Tomato Pedicel Physical Characterization for Fruit-Pedicel Separation Tomato Harvesting Robot

Abstract

To solve the problem of the lack of physical properties of pedicels and the changing pattern for designing the end-effector of tomato harvesting robot and different harvesting modes, research was conducted on the physical properties of tomato pedicels and their change patterns. Using a Universal TA texture analyzer, tensile, three-point bending, and shearing tests were performed on tomato pedicels in the early firm-ripening stage. The tomato variety used was Syngenta Spectrum, cultivated seasonally with two crops per year. Spring crop tomatoes were used in this study. The experimental results provide a theoretical basis for designing tomato harvesting robots across three harvesting modes. Tensile tests measured the pull-off force and tensile strength of the abscission zone with varying diameters. These results are crucial for designing robots using a tensile harvesting mode. The location of the tomato pedicel significantly affects the shearing force. A one-way test was conducted on the shearing part. The results showed that the shearing force and energy required for the proximal pedicel are significantly greater than for the distal pedicel. To reduce the shearing force and energy needed by the end-effector’s shearing mechanism on distal pedicels, a response surface test was conducted. Three factors were examined: shearing speed, angle, and distal pedicel diameter. Design–Expert software optimized these factors to minimize shearing energy and force, leading to the best shearing parameters for different distal pedicel diameters. From the three-point bending tests, the average maximum bending breaking force, bending modulus, and bending strength of the tomato abscission zone were determined. These findings offer a theoretical basis for designing tomato harvesting robots with a bending-type harvesting mode.

1. Introduction

Globally, China is one of the regions with the largest area of tomato cultivation and a wide variety of tomato species. Tomatoes have a sweet taste and rich nutritional value, especially in vitamin C, vitamin A, potassium, and lycopene. They offer many benefits to human health and are loved by the masses. Tomatoes in China’s agricultural production occupy an important position, with planting mainly concentrated in Shandong, Henan, Xinjiang, Inner Mongolia, and other places [1,2,3]. The tomato plant grows luxuriantly, and a single seedling often bears multiple tomato fruits. The ripeness of the tomato fruits on the same plant varies, requiring careful manual identification and harvesting. This harvesting process is highly labor-intensive and requires a high level of physical strength and eyesight on the part of the worker [4,5]. In greenhouses, the environment is more controllable than outdoor cultivation, so agricultural robots have become a new research direction in robotics. At this stage, research for the tomato harvesting robots’ end-effector mainly focuses on grasping, specifically the mechanical characteristics of the contact between the tomato and the end-effector fingers. Researchers overlooked the crucial step in harvesting: the separation of tomato pedicels. They also lacked studies on the biomechanical properties of tomato pedicels. As a result, the machines developed were unable to achieve high-quality and efficient harvesting operations. At present, the harvesting modes of tomato harvesting robots are divided into three kinds: stretching, shearing, and bending [6]. Understanding the physical properties of the pedicel is essential for designing the end-effector of a tomato harvesting robot. This knowledge provides key parameters and a design basis for harvesting equipment. It also offers a theoretical foundation for different fruit-pedicel separation harvesting modes in tomato harvesting robots.

Many researchers and scholars have conducted mechanical tests on stems or branches of various fruits and vegetables. Gangwar et al. [7] developed a multiscale approach to predict macroscale stiffness and strength properties from the hierarchical microstructure of plant stalk materials. The model was validated by performing a series of bending tests on a sample of oat stalks. Xie et al. [8] utilized a universal testing machine in conjunction with the YYD-1 stem strength tester to test the biomechanical properties of stalks in the clamp section of oilseed rape. The results concluded that the stems of rapeseed shoots belong to anisotropic material; the modulus of elasticity and bending strength of the stems are smaller than the modulus of elasticity and bending strength of the branches and trunks. The experimental results can provide basic technical data support for mechanized harvesting of rapeseed shoots and parameter selection. Réquilé et al. [9] investigated the bending stiffness of hemp and flax by performing three-point bending tests and using microstructural analysis of the stalks to determine the fiber distribution and geometry to obtain the longitudinal modulus. Kumar et al. [10] determined the tensile and shear strength of wheat straw to design a low-cost straw combine and analyzed the effect of moisture in straw on tensile and shear strength. He et al. [11] carried out a one-way and response surface test to reduce sesame stems’ shearing power consumption, targeting four factors: shearing speed, shearing edge angle, shearing angle, and shearing part. The results of the study showed that the lowest shearing power consumption was achieved at a blade angle of 20°, a shearing speed of 8.3 mm/s, a shearing angle of 30°, and a shearing position of 200 mm from the root. Gao et al. [12] designed an SHQG-I type facility vegetable harvesting shearing test platform to achieve the best harvesting performance. The test platform can be adjusted to the effective working range of each factor. Using the response surface test method, they optimized the comprehensive influence factors on the shearing of the knife shearing the stem of the vegetable in the harvesting process. The theoretical shearing force under the optimal parameters was finally determined to be 17.9 N, and the actual shearing force was 17.4 N, which ensured a 100% success rate of harvesting. Gao [13] investigated the mechanical shear performance of tomato stems using a one-way test with leaf angle, water content, and sampling range as experimental factors and peak shearing force as evaluation index. In addition, other scholars have researched crop cutting factors, such as wild chrysanthemum stem cutting force characteristics, millet stem shearing test measurements, and tea shearing mechanical properties [14,15,16,17]. In the case of fruit crops, Rajendran et al. [18] chose different shear angles for strawberry stalks for shear tests: the study of stalk shear forces determined the critical conditions for shearing strawberry stalks to be a 30° orientation, a.6° blade wedge angle, and a 15 N force. Wang et al. [19] developed a new 16-channel electronic glove for data acquisition, which they then wore to randomly harvest 60 tomatoes in a greenhouse. The results showed that the total grasping pressure at the moment of fruit-stem separation primarily depended on the intersection angle between the first and second stems of a fruit. In contrast, at the moment of fruit release, the total grasping pressure primarily depended on the fruit size. Ma et al. [20] conducted tensile and bending tests on tomato potting stems at the planting stage using a DF-9000 dynamic and static electronic universal material testing machine, and compression performance tests on tomato potting stems at the planting stage.

