Vertical air selection drying integrated device and intelligent self-adaptive regulation and control method thereof

Abstract

The application discloses a vertical air selection and drying integrated device and an intelligent self-adaptive regulation and control method thereof, and belongs to the technical field of powder material drying and separation. The device comprises a vertical air selection and drying host, a multi-source sensing unit, an execution and regulation unit, an intelligent control unit and a cyclone dust removal unit; a plurality of layers of inclined and staggered vibration beds are arranged in the host to form a broken line-shaped material channel, dry hot air is countercurrently exchanged with the material, and air selection and grading are realized. The multi-source sensing unit collects material, airflow and vibration operation parameters, the intelligent control unit performs online reasoning, cooperative adjustment and feedback updating on the vibration frequency of each layer, the flow of each dry hot air branch and the feeding rate based on time sequence prediction and multi-objective optimization, so that the drying uniformity, the separation stability and the energy consumption level under the condition of material fluctuation are improved. The application is suitable for continuous drying and separation of coal, minerals, chemical and solid waste powder materials.

Description

Technical Field

[0001] This invention relates to the field of powder material drying and sorting technology, specifically to a vertical air-classifying and drying integrated device and its intelligent adaptive control method, and more particularly to a vertical air-classifying and drying integrated device and its control method that achieves coordinated optimization of material drying, particle size classification and operating energy consumption based on multi-source time-series operating condition perception, drying-air-classifying process characteristic prediction, multi-objective optimization and actuator coordinated adjustment. Background Technology

[0002] Coal, minerals, chemical raw materials, solid waste, and other powdery or granular materials typically require dehydration, drying, particle size classification, or a combination of both during processing. The dryness and particle size distribution of the material not only affect subsequent conveying, shaping, storage, and processing, but also are closely related to system heat consumption, power consumption, and operational stability.

[0003] In existing continuous drying equipment, one type of equipment adopts a vertical structure and uses multi-layer bearing or vibration mechanisms to convey materials from top to bottom while simultaneously allowing hot airflow to flow from bottom to top, thus forming countercurrent heat exchange. Some equipment also has sieve holes installed on the bed surface carrying the material. By allowing airflow to pass through the material layer and sieve holes, particles of different sizes or suspension characteristics are classified to a certain extent, thereby achieving a combination of drying and sorting processes.

[0004] However, for equipment that simultaneously performs drying and air separation, there is a clear coupling relationship between the drying and separation processes. For example, changes in vibration frequency and amplitude affect the material's conveying speed, residence time, bed thickness, and particle stratification on the bed surface; the temperature and flow rate of the hot dry air affect both the heat and mass exchange between the material and the airflow, as well as the suspension, flow, and carrying state of fine particles; changes in the feed rate further alter the bed thickness, airflow resistance, heat exchange per unit material, and material residence time. Therefore, changing any one of these operating parameters often simultaneously causes changes in drying efficiency, separation efficiency, and system energy consumption.

[0005] Some existing equipment mainly uses pre-set fixed operating parameters, or the vibration frequency, hot air temperature, hot air flow rate, and feed rate are adjusted by the operator based on the feed status and output test results. When the feed moisture content, particle size distribution, throughput, or heat source conditions change, the fixed parameters are difficult to adapt to the fluctuations in operating conditions in a timely manner, while the adjustment method based on human experience is difficult to comprehensively consider the interaction between different operating parameters.

[0006] Furthermore, the drying-air separation process is continuous, time-varying, and exhibits a certain degree of lag. The current discharge state is not only related to the current control parameters but is also typically influenced by the feed characteristics, material layer condition, hot air condition, and vibration operation conditions in the previous time period. Therefore, controlling the process based solely on a single measurement at a particular moment is insufficient to accurately reflect the dynamic changes throughout the entire drying-air separation process.

[0007] For control methods that use data-driven models, if only the model is used to predict a single quality parameter without further establishing the corresponding control relationship between the prediction results and the multi-layer vibration actuator, the hot dry air regulating mechanism, and the feeding regulating mechanism, it will be difficult to solve the problem of coordinated adjustment between drying effect, sorting effect and energy consumption in vertical air classifier drying equipment.

[0008] Therefore, there is a need for a technical solution that can provide multi-source time-series sensing of material state, airflow state and equipment operating state based on the dynamic coupling characteristics of the vertical air classification and drying process, and further determine the vibration frequency of each vibrating bed, the flow rate of each dry hot air branch and the feed rate based on the process characteristic prediction results, so as to achieve closed-loop adaptive adjustment. Summary of the Invention

[0009] I. Purpose of the Invention The purpose of this invention is to provide a vertical air-separation drying integrated device and its intelligent adaptive control method, so as to improve the problem that when the properties of the feed material and the operating conditions of the equipment change, it is difficult to simultaneously take into account the drying effect, the separation effect and the system energy consumption by fixed parameter control or manual adjustment.

[0010] This invention establishes a multi-source sensing, time-series prediction, multi-objective optimization, and execution control system corresponding to the physical process of vertical air separation and drying. This enables the intelligent control unit to predict the discharge status and system energy consumption corresponding to different combinations of operating parameters based on actual working conditions, and to adjust the vibration frequency of different layers of vibrating beds, the flow rate of each dry hot air branch, and the feeding rate, thereby forming a closed-loop control based on feedback from actual discharge quality and operating status. Technical solution

[0011] To achieve the above objectives, the present invention provides a vertical air-separation and drying integrated device, including a vertical air-separation and drying host, a multi-source sensing unit, an execution and control unit, an intelligent control unit, and a cyclone dust removal unit.

[0012] The vertical air-classifying dryer has a vertical shaft structure, with an airlock feeder and exhaust port at the top and a hot dry air inlet assembly and an airlock unloader at the bottom.

[0013] The vertical air-classifying dryer has multiple layers of vibrating beds arranged from top to bottom. These vibrating beds are inclined and staggered, forming a zigzag material channel for continuous material transport from top to bottom. Each vibrating bed has grading sieve holes on its surface, and adjacent vibrating beds are connected by guide plates.

[0014] After the material to be processed enters the vertical air classifier dryer through the airlock feeder, it moves from top to bottom along the zigzag material channel under the action of gravity, vibration conveying, and the guiding action between adjacent vibrating beds. Dry hot air enters the vertical air classifier dryer through the dry hot air inlet and generally flows in the direction of bottom to top, forming a reverse contact with the material.

[0015] At least a portion of the dry hot air can pass through the classifying sieve holes of the vibrating bed and the material layer located on the vibrating bed to form cross-flow heat transfer; at the same time, part of the airflow flows between adjacent vibrating beds and along the zigzag material channel, thereby exchanging heat and mass with the material and carrying and classifying fine particulate materials.

[0016] In some implementations, the surfaces of each vibrating bed form an angle with the horizontal plane, and at least some adjacent vibrating beds have different grading sieve apertures.

[0017] Furthermore, the aperture of the grading screen can be gradually reduced along the material conveying direction from top to bottom, so that the material can be screened or air-separated layer by layer during the conveying process along the multi-layer vibrating bed.

