A belt conveyor intelligent speed regulation and multi-drive power balance control method

Abstract

The present application relates to the technical field of belt conveyor control, and specifically discloses a kind of belt conveyor intelligent speed regulation and multi-drive power balance control method, three-dimensional space coordinates and corresponding echo amplitude, pulse width and surface texture features of material surface are collected, total weight value is synchronously obtained and timing synchronization and space matching are carried out;Calculate intrinsic reflectivity and surface roughness index;Intrinsic reflectivity and surface roughness index are mapped to heavy component probability weight coefficient, forming two-dimensional virtual density prior distribution;With the prior distribution as a constraint, combine the total weight value with the volume cross-sectional area, and through the deviation amount reverse apportioning and the peak clipping type correction alternative iterative inversion of the instantaneous two-dimensional real density distribution, the density distribution is point by point convolution calculation with the volume cross-sectional area to obtain the lateral unbalance load additional resistance moment, which is added to the torque given from the drive unit as a feedforward quantity, while adjusting the constraint strength according to the residual error.

Description

Technical Field

[0001] This invention relates to the field of belt conveyor control technology, specifically to a method for intelligent speed regulation and multi-drive power balance control of belt conveyors. Background Technology

[0002] As the core equipment for continuous transportation of bulk materials, belt conveyors rely heavily on intelligent speed regulation and multi-drive power balance control technology to ensure safe and economical system operation. In existing control schemes, industrial vision sensing technology has been used to acquire three-dimensional spatial geometric information of the conveyed material surface. By reconstructing the material's cross-sectional profile, the volumetric flow rate is calculated, and the conveyor belt speed is adjusted accordingly to achieve on-demand speed regulation. Simultaneously, the lateral off-center load geometric distribution acquired through vision is also being explored for torque feedforward compensation in multi-drive units.

[0003] In existing belt conveyor control schemes, industrial vision can only acquire three-dimensional geometric shape information of the material surface, while the weight data provided by the belt scale has transmission lag and is only the total cumulative value in the bandwidth direction. The two lack a spatial synchronization mechanism in the density dimension, which makes it impossible to acquire the instantaneous two-dimensional density distribution of the material along the bandwidth direction in real time. This results in a mismatch between the feedforward compensation amount calculated based on geometric off-center load and the actual mechanical load on the shaft end borne by the drive drum. When the material properties change locally, multiple drive units experience power imbalance. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent speed regulation and multi-drive power balance control method for belt conveyors to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for intelligent speed regulation and multi-drive power balance control of a belt conveyor includes the following steps:

[0007] S1: Collect the three-dimensional spatial coordinates of the surface of the conveyed material, as well as the echo amplitude, pulse width and surface texture features corresponding to each three-dimensional spatial coordinate point. At the same time, obtain the total weight value of the conveyed material, and synchronize the spatial coordinates and weight values ​​in time and match their spatial positions.

[0008] S2: Calculate the local normal vector of each point based on the spatial coordinates, normalize and compensate the echo amplitude according to the angle between the normal vector and the incident light to obtain the intrinsic reflectivity, and obtain the surface roughness index based on the rate of change statistics of the normal vector. The points mentioned are each three-dimensional spatial coordinate points.

[0009] S3: The intrinsic reflectivity and surface roughness indices are mapped pixel by pixel along the bandwidth direction to weight coefficients that characterize the probability of heavy components, forming a two-dimensional virtual density prior distribution.

[0010] S4: Using the two-dimensional virtual density prior distribution as a constraint, combined with the total weight value and the volume cross-sectional area determined by the spatial coordinates, a sparse regularized iterative inversion is performed to recover the instantaneous two-dimensional true density distribution of the conveyed material along the bandwidth direction.

[0011] S5: Based on the point-by-point convolution operation of the two-dimensional true density distribution and the volume cross-sectional area, calculate the additional resistance torque generated by the lateral off-center load, and use the additional resistance torque as a feedforward quantity to be superimposed on the torque of the multi-drive unit. Adjust the constraint strength in the sparse inversion according to the residual between the prior distribution and the true density distribution.

[0012] As a further aspect of the present invention: the process of obtaining the intrinsic reflectivity is as follows:

[0013] The tilt attitude of the micro-region where the corresponding point is located is determined based on the local normal vector of each point.

[0014] Using the angle between the tilted orientation and the direction of the incident light as the incident angle, a pre-calibrated reference reflectivity response curve of the material surface in a non-tilted state is obtained.

[0015] Divide the echo amplitude at the corresponding point by the reference reflectivity at the incident angle, and multiply the resulting ratio by the preset diffuse reflection compensation coefficient to obtain the intrinsic reflectivity at the corresponding point.

[0016] As a further aspect of the present invention: the preset process for the diffuse reflection compensation coefficient is as follows:

[0017] Select a local neighborhood window for each point and calculate the average angle between the normal vector of each point within the local neighborhood window and the normal vector of the center point;

[0018] The average included angle is used to characterize the degree of undulation of the local surface. The ratio of the maximum echo intensity of the material under the standard diffuse reflection reference plate condition to the measured echo intensity under the local undulation is multiplied by the reciprocal of the cosine of the average included angle to obtain the diffuse reflection compensation coefficient of the corresponding point.

[0019] As a further aspect of the present invention: the measurement of surface roughness index specifically includes:

[0020] Select a local window centered on each point, encompassing its surrounding neighboring points;

[0021] Calculate the deflection angle of the normal vector of each point within the local window relative to the normal vector of the center point;

[0022] The root mean square (RMS) values ​​of all deflection angles are statistically analyzed, and the information entropy of the normal vector direction distribution within the local window is calculated. The weighted sum of the RMS values ​​and the information entropy is used as the surface roughness index of the corresponding point.

