Design and optimization of a negative pressure rotary cutting safflower filaments harvesting device

Abstract

During the safflower filament harvesting process, the harvesting device plays a pivotal role in determining overall efficiency. This study addresses the current challenges in harvesting devices, such as their bulky structure and the poor cutting efficiency of the blades, by developing a bio-inspired harvesting blade modeled after the pharyngeal teeth of the grass carp. By focusing on the structure of the pharyngeal teeth of the grass carp, digital image processing techniques were employed to extract its geometric features. Regression equations were then formulated to accurately fit both the contour and tooth profiles. Based on these fitted results, a bio-inspired blade design was proposed. To evaluate the blade’s performance, simulations of the cutting process for safflower filaments were carried out using LS-DYNA software. These simulations compared traditional blades, bio-inspired contour blades, and bio-inspired tooth-shaped blades. The preliminary simulation results suggest that the bio-inspired blade exhibits significant advantages in cutting safflower filaments. Further testing was conducted using a dedicated safflower filament cutting performance test platform to compare and analyze the influence of different contour curves and blade edge tilt angles on harvesting efficiency and damage rates. The experimental results revealed that when the blade edge tilt angle ranged from 10° to 30°, the bio-inspired tooth-shaped blade outperformed the bio-inspired contour blade, which, in turn, demonstrated superior performance compared to the traditional contour blade. Specifically, the bio-inspired tooth-shaped blade achieved an average harvesting efficiency of 99.83%, the bio-inspired contour blade 97.68%, and the traditional contour blade 96.51%. To further assess the impact of the bio-inspired tooth-shaped blade’s design parameters on cutting performance, both single-factor and multi-factor experiments were conducted. The results showed that, with a rotational speed of 61.39 rpm, an intake air velocity of 2.44 m/s, and a blade edge tilt angle of 15.32°, the harvesting device reached optimal performance, achieving a harvesting efficiency of 99.97% and a damage rate of 1.53%. Subsequent repeated tests under these optimal conditions yielded an average harvesting efficiency of 98.95% and a damage rate of 1.54%. The relative deviation between these experimental results and the response surface optimization outcomes was less than 2%, thus fully meeting the practical requirements for safflower filament harvesting. The findings of this study not only provide valuable insights for the development of intelligent safflower filaments harvesting robots but also offer a theoretical foundation for the design of bio-inspired blades in agricultural mechanization fields.

Introduction

Safflower (Carthamus tinctorius), an annual herbaceous plant belonging to the Asteraceae family and commonly referred to as safflower thistle, holds significant medicinal value due to the pharmacological properties of its filaments, which promote blood circulation and exhibit anti-tumor, anti-inflammatory, and analgesic effects^1,2. In China, safflower cultivation is primarily concentrated in Tacheng, Emin, and Musar regions of Xinjiang, accounting for over 80% of the national production³. As living standards have improved, the demand for safflower products has risen, leading to an expansion of cultivation areas. However, due to the short harvesting period, low mechanization, and labor-intensive nature of manual harvesting, significant crop losses occur as a result of untimely or incomplete harvesting. Therefore, researching mechanized harvesting solutions is essential for improving efficiency^4,5. Currently, few harvesting machines are available for safflower filament collection, and those that do exist are inefficient and poorly mechanized, highlighting the need for the development of an intelligent, highly efficient harvesting robot.

The safflower harvesting end-effector is the primary working component of the harvesting device, and the method of harvesting significantly influences the quality and efficiency of filament collection. To address these concerns, researchers have studied various harvesting methods and structures, including cutting, pulling, and roller-brush techniques. Zhang Zhenguo et al.⁶ designed a double-acting knife end-effector based on low-speed sectional cutting, optimizing its working parameters through field testing. Guo Hui et al.⁷ developed a reciprocating push-pull harvesting device using a crank-slider mechanism, conducting tests to evaluate its performance. Zhang Zhenguo et al.⁸ also developed a circular arc asymptote-type harvester, utilizing Fluent software to analyze the airflow within the harvesting chamber, followed by second-order orthogonal rotation experiments that yielded promising results. Chen Jun et al.⁹ applied rotary cutting with three rotating blades to collect filaments, while Sun Chao et al.¹⁰ designed a vertical roller-brush harvester specifically for dry safflower collection. Ge Yun et al.¹¹ developed a three-fingered pulling harvester, confirming its practical use in safflower harvesting robots. He Huanhuan et al.¹² simulated human gripping and pulling actions, designing flower-holding plates to grasp and pull filaments.

Given the softness of safflower filaments and the complex harvesting environment, most harvesters utilize negative-pressure fans to collect the separated filaments. Chen Fei et al.¹³ optimized airflow dynamics within the harvesting chamber to enhance filament collection. Zhang Zhenguo et al.¹⁴ modeled the biomechanics of safflower filaments using discrete element methods, providing valuable design references for harvesting machinery. In previous work, researchers have focused on cutting end-effectors for safflower filament harvesting^15,16. However, these designs still suffer from significant damage to the filaments and flowers, which negatively impacts the secondary harvesting process. In summary, although there has been extensive research on safflower filaments harvesters, current technologies continue to face challenges, such as large structural sizes, suboptimal harvesting efficiency and damage rates, as well as the inability to harvest both dry and wet safflower simultaneously. With the continuous development of bionic technology, an increasing number of scholars have discovered that bio-inspired structures can significantly reduce cutting resistance and energy consumption. Consequently, the development of safflower filaments harvesting devices with bio-inspired structures has emerged as a promising new direction for improving harvesting efficiency.

