X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in x-ray intelligent ore sorter

Abstract

The application discloses an X-ray three-dimensional imaging and defect quantitative characterization method for internal defects of ores in an X-ray intelligent ore dressing machine, and relates to the technical field of data processing, and comprises the following steps: acquiring a three-dimensional image of the ore by using an X-ray binocular imaging device; inputting the three-dimensional image of the ore into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and divide internal defect grades of the ore; acquiring current setting parameters of the X-ray intelligent ore dressing machine, constructing an expert experience sequence by using the internal defect characterization data of the ore and the current setting parameters of the X-ray intelligent ore dressing machine; training a multi-agent reinforcement learning model to obtain an internal defect grade separation parameter model; and conveying the ore to be screened to the X-ray intelligent ore dressing machine to obtain different internal defect grade ores by separation. The method can separate different internal defect grade ores, and solves the problem that the setting parameters of the X-ray intelligent ore dressing machine are updated automatically according to the internal defects of the ores.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore dressing machine. BACKGROUND

[0002] The intelligent ore dressing machine can pre-select and discard waste in the ore production process, effectively improving the utilization rate of ores. The intelligent ore dressing machine can be divided into three major mechanisms in structure, which are feeding mechanism, detection and recognition mechanism, and sorting execution mechanism. The feeding mechanism is responsible for uniformly feeding the ores into the detection area; the detection and recognition mechanism uses an X-ray detector to scan the ores and processes the task through image processing algorithms; the sorting execution mechanism accurately controls the sorting device according to the information provided by the detection and recognition mechanism, effectively separates the concentrate and waste ores, and thus improves the efficiency and quality of ore dressing. The whole system realizes the collaborative work between the mechanisms through a highly integrated control system, ensuring the continuity and stability of the ore dressing process.

[0003] A phosphorite photoelectric ore dressing separation process is disclosed in Chinese Patent No. CN 112221657 B, in which the phosphorite raw ore is sieved, the sieved ores are transported to a photoelectric separator by a belt conveyor, the ores are identified by X-ray perspective, each ore is identified by X-ray, and the ores and waste rocks are distinguished by intelligent algorithms. After the identification by the photoelectric separator, the concentrate and tailings are separated by precise hitting with a high-speed air blast gun, and the photoelectric ore dressing of the phosphorite is completed. This method can perform photoelectric ore dressing on phosphorite, but does not consider the internal defects of the ore dressing stone and the problem that the parameter error of the ore dressing machine increases with the use time of the ore dressing machine.

[0004] A scheelite separation method based on XRT rays is disclosed in Chinese Patent No. CN 110046653 B, which includes: transmitting raw ore through an X-ray processing area and a spraying area in sequence and obtaining an X-ray image of the raw ore; processing the X-ray image of the raw ore by using a contrast neural network to separate scheelite ores and calculate the spatial coordinates of the scheelite ores; calculating the falling time of the scheelite ores falling to the position of the spray valve in the spraying area based on the spatial coordinates of the scheelite ores and controlling the spray valve to spray compressed air to the scheelite ores for separation based on the falling time. This method can accurately identify scheelite ores and reduce the discard rate of raw ores, but does not consider the influence of multi-grade division of ores and changes in parameter error of the ore dressing machine on the accuracy of ore dressing.

[0005] Therefore, the present application provides a method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore dressing machine. SUMMARY

[0006] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application proposes an X-ray three-dimensional imaging and quantitative characterization method for internal defects of ores in an X-ray intelligent ore dressing machine, which can solve the problems of multi-grade classification of ores and the influence of parameter error changes of the ore dressing machine on the accuracy of ore dressing.

[0007] The present application provides an X-ray three-dimensional imaging and quantitative characterization method for internal defects of ores in an X-ray intelligent ore dressing machine, which specifically comprises the following steps:

[0008] S1, obtaining a three-dimensional image of the ore by using an X-ray binocular imaging device; performing feature processing on the three-dimensional image of the ore to obtain an internal defect data set of the ore;

[0009] S2, inputting the three-dimensional image of the ore into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and dividing the internal defect grade of the ore; the internal defect characterization model of the ore is trained by the internal defect data set of the ore;

[0010] S3, obtaining the current setting parameters of the X-ray intelligent ore dressing machine, and constructing an expert experience sequence by using the internal defect characterization data of the ore and the current setting parameters of the X-ray intelligent ore dressing machine;

[0011] S4, storing the expert experience sequence into an expert experience pool, and training a multi-agent reinforcement learning model to obtain an internal defect grade separation parameter model of the ore;

[0012] S5, conveying the ore to be screened to the X-ray intelligent ore dressing machine, and based on the internal defect characterization model of the ore and the internal defect grade separation parameter optimization model, separating different internal defect grade ores.

