Development of a Bionic Picking Device for High Harvest and Low Loss Rate Pod Pepper Harvesting and Related Working Parameter Optimization Details

Abstract

Addressing the issues of low harvesting efficiency and high losses in current pod pepper harvesters, this study presents the design of a bionic comb finger pod pepper picking device and an inclined double-spiral bell pepper harvester to improve the harvest rate and reduce loss rate. Through the utilization of discrete element simulation software EDEM, a discrete element model for pod peppers is established. Additionally, a simulation platform for ground drop loss during pod pepper picking is developed, enabling exploration of the movement trajectory and velocity changes of the pod pepper elements. The study also conducts an analysis on the impact of the speed of the picking rollers X₁, the feeding speed of pod peppers X₂, and the spacing between the two picking rollers X₃ on ground drop losses. Based on the results of the single-factor test, the Box–Behnken response surface test was used to optimize the working parameters of the picking device, which resulted in the optimal combination of the working parameters of the picking device: the speed of the picking rollers was 680.41 rpm, the feeding speed of the pod peppers was 0.5 m/s, and the spacing between the two picking rollers was 12 mm, which resulted in the loss rate of pod peppers on the floor of the ground being 3.526%.

1. Introduction

The pepper planting area in Jiangsu Province is mainly concentrated in Feng County and Pei County in Xuzhou City. While there is also a small amount of pepper planting in other areas, the scale is small and the varieties are mixed. According to information, in 2020, Pei County’s pepper planting area covered 17,000 acres, where a total of 158 registered varieties of pepper were grown, of which the “Su Heng Run”, “Su Run”, and “Run La” brands also won the “China Well-known” award. Three brands have won the title of “China Famous Brand”, and cultivated more than 3000 new vocational and technical farmers, the annual output value exceeded 500 million yuan, and effectively driving the local farmers to increase their income by 250 million yuan; in 2021, Feng County pepper planting area reached more than 30,000 acres; the economic output value reached about 150 million yuan, effectively driving farmers to increase income and get rich, and now it has become an important poverty alleviation industry. Pod pepper picking is labor intensive; relative to the value of the product, the cost of labor continues to increase, making the economic benefits of pod pepper planting decline. The farmers planting pod peppers decreased, resulting in pod pepper planting areas continuing to shrink. The development of a small hand-held pod pepper harvester with a high picking rate, low loss rate, and compact structure adapted to small field operations is a key technical problem that needs to be solved in China at present. Therefore, expediting research on key technologies, such as the spiral comb finger picking device, holds substantial significance for advancing China’s pepper industry, fostering increased income and wealth among farmers.

There exists a significant disparity between the research levels of domestic and foreign pepper harvesters. Foreign pepper harvester technology is very mature; pepper harvesting research has a longer history in other countries including the United States [1], Mexico [2,3], South Korea [4], Europe, and other countries. Although the foreign pepper harvester has advanced technology, the price is relatively high, and the domestic pepper planting mode is not applicable, making it difficult to meet the domestic pepper harvesting market. Domestic development of pepper harvesters started late. At present, individual enterprises and research institutions in Xinjiang Province [5,6], Shanxi Province, Jiangsu Province [7], and Guizhou Province [8,9] carry out the development of pepper harvesters. The development of pepper harvesters is mainly used for large field operations. In Guizhou Province, Shandong Province, Henan Province, Jiangsu Province, and other places where pod pepper is grown, pepper planting in small fields and individual households planting mainly rely on manual picking, but the manual picking cost is high and the harvesting cycle is long, affecting the planting of the next season’s crop. In Guizhou Province, Shandong Province, Henan Province, Jiangsu Province, and other places where pod pepper is grown, the pepper harvest is faced with artificial picking costs that increase year by year. The pepper harvest “no machine available” problem is becoming more and more prominent.

The inclined double-helix picking device has the advantages of compact structure, high flexibility, wide picking range, etc., which can achieve the continuous combing action of pepper, and has been studied by a large number of scholars. Among them, in 1997, E. Palau et al. [10] developed a hydraulically-driven single-row pepper harvester, which brushes the peppers from the plants through the reverse rotation of the combing drums and throws the pepper fruits to the conveyor belts on both sides, and the conveyor belts convey the peppers while the fan clears and separates the impurities left in them, and finally conveys the peppers to the bins; in 2018, Byum et al. [11] used discrete element simulation software EDEM to analyze the effect of pepper harvesting and picking components on the effect of pepper picking; In 2021, Yuan et al. determined the main factors affecting pepper harvesting by analyzing the force on the spiral steel bar in contact with the pepper and conducting a field test [12], and determined the range of factors selected in the optimization test by using a one-way trial. In 2024, Du et al. [13] used EDEM to carry out simulation tests on the roller comb-finger-type pepper picking device and carried out simulation analysis through the response surface orthogonal test method to determine the main factors affecting the loss rate of pod peppercorns falling on the ground and their optimal parameter combination values. At present, there is less research on the spiral comb finger picking device, the test results fail to meet the expected performance requirements, the breakage rate and ground loss rate are high, the comb finger structure, material and arrangement still need to be studied, and the structural parameters and movement parameters of the pepper picking device still need to be optimized.

