Design of conical seed separator plate for wide belt seeder based on DEM and experiment

1 Introduction

As one of the three major food crops, wheat yields are closely linked to the national economy and people’s livelihoods (Dadrasi et al., 2023). At present, the world is experiencing the second food security crisis (Yaseen et al., 2024). In China, where the population is large and arable land is scarce, wheat cultivation areas continue to decline (Xi et al., 2020). Therefore, how to increase wheat yields and ensure food security has become a major challenge to the agricultural sector. The traditional method of sowing wheat was broadcast, which resulted in uneven seed distribution and led to issues such as competition for sunlight, nutrients and water (Tobisa and Nakano, 2024; Wang Y. et al., 2024). This method has gradually become unable to meet the demands for high and stable yields (Luo et al., 2022). Wheat wide belt sowing is an innovative method in China to improve production, which is the practice of uniformly distributing wheat seeds into seed furrows at a specified width, ensuring uniform emergence and promoting better crop growth compared to broadcast sowing (Lv et al., 2020). This method not only increases yields but also enhances the crop’s utilization of light, water and fertilizer (Cheng et al., 2021).

At present, many researchers have studied wide belt seeders. Wang et al. (2022) designed a test platform for throwing soil after sowing with wide belt sowing device, which could complete multiple functions at once, such as straw crushing, straw lateral concentration and uniform sowing. Furthermore, a discrete element simulation model of soil throwing after sowing was established to optimize the platform. Zhao et al. (2022) designed the structure of seed metering device and positive negative pressure air chamber for a wide belt sowing device to improve the uniformity of wheat strip sowing, and optimized the design using EDEM-Fluent coupling technology. Zhao P. et al. (2024) proposed a novel front-mounted fertilizer pipe wide-seedbed strip rotary tillage device to improve low straw cleaning efficiency, inadequate soil pulverization, inconsistent sowing depth and high-power consumption. Discrete element simulation models were established for optimization, and the design was validated through field experiments. He et al. (2024) designed a dual-pressure seeder for pre-sowing and seed furrow of wheat in wide sowing belts to solve the problem of poor sowing quality caused by loose soil in wheat seedbed after land preparation, and the results showed that pre-sowing compaction and furrow compaction improved the seedbed soil environment and increased the uniformity of sowing depth. Although there has been extensive research on wide belt sowing devices, the research and development on seed separator plate, as critical component of such devices, remains relatively insufficient. Luo et al. (2023) designed six structures of uniform seed device for wide width seeder, the main parameters of these uniform seed devices include chord length of the ridge, installation inclination, the angle between the chord and tangent of the end of the ridge, span and bottom curve radius. Although they achieved good lateral uniformity in sowing, their mathematical models and structures were overly complex, requiring control of numerous main parameters of uniform seed device, which made them difficult to implement. Therefore, it was necessary to conduct further research on improvements to uniform seed device.

Based on the above research, this paper proposed a conical seed separator plate to solve the problem of non-uniform seed distribution from wheat wide belt sowing device. Firstly, a mathematical model was established based on the conical seed separator plate to determine the critical parameters. Secondly, a discrete element simulation model was developed to optimize these parameters and identify the optimal combination. Finally, the results were validated through indoor and field experiments. This study could provide a reference for the design of seed separator plates for wheat wide belt sowing devices and for improving the lateral uniformity of sowing.

2 Materials and methods

2.1 Agronomic model of wheat wide belt sowing

In general, the sowing width of wheat wide belt is 60–120 mm, the spacing is 70–170 mm, the seed depth of wheat is 30–50 mm, and the fertilizer is 50–80 mm below seed (Wang et al., 2022). In this paper, the wide belt seeder manufactured by Xi’an Ya’ao Agricultural Machinery Co., Ltd. was selected, with a sowing width of 120 mm and a spacing of 150 mm, as shown in Figure 1, which has been widely used in China (Lv et al., 2020).