[0013] In particular, the X-ray binocular imaging device comprises one X-ray source and two flat panel detector arrays.

[0014] In particular, the step of performing feature processing on the three-dimensional image of the ore to obtain an internal defect data set of the ore comprises:

[0015] Step S11: normalizing, image denoising and voxel data enhancement are performed on the three-dimensional image of the ore;

[0016] Step S12: edge positioning is performed on the internal defects of the ore by using an edge detection algorithm to obtain edge data of the defects;

[0017] Step S13: the region growing algorithm is used to segment the defect region to obtain geometric data of the defects;

[0018] Step S14: the internal defects of the processed three-dimensional image of the ore are labeled to obtain an internal defect data set of the ore;

[0019] The ore is a particle size ore obtained after crushing and vibrating screening of raw ore; the raw ore is a concentrate containing only one ore after pre-selection and waste removal.

[0020] In particular, the step of inputting the three-dimensional image of the ore into the internal defect characterization model of the ore to obtain internal defect characterization data of the ore and dividing the internal defect grade of the ore is:

[0021] Step S21: dividing the internal defect data set of the ore into a training set and a validation set, inputting into a 3D CNN network model for training, and obtaining an internal defect characterization model of the ore;

[0022] The ratio of the training set and the validation set is 7:3; the 3D-CNN is composed of an input layer, a convolution layer, a pooling layer, a full connection layer and an output layer, and the output layer outputs internal defect characterization data of the ore;

[0023] Step S22: inputting the three-dimensional image of the ore into the internal defect characterization model of the ore to obtain internal defect characterization data of the ore;

[0024] The internal defect characterization data of the ore includes defect classification results and internal defect volume ratio, the defect classification results include two classification results of having defects and no defects, and the internal defect volume ratio refers to the ratio of the volume of all internal defects in the ore to the volume of the ore;

[0025] Step S23: dividing the internal defect grade of the ore according to the internal defect volume ratio;

[0026] The method for dividing the internal defect grade of the ore is:

[0027] Set a first threshold X1 and a second threshold X2;

[0028] If the internal defect volume ratio is greater than and equal to the first threshold X1, the internal defect grade of the ore is A grade;

[0029] If the internal defect volume ratio is greater than the second threshold X2 and less than the first threshold X1, the internal defect grade of the ore is B grade;

[0030] If the internal defect volume ratio is less than and equal to the second threshold X2, the internal defect grade of the ore is C grade.

[0031] In particular, the current X-ray intelligent ore dressing machine setting parameter is obtained, and the internal defect characterization data of the ore and the current X-ray intelligent ore dressing machine setting parameter are used to construct an expert experience sequence, wherein:

[0032] The current X-ray intelligent ore dressing machine setting parameter includes feeding belt speed, machine body belt speed, vibration amplitude, vibration frequency, vibration direction angle, and nozzle injection pressure;

[0033] The expert experience sequence comprises an ore internal defect state set, an intelligent ore dressing machine parameter action set, a reward value set and a next state set;

[0034] The ore internal defect state set comprises defect classification result data and internal defect volume ratio data;

[0035] The intelligent ore dressing machine parameter action set comprises feeding belt speed data, machine body belt speed data, vibration amplitude data, vibration frequency data, vibration direction angle data, nozzle spray pressure data.

[0036] In particular, the step of storing the expert experience sequence into the expert experience pool and training the multi-agent reinforcement learning model to obtain the ore internal defect grade separation parameter model is:

[0037] Step S31: constructing an expert experience sequence using the ore internal defect representation data and the current X-ray intelligent ore dressing machine setting parameters;

[0038] The ore internal defect state set, the intelligent ore dressing machine parameter action set, the separation accuracy reward value set and the next time ore internal defect state set;

[0039] Step S32: storing the expert experience sequence into the expert experience pool to complete the initialization of the expert experience pool;

[0040] Step S33: training the multi-agent reinforcement learning model using the expert experience pool to obtain the ore internal defect grade separation parameter model.

