多功能智能蔬菜播种机的设计与试验（Design and Experimental Study of a Multifunctional Intelligent Vegetable Seeder）

• Article

• 

26 October 2026

24 Pages

,

,

,

,

,

and

1

School of Mechatronic Equipment, Qingdao University of Technology, Linyi 276038, China

2

School of Mechanical and Automotive Engineering, Qingdao University of Technology, Qingdao 266520, China

3

College of Mechanical and Electrical Engineering, Tarim University, Alar 843300, China

4

School of Mechanical Engineering, Ningxia University, Yinchuan 750021, China

Abstract

Vegetable seeders effectively reduce labor intensity and improve seeding uniformity and operation efficiency. However, existing models suffer from limited functionality, inadequate intelligence, and low energy efficiency. Accordingly, an intelligent vegetable seeder integrating furrow opening, seeding, soil covering, and compaction was developed, with key parameters and working principles of the seed-metering device, furrow opener, and soil-covering and compacting device defined. Its intelligent control system optimized seeding monitoring and execution control via the Kalman filter algorithm. A dual power-supply mode (photovoltaic (PV) + storage battery) ensured operation endurance. Using a Box–Behnken design, the seed-metering device parameters were optimized by response surface methodology, with pocket wheel speed, seed-stripping brush installation angle, and pocket count as factors and seeding qualified rate as the index. The results showed that the factors ranked in descending order of influence were the number of pockets, rotational speed of the pocket wheel, and the installation angle of the seed-stripping brush; the theoretically optimal parameters were pocket wheel speed of 45.70 r·min−1, brush installation angle of 34.71°, and 13.84 pockets, corresponding to a model-predicted peak seeding qualified rate of 99.04%. The proposed seeder achieves precision seeding for small-sized pelleted vegetable seeds, whereas its structural configuration requires further optimization for non-pelleted vegetable seeds.

1. Introduction

Driven by the rapid advancement of global agricultural science and technology and the continuing progress of agricultural modernization, agricultural production is accelerating its transformation from the traditional model dominated by manual and animal labor toward full-process mechanization and intelligence. Research data show that the cultivation area of China’s protected agriculture accounts for more than 80% of the global total; the proportion of vegetable planting area has also exceeded 80%, and the annual vegetable output has surpassed 700 million tons, far higher than that of other countries worldwide [1]. However, the vegetable production sector has long faced prominent bottlenecks such as the rigid rise in labor costs, structural labor shortages, and a low level of precision operations, which constrain the industry’s quality improvement, efficiency increase, and green sustainable development [2]. As the core equipment for the full-process mechanization of vegetable production, the vegetable seeder can efficiently replace manual labor to complete operations such as furrow opening, seed-metering, and soil covering, significantly improving the operating efficiency and seeding accuracy, reducing the production cost and agricultural-input loss per unit area, and holding important application value in optimizing resource utilization, reducing environmental load, and alleviating labor shortage pressure.

Numerous researchers have conducted extensive studies on vegetable seeders. In terms of functional integration, the integrated potato planter developed by Li et al. integrates multiple functions including seeding, furrow opening, mulching, soil covering, and fertilization into one unit [3]. In the field of intelligent seeding, Chen et al. adopted electronic control and pneumatic transmission technology and employed a timing control method to achieve efficient and accurate automatic seeding [4]; however, their seeding accuracy dropped significantly when facing complex environments and terrains. In terms of seed-metering device research, the agricultural machinery team at Akdeniz University in Turkey developed an integrated cylindrical drum device combining a hopper and a seed-metering unit, which integrates seed storage and metering functions within a single rotating drum. However, the seed-metering device serves only the single function of seed discharging and is not equipped with on-board monitoring for key operational quality indicators including seeding depth and seeding qualified rate [5]. For intelligent seeding monitoring, Abdallah Elshawadfy Elwakeel et al. developed an IoT-based precision seed-metering monitoring system. The system uses an Arduino Mega as the main controller and employs infrared sensors to count seeds passing through the metering outlet in real time. Nevertheless, the real-time responsiveness and anti-interference performance of the Arduino Mega board remain limited under harsh field conditions including mechanical vibration and heavy dust [6]. Overall, vegetable seeders are developing toward functional integration, intelligence, and new-energy utilization while ensuring seeding accuracy and efficiency, thereby achieving high efficiency of intelligent agricultural equipment [7].

Building on the foregoing technical basis, an intelligent vegetable seeder was developed in the present study. It features a multifunctional integrated design that incorporates furrow opening, seeding, soil covering, compaction, and seeding monitoring into a single unit. Such a design significantly improves the seeding qualified rate of the vegetable seeder and provides a reference for the development of intelligent agricultural machinery and intelligent agriculture.

2. Materials and Methods

2.1. Structural Design

2.1.1. Overall Machine Structure Design

The seeder as a whole consists of a machine frame, a DC drive motor (Datai New Energy Co., Ltd., Taizhou, Zhejiang, China), a storage battery, a transmission chain, rollers, a pocket-wheel seed-metering device, a furrow opener, a soil-covering device, a dust-proof cover, and a PV panel (JinkoSolar Co., Ltd., Shanghai, China) (Figure 1). These components are integrated into four functional modules: the power system, intelligent control system, furrow-opening and soil-covering system, and seeding system. A longitudinal assembly layout is adopted for the machine, with rollers mounted at both the front and rear ends to improve operational stability when the seeder operates on sloping terrain. Meanwhile, an STM32 microcontroller (STMicroelectronics, Geneva, Switzerland) is embedded as the control core to realize intelligent control, supporting the coordinated operation of multiple functions such as furrow opening, seeding, soil covering, and compaction. The main design parameters of the seeder are summarized in Table 1.

Figure 1. Structure diagram of the intelligent seeder. 1. PV panel; 2. Pocket wheel; 3. Shell; 4. DC drive motor; 5. Roller; 6. Pinion; 7. Gear; 8. Chain; 9. Furrow opener; 10. Soil-covering device; 11. Baffle; 12. Frame; 13. Storage battery.

Table 1. Main parameters of the multifunctional intelligent vegetable seeder.