2.2 Structural design of wide belt sowing device

Wide belt sowing device was the key component of wheat wide belt seeder. The structure of wide belt sowing device was shown in Figure 2, which mainly included seed drop tube, cover plate for seed protection, side plate for seed protection, seed separator plate. The upper end of seed drop tube was connected with seed discharge device. Wheat seeds traveled through seed drop tube to seed separator plate, where they collided with it and then fell randomly onto seedbed. In order to fulfill the requirement of Figure 1, the edge width of seed separator plate was 120 mm. Traditional wide belt sowing device was shown in Figure 2a, whose seed separator plate was flat surface, and the falling wheat seeds showed the situation of “more in the center and less on both sides”, in order to improve the situation, this paper replaced flat surface to conical surface, as shown in Figure 2b.

Seed separated plate was taken from a part of conical surface whose mathematical model was shown in Figure 3. Points A and B were 2 points on the lower edge of side plate for seed protection (when model parameters were determined, the bottom edge width of cone was adjusted to 120 mm according to conical surface and installed on wide belt sowing device), point D was top point of seed separated plate, point C is midpoint of line AB, points A and B were equidistant from point O, given that A (x _A, 0, 0), B (x _B, y _B, 0), C (x _C, y _C, 0) and D (0, 0, z _D). The relationship between points A and B was given by Equation 1.

x B 2 + y B 2 = x A x A = R (1)

The spatial coordinate relationships between points A, C, and D and angles θ and α, fulfill Equation 2.

tan ⁡ α = x c 2 + y c 2 z D cos ⁡ θ = x c 2 + y c 2 x A (2)

Where x _C= (x _A–x _B)/2, mm, y _C = y _B/2, mm.

From Equations 1, 2, when the width of wide belt sowing device was defined, the structure of seed separator plate on wide belt sowing device was determined by the cone radius R (mm), cone angle θ (°) and installation angle α (°).

2.3 Discrete element simulation modeling

As an advanced and reliable discrete element simulation software, EDEM 2020 software (DEM Solutions Ltd., Edinburgh, United Kingdom) had been widely used in agricultural engineering simulation as a bulk kinematic analysis tool to reduce the cost of experiments and decrease the time of development cycle (Wang et al., 2026; Dun et al., 2025). In this paper, the software was used to simulate working process of wide belt sowing device, and the parameters of seed separator plate were optimized by analyzing the distribution of wheat on seedbed. Before analyzing the sowing uniformity, it was necessary to build a sowing model, which included wheat seed model, wide belt sowing model, material and contact parameters.

2.3.1 Wheat seed model

Due to the ellipsoid-like shape of wheat, a large number of studies had been carried out on the use of ellipsoid shape for discrete elemental modeling of wheat (Zhang et al., 2026). The ‘Ningmai 13’ was an improved variety selected and bred by Jiangsu Academy of Agricultural Sciences, which was characterized by high spike rate, fast polarization, high yield, etc. It was widely planted in Jiangsu, Anhui, Hubei and Henan Provinces in China (Huo et al., 2020). In this paper, this variety of wheat seed was selected for three-dimensional modeling. One hundred wheat seeds were randomly selected. Length, width and thickness of wheat seeds were measured several times using vernier caliper, which resulted in average values of 6.33, 3.15 and 3.12 mm, respectively. The width and thickness were approximate, which conformed to the conditions of simplified ellipsoid shape. The one hundred seeds were weighed individually using micrometer electronic scale, the volume of the wheat was measured by pycnometer method (Yan et al., 2022), the wheat density was calculated by dividing the seed weight by the volume. The five spheres method was used to construct the wheat seed model (Petingco et al., 2022; Wang et al., 2026), as shown in Figure 4.

