Vibrating screen parameter self-adaptive control method and system based on machine learning

Abstract

The invention relates to the technical field of parameter control, and discloses a vibrating screen parameter self-adaptive control method and system based on machine learning. The method comprises the following steps: collecting coal characteristic parameters through an image device and a sensor; calculating a load coefficient and a difficulty index according to the parameters, and adjusting the parameters of the dual-motor vibration exciter to control vibration; the spring pre-tightening force and the screen surface inclination angle are adjusted according to vibration feedback, and the screening track is optimized; and the feeding speed and the baffle height are adjusted based on screening efficiency data, and self-adaptive control is achieved. The technical problems that traditional vibrating screen fixed parameter control cannot adapt to multi-coal-type changes, a vibrating screen control system lacks coal material characteristic real-time sensing and analyzing capacity, and a precise mapping and dynamic optimization mechanism is lacked between vibrating parameters and screening efficiency are solved.

Description

Technical Field

[0001] This application relates to the technical field of parameter control, and particularly to a vibration screen parameter adaptive control method and system based on machine learning. Background Art

[0002] A vibrating screen is a key device for material sorting and grading in the coal processing field. The traditional vibrating screen control system adopts a preset parameter control method, determines parameters such as vibration frequency, amplitude, and screen surface inclination according to a specific coal type during the equipment commissioning stage, and keeps these parameters unchanged during the production process. This control method calculates key parameters through empirical formulas and can meet the basic production requirements and achieve the screening of coal materials when dealing with a single coal type and stable working conditions.

[0003] However, with the diversification of coal production processes, the characteristics of coal materials to be processed by vibrating screens are becoming increasingly complex and variable, and the fluctuation ranges of characteristic parameters such as hardness, moisture content, and particle size distribution are constantly expanding. The fixed parameter control method shows obvious deficiencies in the face of such variable coal type characteristics, mainly manifested in three aspects: First, it cannot automatically adjust control parameters according to the changes in coal type characteristics, resulting in a significant decrease in screening efficiency during the coal type conversion process; second, the energy utilization efficiency is low, and the same vibration parameters are used for coal types with different characteristics, causing energy waste; third, there is a lack of real-time monitoring and feedback mechanism for material characteristics and equipment status, and it is unable to respond in a timely manner to screening abnormalities caused by changes in working conditions.

[0004] A deeper technical problem lies in how to effectively identify the characteristics of variable coal types and establish an accurate dynamic control model. Although some sensors are introduced for monitoring in the existing semi-automatic control systems, they lack the ability of multi-source data fusion and cannot comprehensively characterize the characteristics of coal materials. At the same time, the vibrating screen system itself is highly non-linear and coupled, and traditional control strategies are difficult to handle the collaborative optimization problems of multiple objectives such as vibration direction, intensity, and uniformity. In addition, the screening efficiency evaluation and feedback adjustment mechanism are not perfect. Especially when dealing with coal types with high moisture content or wide particle size distribution, it is impossible to achieve precise matching and dynamic adjustment of screening parameters, resulting in large fluctuations in screening process efficiency, high energy consumption, and frequent screen mesh blockages. Summary of the Invention

[0005] This application provides a vibration screen parameter adaptive control method and system based on machine learning, which is used to solve the technical problems that the fixed parameter control of traditional vibrating screens cannot adapt to the changes of multiple coal types, the vibrating screen control system lacks the ability of real-time perception and analysis of coal material characteristics, and there is a lack of accurate mapping and dynamic optimization mechanism between vibration parameters and screening efficiency.

[0006] In a first aspect, the present application provides a method for adaptive control of vibrating screen parameters based on machine learning. The method for adaptive control of vibrating screen parameters based on machine learning includes: collecting data on coal materials at the feeding end of the vibrating screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface to obtain a set of characteristic parameters; calculating the current coal load coefficient and screening difficulty index according to the set of characteristic parameters, and adjusting the phase difference and current input of the double-motor vibrator through a control system to obtain the vibration direction and intensity; adjusting the pre-tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity to obtain the screening trajectory and the material movement speed; and controlling the feeding speed of the feeder and the height position of the longitudinal baffle of the screen box according to the screening efficiency data detected by the undersize and oversize samplers.

[0007] Optionally, the collecting data on coal materials at the feeding end of the vibrating screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface to obtain a set of characteristic parameters includes: Obtaining a multi-angle continuous image sequence of the coal surface through the image acquisition device, identifying the coal block boundaries in the image sequence to obtain a coal particle size distribution characteristic matrix; Performing multi-point scanning on the coal materials using a near-infrared spectral sensor array installed at the front end of the screen surface, extracting the characteristic spectral peak after eliminating background noise by wavelet transform to obtain a coal moisture content distribution map and a mineral composition fingerprint; Separating the signals collected by the strain type weight sensor installed at the bottom of the screen box and the triaxial acceleration sensor on the side wall of the screen box through an adaptive filter to obtain a dynamic load characteristic curve of the coal materials; Reducing the dimension of the coal particle size distribution characteristic matrix, the coal moisture content distribution map and mineral composition fingerprint, and the dynamic load characteristic curve of the coal materials, and performing feature fusion through non-linear embedding to generate a set of characteristic parameters.

[0008] Optionally, the calculating the current coal load coefficient and screening difficulty index according to the set of characteristic parameters, and adjusting the phase difference and current input of the double-motor vibrator through a control system to obtain the vibration direction and intensity includes: Multiplying the particle size distribution characteristic matrix in the set of characteristic parameters by a preset weight coefficient, summing the weighted values of all particle size intervals, and multiplying by the spatial average value of the moisture content distribution map to obtain a first coal load coefficient; Calculating the ratio of the clay mineral content to the hard mineral content based on the mineral composition fingerprint and the dynamic load characteristic curve in the set of characteristic parameters, and generating a first screening difficulty index in combination with the fluctuation coefficient; Apply a non - linear mapping function to the first coal load coefficient and perform weighted fusion with the first screening difficulty index to obtain a second screening difficulty index and a second coal load coefficient; Construct a parameter mapping table according to the second coal load coefficient and the second screening difficulty index, and calculate the phase difference and current input value of the double - motor vibrator through an interpolation algorithm; Convert the phase difference into a phase control signal, convert the current input value into a frequency control signal, and send them to the motor drive unit through a controller to generate a vibration direction and vibration intensity; Use vibration sensors installed on the screen box to measure actual vibration parameters, compare them with the theoretically calculated values, calculate the compensation amount through closed - loop control, and adjust the vibration direction and vibration intensity.

[0009] Optionally, adjusting the pre - tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity to obtain a screening trajectory and a material movement speed, including: Decompose the three - axis acceleration signal in the feedback data of the vibration direction and intensity into horizontal and vertical components, calculate the ratio and phase difference of the two components, judge the elliptical shape of the current vibration trajectory, and trigger the adjustment of the screen surface inclination when the horizontal - vertical ratio deviates from the target value; Conduct a time - domain comparison of the vibration signals of the four support points of the screen box, measure the time difference of the vibration peak values of each support point, judge the uniformity of the screen box vibration, and send an adjustment instruction to the left - side or right - side spring pre - tightening device when the vibration time difference between the left and right support points exceeds a preset threshold; According to the difference between the current vibration trajectory and the target trajectory, control the hydraulic system to extend or shorten the screen surface inclination hydraulic cylinder to change the angle between the screen surface and the horizontal plane until the screen surface inclination reaches the calculated optimal angle; Drive the spring pre - tightening screw to rotate through a servo motor to adjust the spring support stiffness so that the vibration intensity of the screen box reaches the optimal value matching the current coal type until the movement speed of the material on the screen surface meets the preset requirements.

