Coal vibration airflow collaborative sorting method and system based on multi-target collaborative control

Abstract

The invention relates to the technical field of coal sorting, and discloses a coal vibration airflow collaborative sorting method and system based on multi-target collaborative control, and the system comprises a data acquisition unit, a data processing unit and an execution adjustment unit; wherein the data acquisition unit is used for synchronously acquiring multi-dimensional operation data; the data processing unit performs fusion operation on the multi-dimensional operation data to obtain estimated values of bed stratification degree and real-time cutting density, and performs rolling prediction based on a prediction model to obtain an optimization control instruction; the execution adjustment unit receives the optimization control instruction to adjust the apparent wind speed distribution of each subarea to stabilize bed layering, dynamically adjusts the vibration frequency and amplitude to maintain particle migration balance, corrects the longitudinal and transverse inclination angles to suppress backmixing, and adjusts the discharge boundary. The problems of separation boundary drifting and backmixing are effectively avoided, and the stability of clean coal ash content and cutting density is ensured.

Description

Technical Field

[0001] This invention relates to the field of coal sorting technology, and more specifically, to a coal vibration airflow collaborative sorting method and system based on multi-objective collaborative control. Background Technology

[0002] As coal resource development and utilization move towards refinement and cleaner processes, dry separation of fine-grained coal has gradually become a crucial step in coal processing. Vibration-airflow synergistic separation technology, through the combined action of a vibrational force field and an airflow force field, achieves particle stratification and separation based on density differences. It boasts advantages such as water conservation, environmental friendliness, low energy consumption, and strong adaptability, and has been widely applied in coal washing. Especially when processing fine coal ranging from 0 to 13 mm, the coupling effect of airflow fluidization and bed vibration effectively reduces particle adhesion and backmixing, improving clean coal recovery. However, due to differences in coal quality, feed fluctuations, and complex operating conditions, existing vibration-airflow separation equipment still has significant shortcomings in terms of separation boundary stability and coordination of multi-objective operating indicators.

[0003] For example, invention patent CN115350916A describes a vibratory cascade separator for fine coal powder. This separator uses eccentric excitation to generate linear reciprocating vibration on the separator bed, combined with bottom airflow to form a fluidized bed, achieving coal and gangue stratification and cascade separation. Structurally, this device employs a porous corrugated bed surface, adjustable tilt angle, and trapezoidal air distribution plates, which can improve material stratification to some extent. However, this solution still has some problems: it mainly relies on manually setting parameters such as vibration frequency, bed tilt angle, and air distribution aperture ratio, lacking online monitoring and closed-loop control of bed state, particle stratification degree, and real-time cutting density. This leads to easy drift in separation effect when the feed particle size fluctuates or the coal quality changes; although it has adjustable components, the various adjustment parameters are independent of each other, lacking coordinated optimization of multi-zone air supply, vibration frequency amplitude, bed attitude, and discharge boundary, making it difficult to simultaneously achieve stable cutting density, reduced mismatch rate, and energy consumption control; and it does not include environmental indicators such as dust emissions in the control targets, failing to meet the requirements of green operation while ensuring separation quality.

[0004] Therefore, it is necessary to design a coal vibration airflow collaborative separation method and system based on multi-objective collaborative control to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a coal vibration airflow collaborative separation method and system based on multi-objective collaborative control, which aims to solve the problems of current technology such as separation boundary drift, back mixing of light and heavy particles, increased energy consumption and excessive dust emissions.

[0006] In one aspect, the present invention proposes a coal vibration airflow coordinated separation system based on multi-objective cooperative control, comprising: The system comprises a data acquisition unit, a data processing unit, and an execution and adjustment unit; among which, The data acquisition unit includes a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor. The data acquisition unit is used to synchronously acquire multi-dimensional operating data, including bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The data processing unit includes a multi-objective collaborative controller, which performs fusion calculations on the multi-dimensional operating data to obtain estimated values ​​of the bed layering degree and real-time cutting density, and performs rolling predictions based on a prediction model to obtain optimized control commands that constrain energy consumption and dust emissions while minimizing real-time cutting density deviation and mismatch rate. The execution adjustment unit includes a multi-zone air supply unit, a servo eccentric vibration mechanism, a bed surface attitude adjustment mechanism, and a material distribution adjustment mechanism. The multi-zone air supply unit receives the optimized control command and adjusts the apparent wind speed distribution of each zone to stabilize the bed layer stratification. The servo eccentric vibration mechanism dynamically adjusts the vibration frequency and amplitude according to the optimized control command to maintain particle migration balance. The bed surface attitude adjustment mechanism corrects the longitudinal and transverse tilt angles according to the optimized control command to suppress back mixing. The material distribution adjustment mechanism adjusts the discharge boundary according to the optimized control command.

[0007] Furthermore, it also includes: The differential pressure sensor array is installed in the air distribution cavity at the bottom of the sorting host, and corresponds one-to-one with the feeding area, the upper section of the sorting area, the middle section of the sorting area, the lower section of the sorting area, and the gangue channel. Each corresponding area has at least two measuring points to form a differential pressure measurement channel. The airborne vision sensor is installed above the sorting area and at a downward angle of 15° to 35° relative to the bed surface, with the field of view covering the sorting area and the discharge port. The acceleration and acoustic sensors are fixed to the bed frame and the inner wall of the machine casing of the sorting area, and are no more than 50 mm away from the bed surface. The online ash sensors are installed in the clean coal discharge channel and the gangue discharge channel, respectively. The dust sensors are installed in the sampling branch before and after the dust collector in the tail gas pipeline. Each sensor acquisition channel is coded by regional index and bound to the one-to-one mapping relationship of the multi-zone air supply unit.

[0008] Furthermore, when the data acquisition unit is used to synchronously acquire multi-dimensional operational data, it includes: The data acquisition unit synchronizes and aligns the timestamps of all sensors to ensure that the time error between any two signals is no greater than 2 milliseconds. The bed pressure difference signal is sampled at a frequency of at least 200 Hz and, after zero-point drift correction and Kalman filtering, outputs the bed porosity. The airborne vision sensor acquires data at a frequency of at least 60 frames per second. After lens distortion correction, flat field correction, and region interest segmentation, the stratification index is calculated based on the optical density difference between upper and lower regions, texture gradient projection, and the ratio of the vertical migration vectors of the optical flow field. The stratification index is then used to... The particle stratification state is determined; acceleration and acoustic signals are sampled at a frequency of at least 1 kHz and denoised and envelope extracted to assist in judging stratification stability; online ash signals are sampled at a frequency of at least 1 Hz and, after temperature and baseline compensation, the real-time cutting density is calculated based on the ash-density calibration curve; the mismatch rate is calculated based on the joint criterion of the proportion of out-of-bounds particles and online ash deviation; dust emission is output after outlier removal and moving average processing of dust signals; all results are cached with a unified timestamp and regional index, constituting the multi-dimensional operating data.

[0009] Furthermore, when the multi-objective collaborative controller performs fusion calculations on the multi-dimensional operational data to obtain estimates of the bed layering degree and real-time cutting density, it includes: When the multi-objective collaborative controller performs fusion calculations on the multi-dimensional operational data, it uses a unified timestamp and regional index to form a feature vector from the bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. Within a 2-5 second sliding time window, it performs weighted fusion based on noise covariance adaptive estimation. A constrained Bayesian filter is applied to the stratification index and bed porosity to output the bed stratification degree. The fusion weight parameters satisfy a value between 0 and 1, and the sum of the weights is 1. The estimated real-time cutting density is obtained by using the observed values ​​obtained from online ash content and ash-density calibration curves through table lookup and interpolation as measurement items. This, combined with the joint correction of the proportion of out-of-bounds particles, outputs the estimated real-time cutting density, and provides a confidence interval as an estimation confidence index.

[0010] Furthermore, when the multi-objective collaborative controller performs rolling predictions and generates optimized control commands based on the prediction model, it performs prediction simulations on the apparent wind speed, vibration frequency, amplitude, longitudinal tilt angle, lateral tilt angle, and material distribution plate position within a control step size of 1-2 seconds and a prediction time domain of 10-30 seconds. A piecewise affine ARX model is used as the prediction method. Based on the prediction results, the controller aims to minimize the real-time cutting density deviation and mismatch rate. At the same time, it sets constraints on fan energy consumption and dust emissions, and imposes constraints on the numerical range and rate of change of the control variables. Specifically, the apparent wind speed of the zone is limited to 0.4-2.2 m / s, the vibration frequency is limited to 4-12 Hz, the amplitude is limited to 2-8 mm, the longitudinal tilt angle is limited to 0-4 degrees, the lateral tilt angle is limited to -5-5 degrees, and the material distribution plate position is limited to ±20 mm. The upper limit of the change range of the control variables is also set for gradual adjustment.

[0011] Furthermore, after receiving the optimized control command, the multi-zone air supply unit converts the target apparent wind speed of each zone into an air volume distribution command based on the zone index, and generates specific set values ​​according to the correspondence between fan speed and valve opening. The multi-zone air supply unit executes the set values ​​through cascaded control of variable frequency fans and electric valves, employing a cascade strategy of tracking control with apparent wind speed as the controlled variable in the outer loop and stabilizing control with static pressure in the air distribution chamber as the controlled variable in the inner loop. The outer loop outputs the air volume target, and the inner loop outputs the fan speed and valve opening. Both methods employ a gradual adjustment of the upper limit and an S-shaped ramp to suppress transient impacts. During execution, monotonicity constraints are set for each zone to ensure that the apparent wind speed in the sorting zone along the material flow direction does not increase, and the apparent wind speed in the gangue channel is not lower than the apparent wind speed threshold in the feeding zone. When the bed pressure difference fluctuation provided by the data acquisition unit exceeds the limit or the dust emission reaches the threshold, the multi-zone air supply unit triggers abnormal allocation logic, increases the apparent wind speed in the gangue channel and the feeding zone, and correspondingly reduces the apparent wind speed in the lower section of the sorting zone, while keeping the total air volume below the energy consumption constraint.

[0012] Furthermore, after receiving the optimized control command, the servo eccentric vibration mechanism determines the target vibration frequency and amplitude based on the particle migration balance index, which is composed of the stratification index change rate, particle stratification state, and acceleration envelope characteristics. It then uses dual closed-loop control of frequency and amplitude to achieve gradual adjustment. The outer loop uses the particle migration balance index as the controlled variable to give the target vibration parameters, while the inner loop uses the vibration frequency and amplitude as the controlled variables to track the target values ​​through a closed loop via an encoder and a displacement sensor. The vibration parameters are then gradually adjusted with an upper limit and an S-shaped ramp is used to limit transient impacts.