[0018] The multi-source sensing unit is used to collect multi-source time-series operating condition data that can characterize the operating status of the drying-air separation process.

[0019] The multi-source time-series operating condition data includes at least the feed material property data, the discharge quality data obtained in the current control cycle and / or historical control cycles, the dry hot air status data, the exhaust gas status data, and the vibration status data of each layer of vibrating bed.

[0020] The feed physical property data may include feed moisture content and feed particle size distribution.

[0021] The discharge quality data obtained in the current control cycle and / or historical control cycles may include the discharge moisture content and discharge particle size distribution that have been detected in the corresponding control cycles, and are used together with other operating condition data to characterize the dynamic state of the current drying-air classification process.

[0022] The dry hot air status data may include one or more of the following: dry hot air temperature, humidity, flow rate, and pressure.

[0023] The exhaust gas status data may include one or more of the following: exhaust gas temperature, humidity, dust concentration, and particle size distribution.

[0024] The vibration state data may include the vibration frequency of each layer of the vibration bed, which can be obtained from the operating frequency signal of the strain frequency vibration drive device.

[0025] Furthermore, the multi-source sensing unit may also include a material layer thickness sensor disposed at at least one vibrating bed to collect material layer thickness data on the corresponding vibrating bed.

[0026] In a preferred embodiment, material layer thickness sensors can be installed on different layers of vibrating beds to obtain material layer state data distributed along the height direction of the vertical air classifier dryer.

[0027] The execution control unit includes a frequency conversion vibration drive device connected to each layer of the vibrating bed, an electric hot air regulating valve for adjusting the flow rate of each dry hot air branch, and a frequency conversion feeding device connected to the airlock feeder.

[0028] Each layer of the vibrating bed is preferably equipped with an independent variable frequency vibration drive device, so that the vibration frequency of different layers of the vibrating bed can be adjusted independently.

[0029] By changing the vibration frequency of different layers of vibrating beds, the conveying speed, residence time, material layer state, and particle movement state of the corresponding layers of materials can be altered.

[0030] The dry hot air intake assembly may include a main dry hot air inlet located at the bottom of the vertical air classifier dryer.

[0031] Furthermore, the hot dry air intake assembly may also include at least one auxiliary hot dry air inlet located in the lower part of the vertical air-separated drying host.

[0032] The main hot air inlet and the auxiliary hot air inlet can each be equipped with an electric hot air regulating valve and one or more of the following sensors: temperature sensor, humidity sensor, flow sensor, and pressure sensor. Each electric hot air regulating valve is used to adjust the flow rate of the corresponding hot air branch, allowing the flow rate of different hot air branches to be adjusted independently or collaboratively according to the drying-air separation status of the corresponding area.

[0033] In one embodiment, a hot air mixing chamber is disposed upstream of the dry hot air intake assembly. The hot air mixing chamber is equipped with a temperature sensor to detect the temperature of the mixed hot air. The hot air temperature is controlled by a proportional adjustment method on the heat source side. By coordinating the adjustment of the fuel supply, combustion air volume, and cold flue gas mixing volume, the hot air temperature at the outlet of the hot air mixing chamber is maintained within a preset temperature range. The temperature regulation of the hot air mixing chamber is a unified temperature control on the heat source side, while each dry hot air branch mainly adjusts its flow rate independently through a corresponding electric hot air regulating valve.

[0034] The intelligent control unit includes a deep learning adaptive control module.

[0035] The deep learning adaptive control module includes a drying-air separation process characteristic prediction model based on historical operating data and a multi-objective optimization algorithm that works in conjunction with the drying-air separation process characteristic prediction model.

[0036] The drying-air separation process characteristic prediction model is used to establish the time-series mapping relationship between the current and historical process states, candidate operating parameters, and the subsequent output quality and system energy consumption after the action of the candidate operating parameters.

[0037] Specifically, the feed physical property data obtained within the preset time window, the discharge quality data obtained in the current control cycle and / or historical control cycle, the material layer state data, the dry hot air state data, the exhaust gas state data, the vibration state data, and the corresponding candidate operating parameters can be used as model inputs. The discharge moisture content, discharge particle size distribution, and system energy consumption corresponding to the next control cycle or preset prediction period after the candidate operating parameters are applied can be used as model prediction outputs.

[0038] Therefore, the drying-air separation process characteristic prediction model is used to characterize the impact of current and historical operating conditions and candidate operating parameters on subsequent drying effect, separation effect and system energy consumption, rather than to repeatedly predict the output quality data already obtained in the current control cycle.

[0039] The system energy consumption in this invention can be expressed as the thermal energy consumption, electrical energy consumption, or the combined energy consumption calculated by a preset method within a certain time period of the device, or it can be expressed as the energy consumption per unit processing volume.

[0040] Since the drying-air separation process has time continuity and lag, in the preferred embodiment, multi-source time-series operating condition data are organized according to a preset time window to construct a time-series feature sequence that characterizes the current dynamic state of the drying-air separation process.

[0041] The preset time window may include the current control cycle and the operating condition data obtained from several consecutive control cycles or sampling times prior to it. The constructed time-series feature sequence is used to describe the current and historical process states before the implementation of candidate operating parameters. The drying-air separation process characteristic prediction model then predicts the discharge state and system energy consumption in subsequent control cycles or preset prediction periods after the implementation of candidate operating parameters based on the process state and candidate operating parameters.

[0042] The data preprocessing may include one or more of the following: outlier removal, missing data completion, normalization, and sliding time window feature construction.

[0043] The drying-air separation process characteristic prediction model can employ a deep learning network suitable for processing time-series data.

[0044] In some implementations, the prediction model employs an LSTM-attention network.

[0045] In other implementations, GRU-attention networks, Transformer networks, or other data-driven models capable of predicting the output state of the drying-air separation process based on historical time-series states can be employed.

[0046] The multi-objective optimization algorithm searches for and optimizes the set of operating parameters within the preset equipment operating constraints based on the prediction results of different candidate combinations of operating parameters from the real-time operating conditions and the characteristics prediction model of the drying-air separation process.

[0047] The optimization objectives involve at least the discharge moisture content, sorted particle size, and system energy consumption.

[0048] In one implementation, the discharge moisture content deviation, particle size sorting quality index, and unit processing capacity energy consumption index can be constructed separately, and the optimized operating parameter set can be determined by weighted, multi-objective Pareto optimization or other multi-objective optimization methods.

[0049] For example, the optimization objective can be expressed as: This reduces the deviation between the output moisture content and the target moisture content; This increases the proportion of materials that meet the target particle size requirements; and This reduces energy consumption per unit of processing volume.

[0050] Different weights can be set for each optimization objective, or a Pareto solution set can be used for comprehensive selection.

[0051] The candidate control parameters include at least the vibration frequency of different layers of vibrating beds, the flow rate of at least one dry hot air stream, and the feed rate.

[0052] When multiple dry hot air inlets are set, the candidate control parameters can also include the flow rate setpoints for each dry hot air branch.

[0053] The equipment operation constraints may include the allowable hot air temperature range, the internal temperature range of the main unit, the allowable frequency range of the vibrating equipment, the allowable air volume range of the fan, the processing capacity range of the feeding equipment, and other equipment safety operation restrictions.

