A method for on-line detection of fine powder content in classified powder and evaluation of screening efficiency

Abstract

The present application relates to a kind of hierarchical screen fine powder content on-line detection and screening efficiency evaluation method, to solve the problem of existing hierarchical screen to fine powder content detection response lag and low control accuracy.Problems include: by automatic sampling and dynamic weighing to generate high representative powder sample, using high-frequency vibration screening to achieve accurate classification of powder particle size, and combining multiple source operating parameters to establish a self-adaptive correction model of classification efficiency, to dynamically evaluate the screening efficiency in real time, further output intelligent control guidance information based on efficiency deviation, through industrial field bus to drive actuator, to realize the dynamic optimization adjustment of air door opening degree and speed.The method can realize the rapid detection and efficient control of fine powder content of hierarchical screen, improve the classification accuracy, optimize system operation, reduce material loss, and realize the dynamic intelligent optimization of hierarchical screen system.

Description

Technical Field

[0001] This invention relates to the field of online detection and intelligent process control technology in industrial processes, and in particular to a method for online detection of the fine powder content and evaluation of screening efficiency in graded sieve return powder. Background Technology

[0002] Currently, in powder classification and grinding systems in industries such as cement, mining, and metallurgy, the detection of fine powder content in recycled powder and the evaluation of screening efficiency mainly rely on manual timed sampling and offline screening tests. This results in long testing cycles, poor sample representativeness, and severely delayed data feedback. Although process parameters such as air volume and classifier speed are collected by distributed control systems, they lack online coordination with fine powder content and efficiency evaluation, making it difficult to form a closed-loop control system. The few online detection technologies available are limited by sampling blind spots and environmental interference, and the evaluation models are mostly based on static empirical equations, lacking real-time correction and adaptive capabilities based on field data, leading to severe distortion in judgments under dynamic operating conditions.

[0003] Taking a closed-circuit cement grinding system as an example, the traditional process of manual sampling and manual screening is cumbersome, the testing is time-consuming, the sampling volume is limited and easily affected by human error, and it cannot capture instantaneous fluctuations in operating conditions; the delayed feedback causes efficiency assessment to lag behind actual production, parameter adjustment relies on experience for post-production correction, the grading sieve system deviates from the optimal operating conditions for a long time, energy consumption increases and product quality stability decreases.

[0004] The main drawbacks of existing technologies are: offline detection with poor real-time performance, disconnect between evaluation and process control, lack of fully closed-loop collaborative control capabilities, model drift due to lack of real-time correction mechanism, and waste due to the lack of material return. There is an urgent need to develop a high-frequency, real-time online detection method that combines dynamic efficiency evaluation and closed-loop control under material balance constraints, simultaneously addressing key issues such as representative sampling, rapid screening, precise weighing, data fusion, and automatic material return. Summary of the Invention

[0005] This application provides an online detection method for the fine powder content of graded sieve return powder and a method for evaluating sieve efficiency, aiming to solve the problems of slow response and low control accuracy of existing graded sieves in detecting fine powder content.

[0006] This application provides an online detection method for the fine powder content of graded sieve return powder and a method for evaluating sieve efficiency, specifically including: In the preset bypass detection loop of the graded sieve return powder pipeline, the automatic sampling device is used to extract powder samples at set time intervals, and the initial total mass data of the samples is obtained through the dynamic weighing module to generate a powder sample with instantaneous representativeness. The powder sample to be tested is subjected to rapid vibration sieving. The fine powder under the sieve and the coarse powder over the sieve are separated using a standard sieve. The amount of material under the sieve is weighed again. The real-time fine powder content index is calculated based on the difference between the initial total mass data and the amount of material under the sieve. The real-time fine powder content index is integrated with the system air volume, classifier speed and circulating load operating parameters collected from the distributed control system. Based on the material balance principle, a classification efficiency correction model including an adaptive coefficient of working condition is constructed to generate dynamic screening efficiency evaluation results. Based on the dynamic screening efficiency evaluation results, threshold determination and change rate limiting are performed. If the evaluation results deviate from the preset optimal range, intelligent control guidance information including damper opening adjustment and rotor speed setting is generated. Otherwise, the current control parameters remain unchanged. The intelligent control guidance information is output to the actuator via the fieldbus protocol, and the multi-flow control intelligent control strategy is coupled to adjust the opening of the secondary air damper and the auxiliary air volume of the classifier, so as to change the airflow swirl intensity and axial wind speed distribution.

[0007] After the screening test is completed, the self-circulation loop is controlled by a three-way reversing valve to transport the mixture of coarse powder on the screen and fine powder on the screen back to the designated position between the classifier and the mill in the main pipeline without damage, so as to complete the closed-loop collection of materials and avoid material waste.

[0008] The operating condition adaptive coefficient in the graded efficiency correction model is updated based on the system operation status feedback after returning to the main pipeline. The model parameters are then corrected for the real-time fine powder content index of the next detection cycle to achieve dynamic balance optimization of the detection-evaluation-control closed loop.

[0009] The online detection method for the fine powder content and sieving efficiency evaluation of graded sieve return powder provided in this application has the following beneficial effects: A bypass online detection loop is set up in the horizontal section of the grading and sieving powder pipeline. An automatic sampling device is used to periodically capture powder samples. An integrated detection unit performs dynamic weighing and rapid vibration sieving on the samples. The fine powder content index is calculated in real time using the difference method. On this basis, the online detection results are integrated with the process parameters collected by the distributed control system. A grading efficiency correction model with an adaptive coefficient is constructed based on the material balance principle. The sieving efficiency evaluation results are dynamically output. After threshold judgment and change rate limiting processing, control commands are generated and output to the actuator via fieldbus to adjust the opening of the secondary air damper and the auxiliary air volume to change the airflow swirl intensity and axial wind speed distribution. Meanwhile, the entire detection and control process is executed autonomously and cyclically by the controller. The detected material is returned to the main pipeline without damage via the conveying loop, maintaining the material balance of the system. At the same time, the modular structure supports the replacement of screen specifications as needed, realizing continuous monitoring and dynamic optimization of the grading screen's operating status. Attached Figure Description

[0010] Figure 1 This is the main flowchart of an online detection and sieving efficiency evaluation method for graded sieve return powder with fine powder content.

[0011] Figure 2 This is a sub-flowchart of a method for online detection of fine powder content and evaluation of sieving efficiency in graded sieved powder.

[0012] Figure 3 This is another sub-flowchart of a method for online detection of fine powder content and evaluation of sieving efficiency in graded sieve return powder. Detailed Implementation

[0013] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0014] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0015] like Figure 1 As shown, this application provides an online detection method for the fine powder content of graded sieve return powder and a method for evaluating sieve efficiency, specifically including: S1: In the preset bypass detection loop of the graded sieve return powder pipeline, the automatic sampling device is used to intercept powder samples at set time intervals, and the initial total mass data of the samples is obtained through the dynamic weighing module to generate a powder sample to be tested with instantaneous representativeness.

[0016] S2: Perform rapid vibration sieving on the powder sample to be tested, use a standard sieve to separate the fine powder under the sieve from the coarse powder over the sieve, and weigh the material under the sieve again. Calculate the real-time fine powder content index based on the difference between the initial total mass data and the material under the sieve.

[0017] S3: The real-time fine powder content index is integrated with the system air volume, classifier speed and circulating load operating parameters collected from the distributed control system. Based on the material balance principle, a classification efficiency correction model including an adaptive coefficient for operating conditions is constructed to generate dynamic screening efficiency evaluation results.

[0018] S4: Based on the dynamic screening efficiency evaluation results, perform threshold judgment and change rate limiting processing. If the evaluation results deviate from the preset optimal range, generate intelligent control guidance information including damper opening adjustment and rotor speed setting value; otherwise, maintain the current control parameters unchanged.

[0019] S5: The intelligent control guidance information is output to the actuator through the fieldbus protocol, and the multi-flow control intelligent control strategy is coupled to adjust the opening of the secondary air damper and the auxiliary air volume of the classifier, so as to change the airflow swirl intensity and axial wind speed distribution and thus optimize the classifier operating parameters.

[0020] S6: After the screening test is completed, the self-circulation loop is controlled by the three-way reversing valve to transport the mixture of coarse powder on the screen and fine powder on the screen back to the designated position between the classifier and the mill in the main pipeline without damage, so as to complete the closed-loop collection of materials and avoid material waste.

[0021] S7: Update the adaptive coefficient of the working condition in the graded efficiency correction model based on the system operation status feedback after returning to the main pipeline, and correct the model parameters for the real-time fine powder content index of the next detection cycle, so as to achieve dynamic balance optimization of the detection-evaluation-control closed loop.

[0022] Step S1: In the preset bypass detection loop of the graded sieve return powder pipeline, an automatic sampling device is used to extract powder samples at set time intervals, and the initial total mass data of the samples is obtained through a dynamic weighing module to generate a powder sample with instantaneous representativeness. Specifically, this includes: S1.1: Real-time monitoring of powder flow status in the graded sieve return powder pipeline, triggering an automatic sampling command based on a preset time interval to start the sampling execution mechanism of the bypass detection loop.

[0023] By installing microwave Doppler flow velocity sensors or differential pressure transmitters on the wall of the grading screen return powder pipeline, the instantaneous flow velocity signal and pressure fluctuation data of the gas-solid two-phase flow of powder in the main pipeline are collected in real time, which serve as the original input basis for judging the stability of material flow.

[0024] The preset bypass detection circuit is a branch flow path independent of the main pipeline of the classifier and the return powder. It draws out the sampling airflow from the side wall of the horizontal section of the main pipeline through the sampling probe and the three-way interface. After the detection unit completes the weighing and screening process, it returns to the designated position between the classifier and the mill in the main pipeline through the self-circulation loop, forming a closed bypass circulation structure.

[0025] The acquired raw flow velocity signal is processed by moving average filtering to eliminate high-frequency noise interference caused by pipe vibration or airflow turbulence, and a smooth flow velocity curve reflecting the macroscopic flow trend is extracted to ensure that subsequent triggering logic is based on stable operating conditions.

[0026] The smoothed flow velocity data is compared with the preset minimum stable flow velocity threshold. When the flow velocity of multiple consecutive sampling points is detected to be higher than the threshold, it is determined that the current pipeline is in a stable conveying state, and a valid flow status flag is generated to prevent invalid sampling from being performed under abnormal conditions such as pipe blockage or material interruption.

[0027] Provided that the valid flag of the flow status is true, the timestamp of the internal high-precision real-time clock module is read, the time difference between the current time and the last time the sampling command was successfully executed is calculated, and the actual running interval duration is obtained.

[0028] The actual running interval is logically compared with the standard sampling period parameter set by the system. If the actual interval is greater than or equal to the standard period, the time triggering condition is met, and an initial triggering request signal is generated.

[0029] Perform a logical AND operation on the initial trigger request signal and the current equipment readiness status signal of the bypass detection circuit to confirm that the automatic sampling device, dynamic weighing module and rapid screening unit are all in an idle and fault-free state, and eliminate equipment busy lock conflicts.

[0030] After confirming that the equipment is ready and meets both time and status conditions, the main control unit generates a standardized automatic sampling command pulse, which includes a sampling action identifier, a target sampling quantity code, and an execution priority flag.

[0031] The automatic sampling command pulse is sent to the PLC control station of the bypass detection circuit through the fieldbus communication interface, driving the pneumatic or spiral automatic sampling actuator to enter the pre-start stage.

[0032] Through the above-mentioned chain monitoring and logical judgment processing method, the pipeline flow status data of the previous step is transformed into an automatic sampling command with timing accuracy and equipment compatibility. This achieves the expected technical effect of accurately triggering sampling only when the operating conditions are stable and the equipment is available, thus avoiding the disturbance of system stability caused by invalid operations.

