A depth self-adaptive control system based on down pressure and compacting pressure

Abstract

The application discloses a sowing depth self-adaptive regulation and control system based on down pressure and compacting pressure, and relates to the technical field of agricultural intelligent machinery, which comprises an input layer, a perception layer, a decision layer and an execution layer: the input layer acquires soil type, water content and compactness prior parameters; the perception layer adopts a strain gauge type first sensor installed on a furrow opener arm and a pin shaft type second sensor replacing a hinged shaft, and collects down pressure and compacting pressure signals in real time; the decision layer is internally provided with a hybrid decision model of mechanism and deep neural network, calculates an optimal pressure target value through prior parameter feedforward, compares the real pressure value fed back by the sensor, and generates a deviation signal; and the execution layer synchronously drives a first hydraulic cylinder and a second hydraulic cylinder, and realizes the collaborative regulation of down pressure and compacting pressure. The application adopts a feedforward-feedback double closed loop architecture, does not need to directly detect the sowing depth, and can reduce the sowing depth variation coefficient to below 5%, thereby significantly improving the corn emergence rate and yield, and being suitable for precision sowing operation.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent machinery technology, specifically to a seeding depth adaptive control system based on downforce and upforce. Background Technology

[0002] The planting depth of corn directly determines the emergence rate, uniformity of emergence, and final yield. Traditional corn planters use mechanical spring pre-compression passive contouring technology, which has three major drawbacks: spatial heterogeneity of soil type, moisture content, and firmness in the field leads to uncontrolled planting depth, with the coefficient of variation typically exceeding 15%; ignoring the coupling relationship between downforce and roll pressure, relying solely on manual preset of spring pressure, makes it impossible to establish a precise mapping; and open-loop control lacks real-time feedback, requiring mid-course adjustments based on experience, resulting in high labor intensity and low precision.

[0003] In recent years, several proactive control schemes have been proposed. For example, Chinese patent application CN117759582A discloses a hydraulic control system for the sowing unit's downpressure and rolling pressure, which uses a soil firmness sensor for real-time feedback control. However, this scheme is still a single closed-loop feedback architecture, without incorporating prior parameters such as soil type and moisture content as feedforwards. Furthermore, the downpressure and rolling pressure are adjusted independently, making it difficult to suppress sowing depth rebound. Another Chinese patent application CN111373909A proposes a control method based on a deep neural network. However, its input layer consists of real-time detected quantities such as furrow depth, downpressure, and rolling pressure, without considering prior soil information before operation. Moreover, its furrow depth relies on ultrasonic detection, making it susceptible to interference from straw and dust in the field. In addition, existing technologies generally use indirect measurement of downpressure at the depth-limiting wheel, which is easily affected by tire slippage and soil subsidence, resulting in low measurement accuracy. The downpressure and rolling pressure are adjusted independently without establishing a quantitative matching relationship, leading to soil rebound after covering and causing variations in sowing depth.

[0004] Therefore, it is necessary to propose an adaptive seeding depth control system based on downforce and dam pressure to solve the above problems. Summary of the Invention

[0005] Technical Problem to be Solved: The purpose of this invention is to provide an adaptive sowing depth control system based on downforce and rolling pressure, to solve the problems of low measurement accuracy, slow response, difficulty in suppressing sowing depth rebound, and poor reliability in existing corn planter sowing depth control technologies mentioned in the background. Existing technologies mostly measure downforce indirectly at the depth-limiting wheel, which is easily affected by slippage and soil subsidence; direct detection by ultrasonic and angle sensors is easily obscured by dust and straw; single closed-loop feedback control is commonly used, which cannot quickly adapt to the spatial heterogeneity of soil conditions; downforce and rolling pressure are adjusted independently without establishing a quantitative matching relationship, making it impossible to suppress soil rebound after covering; pure mechanism or pure data-driven models have insufficient generalization ability, and the system has many sensors, complex structure, and high failure rate.

