Anti-deviation self-adaptive ore conveying belt deviation correction control system

Abstract

The application discloses a deviation-preventing self-adaptive ore conveying belt deviation correction control system and belongs to the technical field of conveying belt deviation correction control. The system comprises a data acquisition module which synchronously acquires belt speed time sequence data, deviation displacement time sequence data, carrier roller load distribution data and current fluctuation data of the ore conveying belt; a signal preprocessing module which performs multi-scale decomposition on the deviation displacement time sequence data to obtain low-frequency trend components, periodic fluctuation components and random mutation components, and generates a deviation feature triplet; a deviation correction mode recognition module which matches the triplet with a preset deviation mode library to determine a current deviation mode; a deviation correction parameter generation module which generates an initial deviation correction control parameter set according to the current mode, the belt speed and the load data; and a closed-loop correction module which sends the initial parameters to a deviation correction execution mechanism, reacquires deviation displacement data to calculate residual deviation after execution, and updates the initial parameter set to obtain iterative deviation correction control parameters. The application realizes accurate deviation correction control of the ore conveying belt.

Description

Technical Field

[0001] This invention relates to the field of conveyor belt correction control technology, specifically to an anti-deviation adaptive ore conveyor belt correction control system. Background Technology

[0002] Ore conveyor belts are affected by factors such as uneven material distribution, belt speed fluctuations, idler load changes, and instantaneous impacts during transportation or storage. Belt misalignment can manifest as slow deviation, periodic oscillation, or sudden impacts. Related technologies typically utilize misalignment switches, displacement sensors, or mechanically adjusting idlers for detection and correction. Some control systems directly output adjustment amounts based on misalignment displacement thresholds or deviation ratios. However, misalignment displacement time-series data simultaneously contains low-frequency trends, periodic fluctuations, and random abrupt changes. Existing processing methods based on time-domain statistics and threshold judgments struggle to effectively separate these components, resulting in the inability to distinguish between different misalignment modes such as gradual deviation, reciprocating oscillations, and sudden impacts.

[0003] Different misalignment modes require different idler tilt angles, tensions, and actuator speeds. When modes are confused, the correction parameters are prone to mismatch, resulting in correction lag or over-action. Existing solutions lack synchronous utilization of multi-source data such as belt speed, idler load distribution, and current fluctuations, making it difficult to dynamically adjust correction parameters according to actual load conditions. Furthermore, the residual misalignment after correction is not effectively collected and fed back, and there is a lack of iterative update mechanism after the initial correction parameter output, preventing parameter correction based on execution results. When belt speed, load distribution, and current fluctuations change, fixed correction parameters struggle to adapt to changing operating conditions. Related technologies need to separate low-frequency trends, periodic fluctuations, and random abrupt changes from misalignment displacement time-series data, distinguish different misalignment modes based on these characteristics, and perform closed-loop iterative updates based on the residual misalignment after correction execution to match the correction parameters with the current operating conditions. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive ore conveyor belt deviation correction control system. This system performs multi-scale decomposition on the deviation displacement time series data of the ore conveyor belt, forms deviation characteristic triplet and identifies the current deviation mode, generates initial deviation correction control parameters by combining belt speed time series data and idler load distribution data, and performs iterative correction based on the residual deviation amount after execution, so that the deviation correction action is adapted to the deviation mode and load conditions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: As one technical solution of the present invention, an anti-deviation adaptive ore conveyor belt correction control system is provided, including a data acquisition module, a signal preprocessing module, a correction mode recognition module, a correction parameter generation module, and a closed-loop correction module. The data acquisition module is used to synchronously acquire belt speed time-series data, misalignment displacement time-series data, idler load distribution data, and current fluctuation data of the ore conveyor belt in the transportation or storage device; the signal preprocessing module is used to perform multi-scale decomposition on the misalignment displacement time-series data to obtain low-frequency trend components, periodic fluctuation components, and random mutation components, and generate misalignment feature triplets; the misalignment correction mode recognition module is used to match the misalignment feature triplets with a preset misalignment mode library to determine the current misalignment mode; the misalignment correction parameter generation module is used to generate an initial misalignment correction control parameter set based on the current misalignment mode, the belt speed time-series data, and the idler load distribution data; the closed-loop correction module is used to send the initial misalignment correction control parameter set to the misalignment correction actuator, and after execution, reacquire the misalignment displacement time-series data, calculate the residual misalignment, update the initial misalignment correction control parameter set, and generate iterative misalignment correction control parameters. This system performs multi-scale decomposition of deviation displacement into trend, period, and abrupt changes, and combines it with closed-loop iterative correction to achieve adaptive deviation correction when the working conditions of the ore conveyor belt change, thereby improving the accuracy of deviation correction control and operational stability.

[0006] Preferably, the signal preprocessing module is used to divide the deviation displacement time series data into multiple subsequences according to a preset window, perform variational mode decomposition on each subsequence to obtain multiple modal components; calculate the center frequency and energy proportion of each modal component; superimpose the modal component with the lowest center frequency as a low-frequency trend component, superimpose the two modal components with the middle center frequency and the largest energy proportion as periodic fluctuation components, and superimpose the modal component with the largest kurtosis value among the remaining modal components as a random mutation component; calculate the monotonic offset rate based on the difference between the first and last points of the low-frequency trend component within the window, calculate the oscillation period based on the interval between adjacent zero crossings of the periodic fluctuation component, and calculate the mutation amplitude based on the peak value of the random mutation component, summarizing them into the deviation feature triplet. Through the above multi-scale decomposition and feature extraction, the trend offset, periodic oscillation and sudden impact in the deviation displacement can be effectively separated, providing stable feature input for subsequent deviation pattern recognition.

[0007] Preferably, the deviation correction pattern recognition module maps the monotonic offset rate, oscillation period, and abrupt change amplitude in the deviation feature triplet to a three-dimensional feature space to obtain three-dimensional feature points; it then calls a pre-trained deviation pattern classification network to calculate the distance scores between the three-dimensional feature points and the three pattern center vectors corresponding to the gradual offset pattern, reciprocating oscillation pattern, and sudden impact pattern; based on a preset pattern matching threshold and the distance scores, it determines the current deviation pattern; if the minimum distance score is greater than or equal to the pattern matching threshold, the current deviation pattern is marked as an unknown pattern; if the minimum distance score is less than the pattern matching threshold, the pattern corresponding to the minimum distance score is selected as the current deviation pattern. This method can reduce the probability of mismatched deviation patterns and improve the reliability of pattern recognition.

[0008] Preferably, the correction parameter generation module is used to calculate the load difference between the left and right idlers based on the idler load distribution data when the current misalignment mode is a gradual offset mode, convert the load difference into left and right tension adjustment amounts, and map the idler forward tilt angle adjustment amount according to the monotonic offset rate; when the current misalignment mode is a reciprocating oscillation mode, calculate the correction actuator action speed based on the oscillation period; when the current misalignment mode is a sudden impact mode, calculate the upper limit of the idler forward tilt angle adjustment amount and the limit value of the correction actuator action speed based on the sudden change amplitude; when the current misalignment mode is an unknown mode, use the preset default correction parameters as the initial correction control parameter set; otherwise, combine the corresponding calculated left tension adjustment amount, right tension adjustment amount, idler forward tilt angle adjustment amount, action speed, or limit value into the initial correction control parameter set. Generating differentiated correction parameters for different misalignment modes can improve the adaptability of correction parameters to misalignment states and avoid over-correction or under-correction caused by a single fixed parameter.

[0009] Preferably, the closed-loop correction module is used to reacquire the misalignment displacement time sequence data after the correction actuator executes the initial correction control parameter set, according to a preset stabilization waiting time, and extract the steady-state segment after the stabilization waiting time; calculate the difference between the mean of the misalignment displacement in the steady-state segment and the preset centerline as the residual misalignment amount, and simultaneously calculate the standard deviation of the misalignment displacement in the steady-state segment as the fluctuation evaluation value; if the residual misalignment amount is greater than a preset threshold and the fluctuation evaluation value is less than a preset fluctuation tolerance value, then the current correction actuator action speed remains unchanged, the left tension adjustment amount is reduced by a fixed step size and the idler roller forward tilt angle adjustment amount is increased; if the fluctuation evaluation value is greater than or equal to the preset fluctuation tolerance value, then the correction actuator action speed is increased and the iteration step size of the idler roller forward tilt angle adjustment amount is reduced; the updated parameters are recorded as the iterative correction control parameters for the next iteration. By using closed-loop correction to correct the residual misalignment amount and fluctuation online, the adaptive correction capability and parameter convergence of the system under complex working conditions can be improved.