2.3.2 Wide belt sowing model

The wide belt sowing model included wide belt sowing device and seedbed, the wide belt sowing device model included seed drop tube, cover plate for seed protection, side plate for seed protection, seed separator plate. The seed separator plate installed on wide belt sowing device according to the model of Figure 3. The geometric model of wide belt sowing device was created using Creo 10.0 software (Parametric Technology Corporation, Needham, MA, United States), saved as STP format and imported into EDEM 2020 software. Then the particle factory was added for wheat seeds generation, which was placed above seed drop tube and adjusted to virtual state (Figure 5). A seedbed with 300 mm width and 1,400 mm length was modeled with a distance of 10 mm from the bottom edge of seed separator plate, and the symmetry plane of wide belt sowing device was made to coincide with the symmetry plane of seed separator plate.

2.3.3 Material and contact parameters

The EDEM software provided a variety of choices for the contact model of particles: Hertz-Mindlin, Hertz-Mindlin with Bonding, Hertz-Mindlin with Heat Conduction, etc. In this paper, the Hertz-Mindlin (no slip) model was chosen as the inter-material contact model, considering that the safe storage moisture of wheat seeds was less than 12% and the adhesion force could be neglected (Zhang et al., 2026). The material and contact parameters included contact between wheat seeds and wide belt sowing device, seedbed and wheat seeds. The wide belt sowing device was made of 2 mm thick Q235 steel plate and the seedbed was made of soil. Except for density, all other parameters were according to Lu et al. (2023), Wang Z. et al. (2024) and Wang et al. (2026), the material and contact parameters of wheat seeds, wide belt sowing device and seedbed were shown in Table 1.

TABLE 1. Material and contact parameters.

2.3.4 Parameters setting for simulation process

Wide belt sowing device is generally working at a speed from 4 to 6 km h^–1 (Zhao P. F. et al., 2024). In this paper, the speed of wide belt sowing device was set to 4.32 km h^–1. In order to reduce simulation computation time, the wide belt sowing device remained stationary, the seedbed moved backward relative to the wide belt sowing device at a speed of −4.32 km h^–1. According to Luo et al. (2023), the 185 kg hm^–2 was adopted as the amount of wheat wide belt sowing, so the particle factory was set to continuously generate wheat seeds at a rate of 6 g s^–1 and fall freely under its own weight.

2.4 Evaluation method for lateral variation coefficient

Wheat seeds were dropped by gravity onto the conical seed separator plate and then randomly sown on the soil. The quality of 120 mm wide belt sowing was quantified by lateral variation coefficient (LVC). The wheat seeds were uniform in forward direction because the wide belt sowing device traveled at a uniform speed of 1.2 m s^–1, and only the change in position of wheat seeds in lateral direction needed to be considered. After simulation, a 1,200 mm long seedbed was randomly selected and divided into 6 equal segments. The 120 mm wide seedbed was then divided into 6 equal segments, resulting in a total of 36 seed drop areas, as shown in Figure 6. The number of wheat seeds in each area was counted, using x _ij to represent the number of wheat seeds in ith row and jth column area.

The average number of wheat seeds in all areas in row i was calculated by Equation 3,

x i ¯ = 1 6 ∑ j = 1 6 x i j (3)

The average number of wheat seeds in all areas was calculated by Equation 4,

x ¯ = 1 6 ∑ i = 1 6 x i ¯ (4)

The standard deviation of wheat seeds in the area was calculated by Equation 5,

s = 1 5 ∑ i = 1 6 x i − x ¯ 2 (5)

The LVC was calculated by Equation 6, which was used to evaluate uniformity of wide belt sowing,

LVC = s x ¯ × 100 % (6)

2.5 Single factor simulation experiment

According to the above analysis, factors cone radius, cone angle and installation angle significantly influence LVC. Therefore, the influencing factors identified for single factor experiment were cone radius, cone angle and installation angle. According to a large number of pre-experiments, the simulation parameters of cone radius were selected in this paper as 60, 65, 70, 75, 80 and 85 mm, and both cone angle and installation angle were selected as 30°. The simulation parameters of cone angles were selected as 20, 30, 40, 50, 60° and 70°, and the cone radius and installation angle were selected as 60 mm and 30°, respectively. The simulation parameters of installation angle were selected as 25, 30, 35, 40, 45° and 50°, and the cone radius and cone angle were selected as 60 mm and 30°, respectively. Each experiment was repeated three times, totaling 54 experiments.

