China has the highest vegetable production and consumption in the world now, with the vegetable planting area of 20 million hectares and an annual output of more than 700 million tons (National Bureau of Statistics of China, 2019). With the decrease of labor force and the increase of labor cost, many kinds of vegetables such as pepper, tomato and cucumber are transplanted by using the semi-automatic transplanters (Kumar & Raheman, 2011; Zhang et al., 2011), but these semi-automatic transplanters operate at a low speed because the process of seedling transporting and picking still need to be done manually, so we must find some ways to guarantee the high-speed transplanting and low labor using (Zhang et al., 2013).

In order to improve the transplanting efficiency, we must develop automatic pick-up devices to replace the manual work of seedling transporting and picking. As an important part of the transplanter to realize automatic transplanting, the pick-up device should ensure low damage of the plug seedlings during the seedling picking and dropping process, with all the parts operating fast and accurately (Jin et al., 2018). Scholars from all over the world have already developed some high-efficiency automatic pick-up devices (Chen et al., 2014; Thomas & Kumar, 2017; Khadatkar et al., 2018; Ye et al., 2019), which can really realize automatic seedling picking instead of manual work. Actually, these pick-up devices are mainly composed of end-effectors to realize seedling picking and dropping, as well as some other parts to make sure the end-effectors can reach each working position. As the core of the automatic pick-up device, the end-effectors must ensure the function of successfully seedling picking from the plug tray and seedling dropping into the transplanting cup (Han et al., 2019b). As early as 1990, Ting et al. (1990) designed a sliding needle gripper which was adaptable to a wide range of seedling sizes and shapes. Ryu et al. (2001) proposed an end-effector comprised of two air cylinders and fingers with a good transplanting performance. In order to adapt to the seedlings growing in China, Han et al. (2015) developed a pincette-type end-effector with two fingers and four pins to pick the seedlings following the previous design, and Tong et al. (2019) designed a spade end-effector which could complete the seedling picking and reduce the residual substrate in the plug hole.

At present, most seedling picking methods are designed to ensure that the end-effectors can insert into the seedling pot to clamp the seedlings, and the end-effector must break through the adhesion between the seedling pot and the plug tray in order to pick the seedlings out from the plug tray successfully (Han et al., 2013; Ma et al., 2020a). But the inserting and clamping from the end-effector will bring damages to the root system (Liu et al., 2018), and when the clamping force is less than the sum of the gravity of seedling pot with the adhesion between substrate and the plug tray, the seedling pot will be destroyed by the end-effector. To reduce the damage rate of seedling pot, Tong et al. (2019) provided a new seedling picking method, which is the end-effector using four shovel shaped pins to clamp the substrate around the seedling pot instead of inserting into the seedling pot. This method can reduce the damage of the seedling pot to a certain extent, but the damage still exists because the pins will also break out the adhesion between substrate and the plug tray during seedling picking. Comparing the loosened seedlings with the unloosened ones during seedling picking, Yang et al. (1991) found that the success rate of seedling picking reached 96.3% when using the loosened plug seedlings, but the rate reached 50.9% when using unloosened plug seedlings. Based on the idea for seedling picking easily, Ferrari (2018) created a FUTURA automatic transplanter by using a steel needle to push out the plug seedling, which not only improves the efficiency of seedling picking, but also greatly reduces the damage of seedling pot. In order to match the existing end-effector with the seedling push-out device, Gao et al. (2017) simulated and analyzed the working process of the steel needle pushing out the plug seedling by using the EDEM software. Actually, the seedling push-out method has high requirements for the packing property of seedlings. If the packing property is poor, the steel needle will directly thrust into the seedling pot and the root-soil structure of the seedling pot will be destroyed. The survey found that the drainage outlet size of some plastic plug trays in China is not uniform, so the steel needle cannot push out the seedling and sometimes destroy the plug tray, and the matching accuracy of steel needle and drainage outlet of the plug tray is also very high.

Xiang et al. (2007) developed an experiment of throwing rice seedlings by air force for rice transplanter, the air flow generated by air compressor is used to blow the rice seedling into seedling guide tube during the experiment, and this method is not easy to destroy rice seedlings and the structure is simple compared with the push-out device.

Drawing on the successful experience of the rice transplanter, this was aimed to design an automatic pick-up device using air force to blow out vegetable plug seedlings with some flexible clamping end-effectors picking and dropping seedlings. Besides, a control program was designed to match the actions between the air loosening device and the end-effectors, and an experiment of prototype was implemented to analyze the optimum parameters of the pick-up device.

Design requirements

The following requirements were taken in account when designing the automatic pick-up device:

1. Plastic plug tray had 72 (6×12) holes; the upper and lower caliber of one hole were 40×40 mm and 20×20 mm respectively, and the depth of one hole was 45 mm. Chinese soft plastic plug trays were used (provided by Longji Plastics Co., Taizhou, China).

2. The substrates used in the plug trays were composed of peat, vermiculite and perlite at volume ratio of 3:1:1, the volume ratio of substrates to one hole was 1:1.2, and the moisture content of the substrates was 55-65% (Miao et al., 2013). The seedlings used in the test were 25-30 days old, and the height of the seedlings was from ~100 mm to ~200 mm.

3. The whole row seedling picking method was de3 A pick-up device using air force to blow out seedlings high quality. Most of the devices are driven by the air force signed to improve the efficiency of seedling picking and reduce the damage of the seedlings during the working process. As shown in Fig. 1, when the plug tray is in place, the first six seedlings marked F will be picked together with six end-effectors installed in the pick-up device, then the transporting device will move one hole in the transporting direction for another six seedlings (marked S) picking, so the pick-up device can pick one row seedlings during two movements. 4. The end-effector should be vertical to the plug tray when picking seedlings, and the seedlings should be dropped to the seedling cup in the vertical direction. Besides, the end-effector should ensure the integrity of seedlings during the working process. 5. Picking frequency: following the transplanting speed of the transplanter in China, the picking frequency of the pick-up device should not be < 90 plants/min.