During operation of the vegetable seeder, a start command was issued from the remote terminal. The rotational speed of the DC drive motor was regulated according to the preset travel speed of the whole machine. Power was transmitted to the rollers via the chain-drive mechanism to propel the machine forward. Subsequently, the furrow opener completed the seed-furrow cutting operation. Under the control of the intelligent seeding system, the seeding device could automatically adjust operating parameters such as seeding density and seeding depth and simultaneously monitor the seeding qualified rate in real time. The control parameters were dynamically corrected based on the collected operational data and real-time calculation results. Finally, the soil-covering and compaction device performed soil covering and compaction to complete the seeding operation (Figure 2).

Figure 2. Workflow diagram of the intelligent seeder.

2.1.2. Design of the Pocket Wheel Seed-Metering Device

The seed-metering device is the core component of the vegetable seeder, whose performance of the seed-metering device directly determines seeding accuracy and operational quality [8]. A pocket-wheel seed-metering device (shown in Figure 3) was adopted in the present design for the merits of simple structure, low cost, and convenient maintenance. The device was primarily developed for small-sized pelleted vegetable seeds with a particle size of 2.0–2.5 mm [9]. For non-pelleted natural seeds with irregular shapes, such a configuration is not applicable due to distinct seed-filling behaviors inside the pocket wheel. The device mainly consists of a seed-metering device cover, a seed box, a pocket wheel, a seed-stripping brush, and a seed guard plate. During operation, the rollers or the DC drive motor drives the pocket wheel to rotate through the transmission system. Seeds in the seed box fall into the pockets under gravity and mutual squeezing among seeds; as the pocket wheel rotates into the seed-stripping zone, the seed-stripping brush scrapes off the excess seeds outside the pockets, ensuring single-seed or fixed-amount seed-filling. Subsequently, the seeds enter the seed-guarding zone with the pocket wheel, and the seed guard plate prevents the seeds from dropping prematurely. When the pocket wheel rotates to the seed-dropping port, the seeds fall into the seed furrow opened by the furrow opener under the action of gravity and centrifugal force, completing the seeding process.

Figure 3. Structure of the pocket wheel seed-metering device. (a) Physical prototype of the pocket wheel seed-metering device; (b) Three-dimensional model of the pocket wheel seed-metering device. 1. Seed-metering device cover; 2. Seed box; 3. Pocket wheel; 4. Seed-dropping port; 5. Seed guard plate; 6. Seed-stripping brush.

(1)
Diameter of pocket wheel

The diameter of the pocket wheel directly affected the size of the seed-metering device, the number of pockets, and the seed-filling performance. An excessively large diameter increased the overall machine volume and moment of inertia, whereas an excessively small diameter limited the number of pockets and provided insufficient seed-filling time. The diameter of the pocket wheel is generally selected within 80–200 mm [10]; considering the agronomic requirements of vegetable seeding, the seed-metering device structure, the number of pockets, and the minimization of power consumption, the pocket wheel diameter was designed to be 80 mm.

(2)
Number of pockets

The number of pockets should be determined according to the operating speed and the plant spacing requirement. Let the forward speed of the vegetable seeder be v (m·s−1), the rotational speed of the pocket wheel be n (r·min−1), and the plant spacing be s (m); then the number of pockets satisfies the following:

(1)

(3)
Pocket size

The seed-metering performance of the pocket wheel seed-metering device mainly depends on the single-seed rate of the pocket (i.e., no missed seeding and no multiple seeding), and the key lies in the compatibility between the pocket and the seed [11]. To facilitate machining and ensure seed-filling, the pocket was designed as a cylindrical hole with a hemispherical bottom and a 45° chamfer at the hole opening, and the dimensions of the pocket were determined according to the seed size. Taking typical small-sized vegetable seeds such as lettuce, the average seed size after pelleting was approximately 2.0–2.5 mm. The pocket diameter d h was generally taken as 1.1–1.3 times the seed size, and the depth h as 0.7–0.9 times the seed size. Taking the average seed size d s = 2.2 mm [12], then

(2)

(4)
Pocket-wheel transmission and plant-spacing adjustment

A hybrid transmission configuration combining a roller and a motor is adopted in the current design. During normal operation, the rollers drove the seed-metering shaft to rotate via a chain drive; the plant spacing was adjusted by replacing sprockets with different numbers of teeth or by adopting an independently driven stepper motor. When a stepper motor was used, the motor speed was controlled by the microcontroller to precisely match the forward speed with the rotational speed of the pocket wheel, avoiding missed seeding caused by roller slippage. Let the roller diameter be D w = 250 mm; the forward distance per revolution of the roller is π D w = 0.785 m. If the plant spacing is s = 0.1 m, then the transmission ratio between the roller and the pocket wheel sprockets is

(3)

2.1.3. Design of the Furrow-Opening Device

The core-type share furrow opener was selected, which is a typical structure of the acute-angle furrow opener. During operation, the cutting edge at the front end of the furrow opener cut into the soil layer to a preset depth; the cut soil mass was separated along the share surface and gradually lifted, and after rising to the edge of the share wing, it settled and fell back under gravity. Through the complete process of soil cutting, separation, lifting, and settling, a seed furrow structure with stable furrow depth and width was finally formed (Figure 4).

Figure 4. Structure diagram of the furrow opener. (a) Side view of the furrow opener; (b) Bottom view of the furrow opener. 1. Mounting frame; 2. Soil-retaining plate; 3. Furrow opener body; 4. Ridge line of the furrow opener.

The main performance parameters of the furrow opener are the curvature radius of the ridge line, the penetration angle of the furrow opener

, the penetration clearance angle

, the inclined-plane angle β, and the furrow opener width B [13]. Among these, the curvature radius of the ridge line and the penetration angle are the key structural parameters, which directly determine the spatial layout position of the furrow opener, the structural strength, the penetration performance, and the overall furrow-opening resistance of the machine.

A symmetrical furrow opener body is adopted in the present design, with the two inclined moldboard surfaces intersecting to form a ridge-line structure. To reduce furrow-opening resistance to the greatest extent, a circular-arc ridge profile is selected for the opener configuration [14].

A two-dimensional coordinate system for the ridge line is established with the lowest point O of the ridge line as the origin. Two points A(x 1,y 1) and B(x 2,y 2) on the circular-arc curve were selected, where point A is the start point and point B is the end point of the circular-arc curve. O 1 denotes the center of the circular arc, and curve AB serves as the uniformly varying circular-arc ridge line of the furrow opener, as shown in Figure 5a. The equation of ridge line AB of the furrow opener is expressed as follows:

(4)

where R—radius of the arc, mm.

