Intelligent recognition method for mineral fragment grinding energy discrimination and dissociation degree based on particle morphology characteristics

Abstract

A mineral broken grinding energy discrimination and dissociation degree intelligent identification method based on particle morphology characteristics, first, the magnetite is crushed by a falling weight impact testing machine, and the crushed particles are classified by size, and the mineral dissociation instrument is used to detect and record the dissociation degree of the crushed particles with different broken energy and relatively small particle size, and the dissociation degree distribution diagram of different minerals is established according to the detected data, and the suitable broken energy E is determined after analyzing the distribution diagram, then the particles of each size level under the broken energy are scanned by electron microscope, the electron microscope pictures are obtained, the proportion of crystal fracture of each size level is counted, the broken energy E1 corresponding to each size level is calculated by function, and the proportion of crystal fracture of each size level is calculated, and the error is smaller, and the size level is determined as the size level which can be used for broken energy calculation and speculation, the method takes the proportion of crystal fracture as the intermediate data, and the relationship between the dissociation degree of magnetite and the broken energy is obtained, and a feasible way is found for controlling the dissociation degree of magnetite by broken energy.

Description

Technical Field

[0001] This invention relates to the technical field of the dissociation state between different mineral components after ore crushing, and in particular to a method for intelligent identification of mineral crushing energy and degree of dissociation based on particle morphology characteristics. Background Technology

[0002] The primary task of grinding is the dissociation of different mineral components within ore. Grinding reduces ore particle size, thus achieving mineral dissociation. The degree of dissociation of the target mineral directly impacts process selection and separation efficiency. Grinding of metallic ores is a dissociative grinding process. Ore is an aggregate of different minerals. The bonding force at the interface between these aggregates is less than the cohesive force between particles within the minerals. This means that mineral dissociation and particle size reduction may not occur simultaneously. When appropriate energy is applied, dissociation occurs preferentially between different minerals, followed by particle size reduction. Therefore, mineral dissociation behavior is directly influenced by the amount of crushing energy applied.

[0003] The particle size evolution and compositional separation during the crushing process of magnetite ore particles are essentially intergranular and transgranular fractures between the same and different phases. During the crushing process, magnetite ore will generate particles that are completely liberated from magnetite and quartz.

[0004] Fracture at the interfaces of magnetite ore, and fracture at the grain boundaries between different minerals within the ore, can enhance mineral liberation. Mineral interfaces are areas of weak force; under suitable fracturing energy, intergranular cracks will occur along these interfaces, resulting in intergranular fracture. Therefore, for magnetite ore, a higher proportion of intergranular fracture at the interfaces between different minerals will increase the degree of individual mineral liberation.

[0005] The amount of energy applied has a direct impact on the fracture mode of ore. As the crushing energy increases, the fracture mode of magnetite ore is transgranular fracture. The appropriate crushing energy is needed to achieve a suitable degree of mineral liberation after crushing magnetite ore, which is beneficial for the selection and sorting of subsequent processes. It is necessary to find a suitable method to analyze the relationship between crushing energy and mineral liberation. This relationship is of great significance for guiding the crushing process of magnetite ore and selecting the appropriate crushing energy. Summary of the Invention

[0006] The purpose of this invention is to solve the above-mentioned problems and provide a method for intelligent identification of mineral grinding energy and degree of dissociation based on particle morphology characteristics.

[0007] The technical solution of this invention is:

[0008] A method for intelligent identification of mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics includes the following steps:

[0009] a. Use a drop weight impact test machine to crush the magnetite ore sample, and the crushed ore particles are classified by size through sieving;

[0010] b. Use a mineral liberation analyzer to measure the mineral liberation degree of some small-sized particles after size classification, and record it. Through the inventor's preliminary observation, it is found that the mineral liberation degree of the crushed particles of large-sized particles is low, and has no subsequent sorting value and significance. Therefore, small-sized particles are selected for measurement to avoid affecting work efficiency.

