Seed meter, planter and control method for seed meter
By using seed flow detection and speed sensor real-time monitoring, combined with control unit and machine learning model optimization, the problem of not being able to perceive seed flow in real time in existing technologies has been solved, achieving precise adaptive sowing control, reducing reseeding and missed sowing rates, and improving sowing quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INTELLIGENT EQUIPMENT RESEARCH CENTER BEIJING ACADEMY OF AGRICULTURE AND FORESTRY SCIENCES
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-09
AI Technical Summary
Existing mechanical and motor-driven seed scraping devices cannot sense the seed flow at the seed metering outlet in real time, resulting in an inability to intelligently identify and correct complex sowing defects, such as double seeding and missed seeding.
Seed flow detection sensors and speed sensors are used to monitor seed falling in real time. Combined with the control unit, the time interval between the falling of adjacent seeds is calculated. The extension of the seed scraper driven by the motor is adjusted to achieve real-time correction of reseeding and missed seeding. The control parameters are optimized by using a machine learning model.
It achieves precise and adaptive control of the sowing process, significantly reduces the reseeding rate and missed sowing rate, improves the uniformity and consistency of sowing, and reduces the reliance on the operator's experience.
Smart Images
Figure CN122162565A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural machinery technology, and in particular to a seed metering device, a seeder, and a control method for the seed metering device. Background Technology
[0002] Precision seeding is a crucial step in modern agricultural production, aiming to achieve single-seed precision sowing, ensuring uniform seed spacing and avoiding double sowing and missed sowing. Current mechanical seed metering devices generally use a fixed-position seed scraper or blade to remove excess seeds from the suction holes or sprue holes of the seed metering disc. However, this mechanical structure has the following drawbacks: First, when sowing seeds of different crops or varieties, the machine needs to be stopped and the position of the seed scraper manually adjusted due to differences in seed size, shape, and coefficient of friction. This process is time-consuming and labor-intensive, and the accuracy of the adjustment depends on the operator's experience, making it difficult to guarantee optimal results. Second, during the sowing operation, factors such as the seed metering disc speed, machine vibration, and ground undulations will continuously change. A fixed seed scraping position cannot adapt to these dynamic disturbances, easily leading to scraping too deeply (scraping away many seeds and causing missed sowing) or scraping too shallowly (failing to remove excess seeds and causing re-sowing).
[0003] To address the aforementioned issues, some existing technologies have developed electric seed scraping devices driven by motors. However, most of these improved solutions can only perform simple, pre-programmed position settings based on preset operating speeds or seed types. Their control logic is relatively simple, and they still lack real-time perception of the actual seed flow at the seed metering outlet, making it impossible to intelligently identify and effectively correct complex sowing defects. Summary of the Invention
[0004] This invention provides a seed metering device, a seeder, and a control method for the seed metering device, in order to solve the problem that existing motor-driven electric seed scraping devices lack real-time perception of the actual seed flow at the seed metering device outlet, and cannot intelligently judge and effectively correct complex sowing defects.
[0005] This invention provides a seed metering device, comprising: The mechanical execution unit includes a seed metering device body, a seed metering disc driven by a first motor, and a seed scraper driven by a second motor; the seed metering disc is disposed in the seed metering device body, and the seed scraper is disposed on a set side of the seed metering disc for scraping off the seeds on the seed metering disc; The information detection unit includes a seed flow detection sensor for detecting the moment of seed fall and a speed sensor for detecting the speed of travel. A control unit is electrically connected to the first motor, the second motor, the seed flow detection sensor, and the speed sensor. The control unit is configured to control the rotational speed of the first motor based on a set target plant spacing and the travel speed detected by the speed sensor; calculate the time interval between adjacent seed falls based on the seed fall time sequence detected by the seed flow detection sensor; compare the time interval with a qualified time interval calculated from the target plant spacing and the travel speed to determine whether a reseeding or missed seeding event has occurred; and adjust the second motor according to the determination result to control the insertion amount of the working end of the seed scraper into or out of the working area of the seed metering tray; wherein, when a reseeding event is determined to occur, the insertion amount of the seed scraper is increased; and when a missed seeding event is determined to occur, the insertion amount of the seed scraper is decreased.