2.6 Orthogonal simulation experiment

After the range of three factors was determined by single factor simulation experiment, in order to further determine optimal cone radius, cone angle and installation angle. Orthogonal experiment of three factors and five levels was carried out using cone radius, cone angle and installation angle as experimental factors and LVC as index. Table 2 showed the experimental factors and codes.

TABLE 2. Experimental factors and codes.

According to Central Composite Design, the experimental program and results were shown in Table 3. The standard deviation of the nine repeated center points was 0.757, where a smaller value indicates better model precision. The data were analyzed by quadratic polynomial regression by Design Expert 13.0 software (Stat-Ease, Inc., Minneapolis, MN, United States). The correlations and interactions of factors were analyzed by response surface analysis by Origin 2024 software (OriginLab Corporation, Northampton, MA, United States).

TABLE 3. Experimental design and result.

2.7 Indoor validation experiments

In order to verify the design rationality of seed separator plate and reliability of simulation, both indoor and field verification experiments were conducted. The experiment was conducted on a homemade wide belt sowing device (Figure 7). The conveyor belt was 9 m long and 0.7 m wide, with a speed range of 1.5–12 km h^–1. To minimize seed bounce, the conveyor belt was covered with white sand. The sand discharge device could be continuously adjusted within a speed range of 0–60 r min^–1. The conveyor belt speed was set to 4.32 km h^–1, the height of seed separator plate from the conveyor belt was 10 mm. To minimize seed bounce, the speed of sand discharge device was set to 60 r min^–1, and the wheat sowing rate was set to 6 g s^–1. Three independent sowing experiments were conducted with seed reloading and equipment restarting. In each experiment, two samples were randomly selected for each replicate on the conveyor belt of distances of 2, 4 and 6 m from wide belt sowing device. Each sampling in a 1.2 m long section. The schematic diagram of indoor experimental bench was shown in Figure 8.

2.8 Field validation experiments

The experiment was conducted in a field near Babaiqiao Town (118° 59′ E, 31° 38′ N, 22.7 m above sea level), Nanjing, Jiangsu Province, in mid-November 2023 after rice harvest. The area was located in the lower reaches of Yangtze River basin in eastern China. The wheat cultivar used in experiment was ‘Ningmai 13’. Soil bulk density, water content and penetration resistance in 0–20 cm soil depth were shown in Table 4.

TABLE 4. Physical properties in 0–20 cm soil depth.

To ensure uniform sowing of wheat in forward direction and minimize tractor’s disturbance to soil, the self-made field in situ multifunctional test bench was constructed for this study (Figure 9), which mainly included control cabinet, traction motor, movable carriage, lifting leg motor, etc. A 17 kW electric generator was used for power. The traction motor, lifting leg motor and movable carriage could be controlled using touchscreen on control cabinet and wireless remote control. The test bench spanned 6.9 m in length and 1.2 m in width with a speed range of 0–1.5 m s^–1. The movable carriage connected to one chain, two load sensors were equipped between the chain and movable carriage (Figure 10), which could be equipped with various implements to achieve multiple functions such as rotary tilling, trenching, sowing and rolling. The rotary tiller and wide belt seeder were attached to the movable carriage for this experiment. The forward speed of movable carriage was 4.32 km h^–1, the height of seed separator plate from ground was 10 mm, the furrow width and depth for sowing were 12 cm and 4 cm, respectively, the spacing of wide belt sowing was 15 cm and the sowing rate was 6 g s^–1. In order to effectively compare the results of field experiments with those of simulation and indoor experiments, the rice stubble on soil surface was cut close to ground before field experiments.