This article develops a high-clean and low-loss small hand-held pod pepper harvester through the discrete element simulation software EDEM, establishes a discrete element model of pod pepper, builds a simulation platform of ground drop loss of pod pepper picking device, and explores the movement trajectory and speed change of the discrete element model of pod peppers. Taking the rate of ground drop loss of pod peppers as the evaluation metric, the influence of picking roller speed, feeding speed of pod peppers, and spacing between two picking rollers on the ground drop loss of pod peppers is analyzed through a one-factor simulation test, and the better value range of each factor is determined. Based on the results of the single-factor test, the Box–Behnken response surface test was used to optimize the working parameters of the picking device, to analyze the significant relationship between the picking roller speed, feeding speed of pod peppers, and the spacing between the two picking rollers on the ground loss rate of pod peppers, and to explore the optimal working parameters of the picking device.

2. Materials and Methods

2.1. Bionic Comb Finger Pod Pepper Picking Device

As shown in Figure 1, a small hand-held pod pepper harvester was designed for this study. The harvester mainly consists of a frame, a picking device, a conveying device, a row spacing adjustment device, a storage device, a transmission device, a chassis traveling device, and a grain splitter. The frame and other components are mounted on the chassis traveling device through the three-point suspension device. A hand-held tractor with a power of 9.2 kW was used for the chassis-traveling device. The picking device and other working parts are located at the front end of the chassis traveling device, which can realize a number of operations such as picking, conveying, and storing pod peppers. The pod pepper plant enters the comb finger screw picking rollers after harvesting, and a pair of comb finger screw picking rollers rotate to the outside synchronously. As the harvester walks, the comb fingers begin to brush the pod pepper plants, and the combed-down pod pepper fruits are thrown onto the conveyor belt with the rotation of the comb-finger-type screw picking rollers and then transported to the material box through the conveyor belt.

The picking device in this pod pepper harvester is designed as a high-clean and low-loss bionic pod pepper picking device using the human hand as a bionic prototype for reducing the breakage rate of pod pepper harvesting and increasing the clean rate of pod pepper picking. The design of the bionic comb finger is based on the human hand as a bionic prototype, and the structural morphology and movement process of the hand during manual picking of pod peppers were studied using the i-SPEED TR Endoscopic High-Speed Dynamic Analysis System (i-SPEED TR, Olympus Corporation, Tokyo, Japan; this camera is a high-speed microscope with a 1280 × 1024 pixel sensor, capable of reaching a frequency range of 24 fps to 15,000 fps. Its ultra-high sensitivity and remote-control panel CDU make the i-SPEED TR more practical. It can be used to measure distance, speed, acceleration, angular velocity, and angular acceleration, making its functions more flexible) [14], and the following was obtained: when the worker is picking the peppers, the forefinger and the thumb pick the peppers, where the thumb mainly provides pressure, and the static friction between the forefinger and the peppers’ epidermis is generated when the arm applies the picking force. When the static friction force is greater than the connection force between the pod pepper fruit and the pod pepper plant, the pod pepper is picked, and the whole process takes the structural morphology of the index finger when the pod pepper is picked. The extracted images were imported into i-SPEED Suite analysis software for the extraction of index finger structure morphology, and the extracted data were imported into Origin software for curve fitting. The obtained function image was shown in Figure 2a, and the fitting equation was as follows: (1) and the curve-fitting goodness of fit R² = 0.9965, which is close to the value of 1, which indicates that the curve is closer to the structure and morphology of the index finger.

The structural parameters of the bionic comb finger are designed according to the obtained fitted curve, and the bionic comb finger shown in Figure 2b is generated in SolidWorks 3D drawing software by releasing the bionic comb finger with the equation of the fitted curve as the bionic comb finger central curve.

$$y=0.003706x^2+0.003577x+0.288730$$

(1)

where 0 ≤ x ≤ 70.

The picking roller is the core component of pod pepper harvesting, which separates the pod pepper from the stalk by combing. The diameter, length, and comb finger spacing of the picking roller will affect the quality of the pod pepper picking operation. As shown in Figure 3 for the pod pepper picking structure schematic diagram, L is the length of the picking roller, D is the picking roller diameter, L₁ is the pitch, and l is the comb finger spacing. Two rows of bionic comb fingers are arranged in the picking roller according to the helix of the central axis, and through the welded way of fixing the central axis, in order to ensure that the picking roller on the pod peppers has the continuous combing effect, the two-helix starting angle is according to the configuration of 180°, and the design of the adjacent two comb finger spacing of 20 mm, pitch 450 mm. The length and diameter of the picking roller should be designed to ensure that the maximum and minimum height of the pod pepper can be picked, so the design is based on the characteristics of the pod pepper plant.

