A high-precision classification processing method and system for metal powder

Abstract

The present application discloses a high-precision classification processing method and system for metal powder, which relates to the field of material preparation. The method comprises: obtaining image information of a feed inlet; obtaining feed concentration information, feed speed information and initial particle size distribution information of powder particles according to the image information of the feed inlet; determining the initial rotation speed of a centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, feed speed information and initial particle size distribution information of powder particles; obtaining image information of a centrifugal cavity within a preset time period and continuous pressure change information at different positions of the centrifugal cavity, wherein different positions of the centrifugal cavity are evenly distributed based on the length direction of the cavity, and the straight line formed by different positions of the centrifugal cavity is parallel to the cavity axis of the centrifuge; adjusting the feed concentration information and feed speed information according to the image information of the centrifugal cavity; and adjusting the centrifuge rotation speed information and the opening and closing states of different classification outlets according to the continuous pressure change information.

Description

Technical Field

[0001] The present specification relates to the field of material preparation, and more specifically, the present application relates to a method and system for high-precision grading of metal powders. Background Art

[0002] Powder classification technology is widely used in the fields of nanomaterials, metal powders, electronic materials, etc. Its goal is to efficiently separate particles of different sizes. However, existing technologies mainly improve classification efficiency by optimizing mechanical structures, such as improving centrifuge cavity design, adding multi-layer filter screens, and increasing equipment speed. Although these methods have improved classification efficiency to a certain extent, they still have the following shortcomings when facing dynamic classification requirements and nano-particle processing:

[0003] 1. Insufficient dynamic adaptability: The mechanical structure is usually a fixed design, which is difficult to adjust in real time according to the feed conditions or particle characteristics, making it difficult to ensure the classification accuracy in multiple scenarios.

[0004] 2. Lack of real-time monitoring mechanism: Traditional equipment lacks the means to monitor particle concentration, velocity and particle size distribution, and problems such as blockage and agglomeration during the grading process are difficult to detect and deal with in a timely manner.

[0005] 3. Limited ability to classify tiny particles: Nano-metal powders are small in size and highly agglomerated, so existing methods are difficult to achieve efficient classification.

[0006] Therefore, it is necessary to provide a high-precision classification processing method and system for metal powder to at least solve some of the above problems. Summary of the invention

[0007] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description of the Invention section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0008] In a first aspect, the present application proposes a high-precision classification method for metal powder, comprising:

[0009] Obtain image information of the feeding port;

[0010] Obtaining feed concentration information, feed speed information and initial particle size distribution information of the powder particles according to the image information of the feed inlet;

[0011] Determining the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles;

[0012] Obtaining the centrifugal inner cavity image information of the centrifuge within a preset time period and the continuous pressure change information of different positions of the inner cavity of the centrifuge, wherein the different positions of the inner cavity of the centrifuge are evenly arranged based on the length direction of the cavity, and the straight line formed by the different positions of the inner cavity of the centrifuge is parallel to the inner cavity axis of the centrifuge;

[0013] Adjust the feed concentration information and the feed speed information according to the centrifugal inner cavity image information;

[0014] According to the above continuous pressure change information, the centrifuge speed information and the opening and closing states of different classification outlets are adjusted.

[0015] In a feasible implementation manner, the above-mentioned obtaining the feed concentration information, feed speed information and initial particle size distribution information of the powder particles according to the above-mentioned feed inlet image information includes:

[0016] grayscale processing, denoising processing and contrast enhancement processing are performed on the above-mentioned inlet image information to obtain processed inlet image information;

[0017] Performing a regional segmentation process based on the processed feed inlet image information to obtain a material area and a background area, and determining the feed concentration information according to the pixel ratios of the material area and the background area;

[0018] Determine the feed speed information based on the processed feed inlet image information by a feature point tracking method;

[0019] The particle contour information is extracted based on the processed feed port image information to obtain the particle size characteristics, and the initial particle size distribution information is calculated based on the particle size characteristics.

[0020] In a feasible implementation manner, the above-mentioned region segmentation processing is performed based on the processed inlet image information to obtain the material region and the background region, including:

[0021] Calculate the background weight of the feed inlet image after the above processing :

[0022]

[0023] in, Indicates that the grayscale value is between 0 and The cumulative sum of the pixel ratios within the range, P(i) is the probability that a pixel with gray value i appears in the entire image;

[0024] Calculate the foreground weight of the inlet image after the above processing :

[0025]

[0026] Calculate the background mean of the above feed inlet image :

[0027]

[0028] Calculate the foreground mean of the inlet image after the above processing :

[0029]

[0030] According to the above background weight , the above prospect weights , the background mean and the above outlook average Calculate the between-class variance :

[0031]

[0032] Get the above inter-class variance The target grayscale threshold T* at maximum;

[0033] The feed port image information is subjected to region segmentation processing according to the target grayscale threshold T* to obtain the material region and the background region.

[0034] In a feasible implementation manner, the feed speed information is determined by a feature point tracking method based on the processed feed inlet image information, including:

[0035] Use the Sobel operator to calculate the x-direction gradient in the first frame image and the y-direction gradient ;

[0036] At each pixel point of the first frame image above, a target window is used to calculate the cumulative value M within the window:

[0037]

[0038] Perform eigendecomposition on the above accumulated value M to obtain the first eigenvalue and the second eigenvalue ,

[0039] The first eigenvalue and the second eigenvalue The points corresponding to the values ​​that are all greater than the preset threshold T are determined as target feature points;

[0040] Window matching is performed based on the target feature points, and the position information of the target feature points in the second frame image is determined according to the pixel gradient and intensity difference within the window;

[0041] The feed speed information is determined according to the position information of the target feature point in the first frame image, the position information of the target feature point in the second frame image and the time interval information between the two frames of image.