[0013] Furthermore, after receiving the optimized control command, the bed surface attitude adjustment mechanism performs linked corrections on the longitudinal and lateral tilt angles based on the gradient of the stratification index in the material flow direction, the mismatch rate, and the statistical results of the airborne vision of the transgressing particles: when the stratification index decreases along the material flow direction and the mismatch rate increases, the longitudinal tilt angle is increased to prolong the retention and suppress backmixing; when transgressing particles are concentrated in the lateral separation zone, the lateral tilt angle is adjusted to enhance the lateral separation of light and heavy particles; the adjustment of the longitudinal and lateral tilt angles is constrained by the numerical range and the upper limit of the stepwise adjustment, and a dual threshold hysteresis is set to avoid frequent adjustments, and each attitude adjustment triggers a slight compensation for the apparent wind speed target of the adjacent zone to maintain the stability of the bed porosity.

[0014] Furthermore, after receiving the optimized control command, the material distribution adjustment mechanism uses the deviation between the estimated real-time cutting density and the target cutting density, as well as the online ash content comparison of the clean coal and gangue discharge channels, as criteria to perform fine-step adjustments to the position of the material distribution plate: when the real-time cutting density is higher than the target and the online ash content of the clean coal is too high, the material distribution plate is moved towards the gangue side by a preset step distance; when the real-time cutting density is lower than the target and the online ash content of the gangue is too high, the material distribution plate is moved towards the clean coal side by a preset step distance. The fine-step adjustment is constrained by the position range of the material distribution plate and the upper limit of the step adjustment, and a dead zone and a minimum holding time are set to prevent high-frequency oscillation. At the same time, after each position update, a new round of multi-dimensional operating data is read to confirm the adjustment effect and decide whether to continue fine-tuning, thereby stabilizing the discharge boundary.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: by setting up a data acquisition unit, a data processing unit, and an execution adjustment unit, a complete closed-loop control link is constructed. The data acquisition unit uses a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor to synchronously acquire multi-dimensional operating data such as bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The data processing unit performs fusion calculations on the multi-dimensional operating data through a multi-objective collaborative controller and performs rolling predictions in combination with a prediction model. It can output control commands in real time that ensure the optimization of cutting density and mismatch rate while taking into account energy consumption and dust constraints. The execution adjustment unit decomposes the optimized control commands into coordinated adjustment actions of the multi-zone air supply unit, the servo eccentric vibration mechanism, the bed surface attitude adjustment mechanism, and the material distribution adjustment mechanism, thereby realizing dynamic coupling control of airflow distribution, vibration parameters, bed surface attitude, and discharge boundary. Compared with existing solutions, this application effectively avoids the problems of sorting boundary drift and back mixing under coal quality changes, feed fluctuations and complex working conditions, ensures the stability of clean coal ash content and cutting density, and reduces energy consumption and incorporates dust emissions into the control target, demonstrating beneficial effects in sorting accuracy, system stability and green operation.

[0016] On the other hand, this application also provides a coal vibration airflow coordinated separation method based on multi-objective coordinated control, applied to the aforementioned coal vibration airflow coordinated separation system based on multi-objective coordinated control, including: Simultaneously collect multi-dimensional operational data, including bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate and dust emission; The multidimensional operating data is fused and calculated to obtain the estimated values ​​of the bed layering degree and real-time cutting density. Based on the prediction model, rolling prediction is performed to obtain the optimized control command that minimizes the real-time cutting density deviation and mismatch rate while constraining energy consumption and dust emissions. The system receives the optimized control command and adjusts the apparent wind speed distribution of each zone to stabilize the bed stratification; dynamically adjusts the vibration frequency and amplitude according to the optimized control command to maintain particle migration balance; corrects the longitudinal and transverse tilt angles according to the optimized control command to suppress backmixing; and adjusts the discharge boundary according to the optimized control command.

[0017] It is understandable that the coal vibration airflow collaborative separation method and system based on multi-objective collaborative control described above have the same beneficial effects, and will not be elaborated here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a structural block diagram of a coal vibration airflow coordinated sorting system based on multi-objective coordinated control provided in an embodiment of the present invention; Figure 2 A flowchart of a coal vibration airflow collaborative sorting method based on multi-objective collaborative control provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] In traditional vibratory airflow separation systems, the separation process relies on manually preset vibration parameters, air distribution parameters, and mechanical structure parameters, lacking real-time perception of the dynamic characteristics of the bed and the ability to optimize multiple objectives collaboratively. Due to fluctuations in feed particle size distribution or changes in coal quality, key parameters such as bed porosity, particle stratification, and cutting density are prone to deviation, leading to decreased separation accuracy. Furthermore, the lack of dynamic correlation between the apparent wind speed distribution in the separation zone, vibration energy input, and bed attitude adjustment makes it impossible to suppress backmixing and dust diffusion while maintaining stable stratification, resulting in increased energy consumption and deterioration of environmental indicators.

[0021] For example, in the fine coal sorting scenario, the sorting host is initially set with a fixed vibration frequency and amplitude, and the air volume is evenly distributed across each zone of the air distribution chamber, while the bed inclination angle remains static. When the gangue content in the feed suddenly increases, the change in bed density distribution leads to an imbalance in particle migration velocity. Lighter particles in the upper layer cannot be effectively separated, while heavier particles in the lower layer accumulate, causing bed compaction. At this time, the differential pressure sensor data and the stratification index captured by the visual sensor exhibit asynchronous fluctuations, but the existing system cannot integrate multi-source data in real time and generate dynamic control commands. The apparent wind speed in the sorting zone is not adjusted according to the degree of bed compaction, resulting in particle agglomeration in areas of insufficient fluidization and increased dust entrainment in areas of excessive fluidization. The vibration parameters are not corrected according to the particle migration state; if the amplitude is too small, it cannot overcome the friction between particles, and if the amplitude is too large, it causes secondary back mixing. The bed inclination angle is not adjusted in conjunction with the process, resulting in an imbalance between material residence time and lateral separation efficiency, and gangue mixing in the clean coal channel.

[0022] If the above problems are not addressed, the sorting system will operate in an open-loop state for an extended period. The cutting density will fluctuate with operating conditions, deviating from the set threshold, leading to a continuous increase in mismatch rate and resulting in excessive ash content in the clean coal or coal loss along with the gangue. Uneven fluidization of the sorting bed and mismatched vibration energy will exacerbate particle collision and friction. Abnormal fluctuations in static pressure in the air distribution chamber will cause sudden changes in fan load, increasing energy consumption. Uncontrolled dust emissions will not only reduce the lifespan of dust removal equipment but may also trigger environmental monitoring alarms, forcing the system to operate at reduced frequency or even shut down for maintenance, severely impacting continuous production stability and economic efficiency.

[0023] For this, please refer to Figure 1This application proposes a coal vibration airflow collaborative sorting system based on multi-objective collaborative control, comprising: a data acquisition unit, a data processing unit, and an execution and regulation unit. The data acquisition unit includes a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor. The data acquisition unit is used to simultaneously acquire multi-dimensional operational data, including bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The data processing unit includes a multi-objective collaborative controller. The multi-objective collaborative controller performs fusion calculations on the multi-dimensional operational data to obtain estimates of the bed stratification degree and real-time cutting density, and performs rolling predictions based on a prediction model to obtain optimized control commands that minimize real-time cutting density deviation and mismatch rate while constraining energy consumption and dust emission. The execution adjustment unit includes a multi-zone air supply unit, a servo eccentric vibration mechanism, a bed surface attitude adjustment mechanism, and a material distribution adjustment mechanism. The multi-zone air supply unit receives optimization control commands and adjusts the apparent wind speed distribution of each zone to stabilize the bed layer stratification. The servo eccentric vibration mechanism dynamically adjusts the vibration frequency and amplitude according to the optimization control commands to maintain particle migration balance. The bed surface attitude adjustment mechanism corrects the longitudinal and transverse tilt angles according to the optimization control commands to suppress back mixing. The material distribution adjustment mechanism adjusts the discharge boundary according to the optimization control commands.

[0024] The data acquisition unit includes a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor. This unit synchronously collects multi-dimensional operational data, including bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emissions. This refers to the real-time acquisition of different physical parameters during the sorting process using various sensors. The bed pressure differential sensor array calculates bed porosity by measuring changes in the pressure difference of the air distribution chamber. The airborne vision sensor calculates the stratification index by analyzing particle distribution through image analysis. The acceleration and acoustic sensor uses vibration signals to assist in judging stratification stability. The online ash sensor converts the ash content to density to calculate the real-time cutting density. The dust sensor monitors emission concentration. These data collectively reflect the dynamic characteristics of the sorting process, providing a basis for subsequent control. The data processing unit includes a multi-objective collaborative controller. This controller performs fusion calculations on multi-dimensional operational data to obtain estimates of the bed stratification degree and real-time cutting density. Based on a prediction model, it performs rolling predictions to obtain optimized control commands that minimize real-time cutting density deviation and mismatch rate while constraining energy consumption and dust emissions. This involves processing multi-source heterogeneous data through time synchronization and weighted fusion, combining it with a prediction model to generate a control strategy that balances sorting accuracy and operational efficiency. By dynamically adjusting control parameters through rolling optimization, it resolves sorting boundary drift and multi-objective conflict issues. The execution adjustment unit includes a multi-zone air supply unit, a servo eccentric vibration mechanism, a bed surface attitude adjustment mechanism, and a material distribution adjustment mechanism. The multi-zone air supply unit receives optimization control commands and adjusts the apparent wind speed distribution of each zone to stabilize the bed layer stratification. The servo eccentric vibration mechanism dynamically adjusts the vibration frequency and amplitude according to the optimization control commands to maintain particle migration balance. The bed surface attitude adjustment mechanism corrects the longitudinal and transverse tilt angles according to the optimization control commands to suppress back mixing. The material distribution adjustment mechanism adjusts the discharge boundary according to the optimization control commands, which means adjusting the airflow distribution through zoned air supply, optimizing the particle migration path through vibration parameter adjustment, suppressing back mixing through bed surface tilt angle correction, and stabilizing the discharge boundary through fine adjustment of the material distribution plate position, thereby realizing multi-parameter collaborative control of the sorting process.

[0025] This application generates collaborative control commands through multi-dimensional data fusion and multi-objective rolling optimization, dynamically adjusting airflow distribution, vibration parameters, bed surface attitude and material separation boundary. While ensuring sorting accuracy, it reduces energy consumption and dust emissions, solving the problems of poor stability and multi-objective conflict caused by independent parameter adjustment in the prior art.

[0026] The working process and principle of this application are as follows: The data acquisition unit synchronously collects multi-dimensional operating data through a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor. The multi-dimensional operating data includes bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The multi-objective collaborative controller in the data processing unit performs fusion calculations on the collected multi-dimensional operating data to obtain estimates of the bed stratification degree and real-time cutting density. The multi-objective collaborative controller performs rolling predictions based on a predictive model to generate optimized control commands to minimize real-time cutting density deviation and mismatch rate, while simultaneously constraining energy consumption and dust emission. The execution and adjustment unit receives the optimized control commands and executes corresponding adjustments. The multi-zone air supply unit adjusts the apparent wind speed distribution of each zone to stabilize bed stratification. The servo eccentric vibration mechanism dynamically adjusts the vibration frequency and amplitude to maintain particle migration balance. The bed surface attitude adjustment mechanism corrects the longitudinal and transverse tilt angles to suppress back mixing. The material distribution adjustment mechanism adjusts the discharge boundary. This multi-objective collaborative control method enables precise control and optimization of the coal vibration airflow separation process.