[0010] Furthermore, the data acquisition module is used to trigger the synchronous acquisition of the laser ranging sensor array, speed encoder, and weighing sensor using a hardware clock source. After acquisition, a timestamp and conveyor belt running mileage identifier are added to each frame of data, and the belt speed time-series data, belt misalignment displacement time-series data, idler load distribution data, and current fluctuation data under the same timestamp are bound into a synchronous data tuple. A displacement increment symbol sequence is generated based on the belt misalignment displacement difference between adjacent timestamps, and the length of consecutive identical symbols is counted as the drift length. The drift length is multiplied by the current belt speed value in the belt speed time-series data to obtain the cumulative offset distance, and the cumulative offset distance is written into the belt misalignment displacement time-series data. Through multi-source data synchronous acquisition and mileage identifier binding, the correspondence between data in the time and spatial dimensions can be guaranteed in subsequent processing, providing a synchronous data foundation for belt misalignment feature extraction and closed-loop correction.

[0011] Furthermore, the signal preprocessing module is used to grid the idler load distribution data according to the lateral coordinates of the conveyor belt to obtain a lateral load distribution curve; calculate the lateral distance between the centroid of the lateral load distribution curve and the centerline of the conveyor belt to obtain the load offset; synchronously compare the load offset with the monotonic offset rate of the low-frequency trend component within a time window; if the load offset and the monotonic offset rate have the same sign within the set time window and the rate of change of the load offset exceeds a preset rate of change threshold, a load offset warning mark is generated; the load offset warning mark and the deviation feature triplet are sent to the deviation correction pattern recognition module. By jointly judging the load offset and trend offset, a warning mark can be generated before the deviation trend intensifies, improving the system's timeliness in responding to deviation risks.

[0012] Furthermore, the correction parameter generation module is used to acquire the current fluctuation data, extract the average current, current pulsation amplitude, and current change rate from the current fluctuation data; calculate the ratio of the current pulsation amplitude to the average current as the load fluctuation characteristic; when the load fluctuation characteristic is greater than a preset fluctuation threshold, the adjustment amount of the idler roller forward tilt angle is reduced proportionally, and the operating speed of the correction actuator is increased proportionally; when the load fluctuation characteristic is less than or equal to the preset fluctuation threshold, the adjustment amount of the idler roller forward tilt angle remains unchanged; the adjusted adjustment amount of the idler roller forward tilt angle and the operating speed of the correction actuator are written into the initial correction control parameter set. By considering the load change reflected by the current fluctuation, the influence of load fluctuation on the operation of the correction actuator can be reduced, and the stability of the correction parameters under different load conditions can be improved.

[0013] Furthermore, the closed-loop correction module is used to calculate the parameter differences between corresponding items in the iterative correction control parameters of the current iteration and the previous iteration after generating the iterative correction control parameters. If the absolute value of the parameter difference is less than a preset parameter stability threshold, the iterative correction control parameters of the current iteration are marked as stable parameters. Simultaneously, the current fluctuation data is monitored. If a peak exceeding a preset current mutation threshold appears in the current fluctuation data, the action speed of the correction actuator in the stable parameters is increased by a preset increment, and the residual deviation calculation is re-executed; otherwise, the stable parameters are directly sent to the correction actuator. Through parameter stability judgment and current mutation monitoring, unnecessary adjustments can be avoided when the system is close to stability, and action speed compensation can be performed during current mutations, enhancing the system's anti-interference capability.

[0014] Furthermore, the deviation correction pattern recognition module, after determining the current deviation pattern based on the pattern matching threshold, calculates the load difference between the left and right idlers in the idler load distribution data, and dynamically adjusts the pattern matching threshold based on the current belt speed in the belt speed time series data and the load difference. When the current belt speed is greater than a preset high-speed threshold, the matching threshold for the gradual offset pattern is increased; when the load difference is greater than a preset off-center load threshold, the matching threshold for the sudden impact pattern is decreased. The adjusted pattern matching threshold and the current deviation pattern are output together as the pattern recognition result, and the adjusted pattern matching threshold is used for the next deviation feature triplet matching. By dynamically adjusting the pattern matching threshold, the influence of belt speed and off-center load changes on the deviation feature distribution can be adapted, reducing pattern misjudgment and improving the adaptability of pattern recognition under varying operating conditions.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: The time-series data of the deviation displacement is divided into multiple subsequences according to a preset window by the signal preprocessing module, and variational mode decomposition is performed on each subsequence to obtain multiple modal components. Based on the center frequency and energy proportion of the modal components, the modal component with the lowest center frequency is superimposed as a low-frequency trend component, the two modal components with the middle center frequency and the largest energy proportion are superimposed as periodic fluctuation components, and the modal component with the largest kurtosis value among the remaining modal components is superimposed as a random mutation component. The monotonic deviation rate is obtained from the difference between the first and last values ​​of the low-frequency trend component, the oscillation period is obtained from the interval between adjacent zero crossings of the periodic fluctuation component, and the mutation amplitude is obtained from the peak value of the random mutation component, forming a deviation characteristic triplet. This triplet separates the slow deviation, reciprocating oscillation, and instantaneous impact in the deviation displacement, avoiding the misjudgment of periodic oscillation as continuous deviation or sudden impact as slow deviation by a single amplitude or threshold judgment. The deviation correction pattern recognition module maps monotonic offset rate, oscillation period, and abrupt change amplitude to a three-dimensional feature space. It then scores the distance between these parameters and the center vectors corresponding to the gradual offset pattern, reciprocating oscillation pattern, and sudden impact pattern. Based on the pattern matching threshold, it determines the current deviation pattern; unknown patterns are not forcibly matched. Thus, different deviation patterns can correspond to different correction parameter generation paths, reducing correction lag and excessive actions caused by pattern confusion. The gradual offset pattern matching threshold is increased at high speeds, while the sudden impact pattern matching threshold is decreased when the off-center load increases, ensuring that pattern recognition maintains its discriminative ability under high-speed and off-center load conditions.

[0016] The correction parameter generation module generates an initial set of correction control parameters based on the current deviation mode, belt speed timing data, and idler load distribution data. In progressive offset mode, the load difference between the left and right idlers is used to convert the left and right tension adjustment amounts, and the idler forward tilt angle adjustment amount is mapped according to the monotonic offset rate. In reciprocating oscillation mode, the action speed of the correction actuator is calculated based on the oscillation period. In sudden impact mode, the upper limit value of the idler forward tilt angle adjustment amount and the action speed limit value are calculated based on the sudden change amplitude. The ratio of the current pulsation amplitude to the average current in the current fluctuation data is used as a load fluctuation characteristic. When the load fluctuation is large, the idler forward tilt angle adjustment amount is reduced proportionally and the action speed is increased to avoid over-adjustment caused by the superposition of load fluctuations and correction actions. After executing the initial correction control parameters, the closed-loop correction module re-acquires the deviation displacement timing data, extracts the steady-state segment, and calculates the residual deviation amount and fluctuation evaluation value. When the residual deviation exceeds the preset threshold and the fluctuation evaluation value is less than the preset fluctuation tolerance value, the action speed remains constant, and the left tension adjustment is decreased by a fixed step size while the idler roller tilt angle adjustment is increased. When the fluctuation evaluation value is greater than or equal to the preset fluctuation tolerance value, the action speed is increased while the iteration step size of the idler roller tilt angle adjustment is decreased. The initial parameters are iteratively corrected using the measured residual deviation after execution, so that the correction parameters adopt different update directions under steady-state deviation and fluctuation conditions, reducing the problem that the parameters cannot adapt to the actual execution effect after one output. When the difference of each iterated parameter is less than the preset stability threshold, it is marked as a stable parameter. When the current spikes exceeding the preset sudden change threshold, the action speed is increased by a preset increment, so that the parameters can maintain matching with the working conditions under both stable operation and sudden current fluctuations. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a schematic diagram of the module structure of the adaptive ore conveyor belt deviation correction control system. Figure 2 This is a flowchart of the synchronous acquisition of conveyor belt misalignment displacement and calculation of cumulative offset distance; Figure 3 This is a schematic diagram of the time series data of the deviation displacement and its multi-scale decomposition modal components. Figure 4 This is a schematic diagram of the three-dimensional feature space distribution of deviation characteristics; Figure 5 This is a graph showing the relationship between the roller tilt angle adjustment, monotonic offset rate, and current fluctuation. Figure 6It is the convergence curve of the iterative correction control parameters of the closed-loop correction module; Figure 7 It is a time series diagram of the representative value of the deviation displacement and the cumulative offset distance. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1 This invention provides an adaptive ore conveyor belt deviation correction control system, comprising a data acquisition module, a signal preprocessing module, a deviation correction pattern recognition module, a deviation correction parameter generation module, and a closed-loop correction module. The data acquisition module synchronously acquires belt speed time-series data, deviation displacement time-series data, idler load distribution data, and current fluctuation data of the ore conveyor belt in the transportation or storage device. The signal preprocessing module performs multi-scale decomposition on the deviation displacement time-series data to obtain low-frequency trend components, periodic fluctuation components, and random mutation components, and generates deviation feature triplet sets. The deviation correction pattern recognition module matches the deviation feature triplet sets with a preset deviation pattern library to determine the current deviation pattern. The deviation correction parameter generation module generates an initial deviation correction control parameter set based on the current deviation pattern, belt speed time-series data, and idler load distribution data. The closed-loop correction module sends the initial deviation correction control parameter set to the deviation correction actuator, executes it, reacquires the deviation displacement time-series data, calculates the residual deviation, updates the initial deviation correction control parameter set, and generates iterative deviation correction control parameters.