[0033] For example, a standard sampling period of 300 seconds is set, and the minimum stable flow velocity threshold is 15 m / s. A microwave Doppler flow velocity sensor collects data at a frequency of 10 Hz. After filtering with a 5-point moving average, if the flow velocity at 10 consecutive sampling points remains within the range of 18.5 m / s to 22.3 m / s, the flow state is considered valid. When the real-time clock shows that 305 seconds have passed since the last sampling, and the zero-point drift of the weighing sensor is detected to be less than 0.1 grams and the screening motor is not running, the logical AND operation result is true. The main control unit generates a hexadecimal instruction code 0xA5 and sends it to the PLC via the Modbus TCP protocol. After receiving the instruction, the PLC immediately drives the pneumatic three-way valve to switch to the sampling position and starts the screw feeder. This process ensures that sampling is performed during a specific window period when the material flow is stable and the equipment is idle, significantly improving the representativeness of the samples and the service life of the detection system.

[0034] S1.2: Using a pneumatic or spiral automatic sampling device to respond to an automatic sampling command, a powder sample with instantaneous representativeness is intercepted from the main pipeline airflow to generate a powder sample to be tested and transported to the integrated detection unit.

[0035] Receive the automatic sampling command generated by S1.1, parse the sampling timing parameters and target sampling threshold contained in the command, and activate the actuator drive circuit of the pneumatic or spiral automatic sampling device.

[0036] Based on the negative pressure distribution characteristics in the graded sieve return powder pipeline, the opening degree of the pneumatic sampling valve or the rotation speed of the spiral sampler is controlled to establish an instantaneous material flow channel from the main pipeline to the bypass detection circuit, ensuring that the interception action is synchronized with the flow of powder in the pipeline.

[0037] The high-speed airflow carrying powder particles is guided into the sampling probe by the Bernoulli principle. The pressure difference generated by the Venturi effect draws the instantaneously representative powder sample into the bypass pipe, forming a gas-solid two-phase mixed flow.

[0038] An inertial separation chamber is set up in the bypass pipeline. The powder particles are separated from the airflow and deposited in the collection hopper by the coupling effect of centrifugal force and gravity. At the same time, the carrier gas is discharged through the exhaust filter to maintain the system pressure balance.

[0039] The material accumulation height or weight signal in the hopper is monitored in real time. When the amount of sample collected reaches the preset representative quality threshold, the sampling valve is immediately closed or the screw conveyor is stopped to cut off the material source and fix the total sample amount.

[0040] The sealed conveying mechanism is activated, and the powder sample to be tested in the collection hopper is completely pushed to the feed port of the integrated detection unit by using short-time pulse airflow or spiral propulsion, so as to avoid sample residue or graded segregation during the transmission process.

[0041] Through the above processing method, the trigger signal of the previous step is transformed into a powder sample with a defined physical morphology and accurately transported to the weighing stage, achieving the expected technical effect of high representativeness and low interference in the sampling process.

[0042] For example, for a powder return pipeline with a diameter of 800 mm, the sampling interval is set to 5 minutes. The pneumatic sampling valve opens for 2 seconds, using a negative pressure of -5 kPa in the pipeline to extract the sample. The hopper capacity is set to 500 g, and the valve closes when the weighing sensor detects a mass of 495 g. Through inertial separation, more than 98% of the powder is retained. Subsequently, a 0.3 MPa compressed air pulse is used to blow the sample to the detection unit, with the transmission time controlled within 3 seconds to ensure no sample loss and to maintain the original particle size distribution, providing a stable input for the subsequent high-precision weighing in S1.3.

[0043] S1.3: Perform high-precision dynamic weighing on the powder sample to be tested entering the integrated detection unit. Use a weighing sensor to collect real-time weight signals and filter and denoise them to obtain the initial total mass data of the sample.

[0044] The powder sample to be tested is dropped into the load-bearing hopper inside the detection unit, triggering the strain gauge of the high-precision cantilever beam load cell to generate a micro-deformation electrical signal. This signal is then amplified by an instrumentation amplifier to a millivolt level to match the input range of the analog-to-digital converter.

[0045] The amplified analog voltage signal is subjected to high-frequency sampling processing. The original weight data sequence is continuously acquired using a 24-bit high-resolution ADC at a sampling frequency of 1kHz to construct an initial time series dataset containing environmental vibration noise and airflow disturbance interference.

[0046] The original weight data sequence is preprocessed using a moving average filtering algorithm. The window length is set to N, and the average of the original readings of the N collection points before the current time is used as the smoothed weight value at that time. The value range is set to 10 to 20 to balance the response speed and noise suppression effect.

[0047] Stability criterion analysis is performed on the smoothed weight data. The standard deviation of the most recent M sampling points is calculated. When the standard deviation is lower than the preset threshold and the rate of change of the mean approaches zero, the material feeding process is determined to have ended and reached a static equilibrium state.

[0048] The average weight under steady-state conditions is extracted as the final measurement value, and the data is corrected by combining it with a pre-calibrated temperature drift compensation coefficient to eliminate zero-point drift error caused by temperature fluctuations inside the detection unit.

[0049] By using the high-precision dynamic weighing and multi-level filtering methods described above, the powder sample generated in the previous step is transformed into initial total mass data with metrological validity, achieving high-precision mass quantification that resists environmental interference and providing a reliable benchmark for subsequent fine powder content calculation.

[0050] For example, the moving average filter window length N is set to 15, and the number of stability judgment sampling points M is set to 10. When a 100g powder sample falls into the hopper, the ADC collects data at a frequency of 1kHz. After amplification and filtering, the standard deviation of the most recent 10 points is calculated. If the standard deviation is less than 0.05g and the mean change rate is less than 0.01g / s, the current mean is locked at 99.82g. A temperature compensation coefficient of 0.002g / ℃ is introduced. If the ambient temperature deviates from the calibration temperature by 2℃, the corrected initial total mass is 99.86g. This processing method significantly improves the signal-to-noise ratio of the weighing data, ensuring that the mass measurement error is controlled within ±0.1g under strong vibration conditions, and greatly improving the accuracy of subsequent sieving efficiency evaluation data.

[0051] S1.4: Based on the initial total mass data, compare and verify with the preset sample representativeness threshold. If the data is valid, lock the initial total mass data as the standard reference value to generate a sample information package of the powder to be tested with metrological validity.

[0052] The system receives the initial total mass data of the sample after filtering and noise reduction, using it as the core input variable to be verified. It then calls the sample representative threshold range pre-stored in the embedded control unit's memory. This range is determined based on the typical material density of the graded sieve return powder pipeline and the volume calibration of the sampling device. Numerical comparison logic is executed to determine whether the initial total mass data falls between the preset minimum effective mass threshold and the maximum allowable overload threshold. If the initial total mass data is lower than the minimum effective mass threshold, it is determined as insufficient sampling, triggering an anomaly flag and discarding the current data to prevent statistical bias due to insufficient sample size. If the initial total mass data is higher than the maximum allowable overload threshold, it is determined as excessive sampling, triggering an overflow alarm and initiating a venting procedure to avoid exceeding the linear range of the weighing sensor and causing measurement distortion. When the initial total mass data is within the valid range, a data locking operation is performed, marking the value as a standard reference value with metrological validity. A data structure for the powder sample information packet is constructed, encapsulating the standard reference value, acquisition timestamp, device status identifier, and verification flag. The information packet is encoded using CRC cyclic redundancy check to ensure the integrity of the data transmission process. The packaged powder sample information packet is sent to the control interface of the rapid vibrating sieving device via the internal bus interface. Through the above-mentioned data validity verification and information encapsulation processing, the original weighing signal obtained in the previous step is transformed into a standardized powder sample information packet with metrological validity, realizing a precise conversion from physical sampling to a digitally ready state, and providing a reliable data benchmark for subsequent high-precision sieving.

[0053] For example, the representative threshold range for samples is set to a minimum effective mass of 50g and a maximum allowable overload of 500g. The initial total mass data collected by the weighing sensor and filtered is 210.5g. The system compares 210.5g with 50g and 500g to confirm that it is within the valid range. The system locks 210.5g as the standard reference value and generates an information packet containing this mass value, a timestamp of 2023-10-27 10:00:00, and the device ID. The CRC checksum is calculated and appended to the end of the packet. The information packet is transmitted to the screening controller via an RS485 bus. The controller receives and parses the information packet, and after confirming that the verification is correct, sets the ready flag and starts the vibrating feeding mechanism. This process ensures that only samples with qualified mass enter the screening stage, significantly improving the reliability of the test data and the stability of the system operation.

[0054] S1.5: Transmit the generated sample information packet of the powder to be tested to the control interface of the rapid vibrating sieve device to confirm the ready status of the sample and provide a clear input object for the next step of sieving.

[0055] Step S2: Perform rapid vibratory sieving on the powder sample to be tested. Use a standard sieve to separate the undersize fine powder from the oversize coarse powder, and weigh the undersize material again. Calculate the real-time fine powder content index based on the difference between the initial total mass data and the undersize material. Specifically, this includes: S2.1: Perform uniform material distribution on the powder sample to be tested. Use a vibrating feeding mechanism to spread the powder material corresponding to the initial total mass data in a single layer on the surface of a standard sieve to eliminate the material accumulation effect and generate a material layer to be screened with single particle size penetration conditions.

[0056] Receive the sample information packet of the powder to be tested from step S1.5, and extract the locked initial total mass data as the benchmark target value for fabric control.

[0057] The excitation frequency of the vibrating feeder is controlled to be in a low-frequency standby state, driving the electromagnetic vibrator to generate micro-amplitude oscillations, so that the powder material in the storage bin overcomes static friction and enters a flowing state.

[0058] By adjusting the throttling opening at the bottom of the feed trough, the powder flow rate is limited to a preset low flow rate range, preventing the material from impacting the screen in an agglomerated form and causing local accumulation.

[0059] Monitor the material distribution image or weight change rate on the surface of the standard screen. When the material layer thickness is detected to be uniform and without voids, gradually increase the excitation amplitude to the working threshold.

[0060] The powder on the sieve surface is loosened by using a high-frequency, low-amplitude vibration waveform, which breaks the van der Waals forces and electrostatic adsorption between particles, causing fine particles to separate from the gaps between coarse particles and settle down.

[0061] Continue the spreading process until a single layer of material to be screened with uniform thickness and porosity is formed on the screen surface, ensuring that each particle has an independent probability of passing through the screen.

[0062] By using the above-mentioned uniform material distribution method, the discrete powder sample to be tested is transformed into a standardized material layer with single particle size penetration conditions, eliminating the screening blockage effect caused by material accumulation, providing input conditions that conform to the geometric size separation principle for subsequent high-frequency vibrating screening, and significantly improving the accuracy and repeatability of fine powder content detection.

[0063] For example, for a sample of recycled cement raw meal, the initial total mass is set to 500g. The vibrating feeder is started at a low frequency of 2Hz, with the throttling opening set to 30%, and the material flows into the screen at a rate of 20g / s. When the infrared sensor detects that the screen coverage reaches 85%, the excitation frequency is increased to 50Hz, and the amplitude is adjusted to 0.5mm. After 15 seconds of loose spreading, a uniform material layer of about 2mm thickness is formed on the screen surface, with a porosity stabilizing at 0.45. In this state, fine powder particles with a diameter less than 45μm are fully exposed above the screen openings without agglomeration or obstruction, laying the foundation for efficient screening in the next step and effectively avoiding screening errors caused by uneven material distribution in traditional manual methods.

[0064] S2.2: High-frequency vibration screening is performed on the material layer to be screened. A vibrator with a preset frequency drives a standard screen to generate directional mechanical vibration, so that particles with a diameter smaller than the screen hole size pass through the screen to form a fine powder flow under the screen, while particles with a diameter larger than the screen hole size are retained to form a coarse powder pile on the screen, so as to complete the solid phase classification and separation based on geometric size.