[0006] Technical Solution: To achieve the above objectives, the present invention provides the following technical solution: a seeding depth adaptive control system based on downforce and sedation pressure, comprising: The input layer is used to obtain field prior parameters for three dimensions: soil type, moisture content, and firmness. The sensing layer includes a first sensor for real-time acquisition of the soil penetration pressure signal of the trencher and a second sensor for real-time acquisition of the soil compaction pressure signal of the roller. The decision-making layer includes a decision-making model and a calibration model. The decision-making model takes the prior parameters as input and calculates and outputs the optimal downforce target value and the optimal stoichiometric pressure target value. The calibration model is used to convert the signals collected by the first sensor and the second sensor into the actual downforce value and the actual stoichiometric pressure value, respectively. The decision-making layer is also used to compare the actual downforce value with the optimal downforce target value and the actual stoichiometric pressure value with the optimal stoichiometric pressure target value to generate a deviation signal. The execution layer includes a first hydraulic cylinder for adjusting the soil entry pressure of the trencher and a second hydraulic cylinder for adjusting the ground pressure of the press wheel; the execution layer responds to the deviation signal and synchronously drives the first and second hydraulic cylinders to move, so that the actual downward pressure value and the actual pressing pressure value approach the corresponding target values.

[0007] Furthermore, the decision-making model is a hybrid decision-making model combining the "soil-down pressure-suppression pressure" coupling mechanism and a deep neural network, including: The coupling mechanism sub-model establishes a physical mapping relationship between soil firmness, moisture content and trencher entry resistance and compaction resistance of roller based on soil mechanical properties, which is used to output the initial pressure reference value. The deep neural network sub-model uses multiple sets of soil parameters, pressure parameters, and corresponding seeding depth data collected from bench tests and field tests as training sets to perform nonlinear correction on the initial pressure reference value, and outputs the final optimal downpressure target value and optimal hill pressure target value.

[0008] Furthermore, the first sensor is a resistance strain gauge sensor, including at least one strain gauge, mounted on the cutter arm of the trencher, used to sense the bending deformation of the cutter arm caused by the soil penetration resistance and output an electrical signal proportional to the downward pressure; the second sensor is a pin-type sensor, which serves as a hinge pin between the piston rod end of the second hydraulic cylinder and the wheel axle support of the press wheel, so as to directly measure the actual pressing pressure signal of the press wheel on the ground.

[0009] Preferably, the first sensor consists of a full-bridge measurement unit composed of four resistance strain gauges, two of which are symmetrically mounted on the front surface of the root of the cutter arm, and the other two are symmetrically mounted on the rear surface of the root of the cutter arm, arranged along the length of the cutter arm.

[0010] Furthermore, the execution layer also includes a hydraulic dynamic adjustment module. The hydraulic dynamic adjustment module adjusts the oil supply pressure or flow rate of the first hydraulic cylinder and the second hydraulic cylinder according to the magnitude and direction of the deviation signal. When the actual downward pressure value is less than the optimal downward pressure target value, the thrust of the first hydraulic cylinder is increased; when the actual downward pressure value is greater than the optimal downward pressure target value, the thrust of the first hydraulic cylinder is decreased. At the same time, the hydraulic dynamic adjustment module adjusts the thrust of the second hydraulic cylinder synchronously according to the adjustment amount of the downward pressure, so that the pressure and the downward pressure maintain a preset matching ratio.

[0011] Furthermore, the matching ratio is pre-calibrated as a proportionality coefficient K based on soil type and moisture content; wherein, the K value is 0.8 to 1.2 under loam conditions, 1.2 to 1.5 under sandy conditions, and 0.5 to 0.8 under clay conditions.

[0012] Furthermore, the decision-making layer is also connected to a visualization display screen, which is used to display the prior parameters, actual downforce value, actual ballast pressure value, target seeding depth, actual seeding depth estimate, and operation status information in real time, and supports the storage and traceability of operation data.

[0013] An adaptive seeding depth control method based on downforce and damming force includes the following steps: S1: Obtain the field prior parameters of three dimensions: soil type, moisture content and firmness, and call the built-in decision model to calculate the optimal downforce target value and the optimal rolling pressure target value; S2: During the operation, the first sensor installed on the trencher cutter arm collects the soil in-soil pressure signal in real time, and the second sensor, which serves as the hinge pin between the piston rod end of the second hydraulic cylinder and the wheel axle support of the pressing wheel, collects the pressing pressure signal in real time. The pre-calibrated calibration model is called to convert the signals into the actual downward pressure value and the actual pressing pressure value, respectively. S3: Compare the actual downward pressure value with the optimal downward pressure target value, and compare the actual pressing pressure value with the optimal pressing pressure target value to generate a deviation signal; in response to the deviation signal, synchronously drive the first hydraulic cylinder to adjust the trencher's soil entry pressure and the second hydraulic cylinder to adjust the pressing wheel's ground pressure, so that the actual pressure value approaches the target value.