In the study of bio-inspired, Hao Gan et al.¹⁷ explored the energy consumption of three different blade shapes-straight, angled, and serrated-during the harvesting of reed grass. A series of energy consumption experiments were conducted, and the results demonstrated that the serrated blade exhibited the best energy efficiency during the cutting process. Yunlong Cao et al.¹⁸, using the upper jaw of a locust as a bio-inspired model, fitted contour curves and designed a multi-tooth cutting tool for the xylem of broccoli based on the cutting teeth of the locust’s upper jaw. This tool improved stability during the cutting process and effectively reduced tool wear. Zhe Du et al.¹⁹ took the upper jaw of a cricket as a bio-inspired prototype to design a bio-inspired tea-leaf cutting tool, and the experimental results showed that the bio-inspired tool significantly reduced cutting resistance and power consumption. Jinpeng Hu et al.²⁰ employed three-dimensional point cloud reconstruction and machine vision methods to extract edge features from the upper jaw of the Manila locust and conducted cutting tests, which revealed that the bio-inspired blade exhibited lower average cutting force compared to conventional blades. Kunpeng Tian et al.²¹ extracted the cutting tooth profile curve from the cheek teeth of the white-cheeked crested macaque and designed a bio-inspired cutting blade. Comparative experiments with standard blades showed that the bio-inspired blade performed excellently in terms of cutting force and energy consumption. Han Tang et al.²² extracted and optimized the tibial curve of a mantis forelimb and applied it to the cutting edge of both single-disk rotary and linear fixed blades. The experimental results indicated that the optimized bio-inspired tool outperformed conventional tools in cutting performance. These successful studies provide valuable insights for the design of cutting blades for safflower filaments harvesters. Many organisms in nature have evolved optimized morphological features to adapt to their environments, such as enhanced biting, tearing, or cutting capabilities, which effectively process plant-based food. For instance, the grass carp, through long-term natural evolution, has developed a unique pharyngeal tooth structure that allows it to easily cut aquatic plants and effectively chew and grind them. Since safflower is similar to aquatic plants in being a flexible plant, the design of the grass carp’s pharyngeal tooth structure can be adapted to develop a bio-inspired cutting blade suitable for safflower filaments harvesters.

In summary, safflower filament harvesting currently relies predominantly on manual labor, with mechanical harvesting efficiency remaining relatively low. The primary reason for this inefficiency stems from structural design flaws in existing harvesting devices, such as their large size and inadequate blade performance, resulting in a low collection rate of safflower filaments. Additionally, the devices cause considerable damage to the filaments during the harvesting process, thereby negatively affecting the quality of the safflower filaments. Currently, the technology has not yet achieved integrated harvesting of both wet and dry safflower. To address these shortcomings, this paper proposes a biomimetic safflower filament harvesting device inspired by the pharyngeal teeth of grass carp. First, the device was designed to align with the growth characteristics of safflower, and the pharyngeal teeth structure was analyzed through microscopic imaging and software analysis. The teeth’s shape was extracted and optimized using polynomial fitting methods, while MATLAB was employed for image threshold segmentation, extraction, and stitching to determine the optimal blade contour. A finite element model of safflower filaments was then established to simulate the cutting effects of traditional and biomimetic blades using LS-DYNA software. Finally, a test bench was constructed to assess the effects of various blades and edge angles on filament cutting, determining the optimal operational parameters for the harvesting device.

Materials and methods

The structure and working principle of the safflower filaments harvester

This study applies the principles of bionics to design a flower-gathering port inspired by the morphology of safflower fruit and flowers. The blade curvature of the rotary cutter also follows bionic principles. The structure of the rotary cutting safflower harvesting device is shown in Fig. 1. A stepper motor powers the rotary blade, controlling its rotational angle. The mounting frame connects the harvesting device to the robotic arm and links the stepper motor to the flower-gathering chamber. The chamber concentrates the vacuum force generated by the fan onto the flower-gathering port, enhancing the vacuum strength at the port. Under this vacuum, safflowers beneath the device are drawn into the port. The rotary blade, located between the upper end of the port and the lower end of the chamber, cuts the safflower filaments as it rotates.

Fig. 1.

Schematic diagram of the rotary cutting safflower harvesting device. (1) Stepper motor. (2) Mounting frame. (3) Flower-gathering chamber. (4) Support. (5) Rotary blade. (6) Flower-gathering port. (Image source: Created by the author; Software version: SolidWorks2023; URL: https://www.solidworks.com/)

The working principle is as follows: the vacuum force generated by the fan is directed to the port through the chamber. Filaments are drawn into the chamber via the port, while the fruit remains inside. The stepper motor controls the rotary blade, which rotates at a specified angle and speed to cut the filaments at the junction between the fruit and filaments. The vacuum transports the cut filaments through a windpipe to subsequent devices for separation and collection, enabling efficient harvesting.

Design and optimization of biomimetic blades

Grass carp are herbivorous fish that primarily feed on cellulose-rich aquatic plants, distinguishing their pharyngeal teeth from those of other fish species. Located at the junction between the center-rear of the head and the body, grass carp have a pair of pharyngeal bones, one on each side. Each bone contains two rows of pharyngeal teeth: the first row has four or five large teeth, while the second row consists of much smaller ones. The surfaces of these teeth are lined with numerous ridges and grooves, making them well-suited for grinding plant fibers. The chewing process of these teeth is complex, as the teeth on both sides repeatedly meet and separate to cut the plants and then grind them into fine particles. As a result, the cross-sectional contour of the grass carp’s pharyngeal teeth exhibits excellent cutting performance²³. Due to the complexity of the grass carp’s feeding behavior and building on previous research²⁴, this study focuses solely on the cutting efficiency of the pharyngeal teeth. The grass carp’s plant-shearing motion was simplified into a cutting model. Using a segmented design approach, the outline curve of the tooth crown was extracted and fitted to develop a biomimetic blade with a contour curve applied to a rotary cutting knife. Additionally, the edge profile of the chewing surface was extracted and fitted to create a novel biomimetic tooth-profile cutting blade.

As shown in Fig. 2a (Image source: Provided by the Food Science Research Institute of Xinjiang Institute of Technology), each grass carp has one pharyngeal bone on either side. To enhance the accuracy and precision of subsequent curve extractions, we selected the more intact right pharyngeal bone as the specimen for further study. As illustrated in Fig. 2b, each pharyngeal bone features two rows of teeth, with the larger teeth located in the first row. For ease of feature extraction, one of the larger teeth from the first row was chosen as the subject of this study, as depicted in Fig. 2c.