[0010] Optionally, the time - domain comparison of the vibration signals of the four support points of the screen box, measuring the time difference of the vibration peak values of each support point, judging the uniformity of the screen box vibration, and sending an adjustment instruction to the left - side or right - side spring pre - tightening device when the vibration time difference between the left and right support points exceeds a preset threshold, includes: Collect the data of the acceleration sensors installed at the positions of the four support points of the screen box, namely the left - front, right - front, left - rear, and right - rear, perform low - pass filtering on the original data, and extract the peak - point timestamps within each signal period; Calculate the difference between the peak - point timestamps of the diagonal support points to obtain the torsional vibration parameters of the screen box. When the torsional vibration parameters are greater than the first preset threshold, increase the pre - tightening force of the spring on the side with lower torsional stiffness; Calculate the difference in peak timestamps of the same-side fulcrums to obtain the screening box swing vibration parameter. When the swing vibration parameter is greater than the second preset threshold, increase the pre-tightening force of the spring on the side with larger swing. Input the vibration amplitude data of the four fulcrums of the screening box into the unbalance calculation unit to generate the screening box vibration uniformity index. When the screening box vibration uniformity index exceeds the third preset threshold, calculate the adjustment amount of the pre-tightening force of each fulcrum spring. Send the calculated adjustment amount of the spring pre-tightening force to the servo motor controllers of each fulcrum, and drive the servo motor to rotate forward or backward to adjust the spring compression amount until the vibration time difference of the four fulcrums is less than the fourth preset threshold.

[0011] Optionally, the inputting the vibration amplitude data of the four fulcrums of the screening box into the unbalance calculation unit to generate the screening box vibration uniformity index includes: Normalize the vibration amplitude data of the four fulcrums of the screening box, namely the left front, right front, left rear, and right rear fulcrums, to unify the numerical range of each fulcrum and obtain the normalized vibration amplitude set. Calculate the difference between the maximum value and the minimum value in the normalized vibration amplitude set, and divide it by the average value of the vibration amplitudes of the four fulcrums to obtain the first screening box vibration unbalance. Conduct variance analysis on the vibration amplitudes of the four fulcrums, calculate the ratio of the standard deviation to the mean value to obtain the coefficient of variation of the vibration amplitude, and use the coefficient of variation as the second screening box vibration unbalance. Synthesize the first screening box vibration unbalance and the second screening box vibration unbalance through weighted averaging to obtain the screening box vibration uniformity index. According to the numerical range of the screening box vibration uniformity index, divide the vibration state into a uniform zone, a slightly uneven zone, a moderately uneven zone, and a severely uneven zone, and adopt corresponding spring pre-tightening force adjustment strategies for different zones.

[0012] Optionally, the controlling the feeding speed of the feeder and the height position of the longitudinal baffle of the screening box according to the screening efficiency data detected by the undersize and oversize samplers includes: Collect material samples through the oversize sampler installed at the end of the screen surface and the undersize sampler at the discharge port of the screening machine, conduct screening particle size analysis and quality determination on the samples, and calculate the screening efficiency value of the current screening process. Compare the screening efficiency value with the preset target efficiency threshold. When the screening efficiency value deviates from the preset target efficiency threshold by more than the first deviation range, generate a screening state evaluation result. Based on the screening state evaluation result, control the frequency conversion device of the feeder to adjust the feeding speed, change the residence time of the material on the screen surface, and at the same time control the hydraulic cylinder to adjust the height position of the longitudinal baffle of the screening box to adjust the material layer thickness. Monitor the screening efficiency for multiple consecutive cycles under the adjusted working conditions, calculate the change slope of the screening efficiency. When the absolute value of the change slope is less than the stability determination threshold and the efficiency value is within the second deviation range of the preset target efficiency threshold, lock the current feeding speed and the position of the baffle height. When it is detected that the screening efficiency value exceeds the third deviation range of the preset target efficiency threshold or the particle size distribution of the undersize material is abnormal, trigger the start signal of the screen cleaning mechanism and temporarily adjust the vibration intensity parameter to the maximum limit value.

[0013] In a second aspect, the present application provides a vibration screen parameter adaptive control system based on machine learning. The vibration screen parameter adaptive control system based on machine learning includes: An acquisition module for collecting data on coal materials at the feeding end of the vibration screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface to obtain a set of characteristic parameters. A control module for calculating the current coal load coefficient and screening difficulty index according to the set of characteristic parameters, and adjusting the phase difference and current input of the double-motor vibrator through the control system to obtain the vibration direction and intensity. An adjustment module for adjusting the pre-tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity to obtain the screening trajectory and the material movement speed. A feeding module for controlling the feeding speed of the feeder and the position of the longitudinal baffle of the screen box according to the screening efficiency data detected by the undersize material and oversize material samplers.

[0014] In a third aspect, there is provided a vibration screen parameter adaptive control device based on machine learning, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the vibration screen parameter adaptive control device based on machine learning executes the above-mentioned vibration screen parameter adaptive control method based on machine learning.

[0015] In a fourth aspect, there is provided a computer-readable storage medium in which instructions are stored. When it runs on a computer, it causes the computer to execute the above-mentioned vibration screen parameter adaptive control method based on machine learning.

[0016] In the technical solution provided by this application, by installing an image acquisition device above the feed hopper and deploying a material sensor at the front end of the sieve surface, multi-dimensional real-time perception of the characteristics of coal materials is achieved, solving the technical problem that the traditional vibrating screen control system lacks accurate identification of material characteristics; the construction of the characteristic parameter set adopts deep learning image processing and multi-source heterogeneous data fusion technology to accurately extract and characterize key characteristics such as coal particle size distribution, moisture content, and mineral composition, significantly improving the accuracy and stability of coal type identification; based on the coal load coefficient and screening difficulty index calculated from the characteristic parameter set, an accurate mapping relationship is established between the physical characteristics of coal and the vibration control parameters, enabling the phase difference and current input of the double-motor vibrator to be differentially adjusted according to the characteristics of different coal types, effectively solving the core technical problem that the fixed parameter control method cannot adapt to the changes of multiple coal types; the closed-loop feedback mechanism of the vibration direction and intensity realizes precise control of the screening trajectory through the coordinated adjustment of the spring pre-tightening force and the sieve surface inclination angle, ensuring that the movement speed of the material on the sieve surface always remains in the optimal range and overcoming the problem of screening efficiency fluctuations caused by coal type changes; the final screening efficiency monitoring and the adjustment links of the feeding speed and the baffle height construct a complete quality closed-loop control system, enabling the entire screening system to dynamically respond to coal type changes and maintain efficient and stable operation.

