On-line laser detection and classification control system for crushed raw ore particle size distribution

Abstract

The present application belongs to the field of automatic detection technology, and discloses an online laser detection and grading control system for broken raw ore particle size distribution; aiming at solving the problems of halo effect interference caused by fine dust adhesion in complex industrial environment and insufficient detection accuracy of traditional methods. The system obtains dynamic scattering light field data through a laser scanning module, identifies the halo effect area and generates a compensation coefficient using a light field preprocessing module, and realizes accurate reconstruction of particle boundaries in combination with an edge reconstruction module. The particle size distribution analysis module extracts multi-dimensional geometric features based on accurate boundary data and generates a real-time particle size distribution curve. The grading strategy generation module formulates a control strategy according to the deviation characteristics of the curve and the preset standard, and the dynamic control module adjusts the parameters of the crushing equipment in real time to optimize the particle size distribution. The present application effectively overcomes the halo effect interference, improves the detection accuracy and system stability, and breaks through the single limitation of traditional methods, providing comprehensive data support for downstream processes.

Description

Technical Field

[0001] This invention relates to the field of automated detection technology, and more specifically, to an online laser detection and grading control system for the particle size distribution of crushed raw ore. Background Technology

[0002] In the ore processing and beneficiation industry, the particle size distribution of crushed raw ore is a key factor affecting the efficiency of subsequent processes and product quality. Precise particle size distribution detection and classification control can not only optimize the operating parameters of crushing equipment and improve screening efficiency, but also effectively reduce energy consumption, decrease equipment wear, and enhance the quality of the final product. Therefore, developing efficient, real-time online particle size detection and control systems has become a key direction for technological development in the industry.

[0003] One significant problem is the "halo effect" caused by fine dust adhering to the edges of ore particles. During the crushing process, the mechanical crushing and friction of the ore generate a large amount of fine dust, which adheres to the particle surface, forming an irregular scattering medium. When a laser beam illuminates the particle edge, the light intensity is scattered by the dust, resulting in a non-uniform radial diffusion in the edge region, forming a blurred halo area. This halo effect makes it difficult to accurately identify the true boundary of the particle in dynamic scattered light field data. Traditional boundary determination algorithms, such as edge detection methods based on threshold segmentation or simple gradient analysis, often misclassify the halo area as part of the particle boundary, causing the boundary position to deviate from the true value, and thus leading to calculation errors in geometric feature parameters such as particle size and shape. For example, in a crushing environment with high dust concentration, the spatial influence range of the halo effect may reach 10%-30% of the particle diameter, causing small particles to be magnified and identified, while the boundaries of large particles are blurred, seriously affecting the accuracy of the particle size distribution curve. Furthermore, the intensity and diffusion characteristics of the halo effect vary depending on factors such as ore type, particle size range, dust concentration, and laser incident angle. Traditional algorithms lack adaptive adjustment capabilities, leading to instability in detection results.

[0004] Another problem is the limited scope of existing technologies in particle size distribution analysis. Traditional methods primarily focus on particle size distribution, such as calculating particle size distribution histograms using equivalent diameters, but often neglect important geometric features like particle shape factors and boundary complexity. This singular analytical approach cannot comprehensively characterize the properties of the raw ore particle population, especially when specific process requirements need to be met (such as flotation and screening). The lack of comprehensive consideration of particle shape and surface roughness leads to a low degree of alignment between the analytical results and actual process needs. For example, in flotation processes, irregularly shaped particles exhibit different behaviors than regularly shaped particles of the same size, and traditional methods cannot effectively distinguish these differences, limiting the optimization of classification and control strategies.

[0005] Furthermore, in terms of graded control, existing technologies typically employ static or semi-static control strategies, lacking real-time particle size distribution feedback and dynamic adjustment mechanisms. Traditional methods often focus only on the average value or peak position of the particle size distribution, ignoring the uniformity and multimodal characteristics of the distribution, resulting in insufficiently refined control strategies. For example, when the particle size distribution is large and uneven, traditional methods may simply increase the crushing intensity without developing differentiated strategies to optimize the distribution morphology. Simultaneously, in adjusting the operating parameters of crushing equipment, existing technologies largely rely on trial and error or experience-based adjustments, lacking mechanisms for predicting and correcting the adjustment effects. This adjustment method is not only inefficient but may also lead to production fluctuations, increased energy consumption, or accelerated equipment wear. Moreover, the parameter adjustments of crushing and screening equipment often lack synergistic optimization, making it difficult to achieve optimal overall control effects.

[0006] In view of this, the present invention proposes an online laser detection and classification control system for the particle size distribution of crushed raw ore to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an online laser detection and classification control system for the particle size distribution of crushed raw ore, comprising:

[0008] The laser scanning module is used to perform online laser scanning on the conveying flow field of crushed raw ore to obtain dynamic scattered light field data containing the edge features of raw ore particles;

[0009] The light field preprocessing module is used to identify the halo effect region based on the non-uniformity of light intensity distribution in the dynamic scattered light field data, and generate a halo effect compensation coefficient.

[0010] The edge reconstruction module is used to reconstruct the boundary of the halo effect region at the edge of the particles in the dynamic scattered light field data based on the halo effect compensation coefficient and the geometric constraints of laser scattering, so as to obtain accurate particle boundary data.

[0011] The particle size distribution analysis module is used to extract the geometric feature parameters of the raw ore particles based on the precise particle boundary data, and generate a real-time particle size distribution curve.

[0012] The grading strategy generation module is used to determine the grading control strategy of the raw ore particles based on the deviation characteristics between the real-time particle size distribution curve and the preset particle size grading standard.

[0013] The dynamic control module is used to adjust the operating parameters of the crushing equipment in real time based on the graded control strategy to optimize the particle size distribution of the raw ore.

[0014] Furthermore, the method for identifying the halo effect region includes:

[0015] Based on the gradient characteristics of light intensity distribution in the dynamic scattered light field data, regions of abrupt changes in light intensity gradient are extracted as candidate halo effect regions.

[0016] The radius range of light intensity diffusion is calculated based on the radial diffusion characteristics of light intensity within the candidate halo effect region.