[0042] In a feasible implementation, the above-mentioned extraction of particle contour information based on the processed feed port image information to obtain particle size characteristics, and the calculation of the above-mentioned initial particle size distribution information according to the above-mentioned particle size characteristics include:

[0043] Performing expansion operations, corrosion operations and opening operations on the processed inlet image information to obtain an image to be extracted;

[0044] Obtaining the contour information of the particles based on the findContours method for the above-mentioned image to be extracted;

[0045] Extracting the particle size characteristics according to the particle contour information, wherein the particle size characteristics include one or more of the particle area, the particle minimum circumscribed circle diameter, the particle minimum circumscribed rectangle size and the shape factor;

[0046] The initial particle size distribution information is calculated based on the particle size characteristics.

[0047] In a feasible implementation manner, the above-mentioned opening and closing state includes an opening and closing degree,

[0048] The above-mentioned determining the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the above-mentioned feed concentration information, the above-mentioned feed speed information and the above-mentioned initial particle size distribution information of the powder particles includes:

[0049] Calculate particle settling velocity information based on the feed concentration information and the feed velocity information;

[0050] Determine the initial speed of the centrifuge according to the sedimentation velocity information, fluid dynamic viscosity, average particle diameter, particle density, fluid density, and centrifuge rotation radius;

[0051] According to the above initial particle size distribution information, the proportion of particles with different particle diameters is counted;

[0052] The opening and closing degrees of different classification outlets are adjusted according to the proportion of particles with different particle diameters.

[0053] In a feasible implementation manner, the adjusting the feed concentration information and the feed speed information according to the centrifugal cavity image information includes:

[0054] Determine particle accumulation information and particle stratification information according to the centrifugal inner cavity image information;

[0055] The feed concentration information and the feed speed information are adjusted according to the particle accumulation information and the particle stratification information.

[0056] In a feasible implementation manner, the above-mentioned continuous pressure change information includes average pressure information and pressure fluctuation range information;

[0057] According to the above continuous pressure change information, the centrifuge speed information and the opening and closing states of different classification outlets are adjusted, including:

[0058] Adjust the centrifuge speed information according to the average pressure information;

[0059] The opening and closing states of the above-mentioned outlets of the same grade are adjusted according to the above-mentioned pressure fluctuation range.

[0060] In a feasible implementation, it also includes:

[0061] When the adjustment change rates of the centrifuge speed information and the opening and closing states of the different graded outlets are both greater than their respective corresponding preset change rates, a simultaneous graded adjustment strategy is adopted to adjust the centrifuge speed information and the opening and closing states of the different graded outlets.

[0062] In a second aspect, the present application proposes a high-precision metal powder classification processing system, comprising:

[0063] A first acquisition unit, used to acquire image information of the feed inlet;

[0064] A second acquisition unit is used to acquire feed concentration information, feed speed information and initial particle size distribution information of the powder particles according to the image information of the feed inlet;

[0065] A determination unit, used to determine the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles;

[0066] A third acquisition unit is used to acquire the centrifugal cavity image information of the centrifuge within a preset time period and the continuous pressure change information of different positions of the centrifuge cavity, wherein the different positions of the centrifuge cavity are evenly distributed based on the length direction of the cavity, and the straight line formed by the different positions of the centrifuge cavity is parallel to the cavity axis of the centrifuge;

[0067] A first adjustment unit, used for adjusting the feed concentration information and the feed speed information according to the centrifugal inner cavity image information;

[0068] The second adjustment unit is used to adjust the centrifuge speed information and the opening and closing states of different classification outlets according to the above-mentioned continuous pressure change information. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present specification. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0070] Figure 1 A schematic diagram of a process for high-precision classification of metal powder provided in an embodiment of the present application;

[0071] Figure 2 A schematic diagram showing a 100% open and closed state of a graded outlet provided in an embodiment of the present application;

[0072] Figure 3 A schematic diagram of a grading outlet provided in an embodiment of the present application with an opening and closing state of 50%;

[0073] Figure 4 A schematic structural diagram of a high-precision metal powder grading processing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0074] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0075] Figure 1 A schematic diagram of a process for high-precision classification of metal powder provided in an embodiment of the present application, the method may specifically include:

[0076] S110, acquiring image information of the feed inlet.

[0077] For example, a high-resolution industrial camera is used to collect the powder feed flow image at the feed inlet. The image resolution must be high enough to distinguish the contours and distribution of tiny particles.

[0078] S120, obtaining feed concentration information, feed speed information and initial particle size distribution information of the powder particles according to the image information of the feed inlet.

[0079] For example, the pixel ratio of the material area and the background area is calculated through grayscale and area segmentation. For nano metal powder, the concentration information reflects the feeding stability of the powder and can determine whether there is blockage or particle agglomeration caused by excessive concentration.

[0080] The motion trajectory and speed of the particles in the image frame are calculated by the feature point tracking method. For nano-metal powder, a feed rate that is too low may cause particle deposition, while a feed rate that is too high may cause airflow turbulence.

[0081] The particle outline is extracted through edge detection, and the particle size distribution curve is generated by analyzing the characteristics of particle diameter, area, shape factor, etc. For nano metal powder, accurate particle size distribution information is crucial and directly determines the subsequent classification accuracy.

[0082] S130, determining the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles.