[0017] This invention adopts a multi-level machine learning algorithm in the field of coal processing, especially the contributions of the algorithm to each link are significant: the deep convolutional neural network in image processing can accurately identify the coal block boundary from a complex background, solving the problem that the traditional image recognition algorithm is vulnerable to interference in a harsh industrial environment; the non-linear embedding algorithm for multi-source data fusion retains the topological relationship between features of different dimensions, effectively solving the difficulty of characterizing the characteristics of multiple coal types; the parameter optimization mechanism based on the improved grey wolf optimization algorithm introduces the adjustment of the material adaptability weight, greatly improving the global search ability in a complex non-linear system and solving the defect that the traditional optimization method is prone to falling into local optima; the multi-objective evaluation system in the spring pre-tightening force adjustment can balance the contradiction between vibration uniformity and energy consumption, meeting the dual requirements of equipment reliability and economy in the coal screening process. In summary, through the system integration of material characteristic perception, parameter optimization control, vibration state monitoring, and quality closed-loop adjustment, this invention realizes the adaptive control of the vibrating screen for multiple coal types and solves the key problem that the parameters are fixed in the traditional technology and cannot cope with the changes of coal types. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a schematic diagram of an embodiment of the vibration screen parameter adaptive control method based on machine learning in the embodiment of the present application; Figure 2 It is a schematic diagram of an embodiment of the vibration screen parameter adaptive control system based on machine learning in the embodiment of the present application; Figure 3 It is a structural schematic block diagram of the vibration screen parameter adaptive control device based on machine learning in the embodiment of the present invention. Specific embodiments

[0020] The embodiment of the present application provides a vibration screen parameter adaptive control method and system based on machine learning. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiment of the present application will be described below. Please refer to Figure 1 , an embodiment of the vibration screen parameter adaptive control method based on machine learning in the embodiment of the present application includes: Step S101: Collect data on the coal material at the feed end of the vibration screen through an image acquisition device installed above the feed hopper and a material sensor installed at the front end of the screen surface to obtain a characteristic parameter set; Step S102: Calculate the current coal load coefficient and screening difficulty index according to the characteristic parameter set, and adjust the phase difference and current input of the double-motor vibrator through the control system to obtain the vibration direction and intensity; Step S103: Based on the feedback data of the vibration direction and intensity, adjust the pre-tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder to obtain the screening trajectory and the material movement speed; Step S104: Control the feeding speed of the feeder and the height position of the longitudinal baffle of the screen box according to the screening efficiency data detected by the under-screen and over-screen samplers.

[0022] It can be understood that the execution entity of this application can be an adaptive control system for vibrating screen parameters based on machine learning, or it can also be a terminal or a server. Specifically, it is not limited here. In the embodiments of this application, the server is taken as an example of the execution entity for illustration.

[0023] Specifically, coal material characteristic data is collected through multi-source sensors. The high-definition camera device installed above the feed hopper acquires coal surface images at a rate of 30 frames per second. The improved Canny edge detection algorithm is used to identify the boundaries of coal blocks, and the proportion distribution in different particle size ranges is calculated through region segmentation to form a particle size distribution characteristic matrix. The near-infrared spectral sensor array at the front end of the screen detects the spectral reflection characteristics of coal. After using wavelet transform to eliminate background noise, the peak values of characteristic bands are extracted. These peak values are directly related to the moisture content and mineral composition of coal, generating a moisture content distribution map and a mineral composition fingerprint. At the same time, the strain type weight sensor at the bottom of the screen box measures the change in material load, and the three-axis acceleration sensor on the side wall of the screen box records the vibration state. The effective signal and noise are separated through an adaptive filter to obtain the dynamic load characteristic curve of coal materials. These multi-source heterogeneous data are fused through dimensionality reduction and non-linear embedding to generate a comprehensive set of characteristic parameters. When calculating the control index based on the set of characteristic parameters, the particle size distribution characteristic matrix is multiplied by the preset weight coefficient, and the sum is then multiplied by the average moisture content to obtain the first coal load coefficient. At the same time, the ratio of the content of clay minerals to hard minerals is extracted from the mineral composition fingerprint, combined with the dynamic load fluctuation coefficient, to generate the first screening difficulty index. These two primary indices are fused through a non-linear mapping function and weighting to form the second coal load coefficient and the second screening difficulty index, with the latter more accurately reflecting the screenability of the material. Based on these indices, a parameter mapping table is constructed, the optimal dual-motor parameters are calculated, and converted into control signals. For example, when processing high-hardness and low-moisture coal, the calculated load coefficient is relatively high, and the system automatically increases the vibration frequency to 45 Hz and reduces the phase difference to 15 degrees to provide stronger vibration energy; while when processing low-hardness and high-moisture coal, the system reduces the vibration frequency to 35 Hz and increases the phase difference to 30 degrees to enhance the material dispersion effect. After the vibration control parameters are executed, the system continuously monitors the vibration effect. The data from the three-axis acceleration sensor is decomposed into horizontal and vertical components, and the elliptical shape of the vibration trajectory is judged by calculating the ratio and phase difference. The vibration signals of the four support points of the screen box are compared in the time domain, and the time difference of the vibration peak values of each support point is measured to evaluate the vibration uniformity. When it is detected that the horizontal-vertical ratio deviates from the target or the vibration of the support points is uneven, the system automatically adjusts the pre-tightening force of the support springs of the screen box and the inclination angle of the screen surface. The pre-tightening force of the springs is adjusted by driving the pre-tightening screw with a servo motor, and the inclination angle of the screen surface is controlled by the telescopic cylinder. This dual-parameter linkage adjustment ensures that the material forms the best movement trajectory and speed on the screen surface, avoiding the problem of unstable material movement in traditional fixed-parameter control. The system evaluates the screening effect in real time through the oversize sampler installed at the end of the screen surface and the undersize sampler at the discharge port of the screening machine. After the sampler collects the material sample, particle size analysis is carried out to calculate the current screening efficiency value. This value is compared with the preset target threshold to generate the screening state evaluation result. Based on the evaluation result, the control system adjusts the operating frequency of the variable frequency device of the feeder and the height position of the longitudinal baffle of the screen box to change the material feeding speed and the thickness of the material on the screen surface.For example, when the screening efficiency is detected to be lower than the target value, the system reduces the feeding speed and raises the height of the baffle to extend the residence time of the material on the screen surface; when the screening efficiency returns to the target range, the current parameter settings are locked. This closed-loop control mechanism can cope with the fluctuations in screening difficulty caused by changes in coal types, maintain stable production quality and efficiency, and solve the problem of reduced efficiency during coal type conversion in traditional vibrating screens.

[0024] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Obtain a multi-angle continuous image sequence of the coal surface through an image acquisition device, identify the coal block boundaries of the image sequence, and obtain a coal particle size distribution feature matrix; Use the near-infrared spectral sensor array installed at the front end of the screen to perform multi-point scanning on the coal material, extract the characteristic spectral peak after eliminating the background noise by wavelet transform, and obtain the coal moisture content distribution map and the mineral composition fingerprint; Separate the signals collected by the strain type weight sensor installed at the bottom of the screen box and the three-axis acceleration sensor on the side wall of the screen box through an adaptive filter to obtain the dynamic load characteristic curve of the coal material; Reduce the dimensions of the coal particle size distribution feature matrix, the coal moisture content distribution map, the mineral composition fingerprint, and the dynamic load characteristic curve of the coal material, and perform feature fusion through non-linear embedding to generate a set of characteristic parameters.