[0014] Furthermore, the method for constructing the decision model described in step S1 includes: S11: Establish a coupled mechanism sub-model based on soil mechanics theory and output the initial pressure reference value; S12: Conduct bench tests and field tests under different soil conditions, collect multiple sets of soil parameters, downforce values, rolling pressure values ​​and corresponding actual sowing depth data, and construct a training dataset; S13: Using soil parameters from the training dataset as input, and the difference between the initial pressure reference value and the actual optimal pressure value output by the coupling mechanism sub-model as the target output, train the deep neural network sub-model. S14: Add the output of the coupling mechanism sub-model to the correction value of the deep neural network sub-model to obtain the final optimal downforce target value and optimal stabilizing pressure target value.

[0015] Furthermore, the specific logic of the synchronous drive in step S3 is as follows: a matching ratio coefficient K for the downward pressure and the pressing pressure is preset under different soil conditions; when the first hydraulic cylinder is adjusted to change the downward pressure by ΔF1, the second hydraulic cylinder is simultaneously adjusted to change the pressing pressure by ΔF2 = K × ΔF1; wherein, the value of K is pre-calibrated according to the soil type and moisture content, and the value of K is 0.8 to 1.2 under loam conditions, 1.2 to 1.5 under sandy conditions, and 0.5 to 0.8 under clay conditions.

[0016] Beneficial effects: Compared with the prior art, the present invention provides a seeding depth adaptive control system based on downforce and dam pressure, which has the following beneficial effects: 1. Adopting a dual closed-loop control architecture of "prior parameter feedforward + real-time pressure feedback", the response speed is improved by about 40% compared with a pure feedback control system, which can adapt to the spatial heterogeneity of the soil in advance and avoid sudden changes in sowing depth.

[0017] 2. The unique vertical cutter arm direct mechanical sensing (first sensor) avoids interference from depth-limiting wheel slippage and soil subsidence, reducing the measurement error to within ±3%; the second sensor is directly installed at the hinge pin of the hydraulic cylinder piston rod end and the press wheel axle support, directly measuring the real press pressure signal without additional mechanical conversion, resulting in higher measurement accuracy; the mechanism-data hybrid decision model combines physical interpretability with data-driven self-learning ability, improving the model's generalization ability by more than 30% compared to the pure data model.

[0018] 3. The downforce and rolling pressure are synergistically adjusted, and a matching ratio coefficient K is preset for different soil types to effectively suppress the rebound of sowing depth caused by soil rebound after covering. Field comparative tests show that the sowing depth variation coefficient of this system can be reduced to below 5%, significantly better than existing technologies.

[0019] 4. It eliminates the need for easily interfered direct seeding depth detection devices such as ultrasonic sensors and angle sensors, reducing the system failure rate by approximately 60%; it increases corn emergence rate by 10%–15%, yield by 6%–9%, ​​and operational efficiency by 10%–15%. Attached Figure Description

[0020] Figure 1 This is a front view schematic diagram of the mechanical structure of the present invention; Figure 2 This is a block diagram illustrating the working principle of the present invention; Figure 3 This is a flowchart of the control method of the present invention; Figure 4 This is a flowchart of the dual closed-loop control logic of the present invention.

[0021] In the diagram: 1. Furrow opener; 2. Single main beam connecting frame; 3. Parallel four-bar linkage; 4. First hydraulic cylinder; 5. Sowing unit main frame; 6. Seed box; 7. Air suction precision seed metering assembly; 8. Depth limiting wheel; 9. Press wheel support arm; 10. Press wheel; 11. First sensor; 12. Second sensor; 13. Second hydraulic cylinder. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] Example 1: Overall System Structure like Figure 1 As shown, the adaptive sowing depth control system of this embodiment is integrated and installed on the corn precision sowing unit. The sowing unit is arranged with furrow opener 1, sowing mechanism and pressing mechanism in sequence along the working direction.