Fig. 2.

Grass carp tooth outline diagram. (a) Complete pharyngeal bone; (b) Single pharyngeal bone; (c) Tooth outline diagram.

The intricate structure of the pharyngeal region in grass carp plays a crucial role in their feeding process, substantially enhancing the chewing function. Due to the small size of the pharyngeal teeth, an ASV0880-HK830 microscope was used to capture images of the teeth from a cross-sectional perspective, as shown in Fig. 3a. To obtain a complete outline within the limited field of view, the posterior curve of the pharyngeal teeth was aligned approximately parallel to the microscope window. The teeth were rotated counterclockwise by 14.6° to capture the full structure. The measurement results, presented in Fig. 3b, reveal a total tooth surface length of 9.26 mm, with an average ridge spacing of 0.53 mm, a maximum ridge spacing of 0.84 mm, and a minimum ridge spacing of 0.39 mm, gradually decreasing from the outer to the inner regions.

Fig. 3.

Geometric parameter measurements of pharyngeal teeth. (a) Microscope image; (b) Measurement of ridge spacing on pharyngeal teeth.

Extraction and feature analysis of biomimetic curves

To extract the curve of the tooth crown, a 1:1 biomimetic prototype was structurally analyzed using image processing techniques. The collected images were initially processed in Photoshop to remove noise and enhance the tooth outline. Subsequently, MATLAB R2022B was used to extract point data from the outline, generating coordinates and calibrating the scale between image pixels and actual measurements from the microscope. In the images, 827 pixels correspond to 1 mm. To improve curve accuracy, various polynomial fittings were tested, with the seventh-order polynomial providing the best fit. The resulting polynomial is shown in Eq. (1), and the goodness of fit, R², in Eq. (2), exceeds 0.9669, indicating that the fitted curve accurately represents the pharyngeal tooth contour. The specific technical approach is illustrated in Fig. 4, and the fitted outline curve is displayed in Fig. 5.

1

Fig. 4.

Technical route for biomimetic curve extraction and fitting.

Fig. 5.

Fitted outline curve.

Where, p₁ = 0.0000306863349728423, p₂ = − 0.00119140862307744, p₃ = 0.0183890591403376, p₄ = − 0.145967172698205, p₅ = 0.632933154131689, p₆ = − 1.39456492214706, p₇ = 1.29121770732939, p₈ = − 2.70801489412505.

2

Extraction and feature analysis of cutting blade shape

Given the structural differences between the blade and the pharyngeal teeth of grass carp, a more detailed analysis of the crown shape of the pharyngeal teeth is necessary. The crown shape curve can be viewed as a combination of multiple curve segments. In the specimen images collected, the midpoint of the concave arc is used as the division point, and the fitted coordinate origin divides the crown shape curve into 15 segments (one curve near the root, which has a smaller curvature and lacks prominent features, is excluded from this extraction). MATLAB R2022B was used for threshold segmentation, extraction, and further processing of the microscope images. The technical approach is shown in Fig. 4. During curve fitting, polynomial models of varying orders exhibited significant differences in goodness-of-fit (R²), emphasizing the importance of selecting an appropriate polynomial order to improve the fitting accuracy. Polynomial models of orders 2 to 9 were applied to fit 14 curves. By comparing the R² values across different polynomial orders, a fifth-order polynomial was chosen to balance fitting accuracy and overall effectiveness. With the exception of curves 2, 4, and 9, the R² values of the remaining curves exceeded 0.9, effectively capturing the original curve shapes. The original and fitted curves were plotted in the same coordinate system, as shown in Fig. 6.

Fig. 6.

Pharyngeal tooth shape curve and fitted curve.

During the feeding process of grass carp, the first pharyngeal tooth at the front is primarily responsible for cutting food, making it sharper, while the subsequent teeth are relatively smaller and less distinct. To further compare the cutting performance of the tooth shape curves and select the sharpest curve with the smoothest cutting surface for use as the tooth profile of a rotary blade, the second derivatives and curvatures of the 14 extracted curves were analyzed. The second derivative indicates the smoothness of the curve; a more gradual and continuous second derivative suggests a smoother curve, which should result in a smoother cut through plant fibers. A larger curvature (i.e., a smaller radius of curvature) corresponds to a smaller contact area at the blade edge, leading to higher pressure on a single filament during cutting, making the cutting action sharper.

Since the tooth curves were divided at the concave arcs, higher second derivative and curvature values tend to appear at both sides of the curve. However, for cutting safflower filaments, the primary cutting action is exerted by the curve at the tip of the tooth, so this study focuses on the curves near the tooth tips. As shown in Fig. 7, curves 1, 2, 4, 5, 6, 7, 9, 11, 12, and 13 have similar second derivatives and curvature changes. When comparing curve 1 with curve 2, curve 1 exhibits smaller variations in its second derivative and a significantly lower curvature, making it less sharp as it is located at the foremost part of the pharyngeal tooth and is responsible for grasping the food. Therefore, curve 1 was chosen as the first blade segment for the rotary cutter. Curve 3 showed large variations in the second derivative and was thus not considered. Although curve 5 has larger second derivative variations than curve 4, its greater curvature enhances its sharpness, so curve 5, which performs better in cutting, was selected as the second blade segment. Curve 6 and curve 7 have similar curvatures, but curve 6 has a slightly smoother second derivative, and since it is located closer to the front of the tooth, curve 6 was chosen as the third blade segment. Curve 8 exhibited large variations in its second derivative and smaller curvature, making it unsuitable for selection. Compared to curves 11 and 12, curve 9 exhibits a smoother second derivative and greater curvature, while curve 13 corresponds to a smaller, less prominent tooth, so curve 9 was chosen as the fourth blade segment. Curve 10 has a smaller curvature and is located closer to the root of the tooth, with less cutting significance, and was therefore excluded. The polynomial equations for the four selected curves, fitted using a fifth-order polynomial, are shown in Eq. (3).