With the development of harvesting robots, domestic and international research on end-effectors tends to diversify. Guo et al. [21] designed a flexible underactuated end-effector for tomato harvesting and sorting, based on the physical properties of tomatoes. Byun et al. [22] proposed a harvesting end-effector designed to minimize the length of the stem remaining on the fruit by using a hook for stem-cutting. The end-effector uses a hook mechanism to pull the stem of the target fruit towards the cutter. This reduces the length of the remaining stem, which can otherwise cause damage to other fruits during packaging and transportation. Zhang et al. [23] designed a finger with sinusoidal characteristics by studying how the human hand grips a cherry tomato. The end-effector was designed to separate the fruit from the stalk, using the linear motion of a constraint part and a rotating gripper.

At present, most of the measurements of the physical properties of tomato tissues are centered around fruits and potting seedlings, with fewer studies on the physical properties of tomato pedicels. To better adapt to the tomato mechanized orderly harvesting and improve the harvesting success rate, as shown in Figure 1, the objective of this study was to comprehensively determine the physical properties of tomato pedicels. To this end, three tests were employed: (1)Through the abscission zone tensile test, to obtain the abscission zone tensile force and tensile strength, and to analyze the correlation between the tensile force and the diameter of the abscission zone. This provides a theoretical basis for the design of tomato harvesting robots in tensile harvesting mode. (2) Find the optimal shearing part by one-way testing of the shearing part. The shearing response surface test helps determine the optimal shearing angle and shearing speed for the distal pedicel. These findings can guide the design and parameter optimization of the shear mechanism in the end-effector. (3) To obtain the bending breaking force and bending strength of the pedicel through the three-point bending test, and to obtain the correlation between the breaking force and the abscission zone diameter. This provides a theoretical basis for the design of tomato harvesting robots in bending harvesting mode.

2. Materials and Methods

2.1. Tomato Pedicel Sample Preparation

The tomatoes selected for this study were grown in a greenhouse at Yinong Agricultural Base, Changle District, Fuzhou City, Fujian Province. The base is located at longitude 119.47° E, latitude 25.91° N, and altitude four meters above sea level. The tomato variety used was Syngenta Spectrum, cultivated seasonally with two crops per year, divided into spring and winter crops. Tomato pedicels were measured with calipers and manually harvested using scissors. The samples were selected based on the required diameters of the abscission zone, distal pedicel, and proximal pedicel. These tomatoes were harvested at the early firm-ripening stage. After harvesting, they were stored at 5 °C in preservation bags. Each tomato pedicel sample was assigned a unique number. Measurements were taken for abscission zone diameter, distal pedicel diameter, distal pedicel length, proximal pedicel diameter, proximal pedicel length, and weight. The tool used was digital vernier calipers (Mitutoyo Precision Gauge (Shanghai) Co., Ltd., Shanghai, China) (accuracy: 0.01 mm) and sensitive scales (Shanghai LICHEN Instrument Technology Co., Ltd., Shanghai, China) (accuracy: 0.001 g). As shown in Figure 2, the tomato pedicel is comprised of three distinct parts. The proximal pedicel is connected to the peduncle. The second part is the abscission zone, a relatively thick section of the distal pedicel. The third part is the distal pedicel, which connects the abscission zone to the calyx. The distal pedicel and proximal pedicel typically exhibit a curved angle at the abscission zone [24].

2.2. Texture Analyzer

The Universal TA Texture Analyzer (Shanghai Tengba Instrument Technology Co., Ltd., Shanghai, China) was used for the tomato pedicel physical properties test. The maximum test force is 500 N, with an accuracy of 0.0001 g. The maximum test speed is 50 mm/s, with an accuracy of 0.0001 mm. The test travel distance was 0–420 mm, with an accuracy of 0.0001 mm. The companion software for the Universal TA texture analyzer is QCTech3_A2 23.40.00.06.

2.3. Abscission Zone Tensile Test

The majority of the tomato abscission zone diameters were found to be distributed within the range of 5−8 mm, so eight samples each of tomato pedicels with abscission zone diameters of 5−6 mm, 6−7 mm, and 7−8 mm were selected. The pedicel samples were secured to the tensile probe and base. To minimize the vibration and instability of the instrument, a loading rate of 1 mm/s was set, and the tensile force was gradually applied as shown in Figure 3. The maximum tensile force at breakage of the abscission zones was recorded, and the tensile strength Equation (1) was calculated.

$${\sigma }_t={\displaystyle \frac{F_{t\max }}{A_t}}={\displaystyle \frac{4F_{t\max }}{\pi D^2}}$$

(1)

where the ${\sigma }_t$ is the Tensile strength of the pedicel in MPa; $F_{t\max }$ is the maximum tensile force in the test in N; $A_t$ is the cross-sectional area of the abscission zone in mm²; $D$ is the abscission zone cross-section diameter in mm.