[0027] As a preferred embodiment, the solution of this application is implemented as follows: In the data acquisition unit, the bed pressure differential sensor array is arranged in the air distribution cavity at the bottom of the sorting host, corresponding to the feeding area, the upper, middle and lower sections of the sorting area, and the gangue channel. An onboard vision sensor is installed above the sorting area, covering the sorting area and the discharge port from a certain downward angle. Acceleration and acoustic sensors are fixed to the bed surface frame of the sorting area and the inner wall of the machine casing. Online ash sensors are installed in the clean coal and gangue discharge channels respectively. Dust sensors are installed in the sampling branches before and after the exhaust gas pipeline dust collector. Each sensor acquisition channel is coded according to the regional index and a mapping relationship is established with the multi-zone air supply unit. The data acquisition unit synchronizes the time of each sensor, acquiring multi-dimensional operating data such as bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The multi-objective collaborative controller in the data processing unit performs fusion calculations on the multi-dimensional operating data to obtain estimates of the bed stratification degree and real-time cutting density. The multi-objective collaborative controller performs rolling predictions based on a predictive model to generate optimized control commands. In the execution adjustment unit, the multi-zone air supply unit receives the optimized control commands and adjusts the apparent air velocity distribution of each zone via variable frequency fans and electric valves. The servo eccentric vibration mechanism adjusts the vibration frequency and amplitude according to the optimized control commands. The bed surface attitude adjustment mechanism corrects the longitudinal and transverse tilt angles according to the optimized control commands. The material distribution adjustment mechanism adjusts the discharge boundary position according to the optimized control commands.

[0028] Through the above scheme, this application achieves multi-objective coordinated control of the coal vibration airflow separation process. The data acquisition unit synchronously collects multi-dimensional operational data, providing comprehensive status information for the separation process. The multi-objective coordinated controller generates optimized control commands that take into account multiple indicators through fusion calculations and rolling predictions. The execution and regulation unit coordinates the air supply, vibration, bed surface attitude, and material separation boundary according to the optimized control commands. This closed-loop control method can effectively cope with feed fluctuations and coal quality changes, maintain bed stratification stability, reduce backmixing, and improve separation accuracy. Simultaneously, by constraining energy consumption and dust emissions, a balance between separation quality and environmental benefits is achieved. This scheme overcomes the limitations of traditional vibration airflow separation systems with fixed preset parameters and independent adjustment methods, improving the adaptability and stability of the separation process.

[0029] In some of the solutions described above in this application, the data acquisition unit acquires multidimensional operational data through various sensors. However, in practical applications, due to unreasonable sensor placement or lack of regional correspondence in the acquisition channels, data acquisition suffers from problems such as time asynchrony, incomplete spatial coverage, and inaccurate control of the actuator, which in turn affects the stability and control accuracy of the sorting process.

[0030] This application further proposes that a bed pressure differential sensor array be installed in the air distribution cavity at the bottom of the sorting host, corresponding one-to-one with the feeding area, upper section of the sorting area, middle section of the sorting area, lower section of the sorting area, and gangue channel. Each corresponding area has at least two measuring points forming a differential pressure measurement channel. An airborne vision sensor is installed above the sorting area at a 15°–35° downward angle relative to the bed surface, covering the sorting area and the discharge port. Accelerometers and acoustic sensors are fixed to the bed frame and inner wall of the machine casing in the sorting area, at a distance of no more than 50 mm from the bed surface. Online ash sensors are installed in the clean coal discharge channel and the gangue discharge channel, respectively. Dust sensors are installed in the sampling branches before and after the exhaust gas pipeline dust collector. Each sensor acquisition channel is coded by area index and bound to a one-to-one mapping relationship with the multi-zone air supply unit.

[0031] The bed pressure differential sensor array is partitioned within the air distribution chamber at the bottom of the sorting unit, allowing independent monitoring of the bed pressure differential in the feeding zone, sorting zone, and gangue channel. Each corresponding zone has at least two measuring points forming a differential pressure measurement channel, preventing errors from a single measuring point from affecting the measurement results. The airborne vision sensor's downward viewing angle is designed between 15° and 35°, avoiding image distortion caused by vertical viewing angles while ensuring the field of view covers the sorting zone and discharge port, completely capturing the particle movement trajectory. Accelerometers and acoustic sensors are fixed to the bed frame and inner wall of the casing in the sorting zone, with a distance of no more than 50 mm from the bed surface, effectively collecting vibration signals and particle collision acoustic characteristics. Online ash sensors are deployed in the clean coal and gangue discharge channels to directly monitor product quality. Dust sensors are installed in the sampling branches before and after the exhaust gas pipeline dust collector to compare dust concentrations before and after dust removal. Each sensor acquisition channel is mapped one-to-one with the multi-zone air supply unit through regional index coding, ensuring accurate correspondence between data and actuators.

[0032] Specifically, the zoned layout of the bed pressure differential sensor array, combined with multi-point differential pressure channels, accurately reflects changes in bed resistance in different areas, providing a regionalized data foundation for subsequent wind speed control. Airborne vision sensors, optimized for a top-down perspective, fully cover the sorting area and discharge port while avoiding image distortion. Combined with regional index coding, this achieves a direct correlation between visual data and the sorting area. Close-proximity installation of acceleration and acoustic sensors ensures high-fidelity acquisition of vibration signals and particle motion acoustic characteristics, aiding in the assessment of stratification stability. The online ash sensor, installed at the discharge channel, directly reflects the ash difference between clean coal and gangue. Combined with the ash-density calibration curve, it accurately calculates real-time cutting density. The arrangement of front and rear sampling branches for dust sensors effectively monitors dust removal efficiency and emission levels. The mapping relationship between the regional index coding of each sensor channel and the multi-zone air supply unit allows for precise matching of acquired data to the actuators by region, avoiding cross-interference of control commands and improving the coordination and response speed of multi-zone air supply control.

[0033] As a preferred embodiment, the solution of this application is specifically implemented as follows: A differential pressure sensor array is installed in the air distribution chamber at the bottom of the sorting unit, corresponding one-to-one with the feed area, upper section of the sorting area, middle section of the sorting area, lower section of the sorting area, and gangue channel. Each corresponding area has at least two measuring points forming a differential pressure measurement channel. An onboard vision sensor is installed above the sorting area at a 25° downward angle relative to the bed surface, covering the sorting area and the discharge port. Accelerometers and acoustic sensors are fixed to the bed frame and inner wall of the casing, 40 mm from the bed surface. Online ash sensors are installed in the clean coal discharge channel and the gangue discharge channel. Dust sensors are installed in the sampling branches before and after the exhaust gas dust collector. Each sensor's acquisition channel is coded by area index and bound to a one-to-one mapping relationship with the multi-zone air supply unit.

[0034] Specifically, the bed pressure differential sensing array consists of five sets of differential pressure sensors, each set containing two measuring points, corresponding to the feed area, upper section of the sorting area, middle section of the sorting area, lower section of the sorting area, and gangue channel, respectively. The airborne vision sensor uses a high-speed industrial camera, installed 2 meters above the sorting area, facing the bed surface at a 25° downward angle. Accelerometers and acoustic sensors are fixed to the bed frame and the inner wall of the casing, respectively, 40 mm from the bed surface. The online ash content sensor uses a dual-energy X-ray transmission ash analyzer, installed in the clean coal and gangue discharge channels. The dust sensor uses a laser scattering dust concentration meter, installed in the sampling branches before and after the dust collector.

[0035] Furthermore, each sensor acquisition channel is coded by region index. For example, the feed area corresponds to code A, the upper section of the sorting area corresponds to code B, the middle section of the sorting area corresponds to code C, the lower section of the sorting area corresponds to code D, and the gangue channel corresponds to code E. Correspondingly, the fans and valves of each zone in the multi-zone air supply unit also adopt the same coding method, thereby establishing a one-to-one mapping relationship between the sensor acquisition channels and the air supply unit.

[0036] Through the above technical solution, this application achieves comprehensive monitoring of key parameters in the sorting process. The bed pressure differential sensor array accurately reflects changes in the bed state in each area. Airborne visual sensors provide direct observation of particle stratification. Accelerometers and acoustic sensors capture bed surface vibration characteristics. Online ash and dust sensors monitor sorting quality and environmental indicators, respectively. A regional index coding system establishes a mapping relationship between sensors and the air supply unit, laying the foundation for precise zoned control. Therefore, this solution provides comprehensive and accurate data support for multi-objective collaborative control, contributing to improved stability and controllability of the sorting effect.

[0037] In some of the solutions described above in this application, the data acquisition unit collects multidimensional operational data through multiple sensors. However, in actual operation, due to the different sampling frequencies of each sensor and the single data processing method, the collected multidimensional data suffers from problems such as time asynchrony, large signal noise interference, and difficulty in data fusion, which in turn affects the accuracy of subsequent control commands.

[0038] This application further proposes a data acquisition unit for synchronously acquiring multi-dimensional operational data, including: the data acquisition unit synchronizing and clock-aligning all sensors using a unified timestamp, ensuring that the time error between any two signals is no greater than 2 milliseconds. The bed pressure difference signal is sampled at at least 200 Hz and, after zero-point drift correction and Kalman filtering, outputs the bed porosity. An airborne vision sensor acquires data at at least 60 frames per second, and after lens distortion correction, flat-field correction, and region interest segmentation, calculates the stratification index based on the optical density difference between upper and lower layers, texture gradient projection, and the vertical migration vector ratio of the optical flow field, and determines the particle stratification state based on the stratification index. Acceleration and acoustic signals are sampled at at least 1 kHz and, after denoising and envelope extraction, are used to assist in judging stratification stability. The online ash signal is sampled at at least 1 Hz and, after temperature and baseline compensation, is converted to real-time cutting density based on the ash-density calibration curve. The mismatch rate is calculated based on a joint criterion of the proportion of out-of-bounds particles and the online ash deviation. Dust emission data is output after outlier removal and moving average processing. All results are cached with a unified timestamp and regional index, forming multi-dimensional operational data.

[0039] The unified timestamp synchronization uses a hardware clock synchronization module to align the clocks of each sensor, ensuring time consistency across sensor data. Bed pressure difference signals are corrected for zero-point drift to eliminate sensor baseline offset, and Kalman filtering suppresses high-frequency noise. Airborne vision sensors use regional interest delineation to define key areas for focusing and sorting, and the optical flow field migration vector ratio reflects particle motion trends. Acceleration and acoustic signals are extracted via envelope and used to identify bed vibration stability. Online ash signals are compensated for to eliminate environmental interference, and the ash-density calibration curve uses piecewise linear interpolation to improve conversion accuracy. Mismatch rate calculation combines the proportion of out-of-bounds particles with ash deviation to avoid misjudgment based on a single criterion. Dust signals are processed using a moving average to smooth instantaneous fluctuations.