[0065] The material layer to be screened is received after being uniformly distributed. This material layer is laid flat on the surface of a standard screen and has the ability to penetrate a single particle size.

[0066] The vibrator starts according to the preset control command, generates directional mechanical vibration waves with a specific frequency and amplitude, and transmits the vibration energy to the standard screen frame.

[0067] The standard sieve generates high-frequency reciprocating motion under the action of excitation force, which destroys the van der Waals forces and electrostatic adsorption between powder particles, causing the agglomerated particles to deagglomerate and enter a loose suspension state.

[0068] Fine powder particles with a diameter smaller than the sieve aperture size overcome the resistance of the sieve and pass through the sieve aperture under the combined action of gravity and vibration inertia, forming a downward-moving under-sieve fine powder flow.

[0069] Coarse powder particles with a diameter larger than the sieve aperture are trapped on the sieve surface due to geometric size limitations, and migrate towards the discharge port along the spiral trajectory of the sieve surface under the guidance of vibration, forming a pile of coarse powder on the sieve.

[0070] By monitoring the duration of vibration and the rate of change of the mass of the material under the screen, the screening endpoint is determined, ensuring that fine powder is fully screened and coarse powder is not excessively crushed, thus completing solid-phase classification and separation based on geometric dimensions.

[0071] By using high-frequency vibrating sieving, the material layer to be sieved in the previous step is transformed into a clearly separated undersize fine powder stream and oversize coarse powder pile, achieving the expected technical effect of efficient and accurate particle size classification.

[0072] For example, using a 45μm standard sieve, the excitation frequency is set to 50Hz and the amplitude to 2mm. The vibrator drives the sieve to generate a three-dimensional vibration field combining vertical and horizontal forces. Particles smaller than 45μm in the cement powder sample pass through the sieve rapidly within 0.5 seconds, forming a stable fine powder stream; particles larger than 45μm move spirally along the sieve surface to the outlet. After 120 seconds of continuous vibration, if the increase in the mass of the material under the sieve is less than 0.1g / min, the sieving is considered complete. This process effectively avoids the clogging phenomenon of traditional static sieving and significantly improves the fine powder separation efficiency and real-time detection.

[0073] S2.3: Perform dynamic weighing and acquisition processing on the separated undersize fine powder stream. Use a high-precision weighing sensor to monitor and record the cumulative mass value of the undersize fine powder stream falling into the collection container in real time, so as to obtain the mass data of undersize material that characterizes the total amount of fine powder.

[0074] For the undersize fine powder stream that falls into the collection container below after high-frequency vibrating screening, a high-precision weighing sensor integrated into the bottom of the container is activated to acquire the mass signal in real time. The vertical pressure is converted into a weak millivolt-level analog voltage signal using a strain gauge weighing module, which directly reflects the total mass of fine powder accumulated in the collection container at the current moment.

[0075] The acquired raw voltage analog signal is subjected to low-pass filtering and analog-to-digital conversion to eliminate mechanical noise interference caused by vibration transmission and convert it into a stable digital weight sequence to ensure that the signal-to-noise ratio of the data meets the measurement requirements during the dynamic material feeding process.

[0076] A time-mass accumulation curve is constructed based on a digital weight sequence. A sliding window algorithm is used to monitor the slope of the curve in real time. When the mass increment per unit time is detected to be lower than a preset threshold, it is determined that the flow of fine powder under the screen has basically ended, and a weighing lock command is triggered to freeze the current cumulative mass value.

[0077] Zero-point drift correction is performed on the locked cumulative mass value to deduct the tare weight of the collection container and the initial offset of the sensor, thereby obtaining the absolute mass data of the purified undersize fine powder and eliminating the reference error caused by long-term operation.

[0078] Through the above dynamic weighing acquisition and signal processing process, the physical form of the undersize fine powder flow is transformed into accurate undersize material quality data, providing a high-confidence quality benchmark for subsequent calculation of the fine powder content of the returned powder, and realizing the quantification and standardization of the test results.

[0079] For example, a cantilever beam load cell with a range of 5 kg and an accuracy of 0.01 g is selected, and the sampling frequency is set to 100 Hz. During the sieving process, the sensor outputs a 0-10 mV voltage signal in real time, which is converted into a digital value by a 24-bit ADC. A Butterworth low-pass filter with a cutoff frequency of 5 Hz is used to filter out 50 Hz power frequency interference and high-frequency vibration noise. When the mass increment is less than 0.05 g within one second, the sieving is considered complete, and the final reading is locked at 125.30 g. After deducting the tare weight of 20.00 g, the net weight of the undersized fine powder is 105.30 g. This data is directly used for efficiency calculation in step S2.4, ensuring the accuracy of the fine powder content calculation and significantly improving the repeatability accuracy of the online detection system.

[0080] S2.4: Perform difference calculation on the initial total mass data and the undersize mass data, use the embedded computing unit to call the preset mass balance algorithm to calculate the difference between the two to obtain the mass of coarse powder on the screen, and further perform normalization ratio calculation to generate a real-time fine powder content index that characterizes the current working condition.

[0081] The valid undersize material mass data collected and verified in step S2.3 is acquired and input as the minuend into the arithmetic logic unit of the embedded computing unit. The initial total mass data pre-stored in non-volatile memory, which has been locked as the standard reference value for the current detection cycle in step S1.4, is loaded as the subtrahend into the arithmetic register. A high-precision floating-point subtraction operation is performed, and the difference between the initial total mass and the undersize material mass is calculated according to the law of conservation of mass, yielding the oversize coarse powder mass value representing the total amount of particles that did not pass through the sieve. A non-negativity constraint is applied to the calculated oversize coarse powder mass. If the result is negative, an abnormal alarm is triggered and the current cycle data is discarded; if the result is valid, the normalization process begins. A ratio normalization algorithm is used to divide the undersize material mass data by the initial total mass data to construct a dimensionless index reflecting the proportion of fine powder in the sample. The real-time fine powder content index is calculated using the following formula: in, This is a real-time indicator of the fine powder content in the returned powder. For the data on the amount of material passing through the sieve, The initial total mass data is used. The calculated percentage values ​​are rounded to two decimal places to match the accuracy requirements of the subsequent efficiency evaluation model. The rounded real-time fine powder content index is encapsulated into a standard data frame, with a timestamp and equipment status identifier added. Through the above difference calculation and normalized ratio calculation methods, the discrete mass data from the previous step is transformed into a real-time fine powder content index characterizing the current operating condition, achieving a quantitative conversion from physical quantity measurement to process parameter characterization, providing a highly timely input variable for the dynamic efficiency evaluation in step S3.

[0082] For example, the embedded computing unit reads the undersize material mass data collected in step S2.3 as 18.45g and calls the initial total mass data locked in step S1.4 as 50.00g. Subtraction is performed to obtain the coarse powder mass on the sieve as 31.55g, which is confirmed as valid after non-negativity verification. Substituting into the formula, 18.45 divided by 50.00 yields 0.369, which, multiplied by 100%, gives 36.90%. This value is rounded to two decimal places, ultimately generating a real-time fine powder content index of 36.90%. This index is encapsulated into a data frame containing a timestamp of 2023-10-27 10:00:00 and the device ID. This process converts the physical weighing signal into a specific process content index, significantly improving the real-time performance and accuracy of data processing, eliminating manual calculation errors, and ensuring that subsequent efficiency evaluation models can be dynamically corrected based on accurate fine powder content data.

[0083] Step S3: The real-time fine powder content index is integrated with the system air volume, classifier speed, and circulating load operating parameters collected from the distributed control system. Based on the material balance principle, a classification efficiency correction model including an adaptive coefficient is constructed to generate dynamic screening efficiency evaluation results. Specifically, this includes: S3.1: Perform timestamp alignment and outlier removal on the system air volume, classifier speed and circulating load operating parameters collected from the distributed control system to generate a standardized multi-source operating parameter sequence, ensuring the consistency of the data benchmark for subsequent fusion calculations.

[0084] S3.2: Based on the standardized multi-source operating parameter sequence and the real-time fine powder content index, the material balance principle is used to calculate the difference in material particle size distribution at the inlet and outlet of the classifier, so as to generate a static basic efficiency intermediate variable that characterizes the current separation state.

[0085] Receives a multi-source operating parameter sequence that has been standardized by S3.1, including system air volume Q, classifier speed N, and circulating load L, and obtains the real-time fine powder content index C of the returned powder output by S2. fine .

[0086] Based on the principle of material balance, a mass conservation equation for the particle size distribution at the inlet and outlet of the classifier is constructed, and a theoretical constraint relationship between the coarse powder recovery rate and the fine powder penetration rate is established.

[0087] Utilizing real-time return powder fine powder content C fine As the mass fraction of fine powder in the return flow, the absolute throughput of fine powder in the total material entering the classifier is calculated in conjunction with the circulating load L.

[0088] Using Newton's order of efficiency, the separation process is considered as a superposition of ideal separation and non-ideal mixing. The intermediate variable E of the static basic efficiency under the current operating condition is calculated.static .

[0089] The specific calculation formula is as follows: in, This is an intermediate variable for static basic efficiency. The fine powder content of the feed at the mill outlet was estimated using a moving average of historical data. Cyclic load ratio; This refers to the content of fine powder in the recycled powder, as detected online. The formula is an engineering approximation model based on the principle of material balance.

[0090] The calculated E static Perform physical boundary checks to remove outlier values ​​that fall outside the range of 0 to 1, ensuring the validity of the physical meaning of intermediate variables.

[0091] Verified E static The system is correlated with the system air volume Q and rotational speed N to identify the efficiency deviation benchmark value caused by airflow short circuit or mechanical wear.

[0092] By using the above-mentioned material balance derivation and boundary verification methods, the discrete detection data from the previous step is transformed into a static basic efficiency intermediate variable that characterizes the current separation state, thus achieving the expected technical effect of providing high-confidence benchmark data for the subsequent dynamic correction model.

[0093] For example, the system airflow Q is set to 50000 m³ / h, the classifier speed N to 1200 rpm, and the circulating load L to 2.5. The fine powder content C of the returned powder is obtained through online detection. fine The estimated fine powder content (C) in the mill outlet feed is 15%. feed The value is 40%. Substituting into the formula: the numerator is 0.15 × 2.5 = 0.375, and the denominator is 0.40 × (1 + 2.5) = 1.4. The intermediate variable E of the static basic efficiency is calculated. static The value is 0.268. This value is within the valid range of 0 to 1 and is directly passed as input to step S3.3 for the identification of the adaptive coefficient of the working condition, which significantly improves the initial accuracy of the efficiency assessment.

[0094] S3.3: Based on the historical trends of multiple sets of recent online detection data, the parameters of the preset empirical and semi-empirical equations are identified using the recursive least squares algorithm to generate adaptive coefficients for compensating for equipment wear and material property fluctuations.

[0095] The system receives static baseline efficiency intermediate variables and historical data sequences from multiple recent online detections, and constructs a recursive least squares parameter identification framework that includes a time decay factor. It initializes the covariance matrix and parameter vectors, and sets a forgetting factor to assign higher weights to recent data, ensuring the model's rapid response to sudden changes in operating conditions.

[0096] A regression vector for an empirical or semi-empirical equation is constructed, with classifier speed, system airflow, and circulating load as independent variables, and real-time fine powder content as the dependent variable, to establish a linearized parameter estimation model. The output value at the current moment is predicted using the parameter estimates from the previous moment, and the residual error between the predicted and actual measured values ​​is calculated.

[0097] The gain vector is calculated based on the residual error. This gain vector is used to measure the correction weight of information on parameter estimation. The covariance matrix is ​​then adjusted based on this gain vector to reflect the changes in the uncertainty of parameter estimation. The forgetting factor λ is used to control the decay rate of historical data.