[0017] The spatial boundary of the halo effect region is determined based on the radius range and the incident angle of the laser beam.

[0018] The halo effect compensation coefficient is generated based on the non-uniformity of light intensity distribution within the spatial boundary.

[0019] Furthermore, the boundary reconstruction method includes:

[0020] Based on the halo effect compensation coefficient, the light intensity distribution in the halo effect region of the dynamic scattered light field data is adjusted to generate compensated light field data.

[0021] Based on the light intensity gradient characteristics of the particle edges in the compensated light field data, a candidate set of edge points is constructed.

[0022] Based on the geometric constraints of laser scattering, edge points that satisfy the continuity condition in the candidate set are selected to form an initial boundary curve;

[0023] Based on the curvature change trend of the initial boundary curve, abrupt curvature change points are eliminated to obtain the precise particle boundary data.

[0024] Furthermore, the method for extracting the geometric feature parameters includes:

[0025] Based on the precise particle boundary data, construct a polygonal model of the boundary of the raw ore particles;

[0026] Based on the vertex distribution characteristics of the boundary polygon model, the equivalent diameter, shape factor, and boundary complexity of the raw ore particles are calculated, wherein the shape factor is determined by the ratio of the perimeter to the area of ​​the boundary polygon, and the boundary complexity is calculated by the Fourier descriptor of the boundary curve.

[0027] The real-time granularity distribution curve is generated based on the statistical distribution of the equivalent diameter, shape factor, and boundary complexity.

[0028] Furthermore, the method for determining the graded control strategy includes:

[0029] Based on the real-time particle size distribution curve, extract the peak and width features of the particle size distribution;

[0030] The offset direction of the particle size distribution is determined based on the deviation between the peak characteristics and the preset particle size classification standard.

[0031] The uniformity level of particle size distribution is determined based on the matching degree between the width feature and the preset particle size grading standard.

[0032] The graded control strategy is generated based on the offset direction and uniformity level, wherein the graded control strategy includes the speed adjustment range of the crushing equipment and the screen aperture size adjustment step of the screening equipment.

[0033] Furthermore, the method for calculating the radius range of the light intensity diffusion includes:

[0034] Based on the radial distribution of light intensity within the candidate halo effect region, a light intensity attenuation curve is constructed;

[0035] The diffusion boundary of the halo effect is determined based on the inflection point of the light intensity attenuation curve.

[0036] The radius range of the light intensity diffusion is calculated based on the distance between the diffusion boundary and the incident point of the laser beam;

[0037] Based on the dynamic change trend of the radius range, the region with pseudo-halo effect caused by environmental noise is eliminated.

[0038] Furthermore, the method for eliminating curvature abrupt change points includes:

[0039] Based on the local curvature of the initial boundary curve, candidate abrupt change points whose curvature values ​​exceed a preset curvature threshold are extracted;

[0040] Based on the consistency of the light intensity gradient at the boundary points before and after the candidate mutation point, it is determined whether the candidate mutation point is caused by the halo effect residual.

[0041] If the candidate mutation point is caused by the halo effect residual, then the candidate mutation point is smoothed by interpolation.

[0042] If the candidate mutation point is not caused by the halo effect residual, then the candidate mutation point is retained as the true feature point of the particle boundary.

[0043] Furthermore, the method for generating the real-time particle size distribution curve includes:

[0044] Based on the statistical distribution of the equivalent diameter, a histogram of the particle size distribution is constructed;

[0045] Based on the distribution characteristics of the shape factor, the histogram is shaped and corrected.

[0046] Based on the distribution characteristics of the boundary complexity, the histogram is smoothed for complexity.

[0047] The real-time particle size distribution curve is fitted based on the corrected and smoothed histogram.

[0048] Furthermore, the method for determining the uniformity level includes:

[0049] Based on the width characteristics, calculate the width variation coefficient of the particle size distribution;

[0050] The initial uniformity level of the particle size distribution is determined based on the ratio of the width variation coefficient to the preset uniformity threshold.

[0051] The multimodal characteristics of the particle size distribution are determined based on whether there are secondary peaks in the real-time particle size distribution curve.

[0052] If multimodal characteristics exist, the initial uniformity level is adjusted according to the height and width of the secondary peaks to obtain the final uniformity level.

[0053] Furthermore, the method for real-time adjustment of the operating parameters includes:

[0054] Based on the aforementioned graded control strategy, determine the adjustment direction and adjustment range of the crushing equipment speed;

[0055] Based on the adjustment direction, predict the trend of particle size distribution change of the raw ore particles after the rotation speed adjustment;

[0056] Based on the deviation between the changing trend and the preset particle size classification standard, calculate the feedback correction coefficient for rotation speed adjustment;

[0057] Based on the feedback correction coefficient, the rotation speed adjustment range is dynamically adjusted, and the screen aperture size of the screening equipment is updated synchronously to obtain optimized operating parameters.

[0058] The technical effects and advantages of the online laser detection and classification control system for crushed raw ore particle size distribution of this invention are as follows:

[0059] This invention effectively eliminates the interference of the halo effect, significantly improving the reliability of detection results. Even in high-dust-concentration crushing environments, it ensures accurate identification of particle boundaries, thereby significantly reducing particle size distribution deviations caused by boundary misjudgment and guaranteeing product quality stability. Secondly, by solving the halo effect problem, this invention avoids the phenomenon of small particles being magnified or large particle boundaries being blurred, providing more realistic and consistent data support for downstream processes. This effectively improves process efficiency and product performance, especially in scenarios with high particle size accuracy requirements, significantly reducing process fluctuations caused by detection errors. Furthermore, by improving the accuracy of boundary determination, this invention optimizes the precision of the control process, avoiding blind adjustments due to errors, significantly improving production efficiency, while reducing energy waste and equipment wear caused by over- or under-crushing, extending equipment lifespan, and reducing maintenance costs. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the online laser detection and grading control system for the particle size distribution of crushed raw ore according to the present invention. Detailed Implementation

[0061] 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.