[0021] The data acquisition module provides the acquired belt speed timing data, belt misalignment timing data, idler load distribution data, and current fluctuation data to the signal preprocessing module and the correction parameter generation module. The signal preprocessing module performs multi-scale decomposition on the belt misalignment timing data to generate misalignment feature triples. The correction pattern recognition module matches the misalignment feature triples with a preset misalignment pattern library and outputs the current misalignment pattern. The correction parameter generation module generates an initial set of correction control parameters based on the current misalignment pattern, belt speed timing data, and idler load distribution data. The closed-loop correction module sends the initial correction control parameter set to the correction actuator, executes it, reacquires the belt misalignment timing data, calculates the residual misalignment, updates the initial correction control parameter set, and generates iterative correction control parameters.

[0022] In practice, after acquiring the deviation displacement time-series data, the signal preprocessing module divides the data into multiple sub-sequences using a sliding window according to a preset window length and a preset step size. The preset window length is set to 1000 sampling points, and the preset step size is set to 500 sampling points, with a 500-point overlap between adjacent sub-sequences. The sliding window division starts from the first sampling point of the deviation displacement time-series data and slides 500 sampling points at a time until all deviation displacement time-series data is covered. Each sub-sequence contains 1000 sampling points.

[0023] The signal preprocessing module performs variational mode decomposition on each subsequence, obtaining multiple mode components. The number of mode components in the variational mode decomposition is set to 5, the quadratic penalty factor is set to 2000, and the convergence tolerance is set to... The number of modal components is set to 5 because the displacement time series data contains three types of components: trend components, periodic fluctuation components, and random mutation components. Five modal components can cover all three types of components and retain decomposition redundancy. The second-order penalty factor is set to 2000 because when the second-order penalty factor is 2000, the bandwidth and center frequency of the modal components in the decomposition result are well balanced. The convergence tolerance is set to... The basis for this is that it meets the requirements for decomposition accuracy and the computation time on the embedded processor is acceptable.

[0024] The algorithmic architecture of variational mode decomposition (VMD) includes a variational model construction module, a frequency domain update module, and a convergence determination module. The variational model construction module constructs the objective function and constraints that minimize the sum of the bandwidths of all modal components. The frequency domain update module updates each modal component and its center frequency in the frequency domain. The convergence determination module calculates the convergence conditions. The variational model construction module outputs the objective function and constraints to the frequency domain update module. After updating, the frequency domain update module outputs the modal components and their center frequencies to the convergence determination module. If the convergence conditions are not met, the convergence determination module feeds back the updated modal components and their center frequencies to the frequency domain update module for further iteration. If the convergence conditions are met, multiple modal components are output.

[0025] The solution steps for variational mode decomposition are as follows: construct a variational model, minimize the sum of bandwidths of all modal components, and constrain the sum of bandwidths of all modal components to equal the input subsequence; introduce a quadratic penalty factor and Lagrange multipliers to transform the constrained variational problem into an unconstrained variational problem; update each modal component and center frequency in the frequency domain; update the Lagrange multipliers in the time domain; check whether the convergence condition is met, and if not, continue iterating.

[0026] The signal preprocessing module calculates the center frequency and energy percentage of each modal component. The center frequency is obtained by calculating the centroid of the power spectrum of the modal component. A Fast Fourier Transform is performed on each modal component to obtain its spectrum. The square of the spectrum amplitude is calculated as the power spectrum. The first moment of the power spectrum is calculated with frequency on the x-axis and power spectrum on the y-axis to obtain the center frequency. The energy percentage is obtained by calculating the ratio of the energy of each modal component to the total energy of the subsequence. The energy of a modal component is the sum of the squares of its amplitude at each sampling point in the time domain. The total energy of the subsequence is the sum of the squares of its amplitude at each sampling point in the time domain. The energy percentage is equal to the energy of each modal component divided by the total energy of the subsequence.

[0027] The signal preprocessing module superimposes the modal components with the lowest center frequency to form a low-frequency trend component. If there is more than one modal component with the lowest center frequency, these multiple modal components are added together to obtain the low-frequency trend component. The signal preprocessing module superimposes the two modal components with the highest energy proportions and center frequencies in the middle range to form a periodic fluctuation component. Modal components with center frequencies in the middle range, arranged from smallest to largest, and located between the second and fourth positions are considered. The signal preprocessing module superimposes the modal component with the largest kurtosis value among the remaining modal components to form a random abrupt change component. The remaining modal components refer to those not selected for the low-frequency trend component and the periodic fluctuation component. The kurtosis value is obtained by calculating the ratio of the square of the fourth-order cumulant to the square of the second-order cumulant of the modal component.

[0028] The signal preprocessing module calculates the monotonic offset rate based on the difference between the first and last samples of the low-frequency trend component within the window. The monotonic offset rate equals the difference between the value of the last sample point of the low-frequency trend component within the window and the value of the first sample point within the window, divided by the number of sample points within the window. The number of sample points within the window equals the preset window length. The sign of the monotonic offset rate indicates the offset direction, and the absolute value of the monotonic offset rate indicates the rate of offset.

[0029] The signal preprocessing module calculates the oscillation period based on the interval between adjacent zero-crossing points of the periodic fluctuation component. The oscillation period is obtained by multiplying the number of sampling points between two adjacent zero-crossing points in the same direction in the periodic fluctuation component by the sampling period. If the periodic fluctuation component has no zero-crossing points, the oscillation period is set to a preset maximum oscillation period. The preset maximum oscillation period is the preset window length multiplied by the sampling period.

[0030] The signal preprocessing module calculates the amplitude of the sudden change based on the peak value of the random sudden change component. The peak value of the random sudden change component is obtained by calculating the maximum absolute value of the random sudden change component within a window. The amplitude of the sudden change is equal to the peak value of the random sudden change component.

[0031] The signal preprocessing module summarizes the monotonic offset rate, oscillation period, and abrupt change amplitude into a triplet of deviation characteristics.

[0032] After acquiring the idler load distribution data, the signal preprocessing module grids the data according to the conveyor belt's transverse coordinates. The transverse coordinates extend from the conveyor belt centerline to both sides. During gridding, the transverse coordinates are discretized into multiple transverse grid cells, each with a width of 10 mm. The idler load data within each transverse grid cell are accumulated to obtain the transverse load distribution curve. The horizontal axis of the transverse load distribution curve represents the center position of the transverse grid cell, and the vertical axis represents the accumulated load value within that cell.

[0033] The signal preprocessing module calculates the lateral distance between the centroid of the lateral load distribution curve and the centerline of the conveyor belt, thus obtaining the load offset. The formula for calculating the centroid of the lateral load distribution curve is: in, Indicates the centroid of the lateral load distribution curve; Indicates the first The center position of each horizontal grid cell; Indicates the first The cumulative load value within each horizontal grid cell; This indicates the total number of horizontal grid cells. It equals the conveyor belt's transverse width divided by the width of the transverse grid unit. The conveyor belt's transverse width is determined by the conveyor belt model, and the transverse grid unit width is set to 10 mm. The load offset equals... The difference between the lateral coordinate of the conveyor belt centerline and the centerline of the conveyor belt is set to 0.

[0034] The signal preprocessing module synchronously compares the load offset with the monotonic offset rate of the low-frequency trend component within a time window. The time window length is set to a preset window length multiplied by the sampling period. Within the set time window, if the load offset and the monotonic offset rate have the same sign and the rate of change of the load offset exceeds a preset rate of change threshold, a load offset warning flag is generated. The rate of change of the load offset is obtained by dividing the difference between the load offsets of two adjacent sampling points by the sampling period. The preset rate of change threshold is set to 0.5 mm / s, based on the assumption that a rate of change of misalignment caused by changes in roller load exceeding 0.5 mm / s indicates a tendency for off-center loading. If the load offset and the monotonic offset rate have different signs or the rate of change of the load offset does not exceed the preset rate of change threshold, no load offset warning flag is generated.

[0035] The signal preprocessing module sends the load offset warning flag along with the deviation feature triplet to the deviation correction mode recognition module. If no load offset warning flag is generated, only the deviation feature triplet is sent.

[0036] See Figure 3 In the graph, the horizontal axis represents time in seconds, and the vertical axis represents belt misalignment displacement in millimeters. The solid black line represents the original belt misalignment displacement time series data, exhibiting complex characteristics of superimposed trend changes, periodic fluctuations, and random mutations. The dashed blue line represents the low-frequency trend component, showing a clear monotonically increasing trend, with the value increasing from 0 mm to approximately 0.5 mm, reflecting a positive and relatively stable monotonically shifting rate of the conveyor belt misalignment. The dashed red line represents the periodic fluctuation component, exhibiting an approximately sinusoidal waveform with an amplitude maintained at around ±0.2 mm and a period of approximately 0.2 seconds, reflecting a clear reciprocating oscillation characteristic of the conveyor belt misalignment. The dotted green line represents the random mutation component, which is generally close to zero, with significant peak mutations only appearing at 0.2 seconds, 0.4 seconds, 0.6 seconds, and 0.8 seconds, with mutation amplitudes of approximately 0.3 mm, indicating the presence of sudden impact signal components at these four locations. Throughout the entire time domain, the low-frequency trend component is consistent with the changing trend of the deviation displacement data, the periodic fluctuation component corresponds to the periodic fluctuation of the deviation displacement data, and the random mutation component captures the sudden anomalies in the deviation displacement data.