[0040] Specifically, the data acquisition unit first timestamps and synchronizes the data from each sensor, controlling the time error to within 2 milliseconds using a hardware clock module to ensure the timing consistency of subsequent data fusion. The bed pressure difference signal is sampled at a high frequency of 200 Hz, then undergoes zero-point drift correction and Kalman filtering to eliminate sensor errors and random noise, outputting a stable bed porosity. The airborne vision sensor acquires images at 60 frames per second, corrects for lens distortion, compensates for uneven illumination with flat field correction, and then extracts optical features of key areas in the sorting region by dividing the region into interest boxes. A stratification index is generated by calculating the optical density difference between upper and lower layers, the difference in texture gradient projection, and the proportion of the optical flow field migration vector, thereby determining the particle stratification state. Acceleration and acoustic signals are sampled at 1 kHz, and wavelet denoising is used to remove background noise. After envelope extraction, vibration energy distribution is analyzed to assist in evaluating stratification stability. The online ash sensor samples at 1 Hz, and temperature compensation corrects sensor drift. Based on a pre-calibrated ash-density curve, a lookup table and interpolation are performed to output the real-time cutting density. The mismatch rate is calculated by weighting the proportion of out-of-bounds particles with the online ash content deviation, avoiding misjudgments caused by a single indicator. After outlier removal, the dust signal is processed using a moving average with a window length of 5 seconds to output stable dust emission data. All processed data carries a unified timestamp and regional index, and is aligned and cached according to time series to form a multi-dimensional operational dataset, providing high-precision, low-latency input for subsequent control.

[0041] As a preferred embodiment, the solution of this application is specifically implemented as follows: The data acquisition unit synchronizes and aligns the timestamps of all sensors to ensure that the time error between any two signals is no greater than 2 milliseconds. The bed pressure difference signal is sampled at 250 Hz and, after zero-point drift correction and Kalman filtering, outputs the bed porosity. The airborne vision sensor acquires data at 80 frames per second. After lens distortion correction, flat-field correction, and region of interest segmentation, the stratification index is calculated based on the optical density difference between upper and lower layers, texture gradient projection, and the vertical migration vector ratio of the optical flow field. The stratification state of particles is determined based on the stratification index. Acceleration and acoustic signals are sampled at 1.2 kHz and denoised and envelope extracted to aid in determining stratification stability. The online ash signal is sampled at 1.5 Hz and, after temperature and baseline compensation, is converted to real-time cutting density based on the ash-density calibration curve. The mismatch rate is calculated based on a joint criterion of the proportion of out-of-bounds particles and the online ash deviation. The dust signal is processed by outlier removal and moving average to output dust emissions. Each result carries a unified timestamp and a region index-aligned cache, forming multidimensional runtime data.

[0042] Through the above technical solutions, this application achieves high-precision synchronous acquisition and processing of multi-dimensional operational data. Unified timestamp synchronization and clock alignment ensure the temporal consistency of data from different sensors, laying the foundation for subsequent data fusion and multi-objective collaborative control. High-frequency sampling and filtering of bed pressure difference signals improves the accuracy of porosity calculation. High-frame-rate acquisition and image preprocessing of airborne vision sensors enhance the reliability of stratification index calculation. Auxiliary judgment using acceleration and acoustic signals improves the comprehensiveness of stratification stability assessment. Compensation processing and calibration curve conversion of online ash signals improve the accuracy of real-time cutting density estimation. Joint criterion calculation of mismatch rate enhances the objectivity of sorting effect evaluation. Outlier processing of dust signals improves the accuracy of emission monitoring. Unified caching of multi-dimensional operational data provides comprehensive and reliable data support for subsequent multi-objective collaborative control, effectively improving the overall performance and stability of the vibrating airflow collaborative sorting system.

[0043] In some of the above-mentioned schemes in this application, when the multi-objective cooperative controller performs fusion calculations on multi-dimensional operating data, there are problems such as time synchronization errors of different sensor data, signal noise interference, and unreasonable weight allocation of multi-source heterogeneous data fusion. This results in insufficient confidence in the estimated values ​​of bed layering degree and real-time cutting density, affecting the accuracy of subsequent optimized control command generation.

[0044] This application further proposes a multi-objective collaborative controller that, when fusing multi-dimensional operational data, uses a unified timestamp and regional index to construct a feature vector from bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. Weighted fusion based on noise covariance adaptive estimation is performed within a 2-5 second sliding time window. Constrained Bayesian filtering is applied to the stratification index and bed porosity to output the degree of bed stratification. The fusion weight parameters satisfy a value between 0 and 1, with the sum of the weights being 1. The estimated real-time cutting density is obtained by using observed values ​​obtained from online ash content and ash-density calibration curves through table lookup and interpolation as measurement items. This, combined with the joint correction of the proportion of out-of-bounds particles, outputs the estimated real-time cutting density, and a confidence interval is provided as the estimation confidence index.

[0045] The unified timestamp and regional index data alignment method ensures the consistency of multidimensional operational data in the spatiotemporal dimension, eliminating temporal misalignment caused by differences in sensor sampling frequencies. A sliding time window of 2–5 seconds balances the relationship between data update rate and computational load. A weighted fusion method based on adaptive noise covariance estimation dynamically adjusts the fusion weight parameters by calculating the noise statistical characteristics of each sensor signal in real time, suppressing the interference of high-frequency noise on the feature vector. Constrained Bayesian filtering incorporates the physical constraints of the stratification index and bed porosity into the state estimation process, improving the output accuracy of the bed stratification degree through probability distribution correction. The joint correction process for real-time cutting density cross-validates ash data measured by online ash sensors with the proportion of out-of-bounds particles. Utilizing the nonlinear relationship of the ash-density calibration curve, density observations are generated using cubic spline interpolation, and the proportion of out-of-bounds particles is used to compensate for errors in the density estimate.

[0046] Specifically, the multidimensional operational data output by the data acquisition unit is spatially located and temporally synchronized using a unified timestamp and regional index, and then organized into a feature vector containing six dimensions. Within a sliding time window, the noise covariance matrix is ​​updated online using a recursive least squares method to calculate the noise variance of each sensor signal, thereby generating fusion weight parameters inversely proportional to signal quality. Constrained Bayesian filtering introduces physical boundary conditions for the stratification index and bed porosity during state estimation. For example, the stratification index is limited to a range of 0–1, and the bed porosity change rate does not exceed 0.05 per second. The posterior probability density function is corrected by truncating a Gaussian distribution. During real-time cutting density estimation, the online ash sensor outputs ash data every second. After temperature compensation and baseline correction, cubic spline interpolation is performed on the ash-density calibration curve to generate density observations. Simultaneously, the proportion of out-of-bounds particles is statistically analyzed using an airborne vision sensor to determine the percentage of particles whose density deviates from the target value within the sorting area, serving as an auxiliary correction term for linearly weighted correction of the density observations. The final output of the real-time cut density estimate is accompanied by a confidence interval calculated from the Kalman filter covariance matrix, which is used to characterize the reliability of the estimation result.

[0047] As a preferred embodiment, the solution of this application is specifically implemented as follows: When fusing multi-dimensional operational data, the multi-objective collaborative controller uses a unified timestamp and regional index to construct a feature vector from bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. Weighted fusion based on noise covariance adaptive estimation is performed within a 3-second sliding time window. Constrained Bayesian filtering is applied to the stratification index and bed porosity to output the degree of bed stratification. The fusion weight parameters are set to values ​​between 0 and 1, with the sum of the weights equal to 1. The estimated real-time cutting density is obtained by using observations obtained from online ash content and ash-density calibration curves through table lookup and interpolation. This, combined with the joint correction of the proportion of out-of-bounds particles, outputs the estimated real-time cutting density, and a confidence interval is provided as the estimation confidence index.

[0048] Specifically, the multi-objective collaborative controller first performs time alignment and region matching on the collected multi-dimensional operational data. Then, within a 3-second sliding time window, it executes a weighted fusion algorithm based on adaptive estimation of noise covariance for each data point. Specifically, the bed porosity and stratification index are fused using constrained Bayesian filtering to obtain an estimate of the bed stratification degree. The fusion weights are dynamically adjusted based on the measurement accuracy of each sensor, but the weight sum is guaranteed to be 1. For the real-time cutting density, an initial estimate is obtained by combining online ash data with a pre-calibrated ash-density curve, which is then jointly corrected with the proportion of out-of-bounds particles, finally outputting the estimated real-time cutting density and its confidence interval.

[0049] Through the above technical solutions, this application achieves effective fusion of multi-source heterogeneous sensor data, improving the accuracy and reliability of bed stratification and real-time cutting density estimation. The sliding time window and adaptive weight fusion enhance the system's robustness to operational fluctuations. The introduction of constrained Bayesian filtering effectively suppresses the divergence in stratification index and porosity estimates. Simultaneously, the real-time cutting density estimation employs multi-source data correction and confidence interval output, providing a more reliable basis for subsequent control decisions. These improvements contribute to enhancing the stability and sorting accuracy of the vibrating airflow sorting process.

[0050] In some of the solutions mentioned above in this application, the multi-objective cooperative controller has problems such as delayed response of control parameters and insufficient matching degree between the prediction model and the actual working conditions when making rolling predictions. This makes it difficult to stably control the real-time cutting density deviation and mismatch rate during the sorting process, while energy consumption and dust emissions may exceed the constraints.

[0051] This application further proposes a multi-objective cooperative controller that predicts and simulates apparent wind speed, vibration frequency, amplitude, longitudinal tilt angle, lateral tilt angle, and material distribution plate position within a control step size of 1–2 seconds and a prediction time domain of 10–30 seconds. A piecewise affine ARX model is used as the prediction method. Based on the prediction results, the objective is to minimize the real-time cutting density deviation and mismatch rate. Constraints are set on fan energy consumption and dust emissions, and constraints are imposed on the numerical range and rate of change of the controlled variables. Specifically, the apparent wind speed of the zone is limited to 0.4–2.2 m / s, the vibration frequency is limited to 4–12 Hz, the amplitude is limited to 2–8 mm, the longitudinal tilt angle is limited to 0–4 degrees, the lateral tilt angle is limited to -5–5 degrees, and the material distribution plate position is limited to ±20 mm. An upper limit for the change range of the controlled variables is set for gradual adjustment.

[0052] Specifically, the control step size and prediction time domain settings improve response speed by shortening the control cycle, while expanding the prediction window to cover the dynamic delay of the sorting process. The piecewise affine ARX model improves the model's adaptability to various operating conditions by decomposing the nonlinear system into multiple linear sub-intervals. The numerical range of the control parameters is determined based on a combination of the sorting machine's physical limits and sorting efficiency; for example, the lower limit of apparent wind speed prevents bed collapse, while the upper limit prevents particle separation. Gradually adjusting the upper limit limits the magnitude of single-step control parameter changes, preventing sudden parameter changes that could lead to bed instability.