[0062] This application provides an online laser detection and classification control system for the particle size distribution of crushed raw ore. The main components executing this system include, but are not limited to, the following: laser scanning device, edge computing unit, particle size analysis platform, classification controller, crushing equipment, etc., which can be regarded as general computing nodes of this application.

[0063] Please see Figure 1 This invention provides an online laser detection and classification control system for the particle size distribution of crushed raw ore, comprising:

[0064] The laser scanning module is used to perform online laser scanning on the conveying flow field of crushed raw ore to obtain dynamic scattered light field data containing the edge features of raw ore particles;

[0065] The light field preprocessing module is used to identify the halo effect region caused by fine dust adhering to the edge of the raw ore particles based on the non-uniform characteristics of light intensity distribution in the dynamic scattered light field data, and to generate a halo effect compensation coefficient.

[0066] The edge reconstruction module is used to reconstruct the boundary of the halo effect region at the edge of particles in dynamic scattered light field data based on the halo effect compensation coefficient and the geometric constraints of laser scattering, so as to obtain accurate particle boundary data.

[0067] The particle size distribution analysis module is used to extract the geometric characteristic parameters of raw ore particles based on accurate particle boundary data and generate real-time particle size distribution curves.

[0068] The grading strategy generation module is used to determine the grading control strategy for raw ore particles based on the deviation characteristics between the real-time particle size distribution curve and the preset particle size grading standard.

[0069] The dynamic control module is used to adjust the operating parameters of the crushing equipment in real time based on a graded control strategy to optimize the particle size distribution of the raw ore.

[0070] This invention achieves precise detection of the particle size of crushed raw ore through laser scanning technology and advanced light field processing algorithms. It identifies and compensates for the halo effect at the edge of the particles, making the boundary reconstruction more accurate. Based on the accurate particle boundary data, geometric features are extracted to generate a high-precision particle size distribution curve. The crushing equipment parameters are dynamically adjusted according to the deviation between the real-time particle size distribution and the target requirements, forming a closed-loop control. Ultimately, it achieves precise control of the particle size distribution of crushed raw ore and optimization of the production process.

[0071] In this embodiment of the invention, the method for identifying halo effect regions includes:

[0072] Based on the gradient characteristics of light intensity distribution in dynamic scattered light field data, regions of abrupt changes in light intensity gradient are extracted as candidate halo effect regions.

[0073] Calculate the radius range of light intensity diffusion based on the radial diffusion characteristics of light intensity within the candidate halo effect region;

[0074] The spatial boundary of the halo effect region is determined based on the radius range and the incident angle of the laser beam.

[0075] Based on the non-uniformity of light intensity distribution within the spatial boundary, a halo effect compensation coefficient is generated, wherein the halo effect compensation coefficient is inversely proportional to the radius range of light intensity diffusion.

[0076] In this embodiment, gradient analysis is first performed on the dynamic scattered light field data acquired by the laser scanning module to calculate the first derivative of the spatial distribution of light intensity, forming a light intensity gradient field. Regions with abnormally high gradient values ​​are filtered by setting a threshold (e.g., twice the local mean), which typically corresponds to the geometric boundaries of particles or halo effect regions. Connectivity analysis and morphological processing are performed on gradient abrupt change regions to merge adjacent small regions and remove excessively small noise regions, resulting in a series of candidate halo effect regions. Radial light intensity analysis is performed on each candidate region, extracting radial distribution profiles of light intensity along different angles with the region's centroid as the center. Gaussian fitting or an exponential decay model is applied to fit the radial distribution to obtain characteristic parameters of light intensity decay with distance, such as the decay coefficient and characteristic length. The radius range of light intensity diffusion is determined by combining the fitting parameters, i.e., the range where the light intensity value drops to the background noise level. The distance at the sound level defines the spatial range of the halo effect. Considering the incident angle between the laser beam and the particle surface, the spatial distribution characteristics of the halo effect are calculated using geometric optics principles, and the directional preference of the radius range is adjusted. Experimental results show that the closer the incident angle is to perpendicular, the more pronounced the halo effect; therefore, an angle correction factor is introduced to enhance the model's accuracy. Based on the defined spatial boundary, the non-uniformity of light intensity distribution within the boundary is analyzed, including statistical indicators such as standard deviation, entropy value, and Gini coefficient. A halo effect compensation coefficient calculation model is established, which is inversely proportional to the radius of light intensity diffusion; that is, the larger the diffusion radius, the smaller the compensation coefficient, indicating a need for stronger boundary correction. An adaptive threshold adjustment method is applied to dynamically optimize the compensation coefficient based on the type of raw ore, particle size range, and scanning conditions, ensuring effective identification and compensation of the halo effect under different working conditions.

[0077] This embodiment of the method takes into account the physical characteristics of laser scanning in a real industrial environment. Through steps such as gradient analysis, radial diffusion characteristic analysis, and incident angle correction, it achieves accurate identification and quantitative characterization of the halo effect at particle edges. It is suitable for processing crushed raw ore scenarios containing a large amount of fine dust. Compared with traditional edge detection methods, it can effectively distinguish between real boundaries and halo interference, providing reliable preprocessing support for subsequent boundary reconstruction.

[0078] In this embodiment of the invention, the boundary reconstruction method includes:

[0079] Based on the halo effect compensation coefficient, the light intensity distribution in the halo effect region of the dynamic scattered light field data is adjusted to generate compensated light field data.

[0080] Based on the light intensity gradient characteristics of the particle edges in the compensated light field data, a candidate set of edge points is constructed.

[0081] Based on the geometric constraints of laser scattering, edge points that satisfy the continuity condition in the candidate set are selected to form the initial boundary curve;

[0082] Based on the curvature change trend of the initial boundary curve, abrupt curvature change points are eliminated to obtain accurate particle boundary data.