[0023] As a further aspect of the present invention: the process of establishing the two-dimensional virtual density prior distribution is as follows:

[0024] The intrinsic reflectivity and surface roughness indices are arranged pixel-by-pixel along the bandwidth to form two-dimensional feature pairs;

[0025] Divide the value space of two-dimensional feature pairs into multiple combination intervals;

[0026] For each combination interval, the probability ratio of reflectivity and surface roughness to the corresponding heavy component in the preset material sample library is used as the mapping weight of the combination interval.

[0027] The combination interval of each pixel is determined along the bandwidth direction and the corresponding mapping weight is assigned to the pixel to obtain the probability weight coefficient of the heavy component of each pixel. The probability weight coefficients of all pixels constitute a two-dimensional virtual density prior distribution.

[0028] As a further aspect of the present invention: the calculation process of the instantaneous two-dimensional true density distribution is as follows:

[0029] The initial value of the true density at each location is set to be equal to the corresponding value in the two-dimensional virtual density prior distribution;

[0030] Calculate the deviation between the sum of the products of the actual density at each current location and the corresponding cross-sectional area and the total weight value;

[0031] The deviation is back-allocated to each location according to the proportion of the volume cross-sectional area at each location to update the true density value, and the degree of deviation between the updated true density value at each location and the corresponding value in the prior distribution is calculated.

[0032] Based on the degree of deviation, the updated true density values ​​at each location are corrected by peak shaving and valley filling until the deviation is less than a preset threshold. The true density values ​​at each location at this time are then used as the instantaneous two-dimensional true density distribution.

[0033] As a further aspect of the present invention: the peak-shaving and valley-filling correction specifically includes:

[0034] Calculate the difference between the current true density value at each location and the corresponding value in the two-dimensional virtual density prior distribution;

[0035] The absolute value of the difference is used as the degree of deviation of the corresponding position;

[0036] Arrange the deviation of each position in descending order of numerical value, and select the first part of the arrangement as the position to be corrected;

[0037] For each position to be corrected, its current true density value is pulled back towards the corresponding value in the prior distribution. The pull-back amount is the difference multiplied by a preset scaling factor. The pulled-back value is used as the corrected true density value for the corresponding position. Positions that are not truncated retain their current true density value.

[0038] As a further aspect of the present invention: the process of adjusting the constraint strength is as follows:

[0039] Calculate the residuals between the actual density values ​​and the corresponding values ​​in the prior distribution at each location along the bandwidth direction;

[0040] The residuals are partitioned by positive and negative signs to obtain the total amplitude of the positive residual region, the total amplitude of the negative residual region, and the bandwidth ratio occupied by each region.

[0041] When the proportion of the larger of the positive and negative residual regions exceeds the preset proportion threshold, it is determined that the properties of the incoming material have shifted overall, and the constraint strength is reduced.

[0042] When positive and negative residual regions appear alternately and their amplitudes are all lower than the preset amplitude threshold, the prior distribution is determined to be locally inaccurate. At this time, the constraint strength is increased, and the adjusted constraint strength is used to participate in the sparse inversion of the next cycle.

[0043] The beneficial effects of this invention are:

[0044] (1) By normalizing the incident angle and compensating for diffuse reflection in the echo amplitude obtained by the laser scanning device, the interference of the local tilt of the material surface on the reflection intensity measurement is eliminated, so that the obtained intrinsic reflectivity is determined only by the material properties of the material itself. By combining the root mean square of the normal vector deflection angle and the distribution information entropy to quantify the surface roughness, it is possible to distinguish surface states with different undulation amplitudes and orientation disorder. Visual sensing data not only provides geometric shape information, but also reflects the essential differences in the material surface, providing a reliable visual feature basis for subsequent differentiation between light and heavy components.

[0045] (2) The essential reflectivity and surface roughness indices obtained by visual sensing are mapped to a priori density distribution along the bandwidth direction through a preset sample library. This priori distribution is used to constrain the sparse inversion process, and the instantaneous two-dimensional true density distribution of the material along the bandwidth direction is recovered through alternating iterations of reverse distribution of the deviation and peak-shaving and valley-filling correction. The acquisition of this density distribution does not depend on the addition of additional X-ray or microwave detection equipment, but can be achieved by isomorphic fusion of two heterogeneous sensing data, visual and weight, in the density space. This allows the subsequent calculation of additional drag torque to reflect the true gravity distribution of the off-center load, thereby matching the feedforward compensation amount with the actual mechanical load borne by the drive roller. Attached Figure Description

[0046] The invention will now be further described with reference to the accompanying drawings.

[0047] Figure 1 This is a flowchart of the method of the present invention;

[0048] Figure 2This is a flowchart of the process for obtaining intrinsic reflectivity in this invention;

[0049] Figure 3 This is a flowchart illustrating the process of measuring and obtaining surface roughness indices in this invention. Detailed Implementation

[0050] 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, and 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.

[0051] Please see Figure 1 As shown, this invention provides an intelligent speed regulation and multi-drive power balance control method for belt conveyors, comprising the following steps:

[0052] S1: Collect the three-dimensional spatial coordinates of the surface of the conveyed material, as well as the echo amplitude, pulse width and surface texture features corresponding to each three-dimensional spatial coordinate point. At the same time, obtain the total weight value of the conveyed material, and synchronize the spatial coordinates and weight values ​​in time and match their spatial positions.

[0053] S2: Calculate the local normal vector of each point based on the spatial coordinates, normalize and compensate the echo amplitude according to the angle between the normal vector and the incident light to obtain the intrinsic reflectivity, and obtain the surface roughness index based on the rate of change statistics of the normal vector. The points mentioned are each three-dimensional spatial coordinate points.