3

Fig. 7.

Pharyngeal tooth cutting performance comparison chart.

Where,

Curves	p 1	p 2	p 3	p 4	p 5	p 6
1	− 0.24470278	7.471979644	− 18.76902801	14.65892321	− 3.143196884	− 1.797244696
5	68.12554902	− 73.46102355	14.93542352	1.660682774	0.909737607	− 1.934953411
6	− 12.36237589	61.62612945	− 70.82162592	27.04606909	− 2.224525215	− 1.71845039
9	− 180.7510554	198.6064232	− 41.72978005	− 19.30278891	7.381117792	− 1.254675318

Under the fifth polynomial fitting, the goodness-of-fit for curve 1 is 0.92, for curve 5 is 0.97, for curve 6 is 0.98, and for curve 9 is 0.89. The lower goodness-of-fit for curve 9 may be attributed to noise in the data, which affected the fitting process. However, the overall fit remains satisfactory and meets the design and manufacturing requirements. During the cutting of safflower filaments, the blade primarily engages near the apex of the convex arc. To facilitate subsequent design and manufacturing, data points with significant deviations near the curve ends (10% of the x-axis) were excluded to optimize the curves. The four selected curves were then combined and spliced, and the original and optimized curves were plotted within the same coordinate system, as shown in Fig. 8. After trimming the ends, the optimized curves appear smoother than the directly spliced curves, resolving any improper connections at the start and end points due to the curve fitting process. The x-axis coordinate range for each curve was extracted for use in subsequent blade design. Given the complexity of the curves, the smoothed curve data points were exported into an Excel spreadsheet for future blade design development.

Fig. 8.

Comparison of original and optimized curves.

Blade optimization design

Based on studies of experience with lawnmowers^25,26, crop-cutting harvesters²⁷, and cutting-type safflower harvesting devices⁶, the dynamic impact generated by the blade’s rotational motion is a key factor in achieving effective cutting during the rotary process. This impact is influenced not only by the rotational speed of the blade but also by the blade edge geometry and the material properties of the object being cut. Therefore, analyzing the motion characteristics of the rotating blade during the safflower filament cutting process is crucial for improving cutting quality and reducing energy consumption.

During the rotary cutting of safflower filaments, the high-speed rotating blade impacts the filaments. According to the momentum principle:

4

5

Where F is the impact force of the blade on the safflower filaments, N; t is the impact time, s; m is the filament mass, kg; v1 is the velocity of the filament after cutting, m/s; v0 is the velocity of the filament before cutting, m/s; Rp is the distance from the cutting point of the blade to the center of rotation, m; n is the blade speed, rpm.

Safflower filaments are flexible structures with a certain degree of elasticity and tensile strength¹⁴. According to Eq. (4), the force applied to the filaments is inversely proportional to the impact duration, meaning the shorter the impact time, the greater the force during cutting. This duration is related to the blade edge shape. When using a regular arc blade, significant instantaneous stress concentrations at the cutting point can cause the filaments to collapse, negatively affecting the cutting process. A biomimetic blade with serrations larger than the diameter of a single filament disperses the entire filament into clusters. As the blade rotates, the filaments are gradually cut, distributing the impact force and reducing excessive filament collapse, thereby improving the cutting efficiency.

Traditional rotary safflower harvesting blades are primarily straight or arc-shaped, with smooth edge curves. When cutting with high-speed rotating blades, this method can cause filaments to collapse or become damaged. The cutting mechanism of grass carp pharyngeal teeth resembles that of rotating blades, providing inspiration for optimizing blade edge curves. By referencing traditional rotary blades, the pharyngeal tooth curve is optimized for improved cutting performance and reduced filament collapse. Due to the small size of the pharyngeal teeth, they are unsuitable for rotary safflower harvesting. Based on the dimensions of safflower, the maximum cutting radius of the blade is preliminarily designed to be 25 mm, with a blade radius of 22 mm after accounting for the mounting shaft hole. The magnification factor (K) is calculated using Eq. (6), and the fitted curve length along the x-axis is set to 0 ~ 22 mm to meet blade design requirements. SolidWorks 2023 was used for blade modeling.

6

Where r_s is the designed blade radius, mm; r_f is the length of the bionic curve, mm.

To validate the bionic blade’s cutting performance, three models were created: a traditional blade, a bionic contour blade, and a bionic contour blade with serrations. These models are shown in Fig. 9, with a blade edge angle of 20° for each. By comparing the designs, the primary difference between the traditional and bionic blades lies in the edge curves. Traditional blades have an arc-like curve, while the bionic contour blade has a smoother edge, concentrating pressure on the filaments during cutting, thus enhancing the process. The bionic toothed blade further improves cutting by adding serrations, which apply centripetal pressure, causing localized filaments to compress and making them easier to cut due to the serrations.

Fig. 9.

Traditional and bionic blades. (a) Traditional blade. (b) Bionic contour curved blade. (c) Bionic contour and toothed blade. (1) Blade body. (2) Blade shaft. (3) Conventional blade edge. (4) Serrated blade edge.

Results and analysis

Simulation cutting analysis based on LS-DYNA

The performance of safflower filament harvesting is influenced by multiple factors. To evaluate the cutting efficiency of three different blade types and explore the motion patterns of safflower filaments under bionic blades, the LS-DYNA module in ANSYS software was employed for blade performance analysis^28,29. A safflower filament model was created using SolidWorks software. To enhance the realism of the simulation and reduce computational time, non-cutting parts of the model were simplified to simulate the actual cutting state of the filaments. After positioning the blade, filaments, and other components, the filament model was imported into LS-DYNA for further analysis. The structure of safflower filaments is similar to plant fibers, as outlined in GB/T1209.3-2009 “Agricultural Machinery, Cutters, Part 3: Moving Blades, Fixed Blades, and Blade Shafts,” along with the findings of previous authors and relevant researchers. The blade’s rotation speed was set at 60 rpm, with a blade edge angle of 20°. To minimize simulation time, the cutting task was concluded after a 90-degree blade rotation, corresponding to a simulation time of 0.22 s.

