A dynamic sorting method and system based on multi-stage screening and air flow compensation

Abstract

The present application relates to the technical field of screening, and particularly relates to a dynamic sorting method and system based on multi-stage screening and air flow compensation. The method comprises the following steps: step 1: spatial phase screening pretreatment: in a rotary multidirectional screening channel, white rice is made to form a rotary motion through a flow guide structure, the angle between the long axis direction of the rice and the long side of the screen hole is acquired in real time, and the rotation angular velocity of the screen is dynamically adjusted according to the angle; step 2: mass-air flow coupled sorting; step 3: deformation feedback optical sorting; step 4: energy field cooperative control; and step 5: dynamic convergence optimization: the screen hole pass rate, the rice group distribution entropy value and the total energy consumption power of the system are collected, a state space is constructed, and a stability determination algorithm is used to adjust the vibration frequency, the air flow pressure and the light intensity parameters. The method not only improves the screening precision, but also reduces the energy consumption and improves the production efficiency, and has a significant practical application value for the grain processing industry.

Description

Technical Field

[0001] This invention relates to the field of screening technology, and in particular to a dynamic sorting method and system based on multi-stage screening and airflow compensation. Background Technology

[0002] In the grain processing sector, especially in the refining of white rice, efficient sorting technology is a core factor determining the quality and economic benefits of the finished product. Current technologies primarily rely on a combination of mechanical screening and airflow separation, but these methods have the following drawbacks in practical applications:

[0003] Traditional screening equipment uses fixed-shape screen holes and a static tilt angle design, which cannot adapt to the dynamic changes in the aspect ratio of rice grains (such as the mixed flow of whole and broken rice). When there is an angle between the long axis of the rice grains and the direction of the screen hole arrangement, it is easy to cause misscreening (whole rice grains are mistakenly judged as broken rice) and missed screening (broken rice is not effectively separated). More seriously, when rice grains accumulate on the screen surface, they create a pore blockage effect, forcing an increase in vibration intensity, which in turn leads to an increase in the rice grain breakage rate (the measured broken rice rate is generally over 5%).

[0004] Existing airflow sorting devices mostly employ a constant pressure air supply mode, which cannot adjust airflow parameters according to real-time changes in rice grain density and falling velocity. Especially when processing rice grains with similar mass but different internal densities (such as insect-damaged rice and whole rice), a single airflow field is insufficient to form an effective resistance gradient, leading to a sharp decrease in sorting accuracy (the residual rate of light impurities is as high as 8-12%). In addition, the lack of coordinated control between airflow and vibrating sieving results in low energy utilization (more than 40% of the system's energy consumption is wasted in ineffective turbulence). Summary of the Invention

[0005] To achieve the above objectives, the present invention provides a dynamic sorting method and system based on multi-stage screening and airflow compensation. The dynamic sorting method based on multi-stage screening and airflow compensation includes the following steps:

[0006] Step 1: Spatial phase sieving pretreatment: In the rotary multi-directional sieving channel, the white rice is made to rotate through the flow guiding structure, and the angle between the long axis of the rice grain and the long side of the sieve hole is obtained in real time. The rotational angular velocity of the sieve is dynamically adjusted according to the angle.

[0007] Step 2: Mass-Airflow Coupled Separation: Construct a gas-solid two-phase flow mass separation field at the screening outlet, measure the density and initial velocity of the rice grains, calculate the airflow resistance, and configure the elevation angle and airflow velocity of the multi-layer airflow nozzles. Set up a negative pressure adsorption structure at the bottom of the separation zone.

[0008] Step 3: Deformation feedback optical sorting: During the free fall stage of rice grains, a dual-wavelength laser irradiation system is used to capture the intensity distribution of scattered light. Based on the light intensity ratio, the deformed rice grains are determined and sorted through airflow nozzles.

[0009] Step 4: Coordinated control of energy fields: Divide the phase window of the vibration period and calculate the ratio of kinetic energy to potential energy, and adjust the vibration acceleration and airflow pressure synchronously based on the ratio;

[0010] Step 5: Dynamic convergence optimization: Collect the screen aperture throughput, rice grain distribution entropy, and total system power consumption to construct the state space and use a stability judgment algorithm to adjust the vibration frequency, airflow pressure, and light intensity parameters.

[0011] Preferably, step 1 includes:

[0012] The flow guiding structure is a spiral flow guiding plate, whose spiral angle is dynamically adjusted according to the average length-to-diameter ratio of rice, and the surface of the flow guiding plate is covered with a low friction coefficient material.

[0013] The image acquisition device continuously captures the movement trajectory of rice grains, extracts the main direction of the rice grain outline, and calculates the real-time angle between the rice grain outline and the long side of the sieve hole. The frame rate of the image acquisition device is positively correlated with the rice grain flow rate.

[0014] When the real-time included angle exceeds the preset angle threshold, the rotational angular velocity of the screen is increased to above the critical angular velocity, which is calculated based on the screen diameter and the average mass of the rice grains.

[0015] A vibration detection array is set on the back of the screen to collect the vibration signal spectrum generated by the collision of rice grains, identify the energy ratio of the rigid collision characteristic frequency band and the damped collision characteristic frequency band, and trigger the screen amplitude compensation mechanism when the damped collision energy ratio exceeds the preset ratio threshold.

[0016] The amplitude compensation mechanism includes: nonlinearly adjusting the longitudinal vibration amplitude of the screen according to the mapping relationship between the damping collision energy ratio and the screen porosity, and maintaining the transverse tension of the screen within a safe threshold range during the adjustment process.

[0017] Preferably, step 2 includes:

[0018] The multi-layer airflow nozzles are arranged in a multi-layer annular array structure, with each layer containing multiple adjustable nozzles evenly distributed in the circumference, and the radial deflection angles of the nozzles in adjacent layers are staggered.

[0019] When measuring the initial velocity of falling rice grains, a non-contact velocity sensing device is used, whose measurement area covers the entire cross-section of the separation field, and the measurement error caused by occlusion between rice grains is eliminated by a time series correlation algorithm.

[0020] When calculating airflow resistance, the theoretical resistance value is corrected based on the standard deviation of the grain density distribution. The correction factor is exponentially related to the dispersion of the density distribution.

[0021] When configuring the nozzle elevation angle, an iterative approximation algorithm is used to match the actual airflow velocity field with the theoretical resistance gradient field, and the adjustment amount in each iteration does not exceed the set proportion of the nozzle's maximum adjustment range;

[0022] The adsorption intensity of the negative pressure adsorption structure is dynamically adjusted according to the real-time detected impurity concentration. The impurity concentration is calculated by measuring the rate of change of transmitted light intensity using an optical sensor array. The adsorption gas flow rate and the increase in impurity concentration have a non-linear increasing relationship.

[0023] Preferably, step 3 includes:

[0024] The dual-wavelength laser irradiation system includes a first wavelength light source and a second wavelength light source arranged coaxially, and the difference between the two wavelengths is greater than a preset spectral interval threshold.

[0025] The scattered light intensity distribution is collected using a high dynamic range CMOS sensor. Its exposure time is dynamically adjusted according to the shading rate of the falling rice grains, and multi-frame image fusion is performed in each collection cycle to eliminate motion blur.

[0026] When calculating the light intensity ratio, each grain of rice projection area is sampled in sections to exclude noise data caused by edge diffraction effects. The ratio of the effective sampling area to the total projection area is not less than a set threshold.

[0027] When determining the surface deformation of rice grains, a sliding window algorithm is used to analyze the trend of the light intensity ratio data of continuous rice grains. When the rate of change of light intensity ratio exceeds the preset fluctuation threshold, a secondary verification process is initiated.