Figure 5. Schematic diagram of the furrow opener. (a) Schematic of the furrow opener ridge line; (b) Force analysis diagram at point A.

From the geometric relationship shown in the figure, the following relationship was obtained:

(5)

where

—height of the ridge-line AB section, mm;

—opening of the curved section of ridge line AB, mm;

—slip cutting angle at point A, (°);

—slip cutting angle at point B, (°);

—length of segment OO 1, mm.

Combining Equations (4) and (5), the curve radius R of the circular-arc ridge line AB was obtained as follows:

(6)

Then the curve equation of the arc ridge line AB was expressed as follows:

(7)

According to the curve equation of the circular-arc ridge line AB, the ridge line height H, the initial slip cutting angle

, and the final slip cutting angle

all affect the shape of the furrow opener ridge line. To enable the furrow opener to better cut the soil and ensure timely backfilling, the ridge line height H is determined to be 60 mm.

The penetration angle

is the core geometric parameter characterizing the included angle between the working surface of the furrow opener and the ground; the value of the angle not only directly determines the penetration capability of the furrow opener but also significantly affects the structural strength of the curved surface of the front furrow opener body. Point A on the ridge line was taken as the object of the force analysis, and the corresponding force analysis is presented in Figure 5b.

A coordinate system was established with point A as the origin, and according to the static equilibrium, the following relationship was obtained:

(8)

(9)

where

—draft force of the seeder on the furrow opener, N;

—friction force between the soil and the cutting edge surface of the furrow opener, N;

—tangential resistance exerted by the soil on the cutting edge surface of the furrow opener, N;

—normal pressure exerted by the soil on the cutting edge surface of the furrow opener, N;

—gravity of the soil, N;

—friction coefficient;

—penetration angle of the furrow opener, °.

Combining Equations (8) and (9) yields the following:

(10)

To ensure the structural mechanical strength of the furrow opener body surface, and by comprehensively balancing the operational performance and structural reliability, the design value of the penetration angle

was finally determined to be 45° [15].

2.1.4. Design of the Soil-Covering and Compacting Device

The soil-covering device is located behind the furrow opener and was used to backfill the soil turned up during furrow opening into the seed furrow. A double-disc soil coverer is adopted in the current design, with the two discs symmetrically positioned in a herringbone pattern to permit free rotation and reduce soil blockage [16].

The soil-covering disc diameter D is calculated from the following empirical formula based on the soil-covering depth requirement:

(11)

In the formula,

—maximum design soil-covering depth (mm);

—diameter-to-depth ratio coefficient.

The seeding depth of vegetable seeds ranges from 10 to 40 mm, so

was set to 40 mm; the coefficient

of the shallow-tillage stubble disc was taken as 5, from which the disc diameter was obtained as 200 mm.

The thickness of the soil-covering disc δ was selected according to the magnitude of the working load or calculated using the following empirical formula:

(12)

The commonly used range of δ is 3.5–6 mm, and a value of 3.5 mm is specified for δ in the present design. The included angle between the two discs is adjustable within a range of 60–90°; a smaller included angle corresponds to a larger soil covering volume. An included angle of 75° is selected for the configuration, which meets the requirements of shallow seeding and soil covering (Figure 6).

Figure 6. Soil-covering device. 1. Soil-covering bracket; 2. Soil-covering disc; 3. Soil-covering boot.

The press wheel is located behind the soil-covering device and serves to compact the seedbed after soil covering, promoting close contact between seeds and soil and preserving soil moisture. The press wheel is a hollow cast-iron wheel with an outer diameter of 150 mm and a width of 80 mm, and the wheel surface is fitted with a rubber cushion to avoid excessive compaction. The compaction force is steplessly adjustable from 0 to 200 N via a spring and an adjusting screw to accommodate different soil moisture levels. Both ends of the press wheel axle are mounted on profiling rocker arms and can float up and down with terrain variations.

2.1.5. Design of the Power System

As the energy source and driving core of the vegetable seeder, the power system directly influences the machine’s operational performance, endurance, and energy consumption. A dual power supply scheme of PV panel and storage battery is adopted for the seeder [17], and power for traveling and operation is provided by a DC drive motor.

To adapt to diverse operating environments (e.g., sandy soil and clay soil), rollers are adopted as the traveling components of the seeder to deliver stable support, improve field traveling stability, and ensure consistent seed-row alignment. This configuration extends service life and reduces operational costs as well as maintenance frequency. Furthermore, to cope with uneven or sloping fields, the seeder is equipped with a 57HS22 stepper motor (Leisai Intelligent Control Co., Ltd., Shenzhen, China) for traction, paired with a 48 V/40 Ah lithium-iron-phosphate (LiFePO 4) battery pack (Sunwoda Electronic Co., Ltd., Suzhou, Jiangsu, China) to guarantee operational endurance.

2.2. Intelligent Seeding Control

The vegetable seeder collects seeding data in real time through ultrasonic and photoelectric sensors (Omron Corporation, Kyoto, Japan; model: E3Z‑T61) in the data acquisition layer. In the data processing layer, a Kalman-filter-based dual-channel ultrasonic data fusion algorithm is employed to fuse measured data from the two ultrasonic sensors. The processed data are subsequently transmitted to the user interaction layer through the data transmission layer.

Operators can monitor real-time operating conditions and analyze historical data via the remote terminal. Control commands for seeding depth and travel speed are output to drive the actuators to achieve precise seeding. Ultimately, a complete closed-loop system of “perception–feedback–decision–execution” is formed, enabling intelligent agricultural operation with a low miss-seeding rate and high energy efficiency (Figure 7).

Figure 7. Data acquisition and processing flowchart.

The control interface of the remote terminal is shown in Figure 8.

Figure 8. Remote terminal control interface.

Modern seeding control systems are generally fitted with a communication module that can exchange data with agricultural machinery monitoring platforms to realize real-time monitoring and remote control. An STM32 microcontroller is adopted for seeding control of the developed seeder. It can fully cover functional demands including sensor signal acquisition, motor drive control, and data interaction and transmission and is suitable for complex field operation environments.