Structure and working principle of the pick-up device

According to the requirements mentioned above, the pick-up device was mainly composed of a seedling air loosening device, a seedling clamping device, some seedling cups, an automatic control system, an air compressor and the fixed frame (Fig. 2). A seedling transporting device is needed to feed seedlings to the pick-up device. The seedling air loosening device was installed under the lower surface of the tray baseplate in the seedling transporting device, which was used to blow out the seedlings from the plug tray. The seedling clamping device can move to the predetermined positions with the seedlings outside the plug tray. Finally, the seedling cups accept the seedlings dropped from the seedling clamping device. The frame, assembled from some aluminum profiles, is easy to install and with supplied from the air compressor and stepper motor under the arrangement of the automatic control system.

As shown in Fig. 3, the working process is divided into six parts (a-f) for better explanation. Taking one working cycle as an example:

(1) Before seedling picking, the seedling clamping device ([1]; the same blow) is in the seedling dropping position and the seedling air loosening device [6] is in the initial state (Fig. 3a).

(2) When the seedling transporting device [4] feeds the plug tray [2] to the preset position, the seedling clamping device will flip to the seedling picking position, and the seedling air loosening device will push the air nozzle [5] close to the bottom of the plug tray (Fig. 3b). (3) After that, the air nozzle will blow out the plug seedling (marked 3) from the plug tray, and the seedling clamping device will clamp the plug seedling at the same time (Fig. 3c).

(4) After seedling picking, the seedling clamping device will flip back to the seedling dropping position, and the air nozzle will turn back to the initial state to prepare for the next seedling picking (Fig. 3d, e).

(5) Finally, the seedling clamping device will drop the seedling in the vertical direction into the seedling cup [8], and that is the end of one cycle (Fig. 3f);

(6) All the devices will repeat this cycle again until all seedlings are taken out from one plug tray, and the plug tray will drop into the tray collection box [7] for reuse.

Analysis of deformation of drip irrigation

Consider a linear cross section for the drip-irriga-tion tapes (i.e., the tape is collapsed), so no water flows through the tape, and the upper and lower pipe wall arein contact with each other. When water flows through the tape, the upper and lower walls separate, and the cross section changes with water pressure. At low pressure, the tape does not fully expand, so the cross-sectional shape is a “flat” ellipse. Upon increasing the pressure, the cross section becomes more circular. In this paper, we used α as the flattening coefficient to quantify the flatness of the drip-irrigation tape. The expression for α is

α = D D b a (4)

where Db (Da) is the vertical (horizontal) inner diameter of the drip-irrigation tape. A smaller (larger) α means that the drip-irrigation tape is flatter (rounder). The range of α is 0≤α≤1. The cross section corresponding to α = 0.9 is taken as the critical cross section between oval and circular; that is, the critical point where the cross-sectional shape changes qualitatively. The pressure corresponding to the critical cross section is called the critical pressure. For 0≤α≤0.9, the cross-sectional shape was nearly elliptical, and for 0.9≤α≤1, the cross-sectional shape was circular.

Fig. 4 shows the flattening coefficient as a function of pressure for the six drip-irrigation tapes listed in Table 1. Fig. 4 shows that the flattening coefficient increases with pressure for the six drip-irrigation tapes given in Table 1. The increase is rapid at low pressure, which indicates that the cross-sectional shape of the drip irrigation tape changes significantly in this range. When the pressure reaches the critical pressure, the increase becomes less rapid, indicating that the cross-sectional shape stabilizes. As the pressure increases further, the cross-sectional shape remains relatively constant as a function of pressure. The critical pressure for the various tapes is given in Table 4.

Numerical simulation and verification

To verify the reliability of the numerical model for the regimes of both circular and elliptical cross sections, the simulated results of the two numerical tapes were compared with the measured results. Tape B4 had the thinnest walls of the six experimental drip-irrigation tapes, so the cross-section of tape B4 under weak pressure expanded to round, and the measured results of B4 were used to verify the simulation results for a tape with circular crosssection. Tape B6 had the thickest walls so, even for larger pressure, the cross section remained flat. Thus, B6 was used to verify the simulation results for a tape with near-elliptical cross section. The remaining four sets of measured results were used to verify the fit.

Fig. 5 shows the simulated head loss as a function of measured head loss for tapes B4 and B6. The unity slope of these results indicates that the simulated head loss was consistent with the measured result. The average deviation between the measured and simulated results for tape B4 (B6) was 4.3% (8.5%). The larger average deviation for tape B6 is mainly attributed to the fact that tape B4 was circular, so the cross section varied only slightly (i.e., over a relatively small range of flattening coefficient), and the simulated size of the tape was therefore always close to the actual size. The cross section of tape B6 was elliptical, so the measured cross section of tape B6 could vary more (i.e., over a larger range of f lattening coefficient). For tape B6, the measured head loss was slightly greater than the simulated head loss. The relative average deviation was satisfactory, so the simulation results are considered to accurately reflect the measured results.

Analysis of local head loss in drip irrigation

The numerical simulation of local head loss accurately reflects the measured local head loss in drip-irrigation tape. Therefore, we used the simulation results shown in Fig. 6 to analyze the drip-irrigation tape with 18 different cross sections. Fig. 6 shows the simulated local head loss as a function of water-flow velocity for 18 drip-irrigation tapes. The arguments of Eq. (3) were logarithmically processed, and the simulation results were linearly regressed to obtain the following expression of local head loss of drip-irrigation tapes:

The correlation coefficient R2 = 0.92. Table 5 lists the results of the regression analysis, which show that the Reynolds number in the 0.05 confidence interval fell within a probability of 0.58, so the impact of the Reynolds number on local head loss was not significant. The probabilities for the remaining factors were less than 0.01, so these factors were significant. Thus, the local head loss of drip-irrigation tape was related mainly to theshape of the emitter and the lateral diameter of the tape cross section.We thus simplified Eq. (5), and using 9.8 m/s2 for the acceleration due to gravity and a water-flow viscosity co-efficient (for water at 20 °C) of 1.01×10−6, obtained

(6)