[0054] The intelligent control unit constructs the current time-series operating condition state based on the multi-source time-series operating condition data collected within the current control cycle and the preset time window before it. The current time-series operating condition state is then input into the drying-air separation process characteristic prediction model in combination with different candidate operating parameters to predict the discharge moisture content, discharge particle size distribution, and system energy consumption corresponding to the next control cycle or preset prediction period after the action of each candidate operating parameter combination. Based on the above prediction results, the multi-objective optimization algorithm determines the optimized operating parameter set under preset equipment operating constraints.

[0055] After determining the optimized operating parameter set, the intelligent control unit sends the corresponding control commands to the execution and control unit.

[0056] Each layer of variable frequency vibration drive device adjusts the corresponding vibration bed according to the corresponding vibration frequency set value; Each electric hot air regulating valve adjusts the corresponding dry hot air branch according to the flow rate setting value of the corresponding dry hot air branch. The variable frequency feeding device adjusts the material flow rate entering the vertical air classifier dryer according to the feed rate setting value.

[0057] Through the above-mentioned coordinated regulation, the control parameters directly affect the material residence and conveying state, the gas-solid heat exchange state, and the particle air classification state, thereby forming a closed-loop control relationship between the algorithm calculation results and the actual industrial process.

[0058] After the optimized operating parameters are implemented, the multi-source sensing unit continues to collect the actual output quality data and device operating status data obtained after the corresponding control actions, and feeds them back to the intelligent control unit.

[0059] The intelligent control unit compares the actual obtained discharge moisture content, discharge particle size distribution and / or system energy consumption with the prediction results and target values ​​generated by the drying-air separation process characteristic prediction model for the corresponding control cycle, and uses the actual control results as part of the subsequent time-series operating conditions to determine the optimized operating parameters for the next control cycle, thereby forming a closed-loop adaptive control.

[0060] Furthermore, the deep learning adaptive control module may also include an online model update submodule.

[0061] The online model update submodule updates the drying-air separation process characteristic prediction model periodically or when preset trigger conditions are met, based on the new operating data obtained during the closed-loop control process.

[0062] The triggering conditions may include the deviation between the predicted output and the actual output exceeding a preset range, a significant change in the physical properties of the feed material, or the accumulation of a predetermined number of new operating samples.

[0063] Model updates can be performed using incremental learning to update all parameters of the model, or only some parameters can be updated, so that the prediction model can gradually adapt to changes in material properties, equipment status and external operating conditions.

[0064] The intelligent control unit may also include an anomaly warning and safety interlock module.

[0065] The abnormal warning and safety interlock module can determine abnormal operating conditions such as over-temperature, over-pressure, material blockage, and abnormal vibration based on temperature, pressure, material layer thickness, operating status of vibrating equipment, and other equipment status data.

[0066] When preset safety conditions are met, the control authority of the safety interlock module takes precedence over that of the intelligent adaptive control module, and it can suspend or restrict the current optimized control command.

[0067] Depending on the type of anomaly, one or more safety control actions may be performed, such as reducing the feed rate, reducing the flow rate of the corresponding dry hot air branch, reducing the temperature of the hot air carrier gas through the heat source side mixing system, cutting off the dry hot air, stopping at least part of the vibrating bed drive, or interlocking shutdown.

[0068] The air inlet of the cyclone dust removal unit is connected to the exhaust port of the vertical air classifier dryer, and is used to perform gas-solid separation on the fine particulate material carried in the exhaust gas.

[0069] The gas phase outlet of the cyclone dust removal unit can be connected to an induced draft fan, which will maintain a predetermined airflow state inside the vertical air classifier dryer.

[0070] The solid material obtained from cyclone separation can be fed into the material collection pipeline according to the specific production process.

[0071] In one embodiment, the solid material and the material discharged from the airlock unloader at the bottom of the vertical air classifier dryer enter the same subsequent processing system.

[0072] The present invention also provides an intelligent adaptive control method for a vertical air-separation drying integrated device, comprising the following steps: S1. Collect multi-source time-series operating data during the operation of the vertical air separation and drying device.

[0073] The multi-source time-series operating condition data includes at least the feed material property data, the discharge quality data obtained in the current control cycle and / or historical control cycles, the dry hot air status data, the exhaust gas status data, and the vibration status data of the multi-layer vibrating bed.

[0074] In some implementations, the thickness data of each layer or at least part of the vibrating bed is also included.

[0075] S2. Perform data preprocessing on the multi-source time-series operating data, and construct a time-series feature sequence to characterize the current drying-air separation process state according to a preset time window.

[0076] Data preprocessing may include outlier identification and removal, missing data completion, and normalization of data with different dimensions. The preset time window includes operating condition data obtained from the current control cycle and / or several consecutive control cycles or sampling times prior to it.

[0077] S3. Use historical operating data to train a predictive model for the characteristics of the drying-air separation process.

[0078] The model inputs are the feed condition data, equipment operation status data and corresponding operating parameters within a preset historical time window. The outputs are the discharge moisture content, discharge particle size distribution and system energy consumption obtained in the next control cycle or preset prediction period after the corresponding operating parameters are applied. The prediction model characterizes the coupled influence of the current and historical process status and changes in operating parameters on the subsequent drying effect, sorting effect and system energy consumption.

[0079] S4. Input the real-time obtained time-series feature sequence and different candidate operating parameter combinations into the trained drying-air separation process characteristic prediction model to predict the discharge moisture content, discharge particle size distribution and system energy consumption corresponding to the next control cycle or preset prediction period after the action of each candidate operating parameter combination.

[0080] S5. Using the discharge moisture content, target sorting particle size, and system energy consumption as optimization objectives, a multi-objective optimization algorithm is used to search for candidate combinations of operating parameters under equipment operating constraints to determine the optimal set of operating parameters.

[0081] S6. Adjust the vibration frequency of different layers of vibrating beds, the flow rate of at least one dry hot air stream, and the feed rate according to the optimized operating parameter set.

[0082] The above adjustments cause corresponding changes to the material conveying and residence state, gas-solid heat exchange state, and air classification state in the vertical air classifier drying device.

[0083] S7. Collect the output quality data and device operation status data after adjustment, and feed the data back to the intelligent control unit to form a closed-loop control of the drying-air separation process.

[0084] S8. Based on the new operating data generated by closed-loop control, the characteristic prediction model of the drying-air separation process is periodically updated or triggered.

[0085] Furthermore, the safe operating status of the device is continuously monitored during the execution of steps S1 to S8.

[0086] When the temperature, pressure, material layer thickness, or vibration equipment status are detected to reach the corresponding safety conditions, the corresponding intelligent optimization control can be stopped and preset safety control actions can be executed. Beneficial effects

[0087] Compared with control methods that use fixed operating parameters or rely mainly on manual adjustment, the present invention has at least the following advantages: This invention collects feed properties, output quality, hot air status, exhaust gas status, and vibration equipment operating status as interconnected multi-source time-series operating condition data, enabling the control system to not only obtain parameters at a single moment, but also to characterize the drying-air separation process status using dynamic change information within a certain time range.