The fruits of the pod pepper are densely distributed from the minimum to the maximum fruiting height. In order to ensure that all pod peppers within the picking range are picked, the design of the picking roller should be adapted to the growth characteristics of the pod pepper. A schematic diagram of the contact between the bionic comb finger picking device and the pod pepper is shown in Figure 4. Where X₁ represents the picking roller speed and X₂ represents the feeding speed of the pod pepper. From Figure 4, it can be seen that the design of the picking roller is mainly determined by the pod pepper plant height H, the height of the picking roller above the ground H₁, the minimum fruiting height H₂, the installation angle of the picking roller θ, and other factors. According to the pre-measured pod pepper, plant height is 75.99 ± 6.129 cm and the minimum fruit height is 37.90 ± 4.332 cm. Due to the poor road conditions during the harvesting process of pod peppers, the harvester will have bumps and ups and downs, so in order to prevent the picking rollers from touching the ground, in order to ensure that the height of the picking rollers from the ground is less than the minimum fruiting height of the pod pepper, the picking rollers and the ground must leave a gap of 200 mm.

Considering the conveyance of pod peppers and the size of the whole machine structure, if the inclination angle of the picking roller is too large, it will lead to pod pepper accumulation; on the other hand, if the inclination angle is too small, it will increase the length of the whole machine and reduce the field operability. The tilt angle of the picking roller needs to be considered parallel to the conveyor belt, according to the conveyor belt tilt angle design picking roller tilt angle, the conveyor belt tilt angle according to the friction coefficient of the pod pepper fruit and rubber to determine. The previous study shows a pod pepper and rubber maximum friction tilt angle of 31.24° [15]. In order to ensure the compactness of the whole machine structure, the tilt angle of the picking roller is taken as 30°, and the length of the picking roller is calculated to be ≥1100 mm. In order to ensure that the pod peppers will not be missed, the length of the whole machine should not be too large and should be easy to operate in the field; the length of the picking roller is selected as 1200 mm. According to Equation (3), the diameter of picking roller D ≤ 221.9 mm. Considering the width of the harvesting platform and the accessibility of the pod pepper plants, the diameter of the picking roller is selected at 150 mm.

$$L\ge \left(H-H_1\right)/\mathit{sin}\theta$$

(2)

$$H_2\ge H_1+D\mathit{cos}\theta$$

(3)

where

• L: picking roller length, mm;
• H: plant height, mm;
• H₁: picking roller height above ground, mm;
• H₂: minimum result height, mm;
• θ: picking roller installation angle, °.

2.2. Construction of Discrete Element Simulation Platform for Pod Pepper Picking Device

Based on the previous measurements of the three-dimensional dimensions of the pod pepper, a three-dimensional model of the pod pepper was constructed in SolidWorks with the following parameters: The maximum diameter of the fruit was 9.724 mm, the length of the fruit was 54.789 mm, the diameter of the fruit stalk was 2.626 mm, and the length of the fruit stalk was 36.896 mm. The three-dimensional model of the pod pepper was saved as a “.step” file and then imported into EDEM software. The discrete element model of pod pepper was generated using the automatic filling method of the particle template [16,17], as shown in Figure 5. During the filling process of the particle template, the degree of proximity of the discrete element model of pod peppers to the 3D model depends on the number of particles. As the number of particles increases, the simulation computation increases and efficiency decreases. Therefore, under the premise of ensuring that the pod pepper discrete element model is close to the 3D model, it is recommended to set the number of particles filling to 98 to improve the efficiency of the simulation calculation.

To enhance simulation efficiency, simplification of the pod pepper harvester was pursued. A simplified pod pepper picking apparatus was modeled in SolidWorks, featuring dimensions of 1000 mm × 530 mm × 1150 mm. This rendition retained only essential components, notably a pair of picking rollers and the frame. The resulting 3D model of the picking device was saved in “.step” format and subsequently imported into EDEM. In the EDEM software, according to Figure 6, on the basis of the original structure, the pod pepper collection area, the drop area, and the particle factory area are divided, which are used to generate and count the loss of pod peppers dropped on the ground. In the EDEM software, the relevant parameters of the picking device were set to Q235 material properties, and the merge geometry function was used to merge the parts of the picking device into three parts: the picking device shell, the right picking roller, and the left picking roller. The picking rollers are used as moving parts, and the outward rotation speed around the axis is added to the left and right picking rollers.

Set the attribute parameters, contact parameters, and contact model for the models of pod pepper and picking device imported into EDEM [18]. According to the results of previous research, the attributed parameters of the pod pepper and picking device are set as shown in

Table 1. Material parameters of the pod pepper and harvesting device.