[0028] The airflow nozzle sorting includes multi-stage pressure control. The initial pressure is set according to the median of the rice grain mass distribution, and the subsequent pressure adjustment is positively correlated with the confidence level of deformation determination.

[0029] Preferably, step 4 includes:

[0030] The phase window division adopts an adaptive time slicing algorithm. The window duration is proportional to the reciprocal of the current vibration frequency, and the minimum window duration is limited by the signal response delay of the control system.

[0031] When calculating kinetic energy, the trajectory of the rice grain group is reconstructed based on the discrete element simulation model. The instantaneous velocity of each rice grain is processed by Kalman filtering to reduce noise, and the data of rice grains that collide with the screen are excluded.

[0032] Potential energy calculation includes the integral of the force exerted by the airflow pressure field on the rice grain group, and the integration region is dynamically subdivided into grids based on the real-time rice grain distribution density.

[0033] When the ratio of kinetic energy to potential energy exceeds the critical value, a multi-parameter coordinated adjustment strategy is activated: the vibration acceleration adjustment is calculated according to the extent of the ratio exceeding the critical value using a piecewise function, and the airflow pressure adjustment is negatively correlated with the vibration acceleration adjustment.

[0034] All parameter adjustment commands must pass through the timing verification module to ensure that the control signals of the vibration system and the airflow system are executed synchronously within the preset phase difference range.

[0035] Preferably, step 5 includes:

[0036] The screen mesh throughput is calculated by cross-validation using a mass flow meter and an image recognition system, eliminating abnormal data points caused by airflow disturbances.

[0037] The entropy value of the rice grain distribution is calculated using an improved Shannon entropy algorithm, which discretizes the falling trajectory into three-dimensional grid cells. The probability density of each cell is determined by the proportion of rice grains that remain in that cell.

[0038] When constructing the state space, the pore throughput, distribution entropy value and energy consumption power are dimensionless and reduced to the observable space by principal component analysis.

[0039] The stability determination algorithm includes Lyapunov exponent calculation and phase space reconstruction. When the exponent exceeds the divergence threshold, a multi-objective optimization algorithm is activated to synchronously adjust the control parameters.

[0040] During parameter adjustment, the condition number of the system Jacobian matrix is ​​monitored in real time. When the condition number exceeds the preset stability threshold, the current adjustment amount is frozen and the system is switched to the backup control strategy.

[0041] Preferably, the screen amplitude compensation mechanism includes:

[0042] A transfer function model of the damped collision energy ratio and amplitude compensation was established, and its nonlinear coefficients were calibrated through impact tests.

[0043] When performing amplitude compensation, a feedforward-feedback composite control structure is adopted: the feedforward control quantity is calculated based on the transfer function model, and the feedback control quantity is PID adjusted according to the residual of the damped collision energy ratio after compensation.

[0044] Lateral tension monitoring is achieved through fiber optic strain sensors embedded in the edge of the screen, with a sampling frequency higher than a set multiple of the screen vibration fundamental frequency;

[0045] When the lateral tension approaches the safety threshold, an amplitude adjustment cooling cycle is automatically inserted, during which the vibration energy input is gradually reduced until the tension returns to the safe range.

[0046] Preferably, the iterative approximation algorithm includes:

[0047] Initialize the nozzle elevation angle to a preset percentage of the theoretically optimal angle, and set the maximum number of iterations and the convergence accuracy threshold;

[0048] In each iteration, a local optimum is searched using the particle swarm optimization algorithm, and a nozzle mechanical adjustment inertia compensation factor is introduced into the particle velocity update formula.

[0049] The convergence criteria are that the change in the objective function in multiple consecutive iterations is less than the convergence accuracy threshold, and the airflow velocity field uniformity index meets the minimum requirements in the grading standard.

[0050] If the number of iterations reaches the upper limit and convergence is still not achieved, the system will automatically switch to an experience-based mode based on historical data and trigger a system calibration alarm.

[0051] Preferably, the secondary verification process includes:

[0052] For rice grains with excessive light intensity ratio change rate, initiate multispectral scanning, add at least two auxiliary wavelengths for irradiation, and reconstruct a three-dimensional model of surface deformation.

[0053] Input the feature vector of the 3D model into a pre-trained neural network classifier and output a deformation type confidence score;

[0054] When the confidence score is lower than the judgment threshold, the grain of rice is marked as an object to be reviewed and temporarily stored in the buffer isolation area;

[0055] A contact deformation detection device is installed in the buffer isolation zone. The recovery curve of rice grains under pressure is measured by a micro-force sensor array. Finally, the sorting decision is generated by combining the optical judgment and the contact detection results.

[0056] Accordingly, embodiments of the present invention also provide a dynamic sorting system based on multi-stage screening and airflow compensation, used to run the dynamic sorting method based on multi-stage screening and airflow compensation described in the embodiments of the present invention, including:

[0057] The rotary screening pretreatment module includes:

[0058] The spiral guide unit is equipped with a guide plate with a variable spiral angle, and its surface coating friction coefficient is less than the static friction coefficient between rice and metal.

[0059] The screen posture adjustment unit includes a servo motor-driven rotating screen frame, which is rigidly connected to the screen shaft via a coupling.

[0060] The motion trajectory acquisition unit consists of a high-speed industrial camera array. The camera optical axis is installed at an angle to the screen plane, and the output end is connected to the screen attitude adjustment unit.

[0061] The vibration spectrum analysis unit includes a piezoelectric sensor array and a signal conditioning circuit attached to the back of the screen, and its output is connected to the screen amplitude compensation controller.

[0062] The mass-airflow coupled sorting module includes:

[0063] The velocity field measurement unit consists of a laser Doppler velocimeter array, and the measurement area covers the entire cross-section of the airflow separation cavity.

[0064] The multi-layer airflow control unit includes a coaxially nested multi-layer annular nozzle support, with multiple adjustable nozzles evenly distributed circumferentially in each support layer, and the nozzle elevation angle is driven by a stepper motor.

[0065] The negative pressure adsorption unit consists of a centrifugal fan, a cavity pressure sensor, and a solenoid valve array. An optical impurity detection window is provided at the cavity inlet.

[0066] Deformation optical sorting module, including:

[0067] The dual-wavelength light source unit includes a first laser emitter and a second laser emitter mounted coaxially, with the difference between the two wavelengths being greater than a preset spectral interval;

[0068] The scattered light acquisition unit consists of a high frame rate CMOS sensor and an optical beam splitter, with the sensor output connected to the image processing board.

[0069] The pneumatic sorting unit includes a high-pressure air source, a proportional valve, and an array of nozzles, with the nozzle outlet direction orthogonal to the trajectory of the falling rice grains.

[0070] The energy coordination control module includes:

[0071] The vibration phase control unit includes an encoder and a vibration exciter coupled to the screen drive shaft, and the encoder output signal is connected to the phase division processor;

[0072] The airflow field adjustment unit consists of a pressure transmitter, a proportional-integral control valve, and a flow meter. Its control signal is synchronized with the output of the vibration phase control unit.

[0073] The dynamic optimization decision-making module includes:

[0074] A multi-source data acquisition unit connects to a screen porosity sensor, a distribution entropy calculation chip, and a power metering module.

[0075] The central processing unit has a built-in state space reconstruction algorithm and Lyapunov exponent calculation core, and its output is connected to each actuator via a CAN bus.

[0076] The connections between modules are as follows:

[0077] The screen outlet of the rotary screening pretreatment module is connected to the inlet of the mass-air-flow coupling sorting module through an airtight pipe, and the inner wall of the pipe is provided with an electrostatic elimination coating.

[0078] The light source unit and the scattered light acquisition unit of the deformation optical sorting module are synchronized through an optical fiber bundle, and the output end of the image processing board is electrically connected to the control terminal of the pneumatic sorting execution unit.