2.4. Tomato Distal Pedicel Shearing Test

2.4.1. One-Way Test for the Shearing Part

The growth characteristics of the pedicel result in differing shearing performances across its various parts. During the fieldwork conducted on the greenhouse, it was observed that the majority of the tomato distal and proximal pedicel diameters were distributed between 3 and 5 mm. Consequently, twenty samples each of tomato pedicels with a distal pedicel diameter of 4 mm and a proximal pedicel diameter of 4 mm were selected for a one-way test. The shearing angle was set at 90° and the shearing speed at 1 mm/s, as shown in Figure 4. The shearing force and the shearing energy were recorded during the test.

2.4.2. Response Surface Shearing Test Design

One of the most significant variables influencing the shearing process of a tool is the shearing speed. If the shearing speed is low, the shearing force will be excessive, resulting in an uneven shear section. Conversely, if the shearing speed is high, the average shearing force is reduced. However, the instantaneous impact of the shearing force becomes significant. This results in unstable tool kinetics and causes the tomato pedicels to tilt due to the impact. Following the experimental determination of the optimal range of shearing speeds (1–5 mm/s), this value was selected for further investigation. Plant stalks are anisotropic materials, exhibiting different shearing resistances depending on the direction of shearing. A certain shearing angle can reduce the shearing force of the sheared pedicel, but the gradual increase in the shearing angle will lead to the shearing section and the horizontal plane at a certain angle, and the phenomenon of fiber tearing at the fracture, the shearing force will gradually increase [11]. During the previous one-way test, it was found that the tool shearing success rate was low at 0–30°, so the shearing angle was selected to be 30–90°, as shown in Figure 5. In the previous one-way test for the shearing part conducted in the laboratory, it was found that the diameter of the distal pedicel directly affects the shearing force and shearing energy during shearing. During the field visit to the farm, it was found that most of the tomato distal pedicel diameters were distributed between 3−5 mm, so the range of distal pedicel diameters was selected: 3–5 mm.

Tomato pedicels were selected to have a distal pedicel diameter of 3–5 mm, and the pedicel samples were placed on different angles of shear bases. Subsequently, a gradual increase in shearing force was applied to record the maximum shearing force value of the distal pedicels at the time of fracture. The compression mode was selected as the mode of the texture analyzer, and the test stop judgment was set to 10% stop.

The response surface method was chosen to optimize the analysis of the factors influencing harvest shear. The response surface method is more suitable for continuous factor variables, producing more comprehensive and reliable results, and leading to more accurate test results [25,26,27,28,29]. Based on the selected factors and the range of values, the experimental design was carried out using the central composite bounded design in the central composite design. The number of experimental factors m = 3, the number of replicated trials at the center of the 0 level m₀ = 5, the value of the asterisk arm α was taken as 2, and the central composite design factor coding table was calculated, as shown in

Table 1. Factor level coding table.

Level	Factor
	A/(mm/s)	B/(°)	C/(mm)
−2	1	30	3
−1	2	45	3.5
0	3	60	4
1	4	75	4.5
2	5	90	5

Note: A is the shearing speed in mm/s; B is the shearing angle in °; C is the diameter of the distal pedicel in mm, the same as below.

.

2.5. Pedicel Three-Point Bending Test

The majority of tomato abscission zone diameters fall within the range of 5–8 mm, so five pedicels each with abscission zone diameters of 5–6 mm,6–7 mm, and 7–8 mm were selected. The pedicel samples were placed on the three-point bending platform of the texture analyzer. A gradually increasing bending force was applied until the pedicel broke, and the value of the bending force at breakage was recorded. The loading speed was set to 1 mm/s, as shown in Figure 6.

Before the bending test, adjust the spacing between the left and right support fixtures, and set the three-point bending span to 12 mm to stabilize the placement of the pedicel and break the abscission zone. As shown in Figure 6, the pedicel is placed horizontally in the center of the three-point bending lower support, and then the arch bending indenter is mounted on the texture analyzer and the center of the base table, located in the middle of the abscission zone. This placement ensures that the pedicel is under vertical pressure all the way through the bending process. The test ends when the pedicel yields significantly and breaks after the start of the test.

The computer automatically collects data to obtain the maximum bending breaking force and displacement-bending force curve, as shown in Figure 7. From these data, the bending elastic modulus and flexural strength can be calculated according to Equations (2) and (3), respectively.

$$E_b={\displaystyle \frac{F_b{y}^3}{48\Delta H{I}_z}}={\displaystyle \frac{4F_b{y}^3}{3\mathsf{\pi }{D_a}^4\Delta H}}$$

(2)

$${\sigma }_b={\displaystyle \frac{M_{Max}}{W_z}}={\displaystyle \frac{8F_{b}y}{\mathsf{\pi }{D_a}^3}}$$

(3)

where the $E_b$ is the bending modulus in MPa; $\Delta H$ is the maximum bending distance in mm; $I_z$ is the moment of inertia of the fracture section on the neutral axis in mm⁴; $D_a$ denotes the Section diameter before bending with the unit of mm; $F_b$ is the maximum bending breaking force in N; $y$ is the span in mm; ${\sigma }_b$ is the bending strength in MPa.