[0041] Further, the step of storing the expert experience sequence into the expert experience pool and training the multi-agent reinforcement learning model to obtain the ore internal defect grade separation parameter model is:

[0042] Step S41: constructing a prediction network and a target network;

[0043] Step S42: constructing N agents based on the prediction network, the target network and the experience pool, wherein is a variable and ;

[0044] Step S43: initializing the experience pool as the expert experience pool; the experience pool capacity is M;

[0045] Step S44: initializing the prediction network using random prediction network weights , initializing the target network using random target network weights , initializing the parameter state set , batch n and step m;

[0046] Step S45: Randomly batch n obtains expert experience sequences from the experience pool. Training was conducted, among which ;

[0047] Predict the set of internal defect states of the ore input to the network. Intelligent mineral processing machine parameter action set Output prediction function The next moment's intelligent mineral processing machine parameter action set ;

[0048] The target network input is the set of internal defect states of the ore at the next time step. Output target value function , The calculation formula is:

[0049] ;

[0050] In the formula, For the set of defect states inside the ore Intelligent mineral processing machine parameter action set The reward value for the operation effect. Discount factor and , The set of internal defect states of the ore at the next moment The next set of intelligent mineral processing machine parameters and actions was selected. The obtained maximum predicted value function;

[0051] Step S46: Every m steps, predict the weight parameters of the network. Weight parameters copied to the target network ;

[0052] Step S47: Calculate the joint prediction function of the multi-agent system. The calculation formula is:

[0053] ;

[0054] Calculate the joint objective function of multiple agents The calculation formula is:

[0055] ;

[0056] In the formula, The function and the g function are calculated using a BP neural network model;

[0057] The multi-agent loss function is updated to The calculation formula is:

[0058] ;

[0059] Step S48: minimizing the MSE loss function using the gradient descent method , updating the network model parameters and until convergence, obtaining the internal defect grade separation parameter model of the ore;

[0060] The network structure of the prediction network and the target network is the same, and both adopt a BP neural network as a basic network structure; the BP neural network is composed of an input layer, a hidden layer and an output layer;

[0061] The number of neurons in the input layer is equal to the number of internal defect state sets of the ore;

[0062] The number of neurons in the output layer is equal to the number of actions in the intelligent ore dressing machine parameter action set.

[0063] Another aspect of the present application provides an X-ray three-dimensional imaging and quantitative characterization system for internal defects of an ore in an X-ray intelligent ore dressing machine, comprising:

[0064] An ore image processing module: used to acquire an ore three-dimensional image by using an X-ray binocular imaging device; and to perform feature processing on the ore three-dimensional image to obtain an internal defect data set of the ore;

[0065] An ore defect characterization module: used to input the ore three-dimensional image into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and to divide the internal defect grade of the ore; the internal defect characterization model of the ore is obtained by training the internal defect data set of the ore;

[0066] An ore machine parameter acquisition module: used to acquire current X-ray intelligent ore dressing machine setting parameters, and to construct an expert experience sequence by using the internal defect characterization data of the ore and the current X-ray intelligent ore dressing machine setting parameters;

[0067] An ore machine parameter optimization module: used to store the expert experience sequence into an expert experience pool, and to obtain an internal defect grade separation parameter model of the ore by training a multi-agent reinforcement learning model;

[0068] An ore intelligent separation module: used to convey the ore to be screened to the X-ray intelligent ore dressing machine, and to separate different internal defect grade ores based on the internal defect characterization model of the ore and the internal defect grade separation parameter optimization model of the ore.

[0069] Another aspect of the present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the X-ray three-dimensional imaging and quantitative characterization method for internal defects of an ore in an X-ray intelligent ore dressing machine.

[0070] Yet another aspect of the present application provides a readable storage medium, which stores a computer program, the computer program being adapted to be loaded by a processor to implement the steps in the X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in an X-ray intelligent ore dressing machine described above.

[0071] The beneficial technical effects of the present application are:

[0072] The present application uses binocular imaging combined with X-ray technology to obtain a three-dimensional image of the ore, which can accurately reflect the internal structure of the ore compared to a two-dimensional image, thereby improving the accuracy of characterization of internal defects of the ore; the trained internal defect characterization model of the ore can extract key features of internal defects of the ore and automatically classify the internal defects, and the internal defect characterization model of the ore is trained by an internal defect data set, thereby realizing a closed-loop design from data to model; the introduction of the expert experience pool combines artificial experience and intelligent algorithms, so that the system can continuously accumulate and optimize the separation parameters, the construction of the expert experience sequence associates the ore defect characterization data with the ore dressing machine parameters, thereby forming a complete workflow from perception to decision making, and the multi-agent reinforcement learning model enables the system to continuously optimize the ore separation parameters based on experience and real-time feedback, thereby realizing accurate separation based on the defect grade on the one hand, and dynamically adapting to different ore conditions on the other hand, thereby improving the separation efficiency and accuracy of different internal defect grades of the ore, thereby improving the resource utilization rate and reducing the energy consumption and cost in the ore dressing process. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 is a flowchart of an X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in an X-ray intelligent ore dressing machine according to Embodiment 1 of the present application.