The lateral local head loss was related not only to the size of the emitter, but also to the lateral diameter and water-flow velocity. For a round cross section, the hori-zontal diameter equalled the vertical diameter (Da = Db), in which case we obtained

(7)

where D is the diameter of the cross section.To verify the accuracy of Eq. (6), we used it to cal-culate the local head loss for tapes B1, B2, B3, and B5 for both circular and elliptical cross sections. Under low pressure, the cross sections of these four drip-irrigation tapes were elliptical and their dimensions were compa-red against the simulation results for the near-elliptical condition. Under non-low-pressure conditions, the cross sections of the four drip-irrigation tapes were circular, so their dimensions were compared against the simulation results for circular tapes. Fig. 7 plots the simulation re-sults for head loss versus the measured head loss. These results show that the simulation results for the four groups of drip-irrigation tape were consistent with the measured results. The average relative deviation for drip-irrigation tapes B1, B2, B3 and B5 was 8.3% 11.3%, 9.1% and 6.7%, respectively, and the overall average relative devia-tion of the four groups was about 10%.

Fig. 8 is the same as Fig. 7, but for the near-elliptical cross section. Again, the simulation results reflected the measured trend of local head loss of drip irrigation tapes, although a certain systematic deviation appeared, in par-ticular for B3 and B5. The simulated head loss for B1 and B2 was consistent with the measured head loss (average relative deviations were 3.4% and 7.8%, respectively). The average relative deviation between the two was lessthan 10%. A larger average relative deviation was evident for tapes B3 and B5: 25.2% and 23.6%, respectively, and the average relative deviation between the two exceeded 20%.

Thus, for circular drip-irrigation tape, the simulated total head loss was consistent with measurements. For near-elliptical drip-irrigation tape, the simulated total head loss partially deviated from measurements, so the average relative deviation between the simulated and me-asured results was larger than for a circular drip-irrigation tape. This result is attributed to the cross section of the near-elliptical drip irrigation tape, which was approxima-ted as elliptical, although its real shape was certainly irre-gular rather than elliptical. In some cases, this approxima-tion assumes deviations from the actual shape of the cross section of the drip-irrigation tape, resulting in a greater deviation from the measured value. However, because the total average relative deviation of the four types of drip irrigation tape was 15%, we consider that the local head loss of each drip irrigation tape was satisfactorily approxi-mated by Eq. (6).

To analyze the distribution of pressure and velocity in the drip-irrigation tape, we monitored the transverse and longitudinal sections of an emitter. The mechanism of the entire emitter was explored by monitoring the velocity and pressure distributions of the cross section. Fig. 9a shows a small inlet flow disturbance and a large export distur-bance. The strongest disturbance upstream of the emitter was located in the area adjacent to the front end of the emitter. At the bottom of the lateral in this area appeared a large zone where the velocity was reduced in magnitude but not in direction, so the flow remained along the lateral direction. The strongest perturbation downstream of the emitter was near the end of the emitter. As can be seen from Fig. 9, the bottom of the lateral in this area was also a zone of low velocity, but the velocity was also negative, indicating that the downstream emitter caused reflux and a whirlpool. The range of the emitter head disturbance was small, whereas that of the end disturbance was large. The first section upstream of the emitter changed and the velocity distribution changed to the low-velocity regime. At the second section, however, the velocity distributionreturned to the axially symmetric distribution. At the first section downstream of the emitter, the velocity distribu-tion changed to the low-velocity recirculation zone. The low-velocity zone continued to develop in the direction of the water flow. At the seventh section, the velocity distribution had not been restored to the axially symme-tric distribution. The velocity distribution at the emitter was adjusted mainly downstream of the emitter. Fig. 9b shows that the low-pressure zone near the emitter occu-rred mainly in the top region of the tip of the emitter and in the tail region of the tip of the emitter, both of which were downstream of the cross-section deformation.

Case study

Given the expression (6) of the local head loss, we si-mulated a 60-m-long B1 φ16 drip-irrigation tape with in-tegrated in-line emitters and calculate the local head loss in an actual engineering case study. The local head loss of the drip-irrigation tape was calculated separately for a round cross section and an elliptical cross section with flattening coefficient α = 0.76. We assumed uniform dis-charge from each emitter, and each emitter was numbered consecutively starting from the end of the drip-irrigation tape. The discharge was calculated from the end of the drip-irrigation tape by using the step-by-step method. In addition, the Reynolds number was calculated from the discharge. Reynolds numbers below 2000 corresponded to laminar flow, from 2000 to 3000 they corresponded to a transition from laminar to turbulent, and above 3000 they corresponded to turbulent flow. The friction head loss was calculated by using the Darcy-Weisbach formula, the Hagen formula, and the Blasius formula, and the head-loss coefficient for the transition section was fixed at 0.04 (Provenzano et al., 2005; Palau-Salvador et al., 2006; Zi-tterell et al., 2014). The local head loss of the laminar-, transitional-, and turbulent-flow sections was calculated by using Eq. (6). In each section, friction head loss and local head loss were summed to obtain total head loss, and emitter discharge was set at 1, 2, and 4 L/h. The results appear in Table 6.

Table 6 showed that the local head loss and friction head loss of the drip-irrigation tapes varied as a function of cross-sectional shape. When the drip-irrigation tape had a flat cross section, the friction head loss and the local head loss were greater than when the tape had a circular cross section. The friction head loss increased slightly, whereas the local head loss increased by over 50%. The ratio of local head loss to friction head loss was 0.11–0.19 for elliptical tape, which was greater than the ratio 0.08–0.12 for circular tape. For both circular and ellipticalcross sections and for laminar flow, the total head loss decreased with increasing emitter discharge. For turbulent f low, the total head loss increased strongly with increasing emitter discharge. In the design of drip-irrigation tapes, the local head loss was usually estimated to be 10% to 20% of the friction head loss. According to the results presented in Table 5, when the flow pressure gives a circular cross section, the ratio of local head loss to friction head loss was about 10%, but when the cross section was elliptical, the ratio of local head loss to friction head loss fell between 10% and 20%, so we used a median value of 15% to estimate this ratio.

