Research on the Jet Distance Enhancement Device for Blueberry Harvesting Robots Based on the Dual-Ring Model

Abstract

In China, most blueberry varieties are characterized by tightly clustered fruits, which pose challenges for achieving precise and non-destructive automated harvesting. This complexity limits the design of robots for this task. Therefore, this paper proposes adding a jetting step during harvesting to separate fruit clusters and increase the operational space for mechanical claws. First, a combined approach of flow field analysis and pressure-sensitive experiments was employed to establish design criteria for the number, diameter, and inclination angle parameters of two types of nozzles: flat tip and round tip. Furthermore, fruit was introduced, and a fluid–structure coupling method was employed to calculate the deformation of fruit stems. Simultaneously, a mechanical analysis was conducted to quantify the relationship between jet characteristics and separation gaps. Simulation and pressure-sensitive experiments show that as the number of holes increases and their diameter decreases, the nozzle’s convergence becomes stronger. The greater the inclination angle of the circular nozzle holes, the more the gas diverges. The analysis of the output characteristics of the working section indicates that the 8-hole 40° round nozzle is the optimal solution. At an air compressor working pressure of 0.5 MPa, force analysis and simulation results both show that it can increase the picking space for the mechanical claw by about 5–7 mm without damaging the blueberries in the jet area. The final field experiments show that the mean distance for Type I (mature fruit) is 5.41 mm, for Type II (red fruit) is 6.42 mm, and for Type III (green fruit) is 5.43 mm. The short and curved stems of the green fruit are less effective, but the minimum distance of 4.71 mm is greater than the claw wall thickness, meeting the design requirements.

1. Introduction

Blueberries are delicious and nutritious, offering numerous health benefits, which makes them highly popular among consumers [1]. In China, the blueberry planting area and production have both rapidly increased, leading to a thriving planting industry. As a result, China has now become the country with the largest blueberry planting area [2]. The blueberry harvesting process accounts for about 40% of the total workload in the entire production process, making it the most time-consuming and labor-intensive stage [3], which significantly restricts the development of the blueberry industry. Mechanical harvesting can improve efficiency and reduce costs [4]. However, the characteristic of blueberries growing in clusters makes them one of the most challenging fruits and vegetables to harvest using mechanized and automated methods. Therefore, manual harvesting is still the primary method used, leading to low labor efficiency and high labor intensity, which in turn keeps the cost of blueberry cultivation high [5].

To address these challenges, researchers have studied blueberry growth, fruit characteristics, stem mechanical properties, and the connection properties between berries and stems [6,7,8]. Based on this research, various blueberry harvesting machines have been designed. Currently, the most mature and widely used types are the inertial vibration and comb-brush blueberry harvesters [9]. For example, OSKAR 4WD Plus berry harvesters developed by BEI [10], Littau [11], OXBO, and Poland’s JAGODA [12,13] use vibration devices to create contact vibrations with the plants. These vibrations utilize inertial force to separate the stems from the blueberry fruits. Similar harvesting equipment includes the striking-type picking machines [14,15]. These machines primarily use lateral vibrations on the trunk to shake off ripe fruits, making them more suitable for tall, tree-like plants found in regions like North America. However, most blueberries grown in China are low, shrub-like plants, making it difficult to directly adopt this method. Even with single-tooth mechanisms for precise picking, there is a risk of damaging surrounding fruits due to the close clustering of the berries. Pneumatic picking machines [16] use airflow and negative pressure to pick ripe fruits, but they also face the challenge of being unable to distinguish between ripe and unripe fruits, similar to vibration-based equipment.

With the rapid development of artificial intelligence, precise capturing and sorting of ripe and unripe fruits [17], as well as flexible, non-destructive picking, have become major research directions [18,19]. Currently, research on harvesting robots for large single fruits is quite advanced, focusing primarily on fruit recognition and picking path planning. Williams et al. used deep neural networks and stereo matching to detect kiwis [20], enabling robotic picking. Ge et al. proposed a lightweight CNN regression model for full-view coverage of fruit [21]. Li et al. developed a stereo-vision apple recognition and localization algorithm based on multi-task deep learning [22]. Zhou et al. designed a robot that determines apple picking paths using 3D point cloud data [23], achieving a success rate of 82%. Ye et al. developed a lychee picking robot that uses collision-free motion planning [24]. Huang et al. created an autonomous banana-picking robot that employs a corner-turning method for rapid path convergence [25]. For small fruits such as blueberries, cherry tomatoes, and strawberries, the picking process and end-effector selection are particularly important [26]. Liu et al. studied three different cherry tomato picking modes [27], providing a basis for small fruit harvesting. Zhang et al. designed a flexible end effector suitable for cherry tomatoes [28], enabling non-destructive picking. Lemsalu et al. developed a system for real-time recognition of strawberries and flower stems [29], but it performed poorly in dense berry clusters. Cao et al. proposed a strawberry instance segmentation method to enhance recognition and localization capabilities [30]. Tituaña et al. designed a strawberry-picking robot based on the YOLOv4 model [31], achieving a picking success rate of 71.7%. Although there have been some successful cases of harvesting small, soft fruits, most of them still focus on individual fruits.