[0088] This invention establishes a predictive model for the characteristics of the drying-air separation process, which is based on the control variables such as the vibration frequency of each vibrating bed, the flow rate of each dry hot air branch, and the feed rate, as well as the operating conditions such as the temperature of the hot load gas, and the coupling relationship between these variables and the material moisture content, particle size separation, and system energy consumption. This model can characterize the impact of different operating conditions and candidate control parameters on the subsequent output quality and system energy consumption.

[0089] This invention uses a multi-objective optimization algorithm to simultaneously consider the discharge moisture content, particle size sorting index, and system energy consumption. It further determines the vibration frequency of each vibrating bed, the flow rate of each dry hot air branch, and the feeding rate. The optimization results are then sent to the corresponding variable frequency vibration drive device, electric hot air regulating valve, and variable frequency feeding device, thereby converting the data-driven prediction and optimization results into coordinated control of the actual physical equipment.

[0090] Since the vibration frequency of different layers of vibrating beds can be adjusted independently, the flow rate of different dry hot air branches can be adjusted independently or collaboratively, and the feed rate can participate in joint optimization, the vibration frequency of each layer, the flow rate of each dry hot air branch and the feed rate can be differentiated and coordinated according to the material state and operating conditions of different spatial locations of the equipment, thereby improving the adaptability of the device to fluctuations in feed properties and changes in operating conditions.

[0091] This invention forms a closed-loop feedback by combining actual output quality and equipment operating status, and uses new operating data to periodically or trigger updates to the prediction model, which helps to reduce the impact of factors such as changes in material properties and equipment status on the model's prediction performance during long-term operation.

[0092] This invention sets up an anomaly warning and safety interlock mechanism, and sets up independent safety constraints and protection logic outside of intelligent optimization control, so that the intelligent control process can be carried out within the allowable operating range of the equipment, which is conducive to improving the safety and stability of continuous system operation. Attached Figure Description

[0093] Figure 1 This is a schematic diagram of the overall structure of the vertical air-separation and drying integrated device of the present invention; Figure 2 This is a schematic diagram of the functional units and control relationships of the vertical air-separation and drying integrated device of the present invention; Figure 3 This is a schematic diagram of the intelligent adaptive closed-loop control architecture of the present invention; Figure 4 This is a schematic diagram of the intelligent adaptive control method of the present invention.

[0094] Explanation of reference numerals in the attached figures: 1. Vertical air-separation drying unit; 2. Airlock feeder; 3. Exhaust port; 4. Airlock type unloader; 5. Vibrating bed; 6. Material guide plate; 7. Dry hot air inlet; 8. Cyclone dust removal unit; 9. Exhaust fan; 10. Variable frequency vibration drive device; 11. Electric hot air regulating valve; 12. Variable frequency feeding device. Detailed Implementation

[0095] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be noted that the following specific embodiments are used to illustrate the technical solution of the present invention and are not intended to limit the scope of protection of the present invention. Without departing from the technical concept of the present invention, those skilled in the art can adjust the relevant structural parameters and control parameters according to the specific material properties, processing capacity, and heat source conditions.

[0096] like Figure 1 and Figure 2 As shown, this embodiment provides a vertical air-classifying and drying integrated device, including a vertical air-classifying and drying host 1, a multi-source sensing unit, an execution and control unit, an intelligent control unit, and a cyclone dust removal unit 8.

[0097] The vertical air-classifying dryer 1 adopts a vertical shaft structure, and its cross-section can be rectangular, circular, or other cross-sectional forms suitable for arranging multiple vibrating beds 5.

[0098] This embodiment uses a rectangular steel structure shell, and a thermal insulation structure is set on the outside of the shell.

[0099] The vertical air classifier dryer 1 is equipped with an airlock feeder 2 and an exhaust port 3 at the top, and a dry hot air inlet assembly and an airlock unloader 4 at the bottom.

[0100] The airlock feeder 2 is used to reduce the possibility of a large amount of external air entering the main unit through the feeding position during continuous feeding, and to reduce the escape of hot air and dust from the feeding position.

[0101] The airlock discharger 4 is used to continuously discharge the dried material while maintaining the required airflow inside the main unit.

[0102] The vertical air-classifying dryer 1 has multiple vibrating beds 5 arranged from top to bottom inside.

[0103] In this embodiment, five layers of vibrating beds 5 are set up. The adjacent vibrating beds 5 are inclined in opposite directions and are arranged in an alternating manner so that the material falls from the discharge end of the upper layer of vibrating bed 5 into the feed end of the lower layer of vibrating bed 5 in sequence, thereby forming a zigzag material channel.

[0104] A guide plate 6 is installed between adjacent vibrating beds 5.

[0105] The guide plate 6 is used to limit the falling path of the material, so that the material discharged from the upper vibrating bed can stably enter the lower vibrating bed.

[0106] The vibrating bed 5 has multiple grading sieve holes on its surface.

[0107] In this embodiment, the screen aperture size of each vibrating bed 5 gradually decreases along the material's downward movement direction.

[0108] As an example, the sieve aperture of the five-layer vibrating bed 5 can be set to 20mm, 15mm, 10mm, 8mm and 6mm respectively.

[0109] The specific aperture mentioned above is only used to illustrate one implementation method. In actual design, it can be adjusted according to the particle size range of the material to be processed, the target classification particle size, and the airflow velocity.

[0110] Each layer of vibrating bed 5 is connected to a set of variable frequency vibration drive devices 10.

[0111] In this embodiment, a linear vibrating motor is used as the variable frequency vibration drive device 10, and the vibration drive of each layer of the vibrating bed is independent of each other. For example, the vibration frequency can be continuously adjusted within the range of 700 to 1000 times / min. The vibration frequency of different layers of the vibrating bed 5 can be set separately according to the location of each layer, the state of the material, and the overall optimization results. The amplitude of the vibrating bed is determined by the equipment structure and excitation parameters during the design stage and is not used as an online adjustment variable of the intelligent control unit during equipment operation.

[0112] The intelligent control unit can determine the corresponding vibration frequency setting value according to the location of each layer, the state of the material, and the overall optimization results.

[0113] During operation, by adjusting the vibration frequency of a certain layer of vibrating bed 5, the conveying state and residence time of the material through that layer can be changed; under the condition that other conditions remain basically unchanged, increasing or decreasing the vibration frequency can correspondingly change the motion state of the material in that layer.

[0114] Therefore, the vibration frequency of different layers can be used as a spatial distribution control variable that affects drying time, material layer state and particle classification behavior.

[0115] The dry hot air intake assembly includes a main dry hot air inlet 7 located at the bottom of the main unit.

[0116] In this embodiment, two auxiliary dry hot air inlets 7 are also provided in the lower part of the main unit.

[0117] Electric hot air regulating valves 11 are installed at the main hot air inlet and each auxiliary hot air inlet.

[0118] Different dry and hot air branches are also equipped with one or more of the following: temperature sensor, humidity sensor, flow sensor, and pressure sensor.

[0119] The flow rate of different dry hot air branches can be changed by each electric hot air regulating valve 11.