2.6. Determination of Pedicel Moisture Content

Each tomato pedicel sample in the four tests described above was subjected to moisture content determination. The moisture content of tomato pedicels is about 70–80% when measuring the moisture content, the pedicel samples were weighed in an electronic balance, and the pre-drying mass $m_1$ was recorded sequentially. The pedicel samples were put into the drying oven at 105 °C to dry and weighed every 1 h until the mass was no longer reduced, and the current mass $m_2$ was recorded. Then, the formula for calculating the moisture content of the culm was as follows:

$$W={\displaystyle \frac{m_1-m_2}{m_1}}\times 100\%$$

(4)

where the $W$ denotes moisture content with the unit of %; $m_1$ denotes the pedicel raw mass with the unit of g; $m_2$ denotes the pedicel mass after drying with the unit of g.

2.7. Statistical Analysis

To conduct the statistical analysis, the average value of the measured data was obtained. Variance analysis was performed using SPSS 27.0 for Windows at a 5% significance level by using Duncan’s Multiple Range Test. Linear fitting equations for the abscission zone bending breaking force and abscission zone diameter were obtained by linear curve fitting using Origin 8.0. To study the relationships between the physical and mechanical parameters of the pedicel, the Pearson correlation matrix method was used.

3. Results and Discussion

3.1. Abscission Zone Tensile Test

The abscission zone tensile curve is shown in Figure 8. A total of 24 pedicel samples from three groups were tested using the abscission zone tensile test. As shown in

Table 2. Results of abscission zone tensile tests.

Abscission Zone Diameter/(mm)	Average Abscission Zone Diameter/(mm)	Average Moisture Content/(%)	Average Abscission Zone Pull-Off Force/(N)	Tensile Strength/(MPa)
5–6	5.52 ± 0.221	76.376 ± 3.013	40.262 ± 12.437 b	1.691 ± 0.554 a
6–7	6.27 ± 0.207	77.212 ± 1.477	44.781 ± 15.156 b	1.441 ± 0.427 a
7–8	7.75 ± 0.451	76.348 ± 0.461	72.003 ± 23.401 a	1.534 ± 0.511 a

Note: The Duncan multiple range test indicates that values with different letters are significantly different at a 5% probability level.

and Figure 9, the average diameter of the 5–6 mm group was 5.52 mm. The average pull-off force for this group was 40.262 N, and the tensile strength was 1.691 MPa. The average diameter of the 6–7 mm group was 6.27 mm, the average abscission zone tensile breaking force was 44.781 N, and the tensile strength was 1.441 MPa. The average diameter of the 7–8 mm group was 7.75 mm, and the average abscission zone pull-off force was 72.003 N, and the tensile strength was 1.534 MPa. The average abscission zone pull-off force of the 6–7 mm group was 11.22% higher than that of the 5–6 mm group. The average pull-off force of the 7–8 mm group was 78.83% greater than that of the 5–6 mm group and 60.79% greater than that of the 6–7 mm group.

Note: The Duncan multiple range test indicates that values with different letters are significantly different at a 5% probability level.

From Figure 9, it is evident that the average abscission zone tensile breaking force is positively correlated with the diameter. As the diameter of the tomato abscission zone increases, the average tensile breaking force also increases significantly. While the change in tensile strength is not the same, the tensile strength appears to decline first and then rise as the diameter increases, but the magnitude of the change is not large.

The tensile strength of pedicels obtained in this paper is close to the tensile strength of pedicels at the pink stage in the literature [30], which is 1.561 ± 0.187 MPa. The tensile force of pedicels with diameters of 4−6 mm for the “Jiali 14” tomato in the literature [31] is 51.9 N, and that of the pedicels with diameters of 6−8 mm is 65.8 N. There are many reasons for the inconsistency of these test results because the mechanical determination of biomass materials depends on the variety, maturity, and microscopic defects of the samples requested [32].

As can be seen from

Table 3. Correlation table for Abscission zone tensile pull-off force.

	Abscission Zone Diameter	Distal Pedicel Diameter	Proximal Pedicel Diameter	Moisture Content	Abscission Zone Pull-Off Force	Tensile Strength
Abscission zone diameter	1	0.944 **	0.930 **	−0.022	0.640 **	−0.127
Distal pedicel diameter		1	0.912 **	−0.046	0.708 **	0.045
Proximal pedicel diameter			1	−0.125	0.691 **	0.016
Moisture content				1	−0.133	−0.233
Abscission zone pull-off force					1	0.648 **
Tensile strength						1

Note: ** Correlation is significant at a 1% probability level (p < 0.01). * Correlation is significant at a 5% probability level (p < 0.05).

, the correlation between abscission zone tensile strength and abscission zone diameter, distal pedicel diameter, and proximal pedicel diameter was highly significant (p < 0.01). There was no correlation with moisture content. There was no correlation between tensile strength and abscission zone diameter, distal pedicel diameter, and proximal pedicel diameter.

When harvesting tomatoes using a tensile method, the robot’s end-effector tightens around the tomato and applies a force along the pedicel through the robotic arm. This action pulls the abscission zone of the pedicel, with the tensile force at the abscission zone being equal to the tensile force applied to the tomato. Zhang et al. [30], referring to Liu et al. [33], conducted a study on the tomato harvesting process in which different harvesting modes of tomato were mechanically analyzed, and the study showed that the probability of damage to tomato is highest when the clamping force on tomato is high during tomato stretch harvesting. The results of this part of the tensile test can provide a theoretical reference for the tensile-type tomato harvesting robot.