For blueberries, key issues in precise harvesting include accurate ripeness detection, fruit softness, and the clustering problem, which remains one of the major challenges in the development of blueberry-picking robots. Blueberries have a long maturation period, so a cluster can contain both unripe and ripe fruits. During the harvesting process, the flexible mechanical gripper of a robot cannot directly reach into the fruit cluster to pick the ripe fruits. It requires a separation method to appropriately isolate interfering fruits. For mechanical separation methods, Fu et al. designed a kiwi-picking robot that uses a shielding plate to enclose the fruits [32], allowing the clustered kiwis to be individually separated for precise picking. However, this type of shielding plate design cannot be used for blueberry clusters, and mechanical separation methods are prone to causing fruit damage, especially with small, soft fruits. Therefore, our team studied the morphological and structural characteristics of blueberry fruits and stems, as well as the squeezing and friction characteristics of the fruits [33]. Based on this research, we decided to add a jet mechanism to the picking robot to separate ripe fruits from unripe ones, ensuring that the mechanical gripper can effectively reach the fruits. Before the end-effector clamping in the picking process, the separation of the fruit is interfered with by a momentary jet pass, thus satisfying the extension of the end effector. Therefore, selecting the appropriate nozzle and determining the working distance of the jetting mechanism is crucial. This paper analyzes two different nozzle parameters for the clustered blueberry jetting application. Through simulations and experimental testing, we determined the optimal nozzle structure. We also analyzed the force on the fruit stems under jetting conditions and conducted in-depth simulations and field experiments with the chosen nozzle to observe stem deformation and fruit separation under jetting.

2. Nozzle Simulation and Experiments

2.1. Designing Scheme

As shown in Figure 1, the blueberry-picking robot must complete five basic actions in sequence during each picking cycle: identification, positioning, air jetting, gripping, and separation. First, the recognition device identifies the mature fruit and determines the picking target’s location, issuing the corresponding commands. The recognition system assesses the pre-grip force based on the maturity and external characteristics of the fruit to prevent damaging the fruit during gripping. Next, it checks for the presence of interfering fruits; if any are detected, the air jet device separates them from the target fruit. Then, the system determines the spatial position of the picking target by identifying its X, Y, and Z coordinates. The control system moves the robotic arm to the working area, where the air jet device operates to separate the target fruit from interfering fruits. The mechanical claw grips the target fruit, and the end effector’s linkage mechanism then separates the fruit from its stem. Finally, the robotic arm follows an optimized path to place and store the blueberries. Since this study’s primary goal is to provide an air-jetting solution for separating clustered fruits (using specially designed nozzles to apply high-speed jets to the fruit cluster, allowing for accurate separation and enabling the mechanical claw to reach in), the specific functions of other modules are not discussed. Additionally, major components of the air jet device, such as the air compressor and regulating valves, are installed within the chassis of the picking robot, while the nozzle is connected to the robot’s small arm.

2.2. Simulation Modeling

2.2.1. Workspace Modeling

This project focuses on the rabbit-eye blueberries grown in greenhouses in northern China, which have a shape similar to a sphere, being either fully rounded or slightly flattened. We selected 300 blueberries and measured their diameter and thickness. The measurement results are shown in Figure 2. The maximum diameter of the blueberries does not exceed 25 mm, and the maximum thickness does not exceed 15 mm. Additionally, for rabbit-eye blueberries, the diameter of the entire fruit cluster typically does not exceed 40 mm. The jetting device needs to separate the target blueberry from the surrounding ones. Therefore, we established an annular model with a weak force at the center and a strong force at the periphery, as shown in Figure 3a. The nozzle output range must cover all sizes of blueberries. Therefore, we selected a weak force area with a diameter of 25 mm at the center of the nozzle’s circular working region, and a strong force area with a diameter of no less than 40 mm, with a working distance of 40 mm, as shown in Figure 3a.

2.2.2. Nozzle Model Establishment

This study aims to effectively separate the target fruit from the interfering fruit and to design dual-ring working area as described earlier. Commonly used single-hole nozzles (cylindrical or conical) do not meet our requirements. Therefore, we have chosen multi-hole nozzles for this study. Considering the difficulty of manufacturing and the need to achieve the dual-ring working area, we focused on studying and applying standard flat-tip and round-tip nozzles. Among them, Figure 3b shows the prototypes and cross-sectional structural models of the flat and round nozzles used. The design parameters include the number of holes, hole diameter, and hole inclination angle. The hole inclination angle is the angle between the centerline of the nozzle shell and the centerline of the hole. Specific data are provided in

Table 1. Nozzle design parameters.

Parameters	Notation	Values
Internal diameter	D	11 mm
Length of nozzle	L	23 mm
Number of holes	N	6, 8, 10, 12
Diameter of hole	d	1.00, 1.25, 1.5, 1.75, 2.00 mm
Inclination angle of the hole	α	30°, 35°, 40°, 45°

.