[0120] In this embodiment, the temperature of the hot-carrying gas is uniformly regulated by a mixing system located upstream of the dry hot air intake assembly on the heat source side. A hot-carrying gas mixing chamber is provided on the heat source side, and a temperature sensor is installed within the mixing chamber or at its outlet. By coordinating and adjusting the fuel supply, combustion air volume, and cold flue gas mixing volume, the temperature of the mixed hot-carrying gas is maintained within a target range. For example, the hot-carrying gas temperature can be controlled at approximately 200°C. When it is necessary to increase the total flow rate of the hot-carrying gas, the fuel supply and combustion air volume can be increased accordingly, and the cold flue gas mixing volume can be increased to maintain the temperature within the target range while increasing the flow rate. The flow rate of each dry hot air branch is independently regulated by a corresponding electrically operated hot air regulating valve 11.

[0121] After entering the main unit, the dry, hot air generally flows from bottom to top.

[0122] Part of the airflow passes through the material layer through the grading sieve holes on the vibrating bed 5, thereby forming a direct heat and mass exchange between the gas and the material; the other part of the airflow moves upward along the adjacent vibrating beds and the zigzag material channel.

[0123] In this process, the hot airflow provides the heat required for the evaporation of moisture to the wet material on the one hand, and on the other hand, it generates aerodynamic force on the fine particles.

[0124] Therefore, changes in the flow rate of each dry hot air branch affect the heat transfer intensity between the material and the airflow, as well as the suspension, flow through, and carry-over states of fine particles; the temperature of the hot carrier gas, as an important state parameter reflecting the operating conditions of the heat source, participates in the state characterization of the drying-air separation process. Based on this, the present invention uses the vibration frequency of each vibrating bed, the flow rate of each dry hot air branch, and the feed rate as collaborative control variables, and uses the temperature of the hot carrier gas as a state parameter in multi-source time-series operating data to participate in process prediction.

[0125] This embodiment adds a multi-source sensing unit to the device in Embodiment 1.

[0126] The multi-source sensing unit includes a material state sensing section, an airflow state sensing section, and a vibration operation state acquisition section.

[0127] An infeed moisture content detection device and an infeed particle size distribution detection device are installed upstream of the airlock feeder 2 to obtain the moisture content and particle size distribution of the material before it enters the equipment.

[0128] The moisture content detection device can be a microwave moisture content meter or other online moisture content detection equipment suitable for the materials being processed.

[0129] Particle size distribution can be obtained through online image particle size detection, laser particle size detection, or other online particle size detection methods suitable for continuous industrial materials.

[0130] A discharge moisture content detection device and a discharge particle size distribution detection device are installed downstream of the airlock unloader 4 to obtain actual discharge quality data.

[0131] At least some of the vibrating beds 5 are equipped with material layer thickness sensors.

[0132] For example, radar, laser, or other non-contact material layer thickness detection devices can be installed in the middle of each vibrating bed or at a representative location of the material state.

[0133] By measuring the thickness of the material layer at different heights, we can further determine the spatial distribution of the material at different levels inside the main unit.

[0134] Each dry and hot air branch can detect temperature, humidity, flow rate, and pressure.

[0135] The exhaust port 3 can detect exhaust gas temperature, humidity, dust concentration, and particle size distribution in the exhaust gas.

[0136] The vibration frequency of each layer of the vibrating bed (5) can be obtained from the operating frequency signal of the corresponding variable frequency vibration drive device (10) and input into the intelligent control unit as the operating status data of the vibrating bed. Since the amplitude of the vibrating bed (5) is determined by the equipment structure and excitation parameters during the design stage, and no online amplitude adjustment is performed during equipment operation, the amplitude is not used as a real-time acquisition parameter or online adjustment variable in this embodiment. If necessary, the current, power, bearing temperature or other equipment operating status parameters of the corresponding variable frequency vibration drive device (10) can also be acquired.

[0137] The multi-source sensing unit may also include an energy consumption acquisition module, used to collect electrical energy consumption data, thermal energy consumption data and / or energy consumption data per unit processing volume calculated based on the processing volume during the operation of the device.

[0138] All the above-mentioned material, airflow, vibration, equipment status and energy consumption data are correlated according to timestamps to form multi-source time-series operating condition data.

[0139] The intelligent control unit can obtain the above data through industrial communication networks, analog interfaces, or digital interfaces.

[0140] The control unit includes frequency conversion vibration drive devices 10 for each layer, electric hot air regulating valves 11 for each dry hot air branch, and frequency conversion feeding devices 12.

[0141] The intelligent control unit converts the optimization results into control setpoints for each actuator, thus forming: Vibration frequency settings for each layer of the vibrating bed (5); The flow rate setpoints for each dry hot air branch; and Feed rate setting.

[0142] The temperature of the heat-carrying gas is uniformly regulated by the mixing system on the heat source side and is not used as an independent control setting value for each dry hot air branch.

[0143] Thus, a definite control relationship is formed between the model's output variables and the specific industrial actuators.

[0144] like Figure 3 As shown, the intelligent control unit can adopt a structure that combines an industrial controller with an industrial computer or edge computing device.

[0145] Industrial controllers are mainly used for data acquisition, basic equipment control, and safety interlocking, while industrial computing devices are used to perform data processing, process prediction, and multi-objective optimization calculations.

[0146] The deep learning adaptive control module includes a data preprocessing submodule, an offline training submodule, an online inference submodule, and an online model update submodule.

[0147] The data preprocessing submodule first performs quality processing on the collected raw running data.

[0148] Measurement data that clearly exceeds reasonable physical limits can be identified and removed according to pre-defined data validity rules.

[0149] For short-term missing data, linear interpolation, neighbor-based data filling, or other data completion methods can be used.

[0150] Since different measurement variables have different dimensions and orders of magnitude, min-max normalization, standardization, or other data scaling methods can be used to transform the model input variables to a numerical range suitable for model training and prediction.

[0151] To address the time lag characteristic of the drying-air separation process, a time-series feature sequence is constructed using a sliding time window.

[0152] Let the current control period be t. The multi-source operating condition data obtained from the current control period t and several consecutive sampling periods before it are combined in time order to construct a time-series feature sequence that characterizes the dynamic state of the current drying-air separation process. The time-series feature sequence and candidate operating parameters are used as inputs to the drying-air separation process characteristic prediction model.

[0153] The drying-air separation process characteristic prediction model predicts the discharge moisture content, discharge particle size distribution, and system energy consumption in the next control cycle t+1 or the preset prediction period after the candidate operating parameters are implemented, based on the above inputs.

[0154] The inputs to the drying-air separation process characteristic prediction model include time-series operating condition inputs and candidate operating parameter inputs.

[0155] The time-series operating status input may include: feed moisture content, feed particle size distribution, discharge moisture content and discharge particle size distribution obtained in the current control cycle and / or historical control cycle, material layer thickness of different vibrating bed layers, temperature, humidity, flow rate and pressure of different dry hot air branches, exhaust gas temperature, exhaust gas humidity, exhaust gas dust concentration and exhaust gas particle size information, vibration frequency of different vibrating bed layers in current and / or historical operation, current and / or historical feed rate, current and / or historical system energy consumption, and other parameters that can characterize the current and historical operating status of the equipment.