[0054] S3: The intrinsic reflectivity and surface roughness indices are mapped pixel by pixel along the bandwidth direction to weight coefficients that characterize the probability of heavy components, forming a two-dimensional virtual density prior distribution.

[0055] S4: Using the two-dimensional virtual density prior distribution as a constraint, combined with the total weight value and the volume cross-sectional area determined by the spatial coordinates, a sparse regularized iterative inversion is performed to recover the instantaneous two-dimensional true density distribution of the conveyed material along the bandwidth direction.

[0056] S5: Based on the point-by-point convolution operation of the two-dimensional true density distribution and the volume cross-sectional area, calculate the additional resistance torque generated by the lateral off-center load, and use the additional resistance torque as a feedforward quantity to be superimposed on the torque of the multi-drive unit. Adjust the constraint strength in the sparse inversion according to the residual between the prior distribution and the true density distribution.

[0057] In S1, the three-dimensional spatial coordinates of the conveyed material surface, as well as the echo amplitude, pulse width, and surface texture features corresponding to each three-dimensional spatial coordinate point, are collected. Simultaneously, the total weight of the conveyed material is obtained. The spatial coordinates and weight values ​​are synchronized in time and matched in spatial position, specifically including:

[0058] A laser scanning device is installed above the conveyor. This device emits multiple laser beams along the width of the conveyor to perform continuous cross-sectional scanning of the surface of the conveyed material.

[0059] The laser scanning device has a built-in timer that records scanning data once every preset acquisition cycle. Each scan obtains a set of three-dimensional spatial coordinate points that are discretely distributed along the width direction. The value of each coordinate point represents the height information of the material surface at that position.

[0060] When the photoelectric receiving component embedded in the laser scanning device receives the reflected light signal corresponding to each coordinate point, it simultaneously outputs the echo amplitude value and pulse width value of that coordinate point.

[0061] The surface texture features are output by the multi-echo processing unit of the laser scanning device after comparing the differences in pulse broadening from multiple transmissions and receptions at the same coordinate point.

[0062] The aforementioned three-dimensional spatial coordinate points, echo amplitude values, pulse width values, and surface texture feature values ​​are all numbered according to the arrangement order of the coordinate points in the width direction, forming a complete set of raw data of the material surface contour.

[0063] Multiple sets of load-bearing weighing idlers are installed below the weighing section of the conveyor, and each load-bearing weighing idler is connected to the same digital weighing instrument.

[0064] When the conveyed material passes through the weighing section, each load-bearing weighing roller converts the pressure it receives into an electrical signal and sends it to the weighing instrument. The weighing instrument accumulates the electrical signals of each roller and outputs an instantaneous total weight value.

[0065] The weighing instrument outputs an instantaneous total weight value once every acquisition cycle. The time difference between the output time of each instantaneous total weight value and the start trigger time of each scan by the laser scanning device is recorded by the clock circuit built into the weighing instrument.

[0066] A rotary encoder is installed on the shaft end of the drive drum of the conveyor. The rotary encoder outputs pulse signals as the drive drum rotates. The cumulative number of pulse signals is used to characterize the running distance of the conveyor belt from the section where the laser scanning device is located to the section where the weighing section is located.

[0067] The control unit reads the cumulative number of pulses from the rotary encoder in each acquisition cycle, and combines this with the diameter parameter of the drive roller to calculate the actual running distance of the conveyor belt in that acquisition cycle.

[0068] The control unit calculates the transmission delay time required for the material at the scanning section to move to the weighing section by dividing the fixed installation distance between the laser scanning device and the weighing section by the actual running distance mentioned above.

[0069] The control unit uses the transmission delay time as an offset to establish a time correspondence between the instantaneous total weight value output by the weighing instrument in the nth acquisition cycle and a set of three-dimensional spatial coordinate points, echo amplitude values, pulse width values ​​and surface texture feature values ​​obtained by the laser scanning device before the nth acquisition cycle after the transmission delay time.

[0070] The control unit then aligns the spatial positions corresponding to the numbered three-dimensional spatial coordinate points along the width direction with the arrangement positions of the load-bearing weighing rollers along the width direction, so that each numbered three-dimensional spatial coordinate point, echo amplitude value, pulse width value, and surface texture feature value corresponds to a specific bandwidth position and the instantaneous weight contribution component corresponding to the bandwidth position, thus completing the spatial position matching.

[0071] After the above-mentioned timing synchronization and spatial location matching, each set of three-dimensional spatial coordinate points and its corresponding echo amplitude value, pulse width value and surface texture feature value are established in a one-to-one correspondence with an instantaneous total weight value, and each data has a definite bandwidth location identifier.

[0072] Please see Figure 2 or Figure 3 As shown, in S2, the local normal vector of each point is calculated based on the spatial coordinates. The echo amplitude is normalized and compensated according to the angle between the normal vector and the incident ray to obtain the intrinsic reflectivity. The surface roughness index is obtained based on the rate of change statistic of the normal vector. The points mentioned are each three-dimensional spatial coordinate point, specifically including:

[0073] After the laser scanning device, fixedly installed above the conveyor, completes the cross-sectional scanning, each set of three-dimensional spatial coordinate points carries a corresponding echo amplitude value. For any three-dimensional spatial coordinate point, the local normal vector of the micro-region where the point is located is first determined by the least squares fitting plane method based on the spatial coordinates of the point and its surrounding neighboring points. The direction of the local normal vector represents the spatial tilt attitude of the micro-region.

[0074] The local normal vector is compared with the incident ray direction vector preset by the laser scanning device, and the angle between the two is calculated and recorded as the incident angle.