[0098] The parameter estimation vector is corrected using the gain coefficient, and the residual error is weighted and superimposed onto the old parameter values ​​to generate the latest adaptive coefficients for operating conditions, which include equipment wear compensation characteristics. The covariance matrix is ​​then updated to complete the iterative loop closure, as shown in the following formula: in, It is the identity matrix. The updated covariance matrix, Forgetting factor, For the first Gain vector at time step For the first The regression vector at time step, This is the transpose of the regression vector. For the first The covariance matrix at time t, This is an index for discrete sampling times.

[0099] By dynamically iterating through the recursive least squares algorithm, the historical detection data trend is transformed into a real-time updated adaptive coefficient for operating conditions, enabling the graded efficiency model to automatically compensate for fluctuations in material properties and drift in equipment status, thus significantly improving the dynamic accuracy of the evaluation results.

[0100] For example, the forgetting factor λ is set to 0.98, and the initial covariance matrix P(0) is 1000 times the identity matrix. When the classifier speed suddenly changes from 800 rpm to 850 rpm, the regression vector φ(k) is updated in real time. If the residual between the measured fine powder content and the predicted value is 0.02, the correction amplitude is adaptively amplified after weighting by the gain vector. The adaptive coefficient of the working condition converges to a new steady-state value of 0.95 within 3 detection cycles, effectively eliminating the model deviation caused by the speed adjustment, and reducing the dynamic screening efficiency evaluation error to within 0.5%.

[0101] S3.4: Input the intermediate variables of static basic efficiency and the adaptive coefficients of operating conditions into the graded efficiency correction model to perform weighted mapping operation, so as to generate dynamic screening efficiency evaluation results that eliminate operating condition drift error.

[0102] The static basic efficiency intermediate variables generated by S3.2 and the operating condition adaptive coefficients identified by S3.3 are received as the input data source for the graded efficiency correction model.

[0103] The grading efficiency correction model is a dynamic evaluation model built on the principle of material balance. Its inputs are a standardized sequence of multi-source operating parameters such as system air volume, classifier speed, and circulating load, as well as the real-time fine powder content index of the returned powder. The internal architecture includes a static basic efficiency intermediate variable calculation module derived from the material balance equation and an operating condition adaptive coefficient module generated online through a recursive least squares algorithm. After weighted mapping calculation, the output is a dynamic screening efficiency evaluation result that can eliminate the influence of equipment wear and material characteristic fluctuations. The role of this model is to integrate offline detection data with process control parameters in real time, realize the accurate quantification of the actual separation performance of the grading screen, and provide a reliable basis for subsequent intelligent control.

[0104] The specific construction process of the graded efficiency correction model is as follows: call the pre-set graded efficiency correction algorithm, load the empirical-semi-empirical equation structure containing nonlinear compensation terms, and prepare to perform weighted mapping operation.

[0105] Substitute the intermediate variable of static basic efficiency into the main function of the equation to calculate the theoretical screening efficiency benchmark value under the current ideal operating condition assumption.

[0106] The equipment wear compensation factor and material characteristic fluctuation factor in the adaptive coefficient of working conditions are extracted, which correspond to the linear correction term and nonlinear correction term in the equation, respectively.

[0107] The multiplicative coupling method is adopted to multiply the equipment wear compensation factor with the theoretical screening efficiency benchmark value one by one to offset the efficiency reduction deviation caused by screen blockage or rotor blade wear.

[0108] A material property fluctuation factor is introduced to perform additive bias correction on the multiplication result to compensate for the particle size distribution drift error caused by changes in the moisture content or grindability of the material entering the mill, and finally obtain the dynamic screening efficiency evaluation result.

[0109] Perform a numerical rationality check on the calculated dynamic screening efficiency evaluation results to determine whether they are within the physical effective range of 0 to 100%.

[0110] If the value exceeds the valid range, the boundary limiting logic is triggered to clamp it to the preset minimum efficiency threshold or maximum efficiency threshold to prevent abnormal data from diverging.

[0111] The dynamic screening efficiency evaluation results after verification are packaged into a standard floating-point number format and a timestamp is added to identify the current evaluation period.

[0112] Through the weighted mapping operation described above, the static efficiency variables and adaptive coefficients from the previous step are transformed into dynamic screening efficiency evaluation results that eliminate working condition drift errors, thereby achieving accurate quantitative evaluation of the grading performance under various working conditions.

[0113] For example, suppose the intermediate variable of static basic efficiency is 78.5%, the equipment wear compensation factor identified by the recursive least squares algorithm is 0.96, and the material characteristic fluctuation factor is -1.2%. After wear compensation, the efficiency is corrected to 75.36%. Then, the material characteristic bias correction is added, and the final output dynamic screening efficiency evaluation result is 74.16%. After verification, this value is within the valid range.

[0114] Step S4: Based on the dynamic screening efficiency evaluation results, perform threshold determination and change rate limiting. If the evaluation results deviate from the preset optimal range, generate intelligent control guidance information including damper opening adjustment and rotor speed setpoint; otherwise, maintain the current control parameters unchanged. Specifically, this includes: S4.1: Perform a preset optimal interval comparison process on the dynamic screening efficiency evaluation results to generate an efficiency deviation status flag. The input condition is the dynamic screening efficiency evaluation results calculated in real time. The technical processing method is to logically compare its value with the pre-calibrated upper and lower efficiency thresholds. The output target is an efficiency deviation status flag that indicates whether the current working condition is within the normal range.

[0115] The system receives the dynamic screening efficiency evaluation result from S3.4 as a real-time performance quantification indicator under the current operating conditions. It calls the optimal efficiency range parameter set pre-stored in the controller's non-volatile memory. This parameter set includes upper and lower efficiency thresholds calibrated based on historical best operating data. A greater than or equal to logic comparison operation is performed between the real-time dynamic screening efficiency evaluation result and the upper efficiency threshold, generating a high-order out-of-bounds flag. A less than or equal to logic comparison operation is performed between the real-time dynamic screening efficiency evaluation result and the lower efficiency threshold, generating a low-order out-of-bounds flag. A logical OR operation is performed on the high-order and low-order out-of-bounds flags to comprehensively determine whether the current efficiency value deviates from the preset optimal control range. If the logical OR operation result is true, the efficiency deviation status flag is set to an abnormal state, and the current deviation direction is recorded. If the logical OR operation result is false, the efficiency deviation status flag is reset to a normal state, indicating that the current classifier operating parameters do not require adjustment. By using the threshold comparison and logical judgment processing methods described above, the continuously changing dynamic screening efficiency evaluation results are transformed into discrete efficiency deviation status indicators, realizing real-time binary monitoring of the operating status of the grading system and providing clear triggering conditions for subsequent rate-of-change limiting filtering and intelligent control command generation.

[0116] For example, the optimal screening efficiency range of the grading screen system is set to [85%, 92%], i.e., the upper efficiency threshold is 0.92 and the lower efficiency threshold is 0.85. When the dynamic screening efficiency evaluation result calculated by S3.4 is 0.83, it is compared with 0.92, and the high-order out-of-bounds flag is 0; it is compared with 0.85, and since 0.83 is less than 0.85, the low-order out-of-bounds flag is 1. A logical OR operation is performed, 0OR1=1, the result is determined to deviate from the optimal range, the efficiency deviation status flag is set to "too low", and the deviation direction is marked as negative. When the dynamic screening efficiency evaluation result is 0.88, it is compared with 0.92 and compared with 0.85, both result in 0, the logical OR operation result is 0, and the efficiency deviation status flag is reset to "normal". This process ensures that subsequent complex control calculations are only initiated when the efficiency significantly deviates from the set range, avoiding frequent actions of the actuator and significantly improving the stability of the system.

[0117] S4.2: Perform rate-limited filtering based on efficiency deviation status flags to generate smoothed efficiency deviation values. The input conditions are efficiency deviation status flags that are determined to deviate from the preset optimal range and their corresponding historical sequence data. The technical processing method is to use a first-order inertial filtering algorithm to limit the slope and suppress noise of efficiency mutation signals. The output target is a smoothed efficiency deviation value that eliminates instantaneous interference and reflects the true trend.

[0118] The system receives efficiency deviation status flags and their corresponding historical sequence data that are determined to deviate from the preset optimal range, serving as input for rate-of-change limiting filtering. It extracts the dynamic screening efficiency evaluation results of the current detection cycle and the evaluation results of the previous detection cycle, calculating the difference between the two to obtain the instantaneous efficiency change. It then calls preset first-order inertial filtering algorithm parameters, including the filtering time constant and sampling period, to construct a discretized recursive filtering model.

[0119] A first-order inertial filtering algorithm is used to perform a weighted summation of the real-time efficiency deviation value of the current cycle and the filtered output value of the previous cycle. The filtering coefficient α ranges from 0 to 1 and is used to control the degree of influence of historical data on the current smoothing result.

[0120] The filtering coefficient α is set according to the response characteristics of the classifier's mechanical actuator. If the actuator's response is slow, the value of α is decreased to enhance the smoothing effect; if a fast response is required, the value of α is increased. The calculated smoothed efficiency deviation value is compared with a preset rate of change threshold. If the absolute rate of change of the deviation value exceeds the threshold, it is forcibly limited to the threshold range to suppress efficiency spikes caused by sensor noise or instantaneous material fluctuations. The limited efficiency deviation value undergoes sign preservation processing to ensure that the control direction is consistent with the efficiency deviation direction, generating the final smoothed efficiency deviation value used for table lookup mapping. Through a dual processing method of first-order inertial filtering and slope limiting, the efficiency deviation status flag from the previous step is transformed into a smoothed efficiency deviation value that eliminates instantaneous interference and reflects the true trend, achieving the expected technical effect of suppressing control oscillations and improving system stability.

[0121] For example, the filter coefficient α is set to 0.3, the measured screening efficiency in the current cycle is 82%, the smoothed efficiency in the previous cycle is 85%, and the preset optimal efficiency range center value is 88%. The current efficiency deviation is calculated to be -6%, and the deviation in the previous cycle is -3%. Substituting these values ​​into the formula, the smoothed efficiency deviation is calculated to be -4.5%. If the preset change rate threshold is 2% per minute, and the original deviation change rate is 3%, the limiting mechanism is triggered, correcting the output deviation to -4%. This processing effectively filters out the false impressions of high or low efficiency caused by airflow disturbances at the moment of sampling, making the subsequent damper opening adjustment more stable, significantly improving the continuity of the classifier rotor speed adjustment, and avoiding mechanical wear caused by frequent actuator movements.

[0122] S4.3: The smoothed efficiency deviation value is mapped to a multivariable decoupled control model by looking up a table to generate the initial damper opening adjustment and the initial rotor speed setpoint. The input condition is the smoothed efficiency deviation value after filtering. The technical processing method is to call the built-in classifier airflow-speed decoupling mapping matrix to perform nonlinear function calculation. The output targets are the initial damper opening adjustment and the initial rotor speed setpoint corresponding to the secondary air conditioning and speed conditioning, respectively.

[0123] The smoothing efficiency deviation value processed by S4.2 is received and used as the input variable of the multivariable decoupled control model to start the calculation process of coordinated regulation of airflow and speed of the classifier.

[0124] The classifier airflow-speed decoupling mapping matrix, which is pre-set in the embedded controller, is invoked. This matrix is ​​constructed based on historical best operating condition data and characterizes the coupling relationship between the change in the secondary air damper opening and the change in rotor speed on the screening efficiency.

[0125] By substituting the smoothing efficiency deviation value into the nonlinear decoupling function, the independent damper adjustment component and speed adjustment component are separated by inverse matrix operation, thus eliminating the mutual interference effect between the two in the physical execution process.

[0126] The smoothing efficiency deviation and its rate of change are multiplied by the preset proportional gain coefficient and differential gain coefficient, respectively, and then summed to obtain the preliminary damper opening adjustment amount.