[0083] In this embodiment, firstly, based on the halo effect compensation coefficient determined in the previous step, a light intensity distribution adjustment function is designed. This function typically employs an inverse Gaussian transform or radial attenuation compensation. The adjustment function is then applied to the halo effect region in the dynamic scattered light field data to weaken the halo diffusion effect, strengthen the true boundary signal, and generate compensated light field data. The compensation process pays special attention to preserving the texture features and micro-structural information in the original data, avoiding excessive smoothing that leads to detail loss. Multi-scale gradient operators (such as Sobel, Canny, or LoG operators) are applied to the compensated light field data to extract light intensity gradient features. Based on the gradient magnitude and direction information, an adaptive thresholding method is used to select pixels with significant gradients, forming a candidate set of edge points. Density analysis is performed on the edge points in the candidate set to eliminate isolated noise points and overly dense clusters. Geometric constraints on laser scattering are introduced, including scattering angle limitations and the relationship between reflection intensity and incident angle. The system employs physical models such as the scattering characteristics of materials; based on geometric constraints, it calculates the physical rationality score of each candidate edge point, selecting points with high scores as valid edge points; it applies path connection algorithms (such as minimum energy path or B-spline interpolation) to valid edge points, connecting adjacent points to form a continuous initial boundary curve; it calculates the local curvature of each point on the initial boundary curve, constructing a curvature distribution map; it analyzes the curvature change trend, identifying curvature abrupt change points, which usually correspond to unnatural distortions of the boundary or residual effects of halo effects; combining curvature analysis and local light field characteristics, it judges the authenticity of abrupt change points, distinguishing between actual particle shape features and pseudo-features caused by halos; it removes curvature abrupt change points determined as pseudo-features, retains real feature points, and reconstructs boundary segments through smooth interpolation, ultimately forming accurate particle boundary data; the boundary data is stored in the form of closed curves or point sequences, containing spatial coordinates and confidence information, providing a foundation for subsequent geometric feature extraction.

[0084] A physical model of laser scattering was introduced as a constraint, making the boundary recognition process not only dependent on image features but also following the laws of optical scattering, significantly improving the accuracy of boundary recognition in complex industrial environments. The introduction of curvature analysis further optimized boundary details, making the reconstructed particle shape more realistic and providing high-quality basic data for subsequent particle size analysis.

[0085] In this embodiment of the invention, the method for extracting geometric feature parameters includes:

[0086] Based on accurate particle boundary data, construct a polygonal model of the boundary of the raw ore particles;

[0087] Based on the vertex distribution characteristics of the boundary polygon model, the equivalent diameter, shape factor, and boundary complexity of the raw ore particles are calculated. The shape factor is determined by the ratio of the perimeter to the area of ​​the boundary polygon, and the boundary complexity is calculated by the Fourier descriptor of the boundary curve.

[0088] Real-time granularity distribution curves are generated based on the statistical distributions of equivalent diameter, shape factor, and boundary complexity.

[0089] In this embodiment, starting from precise particle boundary data, the Douglas-Peucker algorithm or vertex curvature sampling method is applied to simplify continuous boundary curves into polygonal models with representative vertices, while controlling the simplification error within an acceptable range (usually not exceeding 1% of the original boundary perimeter). For particles with complex shapes, an adaptive sampling strategy can be adopted to retain more vertices in areas with large curvature changes, ensuring that detailed features are not lost. Geometric analysis is performed on the constructed boundary polygonal model to calculate basic geometric attributes such as area, perimeter, centroid position, and orientation. The equivalent diameter is calculated based on the area, i.e., the diameter of a circle with the same area as the particle, which is the main parameter characterizing particle size. The shape factor is calculated, defined as the ratio of the square of the boundary polygon's perimeter to its area (normalized to a comparison with a circle, where a circle is 1). The shape factor reflects the irregularity of the particle; a larger value indicates a more irregular shape. Fourier transform is applied to the boundary curves to extract Fourier descriptors, which represent... Frequency characteristics of boundary shape; the top N low-frequency descriptors and specific high-frequency descriptors are selected to form a feature vector, and the magnitude of the feature vector is calculated as a boundary complexity index; the complexity index reflects the richness of detail and the complexity of shape of the boundary, and is an important feature for distinguishing different crushing mechanisms and ore types; statistical analysis is performed on the equivalent diameter of a large number of particles to construct a particle size distribution histogram, which is usually divided into 10-20 particle size levels; kernel density estimation or parametric fitting methods (such as log-normal distribution, Rosin-Rammler distribution, etc.) are applied to smooth the histogram into a continuous particle size distribution curve; combined with shape factor distribution information, the particle size distribution curve is shape corrected to consider the actual process influence of non-spherical particles; boundary complexity distribution is introduced to further refine the particle size distribution characteristics and distinguish between particle groups with smooth and rough surfaces; the final real-time particle size distribution curve not only contains size distribution information, but also integrates shape and surface characteristics, providing a comprehensive feature characterization for graded control.

[0090] This method extracts multidimensional geometric features of raw ore particles using a boundary polygon model. It not only focuses on traditional particle size distribution but also introduces new features such as shape factor and boundary complexity, comprehensively characterizing the geometric properties of the particle group. In particular, Fourier descriptors are introduced to analyze boundary complexity. The real-time particle size distribution curves generated based on these multidimensional features contain richer information than traditional size-based distribution curves, providing a more accurate basis for subsequent hierarchical control strategies.

[0091] In this embodiment of the invention, the method for determining the hierarchical control strategy includes:

[0092] Based on the real-time particle size distribution curve, extract the peak and width features of the particle size distribution;

[0093] The offset direction of particle size distribution is determined based on the deviation between the peak characteristics and the preset particle size classification standard.

[0094] The uniformity level of particle size distribution is determined based on the degree of matching between the width characteristics and the preset particle size grading standard.

[0095] Based on the offset direction and uniformity level, a graded control strategy is generated, which includes the speed adjustment range of the crushing equipment and the screen size adjustment step of the screening equipment.