[0074] Figure 2 is a step of classifying internal defect grades of ores according to Embodiment 1 of the present application.

[0075] Figure 3 is a step of training a multi-agent reinforcement learning model according to Embodiment 1 of the present application.

[0076] Figure 4 is an X-ray three-dimensional imaging and quantitative characterization system of internal defects of ores in an X-ray intelligent ore dressing machine according to Embodiment 2 of the present application.

[0077] Figure 5 is a schematic structural diagram of an electronic device according to Embodiment 3 of the present application.

[0078] Figure 6 is a schematic structural diagram of a computer readable storage medium according to Embodiment 4 of the present application. DETAILED DESCRIPTION

[0079] To better understand this application, various aspects of this application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely illustrative of exemplary embodiments of this application and are not intended to limit the scope of this application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0080] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustrative purposes only and are not strictly to scale. As used herein, the terms “approximately,” “about,” and similar terms are used to indicate approximation, not degree, and are intended to illustrate inherent deviations in measured or calculated values ​​that will be recognized by one of ordinary skill in the art. Furthermore, the order in which the steps are described in this application does not necessarily indicate the order in which these steps occur in actual operation, unless otherwise expressly defined or deduced from the context.

[0081] It should also be understood that expressions such as "comprising," "including," "having," "containing," and / or "comprising" are open-ended rather than closed-ended expressions in this specification, indicating the presence of the stated features, elements, and / or components, but not excluding the presence of one or more other features, elements, components, and / or combinations thereof. Furthermore, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features, not just individual elements in the list. Additionally, when describing embodiments of this application, the word "may" is used to mean "one or more embodiments of this application." And the term "exemplary" is intended to refer to examples or illustrations.

[0082] Unless otherwise specified, all terms used herein (including engineering and technical terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that, unless expressly stated herein, terms defined in common dictionaries shall be interpreted as having the meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or overly formalized meaning.

[0083] It should be noted that, where there is no conflict, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0084] Example 1

[0085] like Figures 1-3 As shown, this embodiment provides a method for X-ray three-dimensional imaging and quantitative characterization of internal defects in ore in an X-ray intelligent mineral processing machine;

[0086] like Figure 1 As shown, the steps of the X-ray three-dimensional imaging and quantitative characterization method for internal defects of ore in this X-ray intelligent ore beneficiation machine include:

[0087] S1. Obtain three-dimensional images of the ore using an X-ray binocular imaging device; perform feature processing on the three-dimensional images of the ore to obtain a dataset of internal defects in the ore;

[0088] Specifically, the X-ray binocular imaging device includes one X-ray source and two flat panel detector arrays.

[0089] X-ray two-dimensional imaging can only acquire planar projection information of the ore, making it difficult to fully reflect the internal structure and defects of the ore. Furthermore, two-dimensional projection is easily affected by the shape, rotation angle, and complex internal features of the ore, leading to inaccurate detection results for internal defects. By utilizing one X-ray source and two flat-panel detector arrays, two-dimensional projection images of the ore can be acquired from different angles. A reconstruction algorithm is then used to generate a three-dimensional image of the ore, clearly expressing its internal volume, cracks, pores, and other structural features, avoiding misjudgments caused by the loss of two-dimensional projection information.

[0090] In a preferred embodiment, internal defects in the ore, such as cracks, pores, and inclusions, possess three-dimensional characteristics. X-ray three-dimensional imaging can express and extract the complete spatial distribution information of these internal defects, providing data support for subsequent quantitative characterization of ore internal defects. The X-ray three-dimensional imaging in the X-ray intelligent ore beneficiation machine employs, but is not limited to, filtered back-projection algorithms and iterative reconstruction algorithms, to reconstruct three-dimensional ore images from two-dimensional projection data. This improves the accuracy and robustness of ore internal defect detection, especially when internal defects are complex and unevenly distributed.

[0091] In a preferred embodiment, an X-ray binocular imaging device using one X-ray source and two flat panel detector arrays acquires ore projection data from different angles without the need for additional X-ray sources. This simplifies the hardware structure, makes it suitable for different ore sizes and shapes, and has strong adaptability to the input direction and position of the ore. While reducing the hardware complexity and operating cost of the X-ray intelligent ore beneficiation machine, it can generate high-resolution three-dimensional ore images and is a key component of the X-ray intelligent ore beneficiation machine.