[0075] This step involves calibrating the reference reflectivity response curve of the material surface beforehand: Select a standard plate with the same material as the surface of the material to be tested, place it horizontally below the laser scanning device, set an angle adjustment bracket between the laser scanning device and the plate, and tilt the plate to several different angles in sequence, so that the angle between the laser beam and the plate normal is equal to 0 degrees, 5 degrees, 10 degrees, 15 degrees up to 85 degrees, and record the echo intensity value reflected back from the plate surface at each angle. Plot a continuous curve with the angle value on the x-axis and the echo intensity value on the y-axis, which is the reference reflectivity response curve of the material surface in the untilted state.

[0076] The curve is stored in the storage medium of the control unit, and the corresponding reference reflectivity can be obtained by interpolation based on any incident angle value.

[0077] After obtaining the echo amplitude and the incident angle at a certain point, the echo amplitude is divided by the reference reflectivity value corresponding to the incident angle to obtain the preliminary corrected reflection value.

[0078] The initial correction value also needs to be multiplied by a diffuse reflection compensation coefficient to eliminate the secondary scattering of light caused by local undulations on the material surface.

[0079] The diffuse reflection compensation coefficient is preset based on the surface undulation of the local area where the point is located.

[0080] The specific setting method is as follows: take the point as the center, select a neighborhood range with a radius equal to the distance between 10 adjacent points, and all points falling within this range constitute a local window.

[0081] Calculate the angle between the local normal vector of each point within the local window and the normal vector of the center point, and record the average of all the angles as the average angle. The larger the average angle, the more severe the local surface undulations.

[0082] At the same time, the arithmetic mean of the echo amplitude of all points within the local window is obtained and denoted as the local average echo intensity.

[0083] Furthermore, the echo intensity value measured beforehand on a standard diffuse reflection reference plate (which is a metrologically certified Lambertian standard plate) under conditions of perpendicular laser incidence is recorded as the maximum reference intensity. For the current point, its diffuse reflection compensation coefficient is calculated according to formula (1): ;

[0084] in, Indicates the current point The diffuse reflection compensation coefficient; The maximum reference strength; This represents the total number of points contained within the local window. For the first in the window The echo amplitude at each point; For the first in the window The angle between the local normal vector of a point and the normal vector of the center point.

[0085] In formula (1), the first term of the denominator is the average value of the echo amplitude of all points within the window, and the second term of the denominator is the cosine value of the average value of all included angles within the window.

[0086] The physical meaning of this formula is: the ratio of the average local echo intensity to the maximum reference intensity reflects the overall reflectivity of the local surface; while the reciprocal of the cosine of the average angle is used to compensate for the reduction in effective receiving area caused by tilting.

[0087] Multiplying the two gives the diffuse reflection compensation coefficient for that point. Dividing the echo amplitude at the current point by the reference reflectivity and then multiplying by the compensation coefficient yields the intrinsic reflectivity at that point.

[0088] For each point in each set of three-dimensional spatial coordinates, its intrinsic reflectivity is calculated according to the above process.

[0089] The calculation of surface roughness index is also based on the local normal vector information of each point.

[0090] Still using the current point as the center, select a neighborhood range with a radius equal to the distance between 10 adjacent points to form a local window.

[0091] For each point within the window except the center point, calculate the deflection angle between the local normal vector of that point and the local normal vector of the center point. This deflection angle is the angle in three-dimensional space, and its value ranges from 0 degrees to 90 degrees.

[0092] To calculate the root mean square value of all deflection angles within the window, square each deflection angle, sum the results, divide by the total number of points within the window, and finally take the square root.

[0093] To characterize the uniformity of the deflection angle distribution across the entire window, the deflection angle range from 0 degrees to 90 degrees is divided into 9 equal sub-intervals. The number of deflection angles falling into each sub-interval is counted, and this number is divided by the total number of points within the window to obtain the probability value for each sub-interval, denoted as . ,in According to the definition of information entropy, each probability value is multiplied by its natural logarithm, and the result is negative. Then, it is multiplied by (1 divided by) Normalize the information to obtain the normalized information entropy value.

[0094] The root mean square value and the normalized information entropy value are weighted and summed, with weights of 0.6 and 0.4 respectively. The resulting weighted sum is the surface roughness index of that point.

[0095] This process can be uniformly represented by formula (2): ;

[0096] in, Indicates the current point Surface roughness index; For the first in the window The deflection angle of the local normal vector of each point (including the center point itself, whose deflection angle is 0 degrees) relative to the local normal vector of the center point; This represents the total number of points within the window. For the deflection angle to fall into the first The probability value of each subinterval.

[0097] In formula (2), the first term represents the overall magnitude of the deflection amplitude, and the second term represents the degree of disorder in the distribution of the deflection angle. The combination of the two can distinguish between two different surface states: one with gentle undulations but chaotic orientation, and the other with violent undulations but consistent orientation.

[0098] Following the above process, the corresponding surface roughness index is calculated for each point in each set of three-dimensional spatial coordinates.

[0099] In S3, the intrinsic reflectivity and surface roughness indices are mapped pixel-by-pixel along the bandwidth direction to weighting coefficients representing the probability of heavy components, forming a two-dimensional virtual density prior distribution, specifically including:

[0100] After obtaining the essential reflectivity and surface roughness indices of each three-dimensional spatial coordinate point on the material surface of the belt conveyor, it is necessary to fuse the feature information of these two dimensions into a weighting coefficient that can characterize the proportion of heavy components in the material at the bandwidth location of that point.

[0101] For each set of data acquired by cross-sectional scanning, according to the order of the points in the bandwidth direction, an intrinsic reflectivity value and a surface roughness index value located at the same bandwidth position are combined into a two-dimensional feature pair. The two components of the two-dimensional feature pair correspond to the reflectivity and geometric undulation properties of the material at that position, respectively.