Materials	Poisson’s Ratio	Young’s Modulus (MPa)	Density (kg/m3)
Pod pepper	0.406	2.39	767
Harvesting device (Q235)	0.280	2.1 × 103	7850

; according to the results of previous research in the group, the contact parameters between pod pepper–comb finger and pod pepper–pod pepper are set, and the values of the contact parameters are shown in

Table 2. Contact parameters for pod pepper–comb fingers and pod pepper–pod pepper [15].

Materials	Coefficient of Restitution	Coefficient of Static Friction	Coefficient of Rolling Friction
Pod pepper–Pod pepper	0.36	0.51	0.11
Pod pepper–Comb finger	0.42	0.39	0.19

. Given that there is no adhesion between the pod pepper and comb finger and the deformation of the pod pepper is small during the picking process, the Hertz–Mindlin (no slip) model and the Standard Rolling Friction contact model are used in the EDEM [19]. Gravity, simulated with an acceleration of 9.81 m/s² in the positive X-axis direction, was applied to replicate the downward descent of pod peppers during the picking process.

During the harvesting process, the generation point of the pellet plant was set according to the feeding position of the pod pepper. The location of the pellet plant was determined as a rectangular surface with dimensions of 500 mm × 200 mm and central coordinates of (X: −100; Y: 84.84; Z: 900), and the angle between this rectangle and the XY plane was 45°. The parameters of the pellet plant were set according to the number of fruits per pod pepper plant, plant spacing, and machine travel speed. According to the results of the previous study, the average number of fruits per pod pepper plant was 95, and the plant spacing was 25.74 cm. The walking speed of the machine was taken as the average value, and the total number of pod pepper factories generated was set to 256. The generation method of the pod pepper discrete element model was set to dynamic, and a velocity along the positive direction of the Z axis was added to the generated pod pepper discrete element model as the pod pepper feeding velocity. This velocity will be used to study the effect on the ground drop loss of the pod pepper picking device.

Upon completing the preprocessor settings, simulation parameters must be configured in the solver prior to initiating the simulation. In order to ensure structural accuracy, the simulation time should be shortened, and the simulation efficiency should be improved as much as possible. Therefore, the total simulation time is set to 5 s, the time step is set to 19.33%, the data saving interval is set to 0.01 s, and the mesh size is set to 3 R min. Once settings are finalized, initiate the simulation by clicking “Start Simulation”.

After completing the solution, in order to detect the number of pod peppers in the collection and drop zones, refer to Figure 7, and use the Setup Selections analysis function in the Grid Bin Group module in the post-processor to establish detection regions that overlap with the collection and drop zones by detecting the number of fallen pod peppers in each area in real time and using the following formula to calculate the ground fall loss rate of pod peppers (η_s2). In the post-processor, the movement trajectory and speed change of the pod pepper can be visualized, as shown in Figure 8. As illustrated in Figure 8a, pod pepper particles exhibit downward motion post-generation, being directed towards the collection area upon contact with comb fingers. However, some pod peppers fail to engage with comb fingers, consequently falling into the drop area. According to Figure 8b, some of the pod pepper that contacted the comb fingers had a large velocity and may have collided with the side baffles in the subsequent movement, resulting in their ejection to the ground. This part of the pod pepper is more likely to be damaged as it undergoes two collisions.

$${\eta }_{s2}=\frac{n_{s1}}{n_{s1}+n_{s2}}$$

(4)

where

• η_s2: the ground drop loss rate of pod peppers in simulation experiment, %;
• n_s1: the number of pod peppers dropped in the designated area;
• n_s2: the number of pod peppers collected in the designated area.

3. Experiment Design

3.1. Single-Factor Simulation Experiment

By drawing on the research experience of Xinjiang, Inner Mongolia Province, and foreign countries on the inclined double-screw picking device, combining the existing research data on the pod pepper harvester and the previous research results on the collision loss mechanism of pod pepper, the speed of the picking rollers X₁, the feeding speed of pod pepper feeding X₂, and the spacing between the two picking rollers X₃ were determined as the experimental factors to study the ground drop loss of the picking device. Taking the rate of ground drop loss of pod pepper η_s2 as the evaluation index, the experimental range of each factor was selected, and the specific range is shown in

Table 3. Single-factor experimental factors and ranges of ground drop loss in the picking device.

Item	X1 (rpm)	X2 (m/s)	X3 (mm)
Level	400~800	0.2~1.0	5~45

Note: X₁, the speed of the picking rollers. X₂, the feeding speed of pod peppers feeding. X₃, and the spacing between the two picking rollers.

. To ensure the accuracy and reliability of the simulation test of the ground drop loss of pod pepper picking device, each group of simulation tests will be repeated three times, and the final simulation results will be averaged.

