Screening machine operation monitoring method and system

Abstract

The present invention relates to the field of equipment monitoring technology, and discloses a screening machine operation monitoring method and system. The method comprises: performing vibration anti-interference stabilization processing on the real-time collected image of the working area of ​​the screening machine to obtain an operation monitoring image; performing domain-specific visual feature analysis to obtain screen tension state data, material flow state data, and equipment vibration state data; performing coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain an abnormal operation judgment result of the contact sensor; when the abnormal operation judgment result indicates a contact sensor failure, performing visual data replacement compensation processing to obtain target operation parameters; optimizing the screening process parameters based on the target operation parameters, and outputting a control instruction for the screening machine excitation device. The present invention can accurately identify various fault types of the contact sensor and can still maintain optimal screening performance in the event of a sensor failure.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment monitoring, and in particular to a screening machine operation monitoring method and system. Background Art

[0002] Traditional screening machine monitoring methods rely primarily on contact sensors, such as vibration sensors, temperature sensors, and pressure sensors, to obtain equipment operating parameters. These sensors then provide data for fault diagnosis and control decisions. However, screening machines operate in harsh environments. Factors such as dust accumulation, mechanical vibration, and temperature fluctuations can easily cause contact sensors to experience signal drift, loose connections, and decreased measurement accuracy, severely impacting the reliability of the monitoring system.

[0003] A core issue facing existing monitoring technologies is the lack of effective compensation mechanisms for sensor failures, often requiring downtime for repairs or downgraded operation, resulting in production losses. While some research has attempted to incorporate visual monitoring technology, limited by the high-frequency vibration environment of screening machines, traditional visual methods struggle to obtain clear and stable images, making it difficult to accurately extract operational characteristics. Summary of the Invention

[0004] The present invention provides a screening machine operation monitoring method and system, which can accurately identify various fault types of contact sensors and maintain optimal screening performance in the event of sensor failure.

[0005] In a first aspect, the present invention provides a method for monitoring the operation of a screening machine, the method comprising:

[0006] Step S1: performing vibration anti-interference stabilization processing on the real-time collected image of the working area of ​​the screening machine to obtain an operation monitoring image;

[0007] Step S2: performing domain-specific visual feature analysis of the screen area, material area, and equipment area based on the operation monitoring image to obtain screen tension state data, material flow state data, and equipment vibration state data;

[0008] Step S3: performing coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain an abnormal operation determination result of the contact sensor;

[0009] Step S4: When the abnormal operation determination result indicates a contact sensor failure, performing visual data replacement compensation processing based on the material flow state data and the equipment vibration state data to obtain target operating parameters;

[0010] Step S5: Optimize the screening process parameters based on the target operating parameters and output a control instruction for the vibration device of the screening machine.

[0011] Optionally, in a first implementation of the first aspect of the present invention, step S1 includes:

[0012] Step S11: performing fast Fourier transform spectrum analysis on the real-time collected image of the working area of ​​the screening machine to identify the dominant vibration frequency component;

[0013] Step S12: using the dominant vibration frequency component to dynamically calculate state transfer matrix coefficients and an observation noise covariance matrix of a Kalman prediction module, and forming vibration frequency adaptive filtering parameters based on the state transfer matrix coefficients and the observation noise covariance matrix;

[0014] Step S13: inputting the real-time collected image into a vibration anti-interference encoder-decoder network configured with the vibration frequency adaptive filtering parameters to perform image stabilization and reconstruction processing, and outputting an operation monitoring image.

[0015] Optionally, in a second implementation of the first aspect of the present invention, step S13 includes:

[0016] Step S131: configuring the state transfer matrix coefficients in the vibration frequency adaptive filter parameters to the encoder part of the vibration anti-interference encoder-decoder network, performing vibration state prediction on the image features in the encoder part to obtain a multi-level image feature map, the encoder part includes 6 convolutional layers, and each convolutional layer is connected to a Kalman prediction module based on the state transfer matrix coefficients;

[0017] Step S132: configuring the observation noise covariance matrix in the vibration frequency adaptive filtering parameters to the middle layer of the vibration anti-interference encoder-decoder network, and performing Kalman update filtering on the multi-level image feature map using the observation noise covariance matrix to obtain a filtered and purified image feature map;

[0018] Step S133: Inputting the filtered and purified image feature map into the decoder part of the vibration interference suppression encoder-decoder network for upsampling and reconstruction processing to obtain an initial stabilized image, wherein the decoder part includes 6 deconvolution layers, and each deconvolution layer combines the optimal estimation result of the Kalman filter to perform pixel-level reconstruction;

[0019] Step S134: performing pixel displacement compensation processing based on Kalman filter state estimation on the initial stabilized image, and performing reverse correction on the image pixels using the vibration displacement predicted by the state transfer matrix to obtain an operation monitoring image.

[0020] Optionally, in a third implementation of the first aspect of the present invention, step S2 includes:

[0021] Step S21: identifying pixel attributions of screen lines, material particles, and equipment structural components in the operation monitoring image, and dividing the operation monitoring image into a screen area, a material area, and an equipment area based on the pixel attributions;

[0022] Step S22: performing edge detection and line curvature calculation based on the screen area, extracting the screen line profile, and calculating the screen intersection displacement and line length change rate based on the screen line profile to obtain screen tension state data;

[0023] Step S23: performing optical flow tracking and three-dimensional reconstruction based on the material area to obtain material particle motion trajectories, and extracting particle size distribution and flow velocity vectors based on the material particle motion trajectories to obtain material flow state data;

[0024] Step S24: performing background subtraction and displacement detection based on the device area to obtain device vibration state data.

[0025] Optionally, in a fourth implementation of the first aspect of the present invention, step S3 includes:

[0026] Step S31: Modeling the mesh tension-vibration coupling relationship based on the mesh tension state data and the measured data of the screening machine vibration sensor to obtain a mesh tension-vibration coupling relationship model;

[0027] Step S32: inputting the real-time output signals of the vibration sensor, temperature sensor and pressure sensor of the screening machine into the screen tension-vibration coupling relationship model for prediction and correlation deviation analysis to obtain the correlation deviation value of each contact sensor;

[0028] Step S33: performing a threshold comparison and determination on the correlation deviation value to obtain an abnormality determination mark of each contact sensor;

[0029] Step S34: performing fault type identification based on the correlation deviation value and the abnormality determination flag to obtain an operational abnormality determination result of the contact sensor.

[0030] Optionally, in a fifth implementation of the first aspect of the present invention, step S31 includes:

[0031] Step S311: Time-synchronizing the screen tension state data with the measured data of the screening machine vibration sensor to obtain a screen tension-vibration data sequence;

[0032] Step S312: constructing a quadratic polynomial coupling function between screen tension and vibration output based on the physical characteristics of the screening machine vibration response, wherein the vibration sensor output is used as the response variable, the screen tension is used as the excitation variable, and the coupling function includes a square term, a linear term, a constant term, and a random disturbance term;

[0033] Step S313: Based on the screen tension-vibration data sequence, a least squares regression algorithm is used to identify the model parameters of the quadratic polynomial coupling function, and the square term coefficient, linear term coefficient and constant term value are calculated;

[0034] Step S314: establishing a screen tension-vibration coupling relationship model based on the square term coefficient, the linear term coefficient and the constant term value.