[0025] Specifically, an industrial camera array installed above the feed hopper captures coal surface images from different angles at a collection frequency of 30 frames per second and a resolution of 1920×1080 pixels. After the acquired original images are subjected to grayscale conversion, an improved Canny edge detection algorithm is applied to identify the boundaries of coal blocks. This algorithm detects strong edges and weak edges through a double-threshold method, and the thresholds are dynamically determined according to the image histogram, avoiding the problem of edge loss caused by fixed thresholds. After edge connection, a morphological closing operation is performed to fill the edge gaps and obtain the complete coal block contour. Based on the contour information, the equivalent diameter of each coal block is calculated, and the number distribution in different particle size ranges is statistically analyzed to form an n×m-dimensional particle size distribution feature matrix, where n represents the number of particle size grades (usually 5 - 8 grades), and m represents the length of the time series. At the same time, a near-infrared spectral sensor array installed at the front end of the screen scans the coal material at a frequency of 100Hz, collecting reflected spectral data in the wavelength range of 900 - 1700nm. The original spectral data is interfered by ambient light and random noise, and wavelet transform is used for noise reduction processing. Specifically, the db4 wavelet basis is selected, and the spectral data is decomposed into 6 layers to obtain wavelet coefficients of different frequencies. The high-frequency coefficients are processed by the soft-threshold method, retaining the low-frequency information reflecting the material characteristics and removing high-frequency noise. After wavelet reconstruction, the purified spectral data is obtained, and the peak and valley positions of the characteristic bands are extracted. The characteristic peaks related to moisture content are mainly located near 1450nm and 1940nm, and the characteristic peaks related to mineral composition are distributed in the region of 1000 - 1300nm. Through the characteristic peak intensity ratio and displacement amount, the moisture content distribution map and mineral composition fingerprint of the coal are calculated.

[0026] To obtain the dynamic load characteristics of the coal material, a strain-type weight sensor at the bottom of the screen box and a three-axis acceleration sensor on the side wall collect data synchronously at a sampling frequency of 1000Hz. These two signals are mixed with material load information, equipment vibration information, and environmental interference. Adaptive filtering technology is used to separate the effective signals. Specifically, the least mean square error (LMS) adaptive filtering algorithm is used. With the acceleration signal as the reference input and the weight signal as the main input, the filter coefficients are iteratively adjusted to extract the effective components of the material load's impact on the equipment. The filtered signal undergoes envelope detection to obtain the material dynamic load change curve, which reflects the fluidity and resistance characteristics of the material during the screening process.

[0027] Fuse the obtained multi-source heterogeneous data (particle size distribution feature matrix, moisture content distribution map, mineral composition fingerprint, and dynamic load characteristic curve). Since the dimensions and measurement units of various types of data are different, first perform standardization processing on each type of data to map all data to the interval [0, 1]. Then, use the principal component analysis (PCA) method to reduce the dimensionality of the high-dimensional data and select the principal components with a cumulative contribution rate exceeding 95%. Finally, use the t-SNE non-linear embedding algorithm to project the data into a low-dimensional feature space, retain the topological relationship between the data, and generate a compact set of characteristic parameters. This processing method solves the problem of insufficient perception of material characteristics in traditional vibrating screen control, enables the control system to obtain rich material information, and provides a data basis for subsequent intelligent decision-making. Taking a vibrating screen in a coal mine as an example, when the coal type being processed changes from anthracite to bituminous coal, the hardness index in the set of characteristic parameters decreases significantly, and the moisture content index increases. Based on this, the system automatically adjusts the vibration parameters to keep the screening efficiency stable.

[0028] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Multiply the particle size distribution feature matrix in the set of characteristic parameters by a preset weight coefficient, sum the weighted values of all particle size intervals, and multiply by the spatial average value of the moisture content distribution map to obtain the first coal load coefficient; Based on the mineral composition fingerprint and dynamic load characteristic curve in the set of characteristic parameters, calculate the ratio of the clay mineral content to the hard mineral content, and generate the first screening difficulty index in combination with the fluctuation coefficient; Apply a non-linear mapping function to the first coal load coefficient and perform weighted fusion with the first screening difficulty index to obtain the second screening difficulty index and the second coal load coefficient; Construct a parameter mapping table based on the second coal load coefficient and the second screening difficulty index, and calculate the phase difference and current input value of the double-motor vibrator through an interpolation algorithm; Convert the phase difference into a phase control signal, convert the current input value into a frequency control signal, and send them to the motor drive unit through a controller to generate the vibration direction and vibration intensity; Use the vibration sensor installed on the screen box to measure the actual vibration parameters, compare them with the theoretically calculated values, calculate the compensation amount through closed-loop control, and adjust the vibration direction and vibration intensity.

[0029] Specifically, multiply the particle size distribution characteristic matrix by preset weight coefficients, which reflect the influence degree of different particle sizes on the screening difficulty. Coarse particle sizes (such as >10 mm) have lower weights, while fine particle sizes (such as <3 mm) have higher weights because fine materials are more likely to clog the sieve mesh. Sum up the weighted data of each particle size level to obtain a comprehensive particle size index, and then multiply it by the spatial average value of the moisture content distribution map to form the first coal load coefficient. The higher the moisture content, the worse the fluidity of the material, and the corresponding increase in the load coefficient. At the same time, extract the data of clay mineral content and hard mineral content from the mineral composition fingerprint. Clay minerals (such as kaolinite, montmorillonite) make the coal more adhesive, while hard minerals (such as quartz, pyrite) increase the screening resistance. After calculating the ratio of the two and combining it with the fluctuation coefficient of the dynamic load characteristic curve, the fluctuation coefficient represents the severity of the load change and reflects the stability of the material movement on the sieve surface. These two indicators generate the first screening difficulty index through weighted summation.

[0030] The first coal load coefficient is non-linearly mapped through the Sigmoid function, which converts the original value into a value within the range of 0-1, making the control smoother and highlighting the differences in the medium load area. The mapped load coefficient and the first screening difficulty index are weighted and fused, and the weights are dynamically adjusted according to the current focus of the screening task to form a more accurate second screening difficulty index and a second coal load coefficient. Based on these two secondary indicators, a two-dimensional parameter mapping table is constructed, with the load coefficient on the horizontal axis and the difficulty index on the vertical axis. The optimal vibration parameters corresponding to each combination are stored in the table. Use the bilinear interpolation algorithm to obtain the optimal phase difference and current input value of the double-motor vibrator under the current working conditions from the parameter mapping table. The phase difference directly determines the ellipticity of the vibration direction, and the current input value affects the vibration intensity.

[0031] The control system converts the phase difference value into a digital instruction for the phase controller to control the timing trigger time difference between the two motors; at the same time, it converts the current input value into a frequency control signal for the frequency converter to adjust the motor speed and output torque. These control signals are transmitted to the motor drive unit through the fieldbus to generate the expected vibration direction and vibration intensity. The three-axis vibration sensor installed on the sieve box collects vibration data in real time, calculates the actual vibration elliptical trajectory parameters and energy distribution. Compare with the theoretical calculation value to obtain the deviation. The control system calculates the compensation amount through the proportional-integral-differential (PID) algorithm, dynamically adjusts the control output, eliminates the influence of external interference and model error, and stabilizes the vibration parameters in the optimal area. For example, when it is detected that the vibration intensity decreases due to material accumulation, the system automatically increases the current input value. When it is found that the vibration direction deviates from the optimal trajectory, the phase difference is finely adjusted to ensure the highest screening efficiency is always maintained.