[0053] Specifically, the multi-objective collaborative controller collects multi-dimensional operational data within each control step and uses a piecewise affine ARX model to perform rolling predictions of zoned apparent wind speed, vibration parameters, bed tilt angle, and sorting plate position. The prediction time domain covers the complete cycle of the sorting bed's dynamic response, such as the residence time of particles migrating to the discharge port. The model switches to the corresponding linear sub-model based on the current operating conditions, generating a predicted sequence of control quantities for the next 10-30 seconds. In the optimization objective function, real-time cutting density deviation and mismatch rate are used as the main optimization terms, while fan power and dust concentration are used as constraint terms. The optimal combination of control quantities is solved through quadratic programming. The numerical range of control quantities is ensured to guarantee the safe operation of the sorter through hard constraints, such as an upper limit on vibration frequency to avoid mechanical resonance and a range on the sorting plate position to prevent discharge port blockage. The upper limit is gradually adjusted by setting a maximum change in a single step, such as an amplitude adjustment not exceeding 1 mm / s and a longitudinal tilt angle adjustment not exceeding 0.5 degrees / s, to smooth the control quantity trajectory and reduce mechanical impact. In this process, the synergistic effect of the predictive model and the constraints improves the control accuracy of the sorting boundary stability, while meeting the dual constraints of energy consumption and environmental protection indicators.

[0054] As a preferred embodiment, the solution of this application is specifically implemented as follows: When the multi-objective cooperative controller performs rolling predictions based on the predictive model and generates optimized control commands, the control step size is set to 1.5 seconds, and the prediction time domain is set to 20 seconds. Predictive simulations are performed on the apparent wind speed, vibration frequency, amplitude, longitudinal tilt angle, lateral tilt angle, and material distribution plate position for each zone, using a piecewise affine ARX model as the prediction method. Based on the prediction results, the goal is to minimize the real-time cutting density deviation and mismatch rate, while simultaneously setting constraints on fan energy consumption and dust emissions, and imposing constraints on the numerical range and rate of change of the controlled variables. Specifically, the apparent wind speed for each zone is limited to 0.6–2.0 m / s, the vibration frequency to 5–10 Hz, the amplitude to 3–7 mm, the longitudinal tilt angle to 0.5–3.5 degrees, the lateral tilt angle to -4–4 degrees, and the material distribution plate position to ±15 mm. The upper limit for the change in the control quantity is set to be adjusted gradually. For example, the apparent wind speed of each zone should not exceed 0.2 m / s, the vibration frequency should not exceed 0.5 Hz, the amplitude should not exceed 0.5 mm, the tilt angle should not exceed 0.5 degrees, and the position of the material distribution plate should not exceed 2 mm.

[0055] Through the above technical solution, this application achieves multi-objective collaborative optimization control of the sorting process. By setting reasonable control step size and prediction time domain, the real-time performance and prediction accuracy of the control system are ensured. Using a piecewise affine ARX model for prediction effectively describes the nonlinear characteristics of the system. By imposing numerical range and rate of change constraints on the manipulated variables, drastic fluctuations in the control variables are avoided, improving system stability. Simultaneously considering multiple objectives such as real-time cutting density deviation, mismatch rate, energy consumption, and dust emissions, a comprehensive balance between sorting effect, economy, and environmental protection is achieved. This multi-objective collaborative control strategy can effectively cope with feed fluctuations and changes in operating conditions, improving the stability of the sorting boundary and overall operating indicators.

[0056] In some of the solutions mentioned above in this application, when adjusting the apparent wind speed of each zone, the multi-zone air supply unit may be subject to the risk of bed instability due to fluctuations in bed pressure difference or abnormal dust emissions. At the same time, a single control loop is difficult to balance wind speed tracking accuracy and system pressure stability, which may lead to transient shocks or excessive energy consumption.

[0057] This application further proposes that, after receiving optimized control commands, the multi-zone air supply unit converts the target apparent wind speed of each zone into an air volume distribution command based on the zone index, and generates specific setpoints according to the correspondence between fan speed and valve opening. The multi-zone air supply unit executes the setpoints through cascaded control of variable frequency fans and electric valves, employing a cascade strategy of tracking control with apparent wind speed as the controlled variable in the outer loop and stabilizing control with static pressure in the air distribution chamber as the controlled variable in the inner loop. The outer loop outputs the air volume target, and the inner loop outputs the fan speed and valve opening. Both use a gradual adjustment of the upper limit and an S-shaped ramp to suppress transient impacts. During execution, monotonicity constraints are set for each zone to ensure that the apparent wind speed in the sorting zone along the material flow direction does not increase, and the apparent wind speed in the gangue channel does not fall below the apparent wind speed threshold in the feed zone. When the bed pressure difference fluctuation provided by the data acquisition unit exceeds the limit or the dust emission reaches the threshold, the multi-zone air supply unit triggers the abnormal allocation logic, increases the apparent wind speed in the gangue channel and the feed area and correspondingly reduces the apparent wind speed in the lower section of the sorting area, while keeping the total air volume within the energy consumption constraint.

[0058] The process of converting the target apparent wind speed into airflow distribution commands is achieved through a preset wind speed-airflow conversion model, which is based on the equivalent cross-sectional area and resistance characteristics of the air distribution chambers in each zone. In the cascaded control of the variable frequency fan and electric valve, the variable frequency fan is responsible for coarse adjustment of the total airflow, while the electric valve is responsible for fine adjustment of the zone airflow distribution. The outer loop tracking control uses a proportional-integral algorithm, while the inner loop pressure stabilization control uses a fuzzy proportional-integral-differential algorithm. The S-shaped ramp method uses an acceleration limit curve to ensure a smooth transition during the adjustment process. Monotonicity constraints are achieved through inequality constraints in the online optimization algorithm to ensure that the apparent wind speed gradient along the material flow direction in the sorting zone decreases. The abnormal distribution logic is triggered by a pre-set rule base; when the bed pressure difference fluctuation exceeds the preset standard deviation or the dust concentration exceeds the threshold, the wind speed redistribution strategy is activated.

[0059] Specifically, after receiving the optimized control command, the multi-zone air supply unit first maps the target apparent wind speed to the corresponding zone according to the zone index, and calculates the required air volume for each zone through the wind speed-air volume conversion model. The variable frequency fan adjusts its speed according to the total air volume demand, while the electric valves of each zone adjust their opening according to the allocation ratio. The outer loop controller monitors the actual apparent wind speed in real time and compares it with the target value to generate an air volume correction command. The inner loop controller dynamically adjusts the fan speed and valve opening according to the static pressure change in the air distribution chamber to maintain stable system pressure. During the adjustment process, an S-shaped ramp function is used to limit the rate of parameter change to avoid airflow impact caused by step adjustments. The apparent wind speed in the sorting zone along the material flow direction is forcibly constrained to a monotonically non-increasing sequence to prevent particle back-mixing due to sudden changes in local wind speed. When abnormal fluctuations in bed pressure difference or excessive dust emissions are detected, the system automatically increases the wind speed in the gangue channel and feed zone to enhance gangue removal capacity, while reducing the wind speed in the lower section of the sorting zone to reduce fine particle diffusion. During this process, closed-loop control of the total air volume ensures that energy consumption does not exceed the limit. For example, during the process of increasing the wind speed in the gangue channel, the fan speed increases at a rate not exceeding 5% per minute, and the valve opening adjustment step is controlled within 2%.

[0060] As a preferred embodiment, the solution of this application is specifically implemented as follows: After receiving the optimized control command, the multi-zone air supply unit converts the target apparent wind speed of each zone into an air volume distribution command based on the zone index, and generates specific setpoints according to the correspondence between fan speed and valve opening. The multi-zone air supply unit executes the setpoints through cascaded control of variable frequency fans and electric valves. It adopts a cascade strategy of tracking control with apparent wind speed as the controlled variable in the outer loop and stabilizing control with static pressure in the air distribution chamber as the controlled variable in the inner loop. The outer loop outputs the air volume target, and the inner loop outputs the fan speed and valve opening. Both adopt a gradual adjustment of the upper limit and use an S-shaped ramp to suppress transient impacts.

[0061] During execution, monotonicity constraints are set for each zone to ensure that the apparent wind speed in the sorting zone along the material flow direction does not increase, and the apparent wind speed in the gangue channel does not fall below the threshold of the apparent wind speed in the feeding zone. When the bed pressure difference fluctuation provided by the data acquisition unit exceeds the limit or the dust emission reaches the threshold, the multi-zone air supply unit triggers abnormal allocation logic, increasing the apparent wind speed in the gangue channel and the feeding zone and correspondingly reducing the apparent wind speed in the lower section of the sorting zone, while keeping the total air volume below the energy consumption constraint.

[0062] Specifically, the multi-zone air supply unit includes a variable frequency fan, electric valves, and a controller. After receiving the optimized control command, the controller first converts the target apparent wind speed into an air volume distribution command. For example, for the five zones—the feed zone, the upper section of the sorting zone, the middle section of the sorting zone, the lower section of the sorting zone, and the gangue channel—the target apparent wind speeds are set to 1.8 m / s, 1.5 m / s, 1.2 m / s, 0.9 m / s, and 2.0 m / s, respectively. The controller then calculates the corresponding air volume distribution command based on the area of ​​each zone.

[0063] Furthermore, the controller converts the airflow distribution command into setpoints for fan speed and valve opening based on the fan characteristic curve and valve flow coefficient. A cascade control strategy is employed, with the outer loop using apparent wind speed as the controlled variable and the inner loop using the static pressure in the air distribution chamber as the controlled variable. The outer loop controller uses a PID algorithm to calculate the target airflow based on the deviation between the apparent wind speed and the setpoint. The inner loop controller also uses a PID algorithm to adjust the fan speed and valve opening based on the deviation between the static pressure in the air distribution chamber and the target value.

[0064] Therefore, the controller adjusts the fan speed and valve opening by gradually increasing the upper limit, with each adjustment not exceeding 5%, and uses an S-shaped ramp curve for smooth transition to suppress transient impacts. Simultaneously, the controller sets monotonicity constraints on the wind speed of each zone to ensure that the apparent wind speed in the sorting zone decreases along the material flow direction, and that the apparent wind speed in the gangue channel is not less than 1.1 times the apparent wind speed in the feed zone.

[0065] Furthermore, when the detected bed pressure differential fluctuation exceeds a set threshold or dust emissions reach the limit, the controller triggers abnormal allocation logic. For example, it increases the apparent air velocity in the gangue channel and feed area by 10%, while reducing the apparent air velocity in the lower section of the sorting area by 15% to stabilize the bed and suppress dust. During this process, the controller ensures that the total air volume does not exceed energy consumption constraints by adjusting the fan speed and valve opening.