[0011] c. Adjust the impact crushing energy of the drop weight impact test machine and perform steps a-b multiple times;

[0012] d. According to the records of step c, establish a magnetite monomer liberation degree distribution graph under different crushing energies and a quartz monomer liberation degree distribution graph under different crushing energies; the relationship between liberation degree and crushing energy can be more intuitively analyzed;

[0013] e. Analyze the two distribution graphs in step d. The mineral liberation degree of the particles in part A is high, and a certain crushing energy is determined as the appropriate crushing energy E. Focusing on the crushing energy corresponding to the high liberation degree after crushing is more instructive and has better guidance;

[0014] f. Scanning the particles of each size level under the appropriate crushing energy E by electron microscopy to obtain the intergranular fracture proportion in the particle fracture of each size level;

[0015] g. According to the intergranular fracture proportion obtained in step f and the functional relationship between the intergranular fracture proportion and the crushing energy, the corresponding calculated crushing energy E1 of each size level is calculated; according to the known relationship between the intergranular fracture proportion and the crushing energy,

[0016] h. Compare E1 and E to obtain the error range; determine that the error of E1 and E corresponding to part B of the size level is small, and the intergranular fracture proportion of the crushed particles of part B of the size level can be used to calculate and infer the appropriate crushing energy E. According to the known relationship between the intergranular fracture proportion and the crushing energy, the relationship between the mineral liberation degree of the easily sorted particle level and the crushing energy is established. In this way, the mineral liberation degree of the easily sorted particle level can be controlled by the crushing energy in reverse, so as to optimize the crushing process through the crushing energy and improve the crushing effect.

[0017] Preferably, the method is performed on single-particle magnetite ore, and the sample in step a is Φ50x25mm,

[0018] The particle size classification in step a is -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm;

[0019] The mineral dissociation degree of the -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm particle size grades in step b is measured respectively, and recorded;

[0020] The multiple impact crushing energy in step c is 226.01J, 376.69J, 527.36J, 640.37J, 753.38J, 866.38J respectively;

[0021] In step e, when the crushing energy is 753.38J, the dissociation degree of the -0.074+0.038mm,

[0022] The dissociation degree of the -0.038+0.019mm particle grade of the magnetite ore after crushing is higher, which is 64.71%, 83.49% respectively, and is defined as the A part particle size grade, and the suitable crushing energy E is 753.38J;

[0023] In step f, the -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm particle grades of the magnetite ore after crushing at the crushing energy of 753.38J are scanned by electron microscope to obtain the proportion of intergranular fracture in the particle fracture in each particle size grade;

[0024] In step h, the calculated E1 and E of the -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm particle size grades have small errors, and the errors are 3.77%, 5.93%, 1.52% respectively, and are defined as the B part particle size grade.

[0025] Preferably, the method is performed on a particle group of magnetite, and the sample in step a is a particle group of -2.000+1.180mm particle size of magnetite, 150g is taken for each experiment,

[0026] The particle size classification in step a is -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm;

[0027] The mineral dissociation degree of -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm particle size classification in step b is measured respectively, and recorded;

[0028] The multiple impact crushing energy in step c is multiple impact energy, which is 332.59J, 432.36J, 532.14J, 631.25J, 731.69J respectively;

[0029] In step e, when the crushing energy is 432.36J, the dissociation degree of -0.074+0.038mm,

[0030] The dissociation degree of -0.038+0.019mm particle size classification of magnetite after crushing is higher, which is 69.57%, 84.03% respectively, and is defined as A part of particle size classification, and the appropriate crushing energy E is 432.36J;

[0031] In step f, the -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm particle size classification of magnetite after crushing at the crushing energy of 432.36J is scanned by electron microscope to obtain the proportion of intergranular fracture in the particle fracture of each particle size classification;

[0032] In step h, the error between E1 and E of -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm particle size classification is small, and the error is within 5%, which is determined as B part of particle size classification.