The ground drop loss test of the pod pepper picking device was carried out by maintaining the feeding speed of pod pepper at 0.6 m/s and setting the spacing between the two picking rollers at 25 mm. The picking roller speeds of 400 rpm, 500 rpm, 600 rpm, 700 rpm, and 800 rpm were selected for the test. Then, keeping the spacing between the two picking rollers at 25 mm, the picking roller rotation speed was preferably 650 rpm, and the pod pepper feeding speeds were selected to be 0.2 m/s, 0.4 m/s, 0.6 m/s, 0.8 m/s, and 1.0 m/s for the experiments, respectively. Finally, keeping the picking roller speed at 650 rpm and the pod pepper feeding speed at 0.6 m/s, the two picking rollers were selected with a spacing of 5 mm, 15 mm, 25 mm, 35 mm, and 45 mm for the experiments, respectively.

3.2. Box–Behnken Response Surface Experiment

The results of the single-factor simulation test showed that the picking roller rotation speed X₁, the pod pepper feeding speed X₂, and the spacing between the two picking rollers X₃ had a significant effect on the rate of ground drop loss of pod pepper η_s2. Although preliminary preferences for parameters have been made for each factor, only the influence pattern of individual factors on the ground drop loss of pod pepper has been obtained, and the interaction effect between factors has not yet been derived. For this reason, using Design-Expert 13.0 software, the peripheral ranges of preliminary preferred values for picking roller speed X₁, pod pepper feeding speed X₂, and spacing between two picking rollers X₃ were selected, and a three-factor, three-level Box–Behnken response surface optimization test was designed, with the specific experimental factor design shown in

Table 4. Factors and levels of response surface experiment.

Level	X1 (rpm)	X2 (m/s)	X3 (mm)
−1	600	0.5	10
0	650	0.6	15
1	700	0.7	20

Note: X₁, the speed of the picking rollers. X₂, the feeding speed of pod pepper. X₃, and the spacing between the two picking rollers.

below.

To ensure the reliability of the simulation test results, we conducted Box–Behnken response surface optimization tests for ground fall loss of the picking device of pod peppers using the ground fall loss rate of pod peppers, η_s2, as the evaluation index. Three replications were conducted for each set of tests, and the test results were averaged. The specific experimental design and results are shown in

Table 5. Design and results of response surface optimization experiments.

Item	X1 (rpm)	X2 (m/s)	X3 (mm)	ηs2 (%)
1	700	0.7	15	4.73
2	650	0.5	20	4.75
3	700	0.6	20	5.37
4	600	0.5	15	9.02
5	650	0.7	10	6.24
6	700	0.6	10	4.16
7	600	0.6	20	12.52
8	600	0.7	15	14.27
9	700	0.5	15	3.61
10	650	0.7	20	8.83
11	650	0.6	15	6.52
12	650	0.6	15	5.97
13	650	0.6	15	6.23
14	600	0.6	10	9.64
15	650	0.6	15	6.97
16	650	0.6	15	5.14
17	650	0.5	10	4.01

Note: X₁, the speed of the picking rollers. X₂, the feeding speed of pod pepper. X₃, and the spacing between the two picking rollers.

.

3.3. Field Trial Validation

According to the results of the previous research, Jiangsu University and Shandong Dali Agricultural Machinery Co., Ltd., Dezhou, China, jointly modified the three-dimensional and two-dimensional drawings, and completed the processing and assembly of the prototype machine in Dali Agricultural Machinery Co., Ltd. to manufacture the 4JF-1.1 small hand-held pod pepper harvester. The harvester is mounted on a 9.2 kW hand-held tractor and includes a picking device, a conveying device, a material storage device, and a chassis traveling device. The overall manufacturing cost is 25,000 RMB. The relevant technical parameters of this pod pepper harvester are shown in

Table 6. Technical specifications of the 4JF-1.1 small hand-held pod pepper harvester.

Parameters	Type/Value
Structure type	Hand-held
Harvesting rows	2
L × W × H/mm × mm × mm	2400 × 1300 × 1200
Matching power/kW	9.2
Travelling speed/m/s	0.2~1.1
Type of picking device	Bionic comb fingers
Picking device speed/rpm	500~750

. In order to verify the operational performance of the 4JF-1.1 small hand-held pod pepper harvester and the reliability of the simulation test results of the bionic harvesting device, the pod pepper harvesting test was carried out from 15 January 2024 to 18 January 2024 at Jiangsu University.