2.2.3. Finite Element Simulation

First, a finite element analysis of the nozzle’s flow field was conducted, using the Mosaic method for mesh division as shown in Figure 4. In numerical simulations, a grid independence analysis is often used to balance computational accuracy and time. Six mesh settings were tested using the Mosaic method, with grid numbers of 64,000, 75,000, 88,000, 101,000, 112,000, and 125,000. The effectiveness of each grid setting was evaluated based on the relative deviation between different schemes. Taking the results with 125,000 grids as the reference standard, the relative error in exit velocity and pressure is shown in Figure 5. The results indicate that the relative error between the calculations with 112,000 and 125,000 grids is less than 0.5%, which is within an acceptable range. Therefore, this article will select the 125,000 scheme to set the grid numbers.

In Figure 4, D represents the inlet diameter, L represents the nozzle length, and the blue mesh denotes the entire fluid domain. Subsequently, numerical simulation methods were used to calculate the Reynolds number for various nozzles based on Equation (1).

$$Re={\displaystyle \frac{\rho vd}{\mu }}$$

(1)

In the equation, $Re$ represents the Reynolds number, $\rho$ is the air density, $v$ is the average velocity, $d$ is the nozzle diameter, and $\mu$ is the air viscosity. Since the minimum Reynolds number for the nozzle is greater than 10,000, the standard k-ε model was chosen, and both complete turbulence and molecular viscosity can be neglected. Given that the Mach number is 0.286, the fluid can be considered incompressible, so the energy equation does not need to be considered [34]. The solution method used is the SIMPLE algorithm, which requires solving the pressure field through the Poisson equation and uses a steady-state solver. To ensure consistency, the same simulation conditions are applied to all structural models once they are determined.

Considering the actual output power of the air compressor and the correspondence of the simulation, the output pressure and speed of the air compressor were calibrated. A wind speed meter (Testo 417, Testo, Schwarzwald, Germany) was used to measure the wind speed at the interface between the air compressor and the nozzle. The measurement results are shown in

Table 2. Interface wind speed at different compressor output pressures.

Air Compressor Pressures (MPa)	Values (m/s)
0.1	6.11
0.2	7.74
0.3	8.56
0.4	9.11
0.5	9.73

. At an output pressure of 0.5 MPa, the wind speed at the connection point is approximately 9.73 m/s. Therefore, the simulation was set with an intake speed of 9.73 m/s to match the measured speed, and the outlet environmental pressure was set to one standard atmosphere. All nozzles with different structural parameters were set with the same initial boundary parameters.

2.2.4. Theoretical Analysis of the Working Interface

Based on the mechanical gripper’s motion stroke, the working distance in Figure 3a was calculated and set to 40 mm. The nozzle has a multi-hole structure, and the annular area in the working plane involves the coupling effect of the multi-hole airflow. A 2D model of the coupled airflow output characteristics was established, as shown in Figure 6, which includes three stages: the initial section, the basic section, and the dissipation stage.

The output characteristic evaluation indicators include the effective width of the nozzle jet and the sectional velocity of the basic section. The blue dashed line represents the section at the working distance within the basic section, and its length is the effective width. Additionally, the velocity boundaries of the strong and weak force zones should be as circular as possible, and the velocity distribution should be relatively concentrated. In the working area, the speed in the strong force zone must be sufficient to effectively separate the fruits without causing damage, as described by Equation (2).

$$V_w\ge V_e and V_{wmax}\le V_b=min\left\{V_{b1,}{V}_{b2}\right\}$$

(2)

In the equation, $V_b$ is the minimum wind speed that can cause damage to the fruit, $V_{b1}$ is the minimum wind speed that the blueberry skin can withstand, and $V_{b2}$ is the minimum wind speed that the connection between the mature blueberry and the stem can withstand. $V_w$ is the speed in the strong force zone at the working distance, $V_{wmax}$ is the maximum speed at the working distance, and $V_e$ is the effective wind speed, which is the minimum wind speed that can achieve fruit separation. The effective wind speed and the minimum wind speed that can cause damage to the fruit can both be experimentally measured using a wind speed meter. The measurement results are provided in the later section on fruit separation experiments.

In Figure 6, the areas defined by the red and green lines represent the airflow characteristics of two symmetrically positioned holes. The output characteristics of the multi-hole nozzle are influenced by the synergy between the two holes. As shown in Figure 6, the sectional diameter D_C of the airflow intersection region (indicated by the red and green dashed lines) increases with the distance Lc from the nozzle. When Lc is small, D_C increases linearly with Lc. However, when Lc exceeds 6 D, D_C increases more slowly, and beyond 8.5 D, D_C remains almost constant. Compared to a single-hole jet, the dashed line section shows significant contraction.

At different Lc values, the flow velocity V first increases and then decreases, with the peak occurring between 2.5 D and 8.5 D. This range is preferred for the annular model design; thus, in this instance, Lc is set to 40 mm. Notably, the central airflow velocity in the overlapping region is weaker than the velocity at the center of the holes. During actual jetting, when this part is directed at the target fruit, the normal pressure in the central region is small while the pressure in the edge region is high. This forms the mechanical basis for the ideal annular model in Figure 3.