[0035] Optionally, in a sixth implementation of the first aspect of the present invention, step S4 includes:

[0036] Step S41: when the abnormal operation determination result indicates a contact sensor failure, performing a reverse calculation of the material load weight based on the material accumulation thickness, particle coverage area, and particle size distribution parameters in the material flow state data to obtain a visually perceived material load value;

[0037] Step S42: Substituting the visually perceived material load value as an input variable into the screening machine material load-vibration response transfer function to solve the vibration output theoretical value and obtain a theoretical vibration parameter reference value;

[0038] Step S43: Determining a visual compensation strategy and weight distribution scheme based on the abnormal operation determination result of the contact sensor, adopting a pure visual replacement mode when the sensor signal is completely unreliable, and adopting a visual-sensor hybrid compensation mode when the sensor is partially available, and calculating adaptive compensation weight coefficients for visual data and sensor data;

[0039] Step S44: Using the adaptive compensation weight coefficient, the theoretical vibration parameter reference value and the measured data of the available contact sensor are intelligently fused and reconstructed to obtain the target operating parameters.

[0040] Optionally, in a seventh implementation of the first aspect of the present invention, step S42 includes:

[0041] Step S421: determining the gain coefficient, delay time constant, and time constant in the material load-vibration response transfer function of the screening machine according to the model of the screening machine and the current working conditions, and obtaining a set of transfer function parameters;

[0042] Step S422: performing numerical preprocessing and format conversion on the visually perceived material load value to obtain a material load variable;

[0043] Step S423: Based on the transfer function parameter set, the material load variable is input into the screening machine material load-vibration response transfer function to perform time domain response analysis to obtain dynamic response data;

[0044] Step S424: performing numerical mapping and unit normalization on the vibration characteristic quantities in the dynamic response data according to the output standard of the contact sensor to obtain a theoretical vibration parameter reference value.

[0045] Optionally, in an eighth implementation of the first aspect of the present invention, step S5 includes:

[0046] Step S51: extracting the oversize particle size distribution and undersize flow rate from the target operating parameters, calculating the real-time screening efficiency by the ratio of the undersize flow rate to the total flow rate, performing a difference operation between the real-time screening efficiency and the set target value to obtain the particle size deviation and the flow rate deviation;

[0047] Step S52: Based on the dynamic response characteristics of the screening machine, the particle size deviation and flow rate deviation are used as inputs, and the exciting force adjustment amount and frequency adjustment amount are used as outputs to obtain an intelligent control optimization model;

[0048] Step S53: Utilize the intelligent control optimization model to perform current deviation state analysis, obtain excitation adjustment parameters, and convert the excitation adjustment parameters into control instructions for the screening machine excitation device.

[0049] In a second aspect, the present invention provides a screening machine operation monitoring system, the screening machine operation monitoring system comprising:

[0050] A processing module is used to perform vibration anti-interference stabilization processing on the real-time collected images of the working area of ​​the screening machine to obtain an operation monitoring image;

[0051] A feature analysis module is used to perform domain-specific visual feature analysis of the screen area, material area, and equipment area based on the operation monitoring image to obtain screen tension state data, material flow state data, and equipment vibration state data;

[0052] A correlation verification module is used to perform a coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain an abnormal operation judgment result of the contact sensor;

[0053] a substitution compensation module, configured to perform visual data substitution compensation processing based on the material flow state data and the equipment vibration state data to obtain target operating parameters when the abnormal operation determination result indicates a contact sensor failure;

[0054] The output module is used to optimize the screening process parameters based on the target operating parameters and output the control instructions of the screening machine vibration device.

[0055] In the technical solution provided by the present invention, a vibration anti-interference encoder-decoder network configured with vibration frequency adaptive filtering parameters can obtain a clear and stable image of the operation of the screening machine in a high-frequency vibration environment, and adopts a three-domain parallel feature analysis method of the screen area, material area and equipment area to achieve a comprehensive perception of the operating status of the screening machine. A coupling relationship model between the screen tension state data and the measured data of the vibration sensor is established, which can accurately identify various fault types of the contact sensor. When a sensor fails, the visually perceived material load and equipment vibration state data are used to realize intelligent compensation of the failed sensor data through the material load-vibration response transfer function to ensure continuous and stable operation of the screening machine. The fusion weight of the visual data and the sensor data is dynamically adjusted according to the degree of sensor failure. Based on the complete operating parameters after fusion compensation, the excitation parameters are adjusted in real time using an intelligent control algorithm. In the event of a sensor failure, the optimal screening performance can still be maintained, which significantly improves the fault tolerance and environmental adaptability of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 A schematic diagram of an embodiment of a method for monitoring the operation of a screening machine according to an embodiment of the present invention;

[0058] Figure 2 Schematic diagram of an embodiment of a screening machine operation monitoring system in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] An embodiment of the present invention provides a method and system for monitoring the operation of a screening machine. The terms "first," "second," "third," "fourth," and so on (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0060] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the screening machine operation monitoring method in an embodiment of the present invention includes:

[0061] Step S1: performing vibration anti-interference stabilization processing on the real-time collected image of the working area of ​​the screening machine to obtain an operation monitoring image;

[0062] It is understandable that the execution subject of the present invention can be a screening machine operation monitoring system, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example.

[0063] Specifically, high-speed industrial cameras installed in key monitoring areas of the screening machine capture real-time images of the working process, continuously acquiring a high-resolution image stream at a stable frame rate (e.g., 30 fps). Fast Fourier transform spectral analysis is performed on this image stream to separate the periodic disturbance components caused by mechanical vibration in the image sequence, thereby identifying the dominant vibration frequency component in the current screening environment, which ranges from 0.5 Hz to 50 Hz. After identifying the dominant vibration frequency, the core parameters required by the Kalman prediction module are dynamically calculated based on its frequency characteristics. These parameters include the state transition matrix coefficients and the observation noise covariance matrix. These parameters reflect the temporal continuity disturbance and observation deviation characteristics caused by vibration in the image. By mapping the dominant vibration frequency to the structural parameters of the Kalman filter, a set of vibration frequency-adaptive filter control factors is generated. This allows the filtering algorithm to no longer rely on static or preset models, but instead evolves dynamically with the actual operating conditions of the screening machine. After completing the above parameter calculation, the system will introduce the image processing module configured with this set of adaptive filtering parameters into the deep vision network, specifically using a vibration anti-interference encoder-decoder network as the core framework. The network structure consists of a front-end encoder that performs multi-scale compression extraction on the input image features, and a back-end decoder that restores and reconstructs the image features after anti-interference processing. An updated adaptive filtering control unit is embedded in the middle layer to filter out periodic image disturbances caused by the main vibration frequency. The final output operation monitoring image has spatial stability and temporal continuity.