[0127] The initial speed adjustment base amount is obtained by multiplying the smoothing efficiency deviation value by the speed proportional gain coefficient; the compensation amount is obtained by multiplying the damper opening adjustment amount by the decoupling compensation coefficient; the compensation amount is obtained by subtracting the base amount from the compensation amount. The disturbance effect of damper action on the graded particle size field can be eliminated, and the initial rotor speed increment can be generated.

[0128] Based on the calculated preliminary damper opening adjustment and preliminary rotor speed setting, combined with the current real-time operating reference parameters of the classifier, the target damper opening absolute value and target rotor speed absolute value are accumulated and generated respectively.

[0129] Perform a physical feasibility boundary check on the generated target parameters to ensure that the damper opening is within the effective stroke range of 0% to 100% and the rotor speed is within the safe range allowed by the rated frequency of the motor.

[0130] Through the above decoupling mapping and nonlinear operation processing, the smoothing efficiency deviation value of the previous step is transformed into the initial damper opening adjustment amount and the initial rotor speed set value with physical execution significance, so as to realize the expected technical effect of decoupling response and precise control of multivariable control system.

[0131] For example, when the smoothing efficiency deviation value E is -2.5% (indicating that the actual efficiency is 2.5 percentage points lower than the optimal lower limit) and the rate of change dE / dt is -0.1% / s, the controller calls the mapping matrix. After proportional-derivative calculation, a preliminary damper opening adjustment is generated as -3.5%, which suggests closing the secondary damper by 3.5% to enhance the swirl intensity. Assuming the current base speed is 1200 rpm, after decoupling mapping calculation, a preliminary rotor speed increment is generated as -34.7 rpm. The system generates a preliminary control command: reduce the secondary damper opening by 3.5% and reduce the rotor speed by 34.7 rpm. This output effectively compensates for the phenomenon of fine powder becoming coarse due to excessive airflow, and at the same time avoids the problem of reduced throughput caused by simple speed reduction through decoupling terms, significantly improving the targeting and stability of the control.

[0132] S4.4: Perform dead zone judgment and step size constraint processing on the initial damper opening adjustment amount and the initial rotor speed setting value to generate the final intelligent control guidance information. The input conditions are the initial damper opening adjustment amount and the initial rotor speed setting value obtained by decoupling calculation. The technical processing method is to apply dead zone logic to eliminate small disturbances and superimpose the maximum single-step adjustment range limit strategy. The output target is the final intelligent control guidance information that conforms to the action characteristics of the actuator and includes the specific damper opening adjustment amount and rotor speed setting value.

[0133] The system receives the initial damper opening adjustment and the initial rotor speed setting value obtained from the decoupling calculation as input data for dead zone judgment and step size constraint processing.

[0134] The absolute value of the initial damper opening adjustment is calculated and compared with the preset damper action dead zone threshold. If the absolute value of the adjustment is less than the dead zone threshold, the slight disturbance is determined to be system noise or measurement error, and the damper opening adjustment is forcibly set to zero to eliminate invalid high-frequency jitter of the actuator.

[0135] A similar dead-zone logic judgment is performed on the initial rotor speed setting value. If the speed change is lower than the minimum response threshold of the motor drive, the current speed command is kept unchanged to avoid the inverter generating additional harmonic interference due to frequent fine-tuning.

[0136] For valid adjustment amounts that pass the dead zone test, single-step constraint processing is performed. The upper limit of the maximum allowable single-step adjustment range of the actuator is obtained, which is set based on the maximum safe angular velocity of the damper mechanical linkage and the maximum acceleration characteristics of the motor rotor.

[0137] A limiting function is used to truncate the effective adjustment amount to ensure that the final output damper opening adjustment amount does not exceed the maximum allowable step size. If the calculated adjustment amount exceeds the upper limit, it is clamped to the maximum allowable step size value to prevent violent oscillations in the airflow field inside the classifier due to excessively abrupt control commands.

[0138] Apply the same step size limit to the rotor speed setpoint, compare the absolute value of the initial speed adjustment with the maximum allowable single-step change value, take the smaller one as the final step size, and keep its direction consistent with the initial adjustment amount.

[0139] The damper opening adjustment and rotor speed increment, after dead zone elimination and step size constraint, are superimposed on the current damper actual opening reference value and the current rotor speed reference value, respectively, to generate the final intelligent control guidance information containing the specific target position.

[0140] By using dead zone judgment and step size constraint processing, the preliminary decoupling results of the previous step are transformed into final intelligent control guidance information that conforms to the physical characteristics of the actuator and has stability, thereby achieving the expected technical effects of suppressing control oscillations, protecting mechanical components, and improving the smoothness of closed-loop control.

[0141] For example, the preset dead zone threshold for damper operation is 2%, and the maximum single-step adjustment range is 5%. If the initial damper opening adjustment is 1.5%, it is set to zero because it is less than 2%; if the initial adjustment is 8%, it is limited to 5%. The preset maximum single-step change in rotor speed is 10 rpm. If the initial speed adjustment exceeds the maximum allowable step size, it is clamped to the maximum allowable step size before being output. The final generated control command causes the damper to open smoothly by 5%, and the rotor speed to decrease steadily by 10 rpm, effectively avoiding abrupt changes in the airflow field.

[0142] like Figure 2 As shown, step S5: The intelligent control guidance information is output to the actuator via the fieldbus protocol, and the multi-flow control intelligent control strategy is coupled to adjust the opening of the secondary air damper and the auxiliary air volume of the classifier, so as to change the airflow swirl intensity and axial wind speed distribution, thereby optimizing the classifier's operating parameters. Specifically, this includes: S5.1: The intelligent control guidance information is encapsulated and processed using a communication protocol. The control data, which includes the damper opening adjustment and rotor speed setpoint, is converted into a standard industrial control frame using the fieldbus protocol stack to generate a fieldbus control instruction packet with a check bit and address identifier.

[0143] Receive the final intelligent control guidance information output from step S4. This information includes the secondary air damper opening adjustment ΔV after dead zone judgment and step size constraint processing, and the classifier rotor speed setpoint N. set .

[0144] The control and guidance information is processed to standardize the data format, converting floating-point values ​​into fixed-point or integer data formats that conform to the requirements of the industrial fieldbus protocol, so as to ensure the accuracy and compatibility of data transmission.

[0145] Construct the data payload area of ​​the fieldbus control frame, and arrange the converted damper opening adjustment and rotor speed setpoint according to the predefined register mapping address to form the main body of the control data packet to be sent.

[0146] The Cyclic Redundancy Check (CRC) code of the data packet is calculated. The payload data is processed according to a preset polynomial generation algorithm to generate a check bit sequence for detecting transmission errors.

[0147] The check bit sequence, the device address identifier of the target actuator, the function code, and the data length field are sequentially encapsulated into the header and trailer of the data packet, and assembled into a complete standard industrial control frame.

[0148] The assembled control frame is written to the transmit buffer through the fieldbus protocol stack interface, triggering physical layer signal conversion and generating a fieldbus control command packet with complete addressing information and integrity protection mechanism.

[0149] By using the above-mentioned communication protocol encapsulation processing method, the abstract control quantity generated in the previous step is transformed into a digital instruction packet that can be recognized by the underlying actuator and has anti-interference capability, thereby realizing a reliable mapping of control instructions from the logic layer to the physical transmission layer.

[0150] For example, the Modbus TCP protocol is used for encapsulation, with the address of the secondary air actuator of the classifier set to 0x01 and the address of the rotor frequency converter to 0x02. The damper opening adjustment of -5.0% is converted to a 16-bit signed integer -50 (resolution 0.1%), and the rotor speed setting of 800rpm is converted to a 16-bit unsigned integer 8000 (resolution 0.1rpm). According to the Modbus frame structure, the transaction processing identifier, protocol identifier, length field, unit identifier, and function code 0x10 (writing multiple registers) are filled in. CRC16 check is performed on the data payload, generating a checksum of 0xA3B2. The final generated hexadecimal instruction packet is 00010000000901100000000204FFCE1F40A3B2. This instruction packet is sent to the PLC gateway through the Ethernet physical layer. After parsing, it accurately drives the actuator, reducing the communication error rate to zero and ensuring high-reliability transmission of control instructions in environments with strong electromagnetic interference.

[0151] S5.2: Based on the fieldbus control command package, the multi-flow control and intelligent regulation strategy is analyzed. The preset airflow distribution mapping logic is used to calculate the damper opening adjustment amount into the target displacement signal of the secondary air damper actuator, so as to generate the secondary air regulation drive pulse for changing the airflow swirl intensity.

[0152] It should be noted that the multi-flow control air intelligent regulation strategy refers to a control scheme that coordinates and decouples the secondary air and auxiliary air volume of the classifier. Its core is to analyze the classification efficiency deviation into swirling intensity adjustment component and axial wind speed adjustment component, which correspond to two independent control channels: the opening of the secondary air damper and the frequency of the auxiliary air variable frequency, respectively. The cross-coupling effect between the two channels is eliminated by a preset airflow distribution mapping matrix, so that the swirling flow field and the axial flow field change synchronously in the classification zone according to a preset ratio.

[0153] Parse the fieldbus control command packet and extract the secondary air damper opening adjustment value contained therein, which is used as the input variable for the airflow distribution mapping logic.

[0154] The airflow distribution mapping table preset in the controller memory is invoked. This mapping table establishes a nonlinear correspondence between the change in damper opening and the number of actuator displacement pulses.

[0155] Based on the current classifier operating condition index mapping table, find the target displacement reference value that matches the extracted damper opening adjustment amount.

[0156] A mechanical hysteresis compensation coefficient for the actuator is introduced to linearly correct the target displacement reference value, thereby eliminating control errors caused by mechanical transmission backlash.

[0157] The corrected target displacement value is converted into the pulse equivalent of a stepper motor or servo driver, and a specific pulse number sequence is calculated and generated.

[0158] An acceleration smoothing process is applied to the generated pulse sequence to limit the rate of change of the pulse frequency and prevent the actuator from generating mechanical vibration due to instantaneous impact.

[0159] The pulse sequence is encoded into a secondary air conditioning drive pulse signal using PWM modulation technology. This signal directly corresponds to the action command of the secondary air damper actuator.

[0160] Through the above processing method, the damper opening adjustment amount in the previous step is converted into a secondary air conditioning drive pulse with mechanical actuation characteristics, thereby achieving precise and controllable adjustment of the airflow swirl intensity.

[0161] For example, the secondary air damper opening adjustment is set to 5%, and the base displacement pulse count is obtained from a table as 2000 pulses. A mechanical hysteresis compensation coefficient of 1.05 is introduced, resulting in a target displacement of 2100 pulses. The maximum acceleration limit is set to 500 pulses / second², generating a trapezoidal acceleration pulse sequence with a duration of 4.2 seconds. Finally, a gradually changing frequency drive pulse signal is output to the damper actuator, causing the damper to smoothly open to the target position, significantly improving the stability and response accuracy of the airflow field adjustment.

[0162] S5.3: The auxiliary air volume frequency converter is frequency modulated according to the secondary air conditioning drive pulse. The rotor speed setpoint is converted into the frequency converter output frequency reference value using the vector control algorithm to generate the auxiliary air volume frequency converter control signal for adjusting the axial wind speed distribution.

[0163] The rotor speed setpoint is received from the secondary wind regulation drive pulse obtained from step S5.2 and the final intelligent control guidance information generated from step S4.4, and used as the initial input conditions for frequency modulation processing.

[0164] The rotor speed setpoint is calculated using a vector control algorithm. The stator current in the three-phase stationary coordinate system is decomposed into excitation and torque components using coordinate transformation. A dynamic mathematical model of the motor is established to accurately match the load characteristics of the classifier rotor.