[0096] In this embodiment, feature extraction is performed on the real-time particle size distribution curve to identify peak features such as the main peak position, peak height, and half-peak width. These features reflect the main size concentration range of the particle population. Simultaneously, width features such as the standard deviation, skewness coefficient, and kurtosis coefficient of the distribution curve are extracted to characterize the dispersion and morphological characteristics of the particle size distribution. The curve is compared with a preset particle size classification standard, which is typically based on downstream process requirements or product specifications and includes target particle size range and distribution morphology requirements. The deviation between the peak features and the preset standard is calculated, particularly the difference between the main peak position and the target particle size, to determine whether the particle size distribution is generally too large or too small, i.e., the direction of deviation. A deviation degree assessment is introduced, quantifying the deviation as a percentage or standard score as a reference for control intensity. The matching degree between the width features and the preset standard is analyzed, and matching indices such as root mean square error (RMSE) or distribution overlap are calculated. Based on the matching indices, uniformity, etc., are then considered. The particle size distribution is categorized into "highly uniform," "moderately uniform," "non-uniform," and "highly non-uniform." Corresponding grading control strategies are developed for different combinations of offset directions and uniformity levels. For example, for a "large and non-uniform" particle size distribution, it may be necessary to increase the crushing force and strengthen the screening process; while for a "small but highly uniform" particle size distribution, only minor adjustments to the crushing parameters may be needed. Specific strategies include adjusting the crushing equipment's rotational speed (e.g., increasing / decreasing by 5%, 10%, 15%, etc.), generally resulting in finer particles after crushing. Simultaneously, the adjustment step size of the screening equipment's screen aperture (e.g., ±2mm, ±5mm, etc.) is determined to control the upper limit of the finished product's particle size and adjust the product grading accuracy. When developing strategies, the physical limitations and response characteristics of the equipment adjustments are considered to ensure the executability of the control commands. Production efficiency and energy consumption indicators are also taken into account, minimizing energy consumption and equipment wear while meeting particle size requirements.

[0097] This method systematically determines a hierarchical control strategy by analyzing the differences between the characteristics of real-time particle size distribution and target requirements. In particular, it introduces two key dimensions: offset direction and uniformity level, enabling the control strategy to accurately address different types of distribution deviations. Compared with traditional methods, this two-dimensional strategy generation method not only focuses on adjusting the average particle size but also emphasizes optimizing the distribution morphology, thus more comprehensively meeting the downstream processes' requirements for refined particle size distribution.

[0098] In this embodiment of the invention, the method for calculating the radius range of light intensity diffusion includes:

[0099] Based on the radial distribution of light intensity within the candidate halo effect region, a light intensity attenuation curve is constructed.

[0100] The diffusion boundary of the halo effect is determined based on the inflection point of the light intensity attenuation curve.

[0101] The radius of light intensity diffusion is calculated based on the distance between the diffusion boundary and the incident point of the laser beam.

[0102] Based on the dynamic changing trend of the radius range, the pseudo-halo effect area caused by environmental noise is eliminated.

[0103] In this embodiment, for each candidate halo effect region, its center point (usually the point of highest light intensity) is determined as the starting point for radial analysis. Using the center point as the origin, light intensity values ​​are sampled along multiple radial directions (e.g., 8 or 16 evenly distributed directions) to obtain raw data on light intensity variation with distance. Smoothing filters (e.g., Savitzky-Golay filters) are applied to the sampled data in each direction to reduce noise and improve the stability of subsequent analysis. A light intensity attenuation curve is constructed based on the filtered data; this curve typically exhibits exponential or power-law attenuation. Curve fitting methods (e.g., nonlinear least squares method) are applied to fit the measured data into a mathematical model, such as... ,in Let r0 be the central light intensity, r0 be the characteristic attenuation radius, and C be the background constant. Calculate the first and second derivatives of the light intensity attenuation curve to identify the inflection point, i.e., the location where the attenuation rate changes most significantly. The inflection point usually corresponds to the boundary region where the halo effect transitions to the background and is a key feature point for determining the diffusion boundary. In complex cases (such as multiple inflection points), combine the physical model and thresholding method to determine the most reasonable diffusion boundary, such as when the light intensity drops to 10% of the central value. Calculate the distance from the diffusion boundary to the incident point in each radial direction to obtain the directional radius value. Combine the radius values ​​in all directions to calculate the mean, standard deviation, and directional preference to form a complete description of the radius range. Analyze the changing trend of the radius range in continuous multi-frame data to construct a time series model. Identify abnormally fluctuating radius values, which are usually caused by environmental noise (such as dust clouds or changes in illumination). Apply statistical testing methods (such as the 3σ criterion or DBSCAN clustering) to identify and remove pseudo-halo effect regions, whose radius changes do not match the expectations of the physical model. Retain stable, true halo effect regions for subsequent processing to improve the system's anti-interference capability.

[0104] This method achieves accurate calculation of the halo effect radius range through refined radial light intensity analysis and dynamic trend monitoring. In particular, the introduction of inflection point analysis and time series models enables the system to distinguish between genuine particle edge halos and pseudo-halos caused by environmental noise, significantly improving detection reliability in complex industrial environments. This method is highly adaptable, capable of handling halo effects under different types of ores and varying dust concentrations, providing accurate reference information for subsequent boundary reconstruction.

[0105] In this embodiment of the invention, the method for eliminating curvature abrupt change points includes:

[0106] Based on the local curvature of the initial boundary curve, candidate abrupt change points whose curvature values ​​exceed a preset curvature threshold are extracted;

[0107] Based on the consistency of light intensity gradient at the boundary points before and after the candidate mutation point, determine whether the candidate mutation point is caused by the halo effect residual;

[0108] If the candidate mutation point is caused by the halo effect residual, then the candidate mutation point is smoothed by interpolation.

[0109] If the candidate mutation point is not caused by the halo effect residual, then the candidate mutation point is retained as the true feature point of the particle boundary.