[0083] Exemplarily, the initial speed is calculated based on the particle size distribution and sedimentation velocity of the powder particles (calculated by concentration and velocity information) combined with the fluid dynamic parameters of the centrifuge (such as viscosity and density). For nano-metal powders, the initial speed of the centrifuge needs to be higher due to the small particle diameter and low sedimentation velocity. According to the particle size distribution, the proportion of particles in each particle size range is counted, and the opening of each classification outlet is adjusted to accurately control the separation of particles of different particle sizes. For nano-metal powders, it is usually necessary to strictly control the separation of small particles (<100nm) and large particles (>200nm).

[0084] It should be noted that if Figure 2 and Figure 3 As shown, each graded outlet is provided with a filter of corresponding size, and the entire opening is provided with a filter. The opening of each graded outlet is adjusted by adjusting the plate structure covering the filter and adjusting the leaked filter area to achieve the purpose of adjusting the opening of each graded outlet.

[0085] S140, obtaining the centrifugal inner cavity image information of the centrifuge within a preset time period and the continuous pressure change information at different positions of the inner cavity of the centrifuge, wherein the different positions of the inner cavity of the centrifuge are evenly distributed based on the length direction of the cavity, and the straight line formed by the different positions of the inner cavity of the centrifuge is parallel to the inner cavity axis of the centrifuge.

[0086] For example, the particle distribution state in the centrifuge cavity is monitored in real time by a camera device, including particle accumulation and stratification. For nano-metal powders, the cavity image information can reveal whether particle agglomeration affects the grading effect. Pressure sensors are installed at different positions in the centrifuge cavity and evenly arranged along the length of the cavity. Continuously monitoring the pressure changes can reflect the uniformity of particle distribution and the stability of the flow field. For nano-metal powders, pressure fluctuations may indicate particle adhesion, blockage or uneven distribution.

[0087] S150, adjusting the feed concentration information and the feed speed information according to the centrifugal inner cavity image information;

[0088] For example, the amount and height of the particles in the inner cavity image can be used to determine whether the feed concentration needs to be adjusted. If there is too much accumulation, the feed concentration may need to be reduced to reduce particle aggregation. Based on the layered height distribution of the particles in the image, it can be determined whether the particles are correctly separated by particle size. If the layering is not obvious, the feed rate may need to be adjusted to optimize the classification effect.

[0089] S160: According to the above continuous pressure change information, adjust the centrifuge speed information and the opening and closing states of different classification outlets.

[0090] According to the average pressure value provided by the pressure sensor, the centrifuge speed is dynamically adjusted. For nano-metal powder, high speed can improve particle separation, but too high speed may cause particle disorder. According to the pressure fluctuation range, adjust the opening of the classification outlet to ensure the stability of the particle flow at each outlet. Accurately control the outlet opening to avoid classification errors caused by outlet fluctuations, and avoid unnecessary wear on other powder particles due to excessive opening, so as to ensure the quality of the classified particles.

[0091] In summary, the method provided by the present application collects the image information of the feed port in real time through a high-resolution industrial camera, and dynamically obtains the feed concentration, feed speed and particle size distribution information of the powder. Accurately reflect the particle agglomeration and flow state of the nano-metal powder, and provide a scientific basis for the subsequent classification process. Dynamically adjust the speed of the centrifuge and the opening of the classification outlet according to the real-time data. For nano-metal powder, the classification requirements of small particles and large particles are met by precise adjustment, and the classification efficiency and applicability are improved. The distribution state of the particles in the inner cavity is monitored in real time by a camera device, including particle accumulation and stratification information. The pressure sensors are evenly distributed in the inner cavity to monitor the pressure changes, and timely detect the adhesion, blockage or uneven distribution of particles to ensure the smooth operation of the classification process. Dynamically adjust the feed concentration according to the inner cavity image information to reduce particle accumulation and agglomeration. Optimize the feed speed according to the particle stratification information, improve the classification accuracy and avoid particle deposition or disorder. Adjust the centrifuge speed according to the average pressure value of the pressure sensor to optimize particle separation. Accurately adjust the outlet opening according to the pressure fluctuation range to ensure stable particle flow and avoid classification errors. Achieve efficient separation of particles of different particle sizes, especially suitable for nano-metal powders with strict particle size distribution. Avoid particle damage caused by an oversized grading outlet to ensure the quality and uniformity of powder particles after grading. Optimize equipment operating status through dynamic adjustment to reduce unnecessary high-load operation. Accurately control the grading outlet to avoid equipment wear and energy waste, and extend equipment service life. Meet the high requirements for particle grading accuracy and uniformity in fields such as nano-metal powders, high-end electronic conductive materials, and aerospace materials. This application overcomes the shortcomings of traditional mechanical grading methods through real-time data drive, dynamic regulation mechanism, and refined outlet control, and improves the accuracy, efficiency, and adaptability of powder grading. It is suitable for the preparation needs of a variety of high-precision powder materials, and provides advanced solutions for intelligent grading in modern industry.

[0092] In a feasible implementation manner, the above step S120 specifically further includes S1201-S1204:

[0093] S1201, performing grayscale processing, denoising processing and contrast enhancement processing on the above-mentioned inlet image information to obtain processed inlet image information;

[0094] S1202, performing region segmentation processing based on the processed feed inlet image information to obtain a material region and a background region, and determining the feed concentration information according to the pixel ratios of the material region and the background region;

[0095] S1203, determining the feed speed information through a feature point tracking method based on the processed feed inlet image information;

[0096] S1204, extracting particle contour information based on the processed feed port image information to obtain particle size characteristics, and calculating the initial particle size distribution information based on the particle size characteristics.

[0097] Exemplarily, the inlet image information is preprocessed, and the preprocessing operation includes grayscale processing, denoising processing and contrast enhancement processing. Grayscale processing converts the color image (RGB) into a grayscale image, retaining only the brightness information and eliminating color interference. The grayscale value formula is: Gray = 0.299 × R + 0.587 × G + 0.11 × B. Grayscale processing can highlight the light and dark differences in the particle shape.