[0156] The candidate operating parameter inputs may include: the vibration frequency to be adjusted for different layers of vibrating bed, the flow rate to be adjusted for different dry hot air branches, and the feed rate to be adjusted.

[0157] The drying-air separation process characteristic prediction model predicts the discharge moisture content, discharge particle size distribution, and system energy consumption in the next control cycle or preset prediction period after the implementation of the candidate operating parameters, based on the time-series operating condition input and candidate operating parameter input.

[0158] The model's predicted output should include at least: Output moisture content; The discharge particle size distribution or the particle size qualification index determined by the discharge particle size distribution; And system energy consumption or energy consumption per unit of processing volume.

[0159] This embodiment uses an LSTM-attention temporal prediction network.

[0160] LSTM is used to extract the features of the working condition data changing over time during the drying-air separation process, and the attention module is used to assign different weights to the information corresponding to different time positions or different features according to the prediction task.

[0161] The model may include an input layer, at least one LSTM layer, an attention layer, a fully connected layer, and an output layer.

[0162] For example, in a specific training configuration, two LSTM layers can be set up, each with 64 neurons, and the predicted values ​​of water content, particle size quality, and energy consumption can be output through fully connected layers.

[0163] The number of network layers and neurons mentioned above is an example implementation and does not constitute a limitation on other network structures.

[0164] In other implementations, GRU-attention networks or Transformer networks can be used to process multi-source time-series operational data.

[0165] The training data comes from historical operating samples generated by the device under different feeding conditions and different operating parameters.

[0166] Each training sample contains at least the operating condition data within a certain time window, the corresponding control parameters, and the actual output status and energy consumption data after the control parameters are applied.

[0167] Therefore, the predictive model can learn the mapping relationship from "operating condition + control parameters" to "output quality + energy consumption", rather than simply outputting an abstract classification result based on the current operating condition.

[0168] This model structure allows the prediction results to be directly used by subsequent multi-objective optimization modules to compare the impact of different candidate control parameters on the actual drying-air separation process.

[0169] After the prediction model is trained, the online inference submodule periodically obtains the current operating status of the device.

[0170] Multi-objective optimization algorithms generate or search for multiple candidate combinations of running parameters.

[0171] A candidate combination of runtime parameters can be represented as: U = { F,f 1 ,f 2 ,..,f n ,Q 1 ,Q 2 ,...,Q m} Where F represents the feed rate; f i This represents the vibration frequency of the i-th layer of the vibrating bed; Q j denoted by , where represents the flow rate of the j-th dry hot air channel; n represents the number of vibrating beds; and m represents the number of dry hot air branches whose flow rates can be independently adjusted.

[0172] The candidate operating parameter combination U and the current time-series operating condition are input into the drying-air separation process characteristic prediction model to obtain the predicted discharge moisture content, predicted particle size distribution and predicted system energy consumption in the next control cycle or preset prediction period after the corresponding candidate operating parameters are applied.

[0173] Multi-objective optimization can construct the following three evaluation directions: First, reduce the deviation between the predicted output moisture content and the target moisture content; Second, increase the proportion of material within the target particle size range or reduce the deviation between the predicted particle size distribution and the target particle size requirement; Third, reduce system energy consumption or energy consumption per unit of processing volume.

[0174] A comprehensive evaluation function can be established using a linear weighting method, or the Pareto optimal solution set can be obtained directly using methods such as multi-objective particle swarm optimization, NSGA-II, or decomposition-based multi-objective evolutionary algorithms.

[0175] When using a weighted approach, the weights of each target can be set according to actual production needs.

[0176] For example, when product moisture control requirements are high, the weight of the moisture content deviation target can be increased; when grading quality is the main target, the weight of the particle size quality target can be increased; when energy-saving operation requirements are high, the weight of the energy consumption target can be increased.

[0177] The multi-objective optimization process is simultaneously constrained by the safety operation boundary of the equipment.

[0178] For example: The hot air temperature must not exceed the allowable temperature of the corresponding equipment; The vibration frequency must not exceed the rated allowable range of the vibration mechanism; The hot air flow rate must not exceed the allowable range of the fan and piping; The feeding rate must not exceed the allowable processing capacity of the airlock feeder 2 and the main unit.

[0179] After optimization, the intelligent control unit selects and distributes the optimized set of operating parameters that meet the equipment constraints.

[0180] After the control command is issued, each layer of vibrating bed (5) can operate at different vibration frequencies, each dry hot air branch can use different flow rate settings, and the feed rate can be adjusted synchronously.

[0181] Since the three main execution variables mentioned above act on the material conveying time, gas-solid heat exchange and particle sorting state respectively, joint regulation can enable the drying target, sorting target and energy consumption target to be considered in the same physical process.

[0182] After the equipment executes optimized operating parameters, the discharge quality sensor obtains the actual discharge moisture content and particle size distribution, the energy consumption metering device obtains the actual operating energy consumption, and at the same time continues to collect equipment operating status data.

[0183] The intelligent control unit stores the model prediction results and the actual operation results in correspondence, forming new training samples.

[0184] Online model updates can be performed using a periodic update method.

[0185] For example, in one implementation, valid new samples obtained within the time window can be summarized every 24 hours, and the prediction model can be incrementally trained.

[0186] Alternatively, a 12-hour, 48-hour, or other update cycle suitable for the specific production process can be adopted.

[0187] Online model updates can also be implemented using a trigger-based approach.

[0188] For example, when the deviation between the predicted moisture content and the measured moisture content exceeds a preset threshold over multiple consecutive control cycles, a model update is initiated.

[0189] Model updates can train all network weights or freeze the temporal feature extraction part and only update some network parameters in the later part of the model.

[0190] For example, when using an LSTM-attention network, the already trained LSTM temporal feature extraction layer can be retained, and new running samples can be used to incrementally train the fully connected output layer or other parts of the network parameters with a smaller learning rate.

[0191] Through the above methods, the model can gradually adapt to changes in material properties, equipment wear, or changes in heat source conditions during long-term equipment operation.

[0192] like Figure 4 As shown in the figure, this embodiment provides an intelligent adaptive control method for a vertical air-separation drying integrated device.

[0193] S1 collects multi-source time-series operating condition data.

[0194] This includes feed moisture content, feed particle size distribution, discharge moisture content, discharge particle size distribution, thickness of different vibrating bed layers, temperature and flow rate of each dry hot air branch, exhaust gas status, vibration frequency of different layers, and feed rate.

[0195] S2, preprocesses the collected data.

[0196] We perform abnormal data processing, missing value imputation, and data normalization, and construct multi-source time series feature sequences using a sliding time window.

[0197] S3, train the predictive model for the characteristics of the drying-air separation process.

[0198] The historical time series operating conditions and corresponding control parameters are used as model inputs, and the actual discharge moisture content, particle size distribution and system energy consumption obtained under the corresponding control conditions are used as supervised training outputs.