[0079] The output of the vibration phase control unit of the energy collaborative control module is interconnected with the servo motor driver signal of the rotary screening pretreatment module, and the pressure feedback signal of the airflow field adjustment unit is connected to the nozzle controller of the mass-airflow coupling sorting module.

[0080] The multi-source data acquisition unit of the dynamic optimization decision module receives real-time data from each sensor via industrial Ethernet, and the control commands generated by the central processing unit are simultaneously sent to the vibration phase control unit, the airflow field adjustment unit, and the pneumatic sorting execution unit.

[0081] The beneficial effects of this invention are:

[0082] Through the close integration and coordination of the above steps, this sorting method achieves efficient and precise screening of white rice. Each step, through feedback mechanisms and collaborative control, optimizes the sorting effect, ensuring high-quality separation of white rice. This method not only improves screening accuracy but also reduces energy consumption and increases production efficiency, demonstrating significant practical application value for the grain processing industry. Attached Figure Description

[0083] 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 only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0084] Figure 1 This is a flowchart of the steps of the method of the present invention;

[0085] Figure 2 This is a flowchart of step 4 of the method of the present invention;

[0086] Figure 3 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0087] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0088] Please see Figures 1-3 This invention provides a dynamic sorting method based on multi-stage sieving and airflow compensation. In step 1, a rotary multi-directional sieving channel and a flow guiding structure are used to induce rotational motion in the rice grains. This allows for dynamic adjustment of the rice grain's movement path and the angle between the sieve openings. The flow guiding structure can be designed to adjust the angular velocity of the sieve in real time based on the physical characteristics of the rice grains (such as their major axis). This operation ensures that the rice grains enter the sieve at the optimal angle, thereby improving the accuracy and efficiency of the sieving process.

[0089] This step, by dynamically adjusting the angular velocity of the screen, makes the fit between the rice grains and the screen holes more precise, ensuring the high efficiency of the initial screening and providing a stable foundation for subsequent sorting.

[0090] In step 2, a gas-solid two-phase flow mass separation field was constructed, and the airflow resistance was calculated by measuring the density and initial velocity of the rice grains. Subsequently, the elevation angle and airflow velocity of the multi-layer airflow nozzles were configured to enable the airflow to effectively separate the rice grains. A negative pressure adsorption structure was set at the bottom of the separation zone to adsorb impurity rice grains, thereby further improving the purity of the sorting.

[0091] By coupling airflow with sieving, impurities can be effectively removed during the sieving process. The configuration of airflow can be adjusted according to the physical characteristics of rice grains, thereby improving the sorting accuracy and ensuring efficient separation of rice grains.

[0092] In step 3, during the free fall phase of the rice grains, a dual-wavelength laser irradiation system and scattered light intensity distribution capture were employed. The analysis of the light intensity ratio determined whether the rice grains had deformed. Based on the changes in the light intensity ratio, deformed rice grains were sorted using airflow nozzles. Throughout this process, high-precision optical sensors ensured accurate detection of deformed rice grains.

[0093] This step utilizes advanced optical feedback technology to accurately identify minute deformations on the surface of rice grains, thereby further improving screening accuracy. By dynamically adjusting the sorting process of the airflow nozzles, it ensures that deformed rice grains can be effectively separated, reducing screening errors.

[0094] Step 4 involves dividing the vibration period into phase windows and calculating the ratio of kinetic energy to potential energy, thereby synchronously adjusting the vibration acceleration and airflow pressure. This method ensures that the vibration and airflow field during the screening process can be adjusted in real time according to the dynamic changes of the rice grains, thus achieving optimal sorting results.

[0095] By coordinating the control of the energy field, the coordination between vibration and airflow can be effectively adjusted, enhancing the stability of the screening system and ensuring screening efficiency and the accuracy of rice grain separation.

[0096] Finally, in step 5, by collecting data such as the pore throughput of the screen, the distribution entropy of the rice grain population, and the total energy consumption of the system, a state space is constructed, and the vibration frequency, airflow pressure, and light intensity parameters are adjusted using a stability judgment algorithm. This process can automatically adjust various parameters during the operation of the entire screening system to ensure the stability and efficiency of the system.

[0097] This step, through a dynamic convergence optimization algorithm, enables automatic adjustment of system parameters, ensuring that the system can maintain optimal working conditions regardless of changing external environments or the type of rice grains during the screening process, thereby improving screening accuracy and system stability.

[0098] Through the close integration and coordination of the above steps, this sorting method achieves efficient and precise screening of white rice. Each step, through feedback mechanisms and collaborative control, optimizes the sorting effect, ensuring high-quality separation of white rice. This method not only improves screening accuracy but also reduces energy consumption and increases production efficiency, demonstrating significant practical application value for the grain processing industry.

[0099] In one possible implementation, a spiral guide plate is first used as the flow guiding structure. The spiral angle is dynamically adjusted according to the average length-to-diameter ratio of the rice grains. This design allows the guide plate to more effectively guide the rice grains along a predetermined trajectory, avoiding random distribution caused by rotation and collisions. Furthermore, the surface of the guide plate is covered with a low-friction coefficient material, reducing friction between the rice grains and the guide plate, thereby reducing energy loss during movement and improving sorting efficiency.

[0100] This design effectively controls the movement trajectory of rice grains, optimizes the screening effect, reduces damage to rice grains during the screening process, and improves the energy utilization efficiency of the entire system.

[0101] Furthermore, the movement trajectory of the rice grains is captured in real time using an image acquisition device. This device adjusts its frame rate based on changes in the rice grain flow rate to more accurately extract the main direction of the grain outline and calculate the real-time angle between the rice grain and the long side of the sieve opening. This method utilizes dynamic image analysis technology to monitor changes in the direction of rice grain movement in real time and respond promptly.

[0102] When the image acquisition device detects that the real-time angle between the rice grain and the long side of the sieve hole exceeds a preset angle threshold, the system automatically increases the rotational angular velocity of the sieve to above the critical angular velocity. The critical angular velocity is calculated based on the sieve diameter and the average mass of the rice grain. This operation ensures that the rice grain enters the sieve at an appropriate speed, avoiding uneven sorting caused by an inappropriate angle.

[0103] A vibration detection array is installed on the back of the screen, which can collect the vibration signal spectrum generated by the collision of rice grains. By analyzing the energy ratio of the rigid collision characteristic frequency band and the damped collision characteristic frequency band in the spectrum, the system will automatically trigger the screen amplitude compensation mechanism when the damped collision energy ratio exceeds a preset threshold.

[0104] Once the amplitude compensation mechanism is triggered, the system nonlinearly adjusts the longitudinal vibration amplitude of the screen based on the mapping relationship between the damped collision energy ratio and the screen porosity. During this process, it is also crucial to ensure that the transverse tension of the screen remains within a safe threshold range to prevent damage or deformation due to excessive adjustment.

[0105] Through the detailed design and execution of the above steps, this method fully utilizes dynamic adjustment and feedback mechanisms to ensure that each link works efficiently and accurately in tandem. The integration of multiple technologies, including flow guiding structures, image acquisition, vibration monitoring, and compensation mechanisms, effectively improves the screening accuracy, efficiency, and system stability during the white rice sorting process. Each step is closely interconnected, ensuring that rice grains are sorted in the optimal manner, thereby improving the overall performance of the sorting system.

[0106] In one possible implementation, a multi-layered annular array structure is first used to arrange the airflow nozzles. Each layer of nozzles is circumferentially evenly distributed, and the radial deflection angles of nozzles in adjacent layers are staggered. This design helps optimize airflow distribution, allowing the airflow to act more evenly on the rice grains, thereby improving the accuracy of rice grain sorting. Each nozzle can adjust the direction and speed of the airflow to ensure optimal airflow distribution.