For seeding-depth measurement, an ultrasonic sensor was selected for measurement because of its advantages of simple structure, convenient installation, and resistance to dust and interference, making it highly suitable for the field operating environment of the vegetable seeder. Both ultrasonic sensors adopted the US-100 ultrasonic ranging module (TELESKY, Shenzhen, China). During seeding operation, the two ultrasonic sensors synchronously acquire ranging data at a fixed sampling frequency of f = 10 Hz. To eliminate the potential interference of straw and weeds with the seeding operation, an ultrasonic ranging sensor was additionally installed on top of the compacting device; the actual seeding depth was characterized by precisely measuring the vertical distance variation between the ground surface and the sensor after compaction. Specifically, the ultrasonic sensor at the top of the press wheel performed a direct elevation measurement, and the measured distance variation directly corresponded to the seeding depth.

In contrast, the ultrasonic sensor above the depth-control wheel adopted an indirect measurement strategy; instead of directly measuring the distance to the ground, it indirectly reflected changes in seeding depth by monitoring the variation in distance between the sensor and a terrain-simulating plate (designed to simulate and follow changes in the ground contour). The specific principle is shown in Figure 9.

Figure 9. Schematic diagram of indirect seeding depth measurement.

Meanwhile, the seeding depth measurement program adopts a dual-channel ultrasonic data fusion algorithm based on the Kalman filter and fuses the measurement data from the dual-channel ultrasonic sensors through the data fusion algorithm to improve measurement accuracy and reliability.

Assume that the seeding-depth is taken as the system state variable k, i.e., the vertical distance from the ground surface after compaction to the bottom of the seed furrow. The observations of the two ultrasonic sensors are

and

. Here, k denotes the discrete time step corresponding to the k-th sampling instant; A is the state-transition matrix describing the recurrence relation of the seeding-depth state over time; B is the control-input matrix; u(k) is the control quantity of the seeding-depth adjustment mechanism at time step k; w(k) is the system process noise representing state disturbances caused by system-modeling errors, mechanical vibration, and abrupt terrain variation; Q is the process-noise covariance characterizing the magnitude of process-noise intensity. In such a scenario, the variation in seeding depth is generally modeled as a random-walk model, where Q denotes the variance of the random seeding-depth change per unit time. N(0, Q) represents a Gaussian distribution with a mean of 0 and a covariance of Q.

(13)

Let H 1 and H 2 be the observation coefficients of the two ultrasonic sensors, respectively.

and

denote the measurement noises of the two sensors, including random errors originating from sensor self-noise, field clutter, and straw occlusion. R 1 and R 2 are the measurement-noise covariances of the two sensors, reflecting their respective measurement accuracies. The observation equation is as follows:

(14)

The recursive procedure of the Kalman filter begins with the prediction step (time update), which yields the prior state estimate.

The Kalman filter performs the estimation recursively in two sequential stages: prediction and update.

In the prediction step (time update), the prior state estimate is first derived as follows:

(15)

where

is the prior state estimate at time k, and

is the posterior state estimate at time k − 1.

The corresponding prior error covariance matrix is then calculated by the following:

(16)

where

is the prior error covariance matrix at time k, and

is the posterior error covariance matrix at time k − 1.

In the update step (measurement update), the Kalman gain for fusing dual-channel sensor data is computed as follows:

(17)

H denotes the augmented observation matrix, which is formed by splicing the observation matrices of the two sensors row-wise to unify the dual-channel observations into a single observation model. R is the augmented observation-noise covariance matrix, which is a diagonal matrix with off-diagonal elements equal to zero, indicating that the measurement noises of the two sensors are mutually independent, where

,

.

Following the measurement update, the posterior state estimate is obtained as follows:

(18)

where

is the posterior state estimate at time k.

The posterior error covariance matrix is then updated by the following:

(19)

where

is the posterior error covariance matrix.

Let the i-th calibrated depth serve as the reference true value (in mm), and let the output of the depth-measurement system (i.e., the dual-channel fusion estimate) at the corresponding depth be

. A total of N measurement trials were conducted. The root-mean-square error (RMSE) is defined as follows:

(20)

To verify the seeding-depth measurement accuracy of the dual-channel ultrasonic system with Kalman filter data fusion, a calibration test was conducted on a soil-trough test bench. A reference seeding depth of 25 mm, within the designed seeding depth range of the prototype, was set as the calibration benchmark. The seeder operated continuously along a 5 m soil trough, and 50 monitoring points were selected at equal intervals along the seed furrow. The entire test was repeated 3 times, yielding a total of 150 sets of depth comparison data. Two sets of values were recorded simultaneously at each monitoring point: the real-time estimated seeding depth output by the Kalman filter fusion algorithm and the actual seeding depth measured manually after sowing as the ground truth. The root mean square error (RMSE) was calculated according to Equation (20), giving RMSE = 0.525 mm. Normalized against the preset reference seeding depth of 25 mm, the normalized relative root mean square error was 2.1%.

In terms of seeding monitoring, an optical convex lens is integrated into the conventional photoelectric sensor to form the seeding monitoring acquisition module, which ensures that the infrared light emitted by the transmitter fully covers the monitoring area. The operational principle of the monitoring module is illustrated in Figure 10.

Figure 10. Schematic diagram of the photoelectric sensor.

The infrared light emitted by the transmitter can completely cover the monitoring area; during the seed-metering process, the infrared radiation signal was blocked by seeds, causing the infrared receiver to fail to receive the predetermined light and thereby triggering the inversion of the level signal [18]. The actual plant spacing L is derived by precisely recording the time interval Δ t between two level-state transitions and simultaneously acquiring the travel speed v of the seeder measured by the encoder. Specifically, when v·Δ t is less than or equal to 0.5 times the target plant spacing d, the case was judged as a multiple-seeding event; conversely, if the value of v·Δ t exceeded 1.5 times the target plant spacing d, it is regarded as a missed-seeding event (Figure 11). To mitigate measurement interference from dust raised during the seeding operation, the infrared photoelectric sensor is sealed inside a protective box fixed at the end of the seed tube.

Figure 11. Flowchart of the seeding monitoring system. Dark blue: Start and end nodes of the whole workflow, and all steps in the system initialization stage (Step 01). Light blue: Decision diamond boxes and the final statistical output module. Cyan: All operation steps in the pulse detection stage (Step 02), including pulse counting, frequency recording and parameter comparison. Yellow: All threshold judgment links in the status check stage (Step 03). Green: The result output stage (Step 04) and the "Normal sowing" judgment result. Red: The "Missed seeding alarm" judgment result. Light orange: The “Multiple seeding alarm” judgment result.