Within the intersecting range of the two-hole airflow, the flow speed increases with distance, starting from zero in the central region and gradually matching the boundary flow speed at the edges. Therefore, the airflow speed in the middle of the multi-hole nozzle gradually increases while the intensity weakens, eventually equaling the boundary speed. Here, V represents the velocity in the nozzle flow field, and $V_{max}$ represents the maximum velocity across different sections of the nozzle flow field. For a single-hole jet, the jet spreads out at a diffusion angle, with the jet radius increasing linearly over the basic stage. The relationship for the jet radius changes in the basic stage is shown in Equation (3).

$$R=3.4a\left(x_0+s\right)$$

(3)

here $R$ represents the jet radius, $a$ represents the exit cross-sectional coefficient, which is determined by the material and manufacturing process, $x_0$ represents the pole depth, and $s$ represents the turbulence coefficient. Transform the formula to obtain Equation (4):

$${\displaystyle \frac{R}{r_0}}={\displaystyle \frac{x_0+s}{x_0}}=1+{\displaystyle \frac{s}{r_0/tg\beta }}=1+3.4a{\displaystyle \frac{s}{r_0}}$$

(4)

here $r_0$ represents the nozzle radius, $\beta$ represents the jet diffusion angle. Empirical Equation (5) can be obtained by combining Equation (5):

$${\displaystyle \frac{R}{r_0}}=\lambda \left({\displaystyle \frac{as}{r_0}}+0.294\right)$$

(5)

here $\lambda$ represents the outlet sectional coefficient, which is complex for intersecting porous jets. The deformation degree of intersecting jets is generally indicated by $\phi$, as shown in Equation (6). At the initial stage, the value of $\phi$ starts to change after the jets intersect. The value of $\phi$ is determined by the distance between the nozzles, the inclination angle α of the nozzles, and the diffusion angle $\beta$ of each jet’s outer boundary. Until the dissipation stage, where the primary deformation rate $\phi$ becomes constant, the primary deformation rate on any cross-section is equal, making the combined flow resemble a single free jet. At this point, all deformations caused by the intersecting jets disappear.

$$\phi ={\displaystyle \frac{R_x-R}{r_0}}$$

(6)

here $R_x$ represents the radius of the jet after the intersection. The two-dimensional model can be used to verify the output characteristics at the working distance, and the more accurate and complex spatial coupling model is usually calculated using the finite element simulation method, and its flow field analysis can be compared with the planar model calculation results.

2.3. Simulation Results and Analysis

2.3.1. Analysis of the Influence of Different Parameters of Flat Tip Nozzles

The main influencing parameters of flat-tip nozzles are the number of holes and the hole diameter. Figure 7 shows the output airflow speed under different structural parameters with the same boundary conditions. From the simulation in Figure 7a, we can see that increasing the number of holes while keeping the hole diameter at 1 mm reduces the output speed of the airflow. Under the condition of six holes, the output fluid speed can still maintain 30 m/s at 30 mm. As the number of holes increases, the airflow speed decreases. The designed 12-hole nozzle can maintain 30 m/s at 15 mm. Similarly, as shown in Figure 7b, increasing the hole diameter while keeping the number of holes at six reduces the output speed of the airflow. Under the condition of a 1.25 mm hole diameter, the speed distribution is relatively concentrated, with a larger and more noticeable range of high-speed airflow indicated in red. As the hole diameter increases, the speed output characteristics decrease, and the airflow speed within the working range is less than 20 m/s for a 2 mm hole diameter. The trend of output airflow changing with parameter variation can guide structural selection for different application scenarios. For this study, the primary goal is to separate the fruit. The flat-tip nozzle we need has a very short length between the weak and strong force areas. The output airflow spreads out in a triangular pattern, making it challenging to find a suitable working ring in the flow field. However, the six-hole 1 mm flat-tip nozzle is the best choice among flat-tip nozzles because it has fewer holes and a relatively smaller hole diameter.

2.3.2. Analysis of the Influence of Different Parameters on Round Tip Nozzles

The main influencing parameters of round tip nozzles are the number of holes, hole diameter, and hole inclination angle. Figure 8 shows the output airflow speed under different structural parameters of round-tip nozzles with the same boundary conditions. Since the effect of hole diameter on the nozzle’s output speed is well-established (the speed continuously decreases as the hole diameter increases, without affecting the shape of the gas output), Figure 8 only displays simulation results for different parameters with a preferable hole diameter of 1 mm.

From the simulations in Figure 8, we can see that increasing the inclination angle of the holes, while keeping the number of holes constant, makes the gas output characteristics more divergent. Under the condition of six holes with a 45-degree inclination angle, the gas from two holes on the same cross-section already shoots towards both ends. However, we find that increasing the number of holes can compensate for this loss. Compared to 6 holes, increasing the inclination angle for 8, 10, and 12 holes shows significant improvement. For our work, the goal is to use the annular airflow to retain the target fruit while separating the interfering fruit on the outer ring, so we need to find a suitable working ring. From the simulations, it is clear that the speed output characteristics of 6 holes at 35°, 8 holes at 40°, and 10 holes at 45° are better. However, it is difficult to evaluate their actual effect solely from the cross-sectional view, so we will further analyze and discuss the selected nozzle structures.

2.3.3. Simulation Analysis in the Work Area

To analyze the output characteristics on the working area plane, we selected a 40 mm × 40 mm square cross-section at a distance of 40 mm from the nozzle for comparative analysis of the working area. The simulation results of the nozzles with the better output characteristics mentioned above are shown in Figure 9.