To validate the uniformity of conical seed separator plate on seed separation under various conditions, two tillage methods (without rice stubble) were employed: no-till wide belt sowing and rotary tillage wide belt sowing. The rotary tiller speed was 280 rpm (Zhao P. F. et al., 2024). The LVC and emergence rate were monitored. Each tillage treatment was replicated three times. After experiment, the soil around the seeds was carefully cleared away and samples were taken. A 1.2 m long sowing sample was taken from each tillage treatment. After sampling, the sown area was covered with soil and compacted. After the wheat was sown, the emergence rate was then observed as an extended agronomic validation.

For emergence rate, sampling began when more than half of the seedling had developed its first true leaves. The sampling length of seedling was 90 cm (Figure 11). Seedlings were considered to have emerged when they were more than 2 cm above ground surface. Three random samples were taken from each area. The emergence rate was calculated by Equation 7. To improve experimental comparability, no-till wide belt sowing and rotary tillage wide belt sowing were conducted at non-standard sowing depths of 2.5 cm and 6 cm, with three replicates for each treatment, and compared with the sowing depth of 4 cm.

Emergence rate = n / N (7)

Where n is the number of emergences, N is the number of seeds.

3 Results

3.1 Result of single factor simulation experiment

Figure 12a showed that as cone radius increased, the LVC first decreased and then increased. When the cone radius was 75 mm, the LVC reached a minimum of 14.78%, achieving optimal seed distribution. When the cone radius was greater than or less than 75 mm, the LVC increased. Moreover, the LVC showed significant differences at different cone radius. Figure 12b showed as cone angle increased, the LVC first decreased and then increased, reaching a minimum of 21.73% when the cone angle was 50°. Moreover, the LVC showed significant differences at different cone angle. Figure 12c showed as the installation angle gradually increased, the LVC first decreased and then increased, reaching a minimum of 13.76% when the installation angle was 35°. Moreover, the LVC showed significant differences at different installation angle.

3.2 Result of orthogonal simulation experiment

Single factor experimental results indicated that cone angle, cone radius and installation angle significantly influence LVC. To further determine the optimal cone angle, cone radius and installation angle, an orthogonal experiment of three factors was conducted. The significance analysis of LVC regression model was shown in Table 5. Since the P value for partial regression coefficient Rθ was greater than 0.1, indicating that Rθ had no significant influence on LVC, the partial regression coefficient Rθ was excluded. This ensured the regression model remained significant and free from misfit. After excluding the partial regression coefficient Rθ, all other coefficients had a significant influence on LVC, and the primary and secondary order of the influence of each factor on LVC was installation angle, cone angle and cone radius.

TABLE 5. Variance analysis of regressions equation.

* and ** indicate significant at 0.05 and 0.01 probability level, respectively.

As shown in Table 5, the model’s P value was less than 0.0001, indicating a highly significant relationship between the dependent variable and all independent variables in the regression model. The P value of the lack of fit was 0.8003, which was greater than 0.1, indicating that the lack of fit was not significant. Thus, the regression model fitted the simulation results well. The regression equation for each factor with respect to LVC was shown in Equation 8, the R ², adjusted R ² and predicted R ² values for the regression model were 0.9856, 0.9773 and 0.9653, respectively.

LVC = 2216.08313 – 34.38546 R – 12.97853 θ – 34.83131 α + 0.189505 R α + 0.035355 θ α + 0.186473 R 2 + 0.118618 θ 2 + 0.276473 α 2 (8)

3.3 Analysis of interaction effects of factors

According to the LVC mathematical model in Equation 8, the response surfaces for the interaction effects between any two of the three factors, as shown in Figure 13, which showed the interaction between cone angle and cone radius, cone radius and installation angle, and installation angle and cone angle all had a minimum point in the study range. As the values of two parameters increased, the LVC first decreased and then increased, resulting in a minimum value. To obtain optimal structural parameters of the seed separator plate, an optimization analysis was conducted on Equation 8. The optimal parameters for the three factors were cone radius of 74.82 mm, cone angle of 49.61° and installation angle of 34.18°, resulting in a regression predicted LVC of 12.36%.