[0006] According to a seed metering device provided by the present invention, the mechanical execution unit further includes: a cam, wherein an eccentrically arranged protrusion is provided on the end face of the cam; the seed scraper is connected to the output shaft of the second motor through the cam; One end of the seed scraper is provided with an elongated hole, and the protrusion is embedded in the elongated hole to form a sliding fit. The other end of the seed scraper is hinged to the seed metering disc.
[0007] The present invention also provides a method for controlling a seed metering device, comprising: The rotational speed of the first motor is controlled based on the set target plant spacing and the traveling speed detected by the speed sensor. The time interval between adjacent seed falls is calculated based on the seed fall time sequence detected by the seed flow detection sensor. The time interval is compared with the qualified time interval calculated from the target plant spacing and the travel speed to determine whether a reseeding or missed seeding event has occurred. The second motor is adjusted according to the judgment result to control the amount by which the working end of the seed scraper extends into or out of the working area of the seed metering tray; wherein, when it is judged that double seeding has occurred, the seed scraper is controlled to increase the extension amount; when it is judged that missed seeding has occurred, the seed scraper is controlled to decrease the extension amount.
[0008] According to the control method of the seed metering device provided by the present invention, the qualified time interval is: ; ; ; ; in The target plant spacing, The speed of travel, This is the allowable relative error coefficient.
[0009] According to the control method of the seed metering device provided by the present invention, when no reseeding or missed seeding event is detected, a PID control algorithm is used to fine-tune the insertion amount of the seed scraper based on the deviation between the actual average plant spacing and the target plant spacing. When a re-reproduction is detected, the insertion depth of the seed scraper is increased by a first preset step. When it is determined that a missed sowing has occurred, the insertion depth of the seed scraper is reduced by a second preset step.
[0010] According to the seed meter control method provided by the present invention, the size of the first preset step size and the second preset step size are positively correlated with the number of consecutive occurrences of the same type of abnormal event.
[0011] The control method for the seed metering device provided by the present invention further includes: Based on a machine learning model, at least one of the parameters of the PID control algorithm, the first preset step size, the second preset step size, and the reference working position of the seed scraper is optimized according to multiple feature parameters of the operation, so as to maximize the comprehensive score of seeding quality. The characteristic parameters include one or more of the following: traveling speed, seed metering disc rotation speed, seed characteristic parameters, running time, environmental parameters, average plant spacing, plant spacing standard deviation, reseeding rate, and missed seeding rate.
[0012] According to the seeder control method provided by the present invention, the training of the machine learning model includes two stages: offline pre-training and online adaptive updating. The offline pre-training process uses historical job data to perform supervised pre-training on the model. The online adaptive update periodically fine-tunes the model parameters based on data collected in real time during operation.
[0013] According to the seeder control method provided by the present invention, the machine learning model is one of deep neural network, reinforcement learning agent, random forest or gradient boosting decision tree.
[0014] The present invention also provides a seeder, including the above-described seed metering device and / or a control method for performing the above-described seed metering device.
[0015] The seed metering device provided by this invention collects seed flow and travel speed signals in real time through an information detection unit. The control unit can accurately determine reseeding or missed seeding events based on this data and immediately respond by adjusting the insertion depth of the seed scraper. This proactively and promptly corrects abnormalities in the seed metering process, thereby maintaining reseeding and missed seeding rates at low levels and effectively improving the uniformity and consistency of sowing. Furthermore, this invention upgrades the adjustment of the seed scraper from traditional manual, static preset settings to automatic, dynamic adjustment driven by a motor. The control unit makes decisions based on real-time operational feedback, eliminating the impact of changes in operating speed, differences in seed characteristics, or component wear, reducing reliance on operator experience, and achieving precise and adaptive intelligent seed metering. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0017] Figure 1 This is a schematic diagram of the seed metering device provided by the present invention.
[0018] Figure 2 This is a cross-sectional schematic diagram of the seed metering device provided by the present invention.
[0019] Figure 3 This is a partial internal schematic diagram of the seed metering device provided by the present invention.
[0020] Figure 4 This is a flowchart of the control method for the seed metering device provided by the present invention.
[0021] Figure 5 This is a schematic diagram of the control logic of the seed metering device control method provided by the present invention.
[0022] Figure 6 This is a schematic diagram of the seed flow detection sensor signal provided by the present invention.
[0023] Figure label: 1. Seed metering device body; 2. First motor; 3. Seed metering disc; 4. Second motor; 5. Seed scraper; 6. Seed flow detection sensor; 7. Protrusion; 8. Elongated hole. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0025] The following is combined Figures 1-6 This invention describes the seed metering device, the seeder, and the control method for the seed metering device provided by the present invention.