[0033] Preferably, the functional relationship between the proportion of intergranular fracture of each particle size classification in step g and the crushing energy is obtained by the following steps;

[0034] 41. Using a drop-weight impact test machine to crush the magnetite sample, selecting single particle magnetite particles after crushing under the same crushing energy, grading the crushed magnetite particles by sieving, obtaining the fracture pictures of magnetite particles of different particle sizes by scanning electron microscopy, and counting the proportion of intergranular fracture;

[0035] 42. Adjusting the crushing energy of the drop-weight impact test machine, repeating step 41 multiple times, and sequentially obtaining the fracture morphology of different particle size levels of the particle magnetite under different crushing energies, and counting the proportion of intergranular fracture;

[0036] 43. Collecting the data in step 42, drawing points in the coordinate system, fitting a curve according to the points, and establishing a functional relationship between the intergranular fracture proportion of the crushed magnetite particles of different particle sizes and the crushing energy according to the fitted curve. Through this sub-step, the functional relationship between the intergranular fracture proportion and the crushing energy is obtained, which establishes a prerequisite for the execution of the previous step g.

[0037] Preferably, the software picture recognition method in step 41 is used to count the proportion of intergranular fracture, which uses the following process. By establishing a computer software recognition method, the efficiency can be improved, so that the previous step 42 can be quickly performed, and only then does it have industrial application value.

[0038] 51. Crush the magnetite by a drop-weight impact test machine, collect and evenly divide the crushed magnetite particles into N groups, and then perform electron microscopy scanning on each group. N≥1000;

[0039] 52. Manually calibrate the surface morphology of each region in each scanned picture, and calibrate different particle morphologies as one of intergranular fracture or transgranular fracture, and count the number of intergranular fracture particles and transgranular fracture particles in the picture. Repeat this step to obtain N labeled pictures,

[0040] 53. Establish a full convolutional neural network model based on the SSD algorithm in the computer for learning and training the labeled pictures after step 52,

[0041] 54. The majority of the calibrated pictures completed in step 52 are imported into the full convolutional neural network model based on the SSD algorithm on the computer for deep learning training, and after deep learning training, a picture recognition model is formed, and the remaining part of the calibrated pictures is used to test the picture recognition model, if the test accuracy meets the requirements, stop deep learning training, and retain the trained model; if the test accuracy does not meet the requirements, continue deep learning training and testing until the test accuracy meets the requirements, and retain the trained picture recognition model; the model established by the full convolutional neural network model based on the SSD algorithm after learning a large number of artificially calibrated pictures can improve the recognition efficiency, facilitate industrial application, and quickly obtain the results of the crystal fracture proportion.

[0042] 55. The picture recognition model obtained in step 54 is connected with the software in the computer, and the software is used to complete step 41.

[0043] The beneficial effects of the present application are: a mineral fragment grinding energy discrimination and dissociation degree intelligent identification method based on particle morphology characteristics, first, the magnetite is broken by a falling weight impact testing machine, and then the particle size classification is carried out, the mineral dissociation instrument is used to detect and record the dissociation degree of the broken particles of different broken energy and relatively small particle size, the dissociation degree distribution graph of different minerals is established according to the detected data, the appropriate broken energy E is determined after analyzing the distribution graph, and then the broken particles of each particle size level under the broken energy are scanned by an electron microscope, the electron microscope pictures are obtained, the crystal fracture proportion of each particle size level is counted, and the broken energy E1 corresponding to each particle size level is calculated by function, and the crystal fracture proportion of those particle size levels with smaller error after calculation is determined by comparing E1 and E, and the particle size level that can be used for broken energy calculation and speculation is determined. The method takes the crystal fracture proportion as the intermediate data, explores the relationship between the dissociation degree of magnetite and the broken energy, and finds a feasible way to control the dissociation degree of magnetite by broken energy. BRIEF DESCRIPTION OF DRAWINGS