[0098] The material area and the background area in the processed feed image are separated by region segmentation, and the feed concentration information is calculated. Region segmentation can be performed based on the threshold value to separate the material area and the background area, and the number of pixels in the material area and the number of pixels in the background area are counted, and the pixel ratio of the material area and the background area is determined as the feed concentration information, that is, concentration = number of pixels in the material area / total number of pixels. It should be noted that the concentration here is a defined reference concentration, not an actual concentration.

[0099] Obtain particle motion trajectory and velocity information to reflect feed flow status. Use corner detection algorithm in the first frame to extract significant feature points (such as corner points) on the edge of the particle. Track the position of the feature points in the second frame, and calculate the displacement of the feature points based on window matching and pixel intensity difference. Calculate the feed velocity based on the displacement Δx and time interval Δt of the feature points in the first and second frames: v=Δx / Δt

[0100] Extract the geometric characteristics of the particles and calculate their particle size distribution. Perform dilation and erosion operations on the processed image to remove noise and enhance the particle contours. Use the opening operation to eliminate isolated small noise points. Use algorithms such as findContours to extract the boundary contours of the particles. Based on the contour information, calculate the following particle geometric features:

[0101] 1. Area: the number of pixels of the particle.

[0102] 2. Minimum circumscribed circle diameter: indicates the particle size.

[0103] 3. Shape factor: used to evaluate the roundness of particles. Shape factor = 4π × area / circumference 2 .

[0104] According to the particle size data, the particle diameters are grouped and counted to generate a particle size distribution curve.

[0105] In a feasible implementation manner, the above step S1202 specifically includes S12021-S12027:

[0106] S12021, calculate the background weight of the feed port image after the above processing :

[0107]

[0108] in, Indicates that the grayscale value is between 0 and The cumulative sum of the pixel ratios within the range, P(i) is the probability that a pixel with gray value i appears in the entire image;

[0109] S12022, calculating the foreground weight of the feed inlet image after the above processing :

[0110]

[0111] S12023, calculate the background mean of the above-mentioned feed port image :

[0112]

[0113] S12024, calculating the foreground mean of the feed port image after the above processing :

[0114]

[0115] S12025, according to the above background weight , the above prospect weights , the background mean and the above outlook average Calculate the between-class variance :

[0116]

[0117] S12026. Obtain the above inter-class variance The target grayscale threshold T* at maximum;

[0118] S12027. Perform region segmentation processing on the feed port image information according to the target grayscale threshold T* to obtain the material region and the background region.

[0119] Exemplary, background weight Indicates the percentage of pixels in the image that are classified as background under the current threshold T. Foreground weight Indicates the percentage of pixels in the image that are classified as foreground (material area) under the current threshold T. Background mean Indicates the average grayscale value of the background area pixels under the current threshold T. Foreground mean Indicates the average grayscale value of pixels in the material area under the current threshold T. Inter-class variance It is a measure of the difference in grayscale distribution between the background and foreground regions, reflecting the quality of regional segmentation. Maximizing the inter-class variance can achieve the best segmentation effect. Traverse all possible grayscale values ​​T and find the value that makes the inter-class variance The maximum target grayscale threshold T*. According to the target grayscale threshold T*, the image is divided into the material area and the background area. For each pixel in the image, it is classified according to T*:

[0120]

[0121] is the segmented image, 1 represents the material area, and 0 represents the background area.

[0122] The method provided in this embodiment automatically calculates the optimal grayscale threshold through the method of maximizing the inter-class variance, avoiding the uncertainty of manually setting the threshold. The statistical characteristics (weight and mean) of the background and foreground are used to achieve accurate segmentation of the material area and the background area. The segmentation result can be directly used to calculate the pixel ratio of the material area, thereby quantifying the feed concentration. This method is suitable for powder image segmentation with complex backgrounds in different scenarios, and is particularly suitable for nano-metal powders with small particle size and complex distribution.

[0123] In a feasible implementation manner, the above step 1203 specifically includes 12031-12036:

[0124] 12031. Use the Sobel operator to calculate the x-direction gradient in the first frame image. and the y-direction gradient ;

[0125] 12032. At each pixel point of the first frame image, a target window is used to calculate the cumulative value M within the window:

[0126]

[0127] 12033. Perform eigendecomposition on the above cumulative value M to obtain the first eigenvalue and the second eigenvalue ,

[0128] 12034. The first eigenvalue and the second eigenvalue The points corresponding to the values ​​that are all greater than the preset threshold T are determined as target feature points;

[0129] 12035. Perform window matching based on the target feature points, and determine the position information of the target feature points in the second frame image according to the pixel gradient and intensity difference in the window;

[0130] 12036. Determine the feed speed information based on the position information of the target feature point in the first frame image, the position information of the target feature point in the second frame image, and the time interval information between the two frames of images.

[0131] For example, edge features are extracted from an image, and the areas with significant intensity changes in the image are identified by calculating the horizontal and vertical gradients. The Sobel operator is used to calculate the x-direction gradient in the first frame of the image. and the y-direction gradient ;

[0132]

[0133] in, is the original image, is the convolution operation.

[0134] The gradient information in the neighborhood of each pixel is counted to generate a cumulative value matrix M, which is used to describe the feature intensity of the point.

[0135]

[0136] is the sum of squares of the x-direction gradients within the target window, is the sum of squares of the y-direction gradients within the target window, The sum of the products of the x- and y-direction gradients within the target window.