[0032] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Decompose the three-axis acceleration signal in the feedback data of the vibration direction and intensity into horizontal and vertical components, calculate the ratio and phase difference of the two components, judge the elliptical shape of the current vibration trajectory, and trigger the adjustment of the screen inclination when the horizontal-vertical ratio deviates from the target value; Conduct a time-domain comparison of the vibration signals at the four support points of the screen box, measure the time difference of the vibration peaks at each support point, judge the uniformity of the screen box vibration, and send an adjustment instruction to the left or right spring pre-tightening device when the vibration time difference between the left and right support points exceeds the preset threshold; According to the difference between the current vibration trajectory and the target trajectory, control the hydraulic system to extend or shorten the screen inclination hydraulic cylinder, change the angle between the screen and the horizontal plane until the screen inclination reaches the calculated optimal angle; Drive the spring pre-tightening screw to rotate through the servo motor, adjust the spring support stiffness, and make the vibration intensity of the screen box reach the optimal value matching the current coal type until the movement speed of the material on the screen meets the preset requirements.

[0033] Specifically, extract effective information from the vibration feedback data collected by the three-axis acceleration sensor, and decompose the three-dimensional acceleration signal (X, Y, Z) into horizontal component (X-Y plane) and vertical component (Z axis) through coordinate transformation. Apply a band-pass filter to the original signal to remove high-frequency noise and low-frequency drift, and retain the effective vibration signal within the range of 0.5 times to 2 times the vibration frequency. Calculate the amplitude ratio (H / V ratio) of the horizontal composite acceleration to the vertical acceleration. This ratio directly reflects the elliptical shape of the vibration trajectory - the larger the ratio, the flatter the ellipse, and the more the movement trajectory of the material on the screen tends to be horizontal. At the same time, extract the fundamental frequency phases of the two components through Fourier transform and calculate the phase difference. The phase difference determines the tilt angle of the elliptical trajectory. When the H / V ratio deviates from the target value (usually 1.2 - 1.8) calculated according to the characteristics of the current coal type by more than ±10%, trigger the automatic screen inclination adjustment mechanism. For the monitoring of vibration uniformity, the system collects the vibration signals at the four support points of the front left, front right, rear left, and rear right of the screen box. Use the peak detection algorithm to determine the peak time of each support point signal, and calculate the time interval between adjacent periods to ensure the signal period stability. Then, taking the front left support point as the reference, calculate the time differences between the peak appearance times of the other three support points and the reference point respectively, forming three time differences. Ideally, the time difference between the diagonal support points should be half a period (corresponding to a 180-degree phase difference), and the time difference between the same-side support points should be close to zero. When the absolute value of the vibration time difference between the left and right support points exceeds the preset threshold (usually 5% of the vibration period), it indicates that the screen box has torsional non-uniform vibration and the spring pre-tightening force on the corresponding side needs to be adjusted. Specifically, if the left support point is ahead of the right support point, increase the spring pre-tightening force on the right side; otherwise, increase the pre-tightening force on the left side.

[0034] According to the analysis results of the vibration trajectory, when the inclination angle of the screen surface needs to be adjusted, the control system first calculates the optimal angle adjustment amount. The calculation is based on the deviation between the material characteristics and the current H / V ratio, and the optimal inclination angle value is predicted through a pre-trained neural network model. The control signal is converted into a 4-20mA standard current signal, and the hydraulic system is controlled through a proportional valve to drive the extension or shortening of the hydraulic cylinder for the inclination angle of the screen surface. The displacement of the hydraulic cylinder is real-time feedback through a position sensor to form a closed-loop control system, accurately adjusting the angle between the screen surface and the horizontal plane. The hydraulic system adopts a slow start-stop strategy to avoid impact during the angle adjustment process until the inclination angle of the screen surface reaches the calculated optimal angle, with the general accuracy controlled within the range of ±0.5 degrees.

[0035] Based on the analysis results of the vibration uniformity of the fulcrums, the control system sends an adjustment instruction to the spring pre-tightening device. The spring pre-tightening device consists of a servo motor, a reducer, and a lead screw mechanism. The servo motor receives a PWM control signal and drives the pre-tightening screw to rotate forward or backward to change the compression amount of the spring, thereby adjusting the spring support stiffness. The adjustment amount of the pre-tightening force is proportional to the degree of vibration non-uniformity. A segmented adjustment strategy is adopted. After each adjustment, wait for 5-10 vibration cycles to observe the effect and then decide whether to continue the adjustment. When the vibration parameters are consistent with the preset target and the movement speed of the material on the screen meets the requirements, lock the current setting.

[0036] In a specific embodiment, the process of performing the step of comparing the vibration signals of the four fulcrums of the screen box in the time domain may specifically include the following steps: Collect the data of the acceleration sensors installed at the positions of the four fulcrums of the left front, right front, left rear, and right rear of the screen box, perform low-pass filtering on the original data, and extract the peak point timestamps within each signal cycle; Calculate the difference between the peak timestamps of the diagonal fulcrums to obtain the torsional vibration parameters of the screen box. When the torsional vibration parameters are greater than the first preset threshold, increase the pre-tightening force of the spring on the side with lower torsional stiffness; Calculate the difference between the peak timestamps of the same-side fulcrums to obtain the rocking vibration parameters of the screen box. When the rocking vibration parameters are greater than the second preset threshold, increase the pre-tightening force of the spring on the side with larger rocking; Input the vibration amplitude data of the four fulcrums of the screen box into the unbalance calculation unit to generate the vibration uniformity index of the screen box. When the vibration uniformity index of the screen box exceeds the third preset threshold, calculate the adjustment amount of the spring pre-tightening force for each fulcrum; Send the calculated adjustment amount of the spring pre-tightening force to the servo motor controllers of each fulcrum to drive the servo motor to rotate forward or backward to adjust the spring compression amount until the vibration time difference of the four fulcrums is less than the fourth preset threshold.

[0037] Specifically, high-precision piezoelectric acceleration sensors are installed at the four fulcrums of the sieve box, namely the left front, right front, left rear, and right rear. The sampling frequency is set to 1000 Hz to ensure capturing the complete vibration waveform characteristics. The collected raw data contains mechanical noise and high-frequency interference, which is processed by a Butterworth low-pass filter with the cut-off frequency set to 3 times the vibration fundamental frequency, retaining the main vibration information while filtering out high-frequency noise. After filtering, peak detection is performed on the data. The sliding window method and dynamic threshold strategy are used to identify the peak points within each vibration cycle, and the timestamps when the peaks occur are recorded to form the time series data of the four fulcrums. Using the obtained peak timestamp data, the difference in the peak occurrence times of the diagonal fulcrums (left front and right rear, right front and left rear) is calculated. In an ideal state, the diagonal fulcrums should maintain a 180-degree phase difference, corresponding to half of the vibration cycle time. The deviation between the actually calculated time difference and the ideal value forms the torsional vibration parameter, which reflects the degree of torsion of the sieve box around the vertical axis. When the torsional vibration parameter exceeds the first preset threshold (usually 5% of the vibration cycle), it indicates that there is obvious torsional deformation of the sieve box, and the torsional stiffness needs to be adjusted. By analyzing the torsional direction, the side with lower torsional stiffness is determined, and the control system automatically sends an instruction to increase the pre-tightening force to the spring pre-tightening device on this side to balance the torsional effect.