[0026] This invention provides a seed metering device, such as... Figures 1 to 3 As shown, the seed metering device includes: a mechanical execution unit, an information detection unit, and a control unit.
[0027] In this embodiment, the mechanical execution unit includes a seed metering device body 1, a seed metering disc 3, and a seed scraper 5. The seed metering device body 1 serves as the support and sealing structure for the entire device, forming channels and chambers for seed flow. The seed metering disc 3 is located inside the seed metering device body 1 and is driven to rotate by a first motor 2. The seed metering disc 3 has regularly distributed perforations (or suction holes) for carrying single seeds. The rotational speed of the first motor 2 determines the frequency at which the perforations pass through the seed filling area, which is the basis for controlling the seed spacing. The seed scraper 5 is located on the designated side of the seed metering disc 3 (usually tangential to the circumference of the seed metering disc 3 or on the side adjacent to the seed filling area), and its function is to scrape off excess seeds that are adsorbed or filled on the perforations of the seed metering disc 3, exceeding the single seed size. The seed scraper 5 is independently driven by a second motor 4. By controlling the second motor 4, the depth (i.e., the insertion amount) of the working end of the seed scraper 5 into or out of the working area of the seed metering disc 3 can be adjusted. Increasing the insertion depth and scraping force can prevent multiple seeds from being sown (re-sowing); decreasing the insertion depth and scraping force can prevent the absence of seeds in the sizing hole (missed sowing).
[0028] The information detection unit includes a seed flow detection sensor 6 and a velocity sensor. The seed flow detection sensor 6 is installed at the seed outlet channel of the seed metering device. When a single seed falls through the detection area, it may block or reflect light, causing the sensor to generate a corresponding electrical pulse signal. The control unit records the time of each pulse, forming a sequence of seed fall times, thereby detecting the seed fall time. The velocity sensor is used to detect the seeder's travel speed in real time. For example, a Global Navigation Satellite System (GNSS) receiver or an Inertial Measurement Unit (IMU) can be used.
[0029] The control unit, such as a microcontroller unit (MCU), programmable logic controller (PLC), or industrial computer, is electrically connected to the first motor 2, the second motor 4, the seed flow detection sensor 6, and the speed sensor. It internally stores and runs intelligent control algorithms. The control unit is configured to control the rotational speed of the first motor 2 based on the set target plant spacing and the traveling speed detected by the speed sensor; calculate the time interval between adjacent seed falls based on the seed fall time sequence detected by the seed flow detection sensor 6; compare the time interval with a qualified time interval calculated based on the target plant spacing and traveling speed to determine whether a reseeding or missed seeding event has occurred. Based on the determination result, the second motor 4 is adjusted to control the insertion and withdrawal of the working end of the seed scraper 5 into or out of the working area of the seed metering tray 3. Specifically, when reseeding is determined to have occurred, the control unit controls the seed scraper 5 to increase the insertion depth, thereby increasing the scraping force and more thoroughly removing excess seeds from the seed holes, preventing reseeding from the source. When a missed sowing is detected, the control unit controls the seed scraper 5 to reduce the depth of insertion, thereby reducing the scraping force and avoiding scraping off the single seeds that should be carried in the sizing hole, ensuring that the sizing hole can effectively carry the seeds and eliminating the cause of the missed sowing.
[0030] The seed metering device provided by this invention collects seed flow and travel speed signals in real time through an information detection unit. The control unit can accurately determine reseeding or missed seeding events based on this data and immediately respond by adjusting the insertion depth of the seed scraper 5. This proactively and promptly corrects abnormalities in the seed metering process, thereby maintaining reseeding and missed seeding rates at low levels and effectively improving the uniformity and consistency of sowing. Furthermore, this invention upgrades the adjustment of the seed scraper 5 from traditional manual, static preset settings to automatic, dynamic adjustment driven by a motor. The control unit makes decisions based on real-time operational feedback, eliminating the impact of changes in operating speed, differences in seed characteristics, or component wear, reducing reliance on operator experience, and achieving precise and adaptive intelligent seed metering.