2.3. Box–Behnken Design (BBD) Seeding Experiment

In the scenario of single-seed precision seeding, the seeding qualified rate serves as the core evaluation index for assessing the operational accuracy of the seed-metering device and seeding uniformity. Taking the theoretically designed plant spacing as the judgment criterion, this index was used to classify continuously measured adjacent seed-spacing samples. The seeding qualified rate was quantitatively characterized as the percentage of qualified seed-spacing samples among all effective seed-spacing samples [19]. To ensure the comparability of qualified rates across all groups, comparisons of qualified rates were performed using datasets of identical size after sowing the same amount of vegetable seeds for each test.

In the sowing experiment, loose-leaf lettuce seeds were selected and pelleted via a conventional clay-based coating and pelleting process. Standard screening tests showed that the particle size of the pelleted seeds ranged from 2.0 to 2.5 mm, with an average particle size of 2.2 mm, a 1000-seed weight of 3.22 g, a seed moisture content of 6.5–7.5%, wet-basis moisture content of 6.5–7.5%, and dry-basis moisture content of 6.95–8.11%. The seed particle size parameters well matched the designed dimensions of the seed-metering device pockets. The sowing experiment was carried out in the soil trough of the Key Laboratory of Modern Agricultural Machinery Equipment, Linyi Campus of Qingdao University of Technology University of. The experimental soil was typical brown soil collected from local farmland in Linyi, which was dried, crushed, sieved, and backfilled uniformly before the test. Prior to the experiment, the soil physical properties were adjusted to simulate conventional field conditions for vegetable sowing: the soil mass moisture content was 15–18%, the soil particle density was 2.64–2.66 g·cm−3, the soil bulk density was 1.25–1.35 g·cm−3, the soil porosity was 47.2–49.5%, and the soil firmness was 0.8–1.2 MPa (Figure 12).

Figure 12. Photograph of the vegetable seeder prototype.

A three-factor, three-level Box–Behnken design (BBD) seeding experiment was carried out. Based on the related literature, the rotational speed n of the pocket wheel X 1, the installation angle θ of the seed-stripping brush X 2 (the angle between the brush centerline and the vertical line), and the number z of pockets X 3 were selected as the experimental factors [20,21,22,23]. A total of 17 experimental runs were designed using Design-Expert (v13) software. Each test was repeated three times, and the average seeding qualified rate was taken as the experimental result under corresponding parameters. According to the structural design in Section 2.1, the level of the pocket wheel rotational speed was set to 25–55 r·min−1 and that of the number of pockets to 12–18. The seed-metering device achieved favorable working performance when the installation angle of the seed-stripping brush was generally taken as 30° [24]; therefore, its levels were set to 20–40°. The experimental factors and their level values are listed in Table 2.

Table 2. Experimental factors and levels.

3. Results

3.1. Experimental Results

The seeding experiment results were summarized in Table 3; the statistical analysis of the data was completed using Design-Expert software with a Box–Behnken response-surface experiment model. Under operating conditions with different combinations of factor levels, the seed-metering performance of the pocket wheel seed-metering device showed significant differences, with the seeding qualified rate ranging from a minimum of 92.41% to a maximum of 98.51%.

Table 3. Experiment design and results.

3.2. Regression Model and Significance Analysis of Variance

Based on the Box–Behnken response surface model, regression fitting was performed on each experimental factor and performance indicator, and the final regression model obtained is as follows:

The coefficient of determination R 2 of the regression model was 0.9857 within the preset range of experimental parameters. The difference between the predicted R 2 (0.7723) and adjusted R 2 (0.9672) was less than 0.2, demonstrating reasonable agreement between them. The ANOVA significance test of the model returned a p-value smaller than 0.0001, revealing that the established regression model was highly significant.

To further verify the fitting adequacy and statistical validity of the regression model, pure error estimation and lack-of-fit testing were conducted. The results showed that the lack-of-fit F-value was 153.03, with a corresponding p-value of 0.0014, which was statistically significant. Notably, although the lack-of-fit term reached statistical significance, this result was mainly attributed to the extremely small pure error under the well-controlled soil-trough test conditions. The sum of squares of pure error was only 0.0097, and the coefficient of variation of the seeding qualified rate at the central test point was as low as 0.05%. At such a low level of random experimental error, the lack-of-fit F-test becomes overly sensitive to minor deviations from the quadratic response surface. Accordingly, the observed statistical significance here does not indicate poor practical fitting performance of the model.

Residual diagnosis was performed to evaluate the model reliability. In the residuals-versus-predicted-values plot, the residuals were randomly and uniformly distributed on both sides of the zero baseline without funneling or directional drift, which verified that the model satisfied the homoscedasticity assumption with no heteroscedasticity. In the predicted-versus-actual-values plot, all data points closely concentrated around the 45° reference line and presented a high linear correlation, confirming excellent consistency between model predictions and measured values. In the residuals-versus-run-order plot, the residuals fluctuated randomly throughout the experimental procedure without periodic trends, indicating no time-dependent systematic errors during the tests (Figure 13).

Figure 13. Residual diagnostic plots. (a) Residuals vs. predicted; (b) Predicted vs. actual; (c) Actual residuals vs. run. The two upper and lower red solid lines are the default ±3σ residual control limits built into Design-Expert, which are used to identify outlier points, with the boundary values of y = ±4.81963.

Experimental repeatability was verified through five independent repetitive tests at the central point of the Box–Behnken design, corresponding to a pocket wheel rotational speed of 40 r·min−1, a seed-stripping brush installation angle of 30°, and 15 pocket holes. The measured seeding qualified rates for the five replicates were 98.44%, 98.51%, 98.39%, 98.41%, and 98.48%, respectively. Statistical analysis showed that the average seeding qualified rate was 98.45%, with a standard deviation of 0.049% and a coefficient of variation of only 0.05%. The value was considerably lower than the conventional acceptable threshold of 5% in agricultural engineering tests, demonstrating that the prototype operated stably and that the experimental results possessed high stability and excellent repeatability [25]. Meanwhile, it should be acknowledged that the predicted coefficient of determination (Predicted R 2 = 0.7723) of the model falls at the lower end of the acceptable range. The pure error was estimated exclusively from replicate runs at the central design point, which yields limited degrees of freedom. Furthermore, the model was calibrated specifically for pelleted lettuce seeds under the tested brown soil conditions; its generalizability to non-pelleted, irregularly shaped seeds and other soil operating conditions remains to be further validated.