[0024] The furrow opener 1 is a double-disc furrow opener located at the foremost end, with its upper vertical blade arm rigidly connected to the lower front of the seeding unit frame 5. The first sensor 11 (downward pressure detection sensor) is installed at the root of the blade arm to collect the soil penetration pressure signal.

[0025] The rear end of the main beam connecting frame 2 is hinged to the right end of the parallel four-bar linkage 3 via a pin. The left end of the parallel four-bar linkage 3 is hinged to the main frame 5 of the seeding unit. The first hydraulic cylinder 4 (downward pressure hydraulic cylinder) is installed at an angle between the upper and lower connecting rods of the parallel four-bar linkage 3. It drives the parallel four-bar linkage to swing through extension and retraction, thereby driving the seeding unit to move up and down.

[0026] Seed box 6 is fixed above the main frame 5 of the sowing unit, and air suction precision seed metering assembly 7 is installed below it. Depth limiting wheel 8 is hinged to both sides of the front part of the main frame 5 of the sowing unit for initial depth limiting.

[0027] The press wheel support arm 9 is an L-shaped swing arm, with its upper end hinged to the left end of the lower connecting rod of the parallel four-bar linkage 3, and its lower end hinged to the cylinder end of the second hydraulic cylinder 13 (press pressure hydraulic cylinder). The piston rod end of the second hydraulic cylinder 3 is hinged to the wheel axle support of the press wheel 10 through the second sensor 12 (pin-type sensor). That is, the second sensor 12 is directly used as the hinge pin between the piston rod end of the second hydraulic cylinder 13 and the wheel axle support of the press wheel 10. The second hydraulic cylinder 13 is installed vertically and directly adjusts the ground pressure of the press wheel 10 by extension and retraction. The second sensor 12 directly measures the actual press pressure signal of the press wheel 10 on the ground at this connection point.

[0028] Example 2: Sensor Installation and Calibration The first sensor 11 uses four resistance strain gauges to form a full-bridge measurement unit with a sensitivity coefficient of 2.0±1% and a range of 0~5000N. Two strain gauges are attached to the front surface of the cutter arm root, and the other two are attached to the rear surface, arranged along the length of the cutter arm to form a Wheatstone bridge. When the trencher enters the soil and encounters horizontal rearward soil resistance, the cutter arm bends back and forth. The tensile resistance of the strain gauges on the front surface increases, while the compressive resistance of the strain gauges on the rear surface decreases. The bridge output is a voltage signal proportional to the downward pressure. This installation method avoids interference from slippage of the depth-limiting wheel and soil subsidence, and the measurement accuracy can reach ±3%.

[0029] The second sensor 12 is a pin-type pressure sensor with a range of 0–3000 N and an accuracy of ±1%FS. This sensor is directly used as a hinge pin between the piston rod end of the second hydraulic cylinder 13 and the axle support of the press wheel 10. When the second hydraulic cylinder 13 applies pressure to the press wheel 10, the sensor directly outputs a signal proportional to the press pressure, without the need for additional mechanical conversion, resulting in higher measurement accuracy.

[0030] The pressure calibration sub-model was pre-calibrated through bench tests: For downward pressure, the trencher assembly was fixed on the test bench, a horizontal backward standard force (0–5000 N) was applied to the center of the disk, the bridge output voltage was recorded, and a voltage-force mapping relationship was established through linear regression, with a calibration determination coefficient R² ≥ 0.998. For ballast pressure, a vertical downward standard force (0–3000 N) was applied to the ballast wheel, the pin sensor output signal was recorded, a mapping relationship was established, and a calibration determination coefficient R² ≥ 0.995.