Comparison of blade cutting performance on safflower filaments

The simulation results for the bionic serrated blade are shown in Fig. 10, achieving a harvesting efficiency of 100%. As shown in Fig. 10a, when the blade first contacts the safflower filament, it compresses it. As the blade continues to rotate (Fig. 10b), the filament, having inherent strength, is severed when the compressive force exceeds its breaking strength. Uncut filaments are dispersed by the blade tip, and those between the teeth are squeezed together. Figure 10c shows that when the blade reaches a certain position, the uncut filaments can no longer withstand the blade’s pressure, causing them to bend. When the blade edge passes over the filament blocking hole, 5.48% of the filaments are pulled, damaging the filaments between the blade and the blocking hole. As rotation continues (Fig. 10d), a small portion of the filaments are cut by the blade tip, resulting in slight damage and affecting cutting performance.

Fig. 10.

Simulation diagram of biomimetic toothed blade cutting filament. (a) Blade contacts the filament. (b) Blade cuts the filament. (c) Blade pulls the filament fibers. (d) Filament cutting completed. (Image source: Created by the author; Software version: Ansys2023 R1; URL: https://www.ansys.com/).

The cutting process of the bionic contour curve blade is illustrated in Fig. 11. During cutting, the blade’s larger curvature and centripetal envelope trend in the cutting direction further compress the filaments, facilitating the cutting process.

Fig. 11.

Simulation results of the bionic contour curve blade. (a) Blade contacts the filament. (b) Blade cuts the filament. (c) Blade edge cuts the filament. (d) Filament cutting completed. (Image source: Created by the author; Software version: Ansys2023 R1; URL: https://www.ansys.com/)

The simulation results for the traditional blade are displayed in Fig. 12. Both the bionic contour curve blade and the traditional blade achieved a harvesting efficiency of 98%. The filament motion was similar, involving compression cutting, edge cutting, and post-cut compression between the blade and upper baffle. However, the contour curve blade exhibited lower cutting stress and reduced filament compression, indicating superior performance in safflower filament harvesting.

Fig. 12.

Simulation results of traditional blades. (a) Blade contacts the filament. (b) Blade cuts the filament. (c) Blade edge cuts the filament. (d) Filament cutting completed. (Image source: Created by the author; Software version: Ansys2023 R1; URL: https://www.ansys.com/)

Comparison of cutting surface

As shown in Fig. 13a, during the bionic serrated blade’s cutting simulation, the filaments in contact with the teeth were severed, while those on both sides were locally dispersed and compressed, facilitating the cutting process. In Fig. 13b, the bionic contour curve blade compressed uncut filaments toward the center, reducing gaps and improving cutting efficiency. In contrast, Fig. 13c shows that the traditional blade’s compression effect on uncut filaments was less pronounced, with filaments tending to spread outward, negatively impacting the cutting result. These findings suggest that the bionic serrated blade had a cutting advantage in the simulation, while the bionic contour curve blade demonstrated significant inward compression, though further experimental validation is needed to compare their actual cutting performances.

Fig. 13.

Analysis of the cutting cross-section of the three types of blades. (a) Bionic serrated blade. (b) Bionic contour blade. (c) Traditional blade. (Image source: Created by the author; Software version: Ansys2023 R1; URL: https://www.ansys.com/)

Bench test analysis

The simulation tests preliminarily demonstrate the superior performance of bionic blades in cutting safflower filaments. To further verify these findings, both bionic and traditional blades were fabricated, and their cutting performances were compared.

Test materials and equipment

The cutting performance of safflower filaments is influenced by physical properties such as moisture content and tensile strength. Therefore, appropriate safflower plants were selected for testing. The test material used was the safflower variety “Yumin No. 2,” cultivated at the Xinjiang Institute of Technology’s Labor Education Practice Base in late May 2024. Harvesting experiments were conducted in the Intelligent Mechanical Laboratory, utilizing a custom-designed safflower rotating cutting harvesting test platform (Fig. 14), along with bionic and traditional blades (Fig. 15), a balance, an anemometer, and other instruments.

Fig. 14.

Test platform. (1) Safflower filaments. (2) Flower collecting cavity. (3) Stepper motor. (4) Experimental test bench. (5) Collection box. (6) Cyclone separator. (7) Flower conveying pipe. (8) Negative pressure pipeline. (9) Frequency converter. (10) Negative pressure fan.

Fig. 15.

Self-made blades. (a) Traditional blade. (b) Bionic contour blade. (c) Bionic serrated blade.

Test methods

To compare the cutting performance of the three blade types for safflower filament harvesting, the blade edge inclination angle was selected as the experimental factor. The harvesting rate and damage rate were used as evaluation metrics in a single-factor comparison experiment. Based on preliminary simulations and pre-experiments, the blade rotation angle was set to 90 degrees, the blade speed to 52.8 rpm, and the inlet wind speed to 2 m/s.

The harvesting rate (Y₁) and damage rate (Y₂) were used as the evaluation indicators, calculated using the following formulas:

7

8

Where Y₁ is the filament harvesting rate (%); Y₂ is the filament damage rate (%); m₁ is the mass of filaments in the collection box (g); m₂ is the mass of residual filaments on the flower head (g); m₃ is the mass of damaged filaments in the collection box (g).

Experiment evaluation

In the comparison experiment using the harvesting rate as the evaluation criterion, the average harvesting rate of the bionic serrated blade was 99.83%, the bionic contour blade 97.68%, and the traditional blade 96.51%. As shown in Fig. 16a, all three blades exhibited superior cutting performance when the edge inclination angle was less than 15°, likely due to the sharper cutting edge at smaller inclination angles. The harvesting rate of the bionic serrated blade remained high, though it decreased as the inclination angle exceeded 25°. Both bionic blades outperformed the traditional blade, demonstrating that bionic blades offer a significant advantage in safflower filament cutting.