Seedling transporting device In order to cooperate with the seedling clamping device and seedling air loosening device, a seedling transporting device was developed in detail following the design requirements, and the function of the seedlingtransporting device was feeding the plug tray to the required position. As shown in Fig. 4, the seedling transporting device was mainly composed of a chain-type longitudinal transporting device, an air cylinder device for transversal feeding and an automatic control system with some sensors. The seedling transporting device can place two plug trays in the tray baseplate at the same time, and the trays are fixed between two leading tray plates connected with the chains. When the trays are placed into the leading tray rods, the stepper motor drives the chain to rotate, so the device can realize the longitudinal transporting. The supporting tray pipes installed in the leading tray rods are used to better fix the plug tray, so the position accuracy of the plug tray can be guaranteed.

During the first feeding process, the plug tray will move into the pressing tray plate until it reaches the picking position, so the laser beam of the laser sensor will be blocked by the front surface of the tray. Meanwhile, the f lip cylinder will drive the pressing tray device to press down the plug tray, and the seedling air loosening device will extend through the air blowing holes to make the air nozzle fit closely with the bottom of the plug hole, then the first six seedlings will be blown out from the plug tray. The pressing tray rods parallel installed in the pressing tray device are designed to press the edge of the plug holes, which not only ensures the enough downforce to the tray, but also avoids the mutual interference between the end-effectors and the pressing tray rods. Besides, the air blowing holes are designed like an ellipse, and the center of the ellipse is the same as the center of the plug hole, so the position of the plug tray and the seedling air loosening device can be precisely adjusted for better cooperation. As the first six seedlings (half of one row) are picked over, the lateral moving cylinder will drive the device to move to the right in the distance of one hole, so the remaining six seedlings can be picked out. After one row picking, the stepper motor will drive the plug tray to move down the distance of one hole for next picking process, so the seedling picking will be repeated until all seedlings in the plug tray are picked out. Finally, the empty plug tray will drop into the tray collection box, to be fed into the picking position for repeated seedling picking.

Seedling air loosening device

As the core part of the whole device, a seedling air loosening device was designed to blow out the plug seedlings from the plug tray with low damage to the substrate and the root system, which directly affects the performance of the whole device. As shown in Fig. 5, the seedling air loosening device mainly consists of an air blow system and a cylinder, both of them are driven by air force. When the seedling plug tray is fed to the seedling picking position, the flip cylinder in Fig. 4 will extend and drivethe pressing tray device to press the plug tray so that it will be in close contact with the tray baseplate, then the air cylinder in Fig. 5 will push the air blow system to make the air nozzle closely connected with the bottom of the plug tray, and the center of the air nozzle should be aligned with the center of the plug hole one by one, so the plug seedlings can be blown out successfully by the seedling air loosening device.

The number of air nozzles was designed according to the number of the holes arranged in the plug tray, so the seedling air loosening device can blow out half of one row seedlings at a time, which can ensure the efficiency and the quality of seedling picking at the same time. Actually, it requires a large pressure to blow out the seedlings, so it is necessary to ensure the airtightness of the whole device. In order to achieve this goal, we drilled some through-holes directly on the mounting plate by using laser drilling technology; the air nozzles and intake nozzles with G-thread (GB/T7307-2001) were installed at both ends of through-holes, and all of the G-threads were coated with the liquid state PTFE tape in order to improve the airtightness of the nozzles. Besides, some soft suckers were installed at the front end of the air nozzles, which can fit close to the bottom of the plug tray under the pressure, so this method can effectively prevent air leakage during the process of air blowing.

In addition to ensuring the air tightness of the whole device, the structural form and size of the air nozzle are also important to ensure the plug seedling can be blown out successfully. The investigation shows that most of the drainage outlets in the plug tray were not in the center due to a problem of the injection molding process, but the position of the air nozzles was based on the position of the drainage outlets in the standard plug tray, so the air nozzles could not be adjusted adaptively during the working process if the drainage outlets were not in the standard position. In order to ensure that the air nozzle could adapt to the drainage outlets at different places on the plug tray and successfully blow out the plug seedlings, it was necessary to select a reasonable diameter of the blowhole in the air nozzle.

As shown in Fig. 6, D is the diameter of the standard drainage outlet, and D0 is the actual diameter of the blowhole we want to work out. It is well known that PVC plastic plug trays are one-time formed by the injection molding process, and the drainage outlets in the tray are formed by the whole row stamping process, so the diameters of all drainage outlets are the same in standard, but the positions of the drainage outlets will be different (marked 3 in Fig. 6) because of the errors existing during the manufacturing process. In order to ensure the accuracy value of D0, we randomly selected some standard 72-hole plug trays and measured the length of m (Fig. 5) by using a simple random sampling method. First, we used two crosshair laser localizers to mark the center of the standard drainage outlet and the possible drainage outlet in the same tray hole respectively; then, we selected the high-accuracy Total Station to measure the length between the two crosshairs; finally, we substituted the measured value into the Eq. (1):

(1)

where mi is the distance between the center of the standard and the possible drainage outlet of the i-th tray hole.

According to the measurement results, D=7 mm and m=1.73 mm, so the blowhole diameter D0=3.54 mm. In order to select the proper size of the air nozzle, the diame-ter of the blowhole was regulated to 3.5 mm.Compared with the end-effectors inserting and clam-ping the seedlings out from the plug tray (Han et al., 2013), our device uses the air force to blow out the seed-lings. No matter which method is used to pick the seed-lings out from the tray, it is necessary to let the seedlings get rid of the adhesion of the plug tray, so we must calcu-late the air pressure acting on the bottom of the seedling to ensure that there is enough pressure to blow the seed-ling out. According to theory of subsonic free jet (Zhao & Jiang, 1998), we simplified the airflow field during the process of air blowing as shown in Fig. 7. At present, the substrate of seedling is mainly composed of peat, perliteand vermiculite in a certain proportion, so the adhesion is formed by the mixture of substrate, root system and the wall of the tray (Miao et al., 2013). Besides, it takes a lot of force to blow out the seedling, but the force should not be too large to destroy the seedling, so the air force must meet the Eq. (2):