[0137] The window can be set to or , used for feature calculation in local area. Matrix Describes the gradient change characteristics of the current pixel point for subsequent feature point extraction.

[0138] Through matrix decomposition, the eigenvalues ​​of the cumulative value matrix M are extracted to determine the characteristic strength of the pixel points. The cumulative value matrix M is decomposed by eigenvalue to obtain two eigenvalues ​​λ 1 and λ 2 The eigenvalue represents the degree of gradient change in the window area: λ1 describes the characteristic intensity in the x direction, and λ2 describes the characteristic intensity in the y direction. 1 and λ 2 If both are greater than the preset threshold T, the current pixel is considered to be the target feature point. Through eigenvalue decomposition, stable pixels with significant features are screened out.

[0139] Match the same target feature points between two frames and calculate their displacement information. In the second frame, set a search window with the feature point as the center. Find the best matching point based on the grayscale intensity difference of the pixels in the window. Use the optical flow method (such as the Lucas-Kanade algorithm) to calculate the motion trajectory of the feature point:

[0140]

[0141] in, is the displacement of the feature point in the x and y directions. bit time interval.

[0142] According to the position information of the feature points in the two frames of images, combined with the time interval , calculate the movement speed of the particles. The calculation formula is:

[0143]

[0144] in: is the displacement distance of the feature point in the two frames of images, is the time interval between two image frames.

[0145] The method provided in this embodiment uses the Sobel operator and eigenvalue decomposition to accurately extract stable feature points and ensure the reliability of feature point detection. The optical flow method is used to match the feature points, which can adapt to small changes in particle movement. According to the displacement and time interval of the feature points, the particle feed rate is calculated in real time to provide a dynamic adjustment basis for the classification equipment. This method is suitable for different particle shapes and flow states, especially for nanoparticles in complex scenes.

[0146] In a feasible implementation manner, the above step S1204 specifically includes S12041-S12044:

[0147] S12041, performing expansion operation, corrosion operation and opening operation on the processed inlet image information to obtain an image to be extracted;

[0148] S12042, obtaining contour information of particles based on the findContours method for the above-mentioned image to be extracted;

[0149] S12043, extracting the particle size characteristics according to the particle contour information, wherein the particle size characteristics include one or more of particle area, particle minimum circumscribed circle diameter, particle minimum circumscribed rectangle size, and shape factor;

[0150] S12044. Calculate the initial particle size distribution information based on the particle size characteristics.

[0151] Exemplarily, the processed inlet image is subjected to morphological processing to remove noise, enhance the boundary features of the particles, and prepare for subsequent contour extraction. The morphological processing used in this embodiment specifically includes dilation operation, erosion operation and opening operation. The dilation operation uses structural elements to expand the boundaries of particles, fill small holes, enhance the edge continuity of particles, and reduce broken boundaries. The erosion operation uses structural elements to shrink the boundaries of particles, remove small noise points, and eliminate isolated noise pixels in the image. The opening operation performs erosion operation and dilation operation in sequence to eliminate isolated noise while maintaining the overall shape of the particles. The image to be extracted has clear boundaries and less noise, providing high-quality input for subsequent contour extraction.

[0152] Extract the boundary contour of each particle in the image and obtain the shape information of the particle. You can use the findContours method provided by OpenCV to extract the particle contour. The contour point set can describe the boundary of the particle for subsequent size feature calculation. According to the contour information, calculate the geometric characteristics of the particle, including area, diameter, rectangular size and shape factor. According to the size characteristics of the particle, calculate the particle size distribution and generate particle size distribution information.

[0153] According to the size characteristics of the particles, the particle size distribution is statistically analyzed to generate particle size distribution information. Statistics are grouped by preset particle size range:

[0154] Grouping example:

[0155] Count the number of particles in each particle size range and calculate the proportion:

[0156]

[0157] The center value of each particle size interval and the frequency Connect to generate a particle size distribution curve. The number and proportion of particles of each particle size are expressed in a table or curve form, providing a reference for subsequent classification control.

[0158] The method provided in this embodiment eliminates noise and accurately extracts the boundaries of particles through morphological processing and contour detection, ensuring the accuracy of subsequent calculations. Multi-dimensional features such as area, diameter, and shape factor are used to fully reflect the geometric characteristics of particles. Through accurate statistics of particle size distribution, data support is provided for classification equipment to optimize classification effects.

[0159] In a feasible implementation manner, the above-mentioned opening and closing state includes an opening and closing degree,

[0160] The above step S130 specifically includes S1301-S1304:

[0161] S1301, calculating particle settling velocity information according to the feed concentration information and the feed velocity information;

[0162] S1302, determining the initial rotation speed of the centrifuge according to the sedimentation velocity information, fluid dynamic viscosity, average particle diameter, particle density, fluid density, and centrifuge rotation radius;

[0163] S1303, counting the proportions of particles with different particle diameters according to the initial particle size distribution information;

[0164] S1304, adjusting the opening and closing degrees of different classification outlets according to the proportion of particles with different particle diameters.

[0165] Exemplarily, the settling velocity can be calculated based on the following formula:

[0166]

[0167] in, is the particle settling velocity, is the particle diameter, is the particle density, is the fluid density, is the fluid dynamic viscosity, is the gravitational acceleration (usually taken as ).

[0168] Extracting the average particle diameter from feed image information and density The distribution state of particles in the fluid is estimated by feed concentration and velocity. The sedimentation velocity reflects the motion characteristics of particles in the centrifugal field and determines the reasonable range of centrifuge speed.