[0107] By arranging multiple nozzles in an alternating pattern, the unevenness of airflow can be effectively reduced, and the concentrated airflow impact caused by a single nozzle can be avoided, thereby improving the separation effect of white rice grains, especially the adaptability to different characteristics of white rice (such as shape, size, density, etc.) during the sorting process.

[0108] A non-contact velocity sensor was used to measure the initial velocity of the falling rice grains. This device covers the entire cross-section of the separation field and eliminates measurement errors caused by obstruction between rice grains using a time-series correlation algorithm. This design ensures accurate measurement of the falling velocity of each rice grain, providing data support for subsequent airflow adjustment.

[0109] Using a non-contact velocity sensor avoids errors or damage caused by direct contact, and algorithms eliminate errors caused by obstructions, ensuring the accuracy of velocity measurements. Accurate initial velocity data provides a precise basis for airflow adjustment, thereby effectively improving sorting efficiency and precision.

[0110] When calculating airflow resistance, the theoretical resistance value is corrected based on the standard deviation of the rice grain density distribution. The correction factor has an exponential relationship with the dispersion of the density distribution. The larger the standard deviation of the density distribution, the more significant the impact of the correction factor, thus making the calculation of airflow resistance more accurate and better adaptable to the sorting needs of different types of rice grains.

[0111] The revised calculation method for airflow resistance enables the system to handle rice grains of different densities and dynamically adjusts the airflow resistance based on actual conditions, thus achieving more precise sorting. By introducing density correction, missorting of rice grains of different densities is avoided, enhancing the system's versatility and accuracy.

[0112] An iterative approximation algorithm was used to configure the nozzle elevation angle, ensuring that the actual airflow velocity field matched the theoretical drag gradient field. The adjustment amount in each iteration was limited to a set proportion of the nozzle's maximum adjustment range. In this way, the airflow velocity field could be continuously adjusted and optimized, achieving the ideal match between the falling velocity of the rice grains and the drag gradient.

[0113] The iterative approximation algorithm can precisely adjust the nozzle elevation angle, ensuring a high degree of matching between the airflow velocity field and the movement state of the rice grains, thus guaranteeing the efficiency and accuracy of white rice sorting. The introduction of this algorithm improves the system's intelligence level, enabling it to dynamically adapt to different sorting conditions and requirements.

[0114] The adsorption intensity of the negative pressure adsorption structure is dynamically adjusted based on the real-time detected impurity concentration. The impurity concentration is calculated by measuring the rate of change of transmitted light intensity using an optical sensor array. There is a non-linear increasing relationship between the adsorption gas flow rate and the increase in impurity concentration. As the impurity concentration increases, the adsorption intensity automatically increases to ensure efficient impurity removal.

[0115] Dynamically adjusting the adsorption intensity effectively removes impurities and improves the purity of white rice. An optical sensor array provides real-time feedback for adjusting the adsorption intensity, ensuring that the system can adjust the adsorption airflow in real time as the impurity concentration changes during the sorting process to achieve optimal sorting results. This mechanism improves the automation and accuracy of the sorting system.

[0116] Through close coordination and dynamic adjustments, the system achieves precise control over the rice sorting process, from airflow nozzle arrangement, velocity measurement, and airflow resistance correction to nozzle elevation angle adjustment and impurity removal. Each step is adjusted based on the characteristics of the rice grains and real-time feedback, ensuring high efficiency and accuracy in the sorting process. Simultaneously, the system can adapt to different working conditions in real time, demonstrating strong adaptability and intelligent advantages. Through the organic combination of these technologies, this method effectively improves the sorting quality and efficiency of white rice.

[0117] In one possible implementation, the system employs a first-wavelength light source and a second-wavelength light source arranged coaxially, with the wavelength difference between the two exceeding a preset spectral interval threshold. This design ensures that the reflection and scattering responses of the two laser sources differ significantly after penetrating the surface of the rice, which is beneficial for identifying surface structural details and deformation features. Coaxial illumination ensures consistency in imaging the rice grains using the two wavelengths, avoiding errors caused by differences in incident angles.

[0118] The dual-wavelength scheme improves the sensitivity to detect minute deformations on the surface of white rice, making it particularly suitable for determining whether rice grains are damaged, yellowed, or have mechanical cracks.

[0119] Scattered light is captured by a high dynamic range CMOS sensor, and its exposure time is dynamically adjusted according to the shading rate of the rice grains to ensure clear imaging regardless of the differences in the light transmittance of the rice grains. At the same time, the system performs multi-frame image fusion processing in each acquisition cycle to eliminate motion blur caused by the high-speed falling of the rice grains.

[0120] Dynamic exposure combined with multi-frame fusion significantly improves image clarity, enabling accurate capture of rice grain morphology information under different falling speeds and transparency levels, thus enhancing system robustness.

[0121] When calculating the light intensity ratio, the system will perform grid-based sampling of the projected area of ​​each rice grain, automatically eliminating abnormal light spots caused by edge diffraction. To ensure data quality, the area of ​​the effective sampling region must account for no less than a preset threshold (e.g., 85%) of the total projected area of ​​the rice grains, to ensure the reliability of the light intensity data used for judgment.

[0122] By filtering out interfering information by region, the light intensity ratio analysis becomes more accurate, avoiding misjudgment of the grain state due to edge effects and improving the accuracy of deformation recognition.

[0123] To determine whether rice grains are deformed, the system uses a sliding window algorithm to analyze the trend of light intensity ratio data of continuously passing rice grains. If the rate of change of light intensity ratio in a certain segment exceeds a preset fluctuation threshold, it is considered that deformation may exist. At this time, a secondary verification process is triggered, such as switching to higher resolution analysis or combining image morphological features for further judgment.

[0124] Dynamic trend analysis is more adaptable than static threshold judgment, can identify rice grains with slight deformation and irregular shape, and at the same time reduce the false alarm rate, making the entire sorting process more intelligent and accurate.

[0125] In the final sorting stage, the airflow nozzles employ multi-stage pressure control. The initial pressure is set based on the median mass distribution of all detected rice grains to ensure average sorting accuracy. Subsequently, the spray pressure is dynamically adjusted based on the confidence level of each rice grain's deformation assessment; the higher the confidence level, the greater the pressure, ensuring that suspicious deformed rice grains are effectively removed.

[0126] This confidence-linked nozzle pressure control strategy enables "force application on demand," which avoids excessive removal of intact rice grains while ensuring that high-confidence defective rice grains are forcefully separated, thereby improving the system's sorting efficiency and yield.

[0127] By combining dual-wavelength illumination with high-dynamic image acquisition, high-quality light intensity data is obtained. This data is then combined with regional screening and trend analysis algorithms to accurately identify rice grains with surface deformation. Finally, intelligently linked airflow nozzles are used for sorting, forming a data-closed-loop, adaptively adjusted intelligent sorting process. This connection method significantly improves the accuracy of identifying surface defects in white rice and the sorting efficiency, making it an important technological means for high-quality white rice production.

[0128] In one possible implementation, the phase window is divided using an adaptive time-slicing algorithm, which ensures that the duration of the phase window is proportional to the reciprocal of the current rice grain vibration frequency. Since different rice grains have different vibration frequencies, this proportional relationship allows the system to more accurately capture the dynamic characteristics of the rice grains, thereby improving sorting accuracy. Simultaneously, the minimum window duration is limited by the signal response delay of the control system, avoiding information loss or misjudgment caused by excessively short windows.

[0129] The introduction of the adaptive time slicing algorithm makes window division more flexible, enabling real-time response to the vibration state of different rice grains and providing more accurate dynamic characteristic analysis. This helps to better identify the physical characteristics of rice grains, such as surface deformation or size changes.