3.4 Simulation experiment verification

To verify the optimized parameters, three simulation experiments were conducted using the optimal structural parameters, obtaining an average simulation validation LVC of 12.28% ± 0.68%, which differed by 0.08% from regression predicted LVC. Simulation experiments were also conducted on the traditional seed separator plate, obtaining an average LVC of 85.05%. The LVC of conical seed separator plate was improved by 85.56% compared to traditional seed separator plate. The simulation results were nearly consistent with the regression predicted value, demonstrating accuracy of the regression model. The sowing results before and after optimization of the seed separator plate were shown in Figure 14.

3.5 Results of indoor experiment

The sampling area for indoor experiment was shown in Figure 15. According to Equation 6, the average LVC for indoor experiment result was 14.73% ± 0.84%, while the regression predicted LVC was 12.36%. Due to the fact that wheat seeds were modeled as ellipsoids during simulation process, whereas actual wheat seeds were not standard ellipsoids, the accuracy of discrete element model using a five sphere ellipsoidal wheat seed structure was relatively low. These factors account for slightly lower LVC in simulation compared to indoor experiment. However, the deviation between indoor experiment and simulation was small, the simulation results could be considered essentially consistent with indoor experiment results.

3.6 Results of field experiment

The sampling area for field experiment was shown in Figure 16. According to Equation 6, the average LVC for no-till wide belt sowing and rotary tillage wide belt sowing were 16.28% ± 1.12% and 16.89% ± 1.56%, respectively, while the LVC for indoor experiment result was 14.73%. The results of three experiments showed that rotary tillage produced the poorest results, followed by no-till, with the indoor achieving the best results. This demonstrated that a flat seedbed had a significant influence on LVC. Due to the uneven surface of field soil and the high vibration during test bench experiments, the LVC in field experiment was higher than that in indoor experiment, however, the regression predicted LVC was 12.36%, with a difference of less than 5%. This demonstrated the rationality of seed separator plate’s structural design, and the use of a conical surface proved effective in achieving uniform lateral seed distribution.

Figure 17 showed no-till wide belt sowing and rotary tillage wide belt sowing at sowing depths of 2.5, 4 and 6 cm. For all tillage treatments, the emergence rate was significantly highest at a sowing depth of 4 cm compared to 2.5 cm and 6 cm. For the same sowing depth, the emergence rate was significantly higher with rotary tillage than with no-till. For sowing depths of 2.5 cm, 4 cm and 6 cm, the average emergence rate for rotary tillage was 9%, 8% and 20% higher than that for no-till, respectively. This indicated that rotary tillage could promote seed emergence.

4 Discussion

Sowing uniformity has a significant impact on crop growth and emergence (Pereyra et al., 2022; Pareek et al., 2023). Usually, sowing uniformity is measured by the variation coefficient, which is a key parameter for evaluating the design and optimization of sowing device and control system. Albasheer et al. (2024) designed a cylindrical air-assisted drill sowing device for rapeseed, wheat and rice, and validated it through bench tests, the Discrete Element Method (DEM), and field experiments. The results showed that the sowing uniformity coefficient of variation in field experiment was 3.44%, with emergence rate at 88%. Pareek et al. (2021) used the integrated artificial neural network and particle swarm optimization approach to optimize the seed-cell filling performance of an inclined plate seed metering device, aiming to achieve uniform seed distribution, the results showed that the maximum deviation between the predicted values and the experimental results was 2%. Chen et al. (2023) designed a dynamic compensation system for corn seed dropping position to improve the accuracy of plant spacing and row spacing during corn sowing, the results showed that the LVC of plant spacing decreased by an average of 9.16%. Wu et al. (2026) designed a centrifugal drone rice direct sowing device to improve lateral distribution uniformity, the results showed that minimum LVC in bench tests was 16.34% and average LVC was 19.36%. Gierz et al. (2023) designed a deflector to be installed on the distributor head of pneumatic seed drill. The results showed that when the air velocity was 14.4 m/s, the minimum LVC was 5%. According to the above findings, combined with the results of this study, it can be observed that sowing uniformity varies. There are many factors that affect sowing uniformity, such as sowing equipment, sowing methods, seed-cell filling performance of sowing device, soil conditions and seed characteristics, therefore, it is necessary to analyze these factors under specific conditions.