[0044] Fig. 1 is the magnetite monomer dissociation degree distribution graph of the single particle magnetite under different broken energies of the present application;

[0045] Fig. 2 is the quartz monomer dissociation degree distribution graph of the single particle magnetite under different broken energies of the present application;

[0046] Fig. 3 is the error comparison graph of the crystal fracture proportion of each particle size level of the broken single particle magnetite and the calculated broken energy E1 of the present application;

[0047] Fig. 4 is the magnetite monomer dissociation degree distribution graph of the particle group magnetite under different broken energies of the present application;

[0048] Fig. 5 is a single mineral liberation degree distribution diagram of quartz under different crushing energies of the particle group magnetite stone of the present application;

[0049] Fig. 6 is an intergranular fracture proportion and calculation error comparison diagram of crushing energy E1 of each particle size level of the particle group magnetite stone of the present application after crushing; DETAILED DESCRIPTION

[0050] Example 1: see Figs. 1-3 A mineral crushing and grinding energy discrimination and liberation degree intelligent identification method based on particle morphology characteristics, comprising the following steps:

[0051] a. Use a drop weight impact testing machine to crush the magnetite stone sample, and realize particle size classification of the crushed ore particles through screening; this method is for single particle magnetite stone, and the sample in step a is Φ50*25mm particle size classification, which is-1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm;

[0052] b. Use a mineral liberation instrument to measure the mineral liberation degree of the particle size classification after crushing of the four small particle sizes of-0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm 8, -0.038+0.019mm, and record it. Through the preliminary observation of the inventor, it is found that the mineral liberation degree of the crushed particles of the large particle size classification is low, and it does not have subsequent sorting value and significance, so the small particle size is selected for measurement to avoid affecting the work efficiency.

[0053] c. Adjust the impact crushing energy of the drop weight impact testing machine, and perform steps a-b multiple times;

[0054] The multiple impact crushing energies are 226.01J, 376.69J, 527.36J, 640.37J, 753.38J, and 866.38J;

[0055] d. According to the records of step c, establish a single mineral liberation degree distribution diagram of magnetite under different crushing energies and a single mineral liberation degree distribution diagram of quartz under different crushing energies; the relationship between the liberation degree and the crushing energy can be more intuitively analyzed,

[0056] e. Analyze the two distribution maps in step d, and determine a certain crushing energy as the appropriate crushing energy E, because the particle mineral dissociation degree of the A part particle size level is high. The crushing energy corresponding to the high dissociation degree after focusing on crushing is the appropriate crushing energy, which is more instructive. Specifically, when the crushing energy is 753.38 J, the dissociation degree of the crushed particles of the -0.074+0.038 mm and -03038+0.019 mm particle levels of the magnetite ore is relatively high, which is 64.71% and 83.49% respectively. The -0.074+0.038 mm and -03038+0.019 mm particle levels are determined as the A part particle size level, and the appropriate crushing energy E is determined as 753.38 J;

[0057] f. Scan the particles of each particle size level under the appropriate crushing energy E by an electron microscope to obtain the intergranular fracture proportion in the particle fracture of each particle size level. Scan the particles of the -1.180+0.900 mm, -0.900+0.600 mm, -0.600+0.300 mm, -0.300+0.150 mm, -0.150+0.074 mm, -0.074+0.03 mm and -0.038+0.019 mm particle levels of the magnetite ore after crushing under the crushing energy of 753.38 J by an electron microscope to obtain the intergranular fracture proportion in the particle fracture of each particle size level.

[0058] g. According to the intergranular fracture proportion obtained in step f and the functional relationship between the intergranular fracture proportion and the crushing energy, the calculated crushing energy E1 of each particle size level is calculated respectively. The relationship between the intergranular fracture proportion and the crushing energy is known.