[0031] Example 3: Construction and Training of Hybrid Decision Model One of the core innovations of this invention is the hybrid decision-making model combining the "soil-down pressure-suppression pressure" coupling mechanism with deep neural networks, the construction process of which is as follows: 3.1 Coupling Mechanism Sub-model Based on the Mohr-Coulomb strength theory and compaction theory in soil mechanics, the physical equations are established as follows: The resistance of the trencher to soil penetration is: F1 = k1 × B × d × (c + σ × tanφ) Compaction resistance of the tamping wheel: F2 = k2 × A × (1 - e^(-k3 × ρ)) Wherein, k1 is the furrow opener shape coefficient (calibration range 0.8–1.2), B is the furrow opener width (m), d is the sowing depth (m), c is the soil cohesion (kPa), σ is the soil vertical stress (kPa), and φ is the internal friction angle (°); k2 and k3 are the soil compaction coefficients (k2 calibration range 0.5–0.9, k3 calibration range 2–5), A is the ground contact area of ​​the roller (m²), and ρ is the soil compaction degree (g / cm³). All of the above parameters are related to soil type, moisture content, and firmness, and a corresponding relationship database was established through extensive soil mechanics experiments. Given the input parameters, the mechanistic sub-model outputs initial pressure reference values ​​F1_mechanism and F2_mechanism.

[0032] 3.2 Deep Neural Network Sub-model A three-layer BP neural network structure is adopted: 3 nodes in the input layer (soil type encoding value, moisture content, and firmness), 10 nodes in the hidden layer (optimized through experiments), and 2 nodes in the output layer (downward pressure correction value ΔF1_nn and downward pressure correction value ΔF2_nn).

[0033] Three typical soil types—sandy, loam, and clay—were selected in major maize-producing areas across the country. Bench and field experiments were conducted within a moisture content range of 10%–35% and a firmness range of 50–500 kPa, collecting 10,000 sets of valid data. 70% of this data was used as the training set, 15% as the validation set, and 15% as the test set. The difference between the output value of the coupled mechanism sub-model and the actual optimal pressure value was used as the target output to train the neural network. Testing showed that the mean absolute error on the test set was ±35 N for downforce and ±28 N for upforce.

[0034] 3.3 Model Fusion Final target values: F1_target = F1_mechanism + ΔF1_nn; F2_target = F2_mechanism + ΔF2_nn.

[0035] Example 4: Adaptive Control Method for Seeding Depth Step S1: Obtaining Prior Parameters and Initializing the Model Before operation, the driver inputs the target sowing depth (usually 3-5 cm) and the current soil type, moisture content, and compaction of the field via display screen 2. Controller 3 invokes a hybrid decision model to calculate the optimal downforce and compaction pressure target values. For example, with loam soil, 20% moisture content, 200 kPa compaction pressure, and a target sowing depth of 4 cm, the mechanistic sub-model outputs an initial downforce of 1150 N and an initial compaction pressure of 980 N; the neural network outputs correction values ​​of +50 N and +20 N; the final target values ​​are a downforce of 1200 N and a compaction pressure of 1000 N.

[0036] Step S2: Real-time pressure signal acquisition and calibration During the operation of the seeder, the first sensor 11 collects the soil pressure electrical signal at a frequency of 100Hz, and the second sensor 12 (which serves as the hinge pin between the piston rod end of the second hydraulic cylinder 13 and the axle support of the press wheel 10) collects the pressing pressure electrical signal at the same frequency. The signals are filtered, amplified, and then transmitted to the controller 3. The controller 3 calls the pressure calibration sub-model to convert the electrical signal into a real mechanical value.

[0037] Step S3: Dual-parameter synchronous closed-loop control Controller 3 compares the actual downward pressure value with the target downward pressure value to generate a deviation signal; it also compares the actual downward pressure value with the target downward pressure value to generate another deviation signal. The hydraulic dynamic adjustment module synchronously adjusts the oil supply pressure of the two hydraulic cylinders based on the deviation signals. If the actual downward pressure is 1000N < the target value of 1200N, increase the thrust of the first hydraulic cylinder 4 to increase the downward pressure to 1200N; at the same time, according to the matching ratio coefficient K (K=0.8 for soil conditions), simultaneously increase the thrust of the second hydraulic cylinder 13 to increase the pressure to the target value of 1000N.

[0038] If the actual downward pressure is greater than the target value, the thrust of the first hydraulic cylinder is reduced, and the thrust of the second hydraulic cylinder is reduced simultaneously.

[0039] The entire adjustment process has a response time of less than 200ms, enabling it to quickly adapt to changes in soil conditions.