Before commencing the experiment, the test plot was tidied up, and the row spacing of 70.15 cm and plant spacing of 25.35 cm for pod pepper planting were set according to the agronomic requirements for field planting. With reference to the pod pepper harvester industry standard DG/T 114-2019 pepper harvester, a test sample area of 1.7 × 1 m² was designed, and three sample areas were randomly selected in the test site, and the parameters of pod peppers in the sample areas were counted. For each trial, a bionic comb finger and hollow spiral picking device were used for harvesting two rows of pod peppers. Before the start of the trials, the pod pepper harvester was tuned to ensure that the engine speed, picking drum speed, and machine traveling speed were stable. At the end of the test, three 1.7 × 1 m² pod pepper sampling areas were randomly selected, and the net picking rate and ground drop loss rate of pod peppers in the sampling areas on both sides were counted with the middle line of the long side as the boundary, respectively. Subsequently, the mass of unpicked pod peppers and the mass of pod peppers dropped on the ground were tallied, and the total mass of pod peppers in the sampling area was utilized to calculate the net picking rate and ground drop loss rate using the formulas (5) and (6), respectively. The experiment was conducted in triplicate. Figure 9 illustrates the pod pepper harvester.

$${\eta }_1=\frac{{m_1-m}_2}{m_1}$$

(5)

$${\eta }_2=\frac{m_3}{m_1}$$

(6)

where

• η₁: Pod pepper picking rate, %;
• η₂: Rate of loss from ground fall of pod pepper, %;
• m₁: Mean value of quality of pod peppers in the sampling area, g;
• m₂: Quality of unpicked pod peppers in the sampling area, g;
• m₃: Mass of fallen pod peppers on the ground in the sampling area, g.

4. Results and Discussion

4.1. Analysis of the Results of the Single-Factor Simulation Test

When the feeding speed of the pod pepper was 0.6 m/s and the spacing between the two picking rollers was 25 mm, the test results were shown in Figure 10a. When the rotational speed is 400~600 rpm, the ground drop loss of pod peppers is due to the small speed of pod peppers after contacting with the picking rollers, thus falling into the drop zone; when the rotational speed is 600~800 rpm, the ground drop loss of pod peppers is due to the fact that the pod pepper has a large speed after colliding with the comb fingers, and they are ejected into the drop zone after colliding with the side baffle plate. Taken together, the above analysis shows that either too fast or too slow picking roller speeds will lead to an increase in the ground drop loss of pod peppers, and faster speeds will increase the damage rate of pod peppers. Kang et al. designed an inclined auger pepper harvesting test stand [20], which determined the optimum speed for pepper picking by analyzing the effect of speed on the efficiency of pepper harvesting but did not further optimize this speed. As a result, the efficiency of the bench test was low, which increased the development cycle of the pepper harvester.

When the spacing between the two picking rollers is 25 mm and the preferred picking roller speed is 650 rpm, the simulation results are shown in Figure 10b. Pod pepper feeding speed ranges from 0.2 to 1.0 m/s, the rate of ground drop damage of pod peppers increases gradually with the increase of feeding speed. When the feeding speed was in the range of 0.2~0.6 m/s, the rate of ground drop damage of pod peppers increased from 4.75% to 6.19%, with a slower rate of increase. In summary, when the feeding speed of pod peppers reaches the maximum feeding value of the harvesting device, the rate of drop damage increases rapidly. In the actual harvesting process of pod peppers, it is necessary to ensure that the ground drop damage is within an acceptable range, and that efforts are made to improve the feeding speed and picking efficiency. Therefore, the ground drop loss rate was selected to be smaller when the feeding speed of pod peppers was about 0.6 m/s. Du et al. optimized the device parameters by analyzing the effect of feeding speed on drum finger-snapping pepper harvesting [14]. However, the structure of the bionic comb finger harvesting device designed in this article differs significantly from that of the finger-snapping drum-type harvesting device, and further optimization of its structure is still required.

When the picking roller speed is 650 rpm and the pod pepper feeding speed is 0.6 m/s, the simulation results are shown in Figure 10c. In the range of 5~45 mm spacing between two picking rollers, the rate of ground drop loss of pod peppers increased with the increase in spacing. When the spacing is 5 mm, the drop loss rate is a minimum 4.92%; when the spacing is 45 mm, the loss rate is a maximum 19.54%. In summary, the main stem needs to be picked through the spacing of two picking rollers during the harvesting process of pod peppers. According to the measured parameters of the main stem size (the spacing between the two picking rollers should be larger than the diameter of the main stem), the loss rate of ground fall should be reduced as much as possible under the circumstances of ensuring normal picking. Therefore, a spacing of 15 mm was selected as the most suitable. Liu et al. designed an inclined double screw comb-finger-type pod pepper harvesting device that has a high picking rate but, due to the large gap between the two picking rollers, a high ground loss rate [21]. Therefore, by analyzing the effects of picking roller speed, pod pepper feeding speed, and the spacing between the two picking rollers on ground drop loss, the pod pepper picking device was optimized.