[0165] Based on the calculated torque component reference value, combined with the rotational inertia of the classifier rotor and the current real-time speed feedback, the electromagnetic torque command required to maintain the target speed is calculated to ensure the speed response is fast and stable under airflow disturbance.

[0166] Space vector pulse width modulation technology is used to convert the calculated voltage vector reference value into the on / off signal sequence of the inverter switching devices, and generate a high-frequency PWM drive waveform to control the power output of the auxiliary air volume inverter.

[0167] Based on the trend of airflow swirl intensity change indicated by the secondary air conditioning drive pulse, the ramp rate of the inverter output frequency reference value is dynamically adjusted to achieve time-series synchronous coupling of airflow change and speed change.

[0168] Through the above vector control and PWM modulation processing, the rotor speed setpoint is converted into a high-precision auxiliary air volume frequency conversion control signal, realizing the fine control of the axial wind speed distribution inside the classifier, and ensuring the effective separation of powder particles in the centrifugal force field.

[0169] S5.4: The secondary air conditioning drive pulse and the auxiliary air volume frequency conversion control signal are used to synchronously drive the classifier actuator. The mechanical linkage action changes the opening of the secondary air damper and the auxiliary air volume is adjusted in combination with the change of motor speed to generate an optimized classification airflow field with specific swirling intensity and axial wind speed distribution.

[0170] It receives the secondary air conditioning drive pulse and the auxiliary air volume frequency conversion control signal as the synchronous action command input for the classifier actuator.

[0171] The secondary air conditioning drive pulse is amplified and servo positioned and analyzed to drive the pneumatic or electric actuator to drive the mechanical linkage to generate linear displacement, thereby precisely changing the opening angle of the damper at the secondary air inlet.

[0172] Changes in the damper opening directly regulate the velocity and flow rate distribution of the secondary air entering the classifier, thereby altering the swirling intensity and tangential velocity component of the airflow within the classification zone.

[0173] The frequency of the auxiliary air volume variable frequency control signal is demodulated synchronously, and the frequency of the rotating magnetic field of the induced draft fan or auxiliary fan motor is adjusted by the vector frequency converter to achieve a smooth transition of motor speed.

[0174] The axial wind speed of the auxiliary air duct is adjusted in real time according to the change of motor speed, thereby optimizing the suspension state and residence time of powder materials in the classification chamber.

[0175] The changes in the swirling flow field caused by adjusting the damper opening and the changes in the axial flow field caused by adjusting the fan speed are coupled and superimposed in the spatiotemporal domain.

[0176] A dynamic equilibrium model of the three-dimensional airflow field inside the classifier is constructed to ensure that the ratio of swirling intensity to axial wind speed is maintained within the preset optimal separation range.

[0177] Through the coordinated drive of the mechanical linkage action and the change of motor speed, the control signal of the previous step is transformed into an optimized graded airflow field with specific swirling intensity and axial wind speed distribution, so as to achieve precise control of the balance point of centrifugal force of powder particles and airflow resistance, and significantly improve the stability and response speed of screening efficiency.

[0178] For example, when the secondary air adjustment drive pulse increases the damper opening by 5 degrees, the actuator completes the action within 2 seconds, increasing the secondary air inlet cross-sectional area by 10% and raising the tangential wind speed in the grading zone from 15 m / s to 16.5 m / s. Simultaneously, the auxiliary airflow frequency conversion control signal increases the fan frequency from 45 Hz to 48 Hz, the motor speed from 1350 rpm to 1440 rpm, and the axial wind speed from 3 m / s to 3.2 m / s. After adjusting the ratio of swirling intensity to axial wind speed, the centrifugal force on coarse particles is significantly enhanced, the carrying capacity of fine particles is optimized, the measured screening efficiency increases from 85% to 92%, and the system pressure fluctuation is reduced to within ±50 Pa, achieving efficient and stable grading operation.

[0179] S5.5: Based on the optimized classifying airflow field, dynamic classification of powder materials is carried out. The movement trajectory of coarse and fine particles is redistributed by the balance mechanism of centrifugal force and airflow resistance to generate the classifier operation parameters that meet the preset optimal screening efficiency range.

[0180] The optimized classifying airflow field parameters with specific swirling intensity and axial wind speed distribution generated in step S5.4 are received and used as boundary conditions for dynamic classification operations, inputting them into the classifier rotor cavity. After the powder material enters the classification zone with the airflow, it is subjected to the centrifugal force field generated by the high-speed rotating rotor, causing the particles to exhibit a radial movement tendency based on their particle size. Simultaneously, the axial airflow pointing towards the rotor center exerts an inward aerodynamic drag force on the particles, forming a resistance field opposite to the direction of the centrifugal force. A particle force balance model is established; when the centrifugal force and airflow resistance on the particles reach dynamic equilibrium, this particle size is the cutting particle size under the current operating condition. The cutting particle size under the critical equilibrium state is calculated using the following formula: in, For cutting particle size, For gas dynamic viscosity, Radial airflow velocity, Particle density, The rotor angular velocity, Where is the rotor radius. This formula is based on the Stokes laminar flow resistance assumption and is applicable to estimating the cut particle size of micron-sized powder particles in a classifier.

[0181] Based on the calculated cutting particle size, the mixed powder entering the classification zone undergoes trajectory redistribution. Coarse particles with a diameter greater than d_50 are thrown towards the wall and fall into the return powder zone due to the dominance of centrifugal force, while particles with a diameter less than d_50 are... 50 Fine particles, due to the superior airflow resistance, pass through the gaps between the rotor blades and enter the finished product zone. Real-time monitoring of the classifier outlet pressure difference and motor current fluctuations verifies the stability of the airflow field, ensuring a clear separation interface between coarse and fine particles without material mixing. Through the aforementioned balance mechanism between centrifugal force and airflow resistance, the optimized airflow field parameters from the previous step are transformed into particle separation results that conform to the preset optimal screening efficiency range, achieving precise closed-loop control of the classifier's operating parameters.

[0182] For example, the rotor radius R of the classifier is set to 0.5 meters, and the gas dynamic viscosity μ is set to 1.8 × 10⁻⁶. -5 Pa·s, particle density ρ p The value is 2500 kg / m³. When the rotor angular velocity ω output by S5.4 is adjusted to 100 rad / s, the radial airflow velocity v r When the velocity is controlled at 0.8 m / s, the cutting particle size d is calculated by substituting the values ​​into the formula. 50The particle size is approximately 32 micrometers. At this point, coarse powder with a particle size greater than 32 micrometers effectively falls back into the powder return pipeline, while fine powder with a particle size less than 32 micrometers smoothly enters the finished product collection system. On-site operation data shows that under this parameter combination, the pressure difference at the classifier outlet is stable at 1200 Pa, the motor current fluctuation is less than 2%, the screening efficiency is significantly improved, and the finished product fineness qualification rate is greatly increased, effectively solving the efficiency lag problem caused by traditional static control.

[0183] like Figure 3 As shown, step S6: After the screening test is completed, the self-circulation loop is controlled by a three-way reversing valve to transport the mixture of coarse powder on the sieve and fine powder on the undersize back to the designated position between the classifier and the mill in the main pipeline without loss, so as to complete the closed-loop collection of materials and avoid material waste. Specifically, this includes: S6.1: Real-time acquisition and logical judgment processing of material accumulation status signals at the outlet of the detection unit to generate a self-circulating loop start command with execution conditions, ensuring that subsequent actions are triggered only when the screening and weighing data is completely uploaded and there is no residual alarm in the detection chamber.

[0184] The material accumulation status signal at the outlet of the detection unit is collected in real time. The material level sensor installed at the bottom of the detection chamber is used to obtain the current material filling height data in the chamber and generate the original material level monitoring sequence.

[0185] The original material level monitoring sequence is subjected to moving average filtering to eliminate instantaneous signal jitter caused by discontinuous powder flow and extract stable real-time material level feature values.

[0186] The system synchronously reads the control register status of the rapid vibrating screening device, verifies whether the screening timer has returned to zero and whether the data lock flag of the weighing sensor has been set, and generates a status indicator indicating the completion of the detection process.

[0187] The real-time material level characteristic value is logically compared with the preset empty bin threshold. If the material level is lower than the threshold and the detection process completion status is true, the detection bin is determined to be in an emptying state.

[0188] Based on the result of the emptied state determination, combined with the reset state of the residual alarm register in the previous cycle, a logical AND operation is performed to generate a self-loop start instruction that meets the execution conditions.

[0189] The validity of the startup command is verified a second time by using a hardware watchdog mechanism to prevent false triggering caused by electromagnetic interference and ensure the reliability of the command output.

[0190] For example, the level sensor uses a capacitive level switch, with an empty chamber threshold set to 5%. Two minutes after screening, the weighing data locks the flag at position 1, the level sensor feedback voltage corresponds to a level of 3%, and there is no residual alarm. After filtering, the system confirms the stable level, the logical AND operation result is true, and a high-level start command pulse with a pulse width of 100ms is generated, successfully triggering the subsequent three-way valve action to prepare the detection chamber for emptying.

[0191] S6.2: Based on the self-circulation loop start command, the pneumatic actuator of the three-way reversing valve is driven and controlled to switch the valve opening from the bypass detection position to the main return position, so as to establish a closed physical conveying channel connecting the detection unit outlet and the main pipeline inlet.

[0192] Receive the self-circulation loop start command, parse the target valve position status identifier and action timing parameters in the command, and confirm that the current three-way directional valve is in the initial state of the bypass detection position.

[0193] The pneumatic actuator drive module is invoked to generate a PWM pulse width modulation signal based on the preset air pressure control curve, which controls the on / off state of the solenoid valve coil to adjust the intake air flow.

[0194] The system monitors the feedback signal from the displacement sensor installed on the stem of the three-way directional valve to obtain the instantaneous position data of the valve core as it moves from the bypass detection position to the main return position in real time.

[0195] The instantaneous position data is compared with the target threshold of the main return position in a closed loop, and the PWM duty cycle is dynamically adjusted to overcome the back pressure resistance of the pipeline and ensure that the valve core smoothly transitions to the fully open position.

[0196] When the displacement sensor feedback value stabilizes within the allowable error range of the main return position, the valve opening is locked, the drive signal is cut off, and the air circuit is kept pressurized to establish a sealed physical transport channel between the detection unit outlet and the main pipeline inlet.

[0197] Through the above-mentioned pneumatic closed-loop control processing method, the logical start command is converted into precise mechanical valve position action, realizing rapid and leak-free switching of the material return channel, and providing a reliable physical link guarantee for subsequent lossless material return.

[0198] For example, the three-way directional valve is driven by a double-acting cylinder with a target switching time of 1.5 seconds. The controller outputs a 100Hz PWM signal with an initial duty cycle of 80% to provide maximum starting torque. The displacement sensor has a range of 0-100mm, corresponding to a valve rotation angle of 0-90 degrees. When the valve stem displacement reaches 45mm (i.e., 45 degrees, the critical point for main circuit conduction), the PID algorithm intervenes, linearly reducing the duty cycle to 30% to achieve a soft landing and avoid mechanical shock. The final valve position stabilizes at 90±0.5 degrees, the sealing surface clamping force meets the 0.6MPa requirement, the channel establishment time is 1.2 seconds, and the air leakage rate is less than 0.1%, ensuring that fine powder does not overflow or backflow under negative pressure.

[0199] S6.3: Utilize the established closed physical conveying channel to perform frequency conversion speed regulation start-up on the screw conveyor or pneumatic conveying fan integrated in the self-circulating loop, so as to generate a directional conveying power field and convert the mixture of coarse powder on the screen and fine powder on the screen after detection into a continuous flow of return material.