[0038] The difference in the peak timestamps of the same-side fulcrums (left front and left rear, right front and right rear) is calculated to obtain the rocking vibration parameter of the sieve box. Ideally, the same-side fulcrums should vibrate synchronously, and the difference is close to zero. When the rocking vibration parameter exceeds the second preset threshold (usually 3% of the vibration cycle), it indicates that there is a front-back rocking phenomenon of the sieve box. The control system identifies the side with larger rocking and increases the pre-tightening force of the spring on this side to suppress excessive rocking. For the evaluation of vibration amplitude uniformity, the system extracts the vibration amplitude data of the four fulcrums, first performs normalization processing to eliminate the influence of sensor sensitivity differences. Then, the difference ratio between the maximum amplitude and the minimum amplitude, and the ratio of the standard deviation to the mean of the amplitudes at the four points are calculated. These two indicators together constitute the sieve box vibration uniformity index. When this index exceeds the third preset threshold, a comprehensive adjustment mechanism is triggered. The system calculates the adjustment amount based on the amplitude deviation of each fulcrum. The larger the deviation, the greater the adjustment amplitude, forming a spring pre-tightening force adjustment plan. The control system converts the calculated pre-tightening force adjustment amount into a servo motor control signal and transmits it to the servo drivers of each fulcrum through the PROFIBUS fieldbus to control the forward or reverse rotation of the servo motor. The motor drives the lead screw to rotate through a reduction mechanism to precisely adjust the spring compression amount and change the support stiffness. The system monitors the adjustment effect in real time. When the vibration time difference of the four fulcrums is less than the fourth preset threshold (lower than 2% of the vibration cycle) and the vibration amplitude deviation rate is less than 5%, the adjustment process is completed. This method solves the technical problem that traditional vibrating screens cannot accurately control vibration uniformity, and significantly improves the screening efficiency and the service life of the equipment.

[0039] In a specific embodiment, the process of inputting the vibration amplitude data of the four fulcrums of the screening box into the unbalance calculation unit may specifically include the following steps: Normalize the vibration amplitude data of the four fulcrums (front left, front right, rear left, and rear right) of the screening box to unify the numerical range of each fulcrum and obtain a set of normalized vibration amplitudes; Calculate the difference between the maximum and minimum values in the set of normalized vibration amplitudes, divide it by the average value of the vibration amplitudes of the four fulcrums, and obtain the first vibration unbalance of the screening box; Conduct an analysis of variance on the vibration amplitudes of the four fulcrums, calculate the ratio of the standard deviation to the mean value, obtain the coefficient of variation of the vibration amplitudes, and use the coefficient of variation as the second vibration unbalance of the screening box; Synthesize the first vibration unbalance of the screening box and the second vibration unbalance of the screening box through weighted averaging to obtain an index of the vibration uniformity of the screening box; According to the numerical range of the vibration uniformity index of the screening box, divide the vibration state into a uniform zone, a slightly uneven zone, a moderately uneven zone, and a severely uneven zone, and adopt corresponding spring pre-tightening force adjustment strategies for different zones.

[0040] Specifically, normalize the vibration amplitude data collected from the four fulcrums (front left, front right, rear left, and rear right) of the screening box to eliminate the systematic errors caused by differences in sensor sensitivity and changes in installation positions. The normalization uses the maximum-minimum normalization method, that is, subtract the minimum value of the four points from the original vibration amplitude of each fulcrum, and then divide it by the amplitude range of the four points (the maximum value minus the minimum value), so that the numerical values of each fulcrum are uniformly mapped to the 0-1 interval. This processing makes the vibration data at different positions and different times comparable, facilitating subsequent uniformity analysis. Based on the set of normalized vibration amplitudes, calculate the first vibration unbalance of the screening box. This index is obtained by dividing the difference between the maximum and minimum values by the average value, and intuitively reflects the range of the vibration amplitude distribution. A larger unbalance indicates that the vibration of some fulcrums is too strong or too weak, which may lead to uneven stress distribution in the screening box structure and accelerate equipment wear. At the same time, in order to comprehensively evaluate the vibration uniformity, conduct an analysis of variance on the vibration amplitudes of the four fulcrums, calculate the standard deviation and then divide it by the mean value to obtain the coefficient of variation of the vibration amplitudes, that is, the second vibration unbalance of the screening box. This coefficient more sensitively reflects the degree of deviation of each fulcrum from the average value and can detect relatively subtle uneven states.

[0041] Synthesize these two complementary unbalance indexes into a comprehensive uniformity index through weighted averaging. The weighting factor is dynamically adjusted according to the characteristics of coal materials - increase the weight of the first unbalance when processing high-hardness coal types (paying attention to extreme value control), and increase the weight of the second unbalance when processing high-moisture coal types (paying attention to overall uniformity). The synthesized uniformity index comprehensively reflects the vibration state of the screening box and serves as the basis for subsequent control decisions.

[0042] Based on a large amount of experimental data and expert experience, the uniformity index is divided into four intervals: uniform zone (0 - 0.1), slightly non-uniform zone (0.1 - 0.2), moderately non-uniform zone (0.2 - 0.35), and severely non-uniform zone (>0.35). For different zones, differential spring pre-tightening force adjustment strategies are adopted: no adjustment is required in the uniform zone; stability adjustment is adopted in the slightly non-uniform zone, and the pre-tightening force of the fulcrum with the lowest vibration amplitude is adjusted slightly (3 - 5%); contrast adjustment is adopted in the moderately non-uniform zone, reducing the pre-tightening force of the high-amplitude fulcrum (5 - 10%) while increasing the pre-tightening force of the low-amplitude fulcrum (5 - 10%); reconstruction adjustment is adopted in the severely non-uniform zone. First, the pre-tightening forces of the four fulcrums are uniformly reset to the standard value, and then the pre-tightening force difference is re-distributed according to the current material characteristics to make the vibration energy distribution more reasonable. This vibration uniformity evaluation and adjustment mechanism solves the technical problem that the vibration state cannot be accurately monitored and adjusted under the traditional fixed-parameter control method of vibrating screens, and realizes the precise matching of vibration parameters and coal type characteristics through data-driven adaptive control.

[0043] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Collect material samples through the oversize sampler installed at the end of the screen surface and the undersize sampler at the discharge port of the screening machine, conduct screening particle size analysis and quality determination on the samples, and calculate the screening efficiency value of the current screening process; Compare the screening efficiency value with the preset target efficiency threshold. When the screening efficiency value deviates from the preset target efficiency threshold by more than the first deviation range, generate a screening state evaluation result; Based on the screening state evaluation result, control the variable-frequency device of the feeder to adjust the feeding speed, change the residence time of the material on the screen surface, and at the same time control the hydraulic cylinder to adjust the height position of the longitudinal baffle of the screen box to adjust the material layer thickness; Monitor the screening efficiency for multiple consecutive cycles of the adjusted working condition, calculate the screening efficiency change slope. When the absolute value of the change slope is less than the stability determination threshold and the efficiency value is within the second deviation range of the preset target efficiency threshold, lock the current feeding speed and baffle height position; When it is detected that the screening efficiency value exceeds the third deviation range of the preset target efficiency threshold or the undersize particle size distribution is abnormal, trigger the start signal of the screen cleaning mechanism and temporarily adjust the vibration intensity parameter to the maximum limit value.