(2)

where FS is the force required to blow out the seedling, N; G, is the gravity of seedling, N; FN, is the adhesion be-tween seedling and tray, N; FB is the critical force of seed-ling destroyed, N; P0 is the jet pressure at the air nozzle, MPa; and S is the actual area of jet action on the bottom of seedling, m2.In order to ensure the stability and pressure demand of the air system, we chose 6 mm and 8 mm air pipes to connect air compressor and pneumatic components; these air pipes are much larger than the diameter of the blowhole, so when the airflow passes through the blowho-le, it will form a contracted flow effect (Han et al., 2004), the air nozzle is similar to a throttle valve at this time (Fig. 7a). According to Bernoulli equation, we can get:

(3)

where ξ is coefficient of flow resistance; P and P0 are the pressure at the inlet and outlet respectively, MPa; v and v0 are the flow velocity at the inlet and outlet respectively,

m/s; S0 is the surface area of the blowhole, m2; S1 is the surface area of the drainage outlet, m2; and ρ is the density of air flow, kg/m3.According to the jet theory, jet core area (AO’D) mar-ked in Fig. 7b is the main effective area of air flow, and the velocity in this area is the same as v0, so we just need to calculate the jet force acted on the seedling of this area. Because of a certain thickness of the tray (marked h in Fig. 7b), the actual area which the jet acted on the bottom of the seedlings is the line marked 4 in Fig. 7b, so we can easily get the actual action area as

(4)

where r is the radius of the actual action area of the jet, mm; α is the turbulence coefficient, and the air nozzle is a cylindrical tube, so α=0.076.

Han et al. (2015) and Wang et al. (2016) showed that the adhesion force FN can be approximately equal to the detaching force FD which could separate the seedling from the tray, and Miao et al. (2013) proposed that the average detaching force of the cucumber seedling which grew in the 72-hole plug tray is 1.44 N. Besides, we randomly se-lected a group of cucumber seedlings and weighed them, so the average value of gravity was 0.21 N, and we also measured the thickness of the plug tray, so h=0.08 mm. Han et al. (2019a) got the critical force of seedling des-troyed with the help of the puncture test commonly used in food science,, so FB = (3.08 ± 0.56) N. Putting all the data together with the diameter of the blowhole into the Eqs. (2), (3) and (4), the theoretical jet pressure (P0) at the air nozzle was between 0.146 and 0.315 MPa. Accor-ding to the established jet pressure, we can also find out the required outlet pressure of the air compressor and the solenoid valves.

Seedling clamping device

The function of the seedling clamping device is to clamp the seedlings blown out from the seedling air loose-ning device, transporting them into the seedling dropping position and then dropping them into the transplanting device. As shown in Fig. 8, the seedling clamping devi-ce mainly consists of whole row end-effectors, pushing cylinders, flip cylinders and the frame, and the end-effec-tors are the core components of the seedling clamping de-vice which will influence the quality of seedling picking and dropping, so it is important to design and calculate the parameters of the end-effectors in detail.

Compared with the end-effectors developed by Han et al. (2013) and Tong et al. (2019), our end-effector will pick the seedling by clamping the surface around the seedling body, which can avoid the disturbance of the root system and the damage of the seedling body caused by the insertion of the needles.

The working principle of the end-effector is shown in Fig. 9. At first, the end-effector is in the initial state, when a work cycle starts, the seedling air loosening de-vice blows out the seedling from the plug tray (Fig. 9a). As the seedling moves into the range of the grippers, the pushing cylinder gets the signal so that the pull rod will retract and drive the pull plate to move upward, and the grippers contract inward to clamp the seedlings (Fig. 9b). Then the end-effector is driven by the flip cylinder and quickly moves into the dropping position; after that, the cylinder drives the pull rod to extend, so the gripper opens and drops the seedlings (Fig. 9c).

In order to ensure the success rate of seedling picking, it is important to design and optimize the parameters of the end-effector. The simplified designed structure of the end-effector is shown in Fig. 10. By analyzing the struc-ture diagram, we can get the distance between installation points of two picking needles as:

(5)

And then we can get the length of the pull plate as:

(6)

Actually, the main parameters of the end-effectors are mostly determined by the seedling sizes and features. Ac-cording to Ma et al. (2020b), the upper side length a=40 mm, the lower side length b=20 mm, and the cone an-gle of one plug hole β=22.5°. Besides, the height of plug seedlings suitable for transplanting is generally about 100-200 mm and, in order to successfully clamp the seed-lings, we should ensure seedling clamping exceed the biological yield point (compression quantity: 3.46 mm) to provide enough clamping force (Han et al., 2013). So we can easily get the restraint conditions as:

(7)

As shown in Table 1, combining Eqs. (5), (6) and (7), we can get the parameters of the key components in the end-effector, and we designed the end-effector(Fig. 8) in detail according to the parameters. We used the 3D printing technology to build the grippers, which can better protect the seedlings and ensure the accuracy of the grippers. Each gripper has one threaded hole, so we can fasten the gripper and the needle by screws, and also adjust the height of the grippers flexibly according to the demand. Besides, we also installed some limit blocks (marked 7 in Fig. 8) in the pull rod, so the end-effector can pick the seedlings of different sizes by adjusting the position of the limit blocks.