[0059] h. Compare E1 and E to obtain the error range; determine the E1 and E of the B part particle size level, and the error is small, so the intergranular fracture proportion of the particles after crushing of the B part particle size level can be used to calculate and speculate the appropriate crushing energy E. According to the known relationship between the intergranular fracture proportion and the crushing energy, the relationship between the mineral dissociation degree of the easily sorted particle level and the crushing energy is established, so that the mineral dissociation degree of the easily sorted particle level can be controlled by the crushing energy in reverse, so as to optimize the crushing process and improve the crushing effect by the crushing energy. The calculated E1 and E of the -1.180+0.900 mm, -0.900+0.600 mm and -0.600+0.300 mm particle levels have small errors, which are 3.77%, 5.93% and 1.52% respectively, and are determined as the B part particle size level.

[0060] The functional relationship between the intergranular fracture proportion of each particle size level and the crushing energy in step g is obtained by the following steps.

[0061] 41. Using a drop-weight impact test machine to crush the magnetite sample, selecting single particle magnetite particles after crushing under the same crushing energy, grading the crushed magnetite particles by sieving, obtaining the fracture pictures of magnetite particles of different particle sizes by scanning electron microscopy, and counting the proportion of intergranular fracture;

[0062] 42. Adjust the crushing energy of the drop-weight impact test machine, repeat step 41 multiple times, and sequentially count the proportion of intergranular fracture of different particle sizes of the magnetite particles under different crushing energies;

[0063] 43. Collect the data in step 42, plot points in the coordinate system, fit a curve according to the plotted points, and establish a functional relationship between the intergranular fracture proportion of different particle sizes of the crushed magnetite particles and the crushing energy according to the fitted curve. Through this sub-step, the functional relationship between the intergranular fracture proportion and the crushing energy is obtained, which establishes a prerequisite for the execution of the previous step g.

[0064] The method of step 41 uses software picture recognition to count the proportion of intergranular fracture, which can improve efficiency, so that the previous step 42 can be quickly performed and has industrial application value. The software picture recognition method uses the following process,

[0065] 51. Crush the magnetite by a drop-weight impact test machine, collect and evenly divide the crushed magnetite particles into N groups, and then perform electron microscopy scanning on each group. The N scanned pictures are imported into a computer, and N ≥ 1000.

[0066] 52. Manually label the surface morphology of each region in each scanned picture, and label different particle morphologies as one of intergranular fracture or transgranular fracture. Count the number of intergranular fracture particles and transgranular fracture particles in the picture, and repeat the step to obtain N labeled pictures.

[0067] 53. Establish a full convolutional neural network model based on the SSD algorithm in the computer for learning and training the labeled pictures after step 52,

[0068] 54. The majority of the calibrated pictures completed in step 52 are imported into the full convolutional neural network model based on the SSD algorithm on the computer for deep learning training, and after deep learning training, a picture recognition model is formed. The remaining part of the calibrated pictures is used to test the picture recognition model. If the test accuracy meets the requirements, stop the deep learning training, and keep the trained model. If the test accuracy does not meet the requirements, continue the deep learning training and testing until the test accuracy meets the requirements, and keep the trained picture recognition model. The model established by the full convolutional neural network model based on the SSD algorithm after learning a large number of artificially calibrated pictures can improve the recognition efficiency, facilitate industrial application, and quickly obtain the results of the crystal fracture ratio.

[0069] 55. The picture recognition model obtained in step 54 is connected with the software in the computer, and the software is used to complete step 41.

[0070] Example 2: see Figs. 4-6 An intelligent recognition method for mineral crushing energy discrimination and dissociation degree based on particle morphology characteristics, which is used for particle group magnetite ore, includes the following steps:

[0071] a. Use the drop weight impact testing machine to crush the magnetite ore sample. The crushed ore particles are classified by sieving. The sample in step a is a particle group of -2.000+1.180 mm size of magnetite ore. 150g is taken for each experiment. The particle size classification after crushing is -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm;

[0072] b. Use the mineral dissociation instrument to measure the mineral dissociation degree of part of the small particle size classified particles, and record it. In step b, the -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm size levels are measured for mineral dissociation degree, respectively. Through the preliminary observation of the inventor, it is found that the mineral dissociation degree of the large particle size classification of the crushed particles is low, and it does not have subsequent sorting value and significance, so the small particle size is selected for measurement to avoid affecting the work efficiency.