[0169] Based on the sedimentation velocity, the physical properties of the particles and the centrifuge design parameters, the initial speed of the centrifuge is calculated to ensure the best particle separation effect. Centrifugal acceleration formula:

[0170]

[0171] in, is the centrifugal acceleration, is the angular velocity of the centrifuge and is the rotation radius of the centrifuge.

[0172] Initial speed formula:

[0173]

[0174] Speed By angular velocity Conversion:

[0175]

[0176] Through the initial particle size distribution information, the proportion of particles in each particle size range is counted to provide a basis for the opening and closing adjustment of the classification outlet. Group by preset intervals, for example:

[0177]

[0178] Statistical particle ratio: Calculate the number of particles in each particle size range And the proportion:

[0179]

[0180] in: For the The proportion of particles in each particle size range, For the The number of particles in each interval, is the total number of particles.

[0181] The opening and closing degree of the classification outlet is adjusted according to the particle ratio to ensure that particles of different particle sizes are separated to the corresponding outlets. The method for controlling the outlet opening is to cover the filter screen for each classification outlet and set an opening and closing control panel. Adjust the position of the control panel to change the exposed area of ​​the filter screen, thereby controlling the outlet opening.

[0182] Outlet opening formula: Assuming that the target particle ratio of a certain outlet is , the corresponding opening Calculated according to the following formula:

[0183]

[0184] in, For the The opening of the outlet, The opening adjustment coefficient is determined according to the equipment design and particle flow rate. Real-time monitoring of particle distribution changes and dynamic update of the opening , ensuring stable grading effect.

[0185] The method provided in this embodiment dynamically adjusts the centrifuge speed and outlet opening according to the characteristics of the particles to ensure that the particles are effectively separated by particle size. The adaptability of the equipment to complex particle distribution is improved by calculating the sedimentation velocity combined with the analysis of the particle size ratio. The method provided in this embodiment can dynamically adjust the operating parameters according to real-time data, reduce manual intervention, and improve the degree of automation of the grading process. Optimize the speed and outlet opening, reduce energy waste caused by excessive operation, and extend the service life of the equipment. The method provided in this embodiment is particularly suitable for powder materials with complex particle size distribution (such as nano metal powder), meeting the needs of high-precision grading.

[0186] In a feasible implementation manner, step S150 specifically includes S1501 and S1502:

[0187] S1501, determining particle accumulation information and particle stratification information according to the centrifugal inner cavity image information;

[0188] S1502, adjusting the feed concentration information and the feed speed information according to the particle accumulation information and the particle stratification information.

[0189] For example, the accumulation and layered distribution information of particles are extracted from the centrifugal cavity image information to provide basic data for judging the feeding state. The area where particles accumulate in the centrifugal cavity is determined by image segmentation technology (such as gray threshold segmentation or edge detection).

[0190] The height of the particle accumulation area is calculated by the vertical pixel value of the image. The height formula can be: H=max(y 颗粒 )−min(y 颗粒 ), where y 颗粒 is the vertical pixel position of the particle accumulation area.

[0191] The packing density is estimated based on the grayscale intensity of the particle accumulation area. The higher the grayscale value, the greater the particle density is likely to be.

[0192] Analyze the layer distribution of particles from the centrifuge cavity image, and use a histogram to analyze the distribution of particles at different heights. By calculating the uniformity of particle distribution (such as variance), determine whether there is uneven particle stratification. Determine the number of particle layers, and analyze the particle size range of each layer based on the particle size distribution information.

[0193] By dynamically adjusting the feed concentration and speed, the particle classification effect is optimized to prevent excessive accumulation or abnormal stratification.

[0194] If the particle stacking height Exceeding the preset threshold , indicating that the feed concentration is too high and needs to be reduced:

[0195]

[0196] in, is the adjusted feed concentration, is the current feed concentration, If the stacking height exceeds the value, is the first adjustment factor.

[0197] If the bulk density is too high, particles may agglomerate and the feed concentration needs to be reduced. If the threshold is exceeded, it means that the stratification is uneven and the feed speed needs to be adjusted.

[0198] The feed rate adjustment formula can be:

[0199]

[0200] in, is the adjusted feed rate, is the current feed rate, is the second adjustment factor.

[0201] The method provided in this embodiment adjusts the feed concentration by adjusting the particle accumulation information and particle stratification information determined by the centrifugal lumen image information, which can reduce particle accumulation and agglomeration and maintain the fluidity of the particles. The feed speed is optimized according to the stratification information to ensure that the particles are correctly stratified according to the particle size and improve the grading effect. The parameters can be dynamically adjusted according to the real-time data of the lumen image to adapt to different particle distributions and feeding conditions. By optimizing the feed concentration and speed, the load on the centrifuge caused by excessive accumulation is avoided, and the service life of the equipment is increased.

[0202] In a feasible implementation manner, the above-mentioned continuous pressure change information includes average pressure information and pressure fluctuation range information; step S160 specifically includes S1601 and S1602:

[0203] S1601, adjusting the centrifuge speed information according to the average pressure information;

[0204] S1602. Adjust the opening and closing states of the above-mentioned outlets of the same grade according to the above-mentioned pressure fluctuation range.

[0205] For example, by monitoring the average pressure in the centrifuge cavity, the centrifuge speed is dynamically adjusted to optimize the centrifugal classification efficiency. Real-time pressure data is obtained from multiple pressure sensors in the centrifuge cavity, and the average pressure value is calculated. :

[0206]

[0207] in, is the number of pressure sensors, For the The pressure value of a sensor.

[0208] like (preset upper limit), it means that the particles are too concentrated or the centrifugal force is too large, and the centrifuge speed needs to be reduced. (preset lower limit), indicating that the particle distribution is sparse or the centrifugal force is insufficient, and the centrifuge speed needs to be increased.