Automatic control system

As the most important part in the pick-up device, the automatic control system manages and coordinates the operation of various parts, to ensure the automation and the normal operation of the device equipped with bothelectromechanical and pneumatic units. As shown in Fig. 11, the automatic control system is composed of the IPC (industrial computer), a mobile workstation, some sensors, some solenoid valves and some other executive elements. As the center of the whole system, the IPC consists Figure 8. Structure diagram of the seedling clamping device: 1, plug seedling; 2, picking needle; 3, pull plate; 4, pull rod; 5, swing shaft; 6, fixed plate; 7, limit block; 8, end-effector assembling plate; 9, pushing cylinder; 10, pushing cylinder assembling plate; 11, rotation axis; 12, roller; 13, chute; 14, flip cylinder; 15, flip cylinder assembling plate; 16, flip plate; 17, frame. (Fig. 8) in detail according to the parameters. We used the 3D printing technology to build the grippers, which can better protect the seedlings and ensure the accuracy of the grippers. Each gripper has one threaded hole, so we can fasten the gripper and the needle by screws, and also adjust the height of the grippers flexibly according to the demand. Besides, we also installed some limit blocks (marked 7 in Fig. 8) in the pull rod, so the end-effector can pick the seedlings of different sizes by adjusting the position of the limit blocks. Automatic control system As the most important part in the pick-up device, the automatic control system manages and coordinates the operation of various parts, to ensure the automation and the normal operation of the device equipped with both Figure 9. Principle diagram of the end-effector. of some relays, switches and some other components, it receives and transmits signals from the sensors, switches and some judgement parts, so as to control the executive elements to complete the specified actions. The photoelectric sensor is used to monitor the position signal of the plug tray so as to control the rotation and stop the stepping motor. Each pneumatic element is equipped with a solenoid valve and each air circuit has a throttle valve, they are all connected to relays to realize precise control. Following other pick-up devices, we installed magnetic switches on each cylinder, so the system could precisely control the extension and retraction position of the cylinder. Besides, the two flip cylinders were connected to the same solenoid valve, so the two flip cylinders could realize synchronous operations. A mobile workstation was needed so we could easily monitor the real-time running status of the system, and we could also set some key parameters such as picking speed, flow rate, start & off, and so on. Besides, the program was edited in Visual Studio 2018 to run in the workstation, and we also added analog quantity in the program to display the real-time flow rate of the air nozzle from the digital display flowmeter, so we could ensure the accuracy of the seedling loosening.

As shown in Fig. 12, the whole pneumatic system was edited by FluidSIM-P 3.6. All double-acting cylinders were connected with the 2-position, 5-way solenoid valves (Sv1-Sv10); all air nozzles were connected with the 2-position, 3-way solenoid valves (Sv11-Sv16); and all solenoid valves were controlled by the relays. The flip cylinders C1 and C2 were connected to the same solenoid valve, used to swing the end-effectors to the seedling picking or dropping position; the cylinder C3 was used to move the seedling transporting device left and right; C4 was used to pressure the plug tray during the seedling picking process; C5 to push the air nozzles to the blowing position, and C6-C11 to drive the end-effectors opening and closing. The air nozzles N1-N6 were used to blow out the seedlings from the plug tray, and each air nozzle was connected with a digital display flowmeter (P1-P6) to monitor the real-time blowing pressure. Each air path was connected with a throttle valve, so the whole system could adjust different pressures and flows according to different recommendations.

Performance test

According to these data, a prototype of the automatic pick-up device was produced, and an indoor experiment was arranged in order to test the working performance ofPerformance test According to these data, a prototype of the automatic pick-up device was produced, and an indoor experiment was arranged in order to test the working performance ofthe effect of asymmetry of grippers, the two strain gauges were connected to the strain gauge through a half brid-ge circuit (Fig. 13b), and two strain gauges were sym-metrically arranged on the inside and outside of a single gripper (Fig. 13c). After connecting each instrument cir-cuit, we calibrated the dynamic strain measuring sys-tem following Wang & Hu. (2020) (Fig. 13d), and then tested the stress of the seedling during the process of seedling clamping.

Previous studies have shown that water content affects the adhesion between seedlings and plug tray (Ting et al., 1990; Miao et al., 2013), so we tested the effect of the seedling air loosening at different water contents: 50-55%, 55-60%, 60-65%. The air pressure was also the key to the success rate of seedling picking, accor-ding to the above calculation; the theoretical air pressu-re required was between 0.146 MPa and 0.315 MPa, so three levels (0.3 MPa, 0.4 MPa and 0.5 MPa) were se-lected for the air pressure considering the pressure loss in the pneumatic circuit. Besides, the value of airflow also affects the success rate of seedling picking, so we insta-lled a throttle valve in each circuit to adjust the airflow, and we selected 50%, 75% and 100% airflows. In order to ensure that the seedlings are successfully blown out from the tray, we set the solenoid valves (Sv11-Sv16) with a de-lay of 0.2 s, and the we selected a 0.8 MPa air compressor to ensure the stability of airflow. The picking frequency was set at 120 plants/min.

As shown in Table 2, A, B and C were determined as the coding values of air pressure, airflow rate and water content, respectively. An L9(33) orthogonal table was selected in the orthogonal experiments, and each group selected one tray (72 seedlings), taking the ave-rage value as the result data of each group, the seed-ling picking test is shown in Fig. 14. The experimen-tal results were analyzed by SPSS 23.0 with general linear model.Ryu et al. (2001) indicated that the seedlings are not able to survive when their damage degree was > 1/4, so we defined the seedling damage degree (Sd) as:

(8)

where Wcs is the weight of the seedling after blowing out, g; and Wms is the weight of the substrate remained in the tray and scattered outside, g.Taking these factors into consideration, the success ra-tio of seedling picking (Rsp) was defined as

(9)

where Nts is the number of seedlings in total, Nms is the number of missing seedlings, Nds is the seedlings with stem and leaf injury, and Nfs is the seedlings with more than 1/4 damage.

The strain gauges were attached to the grippers, and we fixed one gripper to calibrate the strain gauge (Fig. 13d), then we returned the strain instrument to measure the calibration value. We applied forces to the gripper, increasing 1 N each time, from 0 N to 8 N, then we used the least square method to calculate linear fitting of the data. The fitting results (Fig. 15a), show that the linear relationship between the input voltage (mV) and the applied force (N) was y=56.876x-0.954. After the strain gauge was calibrated, we installed the grippers with strain gauges in the end-effector, and started the seedling clamping device to clamp the seedling, the input voltage was collected all the time during this period until the grippers were completely closed (Fig. 15b).