As shown in Figure 9a, the average speed of the flat tip six-hole nozzle is very high, the speed characteristic decreases slowly, and the roundness of the plane is good. On the working plane 40 mm away from the nozzle, the simulation results show that the speed at the center exceeds 15 m/s, forming a circle with a diameter of about 10 mm. However, this strong convergence is not favorable for this study as it contradicts our design concept of weak force at the center and strong force at the edges. As shown in Figure 9b, the six-hole 35° nozzle has a clear weak force area at the center and strong force areas around it on the 40 mm working plane. The airflow from each hole is distributed in dots across the working section, arranged sequentially according to the position of the holes. This condition meets the requirements of this study, as the nozzle output characteristics are significantly dispersed with minimal interaction between airflows, though the roundness of the shape is somewhat poor, necessitating further consideration. As shown in Figure 9c, the eight-hole 40° nozzle exhibits a distinct ring distribution of the weak force area at the center and strong force areas around it on the 40 mm working plane. The gas speed at the boundary between the weak and strong force areas is about 14 m/s, and the speed in the central weak force area is about 11 m/s. Additionally, the cohesion of each hole is relatively good. For this study, it meets the need for separating the fruit and branches around the ripe fruit. Finally, the 10-hole 45° nozzle shows a significant decrease in speed output characteristics and earlier diffusion stage due to the larger inclination angle and more holes, resulting in greater interaction between the holes. As shown in Figure 9d, the output airflow speed is lower, and the shape is irregular.

In summary, combining the analysis of output characteristics and working cross-sections with different nozzle parameters, the round-tip nozzle proves more advantageous than the flat-tip nozzle. The analysis of the working cross-section indicates that the 8-hole 40° round nozzle performs the best.

2.4. Pressure Sensing Experiment

2.4.1. Construction of the Nozzle Test Bench

The main components of the nozzle test bench include an air compressor, a pressure regulator, solenoid valves, a field-programmable gate array (FPGA) driver board, nozzles, and a sensor testing control board (Figure 10). The air compressor supplies high-pressure air to the test bench, while the pressure regulator controls the air pressure. The solenoid valves are controlled via the FPGA driver board to regulate the opening and closing of the valves for jetting. Uniformly distributed thin-film sensors are mounted on the panel of the testing platform. After jetting, these sensors transmit signals to the computer through the sensor control board for data collection and processing.

2.4.2. Testing and Verification Methods

In this study, the working cross-section is set at a plane 40 mm from the nozzle. The compressor and pressure regulator are controlled to output a pressure of 0.5 MPa. The nozzle is connected to the compressor with a 0.25-inch diameter connection, setting the inlet speed to 9.73 m/s for the pressure sensitivity test. Additionally, during the test, the position of the bracket and sensor panel is kept fixed, and only the nozzles with different parameters are replaced. Each time the nozzle is changed, the compressor is stopped first, then the solenoid valve is closed, and finally, the pressure regulator is turned off. After changing the nozzle, the working conditions are set to the specified settings following the same procedure. Each nozzle experiment is conducted three times, and the computer collects the pressure data through the sensor controller to calculate the average value. By comparing the simulation results and experimenting with different parameters of flat-tip and round-tip nozzles, we analyze the differences between them. To verify the accuracy of the simulation results, a matching chart of simulated wind speed and test pressure for the nozzles is established. Finally, the pressure test board is replaced with blueberry branches and leaves for fruit separation experiments. The differences in mechanical analysis deformation, fluid–structure interaction simulation, and field jetting fruit separation are compared, and the effective separation speed is determined using a wind speed tester.

2.4.3. Pressure Sensitivity Test Results and Analysis

Pressure sensitivity tests were conducted for the relatively better nozzles mentioned above, and the experimental data were recorded with the computer via the sensor control board. The nozzles were labeled as follows: A for the 6-hole flat tip nozzle, B for the 6-hole 35° round tip nozzle, C for the 8-hole 40° round tip nozzle, and D for the 10-hole 45° round tip nozzle.

Under the same simulation settings as mentioned earlier, the air compressor was adjusted to a working pressure of 0.5 MPa. The test environment was not altered, and the same bracket was used to move the pressure sensor panel continuously, recording pressure results. The test procedure consisted of two parts. The first part involved aligning the center of the test panel with the center of the nozzle, maintaining a symmetric position while moving the panel to test the axial pressure variation of the output fluid by changing the distance from the nozzle. The second part involved aligning the center of the test panel with the center of the nozzle, keeping the distance between the panel and nozzle constant while moving the panel laterally to test the pressure variation across the working area concerning the diameter.