[0037] In practical implementation, the deviation correction pattern recognition module acquires the deviation feature triplet. The deviation feature triplet includes the monotonic offset rate, oscillation period, and abrupt change amplitude. The deviation correction pattern recognition module maps the monotonic offset rate, oscillation period, and abrupt change amplitude to a three-dimensional feature space, obtaining three-dimensional feature points. The first coordinate axis of the three-dimensional feature space corresponds to the monotonic offset rate, the second coordinate axis corresponds to the oscillation period, and the third coordinate axis corresponds to the abrupt change amplitude. The first coordinate value of the three-dimensional feature point is the value of the monotonic offset rate, the second coordinate value is the value of the oscillation period, and the third coordinate value is the value of the abrupt change amplitude.

[0038] The deviation correction pattern recognition module calls a pre-trained deviation pattern classification network. This network consists of an input layer, a normalization layer, and a distance calculation layer. The input layer has three input nodes, receiving the first, second, and third coordinate values ​​of a 3D feature point. The normalization layer, connected to the input layer, linearly normalizes the values ​​of the three input nodes, mapping them to the [0,1] interval. The normalization layer stores the minimum and maximum values ​​for each dimension. The minimum and maximum values ​​for the monotonic offset rate, oscillation period, and abrupt change amplitude are obtained from the training samples. The distance calculation layer, connected to the normalization layer, has three output nodes, corresponding to the gradual offset mode, reciprocating oscillation mode, and sudden impact mode, respectively. Each output node is fully connected to the three output nodes of the normalization layer, with connection weights equal to the three components of the corresponding pattern center vector, a bias of 0, and a square root activation function. The output of the distance calculation layer is three distance scores.

[0039] The training steps for the deviation pattern classification network are as follows: Historical deviation feature triplet samples are collected. These samples are derived from historical operational data and are manually labeled with pattern tags, including gradual deviation pattern, reciprocating oscillation pattern, and sudden impact pattern. For all historical deviation feature triplet samples labeled as gradual deviation pattern, the average value along the monotonic deviation rate dimension is calculated as the first component of the gradual deviation pattern center vector; the average value along the oscillation period dimension is calculated as the second component; and the average value along the sudden impact amplitude dimension is calculated as the third component. For all historical deviation feature triplet samples labeled as reciprocating oscillation pattern, the average value along the monotonic deviation rate dimension is calculated as the first component of the reciprocating oscillation pattern center vector; the average value along the oscillation period dimension is calculated as the second component; and the average value along the sudden impact amplitude dimension is calculated as the third component. For all historical deviation feature triplet samples labeled as sudden impact modes, the average value along the monotonic offset rate dimension is calculated as the first component of the sudden impact mode center vector; the average value along the oscillation period dimension is calculated as the second component; and the average value along the abrupt change amplitude dimension is calculated as the third component. The center vectors of the gradual offset mode, the reciprocating oscillation mode, and the sudden impact mode are then incorporated into the connection weights of the distance calculation layer.

[0040] After the deviation correction pattern recognition module calls the deviation pattern classification network, the deviation pattern classification network normalizes the 3D feature points. The normalized 3D feature points are then input into the distance calculation layer. The distance calculation layer calculates a distance score for each output node. The distance score calculation formula is: in, Represents the relationship between three-dimensional feature points and the first Distance score between the center vectors of each pattern; This represents the monotonic offset rate among the normalized 3D feature points. This represents the oscillation period in the normalized three-dimensional feature points; This represents the magnitude of abrupt changes in the normalized three-dimensional feature points; Indicates the first The first component of the pattern center vector corresponding to the monotonic offset rate; Indicates the first The second component of the center vector of each mode, corresponding to the oscillation period; Indicates the first The third component in the center vector of each pattern that corresponds to the mutation magnitude; The values ​​are 1, 2, and 3. Corresponding to progressive offset mode, Corresponding to the reciprocating oscillation pattern, Corresponding to sudden impact mode. For the progressive offset mode distance score, Distance score for reciprocating oscillation mode. Distance rating for sudden impact mode.

[0041] The deviation correction pattern recognition module acquires a preset pattern matching threshold. The initial value of the preset pattern matching threshold is set to 0.5. This threshold is set based on the fact that the distance score from the normalized 3D feature point to the pattern center vector is approximately 1.732 when all three dimensions are at their farthest point. 0.5 falls within a smaller region of the possible range of distance scores, enabling the differentiation between known and unknown patterns. The deviation correction pattern recognition module identifies the minimum distance score from the distance scores of the progressive drift pattern, the reciprocating oscillation pattern, and the sudden impact pattern. If the minimum distance score is greater than or equal to the pattern matching threshold, the current deviation pattern is marked as an unknown pattern. If the minimum distance score is less than the pattern matching threshold, the pattern corresponding to the minimum distance score is selected as the current deviation pattern.

[0042] In some embodiments, after determining the current misalignment mode based on a pattern matching threshold, the deviation correction pattern recognition module calculates the load difference between the left and right idlers in the idler load distribution data. The left idler load is the sum of all loads within the left lateral coordinate range of the conveyor belt in the idler load distribution data, and the right idler load is the sum of all loads within the right lateral coordinate range of the conveyor belt in the idler load distribution data. The load difference between the left and right idlers is equal to the left idler load minus the right idler load. The deviation correction pattern recognition module reads the current belt speed from the belt speed timing data, which is the speed encoder measurement value at the current sampling time.

[0043] The belt alignment mode recognition module has a preset high-speed threshold of 3.0 meters per second, defined as 60% of the conveyor belt's rated speed as a high-speed operating state. It also has a preset off-center load threshold, set to the absolute value of the load difference between the left and right idlers being 20% ​​of their average load. This threshold indicates a significant off-center load when it exceeds 20%. When the current belt speed exceeds the preset high-speed threshold, the matching threshold for the progressive offset mode is increased. This is achieved by multiplying the progressive offset mode matching threshold by a preset proportional coefficient of 1.2. This coefficient is set because progressive offset characteristics are easily confused with normal oscillation during high-speed operation; increasing the matching threshold raises the recognition threshold for the progressive offset mode. When the load difference exceeds the preset off-center load threshold, the matching threshold for the sudden impact mode is decreased. The method to reduce the matching threshold for the sudden impact mode is as follows: multiply the matching threshold for the sudden impact mode by a preset proportional coefficient of 0.8. The preset proportional coefficient of 0.8 is set based on the fact that the probability of the sudden impact mode occurring increases when the off-center load increases. Reducing the matching threshold can improve the sensitivity to the sudden impact mode. When the current belt speed is less than or equal to the preset high-speed threshold, the matching threshold for the gradual offset mode is not adjusted. When the load difference is less than or equal to the preset off-center load threshold, the matching threshold for the sudden impact mode is not adjusted. The matching threshold for the reciprocating oscillation mode remains unchanged during dynamic adjustment.

[0044] The deviation correction pattern recognition module outputs the adjusted pattern matching threshold together with the current deviation pattern as the pattern recognition result, and uses the adjusted pattern matching threshold for the next deviation feature triplet matching.

[0045] See Figure 4 The figure uses monotonic offset rate (in millimeters per second) as the x-axis and oscillation period (in seconds) as the y-axis to show the two-dimensional projection distribution of different deviation modes in the three-dimensional feature space. Blue dots represent sample points of the gradual offset mode, red squares represent sample points of the reciprocating oscillation mode, and green triangles represent sample points of the sudden impact mode. The distribution of sample points of the three modes shows a clear clustering trend. The center vector of each mode is marked with a pentagram of the corresponding color, and the current feature point is represented by a black cross.

[0046] The sample points of the progressive migration mode are mainly concentrated in the range of monotonic migration rate of 0.4 to 1.2 mm / s and oscillation period of 2.0 to 4.5 seconds, indicating that the deviation behavior corresponding to this mode has a large monotonic migration rate and a long oscillation period. The center vector is located in this region, reflecting the typical characteristic value of this mode.

[0047] The sample points of the reciprocating oscillation mode are mainly distributed in the range of monotonic offset rate from 0 to 0.4 mm / s and oscillation period from 0.4 to 1.5 seconds, indicating that this mode is characterized by a small monotonic offset rate and a short oscillation period. The corresponding center vector position is clearly focused in this region, indicating that the oscillation period of this mode is short and the monotonic offset rate is small.