After repeated clamping for three times, we got the average input voltage: 158.906 mV, so the corresponding compressive force on the seedling was 2.805 N. Ren et al. (2020) proposed that tomato seedlings substrate will start to damage when the pressure reaches 3.15 N, so the new designed end-effector can meet the needs of automatic seedling picking.

According to the experiments, the results of the performance test for the pick-up device, results of range analysis, and results of variance analysis are shown in Tables 3, 4 and 5 respectively. As shown in Table 4, the influence on the success ratio ranked A> C> B, and when air pressurewas 0.3 MPa, airflow rate 100%, and water content between 55% and 60%, the highest success rate of seedling picking was obtained. Table 5 shows that the success ratio of seedling picking was extremely significantly affected by air pressure (p<0.01), significantly affected by the water content (0.01<p<0.05), but not significantly affected by the airflow rate (p>0.05). According to the range analysis, the air pressure extremely significantly affected the success ratio of the pick-up device by adopting the seedling air loosening method, and the success ratio could reach 94.9% when air pressure was 0.4 MPa. Although the calculated air pressure was between 0.146 MPa and 0.315 MPa, the success ratio sharp declined to 87.49% when air pressure was 0.3 MPa. The reason is that some of the seedlings have advanced root systems, so the adhesion between these seedlings and tray is too strong, and there is a certain degree of loss in the air circuit, thus the air pressure is too low to blow out these seedlings from the tray, bringing out the low success ratio. When air pressure reached 0.5 MPa, almost all the seedlings could be blown out from the tray, but some seedlings with lighter weight will collide with the end-effector after the blowing because of the overpressure from the air loosening device, which will destroy the substrate of the seedling and sometimes break off the item and the leaf of the seedling. Besides, although a small part of the seedlings with poor root systems can be blown out, too much airflow will blow away part of their substrates, which will also affect the success ratio. Because of the particularity of air blowing, it is important to choose a suitable air pressure to ensure the success ratio of seedling picking. Previous studies showed that water content will affect the strength of the seedling substrate, and also affect the adhesion between seedlings and plug tray (Miao et al., 2013; Han et al., 2015). The test showed that the best success ratio (92.81%) was found when the water content was 55-60%. When the water content was 60%-65%, the success ratio declined to 89.9%, and the seedlings maintained relatively integrity. Yang et al. (1991) proposed that proper water content could reduce the adhesion between seedlings and tray; our experiment showed that when the water content was > 60%, the success ratio declined because the adhesion continued increasing. But when the water content was 50%-55%, the success ratio was 88.27%; the reason may be that the substrate is loose under this water content, especially when the water content is close to or even less than 50%, so the substratewill be broken easily by the airflow, although the seed-ling can be blown out under this water content, the va-lue of transplanting is greatly reduced. The substrate is mainly composed of peat, vermiculite and perlite (Miao et al., 2013), then the water content will significantly affect the physical properties of the substrate, so it is important to control the water content of the seedlings in a reasonable range.

Besides, the effect of the airflow rate on the success ratio is in a rising tendency on the whole, and the success rate reached 91.74% when the airflow rate was 100%. During the experiments, it was found that the time of airblowing was longer with the increase of the airflow rate, so full opening of throttle valve will bring higher success rate of seedling picking in the process of air blowing.

The parameters of the pick-up device were adjus-ted according to the results of the optimal combination from the orthogonal tests, then the performance of the seedling pick-up device was tested and verified. The fi-nal results are shown in Table 6. The tests showed that the average success ratio of the new pick-up device was 96.64%, which proves the practicability of the new de-signed pick-up device, and the air loosening method was also proved. Actually, the correctness of the four-claw end-effectors used in most pick-up devices have been proved in previous studies (Yang et al., 1991; Han et al., 2013; Tong et al., 2019), but the way of inser-ting the substrate to pick the seedlings will eventually cause some damage to the seedlings, so we improved the end-effectors on the basis of these previous studies which will pick the seedling by clamping the surface around the seedling body. With the cooperation of the new designed end-effectors and the air loosening me-thod, the pick-up device could reduce the damage rate of the seedlings and improve the success rate of seedling picking while realizing high-speed working. Besides, we only used 72-hole plug trays in our paper, but the pick-up device could also pick seedlings from 50-hole or 128-hole plug trays by changing the number of the end-effec-tors, and the auto-control system will also match with different plug trays. For the particularity of pneumatic driving mode, we installed a throttle valve in each in-dividual pneumatic path, so we can adjust the airflow independently under different requirements.

The parameters of the pick-up device were adjus-ted according to the results of the optimal combination from the orthogonal tests, then the performance of the seedling pick-up device was tested and verified. The fi-nal results are shown in Table 6. The tests showed that the average success ratio of the new pick-up device was 96.64%, which proves the practicability of the new de-signed pick-up device, and the air loosening method was also proved. Actually, the correctness of the four-claw end-effectors used in most pick-up devices have been proved in previous studies (Yang et al., 1991; Han et al., 2013; Tong et al., 2019), but the way of inser-ting the substrate to pick the seedlings will eventually cause some damage to the seedlings, so we improved the end-effectors on the basis of these previous studies which will pick the seedling by clamping the surface around the seedling body. With the cooperation of the new designed end-effectors and the air loosening me-thod, the pick-up device could reduce the damage rate of the seedlings and improve the success rate of seedling picking while realizing high-speed working. Besides, we only used 72-hole plug trays in our paper, but the pick-up device could also pick seedlings from 50-hole or 128-hole plug trays by changing the number of the end-effec-tors, and the auto-control system will also match with different plug trays. For the particularity of pneumatic driving mode, we installed a throttle valve in each in-dividual pneumatic path, so we can adjust the airflow independently under different requirements.

In conclusion, a whole row automatic pick-up devi-ce for vegetable plug seedlings was developed based on the current demand of vegetable production in China. The device, with seedling air loosening method, com-bines mechatronics with pneumatics system, which can reduce the damage rate of seedlings while realizing au-tomatic high-speed seedling picking and dropping. As a new pick-up device, the prototype was produced, and the performance tests were conducted under actual pro-duction conditions. The test results showed that the suc-cess ratio of seedling picking could exceed 95%, and the picking and dropping efficiency could well meet the demand of automatic transplanting. It is still neces-sary to improve the quality of seedling raising in the future, so that the pick-up device will ensure a higher success ratio.