[0209] The speed reduction formula is:

[0210]

[0211] in:

[0212]

[0213] is the third adjustment factor.

[0214] The formula for increasing the speed is:

[0215]

[0216] in:

[0217]

[0218] is the fourth adjustment factor.

[0219] Real-time pressure data from the centrifuge chamber Calculate the fluctuation range:

[0220]

[0221] in, is the maximum pressure within a certain period of time, The minimum pressure within a certain period of time.

[0222] like (preset fluctuation threshold), indicating that the particle flow is uneven or the outlet load is unbalanced, and the outlet opening and closing state needs to be adjusted.

[0223] like , indicating that the particle flow is concentrated and the outlet opening needs to be optimized.

[0224] Reduce the outlet opening (large fluctuation range):

[0225]

[0226] in: is the adjusted outlet opening, is the current outlet opening, is the fifth adjustment factor.

[0227] Increase outlet opening (small fluctuation range):

[0228]

[0229] in, is the fifth adjustment factor.

[0230] The method provided in this embodiment adjusts the centrifuge speed based on the average pressure to optimize the centrifugal classification efficiency. The classification outlet opening is adjusted by the pressure fluctuation range to reduce the classification error. The outlet opening is dynamically adjusted to ensure uniform flow at each outlet to avoid particle blockage or uneven flow. The system can respond in real time according to pressure changes, improving the intelligence and adaptability of the classification system. It is particularly suitable for situations where the particle flow state changes greatly (such as nanoparticles) to meet high-precision classification requirements.

[0231] In a feasible implementation manner, step S170 is also included:

[0232] S170: When the adjustment change rates of the centrifuge speed information and the opening and closing states of the different graded outlets are both greater than their respective corresponding preset change rates, a simultaneous graded adjustment strategy is adopted to adjust the centrifuge speed information and the opening and closing states of the different graded outlets.

[0233] For example, when the speed of the centrifuge and the rate of change of the opening and closing state of the classification outlet are large, by adjusting the two synchronously, the negative impact on the classification effect is reduced, and the stability of the system and the accuracy of particle classification are maintained. N And the adjustment rate of change of the opening and closing state of the graded outlet R A Both are greater than the preset change rate threshold:

[0234]

[0235] in:

[0236]

[0237]

[0238] is the speed change rate threshold, is the opening and closing state change rate threshold.

[0239] In simultaneous classification adjustment, the centrifuge speed and outlet opening Adjust synchronously according to proportion.

[0240] Adjustment formula:

[0241]

[0242] in: Adjust the scale factor, is the average pressure change, is the change in pressure fluctuation range.

[0243] Adjustment ratio of centrifuge speed and outlet opening and closing status Dynamically set according to the current operating status of the system to avoid over- or under-adjustment. Continuously monitor the centrifuge cavity pressure information and outlet particle flow information to verify the adjusted classification effect. If the pressure fluctuation range is not reduced or the classification efficiency is not improved, optimize the adjustment parameters In the next cycle, the rotation speed and outlet opening are readjusted using the optimized parameters until the pressure fluctuation range and particle classification accuracy reach the target values.

[0244] The method provided in this embodiment improves the system's adaptability to changes by adjusting the rotation speed and outlet opening at the same time when the operating state fluctuates greatly. The simultaneous graded adjustment avoids the accumulation of graded errors that may be caused by adjusting the rotation speed or outlet opening alone. Through joint adjustment, the centrifugal force is matched with the outlet flow rate to improve the accuracy and efficiency of particle separation. The frequent and large adjustments of the rotation speed and opening are avoided to affect the stability of the system operation and extend the service life of the equipment. The dynamic feedback mechanism enables the system to adjust parameters in real time according to the operating state to meet the needs of different particle distributions and flow characteristics.

[0245] In a second aspect, the present application proposes a high-precision metal powder classification processing system, comprising:

[0246] A first acquisition unit 21 is used to acquire image information of the feed inlet;

[0247] A second acquisition unit 22 is used to acquire feed concentration information, feed speed information and initial particle size distribution information of powder particles according to the image information of the feed inlet;

[0248] A determination unit 23, used to determine the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles;

[0249] The third acquisition unit 24 is used to acquire the centrifugal cavity image information of the centrifuge within a preset time period and the continuous pressure change information of different positions of the centrifuge cavity, wherein the different positions of the centrifuge cavity are evenly distributed based on the length direction of the cavity, and the straight line formed by the different positions of the centrifuge cavity is parallel to the cavity axis of the centrifuge;

[0250] A first adjustment unit 25, used for adjusting the feed concentration information and the feed speed information according to the centrifugal inner cavity image information;

[0251] The second adjustment unit 26 is used to adjust the centrifuge speed information and the opening and closing states of different classification outlets according to the above-mentioned continuous pressure change information.

[0252] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A high-precision classification method for metal powder, characterized in that: include: Obtain image information of the feeding port; Acquiring feed concentration information, feed speed information and initial particle size distribution information of powder particles according to the image information of the feed inlet; Determining the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles; Obtaining the image information of the centrifugal inner cavity of the centrifuge within a preset time period and the continuous pressure change information of different positions of the inner cavity of the centrifuge, wherein the different positions of the inner cavity of the centrifuge are evenly arranged based on the length direction of the cavity, and the straight line formed by the different positions of the inner cavity of the centrifuge is parallel to the inner cavity axis of the centrifuge; Adjust the feed concentration information and the feed speed information according to the centrifugal inner cavity image information; According to the continuous pressure change information, the centrifuge speed information and the opening and closing states of different classification outlets are adjusted.