Fig. 16.

Comparison of the cutting performance of three types of blades. (a) Comparison of harvesting rate results. (b) Comparison of damage rate results.

In the damage rate comparison experiment, the bionic serrated blade showed an average damage rate of 0.63%, the bionic contour blade 1.37%, and the traditional blade 1.68%. As shown in Fig. 16b, the distinct blade shape and contour curve of the bionic blades effectively reduced filament damage during cutting. The damage rate of the bionic serrated blade was significantly lower than that of the other two blade types. The traditional blade’s damage rate increased rapidly with an increase in edge inclination angle, indicating that bionic blades were sharper and more effective at cutting safflower filaments at the same inclination angle.

Through this analysis and comparison of the cutting performances, the bionic serrated blade not only demonstrated excellent results in cutting safflower filaments but also significantly improved cutting quality due to its unique design.

Optimization experiment

Experimental method

Based on the previous simulation experiments and harvesting experience, the primary factors influencing safflower harvesting efficiency include blade rotation speed, blade edge angle, and negative pressure wind speed. To determine the effects of these factors on harvesting efficiency under different cutting conditions, a single-factor experiment was designed to identify the optimal working range.

Single-factor experiment and results analysis

To minimize the influence of irrelevant variables, safflower plants were categorized by flower and filament morphology prior to the experiment. According to pre-test results, the value ranges for the single-factor experiments were set as follows: blade rotation speed between 35.2 and 70.4 rpm, blade edge angle from 15° to 35°, and negative pressure wind speed at the suction inlet. Each factor was tested at five levels with equal intervals, and the middle level was used as the experimental value when other factors were held constant. The impact of two types of bionic blades on harvesting rate and damage rate is shown in Fig. 17.

Fig. 17.

Single-factor experimental results. (a) Blade rotation speed. (b) Suction inlet wind speed. (c) Blade edge angle.

From Fig. 17a, it can be observed that when the wind speed at the suction inlet is 2.5 m/s and the blade edge angle is 20°, the harvesting rate initially increases and then decreases as blade rotation speed rises, though the overall variation is small. The damage rate first declines and then increases. At rotation speeds below 35.2 rpm, the extended contact time between the blade and the filaments causes frictional damage, as some filaments are not severed in time. As rotation speed exceeds 35.2 rpm, the harvesting rate gradually increases, and at 52.8 rpm, the damage rate decreases. At this optimal speed, the harvesting rate is 99.87%, with a damage rate of 2.00%. A blade rotation speed between 35.2 and 70.4 rpm provides optimal performance for the device.

Figure 17b shows that at a blade rotation speed of 52.8 rpm and a blade edge angle of 20°, the harvesting rate increases with the suction inlet wind speed, while the damage rate decreases initially and then rises. When the inlet wind speed is below 2.25 m/s, the airflow inside the collection chamber and the convergence inlet is insufficient to assist cutting, causing filament friction against the blade and housing, leading to lower harvesting rates and increased filament damage. When the wind speed is between 2.25 and 2.75 m/s, the cut filaments are swiftly removed from the cutting zone, reducing mutual extrusion during cutting. However, at speeds above 2.75 m/s, excessive airflow within the collection chamber causes the filaments to collide with the inner walls, increasing filament damage. At 2.0 m/s, the harvesting rate is 99.48%, and the damage rate is 1.85%. An inlet wind speed between 2.25 and 2.75 m/s is optimal for the device.

In Fig. 17c, with a blade rotation speed of 52.8 rpm and suction inlet wind speed of 2.5 m/s, the harvesting rate decreases as the blade edge angle increases, while the damage rate rises. When the blade edge angle is below 20°, the damage rate increases with angle, and the overall harvesting rate declines slightly. At these angles, the blade remains sharp, minimizing cutting effects on the filaments, though smaller angles lead to faster blade wear. When the blade edge angle exceeds 25°, the reduced sharpness significantly hampers cutting efficiency, causing a sharp decrease in harvesting rate and a rise in damage rate. At a blade edge angle of 20°, the harvesting rate is 99.29%, and the damage rate is 1.96%. A blade edge angle between 15° and 25° is the optimal operating range for the harvesting device.

Multifactor experiment and results analysis

The experiment was conducted using blade rotational speed (x₁), inlet air velocity (x₂), and blade inclination angle (x₃) as the experimental factors, with harvesting rate (Y₁) and damage rate (Y₂) as the evaluation indicators. The Box-Behnken experimental design method^30,31 was employed, in combination with the results of the single-factor experiments. The coding of experimental factor levels is presented in Table 1.

Table 1.

Experimental factor coding.

Code	Blade rotation speed x1 (rpm)	Suction inlet wind speed x2 (m·s−1)	Blade edge angle x3 (°)
− 1	35.2	2.25	15
0	52.8	2.5	20
1	70.4	2.75	25

Experimental data were analyzed using Design-Expert software, with the results shown in Table 2. The average harvesting rate of safflower filaments using the bionic rotary cutting harvesting device was 97.43%, and the average damage rate was 3.65%. A multivariate regression analysis was performed on the experimental results, yielding regression models for blade rotational speed, inlet air velocity, and blade inclination angle with respect to both the harvesting rate and damage rate. Analysis of variance (ANOVA) was conducted, with the results shown in Table 3.

Table 2.

Experimental results.

No.	Blade rotation speed x1 (rpm)	Suction inlet wind speed x2 (m·s−1)	Blade edge angle x3 (°)	Harvesting rate Y1 (%)	Damage rate Y2 (%)
1	0	0	0	99.5	2.68
2	− 1	0	− 1	97.73	3.11
3	0	− 1	1	94.77	6.71
4	− 1	− 1	0	94.6	5.23
5	0	1	− 1	96.96	4.22
6	1	0	− 1	99.42	1.54
7	0	1	1	95.96	4.58
8	0	0	0	99.58	2.14
9	1	0	1	97.52	4.08
10	0	0	0	99.8	2.34
11	0	0	0	99.31	2.14
12	− 1	1	0	95.61	4.27
13	0	− 1	− 1	97.36	2.87
14	0	0	0	99.73	2.24
15	1	1	0	95.03	4.92
16	1	− 1	0	96.23	4.29
17	− 1	0	1	97.25	4.73

Table 3.