[0031] In some embodiments, such as Figures 1 to 3 As shown, the mechanical execution unit also includes a cam, with an eccentrically positioned protrusion 7 on its end face; the seed scraper 5 is connected to the output shaft of the second motor 4 via the cam. That is, the cam is fixedly mounted on the output shaft of the second motor 4 and can rotate with the output shaft. One end of the seed scraper 5 has an elongated hole 8, and the protrusion 7 is embedded in the elongated hole 8 to form a sliding fit. The protrusion 7 can slide freely within the length range of the elongated hole 8, while simultaneously transmitting the rotational motion of the cam to the seed scraper 5. The other end of the seed scraper 5 is hinged to the seed metering disc 3, forming a swing fulcrum.
[0032] When the second motor 4 receives a command from the control unit, it begins to rotate, driving the cam to rotate synchronously. Since the protrusion 7 on the cam is eccentrically positioned, its rotation trajectory is circular. This circular trajectory, through the sliding engagement between the protrusion and the elongated hole 8 of the seed scraper 5, is transformed into a pushing and pulling action on the seed scraper 5: as the cam rotates, the eccentric protrusion slides within the elongated hole 8, simultaneously pushing (or pulling) the seed scraper 5 to reciprocate around its hinge point at the other end. The oscillating motion of the seed scraper 5 is further transformed into linear displacement of its working end (i.e., the blade tip, the part in contact with the type 3 hole of the seed metering disc) relative to the working area of the type 3 hole of the seed metering disc. By controlling the rotation angle and direction of the second motor 4, the position of the cam can be controlled, thereby continuously and steplessly adjusting the depth (i.e., the insertion amount) of the working end of the seed scraper 5 into the type 3 hole area of the seed metering disc. A larger insertion amount enhances the scraping effect; a smaller insertion amount weakens the scraping effect.
[0033] This invention also provides a control method for a seed metering device, which is used to control the aforementioned seed metering device, such as... Figure 4 As shown, it includes the following steps: Step S410: Control the rotation speed of the first motor according to the set target plant spacing and the traveling speed detected by the speed sensor.
[0034] Step S420: Calculate the time interval between adjacent seed falls based on the seed fall time sequence detected by the seed flow detection sensor.
[0035] Step S430: Compare the time interval with the qualified time interval calculated by the target plant spacing and the travel speed to determine whether a reseeding or missed seeding event has occurred.
[0036] Step S440: Adjust the second motor according to the judgment result to control the extension amount of the working end of the seed scraper into or out of the working area of the seed metering tray.
[0037] Specifically, such as Figure 4 and Figure 5 As shown, the control unit first receives the target plant spacing set by the operator. Ideal average seed spacing From the angular velocity of the seed metering disc ω 1. The seeder's travel speed is detected in real time by speed sensors (such as GNSS receivers). v 1 To decide. The control unit is based on the formula. = Perform reverse calculation, assuming = To determine the required seeding disc angular velocity Where k is the number of holes on the seed metering disc, a fixed mechanical parameter. Through this calculation and control, the system ensures that, under ideal conditions, the theoretical time interval between the seeds carried by each hole on the seed metering disc being sown matches the forward speed of the seeder, thereby forming the preset target plant spacing on the ground.
[0038] The seed stream detection sensor records the timestamp of each seed. , , , , ... Based on this sequence and real-time speed information, the control unit calculates indicators reflecting the actual sowing quality: the algorithm more directly calculates the time interval between adjacent seeds. .
[0039] Actual plant spacing between two adjacent seeds It can be approximated as ,in, Seed number 2 represents the time interval. The average forward speed within the range. It reflects the actual distribution distance of seeds in the field.
[0040] Set a qualified time window This window is calculated based on the target plant spacing and the allowable error range: 1, , , The allowable relative error is, for example, 0.2.
[0041] like Figure 6 As shown, if Then it is determined that in and Seeds that fall at the same time are called "re-sown." Because the time interval between two seeds landing is too short, they will land in almost the same spot. If Then it is determined that in and A "missed seed" occurred. Because the waiting time for the next seed was too long, it meant that a gap appeared in the position where a seed should have been planted.
[0042] Finally, based on the diagnostic results, the corrective actions are executed by directly controlling the second motor to adjust the depth (i.e., the insertion amount) of the seed scraper's working end into or out of the seed metering disc's working area, thus forming a closed-loop control: When a reseeding is detected, the control unit immediately instructs the second motor to operate, controlling the seed scraper to increase its insertion depth, thereby increasing the scraping force and more thoroughly scraping away the excess seeds adsorbed on the seed pores, eliminating the cause of the reseeding at its source.