2. The high-precision classification method for metal powder according to claim 1, characterized in that: The obtaining of feed concentration information, feed speed information and initial particle size distribution information of powder particles according to the feed inlet image information includes: Performing grayscale processing, denoising processing and contrast enhancement processing on the feed inlet image information to obtain processed feed inlet image information; Performing a region segmentation process based on the processed feed port image information to obtain a material region and a background region, and determining the feed concentration information according to the pixel ratios of the material region and the background region; Determining the feed speed information based on the processed feed port image information by a feature point tracking method; The particle contour information is extracted based on the processed feed port image information to obtain the particle size characteristics, and the initial particle size distribution information is calculated according to the particle size characteristics.

3. The high-precision classification method for metal powder according to claim 2, characterized in that: The performing of region segmentation processing based on the processed inlet image information to obtain a material region and a background region includes: Calculate the background weight of the processed feed inlet image : in, Indicates that the grayscale value is between 0 and The cumulative sum of the pixel ratios within the range, P(i) is the probability that a pixel with gray value i appears in the entire image; Calculate the foreground weight of the processed inlet image : Calculate the background mean of the feed port image : Calculate the foreground mean of the processed inlet image : According to the background weight , the foreground weight , the background mean and the outlook mean Calculate the between-class variance : Get the between-class variance The target grayscale threshold T* at maximum; The feed port image information is subjected to region segmentation processing according to the target grayscale threshold T* to obtain the material region and the background region.

4. The high-precision classification method for metal powder according to claim 2, characterized in that: The method of determining the feed speed information based on the processed feed port image information by a feature point tracking method includes: Use the Sobel operator to calculate the x-direction gradient in the first frame image and the y-direction gradient ; At each pixel point of the first frame image, a target window is used to calculate the cumulative value M within the window: Perform eigendecomposition on the accumulated value M to obtain the first eigenvalue and the second eigenvalue , The first eigenvalue and the second eigenvalue The points corresponding to the values ​​that are all greater than the preset threshold T are determined as target feature points; Performing window matching based on the target feature point, and determining the position information of the target feature point in the second frame image according to the pixel gradient and intensity difference in the window; The feeding speed information is determined according to the position information of the target feature point in the first frame image, the position information of the target feature point in the second frame image and the time interval information between the two frames of images.

5. The high-precision classification method for metal powder according to claim 2, characterized in that: The extracting of particle contour information based on the processed feed port image information to obtain particle size characteristics, and calculating the initial particle size distribution information according to the particle size characteristics, comprises: Performing expansion operations, corrosion operations and opening operations on the processed inlet image information to obtain an image to be extracted; Obtaining particle contour information of the image to be extracted based on the findContours method; Extracting the particle size feature according to the contour information of the particle, wherein the particle size feature includes one or more of the particle area, the particle minimum circumscribed circle diameter, the particle minimum circumscribed rectangle size and the shape factor; The initial particle size distribution information is calculated according to the particle size characteristics.

6. The high-precision classification method for metal powder according to claim 1, characterized in that: The opening and closing state includes the opening and closing degree, The determining of the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles comprises: Calculating particle settling velocity information according to the feed concentration information and the feed velocity information; Determining the initial rotation speed of the centrifuge according to the sedimentation velocity information, the fluid dynamic viscosity, the average diameter of the particles, the particle density, the fluid density, and the centrifuge rotation radius; Counting the proportions of particles with different particle diameters according to the initial particle size distribution information; The opening and closing degrees of different classification outlets are adjusted according to the proportion of particles with different particle diameters.

7. The high-precision classification method for metal powder according to claim 1, characterized in that: The adjusting the feed concentration information and the feed speed information according to the centrifugal inner cavity image information includes: Determining particle accumulation information and particle stratification information according to the centrifugal inner cavity image information; The feed concentration information and the feed speed information are adjusted according to the particle accumulation information and the particle stratification information.

8. The high-precision classification method for metal powder according to claim 1, characterized in that: The continuous pressure change information includes average pressure information and pressure fluctuation range information; According to the continuous pressure change information, adjusting the centrifuge speed information and the opening and closing states of different classification outlets includes: adjusting the centrifuge speed information according to the average pressure information; The opening and closing states of the outlets of the same classification are adjusted according to the pressure fluctuation range.

9. The high-precision classification method for metal powder according to claim 1, characterized in that: Also includes: When the adjustment change rates of the centrifuge speed information and the opening and closing states of the different classification outlets are both greater than their respective corresponding preset change rates, a simultaneous classification adjustment strategy is adopted to adjust the centrifuge speed information and the opening and closing states of the different classification outlets.

10. A high-precision metal powder classification processing system, characterized in that: include: A first acquisition unit, used to acquire image information of the feed inlet; A second acquisition unit is used to acquire feed concentration information, feed speed information and initial particle size distribution information of powder particles according to the image information of the feed inlet; A determination unit, used to determine the initial rotation speed of the centrifuge and the opening and closing states of different classification outlets according to the feed concentration information, the feed speed information and the initial particle size distribution information of the powder particles; A third acquisition unit is used to acquire the image information of the centrifugal inner cavity of the centrifuge within a preset time period and the continuous pressure change information of different positions of the inner cavity of the centrifuge, wherein the different positions of the inner cavity of the centrifuge are evenly arranged based on the length direction of the cavity, and the straight line formed by the different positions of the inner cavity of the centrifuge is parallel to the inner cavity axis of the centrifuge; A first adjustment unit, used for adjusting the feed concentration information and the feed speed information according to the centrifugal inner cavity image information; The second adjustment unit is used to adjust the centrifuge speed information and the opening and closing states of different classification outlets according to the continuous pressure change information.

Images