[0199] S4 predicts the discharge status and energy consumption after the candidate operating parameters are applied.

[0200] The current time-series operating conditions and different candidate operating parameter combinations, consisting of the vibration frequency of different layers of vibrating beds, the flow rate of each dry hot air branch, and the feed rate, are input into the drying-air separation process characteristic prediction model to obtain the output quality and system energy consumption prediction results after the effect of each candidate operating parameter combination.

[0201] S5 performs multi-objective optimization.

[0202] Using moisture content, target particle size, and energy consumption as optimization objectives, an optimized set of operating parameters is selected within the constraints of equipment operation.

[0203] S6, execute parameter adjustment.

[0204] The corresponding set values ​​are sent to the vibration drive device 10, electric hot air regulating valve 11 and frequency conversion feeding device 12 of each layer.

[0205] S7 performs closed-loop feedback.

[0206] The system detects the actual output status and equipment operating status after the optimization parameters are applied, and feeds this information back to the intelligent control unit.

[0207] S8, Update the prediction model.

[0208] Based on a predetermined time period or a model prediction deviation trigger condition, the prediction model is incrementally updated using newly added valid data.

[0209] By executing S1 to S8 in a cycle, the control system can continuously re-determine the operating parameters based on changes in the material feed and equipment operating status.

[0210] This invention can be used for lignite, oil shale, mineral powder, iron concentrate powder, and other granular or powdery materials that require continuous drying and particle size classification.

[0211] Taking a type of water-containing powder as an example, when an increase in the moisture content of the feed is detected, the prediction model may predict an increase in the moisture content of the discharge while keeping other operating conditions constant among the candidate parameters.

[0212] The multi-objective optimization module can comprehensively compare candidate measures such as adjusting the flow rate of each dry hot air branch, adjusting the vibration frequency of different layers of vibrating bed, and adjusting the feed rate, based on the target moisture content, particle size requirements, and energy consumption constraints, rather than pre-fixing a certain adjustment measure.

[0213] Similarly, when the particle size distribution of the feed changes, the control system can re-optimize the vibration frequency of different layers of vibrating beds, the flow rate of each dry hot air branch, and the feed rate based on the feedback of the current discharge particle size distribution.

[0214] Therefore, this invention does not rely on a single sensing variable to pre-set a fixed "condition-action" rule, but instead uses a predictive model to estimate the impact of different candidate control parameters on the actual output state and energy consumption, and then controls the physical actuator based on the multi-objective optimization results.

[0215] The optimized control commands generated by the intelligent adaptive control module should be within the constraints of safe equipment operation.

[0216] In one implementation, the safety interlock module independently acquires the host temperature, airflow pressure, material layer thickness, and vibration equipment current or operating status.

[0217] When the detected value enters the warning zone, the intelligent optimization module can be restricted from further increasing the corresponding control quantity.

[0218] When the detected value reaches the shutdown protection condition, the safety interlock module obtains priority control authority.

[0219] For example, security procedures can be performed in the following order: First, reduce or stop the feed; Then reduce or cut off the corresponding hot and dry air; The vibrating bed and induced draft system should be kept running for a certain period of time based on the internal temperature and material condition of the equipment. Stop the relevant vibration equipment and ventilation equipment only after the safety conditions are met.

[0220] The specific safety thresholds and shutdown sequence are determined based on the actual equipment rated parameters and process requirements.

[0221] By combining independent safety interlocks with intelligent optimization control, the adaptive regulation capability can be preserved while preventing model optimization commands from exceeding the actual equipment safety boundaries.

[0222] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Based on the technical concept disclosed in this invention, any equivalent substitutions or conventional adjustments made by those skilled in the art to the number and structure of the vibrating bed, sensor type, time-series prediction model, multi-objective optimization algorithm, model update method, and corresponding operating parameters, as long as they do not depart from the technical concept of this invention formed through multi-source time-series working condition sensing, drying-air separation process characteristic prediction, multi-objective optimization, and closed-loop collaborative control of multiple actuators, should fall within the scope of the corresponding technical solutions of this invention.

Claims

1. A vertical air-separation drying integrated device, characterized in that, It includes a vertical air-separated drying host (1), a multi-source sensing unit, an execution and control unit, an intelligent control unit, and a cyclone dust removal unit (8). The vertical air classifier drying host (1) has a vertical shaft structure. It is equipped with an airlock feeder (2) and a tail gas exhaust port (3) at the top and a dry hot air intake assembly and an airlock unloader (4) at the bottom. The vertical air classifier drying host (1) is equipped with multiple vibrating beds (5) from top to bottom. The multiple vibrating beds (5) are inclined and staggered to form a zigzag material channel. The bed surface of each vibrating bed (5) is equipped with a grading screen. The adjacent two vibrating beds (5) are connected by a guide plate (6). After the dry hot air enters the vertical air classifier drying host (1) through the dry hot air inlet (7), it performs countercurrent heat exchange and air classification with the material conveyed from top to bottom along the zigzag material channel. The multi-source sensing unit is used to collect multi-source time-series operating condition data that characterize the operating status of the drying-air separation process. The multi-source time-series operating condition data includes at least the feed material property data, the discharge quality data obtained in the current control cycle and / or historical control cycle, the dry hot air status data, the exhaust gas status data, and the vibration status data of each layer of vibrating bed (5). The execution control unit includes a frequency conversion vibration drive device (10) connected to each layer of vibrating bed (5), an electric hot air regulating valve (11) for adjusting the flow rate of each dry hot air branch, and a frequency conversion feeding device (12) connected to the airlock feeder (2) to adjust the vibration frequency of each layer of vibrating bed (5), the flow rate of at least one dry hot air branch, and the feeding rate. The intelligent control unit includes a deep learning adaptive control module, which includes a drying-air separation process characteristic prediction model based on historical operating data and a multi-objective optimization algorithm that works in conjunction with the drying-air separation process characteristic prediction model. The drying-air separation process characteristic prediction model is used to predict the discharge moisture content, discharge particle size distribution and system energy consumption corresponding to the next control cycle or preset prediction period after the candidate operating parameters are applied, based on the multi-source time-series operating data and candidate operating parameters. The intelligent control unit is configured to determine an optimized operating parameter set based on real-time multi-source time-series operating condition data, with output moisture content, sorting particle size and system energy consumption as optimization objectives, under preset equipment operating constraints, through the multi-objective optimization algorithm. The optimized operating parameter set includes the vibration frequency of each layer of vibrating bed (5), the flow rate of at least one dry hot air and the feeding rate, and sends the optimized operating parameter set to the execution control unit for adjustment, and then forms a closed-loop feedback based on the adjusted output quality data and operating status data. The air inlet of the cyclone dust removal unit (8) is connected to the exhaust port (3) for gas-solid separation of fine particulate matter carried in the exhaust gas.

2. The vertical air-classification and drying integrated device according to claim 1, characterized in that, The deep learning adaptive control module includes a data preprocessing submodule, an offline training submodule, an online inference submodule, and a model online update submodule; The offline training submodule is used to train the drying-air separation process characteristic prediction model based on historical operating data, and the online inference submodule is used to determine the discharge status and energy consumption corresponding to different candidate operating parameters based on real-time operating data and the drying-air separation process characteristic prediction model.