As shown in Table 4, the rotational speed of the pocket wheel (X 1), the installation angle of the seed-stripping brush (X 2), and the number of pockets (X 3) were all significant influencing factors (p< 0.05). Because X 3,

and

were highly significant influencing factors (p< 0.0001), while

was only a significant influencing factor (p< 0.05), the influence of factors X 1 and X 3 on the seeding qualified rate was greater than that of X 2. In addition, X 3 reached a highly significant level while X 1 only reached a significant level, so the influence of factor X 3 on the seeding qualified rate was greater than that of X 1. In summary, the factors were ranked in descending order of influence: number of pockets, rotational speed of the pocket wheel, and installation angle of the seed-stripping brush.

Table 4. Analysis of variance (ANOVA).

3.3. Response Surface Analysis

The response surface in Figure 14 illustrates the interactive influence of pocket wheel rotational speed and seed-stripping brush installation angle on seeding qualified rate at a fixed pocket count of 15. The surface takes on an overall downward-opening convex paraboloid form with a single distinct peak zone, which verifies that both factors have optimal value ranges for the seeding qualified rate. When the installation angle of the seed-stripping brush was fixed, the rotational speed of the pocket wheel affected the seed-filling process mainly by regulating the motion intensity of the seed population in the seed-filling zone. At low rotational speeds, the tumbling and collision motion of the seed population was weak, the kinetic energy of seeds entering the pockets was insufficient, and the pocket filling was inadequate. Therefore, the seeding qualified rate gradually rose with the increase in rotational speed. When the rotational speed exceeded the critical value, the seed population movement became excessively violent. Under the combined action of centrifugal force and inertia force, seeds tended to detach from the pockets, empty pockets appeared, and the seeding qualified rate decreased accordingly.

Figure 14. Variation of the seeding qualified rate with the rotational speed of the pocket wheel and the installation angle of the seed-stripping brush.

The response surface in Figure 15 illustrates the interactive influence of pocket wheel rotational speed and pocket count on seeding qualified rate at a fixed seed-stripping brush installation angle of 30°. When the rotational speed of the pocket wheel was fixed, the number of pockets affected seed-metering performance mainly by altering the seed-filling sites in each working cycle. resulting in a low seed-filling rate and a relatively low seeding qualified rate. With a moderate increase in the number of pockets, more pocket sites were available to accommodate seeds per unit time, the seed-filling rate was improved, seed-filling performance was enhanced, and the seeding qualified rate rose slowly. Nevertheless, an excessively dense pocket arrangement produced small spacing between adjacent pockets. Seeds competed for limited space within the seed-filling zone, which readily caused multiple seeds to enter the same pocket simultaneously and trigger multiple seeding, and the seeding qualified rate decreased accordingly [26].

Figure 15. Variation of the seeding qualified rate with the rotational speed of the pocket wheel and the number of pockets.

The three-dimensional response surface in Figure 16 visually demonstrates the nonlinear coupling relationship between seeding qualified rate and two key parameters—pocket count and seed-stripping brush installation angle—at a fixed pocket wheel rotational speed of 40 r·min−1. The surface takes on an overall inverted paraboloid form. When the number of pockets was fixed, the installation angle of the seed-stripping brush regulated the seed-stripping force primarily by adjusting the radial penetration depth of the bristles into the pocket surface. A small installation angle led to insufficient bristle penetration and weak seed-stripping capacity, which failed to thoroughly remove redundant seeds adhering to the pocket surfaces and resulted in a high multiple seeding rate. A moderate increase in the installation angle provided appropriate seed-stripping force, which effectively scraped off excess seeds outside the pockets without interfering with the pre-filled seeds inside the pockets. This realized a good coordination between seed-filling and seed-stripping processes and improved the seeding qualified rate. In contrast, an excessively large installation angle caused excessive penetration depth and overlarge seed-stripping force, which easily scraped out the seeds already filled in the pockets and thereby reduced the qualified rate.

Figure 16. Variation of the seeding qualified rate with the installation angle of the seed-stripping brush and the number of pockets.

3.4. Parameter Optimization

To optimize the performance of the seed-metering device and improve the operation quality, with the rotational speed of the pocket wheel X 1, the installation angle of the seed-stripping brush X 2, and the number of pockets X 3 as variables, single-objective optimization was performed. Combined with the previous analysis, calculations using Design-Expert software yielded the theoretically optimal parameter combination of the seed-metering device: a pocket wheel rotational speed of 45.70 r·min−1, a seed-stripping brush installation angle of 34.71°, and 13.84 pockets, corresponding to a model-predicted maximum seeding qualified rate of 99.04%. In consideration of practical engineering implementability, the theoretically optimal parameters are rounded to the following recommended engineering values: a pocket wheel rotational speed of 46 r·min−1, a seed-stripping brush installation angle of 35°, and a pocket count of 14.

(21)

4. Discussion

The present study addresses the core issue of parameter matching for the seed-metering performance of small intelligent vegetable seeders. Systematic verification and optimization were carried out via Box–Behnken response surface experiments, and the results elucidate the parameter influence mechanism of the pocket-wheel seed-metering device [27].

In terms of seed-metering performance, the three-factor quadratic regression model constructed in this study reached a highly significant level, clarifying the descending order of influence of the number of pockets, the rotational speed of the pocket wheel, and the installation angle of the seed-stripping brush on the seeding qualified rate; the optimized parameter combination yields a model-predicted maximum seeding qualified rate of 99.04% as calculated by Design-Expert software. Through multi-parameter collaborative optimization, this study further improved the seed-metering stability of small-sized vegetable seeds and revealed the extremely significant influence of the number of pockets on seed-metering performance, filling the research gap in existing studies, which mostly focused on rotational-speed optimization while lacking coupled analysis of pocket arrangement and seed-filling mechanisms. Meanwhile, the nonlinear variation law presented by the response surface analysis also explained, from the perspectives of seed population motion characteristics and seed-filling mechanics, the internal mechanism for the existence of optimal intervals for each parameter, providing a quantitative basis for the structural parameter design of the pocket wheel seed-metering device adapted to pelletized small-sized vegetable seeds.