[0048] The sample points of the sudden impact mode are distributed in the range of monotonic deviation rate from 0 to 1.3 mm / s and oscillation period from 0 to 0.5 seconds, exhibiting a low oscillation period and a wide range of monotonic deviation rates. The central vector of this mode is clustered in this low oscillation period region, indicating that the mode corresponds to a sudden and brief deviation impact event.

[0049] The current feature point (black cross) is located in the region close to the center vector of the reciprocating oscillation mode, with a monotonic offset rate of about 0.4 mm / s and an oscillation period of about 1.2 seconds, indicating that the current deviation state is more inclined to the reciprocating oscillation mode.

[0050] In practice, the deviation correction parameter generation module acquires the current deviation mode, belt speed timing data, and idler load distribution data. The current deviation mode is output by the deviation correction mode recognition module, while the belt speed timing data and idler load distribution data are provided by the data acquisition module.

[0051] When the current misalignment mode is progressive offset mode, the correction parameter generation module calculates the load difference between the left and right idlers based on the idler load distribution data. The left idler load is the sum of all load values ​​within the left lateral coordinate range of the conveyor belt in the idler load distribution data. The right idler load is the sum of all load values ​​within the right lateral coordinate range of the conveyor belt in the idler load distribution data. The load difference between the left and right idlers is equal to the left idler load minus the right idler load. The correction parameter generation module converts the load difference between the left and right idlers into left tension adjustment and right tension adjustment amounts. The calculation formulas for the left and right tension adjustment amounts are: in, Indicates the left tension adjustment amount; Indicates the right tension adjustment amount; Indicates the tension conversion factor; This indicates the load difference between the left and right idler rollers. The value range is from 0.1 to 0.5. The method of obtaining the tension conversion coefficient is to look up the table based on the conveyor belt width and conveyor belt stiffness level from the preset tension conversion coefficient table, which is stored in the storage unit of the correction parameter generation module. The setting is 0.2. The basis for this setting is that when the conveyor belt width is 1200 mm and the conveyor belt stiffness level is medium, a tension conversion coefficient of 0.2 can keep the variation range of the left tension adjustment and the right tension adjustment range matched with the variation range of the load difference. The unit is Newton. and The unit is Newton.

[0052] In some embodiments, the correction parameter generation module obtains the idler roll tilt angle adjustment amount based on the monotonic offset rate mapping. The monotonic offset rate is derived from the misalignment feature triplet. The idler roll tilt angle adjustment amount is obtained by looking up a table. The correction parameter generation module stores a preset monotonic offset rate mapping table, which divides the monotonic offset rate into multiple intervals, each interval corresponding to an idler roll tilt angle adjustment amount. The monotonic offset rate interval from 0 mm / s to 0.5 mm / s corresponds to an idler roll tilt angle adjustment amount of 0.5 degrees; the monotonic offset rate interval from 0.5 mm / s to 1.5 mm / s corresponds to an idler roll tilt angle adjustment amount of 1.0 degree; the monotonic offset rate interval from 1.5 mm / s to 3.0 mm / s corresponds to an idler roll tilt angle adjustment amount of 1.5 degrees; and the monotonic offset rate interval greater than 3.0 mm / s corresponds to an idler roll tilt angle adjustment amount of 2.0 degrees. The interval division is based on the fact that the idler roll tilt angle adjustment amount increases in a stepwise manner with the increase of the monotonic offset rate. The correction parameter generation module determines the adjustment amount of the roller's forward tilt angle based on the range of the monotonic offset rate.

[0053] When the current deviation mode is a reciprocating oscillation mode, the deviation correction parameter generation module calculates the action speed of the deviation correction actuator based on the oscillation period. The oscillation period is derived from the deviation feature triplet. The action speed of the deviation correction actuator is obtained through a preset oscillation period mapping table. The preset oscillation period mapping table divides the oscillation period into multiple intervals, each interval corresponding to a deviation correction actuator action speed. The oscillation period interval of 0 seconds to 0.5 seconds corresponds to a deviation correction actuator action speed of 20 mm / s; the oscillation period interval of 0.5 seconds to 1.5 seconds corresponds to a deviation correction actuator action speed of 12 mm / s; the oscillation period interval of 1.5 seconds to 3.0 seconds corresponds to a deviation correction actuator action speed of 8 mm / s; and the oscillation period interval of more than 3.0 seconds corresponds to a deviation correction actuator action speed of 5 mm / s. The interval division is based on the principle that the shorter the oscillation period, the higher the required action speed of the deviation correction actuator.

[0054] When the current misalignment mode is a sudden impact mode, the correction parameter generation module calculates the upper limit of the idler roll forward tilt angle adjustment and the limit of the correction actuator's operating speed based on the abrupt change amplitude. The abrupt change amplitude is derived from the misalignment feature triplet. The upper limit of the idler roll forward tilt angle adjustment is obtained through a preset abrupt change amplitude mapping table. This table divides the abrupt change amplitude into multiple intervals, each corresponding to an upper limit of the idler roll forward tilt angle adjustment. An abrupt change amplitude interval of 0 mm to 2 mm corresponds to an upper limit of 1.0 degree for the idler roll forward tilt angle adjustment; an abrupt change amplitude interval of 2 mm to 5 mm corresponds to an upper limit of 2.0 degree; and an abrupt change amplitude interval greater than 5 mm corresponds to an upper limit of 3.0 degree. The interval division is based on the principle that the larger the abrupt change amplitude, the higher the allowable upper limit of the idler roll forward tilt angle adjustment. The limit of the correction actuator's operating speed is obtained through the preset abrupt change amplitude mapping table. The amplitude range of sudden change from 0 mm to 2 mm corresponds to a speed limit of 15 mm / s for the corrective actuator; the amplitude range of sudden change from 2 mm to 5 mm corresponds to a speed limit of 25 mm / s for the corrective actuator; and the amplitude range of sudden change greater than 5 mm corresponds to a speed limit of 35 mm / s for the corrective actuator. The interval division is based on the fact that the larger the amplitude of the sudden change, the higher the speed limit of the corrective actuator.

[0055] When the current misalignment mode is unknown, the correction parameter generation module uses preset default correction parameters as the initial set of correction control parameters. The preset default correction parameters include: left tension adjustment of 0 Newtons, right tension adjustment of 0 Newtons, idler roll forward tilt angle adjustment of 1.0 degree, correction actuator operating speed of 10 mm / s, upper limit of idler roll forward tilt angle adjustment of 2.0 degrees, and limit of correction actuator operating speed of 20 mm / s. The preset default correction parameters are set based on using intermediate-level correction parameters when the misalignment mode is unclear.

[0056] When the current deviation mode is not an unknown mode, the correction parameter generation module combines the calculated left tension adjustment, right tension adjustment, idler roll tilt angle adjustment, and action speed or limit value into an initial correction control parameter set. For the gradual deviation mode, the initial correction control parameter set includes the left tension adjustment, right tension adjustment, and idler roll tilt angle adjustment. For the reciprocating oscillation mode, the initial correction control parameter set includes the correction actuator action speed. For the sudden impact mode, the initial correction control parameter set includes the upper limit of the idler roll tilt angle adjustment and the limit value of the correction actuator action speed.

[0057] In practice, the correction parameter generation module acquires current fluctuation data. This current fluctuation data originates from the data acquisition module. The correction parameter generation module extracts the mean current, current ripple amplitude, and current rate of change from the current fluctuation data. The mean current is the arithmetic mean of the amplitudes of all sampling points of the current fluctuation data within a set time window. The current ripple amplitude is the difference between the maximum and minimum values ​​of the current fluctuation data within the set time window. The current rate of change is the average of the absolute values ​​of the differences between adjacent sampling points of the current fluctuation data within the set time window.

[0058] The correction parameter generation module calculates the ratio of current ripple amplitude to the average current as a load fluctuation characteristic. This load fluctuation characteristic is dimensionless. The correction parameter generation module presets a fluctuation threshold of 0.15. This threshold is set based on the fact that the ratio of current ripple amplitude to the average current is typically below 0.15 during normal load fluctuations in the conveyor belt; a value exceeding 0.15 indicates significant load fluctuation.

[0059] When the load fluctuation characteristics exceed the preset fluctuation threshold, the correction parameter generation module proportionally reduces the idler roll tilt angle adjustment and proportionally increases the correction actuator speed. The reduction ratio of the idler roll tilt angle adjustment is set to 0.8, and the increase ratio of the correction actuator speed is set to 1.2. The reason for setting the reduction ratio of 0.8 and the increase ratio of 1.2 is that reducing the idler roll tilt angle adjustment can avoid excessive angle changes when the load fluctuation is significant, and increasing the correction actuator speed can speed up the response. When the load fluctuation characteristics are less than or equal to the preset fluctuation threshold, the correction parameter generation module keeps the idler roll tilt angle adjustment unchanged.

[0060] The correction parameter generation module writes the adjusted idler tilt angle and correction actuator speed into the initial correction control parameter set. If the initial correction control parameter set originally included the idler tilt angle adjustment, it replaces it with the adjusted idler tilt angle adjustment. If the initial correction control parameter set originally included the correction actuator speed, it replaces it with the adjusted correction actuator speed. If the initial correction control parameter set does not include the idler tilt angle adjustment or the correction actuator speed, the adjusted parameters are written directly.