4.2. Analysis of Box–Behnken Response Surface Test Results

Based on the results of the Box–Behnken simulation test in Table 5, the quadratic linear regression equation (Equation (7)) for the rate of loss of pod peppers by ground fall was obtained by ANOVA and multiple regression analysis in Design-Except 13.0 software as follows:

$$\begin{array}{c}{\eta }_{s2}=258.504-0.88369X_1+149.61X_2+0.8321X_3-0.2065X_1{X}_2-0.00167X_1{X}_3\\ +0.925X_2{X}_3+0.000741X_1^2-11.175X_2^2-0.00387X_3^2\#\end{array}$$

(7)

where

• X₁: The speed of the picking rollers, rpm;
• X₂: The feeding speed of pod pepper, m/s;
• X₃: The spacing between the two picking rollers, mm.

According to the ANOVA results of the regression model in

Table 7. Analysis of variance of dropping loss rate η_s2 for ground pod peppers.

Source	Sum of Squares	DF	Mean Square	F-Value	p-Value
Model	142.34	9	15.82	42.38	<0.0001
X1	95.08	1	95.08	254.78	<0.0001
X2	20.10	1	20.10	53.85	0.0002
X3	6.88	1	6.88	18.44	0.0036
X1×2	4.26	1	4.26	11.43	0.0118
X1×3	0.6972	1	0.6972	1.87	0.2140
X2×3	0.8556	1	0.8556	2.29	0.1738
$X_1^2$	14.46	1	14.46	38.75	0.0004
$X_2^2$	0.0526	1	0.0526	0.1409	0.7185
$X_3^2$	0.0394	1	0.0394	0.1056	0.7547
Residual	2.61	7	0.3732
Lack of Fit	0.7454	3	0.2485	0.5324	0.6842
Pure Error	1.87	4	0.4667
Cor Total	144.95	16

Note: DF, Degree Freedom.

, the p-value of the simulation model for the ground drop loss of the pod pepper picking device is less than 0.0001, indicating that the model is highly significant and well-fitted. Meanwhile, the p-value of the simulation model’s misfit term is 0.6842, which is greater than 0.05, indicating that the misfit is not significant, and the model is well-fitted and not misfit. The coefficient of determination (R²) in the regression model is 0.9820, the adjusted R² is 0.9588, the predicted R² is 0.8976, and adeq precision is 21.4818, which indicated that the simulation model of the ground drop loss rate of pod peppers had high reliability. The model can be used to guide the design of pod pepper harvesters. The p-values of picking roller speed X₁, pod pepper feeding speed X₂, and two picking rollers spacing X₃ are all less than 0.01, indicating that these three factors have a highly significant effect on the ground drop loss rate of pod peppers, and the order of the effect is picking roller speed X₁, pod pepper feeding speed X₂, and two picking rollers spacing X₃, in descending order. In this regression model, the p-value for the interaction of X₁ and X₂ is 0.018 < 0.05, indicating that the interaction of these two factors has a significant effect on the rate of ground drop loss of pod peppers; the p-value of $X_1^2$ is 0.0004 < 0.01, indicating that it has a highly significant effect on the rate of ground drop loss of pod peppers; the p-values of the rest of the terms are all greater than 0.05, indicating that the effect was not significant.

The response surface of picking roller speed X₁ and pod pepper feeding speed X₂ is shown in Figure 11a. When the distance between the two picking rollers is taken as 15 mm and the picking roller speed is determined, the loss rate of ground pod peppers increases with the increase in feeding speed. The magnitude increases and then decreases with the increase in picking roller speed. When the feeding speed of pod peppers is determined, the ground drop loss rate of pod peppers shows a tendency to decrease and then increase with the increase in picking roller speed. The magnitude of change was greater than that of the pod pepper feeding speed, indicating that the picking roller speed played a dominant role in the interaction, which was consistent with the results of the ANOVA. The response surface of the pod pepper feeding speed X₂ and the spacing between the two picking rollers X₃ are shown in Figure 11b. When the picking roller speed was taken at 650 rpm, the rate of ground drop loss of pod pepper increased with the increase in the feeding speed of pod pepper and the spacing between the picking rollers. At a pod pepper feeding speed of 0.4 m/s and a spacing of 6 mm between the two picking rollers, the rate of ground loss of pod pepper was lesser. However, in the actual production of the pod pepper harvester, the machine harvesting efficiency and the planting agronomy of pod peppers need to be considered comprehensively. The response surface of picking roller speed X₁ and two picking roller spacings X₃ is shown in Figure 11c. When the feeding speed of pod peppers is 0.6 m/s and the picking roll speed is determined, the loss rate of pod peppers ground drop increases with the increase in spacing between the two picking rollers. When the spacing between the two picking rollers was determined, the ground drop loss rate of pod peppers reached its minimum when the picking roller speed was around 650 rpm. The response surface showed that the effect of picking roller speed on the ground drop loss of pod peppers was greater than that of two picking roller spacings.