[0200] Upon receiving the self-circulating loop start command and the three-way reversing valve positioning signal, the sealed physical conveying channel is confirmed to be established. The variable frequency drive control module is invoked to read the preset return conveying frequency reference value, which is pre-calibrated based on the material bulk density and pipeline resistance characteristics within the detection unit. Soft-start control is implemented on the screw conveyor drive motor or pneumatic conveying fan, gradually increasing the output frequency according to a linear ramp function to avoid instantaneous high current surges that could cause grid fluctuations or damage to mechanical transmission components. The motor current feedback signal and speed encoder data are monitored in real time. When the actual speed reaches 80% of the preset reference value, the system switches to closed-loop vector control mode to precisely lock the target speed and generate a stable directional conveying power field. Pressure sensor data at the conveying pipeline inlet is monitored. If a pressure surge exceeds a preset threshold, the inverter output frequency is dynamically fine-tuned to maintain constant torque and prevent material blockage or interruption. Under stable conveying conditions, the mixture of coarse powder on the oversize screen and fine powder on the undersize screen in the detection chamber is transformed into a continuous and uniform return flow, ensuring that the material is suspended or spirally propelled in the pipeline.

[0201] By using variable frequency speed control to start the process, the static material after detection is transformed into a continuous return flow with stable kinetic energy, achieving the expected technical effect of returning the material to the main pipeline without damage and with high efficiency.

[0202] For example, for cement raw meal return powder with a particle size distribution of 0-80μm, the initial frequency of the screw conveyor frequency converter is set to 0Hz, and the acceleration time is 5s. The preset reference frequency is 35Hz, corresponding to a screw speed of 1400r / min. At startup, the frequency converter linearly increases the frequency at a slope of 7Hz / s. When the frequency reaches 28Hz, the current feedback stabilizes at 60% of the rated value, and the system switches to vector control mode. If the pipeline pressure sensor detects a sudden increase in pressure from 2.5kPa to 4.0kPa, the controller automatically lowers the frequency to 33Hz and maintains it for 2s. After the pressure drops back to 2.6kPa, it resumes operation at 35Hz. Ultimately, a stable return flow with a velocity of 1.2m / s is formed, with no material blockage alarms throughout the process, and the material return is smooth.

[0203] S6.4: Based on the flow feedback signal of the continuously flowing return material flow, the operating parameters of the conveying power field are dynamically matched and adjusted to overcome pipeline resistance and accurately inject the return material flow into the designated return point of the main pipeline between the classifier and the mill, thus completing the non-destructive closed-loop collection of materials.

[0204] The instantaneous volumetric flow rate signal from the flow meter in the circulation loop is collected, and combined with the return flow density data monitored in real time by the online density meter, the mass flow rate feedback value at the current moment is calculated.

[0205] Based on the deviation comparison between the mass flow feedback value and the preset target conveying rate, a primary control error signal is generated for adjusting the conveying power field.

[0206] The primary control error signal is processed using a proportional-integral-derivative (PID) control algorithm to calculate the frequency adjustment of the screw conveyor frequency converter or pneumatic conveying fan.

[0207] The driving parameters of the actuator are dynamically adjusted according to the frequency adjustment to overcome nonlinear load fluctuations caused by pipeline bend resistance and material accumulation, thereby stabilizing the conveying speed of the return flow.

[0208] Real-time monitoring of pressure sensor data at the main pipeline return point assesses the degree of disturbance to the main system's airflow field during the return material injection process, generating a pressure balance compensation coefficient.

[0209] By adding the pressure balance compensation coefficient to the frequency adjustment, the output torque of the conveying power field is finely adjusted to ensure that the return flow smoothly integrates into the main pipeline airflow in a laminar state, thus avoiding pressure differential oscillations in the main system caused by turbulent impact.

[0210] Through the above dynamic matching and adjustment process, the continuous return flow generated in the previous step is transformed into a non-destructive injection flow with a constant mass flow rate and pressure balance with the main system, thereby achieving the expected technical effect of stable and zero-interference material closed-loop collection.

[0211] For example, with a target conveying rate of 50 kg / min, a flow meter detects an instantaneous volumetric flow rate of 0.04 m³ / min, a densitometer measures a density of 1200 kg / m³, and the calculated actual mass flow rate is 48 kg / min. The initial control error is 2 kg / min. The PID controller parameters are set to Kp=1.5, Ki=0.2, Kd=0.1. After one control cycle, the output frequency adjustment is +2.5 Hz. Simultaneously, the return point pressure sensor shows that the main pipeline pressure fluctuation exceeds ±50 Pa, generating a pressure balance compensation coefficient of -0.5 Hz. Finally, the inverter frequency is adjusted to the original reference frequency plus 2.0 Hz. This adjustment ensures that the return flow stably reaches the target rate within 30 seconds, and the main pipeline pressure difference fluctuation is controlled within ±10 Pa, significantly improving the system's operational stability.

[0212] S6.5: Perform consistency comparison and verification processing between the cumulative quality data of the return flow injected into the designated return point of the main pipeline and the initial sampling quality data to generate a zero material loss confirmation signal and reset the three-way reversing valve to the standby state, ensuring that the representativeness of the sampling in the next detection cycle is not affected by the previous residue.

[0213] The cumulative mass data collected after the return flow in the circulation loop is completed is generated in real time by a high-precision weighing sensor installed at the outlet of the return pipe, serving as the measured value of the material closed-loop collection amount. The initial total mass data of the powder sample to be tested, locked by the dynamic weighing module at the beginning of the current testing cycle, is read. This data has been filtered, denoised, and validated in previous steps, serving as a benchmark reference value for the material input amount.

[0214] A material conservation verification logic is constructed to calculate the absolute difference between the cumulative mass data of the return flow and the initial total mass data, in order to quantify the material loss or gain during the detection process. A zero-loss material judgment threshold is set, which is calibrated based on the accuracy class of the weighing sensor and the residual characteristics of the conveying pipeline, and is used to define the allowable mass deviation range.

[0215] The calculated absolute difference is compared with the zero-loss material judgment threshold. If the absolute difference is less than or equal to the judgment threshold, the material balance is considered to be established in this detection process, and a zero-loss material confirmation signal is generated. If the absolute difference exceeds the judgment threshold, an abnormal alarm is triggered, the deviation value is recorded, and the subsequent model update process is paused until manual intervention or system self-diagnosis is completed. After generating the zero-loss material confirmation signal, a reset command is sent to the pneumatic actuator of the three-way reversing valve, driving the valve to switch from the main return position to the bypass detection standby position, cutting off the physical connection between the detection unit and the main pipeline. The position feedback signal of the three-way reversing valve is monitored to confirm that the valve is fully closed to the standby state, preventing backflow of air from the main pipeline into the detection circuit. The temporary stored data inside the detection unit is cleared, and the zero point of the weighing sensor is reset to prepare for automatic sampling in the next detection cycle.

[0216] Through the above consistency comparison and verification and valve reset control, the return material conveying result of the previous step is transformed into a system status confirmation signal, realizing the rapid reset of the detection circuit and ensuring the representativeness of the sampling in the next cycle, thus ensuring the long-term stable operation of the online detection system.

[0217] For example, the initial total mass data is 500.0g, the cumulative mass data of the return flow is 499.8g, and the absolute difference is 0.2g. The material zero loss judgment threshold is set to 0.5g. Since 0.2g is less than 0.5g, the material balance is determined to be established, and a material zero loss confirmation signal is generated. After receiving the reset command, the three-way reversing valve completes the switch from the return position to the standby position within 2 seconds, and the position feedback signal shows an opening of 0%. The zero-point drift of the weighing sensor is controlled within ±0.01g, ensuring that the sampling error of the next cycle is less than 0.02%, significantly improving the continuity and reliability of the detection data.

[0218] Step S7: Update the adaptive coefficients of the grading efficiency correction model based on the system operating status feedback after returning to the main pipeline, and perform model parameter correction for the real-time fine powder content index of the returned powder in the next detection cycle to achieve dynamic balance optimization of the entire closed loop of detection-evaluation-control. Specifically, this includes: S7.1: Collect and standardize the system operation status feedback data after returning to the main pipeline to obtain a set of real-time operating parameters including system air volume, classifier speed, mill feed rate and system pressure difference, so as to generate a benchmark operating condition input vector for model calibration.

[0219] S7.2: Based on the baseline working condition input vector and the dynamic screening efficiency evaluation results output from the previous detection cycle, the theoretical return fine powder content prediction value is back-calculated using the material balance principle. The theoretical return fine powder content prediction value is then compared with the real-time return fine powder content index measured in this cycle to generate an efficiency residual sequence that characterizes the degree of deviation of the model.

[0220] The input vector of the received baseline working condition and the dynamic screening efficiency evaluation result of the previous detection cycle are used as the initial boundary conditions for the back-calculation of the theoretical fine powder content.

[0221] The preset classifier material balance model is invoked to extract the current system air volume, classifier rotor speed, and circulating load rate as key physical constraint parameters.

[0222] Based on the principle of gas-solid two-phase flow dynamics, and using the same cutting particle size calculation logic as in step S5.5, the theoretical cutting particle size under the current operating conditions is calculated based on key physical constraint parameters such as the current system air volume, classifier rotor speed, and cyclic load rate.

[0223] Based on the theoretical cutting particle size and the measured particle size distribution characteristics, the theoretical screening efficiency value is obtained by integrating the product of the classification probability function and the feed particle size distribution density function using a preset Tromp curve model.

[0224] The calculated theoretical screening efficiency value is mapped and converted into the predicted value of the fine powder content of the theoretical return powder, forming the expected output benchmark of the model.

[0225] Obtain the real-time fine powder content index measured by the online detection loop in this cycle, and perform timestamp alignment processing to eliminate phase errors caused by transmission delay.

[0226] A point-by-point subtraction operation is performed between the theoretical predicted value of the fine powder content of the returned powder and the measured real-time fine powder content index to generate a single-point efficiency residual value.

[0227] The single-point efficiency residual values ​​of multiple consecutive detection cycles are arranged in time series to construct an efficiency residual sequence that characterizes the degree of deviation of the model.

[0228] Through the above chain derivation, the multi-source operating parameters and measured data are transformed into a quantified efficiency residual sequence, which enables accurate capture of model prediction bias and provides a high-confidence error feedback signal for subsequent adaptive coefficient correction.

[0229] For example, the system air volume is set to 50,000 m³ / h, the classifier speed to 1200 rpm, and the circulation load rate to 3.5. Based on the material balance model, the theoretical cutting particle size is calculated to be 45 μm, and the theoretical predicted value of the fine powder content of the returned powder is 18.5%. The real-time fine powder content of the returned powder measured by the online detection device is 19.2%. The single-point efficiency residual is obtained by performing subtraction. The residual values ​​of 10 consecutive detection cycles are used to form an efficiency residual sequence [-0.7, -0.65, -0.8, ...]. This sequence accurately reflects the systematic negative bias of the model caused by liner wear, and significantly improves the parameter convergence speed and correction accuracy of the subsequent gradient descent algorithm.

[0230] S7.3: Perform sliding window statistical analysis and time series feature extraction on the efficiency residual sequence to identify systematic deviation trends caused by changes in material properties or equipment wear, so as to generate deviation feature factors that characterize the drift characteristics of the current operating condition.

[0231] A sliding window truncation process is performed on the efficiency residual sequence. The window length is set to N detection periods. The residual values ​​of the current time t and the previous N-1 periods are extracted from the continuous historical residual data to form a local analysis vector.

[0232] The local analysis vectors are detrended, and the slope of the residual change within the window is calculated using a linear regression fitting method. Fluctuations caused by short-term random noise are removed, and the systematic trend term reflecting long-term drift is retained.

[0233] Statistical moment characteristics are calculated based on the detrended residual data. The arithmetic mean formula is used to calculate the residual mean to characterize the residual center offset. This index reflects the average deviation between the model's predicted values ​​and the measured values. The residual standard deviation is calculated to quantify the data dispersion. An increase in the standard deviation indicates drastic fluctuations in material properties or unstable equipment operation.