[0040] Step S4: Job Status Monitoring and Data Storage All operational data (prior parameters, real-time pressure values, target seeding depth, estimated actual seeding depth, etc.) are displayed in real-time on display screen 2 and stored in the built-in storage unit of controller 3 (at least 100 hours of data). The estimated actual seeding depth is obtained through a seeding depth inversion model, which is established based on experimental data regression: Actual seeding depth = a × F downforce + b × F pressure + c × soil hardness + d. After calibration, the model R² = 0.92, eliminating the need to rely on ultrasonic or displacement sensors to directly detect seeding depth.

[0041] Calibration of the matching ratio coefficient K Field trials were conducted to test the effect of different K values ​​on the stability of sowing depth under different soil types and moisture content combinations, and the optimal range of K values ​​was obtained: Loam (moisture content 15%–25%): K = 0.8–1.2 Sandy soil (moisture content 10%–20%): K = 1.2–1.5 Clay (moisture content 20%–35%): K = 0.5–0.8 In actual control, the controller automatically selects the corresponding K value based on the input prior parameters.

[0042] Comparative test results This system was compared with a conventional mechanical spring-type seeder and a pure feedback hydraulic system (refer to CN117759582A) in the same field (each was repeated 5 times, with 200 sowing depth points measured each time). The results are as follows: Seeding depth variation coefficient: mechanical type 15.6%, pure feedback hydraulic type 9.8%, this invention 4.8%; Seedling emergence rate: Mechanical type 82%, pure feedback hydraulic type 88%, this invention 94%; Production: Mechanical standard 100%, pure feedback hydraulic 105%, this invention 109%.

[0043] The above data verifies the significant progress of this invention.

[0044] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A seeding depth adaptive control system based on downforce and dam pressure, comprising: The input layer is used to obtain field prior parameters for three dimensions: soil type, moisture content, and firmness. The sensing layer includes a first sensor (11) for real-time acquisition of the soil pressure signal of the trencher (1) and a second sensor (12) for real-time acquisition of the pressing pressure signal of the roller (10). The decision layer has a built-in decision model and a calibration model. The decision model takes the prior parameters as input and calculates and outputs the optimal downforce target value and the optimal stoking pressure target value. The calibration model is used to convert the signals collected by the first sensor (11) and the second sensor (12) into the actual downforce value and the actual stoking pressure value, respectively. The decision layer is also used to compare the actual downforce value with the optimal downforce target value and the actual stoking pressure value with the optimal stoking pressure target value to generate a deviation signal. The execution layer includes a first hydraulic cylinder (4) for adjusting the soil entry pressure of the trencher and a second hydraulic cylinder (13) for adjusting the ground pressure of the press wheel (10); the execution layer responds to the deviation signal and synchronously drives the first hydraulic cylinder (4) and the second hydraulic cylinder (13) to move, so that the actual downward pressure value and the actual pressing pressure value approach the corresponding target value.

2. The seeding depth adaptive control system according to claim 1, characterized in that, The decision-making model is a hybrid decision-making model combining the "soil-down pressure-repression pressure" coupling mechanism and a deep neural network, including: The coupling mechanism sub-model establishes a physical mapping relationship between soil firmness, moisture content and trencher entry resistance and compaction resistance of the press wheel (10) based on soil mechanical properties, which is used to output the initial pressure reference value. The deep neural network sub-model uses multiple sets of soil parameters, pressure parameters, and corresponding seeding depth data collected from bench tests and field tests as training sets to perform nonlinear correction on the initial pressure reference value, and outputs the final optimal downpressure target value and optimal hill pressure target value.

3. The seeding depth adaptive control system according to claim 1, characterized in that, The first sensor (11) is a resistance strain gauge sensor, including at least one strain gauge, which is installed on the cutter arm of the trencher to sense the bending deformation of the cutter arm caused by the soil penetration resistance and output an electrical signal proportional to the downward pressure; the second sensor (12) is a pin-type sensor, which is used as a hinge pin between the piston rod end of the second hydraulic cylinder (13) and the wheel axle support of the press wheel (10) to directly measure the actual pressing pressure signal of the press wheel (10) on the ground.