[0064] Step S2: Based on the operation monitoring image, the visual feature analysis of the screen area, the material area and the equipment area is performed to obtain the screen tension state data, the material flow state data and the equipment vibration state data;

[0065] Specifically, based on the high-definition images that have completed the anti-vibration stabilization process, an image semantic parsing process is constructed, and different functional areas in the image are identified and segmented. This process uses a convolutional neural network or a feature attribution mechanism based on the fusion of traditional image processing methods and deep learning to spatially cluster and semantically label image pixels, identify the pixel attribution categories belonging to screen lines, material particles, and equipment structural parts, and on this basis, divide the entire monitoring image into three subdomains: screen area, material area, and equipment area. In the screen area, the geometric contours of the screen lines are extracted by the edge detection algorithm, and the curvature of the screen lines is quantitatively calculated by combining morphological operations and line segment fitting algorithms. At the same time, the position of the line intersection is identified and its displacement changes in continuous frames are tracked to measure the degree of structural deformation caused by the screen tension. The line length change rate and the intersection displacement are extracted and recorded as the core physical indicators of the tension state to constitute the screen tension state data. At the same time, within the material area, an optical flow-based particle tracking module is deployed to dynamically capture the displacement trajectories of material particles between image frames. Grayscale gradient analysis and depth estimation methods are then used to perform a three-dimensional reconstruction of the material accumulation, thereby restoring the particle motion path, height hierarchy, and particle distribution. On this basis, the particle size distribution function and flow velocity vector field are calculated, and material flow status data including material flow rate, direction, and particle size parameters are extracted. Within the equipment area, inter-frame background subtraction technology is used to separate the micro-displacement information of structural components caused by vibration. Sub-pixel position tracking and structural feature alignment algorithms are used to quantify the displacement vectors of feature points on the equipment surface in space, thereby obtaining quantitative data on the amplitude, direction, and frequency of the equipment vibration, forming the equipment vibration status data.

[0066] Step S3: performing coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain the abnormal operation judgment result of the contact sensor;

[0067] Specifically, a coupling relationship model is established to reflect the physical correlation between screen tension and equipment vibration. This model uses screen tension state data as the input variable and vibration sensor measured data as the response variable. A nonlinear fitting method is used to construct a screen tension-vibration coupling function. Polynomial regression or curve fitting methods are introduced into the modeling process to describe the impact of screen tension changes on vibration amplitude and frequency, forming a physical response model with time continuity and state interpretability. Real-time output signals from various contact sensors, such as vibration sensors, temperature sensors, and pressure sensors, are input into the coupling model as actual observation data. These signals are compared hourly with the vibration response values ​​predicted by the model based on screen tension. Correlation deviation values ​​are calculated for each sensor type, reflecting the degree of agreement between the actual sensor signal and the theoretical value derived from the visual state. Each deviation value is compared with a preset tolerance threshold (e.g., ±15%). If the deviation exceeds this range, the sensor is deemed to be operating abnormally, and a corresponding abnormality identification flag is generated. Dynamic analysis of the time series characteristics of the deviations is performed to identify specific patterns of anomalies. A linearly increasing deviation indicates sensor drift, while intermittent jumps indicate poor connection. High-frequency random fluctuations are often associated with dust intrusion or surface contamination. By combining the deviation values ​​and anomaly indicators into a fault identification module, the current operating status of each contact sensor is qualitatively classified and quantitatively evaluated, resulting in an operational anomaly determination result that includes anomaly type, severity, and time stamp.

[0068] Step S4: When the abnormal operation determination result indicates a contact sensor failure, visual data substitution compensation processing is performed based on the material flow state data and the equipment vibration state data to obtain target operating parameters;

[0069] Specifically, after detecting a sensor anomaly, parameters such as material stacking thickness, particle coverage area, and particle size distribution are extracted from the material flow state data. These parameters are high-precision visual information obtained through optical flow tracking and image grayscale gradient reconstruction. Based on this set of parameters, the material load weight is reversely calculated using the preset density constant and stacking geometry model to obtain the material load value under visual perception. This value is the equivalent gravity mapping of the current material acting on the screen surface. The visually perceived material load value is used as the input variable and substituted into the material load-vibration response transfer function model of the screening machine. This model is an empirical-physical hybrid form preset by the system. It is in the form of a delayed first-order inertial system, which includes a gain coefficient, a time constant, and a delay parameter to simulate the typical response behavior of the screen under different loads. By solving the output response of the transfer function, the theoretical vibration parameter reference value is obtained, including the vibration amplitude and the main frequency position. Based on the fault determination results from the previous stage, the system determines the availability of the current contact sensor. If the sensor has completely failed or its output is severely abnormal, it automatically switches to pure vision replacement mode, relying solely on image data to generate target operating parameters. If the sensor has slight offset or partial distortion, a hybrid vision and sensor compensation mode is used. In this mode, a dynamic weight coefficient is further calculated between the visual and sensor data. This weight is determined by multiple factors such as sensor deviation, noise level, and historical data stability to achieve optimal information fusion. Based on the adaptive weight coefficient, the theoretical vibration parameter baseline value and the currently available sensor measured value are weighted and fused. The target operating parameters are then intelligently reconstructed using methods such as linear combination or Kalman filtering to obtain the target operating parameters.

[0070] Step S5: Optimize the screening process parameters based on the target operating parameters and output the control instructions for the screening machine vibration device.

[0071] Specifically, the particle size distribution characteristics of the oversize material and the flow rate of the undersize material are extracted from the target operating parameters, and combined with the total amount of material feed, the real-time screening efficiency is calculated to reflect the proportion of qualified particles that are effectively separated in the current screening process. The real-time screening efficiency is then compared and analyzed with the preset process target value to obtain the current efficiency difference, and the currently detected D 50The difference between the particle size value and the process standard particle size value is calculated to obtain the particle size deviation and flow rate deviation respectively. These two indicators are the main variables affecting the screening performance and are directly input into the control model to drive the adjustment of the excitation parameters. Considering that the screening machine is a typical dynamic system, its response characteristics are affected by multiple factors such as vibration amplitude, frequency, and screen surface loading status. In the control strategy, an intelligent control optimization model is constructed based on the dynamic response characteristics of the screening machine that have been modeled. The model adopts a fuzzy PID control structure, takes the particle size deviation and flow rate deviation as input variables, and uses the 25 control logic rules preset in the rule library for fuzzy reasoning processing to output the excitation force adjustment amount and the excitation frequency adjustment amount. Among them, each control item in the rule library is set according to the process experience and the dynamic behavior mapping relationship. The nonlinear rule structure ensures that the control strategy has good response capabilities to different deviation combinations. The adjustment quantity output by fuzzy reasoning is quantitatively defuzzified, and the control range of the vibration device is limited by the physical constraints of the screening machine. For example, the excitation force is controlled within the range of 70% to 130% of the rated value, and the frequency is adjusted within the range of 15Hz to 25Hz to prevent excessive control or structural impact. The calculated excitation adjustment parameters are converted into a specific control instruction format and sent to the excitation device execution unit of the screening machine, including inverter speed control instructions, eccentric block adjustment commands or hydraulic drive parameters, etc., to ensure that the equipment can change the vibration state as needed, thereby achieving the optimal guidance of particle size distribution and the dynamic balance of material diversion.

[0072] In a specific embodiment, the process of executing step S1 may specifically include the following steps:

[0073] Step S11: performing fast Fourier transform spectrum analysis on the real-time collected image of the working area of ​​the screening machine to identify the dominant vibration frequency component;

[0074] Step S12: using the dominant vibration frequency component to dynamically calculate the state transfer matrix coefficients and the observation noise covariance matrix of the Kalman prediction module, and forming vibration frequency adaptive filter parameters based on the state transfer matrix coefficients and the observation noise covariance matrix;

[0075] Step S13: inputting the real-time collected image into a vibration anti-interference encoder-decoder network configured with vibration frequency adaptive filtering parameters to perform image stabilization and reconstruction processing, and outputting an operation monitoring image.