[0130] The kinetic energy calculation process is based on a discrete element method (DEM) simulation model to reconstruct the motion trajectory of the rice grains during vibration or under the influence of airflow. To accurately obtain the instantaneous velocity of each rice grain, Kalman filtering is used to denoise the motion data, removing errors caused by measurement noise or environmental factors. Furthermore, data from all rice grains that collide with the screen are excluded, ensuring that only the impact of rice grains not in contact with the screen is considered on the sorting process.

[0131] By using Kalman filtering for noise reduction, the system can accurately capture the true trajectory of rice grains in environments with significant noise interference, improving the accuracy of kinetic energy calculation and thus enhancing the stability and reliability of sorting.

[0132] In the potential energy calculation, the system integrates the force exerted by the airflow pressure field on the rice grain group. The integration region is dynamically subdivided into meshes based on the real-time rice grain distribution density. Higher grain density results in finer mesh subdivision to more accurately simulate the effect of airflow on the rice grains. This potential energy calculation reflects the interaction between the airflow and the rice grains, influencing the grain movement trajectory and sorting efficiency.

[0133] Dynamic mesh subdivision ensures that the force of the airflow pressure field on the rice grain group is accurately simulated, thereby optimizing the control of the airflow nozzle and ensuring more efficient interaction between the airflow and rice grains during the sorting process.

[0134] When the ratio of kinetic energy to potential energy exceeds a critical value, the system initiates a multi-parameter coordinated adjustment strategy. This strategy first calculates the adjustment amount of vibration acceleration using a piecewise function based on the extent to which the kinetic energy to potential energy ratio exceeds the critical value. The adjustment amount of airflow pressure is then negatively correlated with the adjustment amount of vibration acceleration; that is, when vibration acceleration increases, airflow pressure decreases accordingly to avoid over-adjustment of the system leading to uneven sorting.

[0135] Through a coordinated adjustment strategy, the vibration system and the airflow system can work together under different dynamic environments, avoiding over-adjustment of a single system. The negative feedback mechanism ensures the smooth operation of the sorting process, improves sorting accuracy, and avoids over-operation.

[0136] All parameter adjustment commands must pass through the timing verification module, which ensures that the control signals of the vibration system and the airflow system are executed synchronously within a preset phase difference range. Through precise timing control, the system ensures that the effects of vibration and airflow complement each other in each adjustment cycle, thereby achieving optimal sorting results.

[0137] Timing verification ensures the synchronization of the vibration system and the airflow system, avoiding interference caused by the time difference between their operations, and ensuring the stability and sorting accuracy of the system.

[0138] Intelligent control of the dynamic sorting process of white rice is achieved through adaptive time slicing, precise analysis of kinetic and potential energy calculations, the introduction of a collaborative adjustment strategy, and the synchronous execution of a timing verification module. The entire sorting process is optimized through precise correlation of phase, energy, and control signals between each sub-step. In this way, the system can not only effectively identify and remove unqualified rice grains, but also ensure the accurate sorting of good rice grains, improving sorting efficiency, quality, and system robustness.

[0139] In one possible implementation, the screen mesh aperture throughput rate, as a key parameter for evaluating the efficiency of rice grains passing through the screen, is used. Actual flow rate data is collected using a mass flow meter and cross-validated with the actual number of rice grains passing through as identified by an image recognition system. This cross-validation mechanism effectively eliminates abnormal counts caused by airflow disturbances, improving the accuracy and reliability of the data.

[0140] To avoid control deviations caused by the failure or reading error of a single sensor, improve the system's ability to respond to actual physical conditions, and provide a high-quality data foundation for subsequent judgments.

[0141] Using an improved Shannon entropy algorithm, the system discretizes the trajectory of rice grains during their fall into three-dimensional grid cells, calculates the proportion of time spent in each cell, and derives the probability density of that cell. The overall entropy value reflects the degree of order in the rice grain distribution and the perturbation state of the system.

[0142] Distribution entropy can serve as an important indicator reflecting the degree of spatial disorder in rice grain clusters, helping to determine whether the screening system deviates from the normal sorting path and aiding in the early identification of abnormal trends.

[0143] Key variables such as pore throughput, distribution entropy, and energy consumption are dimensionless to eliminate interference between different physical quantities. Then, principal component analysis (PCA) is used to reduce the dimensionality of the high-dimensional variable space, mapping it to a low-dimensional observable space for subsequent stability analysis and control judgment.

[0144] Dimensionlessness and PCA dimensionality reduction improve computational efficiency and model generalization ability, simplify the subsequent decision model structure, and at the same time maintain the key dynamic characteristics of the system and improve control response efficiency.

[0145] The system performs Lyapunov exponent calculation and phase space reconstruction on the state space trajectory. When the Lyapunov exponent exceeds the preset divergence threshold, it means that the system is in an unstable state or tends to be chaotic. At this time, a multi-objective optimization algorithm is immediately started to synchronously adjust the control parameters, taking into account objectives such as efficiency, stability and energy consumption.

[0146] The Lyapunov index is highly sensitive and can detect subtle instability trends in the system in advance, thereby enabling early warning and control to ensure that the system remains in a stable operating range for a long time.

[0147] During the adjustment process, the system continuously monitors the condition number of the Jacobian matrix of the control equations, which reflects the system's response sensitivity and numerical stability. When the condition number exceeds a preset stability threshold, it indicates that the system may enter an uncontrollable state due to parameter adjustments. In this case, the system will freeze the current adjustment action and automatically switch to a backup control strategy to ensure operational safety.

[0148] Introducing the Jacobian matrix condition number as a real-time criterion can effectively avoid over-regulation or instability of control signals, while the instant switching of backup strategies ensures that the system has good fault self-healing ability and robustness.

[0149] By integrating and analyzing multi-source data (such as screen throughput, rice grain distribution, and energy consumption), an observable state space is constructed and real-time stability assessment is implemented, thereby achieving multi-objective collaborative optimization and control. Through a mechanism combining the Lyapunov criterion and conditional number monitoring, not only is the system's adaptability to dynamic disturbances improved, but a highly intelligent stability assurance strategy is also provided for the sorting system, effectively enhancing the efficiency, consistency, and long-term stability of white rice sorting.

[0150] In one possible implementation, the system first quantifies the relationship between energy loss and amplitude of the vibration system under different vibration states by establishing a transfer function model between the damped collision energy ratio and the amplitude compensation amount. The nonlinear coefficient of this transfer function needs to be calibrated through impact tests to ensure that the model accurately reflects the energy loss characteristics in actual operation.

[0151] The transfer function model provides a precise basis for amplitude compensation calculation, which allows the compensation amount to be quantitatively adjusted according to the actual energy loss, thereby achieving more precise amplitude control in dynamic environments and avoiding over-compensation or undercompensation.

[0152] In the amplitude compensation process, a feedforward-feedback composite control structure is adopted. The feedforward control, based on a pre-established transfer function model, calculates the theoretical amplitude compensation amount and adjusts it in advance. The feedback control, on the other hand, uses PID control based on the residual of the damped collision energy ratio after compensation to correct the compensation amount in real time. PID control can automatically adjust the system's response speed and accuracy, further optimizing the amplitude adjustment effect.

[0153] The composite control structure combines the predictive capabilities of feedforward with the real-time adjustment capabilities of feedback, enabling the system to make rapid and effective adjustments in the face of vibration changes and disturbances, ensuring that the screen amplitude is always maintained within the optimal operating range.