4. The seeding depth adaptive control system according to claim 3, characterized in that, The first sensor (11) consists of a full-bridge measurement unit composed of four resistance strain gauges, two of which are symmetrically installed on the front surface of the root of the cutter arm, and the other two are symmetrically installed on the rear surface of the root of the cutter arm, arranged along the length of the cutter arm.

5. The seeding depth adaptive control system according to claim 1, characterized in that, The execution layer also includes a hydraulic dynamic adjustment module, which adjusts the oil supply pressure or flow rate of the first hydraulic cylinder (4) and the second hydraulic cylinder (13) according to the magnitude and direction of the deviation signal; when the actual downward pressure value is less than the optimal downward pressure target value, the thrust of the first hydraulic cylinder (4) is increased; when the actual downward pressure value is greater than the optimal downward pressure target value, the thrust of the first hydraulic cylinder (4) is decreased; at the same time, the hydraulic dynamic adjustment module adjusts the thrust of the second hydraulic cylinder (13) synchronously according to the adjustment amount of the downward pressure, so that the pressure and the downward pressure maintain a preset matching ratio.

6. The seeding depth adaptive control system according to claim 5, characterized in that, The matching ratio is pre-calibrated as a proportional coefficient K based on soil type and moisture content; wherein, the K value is 0.8 to 1.2 for loam, 1.2 to 1.5 for sand, and 0.5 to 0.8 for clay.

7. The seeding depth adaptive control system according to claim 1, characterized in that, The decision-making layer is also connected to a visualization display screen, which is used to display the prior parameters, actual downforce value, actual tamping pressure value, target seeding depth, actual seeding depth estimate, and operation status information in real time, and supports the storage and traceability of operation data.

8. A method for adaptive seeding depth control based on downforce and damming force, characterized in that, Includes the following steps: S1: Obtain the field prior parameters of three dimensions: soil type, moisture content and firmness, and call the built-in decision model to calculate the optimal downforce target value and the optimal rolling pressure target value; S2: During the operation, the soil indentation pressure signal is collected in real time by the first sensor (4) installed on the cutter arm of the trencher (1), and the tamping pressure signal is collected in real time by the second sensor (12) which serves as the hinge pin between the piston rod end of the second hydraulic cylinder (13) and the wheel axle support of the tamping wheel (10). The pre-calibrated calibration model is called to convert the signals into the actual downward pressure value and the actual tamping pressure value, respectively. S3: Compare the actual downward pressure value with the optimal downward pressure target value and the actual tamping pressure value with the optimal tamping pressure target value to generate a deviation signal; respond to the deviation signal, synchronously drive the first hydraulic cylinder (4) to adjust the trencher's soil entry pressure and the second hydraulic cylinder (13) to adjust the tamping wheel (10)'s ground pressure so that the actual pressure value approaches the target value.

9. The seeding depth adaptive control method according to claim 8, characterized in that, The method for constructing the decision model in step S1 includes: S11: Establish a coupled mechanism sub-model based on soil mechanics theory and output the initial pressure reference value; S12: Conduct bench tests and field tests under different soil conditions, collect multiple sets of soil parameters, downforce values, rolling pressure values ​​and corresponding actual sowing depth data, and construct a training dataset; S13: Using soil parameters from the training dataset as input, and the difference between the initial pressure reference value and the actual optimal pressure value output by the coupling mechanism sub-model as the target output, train the deep neural network sub-model. S14: Add the output of the coupling mechanism sub-model to the correction value of the deep neural network sub-model to obtain the final optimal downforce target value and optimal stabilizing pressure target value.

10. The seeding depth adaptive control method according to claim 8, characterized in that, The specific logic of the synchronous drive in step S3 is as follows: preset the matching ratio coefficient K of the downward pressure and the pressing pressure under different soil conditions; when the first hydraulic cylinder (4) is adjusted to change the downward pressure by ΔF1, the second hydraulic cylinder (13) is synchronously adjusted to change the pressing pressure by ΔF2 = K × ΔF1; wherein, the value of K is pre-calibrated according to the soil type and moisture content, and the value of K is 0.8 to 1.2 under loam conditions, 1.2 to 1.5 under sandy conditions, and 0.5 to 0.8 under clay conditions.

Images