Analysis of variance. P < 0.01 is extremely significant, marked as **. 0.01 ≤ P ≤ 0.05 is significant, marked as *.

Source	Y1					Y2
	Sum of squares	df	Mean square	F	P	Sum of squares	df	Mean square	F	P
Model	54.87	9	6.10	76.66	< 0.0001**	30.67	9	3.41	34.50	< 0.0001**
x 1	1.13	1	1.13	14.24	0.0069**	0.7875	1	0.7875	7.97	0.0257*
x 2	0.0450	1	0.0450	0.5658	0.4764	0.1540	1	0.1540	1.56	0.2520
x 3	4.46	1	4.46	56.02	0.0001**	8.74	1	8.74	88.42	< 0.0001**
x 1 x 2	1.22	1	1.22	15.35	0.0058**	0.6320	1	0.6320	6.40	0.0393*
x 1 x 3	0.5041	1	0.5041	6.34	0.0399*	0.2116	1	0.2116	2.14	0.1867
x 2 x 3	0.6320	1	0.6320	7.95	0.0258*	3.03	1	3.03	30.64	0.0009**
x 1 2	6.57	1	6.57	82.66	< 0.0001**	1.37	1	1.37	13.83	0.0075**
x 2 2	37.07	1	37.07	466.07	< 0.0001**	13.64	1	13.64	138.04	< 0.0001**
x 3 2	0.5291	1	0.5291	6.65	0.0365*	0.9996	1	0.9996	10.12	0.0155*
Residual	0.5567	7	0.0795			0.6916	7	0.0988
Lack of fit	0.4066	3	0.1355	3.61	0.1234	0.4911	3	0.1637	3.27	0.1414
Pure error	0.1501	4	0.0375			0.2005	4	0.0501
Cor total	55.42	16				31.37	16

According to Table 3, the regression model for harvesting rate (Y₁) was highly significant, with factors x₁, x₃, x₁ × ₂, x₁², and x₂² having strong effects on Y₁, and factors x₁ × ₃, x₂ × ₃, and x₃² showing significant effects. For the damage rate (Y₂), the regression model was also significant, with factors x₃, x₂ × ₃, x₁², and x₂² having strong effects, and x₁, x₁ × ₂, and x₃² showing significant effects. After removing insignificant terms, the regression equations for blade rotational speed, air inlet velocity, and blade edge inclination angle on the harvesting rate (Y₁) and damage rate (Y₂) of safflower filaments were derived as follows:

9

10

For the bionic rotary harvesting device, the regression model for the harvesting rate (Y₁) had a P-value < 0.0001, indicating high significance and strong correlation between predicted and actual values. Similarly, the regression model for damage rate (Y₂) also had a P-value < 0.0001, confirming strong predictive accuracy.

Based on the variance analysis in Table 3 and Eqs. (9) and (10), it can be concluded that blade rotational speed, inlet air speed, and blade inclination angle have interactive effects on both harvesting rate and damage rate. To further explore these effects, response surface plots of the experimental factors were generated using Origin 2022 software, as shown in Fig. 18.

Fig. 18.

Response surface of the interaction factors on the harvesting performance of safflower filaments. (a) The effect of tool speed and air inlet wind speed on net recovery rate. (b) The effect of air inlet wind speed and blade inclination on net recovery rate. (c) The effect of tool speed and air inlet wind speed on damage rate. (d) The effect of air inlet wind speed and blade inclination on damage rate.

Figure 18a demonstrates the effect of the interaction between blade rotational speed and inlet air velocity on the safflower filament harvesting rate when the blade inclination angle is 20°. When the blade speed is fixed at 72 rpm, the harvesting rate initially increases and then decreases as the inlet air velocity increases. A higher air velocity generates stronger suction, allowing more filaments to enter the collection chamber, thereby improving the harvesting rate. However, when the air velocity exceeds 2.6 m/s, the excessive suction causes some filaments to adhere to the inner wall of the collection chamber, resulting in missed cuts and a reduction in the harvesting rate. When the air velocity is fixed at 2.25 m/s, the harvesting rate initially increases with the rise in blade speed but declines at higher speeds. This is because an increase in blade speed reduces the time the filaments are exposed to the blade, decreasing filament collapse and improving the harvesting rate. However, excessive blade speed causes filaments to be drawn into the blade’s rotating space, leading to missed cuts and a reduction in the harvesting rate.

Figure 18b presents the effect of the interaction between inlet air velocity and blade inclination angle on the harvesting rate when the blade speed is 52.8 rpm. When the air velocity is fixed at 2.25 m/s, the harvesting rate decreases as the blade inclination angle increases. A larger inclination angle makes the blade relatively blunter, and filaments are more likely to be pulled into the rotating space, leading to missed cuts and a decrease in the harvesting rate.

Figure 18c illustrates the effect of the interaction between blade rotational speed and inlet air velocity on the damage rate when the blade inclination angle is 20°. When the inlet air velocity is fixed at 2.7 m/s, the damage rate first decreases and then increases with rising blade speed. At lower blade speeds, the filaments are not cut efficiently, leading to damage as they are pulled into the rotating space. As the blade speed increases, the filaments are cut promptly, reducing the damage rate. However, when the blade speed exceeds 60 rpm, some filaments are not quickly removed by the suction, resulting in damage as they are compressed in the rotating space.

Figure 18d shows the effect of the interaction between inlet air velocity and blade inclination angle on the damage rate when the blade speed is 52.8 rpm. When the blade inclination angle is 15°, the damage rate initially decreases and then increases with rising air velocity. At lower air velocities, the suction force is insufficient to counterbalance the impact force of the blade during cutting, leading to filaments being trapped and damaged in the rotating space, causing the damage rate to rise.