Chen J, Wu J, Ye J, Fu X, Fang X, 2014. Dynamic modeling and analysis of seven-bar seedling planting mechanism of transplanting machine. Appl Mech Mat 574: 230-238. https://doi.org/10.4028/www.scientific.net/AMM.574.230Ferrari, 2018. Ferrari Futura automated transplanter [EB/OL]. https://ferraricostruzioni.com/en/tray-planters/8-futura-automated-transplanter.htmlGao G, Wang K, Sun X, 2017. Verification for EDEM simulation of process of jacking tray-seedling by steel needle in grafting machine and parameter optimization. T CSAE 33 (21): 29-35.Han FQ, Xiao YX, Kubota T, Liu J, 2004. Study on the contraction mechanism of free jet. J Eng Thermophys 25 (3): 421-423.Han L, Mao H, Miao X, Hu J, Yang X, 2013. Design of automatic picking up seedling end-effector based on mechanical properties of plug seedlings. T CSAM 44 (11): 260-265.Han L, Mao H, Yan L, Hu J, Huang W, Dong L, 2015. Pincette-type end-effector using two fingers and four pins for picking up seedlings. T CSAM 46 (7): 23-30.Han L, Mao H, Zhao H, Liu Y, Hu J, Ma G, 2019a. Design of root lump loosening mechanism using air jets to eject vegetable plug seedlings. T CSAE 35 (4): 37-45.Han L, Kumi F, Mao H, Hu J, 2019b. Design and tests of a multi-pin flexible seedling pick-up gripper for automatic transplanting. Appl Eng in Agr 35 (6): 949-957. https://doi.org/10.13031/aea.13426Jin X, Li D, Ma H, Ji J, Zhao K, Pang J, 2018. Development of single row automatic transplanting device for potted vegetable seedlings. Int J Agr Biol Eng 11 (3): 67-75. https://doi.org/10.25165/j.ijabe.20181103.3969Khadatkar A, Mathur SM, Gaikwad BB, 2018. Automation in transplanting: a smart way of vegetable cultivation. Curr Sci 115 (10): 1884-1892. https://doi.org/10.18520/cs/v115/i10/1884-1892Kumar GVP, Raheman H, 2011. Development of a walk-behind type hand tractor powered vegetable transplanter for paper pot seedlings. Biosyst Eng 110 (2): 189-197. https://doi.org/10.1016/j.biosystemseng.2011.08.001Liu Y, Mao H, Han L, Xu J, Ma G, 2018. Plug damage detection and parameter optimization of picking up cucumber seedlings from tray cells based on Micro-CT. T CSAE 34 (12): 27-34.

Ma G, Mao H, Bu Q, Han L, Abdul S, Gao F, 2020a. Effect of compound biochar substrate on the root growth of cucumber plug seedlings. Agronomy 10 (8): 1080. https://doi.org/10.3390/agronomy10081080Ma G, Mao H, Han L, Liu Y, Gao F, 2020b. Reciprocating mechanism for whole row automatic seedling picking and dropping on a transplanter. Appl Eng Agr 36 (5): 751-766. https://doi.org/10.13031/aea.13622Miao X, Mao H, Han L, Sun H, Yang X, 2013. Analysis of influencing factors on force of picking plug seedlings and pressure resistance of plug seedlings. T CSAM 44 (1): 27-32.National Bureau of Statistics of China, 2019. http://data.stats.gov.cn/easyquery.htm?cn=C01Ren Z, Wang W, Feng J, Yu J, 2020. Experimental study on mechanical properties of plug seedlings. J Chin Agr Mechan 41 (04):169-174.Ryu KH, Kim G, Han JS, 2001. Development of a robotic transplanter for bedding plants. Agr Eng Res 78 (2): 141-146. https://doi.org/10.1006/jaer.2000.0656Thomas EV, Kumar U, 2017. Kinematic analysis of transplanting mechanisms of selected mechanical rice transplanters. J Agr Eng 54 (3): 1-11.Ting KC, Giacomelli GA, Shen SJ, 1990. Robot workcell for transplanting of seedlings: part Ⅱ- end-effector development. T ASAE 33 (3): 1013-1017. https://doi.org/10.13031/2013.31431Tong J, Shi H, Wu C, Ding Y, Zhao X, Wang R, 2019. Simulation and test of seedling pot grabbing by spade end-effector. T CSAM 50 (8): 107-116.Wang D, Jin X, Ji J, He Z, Yan H, 2016. Design and experiment of fully automatic ejection and picking-up seedlings mechanism of transplanter. J Agr Mechan Res 10 (10): 64-69.Wang Z, Hu X, 2020. Research on deflection measurement of helicopter blade based on strain gauge method. Chin Meas Test 46 (4): 42-46.Xiang W, Luo X, Wang Y, 2007. Experiments on pneumatic ordered throwing transplantation of paddy seedling. T CSAM 38 (7): 67-70.Yang Y, Ting KC, Giacomelli GA, 1991. Factors affecting performance of sliding-needles gripper during robotic transplanting of seedlings. T ASAE 7 (4): 493-498. https://doi.org/10.13031/2013.26251Ye B, Tang T, Yu G, Yi W, Zhang G, 2019. Dynamic analysis of rotary seedling pick-up mechanism of vegetable transplanting machine with counterweight. T CSAM 50 (8): 117-122．

Zhang Z, Wang J, Zhang W, Zhao X, Liu W, 2011. Kinematic analysis and performance experiment of cantilever cup vegetable transplanter. T CSAE 27 (11): 21-25.Zhang Z, Cao W, Wang Q, Zhang P, 2013. Development status and prospect of plug seedlings automatic transplanting machine. J Agr Mechan Res 5 (5): 237-241.Zhao C, Jiang Y, 1998. Gas jet dynamics. Beijing Institute of Technology Press, Beijing, China.