The axial pressure variations of the four nozzles are shown in Figure 11. Nozzle A (6-hole flat tip) exhibited the highest axial output pressure. The axial pressure of A initially increased slightly before rapidly decreasing with an increase in distance, reaching a maximum pressure of 106 kPa at 15 mm. At a working distance of 40 mm, the pressure was still 62 kPa, which could cause fruit damage or detachment. The axial output pressures of B (six-hole 35° round tip) and C (eight-hole 40° round tip) were both below 30 kPa, moderate enough to prevent fruit damage or detachment. Their output pressures were similar, with the weakest force located 35 mm from the nozzle, close to the 40 mm working distance mentioned earlier. The axial output pressure of D (10-hole 45° round tip) showed the lowest point of 10 kPa at 25 mm, with a slow increase in pressure afterward. The longitudinal diameter pressures of the four nozzles on the 40 mm working cross-section are shown in Figure 12. It is clear that nozzle A (6-hole flat tip) had the highest pressure of 90 kPa at the center of the circle, decreasing gradually with an increasing diameter, which contradicts our design concept. It is also evident that nozzle D (10-hole 45° round tip) did not meet the requirements, as the gas speed within a 25 mm diameter at the center was similar, failing to achieve the special requirement for fruit separation. The output pressures of B (six-hole 35° round tip) and C (eight-hole 40° round tip) were lowest at the center of the circle, increasing with the diameter. Their pressures started to decrease after reaching the maximum point, indicating a pressure trend from the center to the periphery, which is suitable for the annular working cross-section. The eight-hole 40° round tip nozzle has advantages in terms of the annular working area and pressure peak. The pressure variation trend of the eight-hole 40° round tip nozzle is also clearer. Additionally, Figure 13 shows the corresponding results of test pressure and simulated speed to further demonstrate the relationship between the test and simulation. For nozzle A, an output speed of 34 m/s corresponds to a test pressure of about 88 kPa, while for B, an output speed of 12.5 m/s corresponds to a test pressure of about 36 kPa. For C, an output speed of 17 m/s corresponds to a test pressure of about 45 kPa, and for D, an output speed of 11 m/s corresponds to a test pressure of about 30 kPa. In conclusion, based on the pressure test results, the ranking is as follows: 8-hole 40° round tip > 6-hole 35° round tip > 10-hole 45° round tip > 6-hole flat tip.

3. Simulation and Experimentation of Fruit Stalk

3.1. Mechanical Analysis of Fruit Separation

During the air jet process, the fruit is mainly subjected to the airflow pressure output from the nozzle. The force analysis is shown in Figure 14, where Q represents the distributed load, and G represents its own gravity.

In this study, we disregard the negligible weight of the fruit stem itself. Additionally, the connecting force Fs between the fruit and its stem is considered. The angle between the blueberry stem outside the picking target and the main branch node is defined as θ. The connection between the stem and the branch node is treated as a cantilever beam in this study. Through mechanical simplification, the deflection formula is derived as follows:

$$EI{w}^{″}=-M\left(l\right)$$

(7)

where $E$ is the elastic modulus of the material, $I$ is the moment of inertia of the section, $w$ is the deflection, and $M$ is the relationship between the bending moment $M$ and the distance $l$ from the endpoint to the fruit. Incorporating the boundary conditions of the cantilever beam, Equation (7) is transformed into Equation (8).

$$EIw=\int [\int \left({\int }_0^{L}Qsin\theta l^{2}dl\right)dl]dl-{\displaystyle \frac{1}{6}}G{L}^{3}cos\theta$$

(8)

where $L$ is the length of the fruit stalk, and the load $Q$ is fitted using a logistic model based on the output characteristics of the nozzle axis distance. The specific parameters are listed in

Table 3. Specific parameters of the load model fit.

Parameter	Value
A1	17.24839
A2	61.26441
x0	11.39066
P	2.4368
Reduced Chi-Sqr	0.08925
R square (COD)	0.99925
Adjusted R-squared	0.9991

, and the model equation is given by formula (9).

$$Q=A_2+{\displaystyle \frac{(A_1-A_2)}{1+(\frac{x}{x_0}{)}^P}}$$

(9)

The relationship between $l$ and $x$ is as follows:

$$x=lsin\theta$$

(10)

The parameter $l$ is the direct evaluation criterion for determining the fruit separation effect. Here is a calculation example for the largest fruit: when the fruit peduncle is in a vertical position, the deformation due to the fruit’s weight is minimal, and the deformation of the peduncle under full load is approximately 10.87 mm. In a horizontal position, the deformation caused by gravity is about 1.559 mm. When the peduncle length is 15 mm, and the fruit is positioned at a 53° angle, the deformation is approximately 7.247 mm. For the fruit on the lower side, the horizontal deformation of the peduncle is about 1.659 mm. With a remaining gap of approximately 5.588 mm, the minimum gap fully accommodates the mechanical claw insertion.

3.2. Simulation Analysis of Fruit Separation

This section conducts a fluid–structure interaction simulation for the optimized 8-hole 40° nozzle with fruit. The fruit model is placed at a distance of 40 mm from the nozzle’s axis on the working cross-section. The simulation uses the maximum fruit as an example, with a peduncle length set to 15 mm. The fruit-peduncle model and its structural and material parameters are detailed in

Table 4. Material parameters of fruit and fruiting stem.

Parameter	Value
Fruit diameter	25 mm
Fruit thickness	15 mm
Fruit density	1.16 g cm−3
Fruit Poisson’s ratio	0.35
Fruit modulus of elasticity	0.225 MPa
Stem length	15 mm
Stem diameter	1.5 mm
Fruiting stem density	38 g cm−3
Fruiting stem Poisson’s ratio	0.38
Fruit stem modulus of elasticity	14.2 MPa
Connectivity of fruiting stem to mature fruits	0.17~0.83 N
Connectivity of fruiting stem to immature fruits	1.64~3.67 N

. The fruit model is imported through the fluid–solid coupling module and uses an unstructured mesh method. This approach has advantages in fluid–solid coupling for the solid domain (force exerted by the fluid on the solid surface). Smaller and more numerous meshes allow for better resolution of boundaries and analysis of external fluid forces.