[0076] Specifically, continuous image frames of the screening machine's working area are captured in real time and recorded continuously at a fixed frame rate (e.g., 30 fps). The resulting image data forms a time-series image sequence, which captures the dynamic state of the machine, the screen surface, and the material. This image sequence is then subjected to frequency domain analysis. The Fast Fourier Transform (FFT) algorithm is used to convert the temporal variations of consecutive pixel values ​​in the image sequence into a spectral representation. By analyzing the peak frequency of the spectral amplitude curve within the range of 0.5 Hz to 50 Hz, the dominant vibration frequency component of the current screening machine is accurately identified. Based on the dominant vibration frequency component, the parameters of the prediction and filtering modules in the image processing are updated, focusing on adjusting two core components of the Kalman filter: the state transfer matrix coefficients and the observation noise covariance matrix. The state transfer matrix models the dynamic behavior of pixel positions or image blocks over time. Its parameters determine the prediction model's responsiveness to the motion trends of consecutive frames. The observation noise covariance matrix, on the other hand, reflects the observation uncertainty caused by vibration during the image acquisition process. When the dominant vibration frequency is low, image disturbances change relatively slowly. The weights of the velocity and acceleration terms in the state transfer matrix are appropriately reduced, while the ability to suppress observation noise covariance is improved. In the case of high-frequency vibration, to ensure prediction accuracy, the dynamic response of the state is improved, and sensitivity to sudden changes is enhanced. This parameter adaptive adjustment process is based on a predefined mapping function between frequency and control coefficients, or is learned and adjusted in real time during system operation to form a set of filter parameters tailored to different vibration frequencies. The configured vibration frequency-adaptive filter parameters are embedded in a deep learning network architecture for image processing, employing a vibration interference mitigation encoder-decoder network as the core processing framework. The front-end encoder portion of this network compresses the input image at multiple scales into a high-dimensional feature representation, extracting edge, structural, and material state features. The decoder portion reconstructs features based on the encoded results, restoring the disturbance-removed image information into a high-definition output image. An intermediate dynamic filtering module is constructed between the encoder and decoder, in which the state transition prediction mechanism and the filter update mechanism are embedded. The Kalman filter algorithm is introduced as the basic engine for time series feature vibration suppression processing, so that the network can dynamically adjust the filtering behavior according to the frequency domain analysis results while processing the image frame by frame.

[0077] In a specific embodiment, the process of executing step S13 may specifically include the following steps:

[0078] Step S131: The state transfer matrix coefficients in the vibration frequency adaptive filter parameters are configured in the encoder part of the vibration interference suppression encoder-decoder network. The encoder part predicts the vibration state of the image features to obtain a multi-level image feature map. The encoder part includes 6 convolutional layers, and each convolutional layer is connected to a Kalman prediction module based on the state transfer matrix coefficients.

[0079] Step S132: configuring the observation noise covariance matrix in the vibration frequency adaptive filter parameters to the middle layer of the vibration anti-interference encoder-decoder network, and performing Kalman update filtering on the multi-level image feature map using the observation noise covariance matrix to obtain a filtered and purified image feature map;

[0080] Step S133: Inputting the filtered and purified image feature map into the decoder part of the vibration interference suppression encoder-decoder network for upsampling and reconstruction processing to obtain an initial stabilized image. The decoder part includes 6 deconvolution layers, and each deconvolution layer combines the optimal estimation result of the Kalman filter to perform pixel-level reconstruction;

[0081] Step S134: performing pixel displacement compensation processing based on Kalman filter state estimation on the initial stabilized image, and performing reverse correction on the image pixels using the vibration displacement predicted by the state transfer matrix to obtain an operation monitoring image.

[0082] Specifically, the dominant vibration frequency values ​​extracted through fast Fourier transform spectrum analysis are converted into state transition matrix coefficients and observation noise covariance matrices in the Kalman filter, and then injected into key modules of the network structure as dynamic parameters. In the encoder, the system uses a six-layer convolutional neural network as the image feature compression channel. Each convolution layer extracts low-level to high-level image representations, such as edges, texture, and shape. Each convolution operation is connected to a state transition prediction module based on the current dominant frequency configuration. This module uses the state transition matrix A to extrapolate the feature representation of the previous moment, predicting the feature change trend of the current moment before observing the next frame. After all six convolution layers are completed, the system uniformly processes the multi-level image feature map output by the encoder in the intermediate connection layer and applies the observation noise covariance matrix to complete the Kalman filter update step. During this process, the residual covariance between the previous state prediction result and the current observation value is jointly analyzed to calculate the Kalman gain, thereby performing the optimal state update operation and obtaining the filtered and purified image feature map. The system inputs the filtered and purified image feature map into the decoder, which is symmetrical to the encoder structure and also consists of six deconvolution layers. Each deconvolution layer not only performs conventional upsampling and spatial restoration operations but also, in combination with the optimal estimation results after the Kalman filter in the previous stage, reconstructs and corrects the position and intensity of each pixel during the restoration process. This ensures that the image avoids displacement blur or structural distortion caused by residual vibration errors when restoring from the compressed state, thus obtaining the initial stabilized image. Pixel displacement compensation is then performed on the initial stabilized image. Based on the current frame pixel displacement prediction stored in the Kalman state estimator, the system reversely shifts each pixel in the initial stabilized image to its theoretical stable position. This means that the global displacement field of the image is corrected using the displacement increments predicted by the state transfer matrix, forming an inverse transformation matrix for the image pixels. Each pixel position is remapped and interpolated to finally output the operation monitoring image.

[0083] In a specific embodiment, the process of executing step S2 may specifically include the following steps:

[0084] Step S21: Identify the pixel attribution of screen lines, material particles, and equipment structural parts in the operation monitoring image, and divide the operation monitoring image into a screen area, a material area, and an equipment area based on the pixel attribution;

[0085] Step S22: performing edge detection and line curvature calculation based on the screen area, extracting the screen line profile, and calculating the screen intersection displacement and line length change rate based on the screen line profile to obtain the screen tension state data;

[0086] Step S23: performing optical flow tracking and 3D reconstruction based on the material area to obtain the material particle motion trajectory, and extracting the particle size distribution and flow velocity vector based on the material particle motion trajectory to obtain material flow state data;

[0087] Step S24: performing background subtraction and displacement detection based on the device area to obtain device vibration state data.

[0088] Specifically, the system utilizes a deep image segmentation model or a multi-channel semantic extraction network based on a modified U-Net architecture to perform feature mapping for each pixel in the input image. Combining color gradients, edge responses, and local motion vectors, the system determines the spatial attribution of three object types: screen lines, material particles, and equipment components. Screen lines appear as regularly spaced, densely packed straight or intersecting lines with periodic shape structures and directional edge features; material particles appear as clusters of varying sizes, blurred edges, and active motion; and equipment components are distributed along the image edges, exhibiting static forms and rigid contours. The system then uses regional morphological factors, boundary contrast, and inter-frame motion consistency for fusion discrimination, dividing the image into three labeled domains: screen, material, and equipment. After region segmentation, the system extracts the screen contour lines within the screen area using the Canny edge detection algorithm or a HED edge network based on directional gradients. The Hough transform accurately identifies line segment coordinates and their orientation, and calculates the curvature change of the screen lines. To quantify the tension state, the system calculates the rate of change of line length relative to a standard screen geometry. It also identifies the locations of intersection nodes of the screen lines and tracks the displacement of these nodes across consecutive image frames, thereby deriving the dynamic offset vector and length deformation factor of the screen intersections. The line length change rate and intersection displacement amplitude can be directly mapped to tension patterns, such as local relaxation and stretching, on the screen, forming the key physical basis for the screen tension state data. In the material area, the system uses optical flow to track particles, selecting either dense optical flow (e.g., the Farnebäck method) or sparse optical flow (e.g., the Lucas-Kanade method) to obtain continuous frame-by-frame motion trajectories between particles. These 2D trajectories are then mapped to a 3D reconstruction module, where particle height and thickness distributions are estimated using image grayscale gradients and shadow depth models, ultimately reconstructing the 3D stacking morphology. Based on the particle trajectory data, the system calculates the displacement paths of multiple particles per unit time, normalizes their length and direction, and constructs a velocity vector field. Particle size is then clustered and classified using distribution statistics to generate particle size distribution data. The above-mentioned velocity vectors and particle size distribution together constitute the material flow state data, reflecting the current material bulk density, dynamic conveying direction and potential change trends in screening efficiency. At the same time, within the equipment area, since most structural components remain relatively stationary, their tiny displacements can be identified through the inter-frame background difference method. The system constructs a vibration-free reference background image of the equipment area, and then performs pixel difference operations on the real-time monitoring image to extract the boundary areas where micro-displacements occur under high-frequency vibrations. Sub-pixel block matching methods (such as phase correlation) are used to calculate the pixel displacement and track its vibration trajectory in the time dimension to obtain the local vibration amplitude and frequency information of the equipment structural parts. This type of information is used to reflect the stability of the support structure and the overall dynamic health status of the screening system, constituting the equipment vibration state data.