[0073] c. Adjust the impact crushing energy of the drop weight impact testing machine, and perform steps a-b multiple times. In step c, the multiple impact crushing energies are 332.59J, 432.36J, 532.14J, 631.25J, and 731.69J, respectively.

[0074] d. According to the record of step c, the magnetite monomer dissociation degree distribution diagram under different crushing energy and the quartz monomer dissociation degree distribution diagram under different crushing energy are established, and the relationship between the dissociation degree and the crushing energy can be more intuitively analyzed

[0075] e. The two distribution diagrams in step d are combined for analysis. The dissociation degree of the particle minerals corresponding to the A part particle size level is high, and a certain crushing energy is determined as the suitable crushing energy E. When the crushing energy is 432.36 J, the dissociation degree of the magnetite stone particles after crushing in the-0.074+0.038 mm and-03038+0.019 mm particle levels is relatively high, which is 69.57% and 84.03% respectively, and the A part particle size level is determined, and the suitable crushing energy E is 432.36 J;

[0076] f. The grain fracture along the crystal of the particles in each particle size level under the suitable crushing energy E is obtained by scanning the particles in each particle size level under the suitable crushing energy E by an electron microscope. The grain fracture along the crystal of the particles in each particle size level is obtained by scanning the particles in the-1.180+0.900 mm, -0.900+0.600 mm, -0.600+0.300 mm, -0.300+0.150 mm, -0.150+0.074 mm, -0.074+0.03 mm and-0.038+0.019 mm particle levels of the magnetite stone after crushing under the crushing energy of 432.36 J by an electron microscope;

[0077] g. According to the grain fracture along the crystal obtained in step f and the functional relationship between the grain fracture along the crystal and the crushing energy, the corresponding calculated crushing energy E1 of each particle size level is calculated. The relationship between the grain fracture along the crystal and the crushing energy is known.

[0078] h. The error range is obtained by comparing E1 and E. The E1 and E of the B part particle size level are small, and the grain fracture along the crystal of the particles after crushing in the B part particle size level can be used to calculate and predict the suitable crushing energy E. The calculated E1 and E of the-1.180+0.900 mm, -0.900+0.600 mm, -0.600+0.300 mm, -0.300+0.150 mm and-0.150+0.074 mm particle levels are small, and the error is within 5%. The B part particle size level is determined. According to the known relationship between the grain fracture along the crystal and the crushing energy, the relationship between the dissociation degree of the easily sorted particle level minerals and the crushing energy is established, so that the dissociation degree of the easily sorted particle level minerals can be controlled by the crushing energy in reverse, so as to optimize the crushing process and improve the crushing effect by the crushing energy.

[0079] The function relationship between the crystal fracture ratio of each particle size level in step g and the crushing energy is obtained by the following steps;

[0080] 41. Crush the magnetite sample by using the falling weight impact test machine, select the single particle magnetite particles after crushing under the same crushing energy, classify the crushed magnetite particles by sieving, obtain the fracture pictures of magnetite particles of different particle size levels by scanning electron microscopy, and count the proportion of intergranular fracture;

[0081] 42. Adjust the crushing energy of the falling weight impact test machine, repeat step 41 multiple times, and sequentially count the proportion of intergranular fracture of different particle size levels of the magnetite particles under different crushing energies;

[0082] 43. Aggregate the data in step 42, draw points in the coordinate system, fit a curve according to the points, and establish a function relationship between the intergranular fracture ratio of different particle size levels of the crushed magnetite and the crushing energy according to the fitted curve. Through this sub-step, the function relationship between the intergranular fracture ratio and the crushing energy is obtained, which establishes a prerequisite for the execution of the previous step g.