Based on the above ANOVA results, in order to further optimize the operational performance of the pod pepper picking device, the minimum value of pod peppers ground drop loss rate was taken as the objective. Within the range of values of each influencing factor, multi-objective optimization and optimal operating parameters were determined using the following Equation (8). The optimization solver in Design-Expert 13.0 solved the regression Equation (8) optimally under this function, and the optimal combination of operating parameters of the picking device was obtained when the loss rate of pod peppers ground drop reached 3.526% as follows: The speed of the picking rollers was 680.41 rpm, the feeding speed of pod pepper was 0.5 m/s, and the spacing between two picking rollers was 12 mm.

$$\begin{array}{c}Minimize\left\{{\eta }_{s2}\left(X_1,X_2,X_3\right)\right\}\end{array}$$

(8)

where

$${600 \mathrm{rpm}\le X}_1\le 700 \mathrm{rpm};{0.5 \mathrm{m}/\mathrm{s}\le X}_2\le 0.7 \mathrm{m}/\mathrm{s};{10 \mathrm{m}\mathrm{m}\le X}_3\le 20 \mathrm{m}\mathrm{m}$$

4.3. Analysis of the Results of Field Trials on Pod Pepper

The 4JF-1.1 small hand-held pod pepper harvester performed stably during the test, and the picking device, conveying device, chassis traveling device, and storage device all worked normally. The net picking rate of pod peppers was high and met the harvesting demand. According to the above test scheme and evaluation indexes, the net harvesting rate and ground fall loss of pod peppers harvested using the bionic comb finger picking device were counted, and the performance indexes were obtained as shown in

Table 8. Experimental performance indicators of the bionic comb-type harvesting device.

Item	m1 (g)	m2 (g)	m3 (g)	η1 (%)	η2 (%)
1	513.7	15.73	41.86	96.94	8.149
2	513.7	24.68	39.17	95.20	7.625
3	513.7	17.35	48.26	96.62	9.395
Average	513.7	19.25	43.10	96.25	8.390
Standard Deviation	0	3.89	3.81	0.76	0.742

Note: η₁, pod pepper picking rate; η₂, rate of loss from ground fall of pod peppers; m₁, mean value of quality of pod peppers in the sampling area; m₂, quality of unpicked pod peppers in the sampling area; m₃, mass of fallen pod peppers on the ground in the sampling area.

.

According to Table 8, the average ground fall loss rate of the bionic comb finger picking device during pod pepper harvesting was determined to be 8.39%. However, through the ground drop loss simulation test, it was found that the ground drop loss rate of the picking device for pod peppers was 3.526% under the optimal combination of operating parameters. There is a certain relative error between the simulation test results and the ground drop loss rate of pod peppers in the actual harvesting test. In practice, the harvesting of pod peppers is affected by a variety of factors, including plant height, fruit shape, and harvester operation. Pod peppers not only come into contact with comb fingers, but also with stalks, pepper leaves, weeds, etc. Meanwhile, other factors such as machine vibration, tilting, and bumps during operation also increase the ground drop loss rate. Since the simulation test excludes external interference and assesses ground drop loss by generating simulated pod pepper particles during picking, there will be relative errors. Overall, the relative error between the simulation test and the real harvesting results of the ground drop loss of the pod pepper picking device is relatively small, which can be used as the basis for the design of the pod pepper harvester.

5. Conclusions

In this study, through the analysis of hand-picking pod peppers, we extracted the structural morphology of fingers, designed a bionic comb finger pod peppers picking device, and developed a high-efficiency and low-loss hand-held pod peppers harvester. Firstly, a discrete meta-model of pod peppers was established, and a simulation platform for the ground drop loss of the pod peppers picking device was built to explore the motion trajectory and speed change of the discrete meta-model of pod peppers. At the same time, the effects of the picking roller speed, the feeding speed of pod peppers and the distance between two picking rollers on the ground drop loss of pod peppers were analyzed. In order to evaluate the loss rate of pod peppers ground drop, a one-factor simulation test was conducted to derive the effects of each factor on the loss rate and to preliminary determine the optimal operating parameters of the picking device. Based on the results of the single-factor analysis, the operating parameters of the picking device were optimized using the Box–Behnken response surface test, and the optimal combination of operating parameters was obtained: a picking roller speed of 680.41 rpm, a pod pepper feeding speed of 0.5 m/s, and a spacing of 12 mm between the two picking rollers. Under these conditions, the pod peppers ground drop loss rate was 3.526%. Through the harvesting test, it was learnt that the net harvesting rate of the pod pepper harvester was 96.25%, and the ground drop loss rate was 8.39%. Although there is a certain relative error between the harvesting test and the simulation test, the error is small. Therefore, the discrete element simulation test results of the harvesting device were verified to be reliable.