[0089] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0090] Step S31: Modeling the mesh tension-vibration coupling relationship based on the mesh tension state data and the measured data of the screening machine vibration sensor to obtain a mesh tension-vibration coupling relationship model;

[0091] Step S32: inputting the real-time output signals of the screening machine vibration sensor, temperature sensor, and pressure sensor into the screen tension-vibration coupling relationship model for prediction and correlation deviation analysis to obtain correlation deviation values ​​of each contact sensor;

[0092] Step S33: performing a threshold comparison and determination on the correlation deviation value to obtain an abnormality determination mark of each contact sensor;

[0093] Step S34: performing fault type identification based on the correlation deviation value and the abnormality determination flag to obtain an operational abnormality determination result of the contact sensor.

[0094] Specifically, the mesh tension-vibration coupling relationship is modeled based on mesh tension state data and measured data from the screening machine's vibration sensors. This framework uses mesh tension state data as the dependent variable input source and raw signals (such as frequency, acceleration, or displacement) collected by the screening machine's vibration sensors as the response targets. Through multiple sampling and time alignment, a mapping model between mesh tension and vibration data is constructed. During the modeling process, polynomial regression, nonlinear least squares, support vector machine regression, or neural networks are used to fit the coupling relationship. For example, a quadratic nonlinear model is used to reflect the accelerating effect of tension changes on vibration output. The model structure is a type of squared term, a linear term, and a constant term. The resulting model is the mesh tension-vibration coupling relationship model. Once the modeling is complete, the system collects data from multiple contact sensors in real time during equipment operation, including acceleration signals from vibration sensors, thermal drift data from temperature sensors, and stress response values ​​from pressure sensors. This real-time data is fed into the previously established coupling model. Using visual inspection results of mesh tension as a reference, the system predicts the theoretical response values ​​that each sensor should output under the current tension state. The system compares the model's predicted values ​​with the corresponding sensor's real-time measured values ​​and calculates the deviation between them. These deviations represent the correlation deviations for each contact sensor. Units are categorized as vibration difference, temperature difference, and pressure difference, depending on sensor type, and are uniformly normalized into percentage deviations. For each sensor's deviation, the system uses a fixed or dynamic threshold comparison mechanism to identify anomalies. A ±15% tolerance is set. When a sensor's deviation exceeds this threshold, it is flagged as abnormal. During this assessment, the system tracks and fits the deviation trends to identify sudden anomalies, long-term drift, or intermittent distortion. If the deviation fluctuates dramatically over a short period of time but maintains little change in the mean, it indicates sporadic disturbances. If the deviation shows a linear growth trend, it indicates gradual sensor degradation. If the deviation fluctuates and is unstable between time periods, it indicates a loose sensor connection, unstable wiring, or data communication interruption. Based on a joint analysis of deviation values ​​and anomaly indicators, the system constructs a fault type identification module. This module uses a pattern recognition algorithm to classify and judge by comparing the manifestation patterns of different types of anomalies in the time and frequency domains. For example, linearly increasing deviations often correspond to component stability degradation caused by internal thermal drift, jump signal anomalies reflect poor contact between the sensor and the controller, and high-frequency noise or random mismatches indicate environmental faults such as dust accumulation, electromagnetic interference, or external impact on the sensor surface. The system combines the identified deviation amplitude, time location, frequency distribution, and error form for feature matching, ultimately outputting an operational anomaly determination result that includes the fault type, impact level, and recommended treatment measures.

[0095] In a specific embodiment, the process of executing step S31 may specifically include the following steps:

[0096] Step S311: Time-synchronize the screen tension state data with the measured data of the screening machine vibration sensor to obtain a screen tension-vibration data sequence;

[0097] Step S312: constructing a quadratic polynomial coupling function between screen tension and vibration output based on the physical characteristics of the screening machine vibration response, wherein the vibration sensor output is used as the response variable, the screen tension is used as the excitation variable, and the coupling function includes a square term, a linear term, a constant term, and a random disturbance term;

[0098] Step S313: Based on the screen tension-vibration data sequence, a least squares regression algorithm is used to identify the model parameters of the quadratic polynomial coupling function, and the square term coefficient, linear term coefficient and constant term value are calculated;

[0099] Step S314: establishing a screen tension-vibration coupling relationship model based on the square term coefficient, the linear term coefficient and the constant term value.

[0100] Specifically, the screen tension state data is time-synchronized with the measured data from the screening machine's vibration sensor. This data, derived from a visual tension indicator calculated using image recognition, has a frame rate characteristic, outputting at 25 to 30 frames per second. Vibration sensors, on the other hand, record physical response signals such as structural acceleration, velocity, or displacement at a higher sampling frequency (e.g., 100Hz to 500Hz). Consequently, the two systems suffer from inconsistent sampling intervals and asynchronous timestamps in their original state. The system uses a unified reference clock, with the master timer built into the screening control system as the unified time base. The timestamps of the visual tension data are aligned with the acquisition time of the sensor data. Linear or spline interpolation is used to interpolate and expand the low-frequency visual data, creating a temporal correspondence with the high-frequency vibration data. This generates a screen tension-vibration data sequence. Based on the response characteristics of the vibration system, a functional relationship between screen tension and vibration output is constructed. Considering that the vibration response of the screen structure is not a simple linear characteristic under the dual influence of material load changes and structural tension changes, but rather exhibits certain nonlinear dynamic behavior, a quadratic polynomial function is used to model its coupling mechanism. This function uses the screen tension value as the excitation variable and the vibration sensor output (such as acceleration amplitude or main frequency amplitude) as the response variable, and is defined in the form of V = α·T 2+β·T+γ+ε, where V represents the current output value of the vibration sensor, T represents the screen tension value, α is the square term coefficient, β is the linear term coefficient, γ is the constant term, and ε represents the disturbance term. In order to identify the parameters α, β, and γ in the above model, the least squares regression algorithm is introduced to fit the model based on the data sequence. This process arranges the screen tension values ​​in time sequence and constructs a model containing T 2 , T and the design matrix X of the constant term, and construct the vibration response value at the corresponding moment as the observation vector Y, and establish the regression equation group Y=Xθ+ε, where θ is the parameter vector to be calculated, that is, the three-dimensional vector containing α, β, and γ. The system calculates the regression loss function ∥Y-Xθ∥ 2 Perform minimization and use the analytical solution θ=(X T X) - 1 X T Y or its regularized deformation form is solved to obtain the optimal fitting model coefficient. Based on the square term coefficient α, linear term coefficient β and constant term γ obtained by the above identification, a screen tension-vibration coupling relationship model is established.

