used a cylinder with a 350 mm radius and 1600 mm height. At t = 0, the upper platen compressed the sample downward at 0.01 m s⁻¹ until failure. The virtual uniaxial compression simulation showed that bond breakage was first initiated in the central region of the specimen and gradually propagated along an inclined plane, forming a typical shear failure pattern consistent with the failure mode observed in the physical uniaxial compression test.The resulting damage schematic and force diagrams are shown in Figs. 4 and 5.

The compressive and tensile strengths were calculated by substituting the peak platen forces into the following equations: 16σc=Fmax⁡πR2,17σt=2Fmax⁡πDL. The calculated uniaxial compressive strength is 2.11 MPa, and the Brazilian splitting strength is 0.47 MPa. These values meet the required strength criteria for soft-rock simulation, indicating that the parameter calibration is complete.

The Archard wear model is a contact model that extends the base interaction model to estimate the depth of wear on geometric surfaces. Developed by John F. Archard in 1953, the model is based on the principle that the volume of material removed from a surface is proportional to the frictional work performed by particles sliding over it. The model defines wear depth using Archard's equation, which incorporates both frictional energy and the geometric characteristics of the surface. Archard's equation is expressed as follows: 18Q=WFndt, where Q is the volume of material removed, dt is the tangential distance traveled, and W is the wear constant initially defined as 19W=KH, where K is a dimensionless constant, and H is the hardness measure of the softest surface (based on Vickers hardness in Pa). This was reduced to a single wear constant for ease of use in EDEM.

As the equations predict volumetric material removal, they are rearranged in the software to yield wear depth at each computational cell. 20weardepth=QA In the EDEM software, we set the wear constant W in Pa⁻¹. In this study, given the limited cutting time, the actual wear generated is minimal and difficult to quantify experimentally. Therefore, the wear constant W = 1 × 10⁻¹² was selected based on literature research to enable qualitative analysis of the cutter's wear behavior.

For the EDEM particle bed generation, spherical particles were used to simplify calculations. The particle radius was set to 15 mm with a standard deviation of 0.5 under a normal distribution, producing particles ranging from 11.25 to 18.75 mm. The bonding contact radius was set to 20 mm. The particle bed dimensions were 2 m (length) × 1.5 m (width) × 1.5 m (height). A total of 199 813 particles were generated using dynamic particle generation.

The cutter model, designed in SolidWorks (Sect. 1), was imported into EDEM. Material properties were defined as follows: Poisson's ratio = 0.3, density = 7800 kg m⁻³, shear modulus = 8 × 10¹⁰ Pa, and Young's modulus = 2.08 × 10¹¹ Pa. The cutter and cutter teeth were positioned appropriately. The initial position is shown in Fig. 6.

Before conducting the numerical simulation of cutter–rock interaction, it is necessary to appropriately define the simulation domain and related computational parameters to ensure both accuracy and efficiency. The setup mainly includes (1) defining the cutter motion, (2) determining the simulation domain, (3) specifying the total simulation time and time step, (4) setting the data output frequency, and (5) defining the mesh size.

Setting the motion state of the cutter head. After importing the cutter-head model, linear translation and rotation were applied to define its motion. The “move-with-body” option was enabled for the rotational motion to ensure that the rotation axis followed the cutter body, and appropriate transverse velocity and rotational speed were assigned according to actual operating conditions.

At 0.8 s, the cutter begins reverse cutting as it contacts the particle bed. As the cutter rotates, it applies force to the bed, breaking the bonds between particles. The rock material beneath the cutter teeth fractures into varying fragment sizes. Some unbroken bonded regions cause rock particles to be ejected outside the cutter's envelope, while others are retained inside the cutter arms. To guide debris flow for subsequent removal, the cutting process continues until 7 s, ensuring each cutter arm completes at least two full cutting cycles. The cutting process is illustrated in Fig. 7.

With a traverse speed of 0.2 m s⁻¹, rotational speed of 30 rpm, and cutting angle of 30°, eight cutter teeth on one cutter arm were selected for wear analysis. The wear of cutter teeth no. 1 through no. 8 is shown in Fig. 8. The maximum wear per tooth is shown in Table 3. Each tooth exhibits a stepped wear pattern due to simultaneous engagement with the rock. Friction between the rock and teeth increases cutting force, leading to rapid wear until the rock fractures. Once broken, the cutter tooth force drops sharply, and the wear curve flattens.