[0110] In this embodiment, firstly, a curvature calculation algorithm, such as the three-point method or B-spline fitting followed by second derivative calculation, is applied to the initial boundary curve to calculate the local curvature at each point on the curve; a curvature distribution histogram is established to determine the range and statistical characteristics of normal curvature; a curvature threshold is set, typically 3-5 times the average curvature or an empirical value based on background knowledge, which needs to be adjusted according to the type of raw ore and its crushing characteristics; points with absolute curvature values ​​exceeding the threshold are extracted as candidate abrupt change points, which typically correspond to sharp corners, depressions, or abnormal fluctuations on the boundary; the light intensity gradient characteristics of the boundary segments before and after each candidate abrupt change point are analyzed, including gradient magnitude, direction, and consistency; the gradient correlation or coherence index in the region near the abrupt change point is calculated to quantify the consistency of the optical properties of the local boundary; highly consistent gradients usually represent continuous true boundaries, while inconsistent gradients may be false boundaries caused by halo effect residuals; simultaneously, the corresponding light field data in the original light field before compensation is analyzed. The system identifies potential halo interference based on location characteristics. By combining gradient consistency and original light field characteristics, the causes of candidate abrupt change points are determined, categorizing them as either "halo effect residuals" or "true feature points." For abrupt change points identified as halo effect residuals, appropriate interpolation methods, such as cubic spline interpolation or local weighted regression, are applied for smoothing. The continuity and smoothness of the preceding and following boundaries are considered during interpolation to ensure a natural transition of the corrected boundaries. Abrupt change points identified as true feature points are retained, as these points may represent the true morphological characteristics of particles, such as fracture surfaces, crystal structures, or mineral junctions. To ensure processing quality, a confidence score mechanism can be introduced to quantitatively assess the reliability of the judgment results. Low-confidence areas are subject to manual review or the application of more complex analytical algorithms. Finally, precise particle boundary data after abrupt change point processing is generated, which retains the true morphological characteristics of the particles while eliminating the residual effects of halo effects.

[0111] This method distinguishes between halo effect residuals and true particle characteristics through light intensity gradient consistency analysis, avoiding over-smoothing and feature loss that may occur with the simple curvature thresholding method. While preserving the true morphological characteristics of the original ore particles, this method effectively eliminates the interference of halo effects, making the reconstructed boundaries more accurate and reliable. Especially when dealing with particles of complex shapes, this differentiated processing strategy can retain important morphological information, providing a high-quality data foundation for subsequent particle size and shape analysis.

[0112] In this embodiment of the invention, the method for generating a real-time particle size distribution curve includes:

[0113] Based on the statistical distribution of the equivalent diameter, a histogram of the particle size distribution is constructed.

[0114] Based on the distribution characteristics of the shape factor, the histogram is shaped and corrected, where particles with higher shape factors have higher weights in the histogram.

[0115] Based on the distribution characteristics of boundary complexity, the histogram is smoothed for complexity, where the particles with higher boundary complexity have larger smoothing windows in the histogram.

[0116] Based on the corrected and smoothed histogram, fit the real-time particle size distribution curve.

[0117] In this embodiment, a large amount of equivalent diameter data of particles is collected, typically requiring 100-1000 particle samples to obtain a statistically meaningful distribution. The equivalent diameter range is divided into an appropriate number of intervals (usually 10-20), and the number of particles in each interval is counted to construct an initial particle size distribution histogram. Traditional particle size statistical indicators, such as median diameter (D50) and cumulative distribution curve, are calculated as basic reference data. The shape factor distribution of all particles is analyzed, and the average shape factor and standard deviation of different particle size intervals are calculated. Based on process knowledge, a relationship model between shape factor and actual process influence is established. For example, in flotation processes, irregularly shaped particles often exhibit different behaviors than regularly shaped particles of the same size. A shape weight function is designed so that particles with higher shape factors (i.e., more irregular shapes) have higher weights in the histogram. A typical weight function is W(SF) = 1 + α × (SF-1), where α is a process-related adjustment coefficient and SF is the shape factor. The shape weight function is applied to correct the initial histogram, making the histogram more uniform and accurate. The histogram better reflects the behavioral characteristics of particle groups in actual processes; it analyzes the distribution characteristics of boundary complexity, calculates the proportion and distribution pattern of different complexity levels; it designs a complexity-related smoothing window function so that the smoothing window corresponding to particles with higher boundary complexity is larger, reflecting the diversity of complex surface particles in the process; it applies variable window smoothing to the shape-corrected histogram to reduce sharp fluctuations in high-complexity regions and enhance the stability and representativeness of the distribution curve; it uses an appropriate fitting model, such as a log-normal distribution, Rosin-Rammler distribution, or a multi-peak Gaussian mixture model, to perform curve fitting on the smoothed histogram; the fitting process uses the least squares method or maximum likelihood estimation to optimize model parameters to best match the actual data; it evaluates the fitting quality by calculating indices such as R² and residual sum of squares to ensure that the fitted curve accurately reflects the actual distribution; the final generated real-time particle size distribution curve contains both basic size distribution information and the influence of shape and surface complexity, which is an enhanced distribution representation that comprehensively reflects particle characteristics.

[0118] This innovative method incorporates two key features—shape factor and boundary complexity—into the generation process of particle size distribution curves, overcoming the limitations of traditional methods that only focus on size distribution. Shape correction ensures that the process influence of irregular particles is reasonably expressed, while complexity smoothing reflects the moderating effect of surface properties on distribution stability. This multi-feature fusion distribution curve contains richer information than traditional curves, enabling more comprehensive guidance for the optimization and control of crushing and classification processes.

[0119] In this embodiment of the invention, the method for determining the uniformity level includes:

[0120] Calculate the width variation coefficient of the particle size distribution based on the width characteristics;

[0121] The initial uniformity level of particle size distribution is determined based on the ratio of the width variation coefficient to the preset uniformity threshold.

[0122] The multimodal characteristics of particle size distribution can be determined by whether there are secondary peaks in the real-time particle size distribution curve.

[0123] If multimodal characteristics exist, the initial uniformity level is adjusted according to the height and width of the secondary peaks to obtain the final uniformity level.