Parameter optimization

To identify the optimal parameter combination influencing the cutting performance of the harvesting device, a multi-objective optimization analysis was conducted with the objectives of maximizing the harvesting rate and minimizing the damage rate. The parameters considered blade rotational speed, inlet air velocity, and blade inclination angle, the optimized model is expressed as follows:

11

After analysis and calculation, the optimal parameter combination was determined to be a blade rotational speed of 61.39 rpm, an inlet air velocity of 2.44 m/s, and a blade inclination angle of 15.32°, resulting in a harvesting rate of 99.97% and a damage rate of 1.53%.

Experimental verification

To validate the optimized parameters, a verification experiment was conducted in the Intelligent Machinery Laboratory at the Xinjiang Institute of Technology. For feasibility, the optimized parameters were rounded to a blade speed of 61 rpm, inlet air velocity of 2.44 m/s, and a blade inclination angle of 15°. The experimental setup is shown in Fig. 19a. The test was repeated three times, using safflower filaments at peak bloom and with uniform plant size as the test materials. The cutting experiment is shown in Fig. 19b, and the harvested safflower fruits are shown in Fig. 19c. The harvesting and damage rates were calculated according to Eqs. (7) and (8) and averaged. The cut sections were clean, and the evaluation results are presented in Table 4.

Fig. 19.

Experimental process. (a) Test scenario. (b) Test process. (c) Harvest effect.

Table 4.

Experimental verification results.

Serial number	Y1 (%)	Y2 (%)
1	98.95	1.57
2	98.64	1.43
3	99.27	1.62
Average	98.95	1.54

Under the optimal combination of parameters, the average harvesting rate was 98.95%, with a damage rate of 1.54%. The relative deviation from the optimization results was less than 2%, confirming that the harvesting device meets the operational requirements for safflower filament harvesting. In comparison to the safflower filament harvesting devices proposed by other teams, the harvesting device designed in this study shows improved harvesting performance, with specific parameters outlined in Table 5. Although this research provides a practically significant solution for the mechanized and intelligent harvesting of safflower filaments, there is still room for further investigation and refinement in areas such as blade material selection, optimization of the serrated blade geometry, realization of modular design, and the versatility of the device. Future research should focus on optimizing the performance and enhancing the efficiency of the harvesting device, exploring its cross-application potential across multiple crops, and advancing the lightweight design of the equipment. Furthermore, as the level of agricultural mechanization and intelligence continues to develop, the further advancement of related technologies will play a crucial role in improving crop harvesting efficiency and reducing labor intensity.

Table 5.

Comparison of harvesting performance.

Team	Y1 (%)	Y2 (%)
Zhang et al.6	91.25	5.57
Luo et al.7	94.63	8.2
Li et al.9	91.47	7.51
Author et al.	98.95	1.54

Conclusions

Addressing the challenges of large device structures and ineffective cutting blades in current safflower filament harvesting devices, this study developed a negative-pressure rotary cutting harvesting device based on biomimetic principles. Key conclusions include:

1. Design and Optimization of Biomimetic Blades: Through digital image processing, the structure of grass carp pharyngeal teeth was extracted, with precise fitting of both the contour and tooth profiles. Blades incorporating contour and tooth curves inspired by grass carp pharyngeal teeth were designed. The performance of two types of biomimetic blades was optimized, confirming the theoretical advantages over traditional blades.

2. Simulation of Cutting Processes: LS-DYNA simulations of traditional, contour, and toothed blades demonstrated that the biomimetic toothed blade effectively distributed filaments, reducing filament collapse and improving cutting. The contour blade induced centripetal compression among filaments, enhancing cutting efficiency. In contrast, traditional blades showed minimal compression and increased filament scattering, leading to poor cutting performance.

3. Performance Comparison Across Blade Types: Comparative experiments on harvesting rate and damage rate showed that the biomimetic toothed blade maintained superior cutting performance with inclination angles between 10° and 30°, outperforming the contour and traditional blades. The biomimetic toothed blade achieved an average harvesting rate of 99.83%, representing a 3.32% improvement over traditional blades, while reducing the average damage rate by 1.18%.

4. Optimization and Verification of Biomimetic Parameters: Single-and multi-factor experiments confirmed that the optimal parameters for the biomimetic toothed blade were a rotational speed of 61.39 rpm, inlet air velocity of 2.44 m/s, and a blade inclination angle of 15.32°. Under these conditions, the harvesting rate reached 99.97%, and the damage rate was 1.53%. Verification experiments yielded an average harvesting rate of 98.95% and a damage rate of 1.54%, with a relative deviation of less than 2%, confirming the effectiveness of the designed harvesting device for safflower filament harvesting.

5. This study provides key technical support for the efficient harvesting of safflower filament and lays a solid theoretical foundation for the development of intelligent harvesting robots for safflower filament. However, improving the harvesting speed of safflower filament remains a critical issue that needs to be addressed urgently. Future research will focus on further exploring and optimizing relevant technologies to achieve higher harvesting efficiency and promote the intelligent process of safflower filament harvesting.

Author contributions

Conceptualization, B.C. and F.D.; methodology, B.C.; software, Q.Y.; validation, Q.Y., K.L. and B.M.; formal analysis, K.L.; investigation, B.C.; resources, B.M.; data curation, Q.Y.; writing—original draft preparation, B.C.; writing—review and editing, F.D.; visualization, F.D.; supervision, B.M.; project administration, B.C.; funding acquisition, F.D. All authors have read and agreed to the published version of the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the second batch of Tianshan Talent Cultivation Plan for Young Talent Support Project (Program No. 2023TSYCQNTJ0040) and the Natural Science Basic Research Program of Shaanxi (Program No. 2023-JC-YB-347).

Data availability

The data used to support the findings of this study are included within the article.

Declarations

Competing interests

The authors declare no competing interests.