[0083] The method of step 41 uses software picture recognition to count the proportion of intergranular fracture, which can improve the efficiency by using computer software recognition method, so that the previous step 42 can be quickly performed, and has industrial application value. The software picture recognition method uses the following process,

[0084] 51. Crush the magnetite by using the falling weight impact test machine, collect and evenly divide the crushed magnetite particles into N groups, and then perform scanning electron microscopy on each group. Import the N scanned pictures into the computer, and N≥1000.

[0085] 52. Manually label the surface morphology of each region in each scanned picture, and label different particle morphologies as one of intergranular fracture or transgranular fracture. Aggregate the number of intergranular fracture particles and transgranular fracture particles in the picture, and repeat the step to obtain N labeled pictures.

[0086] 53. Establish a full convolutional neural network model based on SSD algorithm in the computer for learning and training the labeled pictures after step 52,

[0087] 54. The majority of the calibrated pictures completed in step 52 are imported into the full convolutional neural network model based on the SSD algorithm on the computer for deep learning training, and the picture recognition model is formed after the deep learning training. The remaining part of the calibrated pictures is used to test the picture recognition model. If the test accuracy meets the requirements, stop the deep learning training, and keep the trained model. If the test accuracy does not meet the requirements, continue the deep learning training and testing until the test accuracy meets the requirements, and keep the trained picture recognition model. The model established by the full convolutional neural network model based on the SSD algorithm after learning a large number of artificially calibrated pictures can improve the recognition efficiency, facilitate industrial application, and quickly obtain the results of the crystallographic fracture ratio.

[0088] 55. The software in the computer is connected with the picture recognition model obtained in step 54, and the software is used to complete step 41.

Claims

1. A method for intelligent identification of mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics, comprising the following steps: a. The magnetite ore sample was crushed using a drop impact tester, and the crushed ore particles were then screened to achieve particle size classification. b. Using a mineral liberation analyzer, measure and record the degree of mineral liberation of the smaller particle size fractions after grading. c. Adjust the impact crushing energy of the drop impact tester and repeat steps a and b multiple times; d. Based on the records in step c, establish the distribution map of the degree of liberation of magnetite monomers under different crushing energies, and the distribution map of the degree of liberation of quartz monomers under different crushing energies; e. Analyze the two distribution maps in step d. The particle minerals in part A have a high degree of dissociation. Determine a suitable crushing energy E. f. Under suitable crushing energy E, particles of various size classes are scanned by electron microscopy to obtain the proportion of intergranular fracture in the fracture surface of particles of each size class. g. Based on the intergranular fracture ratio obtained in step f, and based on the functional relationship between the intergranular fracture ratio and the crushing energy, calculate the calculated crushing energy E1 corresponding to each particle size level. h. Compare E1 and E to obtain the error range; determine that the error between E1 and E corresponding to the particle size class of part B is small, and the appropriate crushing energy E can be calculated and estimated by the intergranular fracture ratio of the crushed particles of part B.

2. The intelligent identification method for mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics according to claim 1, characterized in that: This method is applied to single-particle magnetite ore, and the sample in step a is Φ50×25mm. The particle size classification in step a is -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm. In step b, the degree of mineral liberation was measured and recorded for the particle sizes of -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm. The energy of the multiple impact crushing in step c is 226.01J, 376.69J, 527.36J, 640.37J, 753.38J, and 866.38J, respectively. In step e, when the crushing energy is 753.38J, the degree of liberation of the crushed magnetite ore particles of -0.074+0.038mm and -03038+0.019mm is relatively high, at 64.71% and 83.49% respectively. These are classified as particle size class A, and the suitable crushing energy E is determined to be 753.38J. In step f, the magnetite ore particles of the following sizes (-1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm) after crushing at a crushing energy of 753.38J are scanned by electron microscopy to obtain the proportion of intergranular fracture in the fracture surface of each particle size. In step h, the calculated E1 and E errors for particle size levels of -1.180+0.900mm, -0.900+0.600mm, and -0.600+0.300mm are small, with errors of 3.77%, 5.93%, and 1.52% respectively, and are determined to be part B particle size levels.