[0102] Along the bandwidth direction, multiple such two-dimensional feature pairs are arranged sequentially from one edge to the other, with each two-dimensional feature pair corresponding to a discrete pixel position on the bandwidth.

[0103] A material sample library is established in advance. The establishment method of the sample library is as follows: collect multiple material samples from the same mining area as the material to be tested. These samples cover different coal types, gangue content and moisture content. For each sample, use the same laser scanning device as the field in the laboratory environment to obtain its intrinsic reflectivity and surface roughness index. At the same time, use density separation method (such as heavy liquid flotation test) to determine the mass percentage of heavy components (mineral components with density greater than 1.8 grams per cubic centimeter) in the sample.

[0104] Each sample's intrinsic reflectance value, surface roughness index value, and corresponding heavy component mass percentage are recorded as a sample record, and all sample records constitute a pre-defined material sample library.

[0105] The sample library contains no fewer than 200 sample records, covering the range of intrinsic reflectivity from 0.2 to 0.9 and surface roughness index from 0.1 to 0.8.

[0106] In the step of dividing the value space of the two-dimensional feature pairs, the value range of the intrinsic reflectivity from 0.2 to 0.9 is divided into 10 sub-intervals at equal intervals, and the width of each sub-interval is 0.07; the value range of the surface roughness index from 0.1 to 0.8 is divided into 10 sub-intervals at equal intervals, and the width of each sub-interval is 0.07.

[0107] The two dimensions each have 10 sub-intervals that intersect to form 100 combined intervals. Each combined interval is determined by the sub-interval number to which the intrinsic reflectivity belongs and the sub-interval number to which the surface roughness index belongs.

[0108] For example, when the intrinsic reflectivity falls into the 3rd sub-interval and the surface roughness index falls into the 5th sub-interval, the two-dimensional feature pair belongs to the combined interval (3, 5).

[0109] For each of the above 100 combination intervals, traverse all sample records in the preset material sample library, extract all sample records in the essential reflectivity sub-interval range and surface roughness index sub-interval range corresponding to the combination interval, count the number of samples in these sample records with a mass percentage of heavy components greater than 50%, divide by the total number of all sample records falling within the combination interval, and obtain the probability value of the occurrence of heavy components within the combination interval.

[0110] Simultaneously, the number of samples in the combined interval where the mass percentage of heavy components is less than or equal to 50% is counted, and then divided by the total number of all sample records falling within the combined interval to obtain the probability value of the occurrence of non-heavy components in the combined interval.

[0111] The ratio of the probability of the occurrence of heavy components to the probability of the occurrence of non-heavy components is used as the mapping weight of the combination interval.

[0112] If no sample record falls within a certain combination interval, the mapping weight of that combination interval is set to 1.0.

[0113] Following the above process, the mapping weight values ​​corresponding to each of the 100 combined intervals are calculated, forming a lookup table in which the combined interval number and the mapping weight value are matched one-to-one, and stored in the control unit.

[0114] During the mapping process in the bandwidth direction, for a two-dimensional feature pair located at a certain pixel position in the bandwidth direction in the cross-sectional scan data, it is determined which of the 10 sub-intervals of intrinsic reflectivity the intrinsic reflectivity value of the feature pair falls into, and its sub-interval number is recorded.

[0115] Next, determine which of the 10 sub-intervals of the surface roughness index value in the feature pair falls into, and record its sub-interval number; the combined interval number to which the pixel position belongs is determined by the two sub-interval numbers.

[0116] Read the mapping weight value corresponding to the combination interval number from the above lookup table, and assign the mapping weight value to the pixel position as the probability weight coefficient of the heavy component of the material at the pixel position.

[0117] When the weighting coefficient is greater than 1, it means that the probability of heavy components appearing in the material at this bandwidth location is higher than the probability of non-heavy components appearing. The larger the weighting coefficient is, the stronger the tendency for high-density materials to exist at this location.

[0118] A weighting coefficient less than 1 indicates a higher probability of the presence of lightweight components.

[0119] Along the entire bandwidth direction, the above-mentioned attribution determination and weight assignment operations are performed on each pixel position to obtain a one-dimensional sequence consisting of the probability weight coefficients of the heavy components of each pixel position along the bandwidth direction. This one-dimensional sequence is arranged along the bandwidth direction, which is a two-dimensional virtual density prior distribution. Each value in this prior distribution corresponds to a specific pixel position on the bandwidth.

[0120] In S4, using the two-dimensional virtual density prior distribution as a constraint, and combining the total weight value and the volume cross-sectional area determined by spatial coordinates, a sparse regularized iterative inversion is performed to recover the instantaneous two-dimensional true density distribution of the conveyed material along the bandwidth direction, specifically including:

[0121] After obtaining the two-dimensional virtual density prior distribution, the volume cross-sectional area at each bandwidth location, and the instantaneous total weight value, the sparse regularized iterative inversion is performed to recover the instantaneous two-dimensional true density distribution.

[0122] First, initialization is performed, setting the initial value of the true density value of each pixel location in the bandwidth direction to the weight coefficient value corresponding to that location in the two-dimensional virtual density prior distribution.

[0123] After initialization, the iteration loop begins. Before each iteration, for all positions along the bandwidth, the current true density value at each position is multiplied by the corresponding volume cross-sectional area value to obtain the volume density product at that position. The volume density products of all positions are then summed to obtain a cumulative total value.

[0124] Subtract the accumulated total value from the obtained instantaneous total weight value, calculate the absolute value of the difference between the two, and record it as the deviation.

[0125] This deviation represents the degree of deviation between the total weight calculated based on the current density distribution and the total weight actually measured by the sensor at the current iteration step.