[0044] Specifically, samples of oversize and undersize materials are collected at regular intervals by a mechanical sampler installed at the end of the screening surface and an automatic sampling device at the discharge outlet of the screening machine. The sampling period is 5 minutes, and the sampling amount each time is about 500 grams. The collected samples are subjected to particle size classification by an automatic screening system, which consists of 5 layers of standard sieve meshes with corresponding pore sizes of 10 mm, 5 mm, 3 mm, 1 mm, and 0.5 mm respectively. The mass of each particle size fraction is measured through vibrating screening and weight sensors, and the screening efficiency value is calculated. The calculation formula is: Screening efficiency = (mass of the target particle size in the undersize material / total mass of the target particle size in the feed) × 100%. The target particle size is determined according to the production process requirements, and in coal processing, it is usually fine-grained materials smaller than 3 mm. The calculated screening efficiency value is compared with a preset target efficiency threshold, which is preset according to the coal type characteristics and production requirements, generally 85 - 95%. When the screening efficiency deviates from the target efficiency by more than the first deviation range (±5%), the status evaluation mechanism is triggered. The evaluation results are divided into three categories: too low efficiency (less than the target value - 5%), moderate efficiency (within the range of the target value ±5%), and too high efficiency (greater than the target value + 5%). The status evaluation results serve as the basis for subsequent parameter adjustment, and the state of too low efficiency needs to be focused on because it directly affects product quality and production efficiency.

[0045] Based on the evaluation results, the control system performs coordinated adjustment of two parameters. When it is detected that the efficiency is too low, the operating frequency of the variable frequency device of the feeder is simultaneously slowed down (reduced by 10 - 20%) and the height position of the longitudinal baffle of the screen box is increased (lifted by 10 - 30 mm). These two operations work together. On the one hand, it prolongs the residence time of the material on the screening surface, and on the other hand, it increases the thickness of the material layer, strengthens the mutual extrusion and friction of the materials, and promotes the passage of fine-grained materials through the sieve holes. Conversely, when the efficiency is too high (which may be accompanied by a reduction in output), the feeding speed is appropriately increased and the baffle height is reduced to increase the throughput.

[0046] After adjusting the parameters, the system enters the monitoring stabilization stage, continuously monitors the screening efficiency data for 5 - 8 cycles (about 25 - 40 minutes), and calculates the slope of the efficiency change. The slope is obtained through the linear regression method and represents the change trend of the screening efficiency over time. When the absolute value of the slope is less than the stability threshold (usually 0.2% / minute) and the efficiency value is within the second deviation range (±3%) of the target threshold, it is determined that the system has reached a stable state, and the current feeding speed and baffle height settings are locked. The system also has the ability to detect and handle abnormal states. When it detects that the screening efficiency suddenly drops by more than the third deviation range (±10%) or there is an obvious abnormal particle size distribution in the undersize material (such as a sudden increase in the proportion of large - sized materials), it is judged that the screen may be blocked. At this time, a signal to start the screen cleaning mechanism is automatically triggered, and at the same time, the vibration intensity is temporarily adjusted to the maximum limit value (increased by 15 - 25%) and restored to the normal setting after 30 - 60 seconds. This short - term strong vibration helps to dredge the blocked screen holes and restore the normal screening state.

[0047] The above describes the vibration screen parameter adaptive control method based on machine learning in the embodiments of the present application. Next, the vibration screen parameter adaptive control system based on machine learning in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the vibration screen parameter adaptive control system based on machine learning in the embodiments of the present application includes: An acquisition module 201, configured to collect data on coal materials at the feeding end of the vibration screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface, to obtain a characteristic parameter set; A control module 202, configured to calculate the current coal load coefficient and screening difficulty index according to the characteristic parameter set, and adjust the phase difference and current input of the double - motor vibrator through the control system to obtain the vibration direction and intensity; An adjustment module 203, configured to adjust the pre - tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity, to obtain the screening trajectory and the material movement speed; A feeding module 204, configured to control the feeding speed of the feeder and the height position of the longitudinal baffle of the screen box according to the screening efficiency data detected by the undersize and oversize samplers.

[0048] Above Figure 2 The vibration screen parameter adaptive control system based on machine learning in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the vibration screen parameter adaptive control device based on machine learning in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0049] Refer to Figure 3, in the embodiments of the present invention, an adaptive control device for vibrating screen parameters based on machine learning is further provided. The adaptive control device for vibrating screen parameters based on machine learning can be a server, and its internal structure can be as Figure 3 shown. The adaptive control device for vibrating screen parameters based on machine learning includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the adaptive control device for vibrating screen parameters based on machine learning includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the adaptive control device for vibrating screen parameters based on machine learning is used to store the corresponding data in this embodiment. The network interface of the adaptive control device for vibrating screen parameters based on machine learning is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0050] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the adaptive control device for vibrating screen parameters based on machine learning to which the solution of the present invention is applied.

[0051] The present invention also provides a computer-readable storage medium. The computer-readable storage medium can be a non-volatile computer-readable storage medium, and the computer-readable storage medium can also be a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the steps of the adaptive control method for vibrating screen parameters based on machine learning.

[0052] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0053] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a machine learning-based vibrating screen parameter adaptive control device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0054] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A vibration sieve parameter adaptive control method based on machine learning, characterized in that, The method includes: Collecting data on the coal material at the feeding end of the vibrating screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface to obtain a set of characteristic parameters; Calculating the current coal load coefficient and screening difficulty index according to the set of characteristic parameters, and adjusting the phase difference and current input of the double-motor vibrator through a control system to obtain the vibration direction and intensity; Based on the feedback data of the vibration direction and intensity, adjusting the pre-tightening force of the screen box support spring and the telescopic amount of the screen surface inclination hydraulic cylinder to obtain the screening trajectory and the material movement speed; Controlling the feeding speed of the feeder and the height position of the longitudinal baffle of the screen box according to the screening efficiency data detected by the under-screen material and over-screen material samplers.

2. The machine learning-based vibration screen parameter adaptive control method according to claim 1, wherein The collecting data on the coal material at the feeding end of the vibrating screen through an image acquisition device installed above the feeding hopper and a material sensor installed at the front end of the screen surface to obtain a set of characteristic parameters includes: Obtaining a continuous image sequence of multiple angles of the coal surface through the image acquisition device, identifying the coal block boundaries of the image sequence to obtain a coal particle size distribution characteristic matrix; Performing multi-point scanning on the coal material using a near-infrared spectroscopy sensor array installed at the front end of the screen surface, extracting the characteristic spectral peak after eliminating background noise by wavelet transform to obtain a coal moisture content distribution map and a mineral composition fingerprint; Separating the signals collected by the strain-type weight sensor installed at the bottom of the screen box and the triaxial acceleration sensor on the side wall of the screen box through an adaptive filter to obtain a coal material dynamic load characteristic curve; Reducing the dimension of the coal particle size distribution characteristic matrix, the coal moisture content distribution map and mineral composition fingerprint, and the coal material dynamic load characteristic curve, and performing feature fusion through non-linear embedding to generate a set of characteristic parameters.

3. The vibration screen parameter adaptive control method based on machine learning according to claim 1, characterized in that, The calculating the current coal load coefficient and screening difficulty index according to the set of characteristic parameters, and adjusting the phase difference and current input of the double-motor vibrator through a control system to obtain the vibration direction and intensity includes: Multiplying the particle size distribution characteristic matrix in the set of characteristic parameters by a preset weight coefficient, summing the weighted values of all particle size intervals, and multiplying by the spatial average value of the moisture content distribution map to obtain a first coal load coefficient; Based on the mineral composition fingerprint and the dynamic load characteristic curve in the set of characteristic parameters, calculating the ratio of the clay mineral content to the hard mineral content, and generating a first screening difficulty index in combination with the fluctuation coefficient; Applying a non-linear mapping function to the first coal load coefficient, and performing weighted fusion with the first screening difficulty index to obtain a second screening difficulty index and a second coal load coefficient; Constructing a parameter mapping table according to the second coal load coefficient and the second screening difficulty index, and calculating the phase difference and current input values of the double-motor vibrator through an interpolation algorithm; Converting the phase difference into a phase control signal, converting the current input value into a frequency control signal, and sending them to the motor drive unit through a controller to generate the vibration direction and vibration intensity; Measuring the actual vibration parameters using a vibration sensor installed on the screen box, comparing with the theoretical calculated values, calculating the compensation amount through closed-loop control, and adjusting to the vibration direction and vibration intensity.