3. The intelligent identification method for mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics according to claim 1, characterized in that: This method is applied to magnetite ore with a particle size distribution. The sample in step a consists of magnetite ore particles ranging from -2.000 to +1.180 mm, with 150 g taken for each experiment. The particle size classification in step a is -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm. In step b, the degree of mineral liberation was measured and recorded for the particle sizes of -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm. The impact energies in step c are 332.59 J, 432.36 J, 532.14 J, 631.25 J, and 731.69 J, respectively. In step e, when the crushing energy is 432.36J, the degree of liberation of the crushed magnetite ore particles of -0.074+0.038mm and -03038+0.019mm is relatively high, at 69.57% and 84.03% respectively. These are classified as particle size class A, and the suitable crushing energy E is determined to be 432.36J. In step f, the magnetite ore particles of the following sizes (-1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, -0.150+0.074mm, -0.074+0.03mm, and -0.038+0.019mm) after crushing at a crushing energy of 432.36J are scanned by electron microscopy to obtain the proportion of intergranular fracture in the fracture surface of each particle size. In step h, the E1 and E errors of the particle size calculations for -1.180+0.900mm, -0.900+0.600mm, -0.600+0.300mm, -0.300+0.150mm, and -0.150+0.074mm are small, all within 5%, and are determined to be part B particle size levels.

4. The intelligent identification method for mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics according to claim 1, characterized in that: The functional relationship between the intergranular fracture ratio and the fragmentation energy at each particle size level in step g is obtained through the following steps; 41. The magnetite ore sample was crushed using a drop impact tester. The crushed particles of a single magnetite ore under the same crushing energy were selected and classified by sieving. Fracture images of magnetite ore particles of different particle sizes were obtained by scanning electron microscopy, and the proportion of intergranular fracture was counted.

42. Adjust the crushing energy of the drop impact tester and repeat step 41 multiple times. Sequentially examine the fracture morphology of magnetite ore of different particle size grades under different crushing energies and count the proportion of intergranular fracture.

43. Summarize the data from step 42, plot the points in the coordinate system, fit the curve based on the plotted points, and establish the functional relationship between the intergranular fracture ratio of different particle size grades of magnetite ore after crushing and the crushing energy based on the fitted curve.

5. The intelligent identification method for mineral grinding energy discrimination and dissociation degree based on particle morphology characteristics according to claim 4, characterized in that: In step 41, the proportion of intergranular fractures is statistically analyzed using software image recognition. This software image recognition method employs the following process.

51. The magnetite ore was crushed by a drop impact tester. The crushed magnetite ore particles were collected and divided into N groups. Each group was spread out and scanned by electron microscopy. The N scanned images were imported into the computer, N≥1000 images.

52. Manually label the surface morphology of particles in each region of each scanned image. Different particle morphologies are labeled as either intergranular fracture or transgranular fracture. The number of intergranular fractured particles and the number of transgranular fractured particles in the image are then calculated. This step is repeated to obtain N labeled images.

53. Build a fully convolutional neural network model based on the SSD algorithm in a computer to learn and train on the labeled images after step 52.

54. Import most of the labeled images completed in step 52 into a fully convolutional neural network model based on the SSD algorithm on the computer for deep learning training. After deep learning training, an image recognition model is formed. Use the remaining labeled images to test the image recognition model. If the test accuracy meets the requirements, stop deep learning training and keep the trained model; if the test accuracy does not meet the requirements, continue deep learning training and testing until the test accuracy meets the requirements, and keep the trained image recognition model.

55. Establish software in the computer to interface with the image recognition model obtained in step 54 above, and use the software to complete step 41.

Images