Overall, the above-mentioned research results support the design rationality of the intelligent vegetable seeder from the perspective of seed-metering performance and possess theoretical significance and engineering guiding value for perfecting the design theory of small-sized precision seeders and reducing machine-related research-and-test costs. Nevertheless, the seed-metering parameter optimization in this study is completed on the basis of prototype experiments, and the analysis is only conducted for small-size pelleted vegetable seeds. The influences of variations in seed particle size and morphology on seed-metering performance remain unclear, so the applicable boundary of the research conclusions has certain limitations. Future research can further extend the study on seed-metering adaptability for vegetable seeds of multiple varieties and morphological forms. In terms of energy supply, the seeder currently adopts a dual-power supply mode combining photovoltaic (PV) panels and storage batteries. However, the photoelectric conversion efficiency of fixed PV panels is greatly affected by the solar altitude angle and azimuth angle, and their power generation potential has not been fully exploited. In follow-up work, an intelligent solar tracking system will be developed, which adopts a dual-axis altitude-azimuth tracking mechanism to orient the PV panels toward the sun in real time and improve photoelectric conversion efficiency. Furthermore, environmental perception modules including machine vision and LiDAR will be introduced on the basis of the existing STM32 control system. Research on crop row recognition, field boundary detection, and path planning will be carried out to achieve autonomous navigation and automatic obstacle avoidance and promote the iterative upgrading of the whole machine towards higher intelligence [28].

5. Conclusions

This study addressed the drawbacks of single function, insufficient intelligence, and limited endurance of existing vegetable seeders and developed an intelligent vegetable seeder integrating furrow opening, seeding, soil covering, and compaction. Structural design, construction of the intelligent seeding system, and operational performance test verification were systematically completed. In terms of structural design, the core parameters of the pocket wheel seed-metering device, core-share furrow opener, and double-disc soil-covering and compacting device were determined. A dual power-supply mode of “PV panel + storage battery” was adopted to guarantee operation endurance. For intelligent seeding, dual-ultrasonic fusion monitoring of seeding depth and real-time photoelectric detection of multiple and missed seeding were realized based on the STM32 microcontroller and Kalman filter algorithm [29]. The quadratic regression model for the seeding qualified rate established via the Box–Behnken response surface method was highly significant. The descending order of influence of the three factors was the number of pockets, the rotational speed of the pocket wheel, and the installation angle of the seed-stripping brush. The theoretically optimal parameter combination was obtained: a pocket wheel rotational speed of 45.70 r·min−1, a seed-stripping brush installation angle of 34.71°, and 13.84 pockets. This study serves as a reference for the development of intelligent agricultural machinery and intelligent agriculture.

Author Contributions

Conceptualization, Y.Z.; methodology, Y.Z.; validation, J.D.; investigation, F.Y.; resources, F.Y.; writing—original draft preparation, J.D. and G.S.; writing—review and editing, Y.Z. and F.Y.; supervision, X.H. and J.L.; project administration, F.Y. and Y.L.; funding acquisition, F.Y. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Natural Science Foundation of Shandong Province (Grant No. ZR2024QE419), the Science and Technology Planning Project of Xinjiang Production and Construction Corps (Grant No. 2024AB047), the Key R&D Program of Ningxia Hui Autonomous Region (Grant No. 2024BBF02028), and the Key R&D Program of Linyi City (Grant No. 2025027).

Institutional Review Board Statement

Not applicable.

Data Availability Statement

Some or all data, models, or code generated or used during the study are available from the corresponding author by request.

Acknowledgments

The authors are grateful for the experiment site provided by Linyi Intelligent Agricultural Machinery Industry Technology Research Institute and the equipment support provided by Linyi Key Laboratory of Modern Agricultural Machinery Equipment.

Conflicts of Interest

The authors declare that this study received funding from the Department of Science and Technology of Shandong Province, the Xinjiang Production and Construction Corps and the Science and Technology Department of Ningxia. The funders were not involved in the study design, collection, analysis, interpretation of data, the writing of this article, or the decision to submit it for publication.

Abbreviations

The following abbreviations are used in this manuscript:

PV photovoltaic
DC direct current
RMSE root mean square error
BBD Box–Behnken design
ANOVA analysis of variance
df degrees of freedom
CV coefficient of variation