4. The vibration sieve parameter adaptive control method based on machine learning according to claim 1, characterized in that Adjust the pre-tightening force of the vibrating screen box support spring and the telescopic amount of the vibrating screen surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity, and obtain the screening trajectory and the material movement speed, including: Decompose the three-axis acceleration signal in the feedback data of the vibration direction and intensity into horizontal and vertical components, calculate the ratio and phase difference of the two components, judge the elliptical shape of the current vibration trajectory, and trigger the adjustment of the vibrating screen surface inclination when the horizontal-vertical ratio deviates from the target value; Conduct a time-domain comparison of the vibration signals of the four support points of the vibrating screen box, measure the time difference of the vibration peak values of each support point, judge the uniformity of the vibrating screen box vibration, and send an adjustment instruction to the left or right spring pre-tightening device when the vibration time difference between the left and right support points exceeds the preset threshold; According to the difference between the current vibration trajectory and the target trajectory, control the hydraulic system to extend or shorten the vibrating screen surface inclination hydraulic cylinder, change the angle between the vibrating screen surface and the horizontal plane until the vibrating screen surface inclination reaches the calculated optimal angle; Drive the spring pre-tightening screw to rotate through the servo motor, adjust the spring support stiffness, and make the vibration intensity of the vibrating screen box reach the optimal value matching the current coal type until the movement speed of the material on the vibrating screen meets the preset requirements.

5. The vibration sieve parameter adaptive control method based on machine learning according to claim 4, characterized in that The time-domain comparison of the vibration signals of the four support points of the vibrating screen box, measuring the time difference of the vibration peak values of each support point, judging the uniformity of the vibrating screen box vibration, and sending an adjustment instruction to the left or right spring pre-tightening device when the vibration time difference between the left and right support points exceeds the preset threshold, includes: Collect the acceleration sensor data installed at the four support points of the front left, front right, rear left, and rear right of the vibrating screen box, perform low-pass filtering on the original data, and extract the peak point timestamps within each signal period; Calculate the difference in the peak timestamps of the diagonal support points to obtain the torsional vibration parameter of the vibrating screen box. When the torsional vibration parameter is greater than the first preset threshold, increase the pre-tightening force of the spring on the side with lower torsional stiffness; Calculate the difference in the peak timestamps of the same-side support points to obtain the rocking vibration parameter of the vibrating screen box. When the rocking vibration parameter is greater than the second preset threshold, increase the pre-tightening force of the spring on the side with larger rocking; Input the vibration amplitude data of the four support points of the vibrating screen box into the unbalance calculation unit to generate the vibrating screen box vibration uniformity index. When the vibrating screen box vibration uniformity index exceeds the third preset threshold, calculate the adjustment amount of the spring pre-tightening force for each support point; Send the calculated spring pre-tightening force adjustment amount to the servo motor controller of each support point, and drive the servo motor to rotate forward or backward to adjust the spring compression amount until the vibration time difference of the four support points is less than the fourth preset threshold.

6. The machine learning-based vibrating screen parameter adaptive control method according to claim 5, wherein The inputting the vibration amplitude data of the four support points of the vibrating screen box into the unbalance calculation unit to generate the vibrating screen box vibration uniformity index includes: Normalize the vibration amplitude data of the four support points of the front left, front right, rear left, and rear right of the vibrating screen box to unify the numerical range of each support point to obtain the normalized vibration amplitude set; Calculate the difference between the maximum value and the minimum value in the normalized vibration amplitude set, and divide it by the average value of the vibration amplitudes of the four support points to obtain the first vibrating screen box vibration unbalance; Conduct a variance analysis on the vibration amplitudes of the four support points, calculate the ratio of the standard deviation to the mean value to obtain the coefficient of variation of the vibration amplitude, and use the coefficient of variation as the second vibrating screen box vibration unbalance; The vibration unbalance degree of the first sieve box and that of the second sieve box are synthesized by weighted average to obtain the sieve box vibration uniformity index; According to the numerical range of the sieve box vibration uniformity index, the vibration state is divided into a uniform zone, a slightly non-uniform zone, a moderately non-uniform zone, and a severely non-uniform zone, and corresponding spring pre-tightening force adjustment strategies are adopted for different zones.

7. The vibration sieve parameter adaptive control method based on machine learning according to claim 1, characterized in that The control of the feeding speed of the feeder and the height position of the longitudinal baffle of the sieve box according to the screening efficiency data detected by the undersize and oversize samplers includes: Collecting material samples through the oversize sampler installed at the end of the sieve surface and the undersize sampler at the discharge port of the screening machine, performing screening particle size analysis and quality determination on the samples, and calculating the screening efficiency value of the current screening process; Comparing the screening efficiency value with a preset target efficiency threshold, and generating a screening state evaluation result when the screening efficiency value deviates from the preset target efficiency threshold by more than the first deviation range; Based on the screening state evaluation result, controlling the frequency conversion device of the feeder to adjust the feeding speed, changing the residence time of the material on the sieve surface, and at the same time controlling the hydraulic cylinder to adjust the height position of the longitudinal baffle of the sieve box to adjust the material layer thickness; Monitoring the screening efficiency for a continuous number of cycles under the adjusted working conditions, calculating the screening efficiency change slope, and locking the current feeding speed and baffle height position when the absolute value of the change slope is less than the stability determination threshold and the efficiency value is within the second deviation range of the preset target efficiency threshold; When it is detected that the screening efficiency value exceeds the third deviation range of the preset target efficiency threshold or the particle size distribution of the undersize is abnormal, triggering the start signal of the screen cleaning mechanism and temporarily adjusting the vibration intensity parameter to the maximum limit value.

8. A vibration sieve parameter adaptive control system based on machine learning, characterized in that, For implementing the machine learning-based vibration screen parameter adaptive control method according to any one of claims 1-7, the machine learning-based vibration screen parameter adaptive control system includes: A collection module for collecting data on coal materials at the feeding end of the vibration screen through an image collection device installed above the feeding hopper and a material sensor installed at the front end of the sieve surface to obtain a characteristic parameter set; A control module for calculating the current coal load coefficient and screening difficulty index according to the characteristic parameter set, and adjusting the phase difference and current input of the double-motor vibrator through the control system to obtain the vibration direction and intensity; An adjustment module for adjusting the pre-tightening force of the sieve box support spring and the telescopic amount of the sieve surface inclination hydraulic cylinder based on the feedback data of the vibration direction and intensity to obtain the screening trajectory and the material movement speed; A feeding module for controlling the feeding speed of the feeder and the height position of the longitudinal baffle of the sieve box according to the screening efficiency data detected by the undersize and oversize samplers.

9. A vibration sieve parameter adaptive control device based on machine learning, characterized in that, It includes a memory and a processor, the memory stores a computer program that can run on the processor, and when the processor executes the computer program, it implements the machine learning-based vibration screen parameter adaptive control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor is caused to execute the machine learning-based vibration screen parameter adaptive control method according to any one of claims 1 to 7.

Images