[0043] When a missed sowing is detected, the control unit immediately instructs the second motor to reverse its operation, controlling the seed scraper to reduce its insertion depth. This reduces the scraping force and prevents the single seed that should be carried in the seed hole from being scraped off as well, ensuring that the seed hole can reliably carry and release the seed, thus eliminating the cause of the missed sowing.
[0044] In some embodiments, such as Figure 5 As shown, when no reseeding or missed sowing events are detected, a PID control algorithm is used to fine-tune the insertion amount of the seed scraper based on the deviation between the actual average plant spacing and the target plant spacing; when it is determined that reseeding has occurred, the insertion amount of the seed scraper is increased by a first preset step; when it is determined that missed sowing has occurred, the insertion amount of the seed scraper is decreased by a second preset step.
[0045] Specifically, when no anomalies are detected, a slow closed-loop adjustment is used to ensure the actual average plant spacing. Approaching The position of the seed scraper is fine-tuned using an incremental PID algorithm. in, . It is the first k Plant spacing deviation within each control period. 、 and These are the proportional, integral, and differential coefficients, respectively.
[0046] When an anomaly is detected, a fast and forceful correction rule is executed first. Upon detecting a replay event, the control unit immediately moves in the "deeper seed scraping" direction, increasing the depth of the seed scraper by a first preset step. ,Right now This operation aims to scrape away excess seeds more forcefully, immediately eliminating sources of reseeding. Upon detecting a missed seeding event, the control unit immediately moves towards "reduced scraping," decreasing the depth of the seed scraper by a second preset step. ,Right now .
[0047] It should be noted that, to prevent insufficient adjustment in a single instance, the algorithm records the number of consecutive occurrences of the same type of anomaly. The adjustment amount can be compared with... Proportional, that is This accelerates the correction process. Specifically, the first and second preset step sizes are positively correlated with the number of consecutive occurrences of the same type of abnormal event. Simultaneously, steady-state adjustment and abnormal response adjustment work together to form a complete adaptive control algorithm for the seed scraper.
[0048] In some embodiments, the control method further includes: based on a machine learning model, optimizing at least one of the parameters of the PID control algorithm, a first preset step size, a second preset step size, and the reference working position of the seed scraper according to multiple characteristic parameters of the operation, so as to maximize the comprehensive score of seed metering quality; wherein the characteristic parameters include one or more of the following: travel speed, seed metering disc rotation speed, seed characteristic parameters, running time, environmental parameters, average plant spacing, plant spacing standard deviation, reseeding rate, and missed seeding rate.
[0049] In this embodiment, the system state can also be analyzed on a longer time scale, such as a few minutes or a work cycle, and the parameters or baseline settings of the control unit can be slowly adjusted so that the system can learn and adapt to complex nonlinear relationships and eventually tend to the global optimal performance.
[0050] Engineering characteristics: The model's input feature vector Includes multi-dimensional parameters that are strongly correlated with seeding quality: in: : The forward speed of agricultural machinery; Seeding disc rotation speed; The average thousand-seed weight or size grade of the seed population can be preset through the human-machine interface; System uptime reflects slow changes in mechanical wear, lubrication status, etc. Ambient temperature and humidity (optional, obtained via additional sensors); : Recent actual average plant spacing; The coefficient of variation of recent plant spacing reflects the uniformity of sowing. , Recent rebroadcast rate and missed broadcast rate; Output Objective 1 (Parameter Determination): The model outputs the optimal parameter combination for the PID controller. ] and event-driven step size[ ].
[0051] Output Objective 2 (Position Determination): The model directly outputs an optimal reference working position for the seed scraper. Fine-tuning is performed near this reference position, i.e. Meanwhile, the training objective of the model is to maximize the ranking quality. A comprehensive scoring function Q is defined as the reward for reinforcement learning or the label for supervised learning: in , , , These are the weighting coefficients. A higher Q value indicates better overall ranking quality. The model's goal is to learn a mapping function. This maximizes the long-term average Q value.
[0052] In some examples, model training is a continuous optimization process that includes two phases: offline pre-training and online adaptive updates.
[0053] Offline pre-training involves supervised pre-training of the model under laboratory or specific field conditions using a large amount of historical operational data, including seeding effect data under various speeds, seed types, and seed scraper positions. The training label is the comprehensive score Q. This stage aims to provide the model with good initial parameters, enabling it to have basic optimization capabilities from the initial deployment stage.