3.2. Tomato Distal Pedicel Shearing Test

3.2.1. One-Way Test for the Shearing Part

From

Table 4. Comparison of results of shearing part.

Sample	Distal Pedicel Diameter/(mm)	Proximal Pedicel Diameter/(mm)	Moisture Content/%	Shearing Force/N	Shearing Energy/J
Proximal pedicel	4.91 ± 0.693	4.11 ± 0.402	76.465 ± 2.514	62.054 ± 15.727 a	0.256 ± 0.152 a
Distal pedicel	4.06 ± 0.077	3.33 ± 0.389	76.790 ± 1.338	33.241 ± 7.962 b	0.117 ± 0.041 b

Note: The Duncan multiple range test indicates that values with different letters are significantly different at a 5% probability level.

, it can be concluded that the proximal pedicel shearing force is 85.32% greater than the distal pedicel shearing force. Additionally, the proximal pedicel shearing energy is 121.37% greater than the distal pedicel shearing energy.

The proximal pedicel is closer to the peduncle and has a higher degree of lignification compared to the distal pedicel. Therefore, when the harvesting robot selects the harvesting site, it is more efficient to cut the distal pedicel as the shearing part, which helps save energy. Byun et al. [22] suggested the need to have reduced the length of the remaining stalks, which can otherwise cause damage to other fruits during packaging and transportation. Therefore, selecting the distal pedicel as shearing part not only reduces shear energy, but also ensures that the pedicel is sheared as completely as possible. So that the tomato pedicels do not cause damage to other fruits during transportation of post-harvest tomatoes.

The shear curve for the proximal pedicel, shown in Figure 10, displays only one peak. In contrast, the shear curve for the distal pedicel has a second peak. This difference is likely due to the varying degrees of lignification between the proximal and distal pedicels.

3.2.2. Response Surface Shearing Test

Response surface tests were carried out according to

Table 5. Shearing response surface test results.

Test Number	Factor			Shearing Force Y/(N)	Shearing Energy Z/(J)
	Shearing Speed (A)/(mm/s)	Shearing Angle (B)/(°)	Distal Pedicel C/(mm)
1	−1	−1	−1	26.529	0.092
2	1	−1	−1	37.798	0.164
3	−1	1	−1	28.098	0.086
4	1	1	−1	34.061	0.081
5	−1	−1	1	33.130	0.159
6	1	−1	1	34.375	0.197
7	−1	1	1	39.289	0.108
8	1	1	1	33.581	0.091
9	−2	0	0	38.083	0.138
10	2	0	0	47.448	0.164
11	0	−2	0	31.678	0.223
12	0	2	0	34.248	0.099
13	0	0	−2	22.136	0.092
14	0	0	2	31.855	0.110
15	0	0	0	31.580	0.115
16	0	0	0	33.326	0.106
17	0	0	0	32.306	0.132
18	0	0	0	31.982	0.109
19	0	0	0	31.011	0.109

, and the results were analyzed by multifactor ANOVA and multiple regression using Design-Expert (Version: 13.0.1.0 64-bit), and the coded regression equations between shearing force (Y) and shearing speed (A), shearing angle (B), and distal pedicel diameter (C) are as follows:

$$Y=31.97+1.97A+0.5211B+2.08C-1.53AB-2.71AC+0.9416BC+2.65A^2+0.2034B^2-1.29C^2$$

(5)

From the ANOVA table of the experimental data, as shown in

Table 6. ANOVA table for regression model of shearing force.

Source	Sum of Squares	df	Mean Square	F-Value	p-Value
Model	486.87	9	54.1	59.37	<0.0001**
A-Shearing speed	62.01	1	62.01	68.06	<0.0001**
B-Shearing angle	4.34	1	4.34	4.77	0.0568
C-Distal pedicel diameter	69.42	1	69.42	76.19	<0.0001**
AB	18.79	1	18.79	20.62	0.0014**
AC	58.83	1	58.83	64.57	<0.0001**
BC	7.09	1	7.09	7.79	0.021*
A²	166.8	1	166.8	183.08	<0.0001**
B²	0.9802	1	0.9802	1.08	0.3267
C²	39.31	1	39.31	43.14	<0.0001**
Residual	8.2	9	0.9111
Lack of Fit	5.2	5	1.04	1.39	0.3865
Pure Error	3	4	0.7496
Cor Total	495.07	18

Note: **Correlation is significant at a 1% probability level (p < 0.01). * Correlation is significant at a 5% probability level (p < 0.05).

, the F-value of the model is 59.37, p < 0.0001, indicating that the model is highly statistically significant. Meanwhile, the F-value of the lack of fit is 1.39, p = 0.3865 > 0.05, indicating that the equation is not significantly out-of-fit, and has a good fit.

The ANOVA table showed that the effect of each factor on the magnitude of shearing force was in the order of distal pedicel diameter (C) > shearing speed (A) > shearing angle (B). The effect of shearing speed (A) and distal pedicel diameter (C) on the shearing force was highly significant. While the impact of shearing angle (B) on the shearing force did not reach significance, but the p-value was 0.0568, which is close to the significant level. There was a highly significant interaction for AC (shearing speed (A), distal pedicel diameter (C)) as well as AB (shearing speed (A), shearing angle (B)). There was a significant interaction between shearing angle and distal pedicel diameter BC. These three items had a substantial effect on shearing force. The correlation coefficient R² of the model is 0.9834, which indicates that the model is reliable, and the coefficient of variation is 2.87%, which indicates that the model is accurate. In this model, AC, AB, and BC had significant effects on the equation, and there was a significant interaction between the factors, so the interaction effects need to be optimized by the software. The effect of the interaction between factors on shearing force is shown in Figure 11.