[0061] See Figure 5 The horizontal axis of the graph represents the monotonic offset rate in millimeters per second, and the vertical axis represents the adjustment amount of the idler roller's forward tilt angle in degrees. The graph contains three types of data curves: the preset monotonic offset rate mapping table (blue stepped curve), the adjustment amount of the idler roller's forward tilt angle (red scatter dots), and the adjustment amount of the idler roller's forward tilt angle after current fluctuation adjustment (green triangular scatter dots).

[0062] The preset monotonic offset rate mapping table is presented in a stepped shape, divided into four distinct horizontal intervals: the monotonic offset rate interval of 0 to 0.5 mm / s corresponds to an adjustment of approximately 0.5 degrees for the idler roller tilt angle; the interval of 0.5 to 1.5 mm / s corresponds to an adjustment of approximately 1.0 degree; the interval of 1.5 to 3.0 mm / s corresponds to an adjustment of approximately 1.5 degrees; and the interval greater than 3.0 mm / s corresponds to an adjustment of approximately 2.0 degrees. The jump positions of the steps strictly correspond to the preset monotonic offset rate interval divisions, reflecting the stepwise increasing relationship between the idler roller tilt angle adjustment and the monotonic offset rate.

[0063] The red scatter dots represent the actual measured or calculated adjustment amount of the idler roller's forward tilt angle. The dots are densely clustered around the blue stepped curve, indicating that the actual adjustment amount basically conforms to the pattern of the preset mapping table, but there is a certain degree of random fluctuation, reflecting the parameter adjustment details and errors in actual applications.

[0064] The green triangular scatter plots illustrate the changing trend of the idler roll tilt angle adjustment after current fluctuation adjustment. Their values ​​are generally lower than the unadjusted red scatter plots, and are relatively concentrated within each monotonic offset rate range, with small fluctuation amplitudes. This characteristic corresponds to the adjustment strategy of proportionally reducing the idler roll tilt angle adjustment when the load fluctuation characteristic exceeds a threshold, demonstrating the impact and compensation effect of current fluctuations on the idler roll tilt angle adjustment.

[0065] In practice, the closed-loop correction module sends the initial set of correction control parameters to the correction actuator. The correction actuator then performs its actions according to the left tension adjustment, right tension adjustment, idler roll tilt angle adjustment, correction actuator operating speed, and the upper limit of the idler roll tilt angle adjustment or the limit of the correction actuator operating speed in the initial set of correction control parameters.

[0066] After the correction actuator executes the initial set of correction control parameters, the closed-loop correction module reacquires the misalignment displacement timing data according to a preset stabilization waiting time. The preset stabilization waiting time is set to 5 seconds, based on the assumption that the transition time required for the conveyor belt misalignment displacement to reach a steady state after the correction actuator's action is within 5 seconds. The closed-loop correction module reacquires the misalignment displacement timing data from the data acquisition module, which includes a timestamp, conveyor belt mileage markers, and misalignment displacement values.

[0067] The closed-loop correction module extracts the steady-state segment after the stabilization waiting period. The steady-state segment is a 500-sampling-point length of the misalignment displacement time series data, starting from the end of the stabilization waiting period. The length of the steady-state segment is set to 500 sampling points because 500 sampling points can cover the steady-state operation cycle of the conveyor belt and contain sufficient statistical information.

[0068] The closed-loop correction module calculates the average deviation displacement in the steady-state section. The average deviation displacement in the steady-state section is equal to the algebraic sum of all deviation displacement sampling points within the steady-state section divided by the number of deviation displacement sampling points within the steady-state section. The closed-loop correction module calculates the difference between the average deviation displacement in the steady-state section and the preset centerline as the residual deviation. The preset centerline value is set to 0 mm, corresponding to the conveyor belt centerline position. The residual deviation is equal to the average deviation displacement in the steady-state section minus the preset centerline value.

[0069] The closed-loop correction module calculates the standard deviation of the deviation displacement in the steady-state section as the fluctuation evaluation value. The formula for calculating the fluctuation evaluation value is: in, This represents the fluctuation evaluation value, in millimeters. This indicates the number of sampling points for deviation displacement in the steady-state section. It is 500; This indicates the sequence number of the deviation displacement sampling point in the steady-state section. Integer and The value range is 1 to ; Indicates the first segment in the steady-state region The values ​​of each deviation displacement sampling point are in millimeters. This represents the average value of the deviation displacement sampling points in the steady-state section, in millimeters.

[0070] The closed-loop correction module has a preset residual deviation threshold of 2 mm, based on the assumption that the allowable deviation displacement of the conveyor belt does not exceed 2 mm. The closed-loop correction module also has a preset fluctuation tolerance value of 1.5 mm, based on the assumption that a deviation displacement standard deviation exceeding 1.5 mm in the steady-state section indicates significant fluctuations in the correction process.

[0071] If the residual deviation exceeds the residual deviation threshold and the fluctuation evaluation value is less than the fluctuation tolerance value, the closed-loop correction module maintains the current actuator speed unchanged, decreases the left tension adjustment by a fixed step size, and increases the idler roll tilt angle adjustment. The fixed step size for decreasing the left tension adjustment is set to 50 Newtons, based on the principle that a single adjustment of 50 Newtons can gradually correct the residual deviation without causing excessive tension changes. The fixed step size for increasing the idler roll tilt angle adjustment is set to 0.2 degrees, based on the principle that a single adjustment of 0.2 degrees can gradually increase the idler roll tilt angle without causing excessive conveyor belt deflection. If the initial correction control parameter set does not include the left tension adjustment, the left tension adjustment is decreased from 0 Newtons by a fixed step size. If the initial correction control parameter set does not include the idler roll tilt angle adjustment, the idler roll tilt angle adjustment is increased from the current default value by a fixed step size.

[0072] If the fluctuation evaluation value is greater than or equal to the fluctuation tolerance value, the closed-loop correction module increases the operating speed of the correction actuator and decreases the iteration step size of the idler roll tilt angle adjustment. The method for increasing the operating speed of the correction actuator is to increase it by 5 mm / s based on the current operating speed. The basis for setting the increase of 5 mm / s is that increasing the operating speed can accelerate the correction response and suppress fluctuations. The method for decreasing the iteration step size of the idler roll tilt angle adjustment is to decrease the iteration step size of the idler roll tilt angle adjustment from 0.2 degrees to 0.1 degrees. The basis for setting the decrease of the iteration step size is that reducing the single angle adjustment amplitude can avoid over-adjustment leading to increased fluctuations. If the initial correction control parameter set does not include the correction actuator operating speed, then the current correction actuator operating speed is set to the preset default correction actuator operating speed and then increased by 5 mm / s.

[0073] The closed-loop correction module records the updated parameters as the iterative correction control parameters for the next iteration. The iterative correction control parameters include the corresponding items involved in this iteration from the updated left tension adjustment, right tension adjustment, idler roll forward tilt angle adjustment, correction actuator operating speed, upper limit of idler roll forward tilt angle adjustment, and limit of correction actuator operating speed.

[0074] In practice, after generating the iterative correction control parameters, the closed-loop correction module calculates the parameter differences between corresponding items in the current iteration and the previous iteration. These corresponding items include the left tension adjustment, right tension adjustment, idler roll tilt angle adjustment, correction actuator speed, upper limit of the idler roll tilt angle adjustment, and limit of the correction actuator speed. If an item exists only in the current iteration or only in the previous iteration's iterative correction control parameters, that item is not included in the parameter difference calculation. The absolute value of the parameter difference is the absolute value of the difference between the value of the corresponding item in the current iteration and the value of the corresponding item in the previous iteration.

[0075] The closed-loop correction module has preset parameter stability thresholds. The stability thresholds for the left tension adjustment parameter and the right tension adjustment parameter are both set to 50 Newtons. The stability threshold for the idler roller tilt angle adjustment parameter is set to 0.1 degrees. The stability threshold for the correction actuator's operating speed parameter is set to 2 millimeters per second. The stability threshold for the upper limit of the idler roller tilt angle adjustment parameter is set to 0.2 degrees. The stability threshold for the correction actuator's operating speed limit parameter is also set to 2 millimeters per second. These parameter stability thresholds are set based on the principle that when the difference between corresponding terms in two iterations is less than the stability threshold, parameter changes have a minimal impact on the correction effect.

[0076] If the absolute value of the difference between all corresponding parameters involved in the calculation is less than the corresponding parameter stability threshold, the closed-loop correction module will mark the iterative correction control parameter of this iteration as a stable parameter.