3. The vertical air-separation and drying integrated device according to claim 2, characterized in that, The data preprocessing submodule is configured to perform outlier removal, missing data completion, and normalization on the multi-source time-series operating data, and construct a time-series feature sequence reflecting the dynamic state of the drying-air separation process according to a preset time window; the drying-air separation process characteristic prediction model adopts one of LSTM-attention network, GRU-attention network, or Transformer network.

4. The vertical air-classification and drying integrated device according to claim 1, characterized in that, The dry hot air intake assembly includes a main dry hot air inlet located at the bottom of the vertical air classifier dryer (1) and at least one auxiliary dry hot air inlet located in the lower middle part of the vertical air classifier dryer (1). Each dry hot air inlet is equipped with an electric hot air regulating valve (11) for independently regulating the flow rate of the corresponding dry hot air branch. A hot air mixing chamber is provided upstream of the dry hot air intake assembly. The hot air mixing chamber is equipped with a temperature sensor and maintains the temperature of the hot air output from the mixing chamber within a preset temperature range by coordinating and regulating the fuel supply, combustion air volume and cold flue gas mixing volume.

5. The vertical air-classification and drying integrated device according to claim 1, characterized in that, Each layer of vibrating bed (5) forms an inclination angle with the horizontal plane, and the aperture of the grading screen holes decreases layer by layer along the material conveying direction from top to bottom.

6. The vertical air-classification and drying integrated device according to claim 1, characterized in that, Each layer of the vibrating bed (5) is equipped with an independent variable frequency vibration drive device (10), which is a linear vibration motor, so that the vibration frequency of different layers of the vibrating bed (5) can be adjusted independently.

7. The vertical air-classification and drying integrated device according to claim 1, characterized in that, The feed physical property data includes at least the feed moisture content and feed particle size distribution; the discharge quality data includes at least the discharge moisture content and discharge particle size distribution; the dry hot air state data includes at least two of dry hot air temperature, humidity, flow rate and pressure; the exhaust gas state data includes at least two of exhaust gas temperature, exhaust gas humidity, exhaust gas dust concentration and exhaust gas particle size distribution; the vibration state data includes the vibration frequency of each layer of vibrating bed (5).

8. The vertical air-classification and drying integrated device according to claim 7, characterized in that, The multi-source sensing unit also includes a material layer thickness sensor disposed on the surface of at least one vibrating bed (5) for collecting material layer thickness data of the corresponding vibrating bed (5), and the material layer thickness data is used as one of the input data of the drying-air separation process characteristic prediction model.

9. The vertical air-classification and drying integrated device according to claim 2, characterized in that, The online model update submodule is configured to incrementally learn the drying-air separation process characteristic prediction model within a preset time window based on the output quality data and operating status data obtained after closed-loop control, and update at least some of the model parameters of the drying-air separation process characteristic prediction model.

10. The vertical air-classification and drying integrated device according to claim 1, characterized in that, The intelligent control unit also includes an anomaly warning and safety interlock module. The anomaly warning and safety interlock module determines abnormal operating conditions such as over-temperature, over-pressure, material blockage, and vibration based on one or more of the following: temperature, pressure, material layer status, and vibration equipment operating status. When the preset interlock conditions are met, the module controls the corresponding execution equipment to enter a safe operating state or executes an interlock shutdown.

11. An intelligent adaptive control method for a vertical air-separation drying integrated device, characterized in that, Includes the following steps: S1. Collect multi-source time-series operating condition data during the operation of the vertical air separation drying device. The multi-source time-series operating condition data includes at least the feed material property data, the discharge quality data obtained in the current control cycle and / or historical control cycles, the dry hot air status data, the exhaust gas status data, and the vibration status data of the multi-layer vibrating bed. S2. Perform data preprocessing on the multi-source time-series operating data, and construct a time-series feature sequence to characterize the current drying-air separation process state according to a preset time window; S3. Train a drying-air separation process characteristic prediction model using historical operating data. Use the feed condition data, equipment operating status data and corresponding operating parameters within a preset historical time window as model inputs. Use the discharge moisture content, discharge particle size distribution and system energy consumption obtained in the next control cycle or preset prediction period after the corresponding operating parameters are applied as model outputs. The drying-air separation process characteristic prediction model can characterize the coupled influence of the current process status and operating parameter changes on the subsequent drying effect, separation effect and system energy consumption. S4. Input the real-time obtained time-series feature sequence and different candidate operating parameter combinations into the trained drying-air separation process characteristic prediction model to predict the discharge moisture content, discharge particle size distribution and system energy consumption corresponding to the next control cycle or preset prediction period after the action of each candidate operating parameter combination. S5. Taking the discharge moisture content, target sorting particle size and system energy consumption as optimization objectives, under the preset equipment operation constraints, a multi-objective optimization algorithm is used to search for the candidate operating parameter combinations to obtain an optimized operating parameter set. S6. Adjust the vibration frequency of different layers of vibrating beds, the flow rate of at least one dry hot air stream, and the feeding rate according to the optimized operating parameter set, so as to change the material residence and conveying state, the heat exchange state between dry hot air and material, and the air classification state in the vertical air classifier drying device. S7. Collect the output quality data and device operation status data after adjustment, and feed the output quality data and device operation status data back to the intelligent control unit to form a closed-loop control of the drying-air separation process. S8. Based on the operating data generated by closed-loop control, the characteristic prediction model of the drying-air separation process is updated periodically or triggered.

12. The intelligent adaptive control method according to claim 11, characterized in that, In step S2, outlier removal, missing data completion, and normalization are performed on the multi-source time-series operating condition data, and the time-series feature sequence is constructed using a sliding time window; the drying-air separation process characteristic prediction model in step S3 adopts one of LSTM-attention network, GRU-attention network, or Transformer network.

13. The intelligent adaptive control method according to claim 11, characterized in that, The multi-objective optimization algorithm is one of the following: multi-objective particle swarm optimization algorithm, NSGA-II multi-objective genetic algorithm, or decomposition-based multi-objective evolutionary algorithm; the set of optimized operating parameters includes at least the vibration frequency of different layers of vibrating bed, the flow rate of at least one dry hot air stream, and the feed rate.

14. The intelligent adaptive control method according to claim 11, characterized in that, Step S8 includes: calculating the deviation between the discharge moisture content and / or discharge particle size distribution after closed-loop control and the corresponding target value, and incrementally learning and updating at least some model parameters of the drying-air separation process characteristic prediction model based on the effective operating data obtained within the preset time window.

15. The intelligent adaptive control method according to claim 11, characterized in that, During the execution of the intelligent adaptive control, the temperature, pressure, material layer thickness, and vibration equipment operating status of the device are continuously monitored. When the temperature, pressure, material layer thickness, or vibration equipment operating status reaches the corresponding preset safety conditions, the corresponding optimized control command is terminated, and at least one safety control action is executed, including reducing the feed rate, reducing or cutting off the dry hot air, stopping the vibration drive of at least part of the vibrating bed, and interlocking shutdown.

Images