Quantitative analysis of the wear gradient distribution reveals that tooth no. 1, located near the hub, experiences the most wear (average thickness loss on the front/side/rear flanks: 3.56 × 10⁻⁴ mm), while tooth no. 8, located near the outer ring, shows the least wear (average thickness loss: 1.3 × 10⁻⁵ mm) – a nearly 30-fold difference. This aligns with real-world wear behavior. Teeth on the outer ring experience lower contact angles and shallower cuts, which reduce abrasive and adhesive wear. In contrast, hub-side teeth encounter larger cutting angles and deeper cuts, altering the direction of resultant cutting forces. This increases interface friction and prolongs cutting time, intensifying deformation and wear diffusion. Wear results for all six cutter arms are presented in Fig. 9, confirming the consistent pattern of greatest wear at the hub side and the least wear at the outer ring. These findings suggest that hub-side teeth require closer monitoring and timely replacement.

The Relative Wear model in EDEM identifies regions of high-impact (normal) and abrasive (tangential) wear by analyzing relative velocity and contact forces between particles and geometry. While EDEM does not calculate an absolute material removal rate, forward and tangential cumulative contact energies provide insight into wear intensity. These values represent energy from direct particle impacts and sliding along the surface, respectively. In this study, the forward and tangential cumulative contact energies for the most worn cutter tooth were extracted to analyze the wear mechanism. A comparative energy profile is shown in Fig. 10.

The interaction force between the rock and cutter tooth surface can be decomposed into normal force (N) and tangential force (T). The normal force presses the material into the tooth surface, while the tangential force causes it to slide along the surface, generating wear. As shown in the figure, tangential cumulative contact energy exceeds forward energy, indicating that chisel-type abrasive wear is the dominant wear mode. This is consistent with microscopic cutting mechanisms. Similar wear energy patterns were verified for the other cutter teeth.

During post-processing, the moment of maximum cutter tooth engagement was selected, and both force and wear distribution maps were extracted (Fig. 11). These reveal that regions experiencing higher forces closely align with areas of severe wear, particularly at the tooth tip and the lower side. Wear on the upper surface of the tooth was also found to be more pronounced than on the lower surface.

In the above study, the force and wear of the cutter were analyzed under conditions of a traverse speed of 0.2 m s⁻¹, a rotational speed of 30 rpm, and a cutting angle of 30°. In the post-processing stage, the total cutting force and torque of the cutter were examined. These are shown in Fig. 12. During the rock-cutting process, fractured rock pieces detach from the surface, causing the force and torque values to exhibit pronounced oscillations and intermittent zeros. As the cutting deepens, the cutter engages more fully with the material, resulting in a gradual increase in force, consistent with realistic cutting conditions.

For soft rock with a compressive strength of 2.40 MPa and a tensile strength of 0.5 MPa, experimental studies by Zhang et al. (2018) report an average cutting torque of 5464 Nm. The average simulated torque in this study is 6093.3 Nm, with a relative error of 11 %, indicating good agreement between the simulation results and experimental data.

In practical dredger operations, cutter parameters must be adjusted according to working conditions. These include traverse speed, rotational speed, and cutting angle, all of which significantly influence cutting force and torque. In this section, the impact of these parameters is analyzed by varying them individually to explore their effects and to identify general trends.

The selection of working condition parameters was based on typical operating ranges of cutter suction dredgers in soft-rock dredging projects, as reported in engineering practice and previous studies. To ensure the representativeness and practical relevance of the simulations, traverse speeds of 0.15–0.3 m s⁻¹, rotational speeds of 20–40 rpm, and cutting angles of 20–40° were adopted. These ranges correspond to the common operating conditions under moderate soil resistance, allowing for a comprehensive analysis of the influence of each parameter on cutter forces, torque, and energy consumption.

First, keeping the cutter's rotational speed fixed at 30 rpm and the cutting angle at 30°, the traverse speed was varied across four values: 0.15, 0.2, 0.25, and 0.3 m s⁻¹. Simulations were conducted, and data from 2 s onward (when the cutter fully engages with the rock) were analyzed. Figure 13 presents the cutting force at different traverse speeds, Fig. 14 shows the corresponding torque, and Fig. 15 summarizes the average values. The average total force and torque of the cutter at different speeds are shown in Table 4.