[0077] The closed-loop correction module simultaneously monitors current fluctuation data, which originates from the data acquisition module. The closed-loop correction module presets a current surge threshold, set to 0.2 times the average current value. This threshold is based on the assumption that a current spike exceeding the average current value by 20% indicates an abnormal current surge. If a spike exceeding the current surge threshold appears in the current fluctuation data, the actuator speed in the stabilization parameters is increased by a preset increment, and the residual deviation calculation is re-executed. The preset increment is set to 3 mm / s, based on the assumption that increasing the speed by 3 mm / s can appropriately improve the actuator speed to cope with load surges caused by current spikes. If the stabilization parameters do not include the actuator speed, the actuator speed is first set to the preset default speed before increasing the preset increment. If no spike exceeding the current surge threshold appears in the current fluctuation data, the stabilization parameters are directly sent to the actuator.

[0078] When the closed-loop correction module re-executes the residual deviation calculation, it returns to the step of re-acquiring the deviation displacement time series data according to the preset stabilization waiting time, re-extracts the steady-state segment after the stabilization waiting time, recalculates the residual deviation and fluctuation evaluation value, and continues to perform parameter updates based on the recalculated residual deviation and fluctuation evaluation value.

[0079] See Figure 6 In the figure, the horizontal axis represents the number of iterations, and the vertical axis represents the numerical value in millimeters. The blue solid line represents the trend of residual deviation with the number of iterations, and the red dashed line represents the trend of fluctuation evaluation value with the number of iterations. The figure also uses gray dashed lines and gray dotted lines to indicate the preset residual deviation threshold and fluctuation tolerance value, respectively, with threshold values ​​of 2 mm and 1.5 mm.

[0080] As shown in the figure, in the initial iteration stage (iterations 0 to 10), the residual deviation was relatively large, starting at about 5 mm, significantly higher than the preset residual deviation threshold of 2 mm. The initial fluctuation evaluation value was about 3 mm, also higher than the fluctuation tolerance value of 1.5 mm. As the number of iterations increased, the residual deviation showed an overall decreasing trend, gradually approaching and falling below the threshold of 2 mm after about 20 iterations. The fluctuation evaluation value decreased synchronously, stabilizing below 1.5 mm after about 20 iterations.

[0081] During approximately 10 to 20 iterations, the residual deviation decreased rapidly and the fluctuation evaluation value gradually decreased, indicating that the closed-loop correction module effectively adjusted the correction control parameters based on the residual deviation and fluctuation evaluation value, and the correction effect gradually improved. After more than 20 iterations, both the residual deviation and the fluctuation evaluation value remained within the preset threshold range, indicating that the conveyor belt deviation state entered a steady state, and the fluctuation amplitude and residual deviation were within the allowable range.

[0082] In specific implementation, please refer to Figure 2 The data acquisition module uses a hardware clock source to trigger the synchronous acquisition of data from the laser rangefinder array, speed encoder, and load cell. The hardware clock source is located internally within the module and outputs periodic pulse signals with a period of 1 millisecond. This period is set based on the requirement for millisecond-level synchronization accuracy for belt speed timing data, belt misalignment timing data, idler load distribution data, and current fluctuation data. The periodic pulse signals output by the hardware clock source are simultaneously connected to the trigger ports of the laser rangefinder array, the speed encoder, and the load cell. Upon receiving the periodic pulse signal, the laser rangefinder array performs a belt misalignment measurement; the speed encoder performs a belt speed measurement; and the load cell performs an idler load measurement.

[0083] The laser rangefinder array consists of multiple laser rangefinders arranged at equal intervals along the transverse direction of the conveyor belt. The spacing between adjacent laser rangefinders in the array is set to 100 mm, based on the requirement that the detection resolution for lateral belt misalignment should be no less than 100 mm. Each laser rangefinder measures the vertical distance from its installation position to the edge of the conveyor belt and outputs the distance value. The laser rangefinder array simultaneously collects the distance values ​​from all laser rangefinders upon the arrival of each periodic pulse signal, obtaining a set of misalignment displacement data. A speed encoder is installed on the shaft end of the conveyor belt drive roller. The speed encoder reads the rotation angle of the drive roller upon the arrival of each periodic pulse signal, and calculates the current belt speed based on the rotation angle and the diameter of the drive roller. A load cell is installed below the conveyor belt idler rollers. The load cell reads the idler roller load value upon the arrival of each periodic pulse signal, obtaining idler roller load distribution data.

[0084] After data acquisition, the data acquisition module adds a timestamp and conveyor belt mileage identifier to each frame of data. The timestamp is provided by a hardware clock source and is calculated by multiplying the count value of the periodic pulse signal by the pulse period of the hardware clock source; the unit of the timestamp is milliseconds. The conveyor belt mileage identifier is calculated from the cumulative rotation count of the speed encoder; the unit of the conveyor belt mileage identifier is the cumulative circumference of the drive roller, measured in meters. The calculation method for the conveyor belt mileage identifier is: cumulative rotation count of the speed encoder multiplied by the diameter of the drive roller multiplied by pi. The data acquisition module binds the belt speed timing data, misalignment displacement timing data, idler load distribution data, and current fluctuation data at the same timestamp into a synchronous data tuple. The synchronous data tuple contains a timestamp, a conveyor belt mileage identifier, a current belt speed value, a set of misalignment displacement data, a set of idler load data, and a set of current fluctuation data. The four types of data at the same timestamp are synchronously acquired by the same hardware clock source pulse trigger.

[0085] Current fluctuation data is collected by a current sensor located on the drive motor power supply line. The current sensor is also triggered by periodic pulse signals output from a hardware clock source, reading the current value of the drive motor power supply line once each periodic pulse signal arrives. The data acquisition module binds the current value at the same timestamp with belt speed timing data, belt misalignment timing data, and idler load distribution data into a synchronous data tuple.

[0086] The data acquisition module generates a displacement increment symbol sequence based on the deviation displacement difference between adjacent time stamps. The deviation displacement difference is obtained by calculating the difference between the representative deviation displacement values ​​in the time series data of deviation displacement at adjacent time stamps. The representative deviation displacement value is the distance value collected by the laser ranging sensor located directly above the center line of the conveyor belt in the laser ranging sensor array. The deviation displacement difference between adjacent time stamps is equal to the representative deviation displacement value corresponding to the later time stamp minus the representative deviation displacement value corresponding to the earlier time stamp. Each element in the displacement increment symbol sequence is the sign of the deviation displacement difference, which can be positive, negative, or zero. When the deviation displacement difference is greater than 0, the element in the displacement increment symbol sequence is recorded as +1; when the deviation displacement difference is less than 0, the element in the displacement increment symbol sequence is recorded as -1; when the deviation displacement difference is equal to 0, the element in the displacement increment symbol sequence is recorded as 0.

[0087] The data acquisition module calculates the length of consecutive identical symbols as the drift length. Consecutive identical symbols refer to adjacent elements in the displacement increment symbol sequence that have the same symbol. The drift length is equal to the number of consecutive identical symbols in the displacement increment symbol sequence traced back from the current time. If the current displacement increment symbol sequence element is +1, and there are 5 consecutive +1 elements traced back, then the drift length is 5. The drift length does not include elements whose symbols were 0 or opposite to the current symbol before the current time. The data acquisition module recalculates the drift length at each timestamp.

[0088] The data acquisition module multiplies the drift length by the current belt speed value in the belt speed time series data to obtain the cumulative offset distance. The formula for calculating the cumulative offset distance is: in, This indicates the cumulative offset distance, in meters. It represents the drift length, which is the number of consecutive displacement increments with the same sign; This indicates the current belt speed value in the belt speed timing data, in meters per second. This indicates the sampling period, in seconds. Equal to the hardware clock source pulse period, Set to 0.001 seconds, based on the hardware clock source pulse period of 1 millisecond. The method for obtaining this is to count the number of consecutive elements with the same symbol. It is an integer with a minimum value of 0. When there are no consecutive identical symbols in the displacement increment symbol sequence. Take 0. The speed value is obtained by measuring the current belt speed value at the corresponding timestamp using a speed encoder.

[0089] The data acquisition module writes the cumulative offset distance into the drift displacement time series data. This is done by adding a new field named "Cumulative Offset Distance" to the drift displacement time series data. This field is aligned with the timestamp, with each timestamp corresponding to a cumulative offset distance value. If the drift length is 0, the cumulative offset distance is 0. The data acquisition module then provides the drift displacement time series data with the cumulative offset distance written into it to the signal preprocessing module.

[0090] See Figure 7 In the graph, the horizontal axis represents time (seconds), the left side of the vertical axis represents the representative value of the misalignment displacement (mm), and the right side represents the cumulative offset distance (mm). The solid black line represents the representative value of the misalignment displacement, and the dashed red line represents the cumulative offset distance. Starting from 0 seconds, the representative value of the misalignment displacement shows an overall trend of first rising and then falling, reaching a peak of approximately 0.52 mm at about 0.45 seconds, and then gradually decreasing to near 0 mm, reflecting the dynamic change process of the conveyor belt misalignment displacement. The cumulative offset distance remains in a low amplitude range throughout the sampling time, with numerical fluctuations concentrated between 0 mm and 0.012 mm, and there is no obvious continuous increasing trend, reflecting a low number of consecutive identical symbols in the displacement increment symbol sequence, a short drift length, and limited changes in the cumulative offset distance.