[0154] Lateral tension is a key factor affecting screen stability. The system achieves real-time monitoring by embedding fiber optic strain sensors at the screen edge. This sensor can acquire lateral tension change data of the screen at a high frequency, and its sampling frequency is set to a multiple higher than the screen vibration fundamental frequency to ensure accurate capture of minute changes during vibration.

[0155] When the system detects that the lateral tension is approaching the safety threshold, it will automatically insert an amplitude adjustment cooling cycle to prevent the screen from breaking due to excessive tension or affecting normal operation. During this period, the system will gradually reduce the vibration energy input and the vibration amplitude until the lateral tension returns to the safe range. At this point, the cooling cycle ends, and the system will resume normal vibration.

[0156] This screen amplitude compensation mechanism, through precise model calculations, a composite control strategy, and real-time tension monitoring, ensures that the screen maintains a stable and efficient vibration state under different operating conditions. By using a transfer function model calibrated with nonlinear coefficients, combining feedforward and feedback control, and introducing precise monitoring of lateral tension and a cooling cycle, it not only improves the system's stability and safety but also effectively extends the equipment's service life and reduces maintenance costs, thereby ensuring the efficiency, accuracy, and safety of the white rice sorting process.

[0157] In one possible implementation, the algorithm's initial step is to initialize the nozzle elevation angle, setting it as a preset percentage of the theoretically optimal angle. This means that in the initial stage, the nozzle elevation angle is close to the theoretical value, but has not yet fully reached the optimal state. This step also includes setting a maximum number of iterations and a convergence accuracy threshold. The maximum number of iterations ensures that the algorithm does not get stuck in an infinite loop, while the convergence accuracy threshold defines the criteria by which the algorithm determines whether it has reached the optimal solution.

[0158] By setting the nozzle angle close to the theoretical optimal value, the number of iterations can be reduced, thereby improving computational efficiency and accelerating the system's convergence process.

[0159] In each iteration, the system searches for a local optimum using the Particle Swarm Optimization (PSO) algorithm to optimize the nozzle elevation angle. PSO simulates the search behavior of a swarm of particles in nature, thus quickly finding a better solution. To further improve the algorithm's accuracy, a nozzle mechanical adjustment inertia compensation factor is introduced into the particle velocity update formula. This factor takes into account the inertial effect during nozzle adjustment, optimizing the adjustment action to make the particle search process more stable and avoid over-adjustment or excessively slow adjustment.

[0160] By introducing an inertia compensation factor, the algorithm can better simulate the physical characteristics of nozzle adjustment, avoid errors and slow response caused by mechanical inertia, and improve the accuracy of the algorithm and the response speed of the system.

[0161] During the iteration process, the algorithm continuously calculates the change in the objective function and uses convergence criteria to determine whether the convergence standard has been met. Specifically, the convergence standard is: the change in the objective function over multiple consecutive iterations is less than the convergence accuracy threshold, and the uniformity index of the airflow velocity field meets the minimum requirement in the grading standard. This means that if the objective function changes steadily and the airflow uniformity meets the requirements during the iteration process, the algorithm is considered to have converged, and the iteration can be terminated.

[0162] The convergence criterion ensures that the algorithm terminates at an appropriate time, avoiding unnecessary excessive iterations. It not only optimizes computational resources but also ensures algorithm accuracy, preventing unnecessary error accumulation.

[0163] If the algorithm fails to converge within the preset maximum number of iterations, the system will automatically switch to an experience-based mode based on historical data and trigger a system calibration alarm. The experience-based mode provides an approximately optimized nozzle adjustment scheme based on past operational data and historical experience. The system calibration alarm alerts the operator that the system has not fully converged and provides an opportunity for manual intervention.

[0164] When the algorithm fails to converge within the maximum number of iterations, the system automatically switches to empirical mode, preventing the system from stagnating or failing. This mechanism ensures that the system continues to run even when the algorithm fails to converge, and alerts operators to make appropriate manual adjustments or maintenance, thereby improving the system's robustness and reliability.

[0165] This iterative approximation algorithm, by combining particle swarm optimization and nozzle mechanical adjustment inertia compensation factors, can efficiently and accurately optimize the nozzle elevation angle, improve airflow uniformity, and thus enhance the effectiveness of the dynamic rice sorting process. Reasonable initialization settings, inertia compensation during iteration, precise convergence criteria, and an empirical mode switching mechanism for extreme cases ensure that the system can adaptively optimize under different conditions and achieve efficient and stable operation. Through these steps, the algorithm improves sorting accuracy while also enhancing system stability and operational efficiency.

[0166] In one possible implementation, for rice grains with an excessive rate of change in light intensity ratio, the system initiates a multispectral scan. This means that the original single-wavelength light source is replaced or supplemented with at least two auxiliary wavelengths. A multi-wavelength light source can more accurately capture the surface features of the rice grains, especially deformations that respond differently to reflection or absorption characteristics at different wavelengths. In this way, the system can perform a more detailed scan of the rice grain surface, thereby reconstructing a three-dimensional deformation model of the rice grain. This model accurately reflects the deformation of the rice grain surface at different wavelengths, providing fundamental data for subsequent deformation analysis.

[0167] By reconstructing the 3D deformation model, the system extracts the deformation feature vectors of the rice grains and inputs these feature vectors into a pre-trained neural network classifier for analysis. Based on historical data and model training, the neural network can classify the deformation types of the rice grains and output a deformation type confidence score for each grain. This score reflects the neural network's confidence in whether the rice grain conforms to a certain deformation pattern.

[0168] When the confidence score for the deformation type is lower than the judgment threshold, the system considers the deformation type of the rice grain to be uncertain, therefore it is marked as an object to be reviewed and temporarily stored in the buffer isolation area. This process ensures that uncertain rice grains do not directly enter the final sorting decision stage, avoiding possible incorrect sorting.

[0169] Within the buffer zone, the contact deformation detection device uses a micro-force sensor array to perform more detailed physical measurements on the rice grains. The sensor array applies minute pressure and measures the deformation recovery curve of the rice grains after being compressed, i.e., the process by which the rice grains return to their original shape after being compressed. Through this process, the system can further verify the physical properties of the rice grains and determine whether they meet the sorting criteria.

[0170] Ultimately, the system integrates the optical judgment results with the contact detection results, and makes a final sorting decision based on the combined assessment. This decision integrates the deformation information obtained from optical scanning and the physical feedback from contact deformation detection, ensuring the accuracy and consistency of the sorting process.

[0171] The secondary verification process employs a series of refined steps, including multispectral scanning, neural network classification, and micro-force sensor detection, to achieve comprehensive detection and verification of rice grain deformation. By temporarily storing and re-verifying low-confidence rice grains, combined with high-precision physical detection methods, the system significantly improves sorting accuracy, ensuring high-quality output in the rice grain sorting process. This process not only enhances the robustness of the sorting system but also improves its ability to handle complex deformations, making the method more effective in dynamic sorting.

[0172] Accordingly, embodiments of the present invention also provide a dynamic sorting system based on multi-stage screening and airflow compensation, used to run the dynamic sorting method based on multi-stage screening and airflow compensation described in the embodiments of the present invention, including:

[0173] The rotary screening pretreatment module includes:

[0174] The spiral guide unit is equipped with a guide plate with a variable spiral angle, and its surface coating friction coefficient is less than the static friction coefficient between rice and metal.

[0175] The screen posture adjustment unit includes a servo motor-driven rotating screen frame, which is rigidly connected to the screen shaft via a coupling;

[0176] The motion trajectory acquisition unit consists of a high-speed industrial camera array. The camera optical axis is installed at an angle to the screen plane, and the output end is connected to the screen attitude adjustment unit.

[0177] The vibration spectrum analysis unit includes a piezoelectric sensor array and a signal conditioning circuit attached to the back of the screen, and its output is connected to the screen amplitude compensation controller.