[0101] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0102] Step S41: When the abnormal operation determination result indicates a contact sensor failure, reverse calculation of the material load weight is performed based on the material accumulation thickness, particle coverage area, and particle size distribution parameters in the material flow state data to obtain a visually perceived material load value;

[0103] Step S42: Substituting the visually perceived material load value as an input variable into the screening machine material load-vibration response transfer function to solve the vibration output theoretical value and obtain the theoretical vibration parameter reference value;

[0104] Step S43: Determine the visual compensation strategy and weight distribution scheme based on the abnormal operation determination result of the contact sensor. When the sensor signal is completely unreliable, a pure visual replacement mode is adopted. When the sensor is partially available, a visual-sensor hybrid compensation mode is adopted. The adaptive compensation weight coefficients of the visual data and the sensor data are calculated.

[0105] Step S44: Using the adaptive compensation weight coefficient, the theoretical vibration parameter reference value and the measured data of the available contact sensor are intelligently fused and reconstructed to obtain the target operating parameters.

[0106] Specifically, using the abnormality determination result as a trigger signal, if vibration, temperature, or pressure contact sensors exhibit persistent drift, abnormally fixed output, or significant decoupling from the coupling model, the visual compensation path is activated and the real-time updated material flow status data from the current image analysis module is utilized. This includes thickness data reflecting vertical material accumulation, coverage data indicating the degree of material coverage per unit screen surface, and particle size distribution data derived from particle edge contours extracted from the image. By integrating these three data sets, the system can inversely estimate the total material mass currently supported by the screen surface based on spatial geometry and structural distribution patterns, deriving a visually perceived material load value representing the current actual load level. This visually perceived material load value is then substituted as an input variable into the screening machine's material load-vibration response transfer function, which is calibrated at the factory using standard materials, standard tension, and typical excitation parameters. This function characterizes the machine's vibration response under specific loads. By inputting the current material load level into this model, the corresponding theoretical vibration parameter baseline is derived, serving as a reference for the vibration characteristics that the screening machine should achieve under ideal, fault-free conditions. The system selects a compensation mode based on the results of the anomaly recognition module in the previous stage. If the contact sensor is completely unavailable, the system directly enters a pure visual replacement mode, using the theoretical vibration values ​​derived above as the current operating parameter input. If the sensor still has partial availability, such as only slight drift or periodic fluctuations, the system initiates a fusion mode, establishing a weighting mechanism between visual and sensor data. This weighting mechanism is calculated in real time by the system's adaptive adjustment module. Its core basis is the abnormal amplitude of the current sensor signal, its historical stability, and the degree of deviation from the visual baseline, while also considering visual confidence indicators such as material image clarity and particle edge integrity. The system integrates this information and outputs a set of weight coefficients to control the data fusion ratio. After completing the weighting calculation, the system fuses the theoretical vibration values ​​with the raw sensor output to ultimately obtain the target operating parameters for the current operating conditions.

[0107] In a specific embodiment, the process of executing step S42 may specifically include the following steps:

[0108] Step S421: determining the gain coefficient, delay time constant, and time constant in the material load-vibration response transfer function of the screening machine according to the model of the screening machine and the current working conditions, and obtaining a set of transfer function parameters;

[0109] Step S422: performing numerical preprocessing and format conversion on the visually perceived material load value to obtain a material load variable;

[0110] Step S423: Based on the transfer function parameter set, the material load variable is input into the screening machine material load-vibration response transfer function to perform time domain response analysis to obtain dynamic response data;

[0111] Step S424: performing numerical mapping and unit normalization on the vibration characteristic quantities in the dynamic response data according to the output standard of the contact sensor to obtain a theoretical vibration parameter reference value.

[0112] Specifically, the system calls the matching transfer function template library based on parameters such as the screening machine model, excitation method, installation type, and material characteristics, and dynamically selects the appropriate gain coefficient, delay time constant, and time constant based on the current operating conditions (including material type, particle size distribution, humidity, screen tension range, and screen inclination). These parameters are obtained through calibration experiments, simulation tests, or on-site identification when the equipment leaves the factory, and are stored in the system in a database. They are automatically called through device coding to form a targeted set of transfer function parameters. Numerical preprocessing and format conversion are performed on the visually perceived material load values. The visual load data is obtained by inverse deduction of the stacking thickness, particle coverage area, and particle size distribution parameters in the image. The original output is represented in floating-point form, and the unit dimension is inconsistent with the excitation variable format expected by the transfer function input. Therefore, numerical preprocessing and format conversion operations are performed on it. This process involves rounding the raw visual load data, limiting its amplitude, removing outliers, and converting its dimensions to ensure that the resulting material load variable falls within the numerical range required by the screener response model. Units are also standardized, such as converting mass from kilograms to the system's defined nominal units or normalizing the visual estimation results to a percentage input range, ensuring that the load variable becomes a valid excitation in the transfer function. The system then inputs the processed material load variable into the screener's material load-vibration response transfer function and performs a time-domain response analysis based on the aforementioned parameter set. This analysis uses numerical integration, iterative differential analysis, or convolution to simulate the system's response to the load change. Dynamic response data, including timestamps, response amplitudes, and vibration trends, is generated to describe the vibration amplitude evolution of the screener structure under the load disturbance. The generated dynamic response data is organized in a time series format, from which the system extracts vibration characteristics such as steady-state vibration amplitude, response time, maximum jump variation, or frequency principal components. The response characteristics are formatted and aligned to standards, and a numerical mapping and unit normalization mechanism is introduced to ensure a one-to-one correspondence between the vibration characteristics and the standard output format of the sensor. If the contact sensor output is a milliacceleration value, the system maps the response characteristics to an equivalent acceleration range and performs a linear transformation based on the sensor range, output accuracy, and calibration coefficient. If the sensor output is frequency or vibration energy density, the system converts the response result into a standard interval corresponding to the frequency distribution amplitude or energy unit, and unifies indicators such as the time integration window and filter bandwidth to ensure that the two have the same evaluation dimension. The response data after the final conversion is defined as the theoretical vibration parameter baseline value.

[0113] In a specific embodiment, the process of executing step S5 may specifically include the following steps:

[0114] Step S51: extracting the oversize particle size distribution and undersize flow rate from the target operating parameters, calculating the real-time screening efficiency by the ratio of the undersize flow rate to the total flow rate, performing a difference operation between the real-time screening efficiency and the set target value to obtain the particle size deviation and flow rate deviation;

[0115] Step S52: Based on the dynamic response characteristics of the screening machine, the particle size deviation and flow rate deviation are used as inputs, and the exciting force adjustment amount and frequency adjustment amount are used as outputs to obtain an intelligent control optimization model;

[0116] Step S53: Utilize the intelligent control optimization model to perform current deviation state analysis, obtain excitation adjustment parameters, and convert the excitation adjustment parameters into control instructions for the screening machine excitation device.