The fluid–structure interaction simulation results for three different fruit orientations (horizontal, vertical, and 53° angle) are shown in Figure 15a–c. These figures clearly illustrate the interaction between the fluid and the fruit. In the software, the end of the peduncle is modeled as a cantilever beam. The airflow from the nozzle exerts forces on the fruit in different orientations, causing deformation. The stress and strain for the horizontal fruit are shown in Figure 15d,g. Since the airflow from the eight-hole nozzle is symmetrically directed, the surrounding pressure is evenly balanced. However, the deformation of the fruit peduncle is downward due to the influence of gravity. The maximum strain is located at the peduncle, where the effect of the airflow load increases after the fruit sags under gravity, resulting in a maximum strain of 0.00496. The total deformation is approximately 1.74 mm, which shows a slight deviation from the theoretical calculation. The stress and strain for the vertical fruit are shown in Figure 15e,h. The stress in the vertical orientation is mainly concentrated in the lower left part of the fruit facing the nozzle. The stress on the back of the fruit and peduncle is relatively small. Therefore, the deformation of the vertical fruit is primarily concentrated at the peduncle, manifesting as an inclination away from the nozzle. As shown in Figure 15f, the maximum strain is 0.0193, and the total deformation is approximately 9.68 mm. Due to the reduced load after bending in the vertical position, combined with the effects of gravity and the pressure difference between the upper and lower parts of the fruit, the deformation deviates from the theoretical calculation.

For the fruit at a 53° angle, the stress and strain are shown in Figure 15f,i. The stress is mainly concentrated on the upper part of the fruit, with other areas having relatively balanced stress. The peduncle deforms upward. As shown in Figure 15i, the maximum strain is 0.014, and the total deformation is approximately 7.12 mm. This is due to the increased load after upward deformation, with higher airflow velocity on the upper side and lower velocity on the lower side, leading to a greater deformation than the theoretical calculation.

Considering the deformation of 1.74 mm in the horizontal case, the minimum gap for the fruit cluster on the lower side is about 5.38 mm, which differs by 0.2 mm from the ideal calculated gap value. However, the results satisfy the requirements for the mechanical claw insertion.

3.3. Experimental Verification of Fruit Separation

To further validate the accuracy of the simulation and theoretical analysis, we conducted separation experiments on blueberry clusters using the previously optimized eight-hole 40° round nozzle. The results of the separation experiments are shown in Figure 16.

Figure 16 illustrates the interference of fruits at different ripeness stages (green, ripe, and mature) with the target fruit to be picked. To prevent damage to the fruit by the mechanical claw, the gap for fruit separation after spraying should be greater than 40 mm. The nozzle was fixed on a support stand and positioned 40 mm from the target fruit cluster for the spraying experiments. Figure 16a shows the interference of green and ripe fruits with mature fruit, where the red box represents the target fruit to be picked. The separation results are shown in Figure 16b. Due to the hardness of green fruits and their smaller airflow range, but with softer peduncles compared to ripe fruits, the deformation is similar to that of ripe fruits. Under the airflow, the fruits did not fall off, and the gap between fruits measured with a caliper was approximately 6.53 mm. Figure 16c shows the interference of mature fruits with the target ripe fruit, and Figure 16d shows the separation results between mature fruits. With forward air jetting on the target fruit, the side fruit spacing is approximately 5.08 mm. A 0.3 mm error, corresponding to a relative error of 5.9%, arises due to differences between the real blueberry cluster and the simulation. This error is caused by discrepancies in fruit size, stem length, and simulation settings, as well as the fruit’s orientation in actual conditions. Despite the errors, the overall results still meet the requirements for the mechanical claw to reach in. In addition to the specific cases shown in the figures, we conducted 30 sets of spraying experiments, with detailed data presented in

Table 5. Experimental results of different types of disturbances.

No.	Type of Disturbance	Initial Spacing (mm)	Blowing Spacing (mm)
1	II	0.34	5.67
2	II, III	1.40	6.11
3	I, III	0.21	5.42
4	II, III	1.32	5.95
5	II, III	0.38	5.78
6	I	0.00	5.08
7	II	1.68	6.24
8	II	0.71	6.01
9	II, III	1.61	6.53
10	II	0.15	5.88
11	I	0.00	5.34
12	II	0.00	5.62
13	II, III	1.37	6.36
14	I	0.56	5.26
15	II	0.00	5.73
16	I, II	0.00	5.11
17	II	0.17	5.99
18	I, II	1.53	6.89
19	II	1.34	6.42
20	I	1.00	5.81
21	II, III	0.98	6.17
22	I	0.82	5.65
23	III	0.00	4.96
24	II	0.56	6.06
25	I	0.00	5.21
26	II, III	1.21	6.12
27	II	0.43	5.89
28	III	0.00	4.76
29	III	0.28	5.43
30	I	0.00	5.55
Arithmetic mean		0.60	5.766

Ripe fruit type I, red fruit type II, and green fruit type III.