The interaction of shearing angle and shearing speed on shearing force at distal pedicel diameter of 4 mm is shown in Figure 11a. When the shearing angle is fixed, an increase in shearing speed results in a decrease in shearing force, which then increases.; when the shearing speed is fixed, an increase in shearing angle results in a gradual increase in shearing force, with the two variables exhibiting an approximate linear relationship.

When the shearing speed is 3 mm/s, the interaction of distal pedicel diameter and shearing angle on the shearing force is shown in Figure 11b. When the distal pedicel diameter is fixed, the shearing force decreases gradually as the shearing angle increases. This relationship can be approximated by a linear function; when the shearing angle is fixed, the shearing force increases significantly as the diameter increases, and then the increase rate slows down. The F-value of the diameter of the distal pedicel was 76.19, so the effect of the diameter of the distal pedicel on the shearing force was more significant than the shearing angle.

At a shearing angle of 60°, the interaction of distal pedicel diameter and shearing speed on shearing force is shown in Figure 11c. When the distal pedicel diameter is fixed, the shearing force gradually increases with the increase of shearing speed; when the shearing speed is fixed, with the increase of distal pedicel diameter the shearing force increases significantly.

The response surface tests were conducted in accordance with the specifications outlined in Table 5. The resulting data were subjected to a multifactor analysis of variance (ANOVA) and multiple regression analysis using the Design–Expert software(Version: 13.0.1.0 64-bit). The coded regression equations between shearing energy (Z) and shearing speed (A), shearing angle (B), and distal pedicel diameter (C) are as follows:

$$Z=0.1121+0.0087A-0.0309B+0.00105C-0.0165AB-0.0057AC+0.0085BC+0.0084A^2+0.0109B^2-0.0041C^2$$

(6)

The ANOVA of the experimental data (

Table 7. ANOVA table for regression model of shearing energy.

Source	Sum of Squares	df	Mean Square	F-Value	p-Value
Model	0.0266	9	0.003	17.16	0.0001 **
A-Shearing speed	0.0012	1	0.0012	7.11	0.0258 *
B-Shearing angle	0.0153	1	0.0153	88.51	< 0.0001 **
C-Distal pedicel diameter	0.0018	1	0.0018	10.24	0.0108 *
AB	0.0022	1	0.0022	12.64	0.0062 **
AC	0.0003	1	0.0003	1.53	0.2467
BC	0.0006	1	0.0006	3.35	0.1003
A²	0.0017	1	0.0017	9.76	0.0123 *
B²	0.0028	1	0.0028	16.4	0.0029 **
C²	0.0004	1	0.0004	2.28	0.1652
Residual	0.0016	9	0.0002
Lack of Fit	0.0011	5	0.0002	2.03	0.2566
Pure Error	0.0004	4	0.0001
Cor Total	0.0282	18

Note: ** Correlation is significant at a 1% probability level (p < 0.01). * Correlation is significant at a 5% probability level (p < 0.05).

) shows that the F-value of the model is 17.16, p = 0.0001 < 0.01, which indicates that the model is significant; the F-value of the lack of fit is 2.03, p = 0.2566 > 0.05, which indicates that the equation is not significantly out-of-fit, and has a good fit.

The ANOVA table shows that the effect of each factor on the size of Shearing energy Z is in the order of shearing angle (B) > distal pedicel diameter (C) > shearing speed (A). The effect of shearing angle (B) on shearing energy is highly significant, and the impact of shearing speed (A) and distal pedicel diameter (C) on shearing energy (Z) is significant. There was a significant interaction between shearing speed and shearing angle AB, which had a large effect on shearing energy. There was a significant interaction between the factors, so the interaction influences need to be optimized by software. The correlation coefficient R² was 0.9449 and the coefficient of variation was 10.50%, indicating that the model was reliable. The effect of the interaction between factors on shearing energy is shown in Figure 12.

The interaction of shearing angle and shearing speed on shearing energy at a distal pedicel diameter of 4 mm is shown in Figure 12a. In the context of a fixed shearing angle, an increase in shearing speed results in a gradual rise in shearing energy; when the shearing speed is held constant, an increase in shearing angle leads to a gradual decline in shearing energy.

The interaction of distal pedicel diameter and shearing angle on shearing energy at a shear rate of 3 mm/s is shown in Figure 12b, which shows that when the distal pedicel diameter is fixed, the shearing energy decreases significantly with the increase of shearing angle, and the two have an approximately linear relationship; when the shearing angle is fixed, the shearing energy increases gradually with the increase of diameter. The F-value of distal pedicel diameter is 10.24, so the effect of shearing angle on shear energy is more significant than distal pedicel diameter.

When the shearing angle is 60°, the interaction of distal pedicel diameter and shearing speed on shearing energy is shown in Figure 12c. When the premise of distal pedicel diameter is fixed, the shearing energy gradually increases with the increase of shear speed; when the premise of shearing speed is fixed, with the increase of distal pedicel diameter shearing energy increases. The effect of distal pedicel diameter on shear energy is more significant than shearing speed.

In this response surface test, the distal pedicel diameter of the tomato is not a parameter to be optimized, which is randomly distributed in the farm. Using the test optimization function of Design–Expert software, the objectives were to minimize shearing energy and shearing force. Other factors were optimized within the range of test levels, and the optimized values were then normalized. The optimal shearing parameters and the shearing test results under the conditions of different distal pedicel diameters were finally obtained as shown in

Table 8. Collection of optimal shearing parameters.