[0117] Specifically, the target operating parameters are analyzed to extract the particle size distribution information and material flow data directly related to the screening process. The particle size distribution is derived from the spatial distribution image of the material particles on the screen, including parameters such as the median particle size, maximum particle size and distribution dispersion. The material flow under the screen is calculated by the particle flow velocity and density formed by the material passing through the screen into the lower channel. The system establishes a mapping relationship between the image change rate and the particle density calibration to estimate the actual conveying volume of the material under the screen in real time. At the same time, the total material flow is extracted from the target parameters, that is, the total material flow including the particles that have not passed the screen and the separated particles under the screen, and the ratio between the two is calculated. This ratio represents the real-time screening efficiency and reflects the overall level of the current material classification effect. The real-time screening efficiency is compared with the target value set for the operation to obtain the difference result. The difference reflects the degree of deviation of the current operating status from the target indicator from the perspective of the screening process. At the same time, the median or representative characteristic value of the oversize particle size distribution is extracted and compared with the target particle size required for the screening task. This generates a particle size deviation measure, revealing the impact of screen tension, material moisture, or excitation parameters on the particle size screening performance. Flow rate deviation, on the other hand, reflects the real-time gap between the undersize flow rate and the target capacity, and can indicate operational issues such as insufficient excitation force, material blockage, or uneven distribution. Based on these two types of deviations, the system feeds them as input variables into an intelligent control optimization model. This model, constructed based on the structural dynamic response characteristics of the screening machine, maps the input particle size deviation and flow rate deviation to the adjustment dimensions of the excitation control strategy and outputs actionable excitation force and frequency adjustments. The model design incorporates fuzzy control logic, an empirical rule library, and dynamic gain parameters to ensure flexible response and adjustable versatility for different deviation combinations. For example, when the particle size deviation is large and the flow rate deviation is small, the system moderately increases the frequency to optimize particle jumping trajectories. When the flow rate deviation is large and the particle size match is good, the excitation force is increased to enhance material penetration. The model optimizes parameters and updates weights through continuous operation and data accumulation, giving it the ability to adaptively evolve and adapt to different screen specifications, material types, and process rhythms. After completing the deviation analysis, the system generates a set of excitation adjustment parameters based on the excitation force and frequency adjustment output by the intelligent control optimization model, and converts them into the control instruction format required by the screening machine's excitation device. The control instruction includes an adjustment signal, an electrical drive quantity, a frequency adjustment code, or a hydraulic control parameter. The specific format is determined by the structure of the equipment's excitation system. For example, if a motor eccentric block drive system is used, the output will be sent to the inverter control unit; if a hydraulic exciter is used, the adjustment value will be converted into a valve control pressure instruction. The instruction output frequency is set to a fixed control cycle, such as updating every 100 milliseconds, to ensure that the control behavior has sufficient real-time performance and response accuracy. The system makes real-time corrections to the screening efficiency within each control cycle, thereby promoting continuous optimization of the excitation state.

[0118] The above describes the operation monitoring method of the screening machine in the embodiment of the present invention. The following describes the operation monitoring system of the screening machine in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a screening machine operation monitoring system includes:

[0119] The processing module 201 is used to perform vibration anti-interference stabilization processing on the real-time collected image of the working area of ​​the screening machine to obtain an operation monitoring image;

[0120] Feature analysis module 202, for performing domain-specific visual feature analysis of the screen area, material area, and equipment area based on the operation monitoring image, to obtain screen tension state data, material flow state data, and equipment vibration state data;

[0121] The correlation verification module 203 is used to perform coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain the abnormal operation judgment result of the contact sensor;

[0122] The substitution compensation module 204 is configured to perform visual data substitution compensation processing based on the material flow state data and the equipment vibration state data to obtain target operating parameters when the operation abnormality determination result indicates a contact sensor failure;

[0123] The output module 205 is used to optimize the screening process parameters based on the target operating parameters and output the control instructions of the vibration device of the screening machine.

[0124] Through the collaborative efforts of these components, a vibration-interference-resistant encoder-decoder network, configured with adaptive vibration frequency filtering parameters, dynamically adjusts image stabilization parameters based on the screen machine's real-time vibration frequency, effectively eliminating image blur and distortion caused by high-frequency vibration during visual monitoring. By decomposing the screen machine's monitoring area into the screen, material, and equipment regions for parallel feature extraction, the system simultaneously captures multi-dimensional information, including screen tension, material flow, and equipment vibration. This provides comprehensive perception of the screen machine's operating status, delivering greater information richness than single monitoring targets. A coupling relationship model is established between screen tension data and measured vibration sensor data. Through a cross-validation mechanism, the system accurately identifies various contact sensor fault types, such as signal drift, loose connections, and dust accumulation, improving the accuracy and reliability of fault detection. In the event of a contact sensor failure, the system utilizes visually perceived material load and equipment vibration data to calculate theoretical values ​​using the material load-vibration response transfer function. This intelligently compensates for the failed sensor data, ensuring continuous and stable operation of the screen machine even in the event of partial sensor failure. The fusion weight of visual and sensor data is dynamically adjusted based on the severity of the sensor failure. Pure visual compensation is used in the event of complete sensor failure, while hybrid compensation is employed in the event of partial anomalies, improving the overall quality and credibility of the monitoring data. Based on the complete operating parameters after fusion and compensation, a fuzzy PID control algorithm is used to adjust the excitation force and frequency in real time according to the screening efficiency deviation, achieving dynamic optimization of the screening process and maintaining optimal screening performance even in the event of a sensor failure.

[0125] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0126] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a screening machine operation monitoring device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0127] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A screening machine operation monitoring method, characterized in that: include: Step S1: performing vibration anti-interference stabilization processing on the real-time collected image of the working area of ​​the screening machine to obtain an operation monitoring image; Step S2: performing domain-specific visual feature analysis of the screen area, material area, and equipment area based on the operation monitoring image to obtain screen tension state data, material flow state data, and equipment vibration state data; Step S3: performing coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain the abnormal operation judgment result of the contact sensor, including: step S31: performing screen tension-vibration coupling relationship modeling based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain a screen tension-vibration coupling relationship model; step S32: inputting the real-time output signals of the screening machine vibration sensor, temperature sensor and pressure sensor into the screen tension-vibration coupling relationship model to perform prediction and correlation deviation analysis to obtain the correlation deviation value of each contact sensor; step S33: performing threshold comparison judgment on the correlation deviation value to obtain the abnormal judgment mark of each contact sensor; step S34: performing fault type identification based on the correlation deviation value and the abnormal judgment mark to obtain the abnormal operation judgment result of the contact sensor; Step S4: When the abnormal operation determination result indicates a contact sensor failure, performing visual data replacement compensation processing based on the material flow state data and the equipment vibration state data to obtain target operating parameters, including: Step S41: When the abnormal operation determination result indicates a contact sensor failure, performing reverse calculation of the material load weight based on the material stacking thickness, particle coverage area and particle size distribution parameters in the material flow state data to obtain a visually perceived material load value; Step S42: Substituting the visually perceived material load value as an input variable into the screening machine material load-vibration response transfer function to solve the vibration output theoretical value to obtain a theoretical vibration parameter reference value; Step S43: Determining a visual compensation strategy and a weight distribution scheme based on the abnormal operation determination result of the contact sensor, adopting a pure visual replacement mode when the sensor signal is completely unreliable, and adopting a visual-sensor hybrid compensation mode when the sensor is partially available, and calculating an adaptive compensation weight coefficient of visual data and sensor data; Step S44: Using the adaptive compensation weight coefficient to intelligently fuse and reconstruct the theoretical vibration parameter reference value and the measured data of the available contact sensor to obtain the target operating parameters; Step S5: Optimize the screening process parameters based on the target operating parameters and output a control instruction for the vibration device of the screening machine.