[0066] Through the above technical solutions, this application achieves precise control and dynamic adjustment of multi-zone air supply. The cascade control strategy improves the accuracy and response speed of apparent wind speed control. The use of gradual adjustment and S-shaped ramp effectively suppresses transient impacts and avoids bed disturbance. The established monotonicity constraints and abnormal allocation logic further ensure the stability of bed stratification. Simultaneously, energy-saving operation is achieved through total air volume constraints. This multi-objective collaborative control method, while ensuring sorting effectiveness, also considers energy consumption and environmental protection requirements, improving the overall performance and adaptability of the vibrating airflow sorting system.

[0067] In some of the solutions described above in this application, the servo eccentric excitation mechanism lacks a quantitative assessment of the particle migration equilibrium state when adjusting vibration parameters, resulting in a lag in the adjustment of vibration frequency and amplitude, which cannot effectively suppress local accumulation or excessive diffusion during particle stratification.

[0068] This application further proposes that after receiving the optimized control command, the servo eccentric vibration mechanism determines the target vibration frequency and amplitude based on the particle migration balance index, which is composed of the stratification index change rate, particle stratification state, and acceleration envelope characteristics. It adopts dual closed-loop control of frequency and amplitude to achieve gradual adjustment. The outer loop uses the particle migration balance index as the controlled variable to give the target vibration parameters, while the inner loop uses the vibration frequency and amplitude as the controlled variables to track the target values ​​through the encoder and displacement sensor in a closed loop. It also applies a gradual adjustment upper limit to the vibration parameters and uses an S-shaped ramp to limit transient impacts.

[0069] The particle migration balance index is constructed by fusing the stratification index change rate, particle stratification state classification results, and acceleration envelope spectrum characteristics. The stratification index change rate reflects the dynamic trend of bed stratification velocity, particle stratification state is classified using particle distribution characteristics captured by a visual sensor, and acceleration envelope characteristics are used to extract the resonant frequency offset of the bed skeleton through vibration signal spectrum analysis. The outer loop controller compares the particle migration balance index with preset thresholds to generate target values ​​for vibration frequency and amplitude. The inner loop controller uses an encoder to provide real-time feedback of the actual rotational speed of the vibration motor and a displacement sensor to measure the actual displacement of the excitation block, performing closed-loop tracking with the target frequency and amplitude respectively. The upper limit for gradual adjustment is set to ensure that the frequency adjustment does not exceed 0.5 Hz and the amplitude adjustment does not exceed 0.3 mm within each control cycle. The S-shaped ramp method uses an acceleration curve to smoothly transition the vibration parameters during adjustment.

[0070] Specifically, when the particle migration balance index detects a decrease in stratification velocity and the appearance of low-frequency components in the acceleration envelope spectrum, the outer loop controller determines that particle migration is hindered and generates a target value for increasing the vibration frequency. The inner loop controller obtains the actual rotational speed of the vibrating motor through the encoder, and outputs a frequency converter adjustment signal after proportional-integral calculation of the difference with the target frequency, so that the motor speed gradually increases to the target value in an S-shaped curve. At the same time, the displacement sensor monitors the displacement amplitude of the excitation block in real time. When the target amplitude value increases, the amplitude is controlled in a closed loop by adjusting the mass distribution of the eccentric block. In this process, the gradual adjustment of the upper limit of the vibration parameters avoids bed flow instability caused by sudden changes, and the S-shaped ramp method effectively suppresses the interference of mechanical impact on the sensor signal. Through the synergistic effect of the particle migration balance index and the dual closed-loop control, the vibration parameters and the dynamic response of the bed are matched in real time, maintaining the dynamic balance of the particle migration process.

[0071] As a preferred embodiment, the solution of this application is specifically implemented as follows: After receiving optimized control commands, the servo-driven eccentric vibration mechanism determines the target vibration frequency and amplitude based on a particle migration balance index jointly composed of the stratification index change rate, particle stratification state, and acceleration envelope characteristics. Specifically, the stratification index change rate is obtained by calculating the difference in stratification indices within adjacent time windows and dividing by the time interval; the particle stratification state is determined by processing images acquired by an airborne vision sensor; and the acceleration envelope characteristics are obtained by extracting the envelope from signals acquired by an accelerometer. After normalization, these three parameters are weighted and summed according to preset weights to obtain the particle migration balance index.

[0072] Furthermore, a dual closed-loop control system for frequency and amplitude is employed to achieve gradual adjustment. The outer loop uses the particle migration balance index as the controlled variable to set the target vibration parameters, while the inner loop uses vibration frequency and amplitude as controlled variables to track the target values ​​through an encoder and displacement sensor in a closed loop. Specifically, the outer loop controller calculates the target vibration frequency and amplitude using a PID algorithm based on the deviation between the particle migration balance index and the set target value. The inner loop controller receives the target vibration frequency and amplitude and uses a PID algorithm to control the servo motor speed and eccentric block position, achieving precise adjustment of the actual vibration frequency and amplitude.

[0073] Therefore, progressively increasing upper limits are applied to the vibration parameters using an S-shaped ramp to limit transient impacts. For example, the upper limit for frequency adjustment is set at 0.5 Hz, and the upper limit for amplitude adjustment is set at 0.2 mm / s. During adjustment, an S-shaped curve is used as the adjustment trajectory, ensuring that the changes in frequency and amplitude are relatively gradual at the beginning and end, and faster in the middle stage, thereby effectively suppressing transient impacts during the adjustment process.

[0074] Through the above technical solution, this application achieves precise closed-loop control of vibration parameters, effectively improving particle stratification stability. By using a particle migration balance index as the control basis, the system can dynamically adjust vibration parameters according to the actual sorting conditions, thus enhancing adaptability. Simultaneously, the application of gradual adjustment and the S-shaped ramp method effectively reduces transient impacts during vibration parameter adjustment, improving the stability and reliability of equipment operation. Furthermore, this solution achieves fine-tuning of particle migration behavior by coordinating vibration frequency and amplitude, thereby improving sorting accuracy and efficiency.

[0075] In some of the solutions mentioned above in this application, the adjustment of the bed tilt angle is based on a single parameter for unidirectional adjustment, without considering the spatial distribution differences of the material flow into the upper layer state and the particle boundary crossing phenomenon in the lateral dividing zone, resulting in insufficient backmixing suppression effect and decreased sorting efficiency.

[0076] This application further proposes that, after receiving optimized control commands, the bed surface attitude adjustment mechanism performs linked corrections on the longitudinal and lateral tilt angles based on the gradient of the stratification index along the material flow direction, the mismatch rate, and the statistical results of cross-boundary particles observed by airborne vision: when the stratification index decreases along the material flow direction and the mismatch rate increases, the longitudinal tilt angle is increased to prolong retention and suppress backmixing. When cross-boundary particles are concentrated in the lateral separation zone, the lateral tilt angle is adjusted to enhance the lateral separation of light and heavy particles. The adjustments of both the longitudinal and lateral tilt angles are constrained by numerical ranges and stepwise adjustment upper limits, and dual threshold hysteresis is set to avoid frequent adjustments. Furthermore, each attitude adjustment triggers a small compensation for the apparent wind speed target of adjacent zones to maintain the stability of the bed porosity.

[0077] The stratification index gradient analysis calculates the spatial change rate using the difference in stratification indices among three consecutive sorting zones along the material flow direction. When the absolute value of the gradient exceeds a preset threshold, longitudinal tilt correction is triggered. Cross-boundary particle statistics track particle trajectories in the transverse zoning zone using an onboard vision sensor, combining ash content data to identify the proportion of light particles entering the gangue channel or heavy particles mixing into the clean coal channel. A dual-threshold hysteresis mechanism sets the threshold for initiating adjustment higher than the threshold for stopping adjustment, ensuring stable adjustment actions during parameter fluctuations. Micro-compensation converts the tilt adjustment amount into the apparent wind speed increment of adjacent zones according to a preset ratio, with the compensation range controlled within ±3% of the original target value.

[0078] Specifically, when the stratification index decreases along the material flow direction and the mismatch rate exceeds the set threshold, the longitudinal tilt angle increases at a rate of 0.1 degrees / second, with a maximum adjustment of no more than 4 degrees. This adjustment prolongs the residence time of the material in the sorting zone, enhancing the density-based stratification process. When the proportion of out-of-bounds particles in the transverse zoning zone exceeds the threshold, the transverse tilt angle tilts towards the particle accumulation side at a rate of 0.05 degrees / second, with a maximum adjustment limited to ±5 degrees. After each tilt angle adjustment, an increment of 0.2-0.5 m / s is injected into the apparent wind speed target value of adjacent zones based on the adjustment amount, maintaining bed porosity stability through the multi-zone air supply unit. During the adjustment process, if the stratification index gradient or the proportion of out-of-bounds particles fluctuates within the dual-threshold hysteresis range, the current tilt angle setting remains unchanged to prevent frequent actuator movements.

[0079] As a preferred embodiment, the solution of this application is implemented as follows: When the stratification index is detected to be decreasing along the material flow direction and the mismatch rate exceeds a preset threshold, the longitudinal tilt angle is gradually increased to prolong the material residence time and reduce backmixing. If the airborne vision sensor detects that out-of-bounds particles are concentrated in the lateral separation zone, the lateral tilt angle is dynamically adjusted to enhance the lateral separation of light and heavy particles. During the tilt angle adjustment process, the longitudinal tilt angle is limited to the range of 0 to 4 degrees, and the lateral tilt angle adjustment range is constrained to between -5 degrees and 5 degrees. After each adjustment, the apparent wind speed of adjacent zones is synchronously compensated with a small amount. The compensation amount is dynamically calculated based on the current change in bed porosity to ensure the stability of the bed flow. The tilt angle adjustment mechanism adopts a dual-threshold hysteresis control logic. When the gradient change of the stratification index is lower than the first threshold or higher than the second threshold, a tilt angle hold or reverse adjustment command is triggered to avoid frequent actuator operation.

[0080] Through the above technical solutions, this application achieves active suppression of backmixing and improved separation efficiency of light and heavy particles. By linking and correcting the longitudinal and lateral tilt angles, the residence time of particles in the sorting zone is effectively extended, reducing secondary mixing of already stratified particles. Dynamic adjustment of the lateral tilt angle based on visual statistics enhances the lateral separation effect and reduces the probability of cross-contamination between light and heavy materials. The combination of dual-threshold hysteresis control and adjacent area wind speed compensation mechanism ensures adjustment accuracy while avoiding bed stability degradation caused by frequent orientation adjustments, maintaining continuous and stable operation of the sorting process.

[0081] In some of the solutions described above in this application, the bed tilt mechanism corrects the longitudinal and lateral tilt angles according to the optimized control command to suppress back mixing. However, in actual operation, simply adjusting the tilt angle may lead to a decrease in stratification stability due to dynamic changes in bed porosity. At the same time, frequent adjustment of the tilt angle can easily cause oscillation of the actuator, affecting the continuity of sorting.