[0126] Then, it is determined whether the deviation is less than the preset convergence threshold. This convergence threshold is determined according to the range of the weighing sensor of the conveyor and is set to 5 / 1000 of the instantaneous total weight value. If the deviation is less than the threshold, it is determined that the iteration has converged, the iteration process is terminated, and the density values ​​of each position are output as the instantaneous two-dimensional true density distribution.

[0127] If the deviation is greater than or equal to the threshold, a density value update operation is performed.

[0128] In the density value update operation, the sum of the volume cross-sectional area values ​​at all locations along the bandwidth direction is calculated to obtain the total volume cross-sectional area.

[0129] For each location along the bandwidth direction, divide the volume cross-sectional area of ​​that location by the total volume cross-sectional area to obtain the volume percentage coefficient for that location.

[0130] Multiply the deviation by the volume ratio coefficient at that location to obtain an allocated correction amount.

[0131] Determine the sign of the difference between the accumulated total value and the instantaneous total weight value:

[0132] If the difference is positive, it means that the total weight calculated by the current density distribution is greater than the actual total weight. Then, the apportionment correction amount is subtracted from the current true density value at that location.

[0133] If the difference is negative, it means that the total weight calculated by the current density distribution is less than the actual total weight. In this case, the absolute value of the apportionment correction amount is added to the current true density value at that location.

[0134] After all locations along the bandwidth have completed the above addition and subtraction operations according to their respective apportioned correction amounts, the density values ​​of all locations are updated, resulting in a set of updated true density values.

[0135] After the update is completed, the density values ​​at each location need to be corrected by peak smoothing and valley filling.

[0136] The correction process is as follows: for each position in the bandwidth direction, calculate the updated true density value of that position and subtract the corresponding value in the two-dimensional virtual density prior distribution to obtain the difference for that position.

[0137] Take the absolute value of the difference as the degree of deviation between the current density value and the prior value at that location.

[0138] Sort all positions by their degree of deviation from largest to smallest, select the top 20% of positions as positions to be corrected, and the remaining 80% as positions not to be corrected.

[0139] For locations that are determined not to be corrected, their current true density values ​​remain unchanged, and no correction operation is performed.

[0140] For each position identified as a position to be corrected, the difference corresponding to that position is multiplied by a preset pullback ratio coefficient to obtain a pullback amount. The pullback ratio coefficient is 0.3.

[0141] Determine the sign of the difference at that position:

[0142] If the difference is positive, it means that the current density value is greater than the prior value. Then, the pullback amount is subtracted from the current true density value at that location.

[0143] If the difference is negative, it means that the current density value is less than the prior value. Then, the absolute value of the pullback amount is added to the current true density value at that location.

[0144] After the above pullback operation, the true density value at this position is corrected. The corrected value is located at a certain position between the current value and the prior value, that is, the pullback towards the prior value is achieved.

[0145] After all the positions identified as needing correction in the bandwidth direction have undergone the above pullback correction, the density values ​​of all positions constitute a set of corrected true density values.

[0146] After completing the peak-shaving and valley-filling correction, the corrected true density values ​​at each location are used as the new values ​​for the current iteration step. The process then returns to the deviation calculation step, recalculates the deviation between the accumulated total value and the instantaneous total weight value, and compares it with the convergence threshold.

[0147] This process is repeated iteratively. Each iteration performs a complete process of deviation calculation, deviation reversal and update, deviation degree calculation, and peak shaving and valley filling correction, until the deviation is less than the preset convergence threshold. The iteration stops when the deviation is less than the preset convergence threshold. The true density values ​​of each position obtained after the last iteration correction are output as the instantaneous two-dimensional true density distribution of the conveyed material along the bandwidth direction of the cross section.

[0148] In S5, the additional drag torque generated by the lateral eccentric load is calculated based on the pointwise convolution operation of the two-dimensional true density distribution and the volume cross-sectional area. This additional drag torque is then used as a feedforward quantity and superimposed on the torque setpoint of the multi-drive unit. Furthermore, the constraint strength in the sparse inversion is adjusted based on the residual between the prior distribution and the true density distribution. Specifically, this includes:

[0149] After obtaining the instantaneous two-dimensional true density distribution of each pixel position in the bandwidth direction through sparse regularization iterative inversion, the two-dimensional true density distribution is multiplied by the volume cross-sectional area value of each position obtained in S4 at the corresponding position. That is, for each pixel position in the bandwidth direction, the true density value of the position is multiplied by the volume cross-sectional area value of the position to obtain the mass distribution value of the position.

[0150] Multiply the mass distribution value at each location along the belt width direction by the distance of that location relative to the center line of the conveyor belt, and sum the product results of all locations to obtain the unbalanced torque value generated by the lateral eccentric loading.

[0151] Divide the unbalanced torque value by the effective width of the conveyor belt to obtain the equivalent additional resistance torque value caused by the lateral eccentric load.

[0152] This additional resistance torque value reflects the differentiated load disturbances generated on each drive roller when the material is unevenly distributed along the bandwidth. The control unit uses this additional resistance torque value as a feedforward compensation amount and adds it to the torque command signal of the slave drive unit. Specifically, the addition method is to algebraically add the current torque command value of the slave drive unit to this additional resistance torque value. If the off-center load direction is towards the slave drive unit, the additional resistance torque is positive, increasing the torque output of the slave drive unit; if the off-center load direction is away from the slave drive unit, the additional resistance torque is negative, decreasing the torque output of the slave drive unit. After the above addition, the slave drive unit can obtain a torque adjustment command that matches the actual off-center load condition before the weight sensor reports dynamic impact.