2. The screening machine operation monitoring method according to claim 1, characterized in that: Step S1 includes: Step S11: performing fast Fourier transform spectrum analysis on the real-time collected image of the working area of ​​the screening machine to identify the dominant vibration frequency component; Step S12: using the dominant vibration frequency component to dynamically calculate state transfer matrix coefficients and an observation noise covariance matrix of a Kalman prediction module, and forming vibration frequency adaptive filtering parameters based on the state transfer matrix coefficients and the observation noise covariance matrix; Step S13: inputting the real-time collected image into a vibration anti-interference encoder-decoder network configured with the vibration frequency adaptive filtering parameters to perform image stabilization and reconstruction processing, and outputting an operation monitoring image.

3. The screening machine operation monitoring method according to claim 2, characterized in that: Step S13 includes: Step S131: configuring the state transfer matrix coefficients in the vibration frequency adaptive filter parameters to the encoder part of the vibration anti-interference encoder-decoder network, performing vibration state prediction on the image features in the encoder part to obtain a multi-level image feature map, the encoder part includes 6 convolutional layers, and each convolutional layer is connected to a Kalman prediction module based on the state transfer matrix coefficients; Step S132: configuring the observation noise covariance matrix in the vibration frequency adaptive filtering parameters to the middle layer of the vibration anti-interference encoder-decoder network, and performing Kalman update filtering on the multi-level image feature map using the observation noise covariance matrix to obtain a filtered and purified image feature map; Step S133: Inputting the filtered and purified image feature map into the decoder part of the vibration interference suppression encoder-decoder network for upsampling and reconstruction processing to obtain an initial stabilized image, wherein the decoder part includes 6 deconvolution layers, and each deconvolution layer combines the optimal estimation result of the Kalman filter to perform pixel-level reconstruction; Step S134: performing pixel displacement compensation processing based on Kalman filter state estimation on the initial stabilized image, and performing reverse correction on the image pixels using the vibration displacement predicted by the state transfer matrix to obtain an operation monitoring image.

4. The screening machine operation monitoring method according to claim 1, characterized in that: Step S2 includes: Step S21: identifying pixel attributions of screen lines, material particles, and equipment structural components in the operation monitoring image, and dividing the operation monitoring image into a screen area, a material area, and an equipment area based on the pixel attributions; Step S22: performing edge detection and line curvature calculation based on the screen area, extracting the screen line profile, and calculating the screen intersection displacement and line length change rate based on the screen line profile to obtain screen tension state data; Step S23: performing optical flow tracking and three-dimensional reconstruction based on the material area to obtain material particle motion trajectories, and extracting particle size distribution and flow velocity vectors based on the material particle motion trajectories to obtain material flow state data; Step S24: performing background subtraction and displacement detection based on the device area to obtain device vibration state data.

5. The screening machine operation monitoring method according to claim 1, characterized in that: Step S31 includes: Step S311: Time-synchronizing the screen tension state data with the measured data of the screening machine vibration sensor to obtain a screen tension-vibration data sequence; Step S312: constructing a quadratic polynomial coupling function between screen tension and vibration output based on the physical characteristics of the screening machine vibration response, wherein the vibration sensor output is used as the response variable, the screen tension is used as the excitation variable, and the coupling function includes a square term, a linear term, a constant term, and a random disturbance term; Step S313: Based on the screen tension-vibration data sequence, a least squares regression algorithm is used to identify the model parameters of the quadratic polynomial coupling function, and the square term coefficient, linear term coefficient and constant term value are calculated; Step S314: establishing a screen tension-vibration coupling relationship model based on the square term coefficient, the linear term coefficient and the constant term value.

6. The screening machine operation monitoring method according to claim 1, characterized in that: Step S42 includes: Step S421: determining the gain coefficient, delay time constant, and time constant in the material load-vibration response transfer function of the screening machine according to the model of the screening machine and the current working conditions, and obtaining a set of transfer function parameters; Step S422: performing numerical preprocessing and format conversion on the visually perceived material load value to obtain a material load variable; Step S423: Based on the transfer function parameter set, the material load variable is input into the screening machine material load-vibration response transfer function to perform time domain response analysis to obtain dynamic response data; Step S424: performing numerical mapping and unit normalization on the vibration characteristic quantities in the dynamic response data according to the output standard of the contact sensor to obtain a theoretical vibration parameter reference value.

7. The screening machine operation monitoring method according to claim 1, characterized in that: Step S5 includes: Step S51: extracting the oversize particle size distribution and undersize flow rate from the target operating parameters, calculating the real-time screening efficiency by the ratio of the undersize flow rate to the total flow rate, performing a difference operation between the real-time screening efficiency and the set target value to obtain the particle size deviation and the flow rate deviation; Step S52: Based on the dynamic response characteristics of the screening machine, the particle size deviation and flow rate deviation are used as inputs, and the exciting force adjustment amount and frequency adjustment amount are used as outputs to obtain an intelligent control optimization model; Step S53: Utilize the intelligent control optimization model to perform current deviation state analysis, obtain excitation adjustment parameters, and convert the excitation adjustment parameters into control instructions for the screening machine excitation device.

8. A screening machine operation monitoring system, characterized in that: A method for monitoring the operation of a screening machine according to any one of claims 1 to 7, comprising: A processing module is used to perform vibration anti-interference stabilization processing on the real-time collected images of the working area of ​​the screening machine to obtain an operation monitoring image; A feature analysis module is used to perform domain-specific visual feature analysis of the screen area, material area, and equipment area based on the operation monitoring image to obtain screen tension state data, material flow state data, and equipment vibration state data; A correlation verification module is used to perform coupling correlation verification based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain an abnormal operation judgment result of the contact sensor, including: modeling the screen tension-vibration coupling relationship based on the screen tension state data and the measured data of the screening machine vibration sensor to obtain a screen tension-vibration coupling relationship model; inputting the real-time output signals of the screening machine vibration sensor, temperature sensor and pressure sensor into the screen tension-vibration coupling relationship model to perform prediction and correlation deviation analysis to obtain a correlation deviation value of each contact sensor; performing threshold comparison judgment on the correlation deviation value to obtain an abnormal judgment mark of each contact sensor; identifying the fault type based on the correlation deviation value and the abnormal judgment mark to obtain an abnormal operation judgment result of the contact sensor; The substitution compensation module is used to perform visual data substitution compensation processing based on the material flow state data and the equipment vibration state data to obtain target operating parameters when the operation abnormality determination result indicates a contact sensor failure, including: when the operation abnormality determination result indicates a contact sensor failure, performing reverse calculation of the material load weight based on the material stacking thickness, particle coverage area and particle size distribution parameters in the material flow state data to obtain a visually perceived material load value; substituting the visually perceived material load value as an input variable into the screening machine material load-vibration response transfer function to solve the vibration output theoretical value to obtain a theoretical vibration parameter reference value; determining a visual compensation strategy and a weight distribution scheme based on the operation abnormality determination result of the contact sensor, adopting a pure visual substitution mode when the sensor signal is completely unreliable, and adopting a visual-sensor hybrid compensation mode when the sensor is partially available, and calculating an adaptive compensation weight coefficient of visual data and sensor data; using the adaptive compensation weight coefficient to intelligently fuse and reconstruct the theoretical vibration parameter reference value and the measured data of the available contact sensor to obtain the target operating parameters; The output module is used to optimize the screening process parameters based on the target operating parameters and output the control instructions of the screening machine vibration device.

Images