[0082] This application further proposes a bed tilt mechanism that uses a linkage correction between the longitudinal and lateral tilt angles based on the gradient of the stratification index along the material flow direction, the mismatch rate, and statistical results of cross-boundary particles obtained by airborne vision. When the stratification index decreases along the material flow direction and the mismatch rate increases, the longitudinal tilt angle is increased to prolong retention and suppress backmixing. When cross-boundary particles are concentrated in the lateral separation zone, the lateral tilt angle is adjusted to enhance the lateral separation of light and heavy particles. The adjustment of both the longitudinal and lateral tilt angles is constrained by numerical ranges and stepwise adjustment limits, and a dual-threshold hysteresis is set to avoid frequent adjustments. Each tilt adjustment triggers a small compensation for the apparent wind speed target of adjacent zones to maintain the stability of the bed porosity.

[0083] The longitudinal tilt adjustment is triggered by detecting the gradient change of the stratification index along the material flow direction. When the gradient decreases beyond a threshold, incremental longitudinal tilt adjustment is initiated, with an upper limit of 4 degrees and an adjustment step not exceeding 0.5 degrees. The lateral tilt adjustment is based on the statistical proportion of cross-boundary particles in the lateral zoning zone determined by airborne vision. When the proportion exceeds a set threshold, lateral tilt correction is initiated, with the adjustment range limited to -5 to 5 degrees and an adjustment step not exceeding 0.3 degrees. A dual-threshold hysteresis mechanism sets independent thresholds for initiating and stopping adjustment. The longitudinal tilt adjustment initiation threshold is a 10% decrease in the stratification index gradient, and the stop threshold is a gradient rebound to 5%. The lateral tilt adjustment initiation threshold is when the proportion of cross-boundary particles exceeds 8%, and the stop threshold is when the proportion falls back to 5%. The apparent wind speed compensation for adjacent zones is ±3% of the current target value, with the compensation direction opposite to the tilt adjustment direction.

[0084] Specifically, when the stratification index is detected to be decreasing along the material flow direction and the mismatch rate exceeds a preset threshold, the longitudinal tilt angle is gradually increased in 0.5-degree increments to increase the bed surface inclination, thereby prolonging the material residence time and suppressing backmixing. Simultaneously, the apparent wind speed target value of adjacent zones is increased by 3% to compensate for fluctuations in bed porosity caused by tilt angle changes. When the airborne vision identifies that the proportion of out-of-bounds particles in the lateral zoning zone exceeds 8%, the lateral tilt angle is adjusted towards the particle concentration side in 0.3-degree increments to promote lateral separation of light and heavy particles. At this time, the apparent wind speed target value of adjacent zones is decreased by 3% to balance the bed fluidization state. The dual-threshold hysteresis mechanism avoids frequent adjustments due to measurement noise or instantaneous fluctuations by setting the difference between the start and stop thresholds, ensuring that the adjustment action is triggered only when there is a continuous deviation. Numerical range and step-by-step adjustment upper limit constraints prevent the actuator from exceeding limits or abruptly changing, ensuring a smooth adjustment process. By linking tilt angle adjustment with wind speed compensation, backmixing is suppressed and separation is enhanced while maintaining stable bed porosity and improving the continuity of the sorting process.

[0085] As a preferred embodiment, the solution of this application is implemented as follows: The material distribution adjustment mechanism performs fine adjustment of the material distribution plate position based on the deviation data between the estimated real-time cutting density and the target cutting density, combined with the comparison of the online ash content detection results of the clean coal discharge channel and the gangue discharge channel. When the real-time cutting density is higher than the target value and the ash content detection value of the clean coal channel exceeds the set range, the material distribution plate moves towards the gangue side at a preset step distance. When the real-time cutting density is lower than the target value and the ash content detection value of the gangue channel exceeds the allowable threshold, the material distribution plate moves towards the clean coal side at the same step distance. A dead zone range is set during the adjustment process, and the adjustment command is not triggered when the deviation is within the dead zone. After each adjustment action is completed, the current state is maintained for at least one complete control cycle. After a new round of online ash content data and cutting density estimation value are updated, it is determined whether to continue the subsequent fine-tuning operation based on the latest data, thereby avoiding high-frequency malfunctions caused by mechanical vibration or signal fluctuations.

[0086] Through the above technical solution, this application achieves dynamic and stable control of the sorting boundary, effectively suppressing the cutting density deviation caused by coal quality fluctuations or changes in operating conditions, and reducing the mismatch rate of clean coal and gangue products. By introducing dead zone and hold time constraint mechanisms, equipment wear and control system oscillations caused by frequent actuator movements are avoided, ensuring continuous and stable operation of the sorting process.

[0087] In the above embodiments, a complete closed-loop control link is constructed by setting up a data acquisition unit, a data processing unit, and an execution adjustment unit. The data acquisition unit uses a bed pressure differential sensor array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor to synchronously acquire multi-dimensional operational data such as bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate, and dust emission. The data processing unit performs fusion calculations on the multi-dimensional operational data through a multi-objective collaborative controller and performs rolling predictions in conjunction with a prediction model. It can output control commands in real time that ensure the optimization of cutting density and mismatch rate while taking into account energy consumption and dust constraints. The execution adjustment unit decomposes the optimized control commands into coordinated adjustment actions for the multi-zone air supply unit, the servo eccentric vibration mechanism, the bed surface attitude adjustment mechanism, and the material distribution adjustment mechanism, thereby realizing dynamic coupling control of airflow distribution, vibration parameters, bed surface attitude, and discharge boundary. Compared with existing solutions, this application effectively avoids the problems of sorting boundary drift and back mixing under coal quality changes, feed fluctuations and complex working conditions, ensures the stability of clean coal ash content and cutting density, and reduces energy consumption and incorporates dust emissions into the control target, demonstrating beneficial effects in sorting accuracy, system stability and green operation.

[0088] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a coal vibration airflow collaborative separation method based on multi-objective collaborative control, applied to the aforementioned coal vibration airflow collaborative separation system based on multi-objective collaborative control, including: S100: Synchronously collects multi-dimensional operating data, including bed porosity, stratification index, particle stratification state, real-time cutting density, mismatch rate and dust emission.

[0089] S200: Performs fusion calculations on multi-dimensional operating data to obtain estimates of the bed layering degree and real-time cutting density, and performs rolling predictions based on the prediction model to obtain optimized control commands that constrain energy consumption and dust emissions while minimizing real-time cutting density deviation and mismatch rate.

[0090] S300: Receives optimized control commands to adjust the apparent wind speed distribution in each zone to stabilize bed stratification. Dynamically adjusts vibration frequency and amplitude according to optimized control commands to maintain particle migration balance. Corrects longitudinal and transverse tilt angles according to optimized control commands to suppress backmixing. Adjusts discharge boundaries according to optimized control commands.

[0091] Specifically, when synchronously acquiring multi-dimensional operational data, the data acquisition unit synchronizes the timestamps of each sensor and aligns the clocks, ensuring a time error of no more than 2 milliseconds. The bed pressure difference signal is corrected for zero-point drift and Kalman filtered to output the bed porosity. Images acquired by the airborne vision sensor are stratified after distortion correction and region segmentation. The online ash signal is converted into real-time cutting density after temperature compensation, and the dust signal is output as dust emission after outlier removal. When performing fusion calculations on the multi-dimensional operational data, weighted fusion based on noise covariance adaptive estimation is performed within a sliding time window. A constraint Bayesian filter is applied to output the bed stratification degree, and the real-time cutting density estimate is corrected jointly by online ash and the proportion of out-of-bounds particles. When performing rolling prediction based on the prediction model, a piecewise affine ARX model is used to predict and simulate the control variables within a control step size of 1–2 seconds and a prediction time domain of 10–30 seconds. The goal is to minimize the real-time cutting density deviation and mismatch rate, constraining fan energy consumption and dust emissions, and generating optimized control commands. After receiving the optimization control command, the multi-zone air supply unit adjusts the apparent wind speed distribution of the zones, the servo eccentric excitation mechanism dynamically adjusts the vibration frequency and amplitude, the bed surface attitude adjustment mechanism corrects the longitudinal and lateral tilt angles, and the material distribution adjustment mechanism adjusts the position of the material distribution plate.

[0092] Specifically, during synchronous data acquisition, each sensor aligns its data using a unified timestamp and regional index encoding, ensuring spatiotemporal consistency of parameters such as bed porosity, stratification index, and real-time cutting density. Fusion computation employs weighted fusion within a sliding time window and Bayesian filtering to eliminate sensor noise interference and improve the accuracy of stratification degree and cutting density estimation. Rolling prediction, based on a piecewise affine ARX model, simulates the changing trends of apparent wind speed, vibration parameters, tilt angle, and material distribution plate position within the prediction time domain. Multi-objective optimization generates control commands that balance sorting accuracy and operational constraints. During adjustment, the multi-zone air supply unit adjusts airflow distribution according to the regional index mapping relationship; the servo eccentric vibration mechanism tracks target vibration parameters through dual closed-loop control; the bed surface attitude adjustment mechanism corrects the tilt angle based on the stratification index gradient and mismatch rate; and the material distribution adjustment mechanism performs step-by-step fine-tuning based on cutting density deviation and ash content comparison. This method stabilizes the sorting boundary in real time when the feed fluctuates by synchronously collecting data, fusing and estimating, rolling prediction and multi-mechanism collaborative execution, simultaneously reducing cutting density deviation and mismatch rate, and controlling energy consumption and dust emissions.

[0093] As a preferred embodiment, the specific implementation of this application is as follows: Multi-dimensional operational data is synchronously collected through a bed pressure differential sensing array, an airborne vision sensor, an acceleration and acoustic sensor, an online ash sensor, and a dust sensor. The bed pressure differential signal is processed by Kalman filtering to output the bed porosity. The images acquired by the airborne vision sensor are subjected to distortion correction and optical flow field analysis to calculate the stratification index. The online ash signal is converted into real-time cutting density through an ash-density calibration curve. Timestamp synchronization and sliding window fusion operations are performed on the multi-dimensional operational data. A constrained Bayesian filtering algorithm is used to estimate the bed stratification degree, and the estimated real-time cutting density is jointly corrected by ash observations and the proportion of out-of-bounds particles. Based on a piecewise affine ARX prediction model, rolling optimization is performed on the zoned apparent wind speed, vibration parameters, bed surface inclination angle, and material distribution plate position within the control step size to generate optimized control commands that minimize cutting density deviation and mismatch rate while constraining fan energy consumption and dust emissions. The multi-zone air supply unit adjusts the air volume distribution of each zone according to the optimization instructions. The servo eccentric vibration mechanism tracks the target vibration parameters through dual closed-loop control. The bed surface attitude adjustment mechanism corrects the tilt angle based on the stratification index gradient and out-of-bounds particle statistics. The material distribution adjustment mechanism performs step fine-tuning according to the ash content comparison results to stabilize the discharge boundary.