[0091] As described in the above embodiment, the data acquisition module synchronously triggers the laser ranging sensor array to collect belt misalignment displacement data via a hardware clock source. The cumulative drift length is multiplied by the current belt speed and the sampling period to obtain the cumulative offset distance. The low amplitude and small fluctuations of the cumulative offset distance in the figure indicate that the conveyor belt misalignment direction changes frequently within this time period. The short drift length represents that the conveyor belt misalignment has not been continuously and stably offset in a single direction. The change in the representative value of the misalignment displacement reflects the real-time displacement dynamics of the conveyor belt edge measured by the laser ranging sensor array. The peak value of the curve and the subsequent decline correspond to the acceleration and correction process of the misalignment.

[0092] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An adaptive ore conveyor belt deviation correction control system, characterized in that, The system includes: The data acquisition module is used to synchronously collect belt speed time data, misalignment displacement time data, idler load distribution data, and current fluctuation data of the ore conveyor belt in the transportation or storage device. The signal preprocessing module is used to perform multi-scale decomposition on the deviation displacement time series data to obtain low-frequency trend components, periodic fluctuation components and random mutation components, and generate deviation feature triplet. The deviation correction pattern recognition module is used to match the deviation feature triplet with a preset deviation pattern library to determine the current deviation pattern; The deviation correction parameter generation module is used to generate an initial set of deviation correction control parameters based on the current deviation mode, the belt speed timing data, and the idler load distribution data. The closed-loop correction module is used to send the initial set of correction control parameters to the correction actuator, and after execution, reacquire the deviation displacement time series data, calculate the residual deviation, update the initial set of correction control parameters, and generate iterative correction control parameters.

2. The anti-deviation adaptive ore conveyor belt correction control system according to claim 1, characterized in that, The signal preprocessing module is used to divide the deviation displacement time series data into multiple subsequences according to a preset window, and perform variational mode decomposition on each subsequence to obtain multiple modal components; Calculate the center frequency and energy percentage of each modal component; The modal component with the lowest center frequency is superimposed as a low-frequency trend component; the two modal components with the middle center frequency and the largest energy proportion are superimposed as a periodic fluctuation component; and the modal component with the largest kurtosis value among the remaining modal components is superimposed as a random mutation component. The monotonic offset rate is calculated based on the difference between the first and last values ​​of the low-frequency trend component within the window, the oscillation period is calculated based on the interval between adjacent zero crossings of the periodic fluctuation component, and the mutation amplitude is calculated based on the peak value of the random mutation component. These are then summarized into the deviation feature triplet.

3. The anti-deviation adaptive ore conveyor belt correction control system according to claim 2, characterized in that, The deviation correction pattern recognition module is used to map the monotonic offset rate, oscillation period and abrupt change amplitude in the deviation feature triplet to a three-dimensional feature space to obtain three-dimensional feature points. The pre-trained deviation pattern classification network is invoked to calculate the distance score between the three-dimensional feature points and the center vectors of the three patterns corresponding to the progressive deviation pattern, the reciprocating oscillation pattern, and the sudden impact pattern. The current deviation pattern is determined based on the preset pattern matching threshold and the distance score; If the minimum distance score is greater than or equal to the pattern matching threshold, the current deviation pattern is marked as an unknown pattern. If the minimum distance score is less than the pattern matching threshold, then the pattern corresponding to the minimum distance score is selected as the current deviation pattern.

4. The anti-deviation adaptive ore conveyor belt correction control system according to claim 3, characterized in that, The correction parameter generation module is used to calculate the load difference between the left and right idlers based on the idler load distribution data when the current deviation mode is the progressive offset mode, convert the load difference into the left tension adjustment amount and the right tension adjustment amount, and obtain the idler forward tilt angle adjustment amount based on the monotonic offset rate mapping. When the current deviation mode is a reciprocating oscillation mode, the action speed of the correction actuator is calculated based on the oscillation period; When the current deviation mode is a sudden impact mode, the upper limit of the roller forward tilt angle adjustment and the limit of the correction actuator action speed are calculated based on the sudden change amplitude. When the current deviation mode is unknown, the preset default deviation correction parameters are used as the initial deviation correction control parameter set. Otherwise, the calculated left tension adjustment, right tension adjustment, idler roll tilt angle adjustment, action speed, or limit value will be combined into the initial set of correction control parameters.

5. The anti-deviation adaptive ore conveyor belt correction control system according to claim 4, characterized in that, The closed-loop correction module is used to reacquire the deviation displacement time series data according to a preset stabilization waiting time after the deviation correction actuator executes the initial deviation correction control parameter set, and to extract the steady-state segment after the stabilization waiting time. The difference between the mean value of the deviation displacement in the steady-state section and the preset centerline is calculated as the residual deviation amount. At the same time, the standard deviation of the deviation displacement in the steady-state section is calculated as the fluctuation evaluation value. If the residual deviation is greater than the preset threshold and the fluctuation evaluation value is less than the preset fluctuation tolerance value, then keep the current correction actuator action speed unchanged, decrease the left tension adjustment amount by a fixed step size and increase the idler roller forward tilt angle adjustment amount. If the fluctuation evaluation value is greater than or equal to the preset fluctuation tolerance value, then increase the action speed of the correction actuator and decrease the iteration step size of the roller forward tilt angle adjustment. The updated parameters are recorded as the iterative correction control parameters for the next iteration.

6. The anti-deviation adaptive ore conveyor belt correction control system according to claim 1, characterized in that, The data acquisition module is used to trigger the synchronous acquisition of the laser rangefinder array, speed encoder and weighing sensor with a hardware clock source; After collection, a timestamp and conveyor belt running mileage identifier are added to each frame of data, and the belt speed time sequence data, the misalignment displacement time sequence data, the idler load distribution data and the current fluctuation data under the same timestamp are bound into a synchronous data tuple; A displacement increment symbol sequence is generated based on the deviation displacement difference between adjacent timestamps. The length of consecutive identical symbols is counted as the drift length. The drift length is multiplied by the current belt speed value in the belt speed time series data to obtain the cumulative offset distance. The cumulative offset distance is then written into the deviation displacement time series data.

7. The anti-deviation adaptive ore conveyor belt correction control system according to claim 2, characterized in that, The signal preprocessing module is used to grid the idler load distribution data according to the transverse coordinate of the conveyor belt to obtain the transverse load distribution curve. Calculate the lateral distance between the centroid of the lateral load distribution curve and the centerline of the conveyor belt to obtain the load offset; The load offset is compared synchronously with the monotonic offset rate of the low-frequency trend component within a time window. If the load offset and the monotonic offset rate have the same sign within the set time window and the rate of change of the load offset exceeds a preset rate of change threshold, a load offset warning mark is generated. The load offset warning marker and the deviation feature triplet are sent together to the deviation correction pattern recognition module.

8. The anti-deviation adaptive ore conveyor belt correction control system according to claim 4, characterized in that, The correction parameter generation module is used to acquire the current fluctuation data and extract the mean current, current pulsation amplitude and current change rate from the current fluctuation data. The ratio of the current ripple amplitude to the average current value is calculated as a load fluctuation characteristic. When the load fluctuation characteristic is greater than the preset fluctuation threshold, the adjustment amount of the roller forward tilt angle is reduced proportionally, and the action speed of the correction actuator is increased proportionally. When the load fluctuation characteristic is less than or equal to the preset fluctuation threshold, the forward tilt angle adjustment of the idler roller remains unchanged; The adjusted idler roll tilt angle and the corrective actuator speed are written into the initial corrective control parameter set.

9. The anti-deviation adaptive ore conveyor belt correction control system according to claim 5, characterized in that, The closed-loop correction module is used to calculate the parameter difference between each corresponding item in the iterative correction control parameters of the current iteration and the previous iteration after generating the iterative correction control parameters. If the absolute value of all the parameter differences is less than the preset parameter stability threshold, then the iterative correction control parameter of this iteration is marked as a stable parameter. Simultaneously monitor the current fluctuation data. If a peak exceeding the preset current mutation threshold appears in the current fluctuation data, increase the action speed of the correction actuator in the stabilization parameters by a preset increment and re-execute the residual deviation calculation; otherwise, directly send the stabilization parameters to the correction actuator.

10. The anti-deviation adaptive ore conveyor belt correction control system according to claim 3, characterized in that, The deviation correction mode recognition module is used to calculate the load difference between the left idler and the right idler in the idler load distribution data after determining the current deviation mode according to the mode matching threshold, and dynamically adjust the mode matching threshold according to the current belt speed of the belt speed time sequence data and the load difference. When the current belt speed is greater than the preset high-speed threshold, the matching threshold of the progressive offset mode is increased; When the load difference is greater than the preset off-center load threshold, the matching threshold of the sudden impact mode is reduced. The adjusted pattern matching threshold is output together with the current deviation pattern as the pattern recognition result, and the adjusted pattern matching threshold is used for the next deviation feature triplet matching.

Images