. For Type I, the maximum distance was 5.81 mm, the minimum distance was 5.08 mm, and the average distance was 5.41 mm, with a small deviation. For Type II, the maximum distance was 6.42 mm, the minimum distance was 5.62 mm, and the average distance was 5.95 mm, with a small deviation. For Type III, the maximum distance was 5.43 mm, the minimum distance was 4.76 mm, and the average distance was 5.2 mm, with a larger deviation. Despite the short and poorly bent peduncles of green fruits, the minimum distance of 4.71 mm is still greater than the claw thickness, meeting the design requirements.

Additionally, to calibrate the effective wind speed mentioned in the design, a wind speed meter was placed below the fruit cluster during the spraying separation process. With a required fruit separation gap of 4 mm, the nozzle output speed must be at least 9.2 m/s to be considered effective. The spraying device was directed at the back of the fruit, and the wind speed meter was gradually raised with increasing output pressure. When the airflow reached approximately 23.5 m/s, the fruit separated from the peduncle and flew out. With blueberries fixed in a fixture and direct airflow applied at a speed of 23.5 m/s, the fruit skin did not rupture, indicating that the maximum wind speed that the fruit can withstand is 23.5 m/s.

4. Discussion

With the rise in labor costs and advancements in agricultural technology, the mechanization of crop harvesting is undoubtedly a future trend in agriculture. However, blueberry harvesting has lagged in the development of picking robots due to the clustered nature and delicate texture of the fruit.

This study addresses the challenge of accurately picking clustered fruits by proposing the addition of a spraying mechanism in the picking robot. The spraying action increases the spacing between blueberry clusters to facilitate the insertion of the mechanical claw. We compared the output characteristics of different nozzle airflows and assessed the impact of various parameters on airflow. This was validated through pressure sensitivity tests, and we established a relationship between nozzle flow speed and output pressure. We modeled the working area and two-dimensional output flow field for the northern rabbit-eye blueberry used in this study, analyzing effective wind speeds and working ranges. By comparing and analyzing, we selected the most suitable nozzle structure and performed theoretical mechanical analysis. We then imported the fruit peduncle model for fluid–solid coupling simulations and conducted fruit separation experiments at a local plantation to compare the results of different interference fruits and separation experiments. The comparison of mechanical analysis, fluid–solid coupling simulations, and fruit separation experiments provided a more accurate reference for future work.

The spraying device and working area settings are designed for rabbit-eye blueberries, with the working area set based on the maximum fruit size for separation. In actual operations, not all fruits are large. We found that small fruits, due to their high hardness, can be pushed away by the mechanical claw without causing injury to the fruit. In cases where fruit peduncles are short and the fruit hardness is high, the separation gap may not be sufficient. In such cases, if the mechanical claw cannot be inserted, the problem can be improved by increasing the output pressure or using nozzles with more apertures. When harvesting different varieties of blueberries, nozzles or working conditions can be adjusted based on the analysis of specific characteristics to achieve effective fruit separation.

In the future, we will continue to refine the structure of the picking robot. For the spraying device, adjustments will be made based on different varieties, fruit orientations, and densities. Different working conditions will correspond to different nozzle structures and operational settings. Additionally, we will incorporate airflow control into visual recognition systems to address issues where the fruit’s front surface is ripe, but the back or sides are not, affecting the robot’s visual assessment.

5. Conclusions

This study on blueberry cluster separation through theoretical calculations and fluid–solid coupling simulations has led to the following conclusions:

• To address the challenge of tightly clustered blueberries that are difficult for mechanical claws to harvest, we proposed a design using a spraying device to separate target fruits from interfering fruits;
• We established a circular model of the working end face in the spatial flow field. By designing high-pressure and low-pressure zones, we achieved functional separation of the airflow. Finite element analysis shows that the effect of air jet separation is directly related to the nozzle structure type and parameters. A solution using an eight-hole 40° round-head nozzle effectively meets the needs of dispersing clustered blueberries;
• Fluid–solid coupling simulation indicates that the fruit’s orientation significantly affects the air jet separation effect. For example, for fruit with a stem length of 15 mm, the horizontal deformation is about 1.74 mm, the total vertical deformation is approximately 9.68 mm, and the deformation at a 53° horizontal angle is about 7.12 mm, with a minimum gap of about 5.38 mm. Therefore, to achieve more accurate results, it is necessary to consider the mechanical parameters of the fruit stem and the impact of gravity. Additionally, the target fruit in the low-pressure area should be kept as horizontal as possible, and the horizontal orientation angle of the fruit in the high-pressure area should be maximized;
• Experimental data comparison shows a 10.0% error in theoretical calculations and a 5.9% error in simulation calculations, with the fluid–solid coupling model being more accurate. These errors are due to differences in fruit size, stem length, and simulation parameters, as well as the impact of the fruit’s orientation in actual working conditions. Despite these errors, the overall results meet the design requirements;
• The research methods and calculation models in this study provide a foundational approach for implementing mechanical claw gripping of blueberries and offer insights for separating and harvesting other similar clustered fruits.