[0054] Online adaptive updates are a feature where the system continuously collects new status data during actual operation. Together with the corresponding real-time seeding quality score Q, a dynamically growing experience replay buffer is formed. The control unit periodically (e.g., after each acre of work is completed or at regular intervals) samples batches of data from the buffer to fine-tune or incrementally learn the model parameters. This mechanism enables the system to adapt to specific fields, specific seed batches, or unique working conditions that change with mechanical wear, achieving true personalized adaptation.
[0055] It should be noted that machine learning models may adopt, but are not limited to, one or more of the following architectures: Deep Neural Networks (DNNs) are suitable for learning complex nonlinear mappings between high-dimensional features.
[0056] Reinforcement Learning (RL) agents: For example, algorithms based on the Actor-Critic framework treat the control of the seed scraper as a sequential decision-making process, learning the optimal strategy through interaction with the environment (seedling system).
[0057] Random Forest or Gradient Boosting Decision Tree (GBDT): These ensemble learning methods have relatively relaxed requirements for feature engineering and are highly efficient in training and inference, making them suitable for real-time applications in embedded devices.
[0058] Bayesian optimization can be used for efficient global optimization in the parameter space, and is particularly suitable for offline or online tuning of lower-level controller parameters.
[0059] The model uses the aforementioned feature vectors As input, the algorithm undergoes nonlinear transformation through multiple hidden layers or complex decision rules, ultimately outputting the optimization objective.
[0060] In addition, the input feature vector It includes not only the original observations, but also their statistics and derived characteristics. For example, the recent actual plant spacing. It can be expanded to multiple time windows, such as the moving average and coefficient of variation for the most recent 10 seconds, 30 seconds, and 1 minute; replay rate. and missed broadcast rate This can be calculated as the proportion of anomalous events among a certain number of recent seeds (e.g., 100 seeds). This processing can provide the model with richer temporal and statistical information.
[0061] The model can output multiple objectives simultaneously, such as PID parameters and reference position, and ensure the coordination between these outputs. For example, when the model determines that more aggressive control is needed under the current operating condition, it may simultaneously output larger parameters. and smaller This aims to achieve a new balance between rapid response and stability. The model learns... The complex relationship between the optimal output combination and the control effect is better than that achieved by manually setting parameters.
[0062] When the new parameters are calculated Or a new reference position Instead of immediately overwriting the current value, a smooth transition strategy is employed. This involves using a first-order low-pass filter or setting a gradual rate to allow the parameter to transition slowly from the old value to the new value. ,in It is a small gain factor, such as 0.05. This avoids oscillations in the lower-level control system caused by sudden changes in model output, ensuring the smoothness of system switching. Through the above machine learning framework, the system has been upgraded from a simple control unit to an intelligent agricultural equipment capable of learning from data, accumulating experience, and continuously improving itself, achieving significant technological progress.
[0063] This invention also provides a seeder, which includes the above-described seed metering device or the above-described control method for the seed metering device. Seeding device (see attached image). Figures 1 to 3 For related textual descriptions and control methods of the seed metering device, please refer to [link / reference]. Figures 4 to 6 The relevant textual descriptions will not be repeated here.
[0064] In summary, compared with the prior art, the embodiments of the present invention have the following significant advantages: 1. Closed-loop control and real-time correction: By directly detecting the actual seed flow at the seed metering device outlet, a closed-loop feedback control system was constructed, enabling control decisions to be based on real seed metering effect data, thus achieving a leap from passive prevention to active detection and real-time correction.
[0065] 2. The control unit adopts a two-layer control architecture that combines fast rule response at the lower layer with machine learning optimization at the upper layer. It can quickly and effectively correct sudden replay and missed replay events, and can also adaptively learn the complex nonlinear relationship between different operating conditions, seed characteristics and optimal seeding parameters through machine learning models to achieve long-term global optimal control.
[0066] 3. It can dynamically eliminate factors that lead to reseeding and missed sowing at the source, significantly reduce the reseeding rate and missed sowing rate, improve sowing uniformity, and lay the foundation for high and stable crop yields.
[0067] 4. Applicable to various types of orifice seed metering devices (such as air suction type, air blowing type, etc.), it can be adapted to different crops and seeds by adjusting parameters, making it highly versatile. At the same time, the lower-level control combining rule base and PID ensures the basic reliability and stability when the machine learning model has not been fully trained or encounters unseen operating conditions.