Distal Pedicel Diameter/(mm)	Shearing Speed (A)/(mm/s)	Shearing Angle (B)/(°)	Shearing Force (Y)/(N)	Shearing Energy (Z)/(J)
3	2.6	68.5	19.556	0.067
3.5	2.2	62.8	26.715	0.090
4	3.2	77.9	32.957	0.089
4.5	3.5	77.6	34.163	0.081
5	3.9	71.0	30.881	0.081

.

While collecting the samples, it was found that most of the tomato distal pedicels were between 3.5–4 mm in diameter, so 3.5 mm as well as 4 mm were selected for both diameters to experimentally validate the response surface model. Each set of experiments was repeated four times, as shown in

Table 9. Table of validation test results.

	Sample	Abscission Zone Diameter/(mm)	Distal Pedicel Diameter/(mm)	Proximal Pedicel Diameter/(mm)	Moisture Content/(%)	Shearing Force/(N)	Shearing Energy/(J)
3.5 mm	1	4.45	3.51	3.29	75.776	30.854	0.099
	2	5.53	3.57	3.17	77.947	26.500	0.056
	3	5.56	3.53	3.29	75.458	29.834	0.075
	4	4.30	3.52	3.18	79.923	28.108	0.094
	Mean	4.96	3.53	3.23	77.276	28.824	0.081
4 mm	1	6.31	4.09	3.78	79.464	31.58	0.115
	2	6.02	4.06	3.76	79.024	29.854	0.091
	3	7.06	4.15	3.59	78.611	34.738	0.090
	4	5.53	4.04	3.23	78.305	34.875	0.079
	Mean	6.23	4.09	3.59	78.851	32.762	0.094

.

As shown in Table 9, the average shearing force obtained from the 3.5 mm diameter validation test is 28.824 N, which differs by 7.89% from the model prediction. The shearing energy is 0.081 J, which is 10% different from the model prediction. The shear effect is good, the fracture is flush, and the success rate of the shearing test reaches 100%, which can meet the requirements. The average value of shear force obtained from the 4 mm diameter validation test is 32.762 N, which is 0.59% different from the model prediction, and the shearing energy is 0.094 J, which is 5.62% different from the model prediction. The shear effect is good, the fracture is flush, and the success rate of the shearing test reaches 100%, which can meet the requirements.

The optimized parameters in Table 8 can serve as a reference for shearing tomato distal pedicels of various diameters. They also guide the design and parameter optimization of the shear end-effector in tomato harvesting robots. Gao et al. [12] conducted tests using the response surface test method and analyzed the test results using Design–Expert software to determine the optimal operating parameters of the vegetable harvester: shearing speed of 675 mm/s, chipping angle of 4.85°, clamping distance of 98.5 mm, and clamping angle of 64.5°. Liu et al. [15] also determined the shear parameters of the tool for shearing wild chrysanthemums by means of a central combination of simulation tests using shearing force as well as power consumption as indicators: cutting-edge angle of 21°, the cutting angle of 66°, and the reciprocating speed of 1290 mm/s.

3.3. Pedicel Three-Point Bending Test

From Figure 13, the three-point bending curve shows a distinct peak representing the abscission zone bending breaking force. This force is positively correlated with the diameter of the abscission zone. The curve’s trend aligns with this correlation. As can be seen from the figure, with the increase of bending loading displacement, the pedicel is crushed to the pressure density stage, the stage curve is approximately linear; after reaching the yield state, the abscission zone is broken. The bending test results are shown in

Table 10. Table of bending test results.

Test Size	Number of Tests	Moisture Content/(%)	Bending Breaking Force/(N)	Bending Modulus/(MPa)	Bending Strength/(MPa)
5−6 mm	1	78.806	22.694	7.145	5.485
	2	77.606	20.223	7.255	4.801
	3	76.695	25.009	8.339	5.565
	4	75.617	30.697	5.961	6.486
	5	73.077	13.299	6.844	3.138
	Mean	76.360 ± 2.178	22.384 ± 6.389 b	7.109 ± 0.855 a	5.095 ± 1.247 a
6−7 mm	1	76.754	32.835	4.948	3.921
	2	75.000	37.111	3.795	4.926
	3	75.758	32.678	3.942	3.413
	4	75.487	42.231	3.442	4.161
	5	73.860	42.035	2.948	5.164
	Mean	75.372 ± 1.060	37.378 ± 4.691 a	3.815 ± 0.740 b	4.317 ± 0.722 a
7−8 mm	1	75.042	29.442	2.627	2.613
	2	74.174	51.685	2.007	2.657
	3	74.426	33.993	2.529	3.017
	4	74.423	42.348	2.906	3.020
	5	74.331	50.567	3.969	3.990
	Mean	74.479 ± 0.331	41.607 ± 9.853 a	2.808 ± 0.726 b	3.059 ± 0.555 b

Note: The Duncan multiple range test indicates that values with different letters are significantly different at a 5% probability level.

.

From Table 10, the range of bending breaking force, bending modulus, and bending strength of the pedicels were 13.299−51.685 N, 2.007−8.339 MPa, and 2.613−6.486 MPa, respectively. The average maximum bending breaking force, average bending modulus, and average bending strength of the offsets with an abscission zone diameter of 5−6 mm were 22.384 ± 6.389 N, 7.109 ± 0.855 MPa, and 5.095 ± 1.247 MPa, respectively.