[0092] Acquiring three-dimensional images of the ore using an X-ray binocular imaging device is the starting point of the entire process, determining the accuracy of subsequent internal defect characterization and intelligent sorting results. The three-dimensional images of the ore can completely describe the spatial distribution of internal defects, enabling the internal defect characterization model to extract features more accurately and classify different internal defect levels, thereby improving the accuracy of quantitative characterization of internal defects and the accuracy of intelligent sorting of different internal defect levels.

[0093] In particular, the step of performing feature processing on the ore three-dimensional image to obtain an ore internal defect dataset is:

[0094] Step S11: normalizing, image denoising and voxel data enhancement on the ore three-dimensional image;

[0095] Step S12: edge positioning on the ore internal defect by using an edge detection algorithm to obtain edge data of the defect;

[0096] Step S13: segmentation on the defect region by using a region growing algorithm to obtain geometric data of the defect;

[0097] Step S14: labeling the processed ore three-dimensional image internal defect to obtain an ore internal defect dataset;

[0098] The ore is a particle size ore obtained after crushing and vibrating screening of raw ore; the raw ore is a concentrate containing only one kind of ore after pre-selection and waste rejection.

[0099] S2, inputting the ore three-dimensional image into an ore internal defect representation model to obtain ore internal defect representation data and dividing ore internal defect grades; the ore internal defect representation model is trained by the ore internal defect dataset;

[0100] In particular, referring to Figure 2 , the step of inputting the ore three-dimensional image into the ore internal defect representation model to obtain ore internal defect representation data and dividing ore internal defect grades is:

[0101] Step S21: dividing the ore internal defect dataset into a training set and a validation set, inputting into a 3D CNN network model for training to obtain an ore internal defect representation model;

[0102] The training set and the validation set are in a ratio of 7:3; the 3D-CNN is composed of an input layer, a convolution layer, a pooling layer, a full connection layer and an output layer, and the output layer outputs ore internal defect representation data;

[0103] Step S22: inputting the ore three-dimensional image into the ore internal defect representation model to obtain ore internal defect representation data;

[0104] The ore internal defect representation data includes defect classification results and internal defect volume ratios, the defect classification results include two classification results of having defects and no defects, and the internal defect volume ratio refers to a ratio of all internal defect volumes in the ore to the ore volume;

[0105] Step S23: dividing ore internal defect grades according to the internal defect volume ratio;

[0106] The method for dividing the internal defect grade of the ore is:

[0107] A first threshold X1 and a second threshold X2 are set;

[0108] If the internal defect volume ratio is greater than and equal to the first threshold X1, the internal defect grade of the ore is A grade;

[0109] If the internal defect volume ratio is greater than the second threshold X2 and less than the first threshold X1, the internal defect grade of the ore is B grade;

[0110] If the internal defect volume ratio is less than and equal to the second threshold X2, the internal defect grade of the ore is C grade.

[0111] By dividing the internal defect data set of the ore into a training set and a validation set and inputting them into a 3D-CNN network model for training, an internal defect characterization model of the ore is obtained, which can effectively learn and extract the feature representation of the internal defect of the ore, so as to realize accurate characterization of the internal defect of the ore. The three-dimensional convolutional neural network 3D-CNN can effectively process the three-dimensional image data of the ore, capture the spatial information and defect structure features of the ore in the image, adapt to different types and scales of ore data, and include the volume information of the internal defect of the ore in the model training and characterization process. Through multi-layer convolution and pooling operations, high-level feature representation in the data is learned and extracted, so as to improve the recognition and characterization ability of the model for the internal defect of the ore.

[0112] According to the internal defect characterization data obtained by the internal defect characterization model of the ore, including the defect classification result and the internal defect volume ratio data, the internal defect grade of the ore is divided according to the preset thresholds X1 and X2. The grade division method based on the internal defect volume ratio can objectively evaluate the internal defect of the ore. The ore is classified and processed according to different internal defect grades, which is beneficial to take different sorting strategies for different grades of ore, improve the sorting efficiency and accuracy. At the same time, the division of the internal defect grade is also helpful for monitoring and controlling the quality problems in the production process of the ore, providing quantitative internal defect characterization and grade evaluation for the overall scheme, which is beneficial to improve the intelligent level of the ore sorting system and the production management efficiency.