[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A seed metering device, characterized in that, include: The mechanical execution unit includes a seed metering device body, a seed metering disc driven by a first motor, and a seed scraper driven by a second motor; the seed metering disc is disposed in the seed metering device body, and the seed scraper is disposed on a set side of the seed metering disc for scraping off the seeds on the seed metering disc; The information detection unit includes a seed flow detection sensor for detecting the moment of seed fall and a speed sensor for detecting the speed of travel. The control unit is electrically connected to the first motor, the second motor, the seed flow detection sensor, and the speed sensor; the control unit is configured to control the rotational speed of the first motor according to a set target plant spacing and the traveling speed detected by the speed sensor. Based on the seed falling time sequence detected by the seed flow detection sensor, the time interval between the falling of adjacent seeds is calculated; The time interval is compared with the qualified time interval calculated from the target plant spacing and the traveling speed to determine whether a reseeding or missed sowing event has occurred; the second motor is adjusted according to the determination result to control the amount by which the working end of the seed scraper extends into or out of the working area of the seed metering tray; wherein, when it is determined that a reseeding has occurred, the seed scraper is controlled to increase the extension amount; when it is determined that a missed sowing has occurred, the seed scraper is controlled to decrease the extension amount.
2. The seed metering device according to claim 1, characterized in that, The mechanical actuation unit further includes: a cam, the end face of which is provided with an eccentrically arranged protrusion; the seed scraper is connected to the output shaft of the second motor through the cam; One end of the seed scraper is provided with an elongated hole, and the protrusion is embedded in the elongated hole to form a sliding fit. The other end of the seed scraper is hinged to the seed metering disc.
3. A control method based on the seed metering device according to claim 1 or 2, characterized in that, include: The rotational speed of the first motor is controlled based on the set target plant spacing and the traveling speed detected by the speed sensor. The time interval between adjacent seed falls is calculated based on the seed fall time sequence detected by the seed flow detection sensor. The time interval is compared with the qualified time interval calculated from the target plant spacing and the travel speed to determine whether a reseeding or missed seeding event has occurred. The second motor is adjusted according to the judgment result to control the amount by which the working end of the seed scraper extends into or out of the working area of the seed metering tray; wherein, when it is judged that double seeding has occurred, the seed scraper is controlled to increase the extension amount; when it is judged that missed seeding has occurred, the seed scraper is controlled to decrease the extension amount.
4. The control method for the seed metering device according to claim 3, characterized in that, The qualified time interval is: ; ； ； ； in The target plant spacing, The speed of travel, This is the allowable relative error coefficient.
5. The control method for the seed metering device according to claim 3, characterized in that, When no reseeding or missed seeding events are detected, a PID control algorithm is used to fine-tune the insertion depth of the seed scraper based on the deviation between the actual average plant spacing and the target plant spacing. When a re-reproduction is detected, the insertion depth of the seed scraper is increased by a first preset step. When it is determined that a missed sowing has occurred, the insertion depth of the seed scraper is reduced by a second preset step.
6. The control method for the seed metering device according to claim 5, characterized in that, The size of the first preset step size and the second preset step size are positively correlated with the number of consecutive occurrences of the same type of abnormal event.
7. The control method for the seed metering device according to claim 5, characterized in that, Also includes: Based on a machine learning model, at least one of the parameters of the PID control algorithm, the first preset step size, the second preset step size, and the reference working position of the seed scraper is optimized according to multiple feature parameters of the operation, so as to maximize the comprehensive score of seeding quality. The characteristic parameters include one or more of the following: traveling speed, seed metering disc rotation speed, seed characteristic parameters, running time, environmental parameters, average plant spacing, plant spacing standard deviation, reseeding rate, and missed seeding rate.
8. The control method for the seed metering device according to claim 7, characterized in that, The training of the machine learning model includes two stages: offline pre-training and online adaptive updating. The offline pre-training process uses historical job data to perform supervised pre-training on the model. The online adaptive update periodically fine-tunes the model parameters based on data collected in real time during operation.
9. The control method for the seed metering device according to claim 7 or 8, characterized in that, The machine learning model is one of the following: deep neural network, reinforcement learning agent, random forest, or gradient boosting decision tree.
10. A seeder, characterized in that, Includes the seed metering device as described in claim 1 or 2 and / or the control method for the seed metering device as described in any one of claims 3 to 9.