The results clearly demonstrate that traverse speed significantly affects cutter force. As traverse speed increases from 0.15 to 0.3 m s⁻¹, average cutting force rises from 10 227.7 to 39 880.7 N, while average torque increases from 3718.6 to 10 369.2 Nm. This nonlinear increase follows a quadratic relationship: as traverse speed increases, so does cutting thickness, leading to a squared growth in cutting force (i.e., F∝v2). Therefore, in conditions involving excessive cutter loading, reducing traverse speed can effectively lower the resultant force on the cutter head and improve operational safety.

Next, cutter rotational speeds of 20, 25, 30, 35, and 40 rpm were tested at fixed traverse speeds of 0.2 and 0.25 m s⁻¹, with a constant cutting angle of 30°. The force and torque data under these conditions are shown in Figs. 16 and 17, and the average values are compared in Fig. 18. Average total force and torque data are shown in Tables 5 and 6.

By comparing cutter forces under different rotational speeds at traverse speeds of 0.2 and 0.25 m s⁻¹, the total cutting force and torque were obtained. It was observed that both total force and torque decrease as the cutter's rotational speed increases. This phenomenon arises from the influence of cutting thickness on the rock crushing force. As the cutter's rotational speed increases, the frequency of contact between the cutter teeth and the rock increases per unit time, leading to a decrease in the cutting thickness per tooth. At lower rotational speeds, the cutter experiences higher cutting forces and torque. As the speed increases, the load on the cutter decreases and gradually stabilizes, with the relationship between cutting force and rotational speed approximated as F∝n-1. Cutter speed also significantly affects fragment size: higher speeds produce smaller rock fragments, which facilitates debris transport. Compared with traverse speed, the cutter's rotational speed has a smaller impact on cutter load. Therefore, it is recommended to first calibrate the traverse speed to control peak loads, followed by adjustments to the rotational speed to optimize energy consumption in practical applications. The following section analyzes the effect of cutting angle on cutter performance. The cutting forces and torques under different cutting angles are shown in Fig. 19. Average total force and torque under different cutting angles are shown in Table 7.

When only the cutter angle is varied, the average cutting force and torque remain within a normal fluctuation range at angles of 10, 20, and 30°, suggesting that the cutting angle has limited influence on average force and torque. However, the fluctuations in force and torque are more pronounced at smaller cutting angles. The cutter operates more smoothly at a cutting angle of 30°, indicating improved operational stability under this condition.

According to the calculation based on the results under various operating conditions, the optimal scenario for minimizing force and wear on the cutter involves a lower traverse speed and higher rotational speed. However, excessively high rotational speed can increase power consumption, and excessively low traverse speed can reduce dredging efficiency. Therefore, operational parameters should be selected based on a balanced consideration of energy efficiency, equipment wear, and project timelines.

Notably, force and torque alone are not sufficient indicators of cutting efficiency. A more comprehensive metric is specific cutting energy, defined as the energy consumed per unit volume of removed material. It is a key parameter for evaluating cutting performance and is calculated as 21η=2πnT+FnνSν, where n is the rotational speed of the cutter, T is the torque of the cutter, Fn is the transverse cutting force of the cutter, v is the transverse speed of the cutter, and S is the cutting area of the cutter. As shown in Fig. 20, the cutter cutting area S was calculated as 0.699 m² using EDEM post-processing. The transverse cutting forces under different conditions are listed in Table 8.

Substituting the relevant values into Eq. (21), the specific energy results are shown in Fig. 21 and Table 9.

Analysis of the specific energy values shows that at lower rotational speeds, specific energy increases noticeably with speed, but the rate of change levels off at higher speeds. Conversely, an increase in traverse speed leads to a significant rise in specific energy. From the results, the optimal operating condition for cutting soft rock is a traverse speed of 0.2 m s⁻¹ and a rotational speed of 20 rpm, yielding the lowest specific energy of 130.5 kJ m⁻³. Under this condition, cutter force and wear are also minimal, making it the most efficient configuration for soft-rock cutting.