[0153] After each iterative inversion of the two-dimensional true density distribution, the final output true density value at each location is compared one by one with the weight coefficient value of the corresponding location in the two-dimensional virtual density prior distribution. For each pixel location in the bandwidth direction, the true density value at that location is calculated by subtracting the weight coefficient value of the corresponding location in the prior distribution, and the difference is recorded as the residual value at that location. A positive residual value indicates that the true density at that location is higher than the prior value, and a negative residual value indicates that the true density at that location is lower than the prior value.

[0154] Traverse all positions in the order from one edge to the other along the bandwidth direction, record the positive and negative signs of the residual values, and form a sign sequence distributed along the bandwidth direction.

[0155] Based on this symbol sequence, adjacent positions with the same symbol are merged into a continuous region, called the same-sign continuous region.

[0156] Each consecutive region with the same sign has a unique sign attribute (positive or negative) and the bandwidth width it occupies, which is obtained by multiplying the number of pixel positions contained in the region by the bandwidth interval width corresponding to each pixel position.

[0157] Count all consecutive regions with the same positive sign, add up the bandwidth occupied by each region to get the total width occupied by the positive residual region; at the same time, take the absolute value of the residual value in each positive residual region and add it up to get the total amplitude of the positive residual region.

[0158] Count all consecutive regions with the same negative sign, add up the bandwidth occupied by each region to get the total width occupied by the negative residual region; then add up the absolute values ​​of the residual values ​​in each negative residual region to get the total amplitude of the negative residual region.

[0159] The sum of the total width occupied by the positive residual region and the total width occupied by the negative residual region equals the total effective bandwidth of the conveyor belt.

[0160] The total number of consecutive regions with the same sign in the statistical bandwidth direction, and the number of times the symbols of adjacent regions in the symbol sequence switch (i.e., the number of times switching from a positive region to a negative region or vice versa), the ratio of this number of switching to (the total number of consecutive regions with the same sign minus 1), is denoted as the symbol alternation frequency.

[0161] The determination and adjustment of constraint strength are performed according to the following logic: First, compare the total amplitude of the positive residual region with the total amplitude of the negative residual region, and record the one with the larger amplitude as the amplitude dominance side, and obtain the total occupied width corresponding to the amplitude dominance side.

[0162] Calculate the total occupied width and divide it by the total effective bandwidth of the conveyor belt to obtain the occupancy ratio of the side with the greater amplitude.

[0163] When the occupancy ratio is greater than the preset overall offset ratio threshold and the symbol alternation frequency is less than 0.3, it is determined that the overall average density of the current incoming material has shifted relative to the prior distribution. In this case, the constraint strength value used for the next cycle of sparse inversion will be reduced to 0.8 times the current constraint strength value to reduce the constraint of the prior distribution on the inversion result and make the inversion result more dependent on the pull of the measured total weight value.

[0164] When the symbol alternation frequency is greater than or equal to 0.7, and the total amplitude of the positive residual region and the total amplitude of the negative residual region are both less than the preset local misalignment amplitude threshold, it is determined that the prior distribution is misaligned in multiple alternating local regions. In this case, the constraint strength value used for the next cycle of sparse inversion is increased to 1.25 times the current constraint strength value to strengthen the constraint of the prior distribution on the inversion results and suppress the over-adjustment noise introduced by iterative inversion in local regions.

[0165] If neither of the above two judgment conditions is met, the current constraint strength value remains unchanged and is directly used for the sparse inversion of the next cycle.

[0166] The overall offset ratio threshold is set to 65%, and the local misalignment amplitude threshold is determined based on the average absolute deviation between the true density value and the prior value at each location during the historical iteration process. Specifically, it is 1.5 times the average absolute deviation, and the initial value is set to 0.1.

[0167] The adjusted constraint strength value will play a role in the sparse regularized iterative inversion of the next cross-section data. Specifically, when calculating the reverse allocation update of the deviation, the volume proportion coefficient of each position is multiplied by the constraint strength value, and then multiplied by the deviation to obtain the allocation correction amount, thereby realizing the control of the constraint strength on the inversion update range.

[0168] The working principle of this invention is as follows: A laser scanning device acquires the three-dimensional spatial coordinates of the material surface and its corresponding echo amplitude, pulse width, and surface texture features, while simultaneously obtaining the instantaneous total weight value of the weighing section. After distance measurement by a rotary encoder, the spatial coordinates and weight values ​​are synchronized in time and matched in bandwidth position. The echo amplitude is normalized and compensated based on the angle between the local normal vector of each coordinate point and the incident light ray to obtain the intrinsic reflectivity. The surface roughness index is obtained by weighting the root mean square of the normal vector deflection angle with the distributed information entropy. The intrinsic reflectivity and surface roughness index are then combined along the bandwidth direction. Two-dimensional feature pairs are formed, and the probability ratios of heavy components in each feature interval are statistically analyzed through a preset sample library to form a two-dimensional virtual density prior distribution. Using this prior distribution as a constraint, combined with the total weight value and volume cross-sectional area, the instantaneous two-dimensional true density distribution is inverted through alternating iterations of deviation back apportionment and peak-shaving and valley-filling correction. The lateral off-center load additional resistance torque is calculated by convolving the true density distribution with the volume cross-sectional area point by point, and is used as a feedforward quantity to be superimposed on the torque given by the drive unit. At the same time, the constraint strength of the next cycle of sparse inversion is adjusted according to the amplitude and bandwidth ratio of the positive and negative regions of the residuals of the true density and the prior distribution.