[0094] Through the above technical solution, this application realizes real-time perception of the bed layer stratification state and multi-objective dynamic collaborative control during the sorting process, effectively suppresses the cutting density drift and back mixing phenomenon caused by feed fluctuations, reduces energy consumption and controls dust emissions while maintaining stable sorting accuracy, and solves the problems of unstable sorting boundaries and difficulty in coordinating multiple indicators in the prior art.

[0095] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A coal vibration airflow collaborative sorting system based on multi-target collaborative control, characterized in that, Comprise: Data acquisition unit, data processing unit and execution adjustment unit; wherein, The data acquisition unit comprises bed pressure difference sensing array, on-board visual sensor, acceleration and acoustic sensor, online ash content sensor and dust sensor, the data acquisition unit is used for synchronously collecting multi-dimensional operation data, the multi-dimensional operation data comprises bed voidage, stratification index, particle stratification state, real-time cutting density, mismatch rate and dust emission; The data processing unit comprises multi-target cooperative controller, the multi-target cooperative controller carries out fusion operation on the multi-dimensional operation data, obtains the estimated value of bed stratification degree and real-time cutting density, and carries out rolling prediction based on the prediction model, and obtains the optimization control instruction of minimizing real-time cutting density deviation and mismatch rate while constraining energy consumption and dust emission; The execution adjustment unit comprises multi-zone air supply unit, servo eccentric excitation mechanism, bed surface posture adjustment mechanism and material distribution adjustment mechanism, wherein the multi-zone air supply unit receives the optimization control instruction to adjust the apparent wind speed distribution of each partition to stabilize the bed stratification, the servo eccentric excitation mechanism dynamically adjusts the vibration frequency and amplitude according to the optimization control instruction to maintain the balance of particle migration, the bed surface posture adjustment mechanism corrects the longitudinal and transverse inclination according to the optimization control instruction to suppress back mixing, and the material distribution adjustment mechanism adjusts the discharge boundary according to the optimization control instruction.

2. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 1, characterized in that, Also comprise: The bed pressure difference sensing array is arranged in the air distribution cavity at the bottom of the sorting host, and corresponds to the feeding zone, the upper section of the sorting zone, the middle section of the sorting zone, the lower section of the sorting zone and the gangue channel one by one, at least two measuring points are provided in each corresponding zone to form a differential pressure measuring channel; the on-board visual sensor is installed above the sorting zone and forms a 15°-35° downward viewing angle relative to the bed surface, the field of view covers the sorting zone and the discharge port; the acceleration and acoustic sensor is fixed to the bed surface skeleton and the inner wall of the machine shell, and is not more than 50 millimeters away from the bed surface; the online ash content sensor is installed in the clean coal discharge channel and the gangue discharge channel respectively; the dust sensor is installed in the sampling branch before and after the dust remover of the tail gas pipeline; each sensor acquisition channel is indexed and coded according to the region and is bound with the one-to-one mapping relationship of the multi-zone air supply unit.

3. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 2, characterized in that, When the data acquisition unit is used for synchronously collecting multi-dimensional operation data, it comprises: The data acquisition unit synchronizes the time stamps of each sensor and aligns the clocks, so that the time error of any two signals is not greater than 2 milliseconds; the bed pressure difference signal is sampled at a frequency of not less than 200 Hz, and after zero drift correction and Kalman filtering, the bed voidage is output; the on-board vision sensor collects images at a frequency of not less than 60 frames per second, after lens distortion correction, flat field correction and region of interest division, the stratification index is calculated based on the optical density difference between the upper and lower layers, the texture gradient projection and the upward and downward migration vectors of the optical flow field, and the particle stratification state is determined according to the stratification index; the acceleration and acoustic signals are sampled at a frequency of not less than 1 kHz, and after denoising and envelope extraction, they are used to assist in judging the stratification stability; the online ash content signal is sampled at a frequency of not less than 1 Hz, and after temperature and baseline compensation, the real-time cutting density is converted according to the ash content-density calibration curve; the mismatch rate is calculated based on the joint criterion of the out-of-bound particle proportion and the online ash content deviation; the dust signal is processed by outlier rejection and sliding mean to output the dust emission; each result is aligned and cached with a unified time stamp and region index to form the multi-dimensional operation data.

4. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 1, characterized in that, The multi-target cooperative controller performs fusion operation on the multi-dimensional operation data to obtain the estimated values of the bed stratification degree and the real-time cutting density, including: When the multi-target cooperative controller performs fusion operation on the multi-dimensional operation data, the bed voidage, stratification index, particle stratification state, real-time cutting density, mismatch rate and dust emission are combined into a feature vector with a unified time stamp and region index, and a weighted fusion based on noise covariance adaptive estimation is performed within a 2-5 second sliding time window; the stratification index and bed voidage are subjected to constrained Bayesian filtering to output the bed stratification degree, and the fusion weight parameter satisfies the value between 0 and 1 and the sum of the weights is 1; the estimated value of the real-time cutting density is obtained by looking up and interpolating the observation value of the online ash content and the ash content-density calibration curve as a measurement item, and the estimated value of the real-time cutting density is output by joint correction of the out-of-bound particle proportion, and a confidence interval is given as an estimation confidence index.

5. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 4, characterized in that, When the multi-target cooperative controller performs rolling prediction based on the prediction model and generates the optimized control instructions, the partition apparent wind speed, vibration frequency, amplitude, longitudinal inclination, transverse inclination and partition plate position are predicted and simulated within the range of control step length of 1-2 seconds and prediction time domain of 10-30 seconds, and a piecewise affine ARX model is used as the prediction method; based on the prediction results, the real-time cutting density deviation and the mismatch rate are minimized as the target, while the fan energy consumption and dust emission are set as the constraint conditions, and the numerical range and change rate of the control quantity are constrained, wherein the partition apparent wind speed is limited to 0.4-2.2 m / s, the vibration frequency is limited to 4-12 Hz, the amplitude is limited to 2-8 mm, the longitudinal inclination is limited to 0-4 degrees, the transverse inclination is limited to -5-5 degrees, the partition plate position is limited to ±20 mm, and the upper limit of the change amplitude of the control quantity is gradually adjusted.

6. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 5, characterized in that, The multi-zone air supply unit converts the target apparent wind speed of each partition into air volume distribution instructions based on the corresponding relationship of the region index after receiving the optimization control instructions, and generates specific set values based on the corresponding relationship of the fan speed and the valve opening. The multi-zone air supply unit executes the set values through the cascade control of the variable frequency fan and the electric valve, adopts the cascade strategy of the outer ring tracking control with the apparent wind speed as the controlled quantity and the inner ring stable pressure control with the air distribution cavity static pressure as the controlled quantity, the outer ring outputs the air volume target, and the inner ring outputs the fan speed and the valve opening, both of which adopt the step-by-step adjustment upper limit and the S-shaped ramp way to suppress the transient impact; the monotonicity constraint is set for each partition during the execution process, so that the apparent wind speed of the sorting zone along the material flow direction does not increase, and the apparent wind speed of the gangue channel is not lower than the apparent wind speed threshold of the feeding zone; when the bed layer pressure difference fluctuation provided by the data acquisition unit exceeds the limit or the dust emission reaches the threshold, the multi-zone air supply unit triggers the abnormal distribution logic, increases the apparent wind speed of the gangue channel and the feeding zone, and correspondingly reduces the apparent wind speed of the lower section of the sorting zone, while keeping the total air volume not exceeding the energy consumption constraint.

7. The coal vibration airflow collaborative sorting system based on multi-target collaborative control according to claim 5, characterized in that, The servo eccentric excitation mechanism determines the target vibration frequency and amplitude according to the particle migration balance index composed of the stratification index change rate, the particle stratification state and the acceleration envelope characteristics after receiving the optimization control instructions, and realizes step-by-step adjustment by adopting frequency and amplitude double closed-loop control, wherein the outer ring uses the particle migration balance index as the controlled quantity to give the target vibration parameters, the inner ring respectively uses the vibration frequency and amplitude as the controlled quantity to track the target value through the encoder and displacement sensor closed loop, and applies step-by-step adjustment upper limit and S-shaped ramp way to limit transient impact.

8. The coal vibrated airflow collaborative sorting system based on multi-target collaborative control according to claim 5, characterized in that, The bed surface posture adjusting mechanism adjusts the longitudinal inclination and the transverse inclination based on the gradient of the stratification index along the material flow direction, the mismatch rate and the statistical results of the out-of-bound particles by the on-board vision after receiving the optimization control instructions: when the stratification index decreases along the material flow direction and the mismatch rate increases, the longitudinal inclination is increased to prolong the residence time and suppress the back mixing; when the out-of-bound particles concentrate in the lateral zoning area, the transverse inclination is adjusted to enhance the lateral separation of light and heavy particles; the adjustment of the longitudinal inclination and the transverse inclination is constrained by the numerical range and the step-by-step adjustment upper limit, and double threshold hysteresis is set to avoid frequent adjustment, and each posture adjustment triggers a small amount of compensation of the apparent wind speed target of the adjacent partition to maintain the stability of the bed void fraction.

9. The multi-objective coordinated control based coal vibrated airflow coordinated sorting system according to claim 5, characterized in that, The material distribution adjustment mechanism, after receiving the optimization control instruction, performs fine step-by-step fine tuning of the material distribution plate position based on the deviation of the estimated value of real-time cutting density from the target cutting density and the online ash content of the clean coal and gangue discharge channels: when the real-time cutting density is higher than the target and the online ash content of the clean coal is high, the material distribution plate is moved to the gangue side by a preset step distance; when the real-time cutting density is lower than the target and the online ash content of the gangue is high, the material distribution plate is moved to the clean coal side by a preset step distance; the fine step-by-step fine tuning is constrained by the range of the material distribution plate position and the upper limit of step-by-step adjustment, and a dead zone and a minimum holding time are set to prevent high-frequency oscillation, while new multi-dimensional operation data are read after each position update to confirm the adjustment effect and decide whether to continue fine tuning, thereby stabilizing the discharge boundary.

10. A coal vibration airflow collaborative separation method based on multi-target collaborative control, applied to the coal vibration airflow collaborative separation system based on multi-target collaborative control according to any one of claims 1-9, characterized in that, It comprises: Synchronous acquisition of multi-dimensional operation data, including bed voidage, stratification index, particle stratification state, real-time cutting density, mismatch rate and dust emission; Fusion operation on the multi-dimensional operation data to obtain the estimated value of the bed stratification degree and the real-time cutting density, and rolling prediction based on a prediction model to obtain the optimization control instruction for minimizing the deviation of the real-time cutting density and the mismatch rate while constraining energy consumption and dust emission; Adjusting the apparent wind speed distribution of each partition to stabilize the bed stratification according to the optimization control instruction; According to the optimization control instruction, dynamically adjusting the vibration frequency and amplitude to maintain the balance of particle migration; according to the optimization control instruction, correcting the longitudinal and transverse inclination angles to suppress back mixing; and adjusting the discharge boundary according to the optimization control instruction.

Images