[0178] The mass-airflow coupled sorting module includes:

[0179] The velocity field measurement unit consists of a laser Doppler velocimeter array, and the measurement area covers the entire cross-section of the airflow separation cavity.

[0180] The multi-layer airflow control unit includes a coaxially nested multi-layer annular nozzle support, with multiple adjustable nozzles evenly distributed circumferentially in each support layer, and the nozzle elevation angle is driven by a stepper motor.

[0181] The negative pressure adsorption unit consists of a centrifugal fan, a cavity pressure sensor, and a solenoid valve array. An optical impurity detection window is provided at the cavity inlet.

[0182] Deformation optical sorting module, including:

[0183] The dual-wavelength light source unit includes a first laser emitter and a second laser emitter mounted coaxially, with the difference between the two wavelengths being greater than a preset spectral interval;

[0184] The scattered light acquisition unit consists of a high frame rate CMOS sensor and an optical beam splitter, with the sensor output connected to the image processing board.

[0185] The pneumatic sorting unit includes a high-pressure air source, a proportional valve, and an array of nozzles, with the nozzle outlet direction orthogonal to the trajectory of the falling rice grains.

[0186] The energy coordination control module includes:

[0187] The vibration phase control unit includes an encoder and a vibration exciter coupled to the screen drive shaft, and the encoder output signal is connected to the phase splitting processor;

[0188] The airflow field adjustment unit consists of a pressure transmitter, a proportional-integral control valve, and a flow meter. Its control signal is synchronized with the output of the vibration phase control unit.

[0189] The dynamic optimization decision-making module includes:

[0190] A multi-source data acquisition unit connects to a screen porosity sensor, a distribution entropy calculation chip, and a power metering module.

[0191] The central processing unit has a built-in state space reconstruction algorithm and Lyapunov exponent calculation core, and its output is connected to each actuator via a CAN bus.

[0192] The connections between modules are as follows:

[0193] The screen outlet of the rotary screening pretreatment module is connected to the inlet of the mass-air-flow coupling sorting module through an airtight pipe, and the inner wall of the pipe is provided with an electrostatic elimination coating.

[0194] The light source unit and the scattered light acquisition unit of the deformation optical sorting module are synchronized through an optical fiber bundle, and the output end of the image processing board is electrically connected to the control terminal of the pneumatic sorting execution unit.

[0195] The output of the vibration phase control unit of the energy collaborative control module is interconnected with the servo motor driver signal of the rotary screening pretreatment module, and the pressure feedback signal of the airflow field adjustment unit is connected to the nozzle controller of the mass-airflow coupling sorting module.

[0196] The multi-source data acquisition unit of the dynamic optimization decision module receives real-time data from each sensor via industrial Ethernet, and the control commands generated by the central processing unit are simultaneously sent to the vibration phase control unit, the airflow field adjustment unit, and the pneumatic sorting execution unit.

[0197] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0198] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic sorting method based on multi-stage screening and airflow compensation, characterized in that, Includes the following steps: Step 1: Spatial phase sieving pretreatment: In the rotary multi-directional sieving channel, the white rice is made to rotate through the flow guiding structure, and the angle between the long axis of the rice grain and the long side of the sieve hole is obtained in real time. The rotational angular velocity of the sieve is dynamically adjusted according to the angle. Step 2: Mass-Airflow Coupled Separation: Construct a gas-solid two-phase flow mass separation field at the screening outlet, measure the density and initial velocity of the rice grains, calculate the airflow resistance, and configure the elevation angle and airflow velocity of the multi-layer airflow nozzles. Set up a negative pressure adsorption structure at the bottom of the separation zone. Step 3: Deformation feedback optical sorting: During the free fall stage of rice grains, a dual-wavelength laser irradiation system is used to capture the intensity distribution of scattered light. Based on the light intensity ratio, the deformed rice grains are determined and sorted through airflow nozzles. Step 4: Coordinated control of energy fields: Divide the phase window of the vibration period and calculate the ratio of kinetic energy to potential energy, and adjust the vibration acceleration and airflow pressure synchronously based on the ratio; Step 5: Dynamic convergence optimization: Collect the screen mesh throughput, rice grain distribution entropy, and total system power consumption to construct the state space and use a stability judgment algorithm to adjust the vibration frequency, airflow pressure, and light intensity parameters; Step 1 includes: The flow guiding structure is a spiral flow guiding plate, whose spiral angle is dynamically adjusted according to the average length-to-diameter ratio of rice, and the surface of the flow guiding plate is covered with a low friction coefficient material. The image acquisition device continuously captures the movement trajectory of rice grains, extracts the main direction of the rice grain outline, and calculates the real-time angle between the rice grain outline and the long side of the sieve hole. The frame rate of the image acquisition device is positively correlated with the rice grain flow rate. When the real-time included angle exceeds the preset angle threshold, the rotational angular velocity of the screen is increased to above the critical angular velocity, which is calculated based on the screen diameter and the average mass of the rice grains. A vibration detection array is set on the back of the screen to collect the vibration signal spectrum generated by the collision of rice grains, identify the energy ratio of the rigid collision characteristic frequency band and the damped collision characteristic frequency band, and trigger the screen amplitude compensation mechanism when the damped collision energy ratio exceeds the preset ratio threshold. The amplitude compensation mechanism includes: nonlinearly adjusting the longitudinal vibration amplitude of the screen according to the mapping relationship between the damping collision energy ratio and the screen porosity, and maintaining the transverse tension of the screen within a safe threshold range during the adjustment process.

2. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 1, characterized in that, Step 2 includes: The multi-layer airflow nozzles are arranged in a multi-layer annular array structure, with each layer containing multiple adjustable nozzles evenly distributed in the circumference, and the radial deflection angles of the nozzles in adjacent layers are staggered. When measuring the initial velocity of falling rice grains, a non-contact velocity sensing device is used, whose measurement area covers the entire cross-section of the separation field, and the measurement error caused by occlusion between rice grains is eliminated by a time series correlation algorithm. When calculating airflow resistance, the theoretical resistance value is corrected based on the standard deviation of the grain density distribution. The correction factor is exponentially related to the dispersion of the density distribution. When configuring the nozzle elevation angle, an iterative approximation algorithm is used to match the actual airflow velocity field with the theoretical resistance gradient field, and the adjustment amount in each iteration does not exceed the set proportion of the nozzle's maximum adjustment range; The adsorption intensity of the negative pressure adsorption structure is dynamically adjusted according to the real-time detected impurity concentration. The impurity concentration is calculated by measuring the rate of change of transmitted light intensity using an optical sensor array. The adsorption gas flow rate and the increase in impurity concentration have a non-linear increasing relationship.

3. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 2, characterized in that, Step 3 includes: The dual-wavelength laser irradiation system includes a first wavelength light source and a second wavelength light source arranged coaxially, and the difference between the two wavelengths is greater than a preset spectral interval threshold. The scattered light intensity distribution is collected using a high dynamic range CMOS sensor. Its exposure time is dynamically adjusted according to the shading rate of the falling rice grains, and multi-frame image fusion is performed in each collection cycle to eliminate motion blur. When calculating the light intensity ratio, each grain of rice projection area is sampled in sections to exclude noise data caused by edge diffraction effects. The ratio of the effective sampling area to the total projection area is not less than a set threshold. When determining the surface deformation of rice grains, a sliding window algorithm is used to analyze the trend of the light intensity ratio data of continuous rice grains. When the rate of change of light intensity ratio exceeds the preset fluctuation threshold, a secondary verification process is initiated. The airflow nozzle sorting includes multi-stage pressure control. The initial pressure is set according to the median of the rice grain mass distribution, and the subsequent pressure adjustment is positively correlated with the confidence level of deformation determination.

4. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 3, characterized in that, Step 4 includes: The phase window division adopts an adaptive time slicing algorithm. The window duration is proportional to the reciprocal of the current vibration frequency, and the minimum window duration is limited by the signal response delay of the control system. When calculating kinetic energy, the trajectory of the rice grain group is reconstructed based on the discrete element simulation model. The instantaneous velocity of each rice grain is processed by Kalman filtering to reduce noise, and the data of rice grains that collide with the screen are excluded. Potential energy calculation includes the integral of the force exerted by the airflow pressure field on the rice grain group, and the integration region is dynamically subdivided into grids based on the real-time rice grain distribution density. When the ratio of kinetic energy to potential energy exceeds the critical value, a multi-parameter coordinated adjustment strategy is activated: the vibration acceleration adjustment is calculated according to the extent of the ratio exceeding the critical value using a piecewise function, and the airflow pressure adjustment is negatively correlated with the vibration acceleration adjustment. All parameter adjustment commands must pass through the timing verification module to ensure that the control signals of the vibration system and the airflow system are executed synchronously within the preset phase difference range.

5. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 4, characterized in that, Step 5 includes: The screen mesh throughput is calculated by cross-validation using a mass flow meter and an image recognition system, eliminating abnormal data points caused by airflow disturbances. The entropy value of the rice grain distribution is calculated using an improved Shannon entropy algorithm, which discretizes the falling trajectory into three-dimensional grid cells. The probability density of each cell is determined by the proportion of rice grains that remain in that cell. When constructing the state space, the pore throughput, distribution entropy value and energy consumption power are dimensionless and reduced to the observable space by principal component analysis. The stability determination algorithm includes Lyapunov exponent calculation and phase space reconstruction. When the exponent exceeds the divergence threshold, a multi-objective optimization algorithm is activated to synchronously adjust the control parameters. During parameter adjustment, the condition number of the system Jacobian matrix is ​​monitored in real time. When the condition number exceeds the preset stability threshold, the current adjustment amount is frozen and the system is switched to the backup control strategy.

6. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 5, characterized in that, The screen amplitude compensation mechanism includes: A transfer function model of the damped collision energy ratio and amplitude compensation was established, and its nonlinear coefficients were calibrated through impact tests. When performing amplitude compensation, a feedforward-feedback composite control structure is adopted: the feedforward control quantity is calculated based on the transfer function model, and the feedback control quantity is PID adjusted according to the residual of the damped collision energy ratio after compensation. Lateral tension monitoring is achieved through fiber optic strain sensors embedded in the edge of the screen, with a sampling frequency higher than a set multiple of the screen vibration fundamental frequency; When the lateral tension approaches the safety threshold, an amplitude adjustment cooling cycle is automatically inserted, during which the vibration energy input is gradually reduced until the tension returns to the safe range.

7. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 2, characterized in that, The iterative approximation algorithm includes: Initialize the nozzle elevation angle to a preset percentage of the theoretically optimal angle, and set the maximum number of iterations and the convergence accuracy threshold; In each iteration, a local optimum is searched using the particle swarm optimization algorithm, and a nozzle mechanical adjustment inertia compensation factor is introduced into the particle velocity update formula. The convergence criteria are that the change in the objective function in multiple consecutive iterations is less than the convergence accuracy threshold, and the airflow velocity field uniformity index meets the minimum requirements in the grading standard. If the number of iterations reaches the upper limit and convergence is still not achieved, the system will automatically switch to an experience-based mode based on historical data and trigger a system calibration alarm.

8. The dynamic sorting method based on multi-stage screening and airflow compensation according to claim 3, characterized in that, The secondary verification process includes: For rice grains with excessive light intensity ratio change rate, initiate multispectral scanning, add at least two auxiliary wavelengths for irradiation, and reconstruct a three-dimensional model of surface deformation. Input the feature vector of the 3D model into a pre-trained neural network classifier and output a deformation type confidence score; When the confidence score is lower than the judgment threshold, the grain of rice is marked as an object to be reviewed and temporarily stored in the buffer isolation area; A contact deformation detection device is installed in the buffer isolation zone. The recovery curve of rice grains under pressure is measured by a micro-force sensor array. Finally, the sorting decision is generated by combining the optical judgment and the contact detection results.

9. A dynamic sorting system based on multi-stage screening and airflow compensation, used to run the dynamic sorting method based on multi-stage screening and airflow compensation as described in any one of claims 1-8, characterized in that, include: The rotary screening pretreatment module includes: The spiral guide unit is equipped with a guide plate with a variable spiral angle, and its surface coating friction coefficient is less than the static friction coefficient between rice and metal. The screen posture adjustment unit includes a servo motor-driven rotating screen frame, which is rigidly connected to the screen shaft via a coupling; The motion trajectory acquisition unit consists of a high-speed industrial camera array. The camera optical axis is installed at an angle to the screen plane, and the output end is connected to the screen attitude adjustment unit. The vibration spectrum analysis unit includes a piezoelectric sensor array and a signal conditioning circuit attached to the back of the screen, and its output is connected to the screen amplitude compensation controller. The mass-airflow coupled sorting module includes: The velocity field measurement unit consists of a laser Doppler velocimeter array, and the measurement area covers the entire cross-section of the airflow separation cavity. The multi-layer airflow control unit includes a coaxially nested multi-layer annular nozzle support, with multiple adjustable nozzles evenly distributed circumferentially in each support layer, and the nozzle elevation angle is driven by a stepper motor. The negative pressure adsorption unit consists of a centrifugal fan, a cavity pressure sensor, and a solenoid valve array. An optical impurity detection window is provided at the cavity inlet. Deformation optical sorting module, including: The dual-wavelength light source unit includes a first laser emitter and a second laser emitter mounted coaxially, with the difference between the two wavelengths being greater than a preset spectral interval; The scattered light acquisition unit consists of a high frame rate CMOS sensor and an optical beam splitter, with the sensor output connected to the image processing board. The pneumatic sorting unit includes a high-pressure air source, a proportional valve, and an array of nozzles, with the nozzle outlet direction orthogonal to the trajectory of the falling rice grains. The energy coordination control module includes: The vibration phase control unit includes an encoder and a vibration exciter coupled to the screen drive shaft, and the encoder output signal is connected to the phase splitting processor; The airflow field adjustment unit consists of a pressure transmitter, a proportional-integral control valve, and a flow meter. Its control signal is synchronized with the output of the vibration phase control unit. The dynamic optimization decision-making module includes: A multi-source data acquisition unit connects to a screen porosity sensor, a distribution entropy calculation chip, and a power metering module. The central processing unit has a built-in state space reconstruction algorithm and Lyapunov exponent calculation core, and its output is connected to each actuator via a CAN bus. The connections between modules are as follows: The outlet of the rotary screening pretreatment module is connected to the inlet of the mass-air-flow coupling sorting module through an airtight pipe, and the inner wall of the pipe is provided with an electrostatic elimination coating. The light source unit and the scattered light acquisition unit of the deformation optical sorting module are synchronized through an optical fiber bundle, and the output end of the image processing board is electrically connected to the control terminal of the pneumatic sorting execution unit. The output of the vibration phase control unit of the energy collaborative control module is interconnected with the servo motor driver signal of the rotary screening pretreatment module, and the pressure feedback signal of the airflow field adjustment unit is connected to the nozzle controller of the mass-airflow coupling sorting module. The multi-source data acquisition unit of the dynamic optimization decision module receives real-time data from each sensor via industrial Ethernet, and the control commands generated by the central processing unit are simultaneously sent to the vibration phase control unit, the airflow field adjustment unit, and the pneumatic sorting execution unit.

Images