References

1. Wei, D.Z. Design and experiment of leek seeder with deep trench and shallow soil covering. Master’s Thesis, Shenyang Agricultural University, Shenyang, China, 2025. [Google Scholar] [CrossRef]
2. Liu, H.; Du, Z.; Guo, X.; Li, H.; Ma, X.Y.; Liao, Y.T. Research progress on precision direct seeding technology and equipment for vegetable seeds. J. Huazhong Agric. Univ. (Nat. Sci. Ed.)2025, 44, 25–36. [Google Scholar] [CrossRef]
3. Li, J.; Sun, W.; Wang, H.; Wang, J.; Simionescu, P.A. An integrated potato-planting machine with full-film mulching and ridged row soil covering. Agriculture2024, 14, 860. [Google Scholar] [CrossRef] [Scilit]
4. Chen, G.; Yao, Y.; Yi, L.; Yin, X.; Du, J.; Chong, J. Automatic precision planting mechanism of garlic seeder. Agriculture2025, 15, 849. [Google Scholar] [CrossRef] [Scilit]
5. Valentin, M.T.; Białowiec, A.; Karayel, D.; Jasinskas, A.; Ciolkosz, D.; Lavarias, J.A. Investigation of the performance of a cylindrical hopper and metering device of a carrot seeder. Sci. Rep.2023, 13, 813. [Google Scholar] [CrossRef] [Scilit] [PubMed]
6. Elwakeel, A.E.; Elbeltagi, A.; Dewidar, A.Z.; Salem, A.; Abdeen, M.A. Development and laboratory evaluation of a novel IoT-based electric-driven metering system for high precision garlic planter. PLoS ONE2024, 19, e0317203. [Google Scholar] [CrossRef] [Scilit] [PubMed]
7. Chen, Z. Research progress on key technologies of precision planter. Agric. Eng.2025, 15, 1822. [Google Scholar] [CrossRef]
8. Wang, X.; Hu, L.; Wu, H.; Gao, X.; Wang, G.; Chen, Y. Current development status of peanut seed metering devices. Agric. Eng.2026, 8, 221. [Google Scholar] [CrossRef] [Scilit]
9. Bozdar, B.; Ahmed, N.; Meghwar, M.R.; Zhu, Z.; Talpur, A.; Li, Z.H. Seed pelleting technologies: Paving the way for resilient and sustainable future farming. Food Energy Secur.2026, 15, e70193. [Google Scholar] [CrossRef] [Scilit]
10. Rao, G.; Zhao, W.Y.; Shi, L.R.; Chen, X.; Sun, B.G.; Wang, Z.; Li, H. Design and DEM simulation analysis of shaped groove eyelet wheel rapeseed precision seed dispenser. Int. J. Agric. Biol. Eng.2025, 18, 102–116. [Google Scholar] [CrossRef] [Scilit]
11. Qu, Z.; Liu, L.; An, X.; Lu, Q.; Ding, L.; Yu, Y. Optimization design and experiment of soybean cell-wheel seed metering device based on EDEM. J. Agric. Mech. Res.2023, 45, 103–110. [Google Scholar] [CrossRef]
12. Wang, B.S.; Wang, L.; Liao, Y.T.; Wu, C.; Cao, M.; Liao, Q.X. Design and test of seeding wheels of precision hole-seeding centralized metering device for small particle size seeds. Trans. Chin. Soc. Agric. Mach.2022, 53, 64–75. [Google Scholar] [CrossRef]
13. Zhang, T. Research on optimization design and quantitative evaluation system of ditching and soil-covering device for vegetable seeder. Master’s Thesis, Shandong Agricultural University, Tai’an, China, 2024. [Google Scholar] [CrossRef]
14. Liu, T. Optimization design and experiment of core-share seed drill furrow opener suitable for heavy clay soil. Master’s Thesis, Hebei Agricultural University, Baoding, China, 2024. [Google Scholar] [CrossRef]
15. Hao, J.; Qu, P.; Liu, T.; Zhao, J.; Yin, C.; Zhao, J.; Wang, X. Optimized design and testing of core-share seed drill furrow opener suitable for heavy clay. Trans. Chin. Soc. Agric. Mach.2026, 57, 121–130. [Google Scholar] [CrossRef]
16. Yuan, H. Design and test of rape bare-seedling opposite-row transplanting machine. Trans. Chin. Soc. Agric. Eng.2025, 41, 112–121. [Google Scholar] [CrossRef]
17. Sultan, U.; Khan, A.A.; Husnain, S.N.; Zohaib, M.; Hashim, N. Multi-crop high efficiency seed drill with solar hybrid seed-metering: A step toward precision and sustainability. Front. Sustain. Food Syst.2023, 6, 1081122. [Google Scholar] [CrossRef] [Scilit]
18. Liu, Z.; Xia, J.; Hu, M.; Du, J.; Luo, C.; Zheng, K. Design and analysis of a performance monitoring system for a seed-metering device based on pulse width recognition. PLoS ONE2021, 16, e0261593. [Google Scholar] [CrossRef] [Scilit] [PubMed]
19. Wang, C.Y.; Wang, X.Z.; Yang, Q.Z.; Zhang, R. Design and experiment of an online monitoring system for peanut seeding parameters based on IoT. Int. J. Agric. Biol. Eng.2025, 18, 140–148. [Google Scholar] [CrossRef] [Scilit]
20. Yu, J.Q.; Wang, G.; Xin, N.; Fu, H. Working process and performance simulation of cell-wheel seed-metering device. Trans. Chin. Soc. Agric. Mach.2011, 42, 83–87, 101. Available online: https://www.j-csam.org/jcsam/article/abstract/20111215 (accessed on 18 October 2026).
21. Liu, T.; He, R.Y.; Lu, J.; Zou, Y.; Zhao, M.M. Simulation and experiment on seeding performance of nest-eye-wheel rapeseed seed-metering device based on EDEM. J. South China Agric. Univ.2016, 37, 126–132. [Google Scholar] [CrossRef]
22. Yu, J.Y.; Lu, C.Y.; Wei, R.X.; Fu, X.L.; Li, K.; Tang, Z.J.; Wang, F.H. Performance simulation test of wheat precision seed-metering device based on discrete element method. Jiangsu Agric. Sci.2018, 46, 225–228. [Google Scholar] [CrossRef]
23. Lu, Y.L.; Jiang, J.T.; He, X.N.; Wang, D.Q.; Zheng, X.S. Design and experiment of Cyperus esculentus cell-wheel seed-metering device. J. Agric. Mech. Res.2023, 45, 156–163. [Google Scholar] [CrossRef]
24. Gao, X.; Cui, T.; Zhou, Z.; Yu, Y.; Xu, Y.; Zhang, D.; Song, W. DEM study of particle motion in novel high-speed seed-metering device. Adv. Powder Technol.2021, 32, 1438–1449. [Google Scholar] [CrossRef] [Scilit]
25. Qian, C.; Qin, W.; Jiang, Y.C.; Huang, Z.S.; Zhang, M.H.; Wang, Z.M.; Yang, W.W.; Zang, Y. Parameter optimization and experiment of the vacuum seed meter with double holes for direct seeding of rice. Int. J. Agric. Biol. Eng.2025, 18, 89–99. [Google Scholar] [CrossRef] [Scilit]
26. Wang, J.W.; Qi, X.; Xu, C.S.; Wang, Z.M.; Jiang, Y.M.; Tang, H. Design evaluation and performance analysis of the inside-filling air-assisted high-speed precision maize seed-metering device. Sustainability2021, 13, 5483. [Google Scholar] [CrossRef] [Scilit]
27. Zheng, G.; Qi, B.; Zhang, W.Y.; Shao, W.X.; Zhang, L.; Wang, Y.X.; You, Q.D. Engineering discrete element method-based design and optimization of the key components of a spoon-wheel spinach seed-metering device. Agronomy2024, 14, 2096. [Google Scholar] [CrossRef] [Scilit]
28. Shi, J.Y.; Bai, Y.H.; Diao, Z.H.; Zhou, J.; Yao, X.B.; Zhang, B.H. Row detection based navigation and guidance for agricultural robots and autonomous vehicles in row-crop fields: Methods and applications. Agronomy2023, 13, 1780. [Google Scholar] [CrossRef] [Scilit]
29. Zhao, N.; Zhao, B.; Yi, S.; Zhou, Z.; Che, G. Research on a sowing depth detection system based on an improved adaptive Kalman filtering method. Electronics2022, 11, 3802. [Google Scholar] [CrossRef] [Scilit]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.

Article Metrics

Citations

Multiple requests from the same IP address are counted as one view.