[0169] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for intelligent speed regulation and multi-drive power balance control of a belt conveyor, characterized in that, Includes the following steps: S1: Collect the three-dimensional spatial coordinates of the surface of the conveyed material, as well as the echo amplitude, pulse width and surface texture features corresponding to each three-dimensional spatial coordinate point. At the same time, obtain the total weight value of the conveyed material, and synchronize the spatial coordinates and weight values ​​in time and match their spatial positions. S2: Calculate the local normal vector of each point based on the spatial coordinates, normalize and compensate the echo amplitude according to the angle between the normal vector and the incident light to obtain the intrinsic reflectivity, and obtain the surface roughness index based on the rate of change statistics of the normal vector. The points mentioned are each three-dimensional spatial coordinate points. S3: The intrinsic reflectivity and surface roughness indices are mapped pixel by pixel along the bandwidth direction to weight coefficients that characterize the probability of heavy components, forming a two-dimensional virtual density prior distribution. S4: Using the two-dimensional virtual density prior distribution as a constraint, combined with the total weight value and the volume cross-sectional area determined by the spatial coordinates, a sparse regularized iterative inversion is performed to recover the instantaneous two-dimensional true density distribution of the conveyed material along the bandwidth direction. S5: Based on the point-by-point convolution operation of the two-dimensional true density distribution and the volume cross-sectional area, calculate the additional resistance torque generated by the lateral off-center load, and use the additional resistance torque as a feedforward quantity to be superimposed on the torque of the multi-drive unit. Adjust the constraint strength in the sparse inversion according to the residual between the prior distribution and the true density distribution.

2. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 1, characterized in that, The process of obtaining intrinsic reflectivity is as follows: The tilt attitude of the micro-region where the corresponding point is located is determined based on the local normal vector of each point. Using the angle between the tilted orientation and the direction of the incident light as the incident angle, a pre-calibrated reference reflectivity response curve of the material surface in a non-tilted state is obtained. Divide the echo amplitude at the corresponding point by the reference reflectivity at the incident angle, and multiply the resulting ratio by the preset diffuse reflection compensation coefficient to obtain the intrinsic reflectivity at the corresponding point.

3. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 2, characterized in that, The process for setting the diffuse reflection compensation coefficient is as follows: Select a local neighborhood window for each point and calculate the average angle between the normal vector of each point within the local neighborhood window and the normal vector of the center point; The average included angle is used to characterize the degree of undulation of the local surface. The ratio of the maximum echo intensity of the material under the standard diffuse reflection reference plate condition to the measured echo intensity under the local undulation is multiplied by the reciprocal of the cosine of the average included angle to obtain the diffuse reflection compensation coefficient of the corresponding point.

4. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 1, characterized in that, The surface roughness index obtained by the measurement specifically includes: Select a local window centered on each point, encompassing its surrounding neighboring points; Calculate the deflection angle of the normal vector of each point within the local window relative to the normal vector of the center point; The root mean square (RMS) values ​​of all deflection angles are statistically analyzed, and the information entropy of the normal vector direction distribution within the local window is calculated. The weighted sum of the RMS values ​​and the information entropy is used as the surface roughness index of the corresponding point.

5. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 1, characterized in that, The process of establishing the two-dimensional virtual density prior distribution is as follows: The intrinsic reflectivity and surface roughness indices are arranged pixel-by-pixel along the bandwidth to form two-dimensional feature pairs; Divide the value space of two-dimensional feature pairs into multiple combination intervals; For each combination interval, the probability ratio of reflectivity and surface roughness to the corresponding heavy component in the preset material sample library is used as the mapping weight of the combination interval. The combination interval of each pixel is determined along the bandwidth direction and the corresponding mapping weight is assigned to the pixel to obtain the probability weight coefficient of the heavy component of each pixel. The probability weight coefficients of all pixels constitute a two-dimensional virtual density prior distribution.

6. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 1, characterized in that, The calculation process for the instantaneous two-dimensional true density distribution is as follows: The initial value of the true density at each location is set to be equal to the corresponding value in the two-dimensional virtual density prior distribution; Calculate the deviation between the sum of the products of the actual density at each current location and the corresponding cross-sectional area and the total weight value; The deviation is back-allocated to each location according to the proportion of the volume cross-sectional area at each location to update the true density value, and the degree of deviation between the updated true density value at each location and the corresponding value in the prior distribution is calculated. Based on the degree of deviation, the updated true density values ​​at each location are corrected by peak shaving and valley filling until the deviation is less than a preset threshold. The true density values ​​at each location at this time are then used as the instantaneous two-dimensional true density distribution.

7. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 6, characterized in that, The peak-shaving and valley-filling correction specifically includes: Calculate the difference between the current true density value at each location and the corresponding value in the two-dimensional virtual density prior distribution; The absolute value of the difference is used as the degree of deviation of the corresponding position; Arrange the deviation of each position in descending order of numerical value, and select the first part of the arrangement as the position to be corrected; For each position to be corrected, its current true density value is pulled back towards the corresponding value in the prior distribution. The pull-back amount is the difference multiplied by a preset scaling factor. The pulled-back value is used as the corrected true density value for the corresponding position. Positions that are not truncated retain their current true density value.

8. The intelligent speed regulation and multi-drive power balance control method for a belt conveyor according to claim 1, characterized in that, The process of adjusting the constraint strength is as follows: Calculate the residuals between the actual density values ​​and the corresponding values ​​in the prior distribution at each location along the bandwidth direction; The residuals are partitioned by positive and negative signs to obtain the total amplitude of the positive residual region, the total amplitude of the negative residual region, and the bandwidth ratio occupied by each region. When the proportion of the larger of the positive and negative residual regions exceeds the preset proportion threshold, it is determined that the properties of the incoming material have shifted overall, and the constraint strength is reduced. When positive and negative residual regions appear alternately and their amplitudes are all lower than the preset amplitude threshold, the prior distribution is determined to be locally inaccurate. At this time, the constraint strength is increased, and the adjusted constraint strength is used to participate in the sparse inversion of the next cycle.

Images