The average bending breaking force, average bending modulus, and average bending strength of the abscission zone diameter 6−7 mm were 37.378 ± 4.691 N, 3.815 ± 0.740 MPa, and 4.317 ± 0.722 MPa, respectively; and the average bending breaking force, average bending modulus, and average bending strength of the abscission zone diameter 7−8 mm were 41.607 ± 9.853 N, 2.808 ± 0.726 MPa, and 3.059 ± 0.555 MPa, respectively. The bending breaking force of the pedicels increased with increasing diameter, while the bending strength as well as the bending modulus decreased with increasing diameter.

As can be seen from

Table 11. Three-point bending correlation table.

	Abscission Zone Diameter	Distal Pedicel Diameter	Proximal Pedicel Diameter	Moisture Content	Bending Breaking Force	Bending Modulus	Bending Strength
Abscission zone diameter	1	0.817 **	0.880 **	−0.508	0.841 **	−0.893 **	−0.700 **
Distal pedicel diameter		1	0.839 **	−0.317	0.759 **	−0.721 **	−0.536 *
Proximal pedicel diameter			1	−0.352	0.796 **	−0.787 **	−0.543 *
Moisture content				1	−0.403	0.590 *	0.506
Bending breaking force					1	−0.763 **	−0.236
Bending modulus						1	0.607 *
Bending strength							1

Note: ** Correlation is significant at a 1% probability level (p < 0.01). * Correlation is significant at a 5% probability level (p < 0.05).

, the bending breaking force correlates very significantly (p < 0.01) with the abscission zone diameter, and there is no correlation with the moisture content. The bending modulus was highly significant (p < 0.01) in correlation with the diameter of the abscission zone and significant (p < 0.05) in correlation with the moisture content, and the bending strength was highly significant (p < 0.01) in correlation with the diameter of the abscission zone and not correlated with the moisture content.

The fitted curve of the regression equation of the abscission zone diameter versus the bending breaking force was obtained by Origin 8.0 and is shown in Figure 14. The regression equation of the abscission zone diameter versus the bending breaking force was obtained as follows:

$$y=-21.364+8.710x$$

(7)

R² = 0.707, F = 31.383, p < 0.001; x ∈ (5, 8.5).

The coefficient of determination of the regression equation is 0.707, p < 0.001, so this equation reaches the level of significance. Therefore, this equation effectively represents the relationship between the abscission zone diameter and the abscission zone bending breaking force. It can be used to predict the bending breaking force of the abscission zone.

Pedicel bending mode tomato harvesting robots typically clamp the tomato by an end-effector and subsequently rotate it through the last axis of the robotic arm to bend the pedicel and separate it. Liu et al. [34] concluded that the success rate of bending harvesting increased with the increase of the bending pendulum angle, with an average success rate of 72.9%, while fruit with longer pedicels was more difficult to fracture. However, as the bending angle of the pendulum increased, so did the number of accidents in harvesting, such as the possibility of loss of fruit bunches by pulling them off the plant. At the same time, neighboring fruits, interference with the end-effector, and accidental bruising also occur. The range of abscission zone bending fracture force derived in this test was between 13.299–51.685 N, and the fracture success rate was only 42.83%. They concluded that, taking into account the convenience of action implementation, the success rate, and the probability of accidents, a 30–40° upward bending is the ideal bending harvesting mode. The bending physical properties of tomato pedicels obtained in the three-point bending test can provide a theoretical basis for the bending-type tomato harvesting robot.

4. Conclusions

The tensile tests revealed that the pull-off force and tensile strength of the abscission zone increase with stem diameter, with significant correlations between tensile strength and the abscission zone, distal, and proximal pedicel diameters (p < 0.01), but no correlation with moisture content. The test results provide a theoretical basis for the design of tomato harvesting robots in tensile harvesting mode.

For the shearing tests, the proximal pedicel required 85.32% more force and 121.37% more energy than the distal pedicel, suggesting that cutting the distal pedicel is more energy efficient. The end-effector choice cuts the distal pedicel with less residue on the fruit. This way, the pedicels will not cause damage to other fruits during transportation after the tomato harvest. Optimal shearing parameters for distal pedicel diameters of 3.5 mm and 4 mm were established, with differences of less than 10% between experimental and predicted values. The shearing effect was satisfactory, with flush fractures and a 100% success rate, providing a reference for optimizing the shearing end-effector design.

In the three-point bending test, the bending strength and modulus of the abscission zone were found to decrease as the diameter increased. Significant correlations were observed between bending force, modulus, and abscission zone diameter (p < 0.01), while moisture content only significantly impacted the bending modulus (p < 0.05). The experimental results provide a theoretical basis for the design of tomato harvesting robots with a bending-type harvesting mode.

Overall, the success of bending-type harvesting was not high relative to the shearing-type harvesting mode as well as the tensile-type harvesting mode. Combined with the study of Zhang et al., the probability of tomato damage during tomato stretch harvesting is highest when the tomato clamping force is high. In summary, the more recommended harvesting mode in this paper is the shear harvesting mode.

In upcoming research, the results of physical testing will be applied to the design of an optimized tomato-harvesting end-effector. This process will involve selecting materials and developing a new series of mechanical tests, including contact experiments on tomato fruits and stalks using various materials. These experiments will inform the design process and lead to a comprehensive study of picking efficiency through field trials with the new end-effector.