[0234] The skewness and kurtosis features of the residual sequence are extracted. Skewness reflects the asymmetry of the distribution to identify unidirectional wear trends, and kurtosis reflects the sharpness of the distribution to capture sudden working condition impacts, forming a multidimensional time series feature vector.

[0235] Principal component analysis is used to reduce the dimensionality of the multidimensional time series feature vector, and the principal component with the highest contribution rate is extracted as the deviation feature factor. This factor comprehensively characterizes the systematic deviation trend caused by changes in material properties or equipment wear.

[0236] By employing a sliding window statistical and temporal feature extraction approach, the efficiency residual sequence from the previous step is transformed into a low-dimensional and information-dense deviation feature factor. This achieves accurate quantification of operating condition drift characteristics, providing a high signal-to-noise ratio input for subsequent adaptive correction of model parameters. For example, the sliding window length N is set to 20, and efficiency residual data from the most recent 20 cycles are collected. The calculated residual mean μ is 0.05, standard deviation σ is 0.02, skewness is 1.2, and kurtosis is 3.5. Principal component analysis extracts the first principal component as the deviation feature factor, with a value of 0.85. This factor accurately reflects the continuous decline in efficiency caused by classifier liner wear, significantly improving the model's sensitivity to long-term operating condition drift.

[0237] S7.4: Execute the gradient descent iterative optimization algorithm based on the deviation characteristic factor, adjust the weight parameters and offsets inside the graded efficiency correction model, and calculate the latest working condition adaptive coefficients to adapt to the current variable working conditions, so as to complete the online self-learning update of the model parameters.

[0238] Receive the deviation characteristic factor generated by S7.3, which represents the drift characteristics of the current operating condition, and use it as the error feedback input vector of the gradient descent algorithm.

[0239] A loss function is constructed with the objective of minimizing the sum of squared efficiency residuals. Half of the sum of squared differences between the measured fine powder content of each sampling point within the sliding window and the model prediction is used as a quantitative indicator of the deviation of the current model.

[0240] The gradient vector of the loss function with respect to the adaptive coefficients of the operating conditions is calculated. This gradient is obtained by summing the product of the prediction error at each sampling point and the partial derivative of the model output with respect to the coefficients, reflecting the sensitivity of the prediction results to the influence of changes in each parameter.

[0241] Introducing the learning rate η controls the step size of the update parameter to prevent model oscillation or divergence due to excessive gradient.

[0242] Perform iterative update calculations for the adaptive coefficients under operating conditions, moving the old coefficients along the negative gradient direction by a specified step size.

[0243] The coefficient update rule is expressed using the following MathML formula: in, The adaptive coefficient for the current operating condition. For the latest updated coefficients, The learning rate constant is preset. loss function For coefficients The gradient vector.

[0244] Perform boundary constraint checks on the updated adaptive coefficients to ensure that the parameter values ​​are within a reasonable range that is physically permissible.

[0245] If the coefficients exceed the preset upper and lower limits, they are truncated to the boundary values ​​to maintain the numerical stability of the graded efficiency correction model.

[0246] By using gradient descent iterative optimization, the deviation characteristic factors from the previous step are transformed into the latest adaptive coefficients that fit the current variable working conditions, thereby achieving online self-learning updates of model parameters and dynamic improvement of screening efficiency evaluation accuracy.

[0247] For example, the number of samples N in the sliding window is set to 10, and the learning rate η is set to 0.01. When the root mean square error of the efficiency residual sequence is detected to be 0.05, the gradient value is calculated to be -0.02. According to the update formula, the original adaptive coefficient of 0.85 is adjusted to 0.85 - 0.01 * (-0.02) = 0.8502. After boundary checks, it is confirmed that this value is within the range of [0.5, 1.5] and is retained. In the next cycle prediction, the updated model significantly reduces the deviation between the theoretical fine powder content and the measured value, effectively compensating for the drift in classification performance caused by mill liner wear.

[0248] S7.5: Write the latest working condition adaptive coefficient into the parameter storage area of ​​the graded efficiency correction model, replace the old coefficient, and activate the new version of the graded efficiency correction model so that it can accurately calculate the real-time fine powder content index based on the updated parameters in the next detection cycle, and realize the dynamic balance optimization of the detection-evaluation-control closed loop.

[0249] The system receives the latest adaptive coefficient vector, calculated through gradient descent iterative optimization. This vector contains model weight parameters adjusted for the current material characteristics and equipment status. Write protection is removed from the parameter storage area of ​​the graded efficiency correction model to ensure the storage medium is writable and prevent data conflicts. The latest adaptive coefficients are serialized and encapsulated according to a preset data structure format, generating a binary data stream that meets memory alignment requirements. An atomic write instruction overwrites the serialized coefficient data stream to a specified address segment in the parameter storage area, replacing the old coefficient values ​​left over from the previous detection cycle, ensuring the integrity and consistency of parameter updates. Verification and comparison are performed on the storage area after the write operation. The newly written coefficient data is read and its cyclic redundancy check (CRC) code is calculated. If the verification result matches the expected value, the write is confirmed successful; otherwise, a retry mechanism is triggered to ensure data reliability. The control logic switch for the new version of the graded efficiency correction model is activated, pointing the parameter pointer of the model's computation kernel to the updated coefficient storage address, enabling the model to automatically load the latest parameters in the next calculation cycle. The model version status flag is locked to prevent dynamic parameter jumps within a single detection cycle, ensuring the stability of the evaluation process. By using the above-mentioned parameter hot update and model activation processing methods, the latest working condition adaptive coefficients obtained from offline optimization are transformed into effective calculation parameters of the online evaluation model in real time. This enables the graded efficiency correction model to adaptively track changing working conditions, ensuring a significant improvement in the calculation accuracy of the real-time fine powder content index in the next detection cycle. This achieves the expected technical effect of dynamic balance optimization in the detection-evaluation-control closed loop.

[0250] For example, the latest adaptive coefficient vector is [0.98, 1.02, -0.05], corresponding to the airflow correction factor, speed correction factor, and differential pressure offset, respectively. The parameter storage area is located at addresses 0x4000 to 0x400C, stored in 32-bit floating-point format. An atomic write operation is performed, writing the coefficient 0.98 to 0x4000, 1.02 to 0x4004, and -0.05 to 0x4008. The CRC32 checksum is calculated to be 0xA1B2C3D4, and the model is activated after verification. In the next cycle, the model uses the new coefficients to calculate the screening efficiency, reducing the error from 3.5% to 0.8%, significantly improving the evaluation accuracy, making the control commands more precise, and greatly enhancing the system's operational stability.

[0251] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0252] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the element or object preceding “comprising” or “including” encompasses the element or object listed following “comprising” or “including” and its equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0253] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for online detection of fine powder content and evaluation of sieving efficiency in graded sieve return powder, characterized in that, include: In the preset bypass detection loop of the graded sieve return powder pipeline, the automatic sampling device is used to extract powder samples at set time intervals, and the initial total mass data of the samples is obtained through the dynamic weighing module to generate a powder sample with instantaneous representativeness. The powder sample to be tested is subjected to rapid vibration sieving. The fine powder under the sieve and the coarse powder over the sieve are separated using a standard sieve. The amount of material under the sieve is weighed again. The real-time fine powder content index is calculated based on the difference between the initial total mass data and the amount of material under the sieve. The real-time fine powder content index is integrated with the system air volume, classifier speed and circulating load operating parameters collected from the distributed control system. Based on the material balance principle, a classification efficiency correction model including an adaptive coefficient for working conditions is constructed to generate dynamic screening efficiency evaluation results. Based on the dynamic screening efficiency evaluation results, threshold determination and change rate limiting are performed. If the evaluation results deviate from the preset optimal range, intelligent control guidance information including damper opening adjustment and rotor speed setting is generated; otherwise, the current control parameters remain unchanged. The intelligent control guidance information is output to the actuator via the fieldbus protocol, and the multi-flow control intelligent control strategy is coupled to adjust the opening of the secondary air damper and the auxiliary air volume of the classifier, so as to change the airflow swirl intensity and axial wind speed distribution.

2. A method for online detection of fine powder content and evaluation of sieving efficiency in graded sieve return powder, characterized in that, Also includes: After the screening test is completed, the self-circulation loop is controlled by the three-way reversing valve to transport the mixture of coarse powder on the screen and fine powder on the screen back to the designated position between the classifier and the mill in the main pipeline without damage, so as to complete the closed-loop collection of materials and avoid material waste. The operating condition adaptive coefficient in the graded efficiency correction model is updated based on the system operation status feedback after returning to the main pipeline. The model parameters are then corrected for the real-time fine powder content index of the next detection cycle to achieve dynamic balance optimization of the detection-evaluation-control closed loop.

3. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The process involves integrating the real-time fine powder content index with system airflow, classifier speed, and circulating load operating parameters collected from the distributed control system. Based on the material balance principle, a classification efficiency correction model including an adaptive coefficient is constructed to generate dynamic screening efficiency evaluation results. Specifically, this includes: The system air volume, classifier speed and circulating load operating parameters collected from the distributed control system are timestamped and outlier removed to generate a standardized multi-source operating parameter sequence, ensuring the consistency of the data benchmark for subsequent fusion calculations. Based on the standardized multi-source operating parameter sequence and the real-time fine powder content index, the material particle size distribution difference between the inlet and outlet of the classifier is calculated using the material balance principle to generate a static basic efficiency intermediate variable characterizing the current separation state. Based on the historical trends of multiple sets of recent online detection data, the recursive least squares algorithm is used to identify the parameters of the preset empirical and semi-empirical equations in order to generate adaptive coefficients for compensating for equipment wear and material property fluctuations. The intermediate variables of static basic efficiency and the adaptive coefficients of operating conditions are input into the graded efficiency correction model to perform weighted mapping operations, so as to generate dynamic screening efficiency evaluation results that eliminate operating condition drift errors.

4. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The threshold determination and rate-of-change limiting processing based on the dynamic screening efficiency evaluation results are performed. If the evaluation results deviate from the preset optimal range, intelligent control guidance information including damper opening adjustment and rotor speed setpoint is generated; otherwise, the current control parameters are maintained unchanged. Specifically, this includes: The dynamic screening efficiency evaluation results are compared with the preset optimal interval to generate an efficiency deviation status indicator. Based on the efficiency deviation status flag, a rate-limiting filtering process is performed to generate a smoothed efficiency deviation value; The smoothing efficiency deviation value is used to look up a table and map it to a multivariable decoupled control model to generate the initial damper opening adjustment and the initial rotor speed setpoint. The initial damper opening adjustment and the initial rotor speed setting are subjected to dead zone judgment and step size constraint processing to generate the final intelligent control guidance information.

5. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The automatic sampling device extracts powder samples from the main pipeline airflow to generate the powder sample to be tested.

6. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The initial total mass data is compared and verified with a preset sample representativeness threshold. If the data is valid, the initial total mass data is locked as the standard benchmark value.

7. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, Before performing the rapid vibrating sieving process, the powder sample to be tested is spread in a single layer on the surface of the standard screen using a vibrating feeding mechanism to eliminate the material accumulation effect.

8. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, Dynamic weighing and data acquisition are performed on the separated undersize fine powder stream to monitor the cumulative mass value falling into the collection container in real time, so as to obtain the mass data of the undersize material.

9. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The difference between the initial total mass data and the undersize mass data is calculated, and the normalization ratio is further calculated to generate the real-time fine powder content index.

10. The method for online detection of fine powder content and evaluation of sieving efficiency of graded sieve return powder according to claim 1, characterized in that, The damper opening adjustment amount in the intelligent control guidance information is calculated as the target displacement signal of the secondary air damper actuator to change the airflow swirl intensity, and the rotor speed setting value is converted into the inverter output frequency reference value to adjust the axial wind speed distribution.

Images