[0124] In this embodiment, based on the width characteristics of the particle size distribution curve, such as standard deviation, half-peak width, and 90% span (D90-D10), quantitative indicators reflecting the degree of dispersion are calculated. These indicators are standardized, and the width variation coefficient is calculated. This coefficient is usually defined as the ratio of the standard deviation to the mean, reflecting the relative dispersion. Preset uniformity thresholds are set. These thresholds are determined based on industry standards, process requirements, or historical experience, and are usually divided into multiple levels corresponding to different uniformity requirements. The ratio of the width variation coefficient to the preset threshold is calculated, and the initial uniformity level is determined according to the ratio range, such as "highly uniform" (ratio < 0.8), "moderately uniform" (0.8 ≤ ratio < 1.2), "non-uniform" (1.2 ≤ ratio < 1.5), and "highly non-uniform" (ratio ≥ 1.5). Peak analysis is performed on the real-time particle size distribution curve, and peak detection algorithms (such as those based on second derivative or wavelet transform) are used to identify the main peak and possible secondary peaks. Peak significance indicators are defined, such as peak height ratio (secondary peak height / main peak height) or valley depth. Used to determine the significance of multimodal characteristics; when the peak significance exceeds the threshold, multimodal characteristics are determined to exist, which usually means that the particle population contains multiple size concentration areas, possibly the result of insufficient crushing or uneven raw materials; analyze the characteristics of secondary peaks, including relative height (ratio of height to the main peak), relative width (ratio of width to the main peak), and distance (distance from the main peak); establish a quantitative scoring model for multimodal characteristics, such as Score=w1×Height+w2×Width+w3×Distance, where w1, w2, and w3 are weighting coefficients; adjust the initial uniformity level according to the multimodal score, usually the significant multimodal characteristics will lead to a decrease in the uniformity level, such as adjusting from "moderately uniform" to "non-uniform"; in special cases, such as when the secondary peak is very small and close to the main peak, only slight adjustments may be made or the original level may be maintained; the finally determined uniformity level serves as an important basis for the generation of graded control strategies, directly affecting the intensity and direction of control.

[0125] This method comprehensively evaluates the uniformity level of particle size distribution by combining the width variation coefficient and multimodal characteristics. In particular, the introduction of multimodal analysis enables the system to identify and handle complex distribution patterns, rather than being limited to simple unimodal distribution evaluation. This method is particularly effective in detecting anomalies in the crushing process (such as partially incomplete crushing or the separation of ores of different hardness), providing important decision-making basis for precise control.

[0126] In this embodiment of the invention, the method for real-time adjustment of operating parameters includes:

[0127] Based on the graded control strategy, determine the direction and magnitude of the speed adjustment for the crushing equipment;

[0128] Based on the adjustment direction, predict the trend of particle size distribution change of the raw ore particles after the rotation speed adjustment;

[0129] Based on the deviation between the changing trend and the preset particle size classification standard, calculate the feedback correction coefficient for speed adjustment;

[0130] Based on the feedback correction coefficient, the rotation speed adjustment range is dynamically adjusted, and the screen aperture size of the screening equipment is updated synchronously to obtain optimized operating parameters.

[0131] In this embodiment, firstly, the adjustment direction (increase or decrease) and initial adjustment range of the crushing equipment speed are determined according to the graded control strategy, such as "increase speed by 10%" or "decrease speed by 15%". An empirical or mathematical model is established between the crushing equipment speed and particle size distribution. Generally, increasing the speed leads to finer product particle size, but the specific relationship depends on factors such as equipment type and ore properties. A machine learning model (such as a random forest or neural network) trained on historical data is used to predict the particle size distribution change trend after a given speed adjustment. The prediction results include multiple aspects such as the shift of the main peak position, changes in distribution width, and the impact on output. The predicted change trend is compared with the preset particle size classification standard to calculate potential deviations after adjustment, such as excessively fine particles or increased inhomogeneity. Based on deviation analysis, a feedback correction coefficient for speed adjustment is calculated. This coefficient is used to optimize the initial adjustment range and avoid over- or under-adjustment. The calculation of the feedback correction coefficient considers multiple factors, including the magnitude of the prediction deviation, the urgency of the adjustment, and the equipment response characteristics. The formula is as follows: The formula `Adjusted = Initial × (1 + Feedback)` calculates the final speed adjustment range, where `Feedback` is the feedback correction coefficient, `Adjusted` refers to the final speed adjustment range after calculation using the feedback correction coefficient, and `Initial` refers to the initial speed adjustment range determined according to the graded control strategy, such as 10% in "increase speed by 10%". Simultaneously, based on the corrected speed adjustment, the optimal screen aperture size for the screening equipment is calculated to ensure coordinated optimization of the crushing and screening processes. The screen aperture size adjustment follows certain rules, such as appropriately reducing the screen aperture size when the crushing speed increases to maintain product specification consistency. Complete operating parameter adjustment instructions are generated, including the precise speed setting value for the crushing equipment, the screen aperture size adjustment value for the screening equipment, the adjustment time schedule, and a description of the expected effects. A smooth transition mechanism is designed to avoid the impact of sudden parameter changes on equipment and production, such as a step-by-step adjustment or gradual adjustment strategy. After the adjustment is implemented, the system continues to monitor changes in particle size distribution, forming a closed-loop control and continuously optimizing operating parameters.

[0132] This method achieves precise adjustment of crushing equipment operating parameters through a predictive model and a feedback correction mechanism. The predictive model provides a forward-looking assessment of the adjustment effect, while the feedback correction coefficient ensures the rationality of the adjustment range, avoiding production fluctuations and resource waste that may result from traditional trial-and-error methods. In particular, the synergistic optimization of crushing and screening equipment parameters forms an integrated hierarchical control strategy, significantly improving the system's accuracy and stability in controlling particle size distribution.

[0133] Through the detailed implementation methods described above, high-precision online detection and intelligent control of raw ore particle size distribution are achieved. The system overcomes the halo effect interference caused by fine dust adhering to particle edges in traditional methods, acquiring accurate particle boundary data through innovative light field processing and boundary reconstruction algorithms. Based on multidimensional geometric feature analysis, the system generates an enhanced particle size distribution curve containing information on size, shape, and complexity, providing comprehensive data support for graded control. The dynamic control module, through predictive models and feedback mechanisms, achieves coordinated optimization of crushing and screening equipment parameters, forming a high-precision closed-loop control. Compared to traditional methods, this system exhibits higher detection accuracy and control stability in complex industrial environments.