In addition, sowing depth also has a significant impact on wheat growth and emergence (Chen et al., 2025). If there is insufficient moisture in the topsoil, shallow sowing may result in poor emergence. Deeper sowing could improve crop nutrition, stability and yield. However, sowing too deeply can have adverse effects on crop growth, leading to reduced yields. Therefore, the appropriate plowing depth is determined by soil moisture, optimal soil temperature, soil aeration and the atmospheric humidity required for seed emergence and seedling growth (Masilamani et al., 2023). Sun et al. (2023) developed a prediction model to predict corn emergence at different soil depths, the results showed that appropriate sowing depth, soil moisture, bulk density, soil texture and air temperature can improve corn emergence rates. Tar et al. (2024) found that a sowing depth of 3 cm resulted in higher spring yields, with shallow sowing being most suitable for springs with high rainfall, while deep sowing was more suitable for dry summers. This study also found that an appropriate sowing depth promotes wheat emergence.

This study optimized the design of seed separator plate using LVM as evaluation criteria, however, it has some limitations, such as:

• This study showed that a conical seed separator plate achieved effective sowing at a speed of 1.2 m/s and a width of 12 cm, however, the sowing dynamics of the conical seed separator plate at different widths and speeds were not further investigated.
• In order to make the field soil similar to the simulated soil without the influence of rice stubble, the experiments were conducted on fields where rice stubble had been removed, however, no experiments were conducted in more complex environments where stubble remained.
• Due to the limitations in the research design and experimental conditions, this research focused on one wheat variety and soil condition. However, different wheat varieties and soil conditions will be investigated in future experiments.

5 Conclusion

To solve the problem of non-uniform sowing in wide belt, the structural design of seed separator plate and the parameter optimization of LVC were conducted through mathematical models, single factor experiment, orthogonal simulation experiment, indoor experiment and field experiment. The specific conclusions are as follows:

• A conical seed separator plate was proposed, and a mathematical model of the plate’s structure was established with variables cone radius, cone angle and installation angle.
• A discrete element simulation model was developed to evaluate wheat sowing uniformity using LVC, and single factor simulation experiment with three variables were conducted to determine the optimal cone radius ranged from 70 to 80 mm, cone angle ranged from 40° to 60° and installation angle ranged from 30° to 40°.
• Orthogonal simulation experiment showed that the primary and secondary order of the influence of each factor on LVC was the installation angle, cone angle and cone radius. The corresponding parameter values after optimization were 74.82 mm for cone radius, 49.61° for cone angle and 34.18° for installation angle, with a regression predicted LVC of 12.36%.
• To verify the results of structural parameter optimization of seed separator plate, simulations, indoor experiments and field experiments with no-till wide belt sowing and rotary tillage wide belt sowing were conducted. The results showed that average simulation validation LVC was 12.28%, improving by 85.56% compared to traditional seed separator plate (the conclusion was restricted to comparison obtained under simulated conditions). The average LVC in indoor experiment was 14.73%, the average LVC in no-till wide belt sowing field experiment was 16.28%, and the average LVC in rotary tillage wide belt sowing field experiment was 16.89%. The results from indoor and field experiments were generally consistent with those from the simulation validation experiments, which demonstrated reliability of structural optimization of seed separator plate. For emergence rate, the sowing depth of 4 cm was the highest. Rotary tillage promoted emergence than no-till.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmech.2026.1932439/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Stanislav N. Gorb, University of Kiel, Germany

Reviewed by: Chaitanya Madhaw Pareek, Woxsen University School of Business, India

Song Wang, Heilongjiang Bayi Agricultural University, China