[0134] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0135] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0136] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An online laser detection and grading control system for the particle size distribution of crushed raw ore, characterized in that, include: The laser scanning module is used to perform online laser scanning on the conveying flow field of crushed raw ore to obtain dynamic scattered light field data containing the edge features of raw ore particles; The light field preprocessing module is used to identify the halo effect region based on the non-uniformity of light intensity distribution in the dynamic scattered light field data, and generate a halo effect compensation coefficient. The step of identifying halo effect regions and generating halo effect compensation coefficients based on the non-uniformity of light intensity distribution in the dynamic scattered light field data includes: Based on the gradient characteristics of light intensity distribution in the dynamic scattered light field data, regions of abrupt changes in light intensity gradient are extracted as candidate halo effect regions. The radius range of light intensity diffusion is calculated based on the radial diffusion characteristics of light intensity within the candidate halo effect region. The spatial boundary of the halo effect region is determined based on the radius range and the incident angle of the laser beam. The halo effect compensation coefficient is generated based on the non-uniformity of light intensity distribution within the spatial boundary. The edge reconstruction module is used to reconstruct the boundary of the halo effect region at the edge of the particles in the dynamic scattered light field data based on the halo effect compensation coefficient and the geometric constraints of laser scattering, so as to obtain accurate particle boundary data. The particle size distribution analysis module is used to extract the geometric feature parameters of the raw ore particles based on the precise particle boundary data, and generate a real-time particle size distribution curve. The step of extracting geometric feature parameters of the raw ore particles based on the precise particle boundary data and generating a real-time particle size distribution curve includes: Based on the precise particle boundary data, construct a polygonal model of the boundary of the raw ore particles; Based on the vertex distribution characteristics of the boundary polygon model, the equivalent diameter, shape factor, and boundary complexity of the raw ore particles are calculated, wherein the shape factor is determined by the ratio of the perimeter to the area of ​​the boundary polygon, and the boundary complexity is calculated by the Fourier descriptor of the boundary curve. The real-time granularity distribution curve is generated based on the statistical distribution of the equivalent diameter, shape factor, and boundary complexity. The grading strategy generation module is used to determine the grading control strategy of the raw ore particles based on the deviation characteristics between the real-time particle size distribution curve and the preset particle size grading standard. The dynamic control module is used to adjust the operating parameters of the crushing equipment in real time based on the graded control strategy to optimize the particle size distribution of the raw ore.

2. The system according to claim 1, characterized in that, The boundary reconstruction method includes: Based on the halo effect compensation coefficient, the light intensity distribution in the halo effect region of the dynamic scattered light field data is adjusted to generate compensated light field data. Based on the light intensity gradient characteristics of the particle edges in the compensated light field data, a candidate set of edge points is constructed. Based on the geometric constraints of laser scattering, edge points that satisfy the continuity condition in the candidate set are selected to form an initial boundary curve; Based on the curvature change trend of the initial boundary curve, abrupt curvature change points are eliminated to obtain the precise particle boundary data.

3. The system according to claim 1, characterized in that, The method for determining the hierarchical control strategy includes: Based on the real-time particle size distribution curve, extract the peak and width features of the particle size distribution; The offset direction of the particle size distribution is determined based on the deviation between the peak characteristics and the preset particle size classification standard. The uniformity level of particle size distribution is determined based on the matching degree between the width feature and the preset particle size grading standard. The graded control strategy is generated based on the offset direction and uniformity level, wherein the graded control strategy includes the speed adjustment range of the crushing equipment and the screen aperture size adjustment step of the screening equipment.

4. The system according to claim 1, characterized in that, The method for calculating the radius range of light intensity diffusion includes: Based on the radial distribution of light intensity within the candidate halo effect region, a light intensity attenuation curve is constructed; The diffusion boundary of the halo effect is determined based on the inflection point of the light intensity attenuation curve. The radius range of the light intensity diffusion is calculated based on the distance between the diffusion boundary and the incident point of the laser beam; Based on the dynamic change trend of the radius range, the region with pseudo-halo effect caused by environmental noise is eliminated.

5. The system according to claim 2, characterized in that, The method for eliminating curvature abrupt change points includes: Based on the local curvature of the initial boundary curve, candidate abrupt change points whose curvature values ​​exceed a preset curvature threshold are extracted; Based on the consistency of the light intensity gradient at the boundary points before and after the candidate mutation point, it is determined whether the candidate mutation point is caused by the halo effect residual. If the candidate mutation point is caused by the halo effect residual, then the candidate mutation point is smoothed by interpolation. If the candidate mutation point is not caused by the halo effect residual, then the candidate mutation point is retained as the true feature point of the particle boundary.

6. The system according to claim 1, characterized in that, The step of generating the real-time granularity distribution curve based on the statistical distribution of the equivalent diameter, shape factor, and boundary complexity includes: Based on the statistical distribution of the equivalent diameter, a histogram of the particle size distribution is constructed; Based on the distribution characteristics of the shape factor, the histogram is shaped and corrected. Based on the distribution characteristics of the boundary complexity, the histogram is smoothed for complexity. The real-time particle size distribution curve is fitted based on the corrected and smoothed histogram.

7. The system according to claim 3, characterized in that, The method for determining the uniformity level includes: Based on the width characteristics, calculate the width variation coefficient of the particle size distribution; The initial uniformity level of the particle size distribution is determined based on the ratio of the width variation coefficient to the preset uniformity threshold. The multimodal characteristics of the particle size distribution are determined based on whether there are secondary peaks in the real-time particle size distribution curve. If multimodal characteristics exist, the initial uniformity level is adjusted according to the height and width of the secondary peaks to obtain the final uniformity level.

8. The system according to claim 1, characterized in that, The real-time adjustment method for the operating parameters includes: Based on the aforementioned graded control strategy, determine the adjustment direction and adjustment range of the crushing equipment speed; Based on the adjustment direction, predict the trend of particle size distribution change of the raw ore particles after the rotation speed adjustment; Based on the deviation between the changing trend and the preset particle size classification standard, calculate the feedback correction coefficient for rotation speed adjustment; Based on the feedback correction coefficient, the rotation speed adjustment range is dynamically adjusted, and the screen aperture size of the screening equipment is updated synchronously to obtain optimized operating parameters.

Images