[0113] S3, obtaining current X-ray intelligent ore sorting machine setting parameters, and constructing an expert experience sequence using the internal defect characterization data of the ore and the current X-ray intelligent ore sorting machine setting parameters;

[0114] In particular, the current X-ray intelligent ore sorting machine setting parameters are obtained, and the expert experience sequence is constructed using the internal defect characterization data of the ore and the current X-ray intelligent ore sorting machine setting parameters, wherein:

[0115] The current X-ray intelligent ore dressing machine setting parameters include feeding belt speed, machine body belt speed, vibration amplitude, vibration frequency, vibration direction angle, and nozzle injection pressure.

[0116] The expert experience sequence includes an ore internal defect state set, an intelligent ore dressing machine parameter action set, a reward value set, and a next state set.

[0117] The ore internal defect state set includes defect classification result data and internal defect volume ratio data.

[0118] The intelligent ore dressing machine parameter action set includes feeding belt speed data, machine body belt speed data, vibration amplitude data, vibration frequency data, vibration direction angle data, and nozzle injection pressure data.

[0119] In a preferred embodiment, the ore internal defect characterization data and the intelligent ore dressing machine setting parameters are combined to construct an expert experience sequence, realizing comprehensive analysis and modeling of the ore internal defect state and the ore dressing machine parameters. As an experience model, the expert experience sequence integrates ore internal defect data and intelligent ore dressing machine parameter data, can provide data-driven decision support for subsequent multi-agent reinforcement learning of the X-ray intelligent ore dressing machine, strengthen the coupling relationship between ore quality and machine parameters, and improve the intelligent level of the ore separation system and the production management efficiency.

[0120] S4, store the expert experience sequence into an expert experience pool, train a multi-agent reinforcement learning model to obtain an ore internal defect grade separation parameter model;

[0121] In particular, with reference to Figure 3 , the step of storing the expert experience sequence into the expert experience pool and training the multi-agent reinforcement learning model to obtain the ore internal defect grade separation parameter model is:

[0122] Step S31: construct an expert experience sequence using the ore internal defect characterization data and the current X-ray intelligent ore dressing machine setting parameters ;

[0123] The is an ore internal defect state set, is an intelligent ore dressing machine parameter action set, is a separation accuracy reward value set, is a next time ore internal defect state set.

[0124] Step S32: store the expert experience sequence into an expert experience pool to complete expert experience pool initialization;

[0125] Step S33: training the multi-agent reinforcement learning model by using the expert experience pool to obtain the internal defect grade separation parameter model of the ore.

[0126] Further, referring to Figure 3 , the step of storing the expert experience sequence into the expert experience pool, training the multi-agent reinforcement learning model to obtain the internal defect grade separation parameter model of the ore is:

[0127] Step S41: constructing a prediction network and a target network;

[0128] Step S42: constructing N agents based on the prediction network, the target network and the experience pool, wherein is a variable and ;

[0129] Step S43: initializing the experience pool as the expert experience pool; the capacity of the experience pool is M;

[0130] Step S44: initializing the prediction network by using random prediction network weights , initializing the target network by using random target network weights , initializing the parameter state set , the batch n and the step m;

[0131] Step S45: obtaining the expert experience sequence from the experience pool by using the random batch n for training, wherein ;

[0132] The prediction network inputs the internal defect state set of the ore and the intelligent ore dressing machine parameter action set , and outputs the predicted value function and the intelligent ore dressing machine parameter action set of the next moment;

[0133] The target network inputs the internal defect state set of the next moment, and outputs the target value function , The calculation formula is:

[0134] ;

[0135] In the formula, is the operation effect reward value of the intelligent ore dressing machine parameter action set when the internal defect state set of the ore is is a discount factor and , is the internal defect state set of the ore of the next momentThe next time intelligent ore dressing machine parameter action set is selected The maximum prediction value function is obtained

[0136] Step S46: The weight parameters of the prediction network are copied to the target network every m steps The weight parameters of the target network are copied to the target network ;

[0137] Step S47: The joint prediction value function of the multi-agent is calculated The calculation formula is:

[0138] ;

[0139] The joint target value function of the multi-agent is calculated The calculation formula is:

[0140] ;

[0141] In the formula, The function and the g function are calculated by the BP neural network model

[0142] The multi-agent loss function is updated to The calculation formula is:

[0143] ;

[0144] Step S48: The gradient descent method is used to minimize the MSE loss function The network model parameters and are updated until convergence, and the internal defect grade separation parameter model of the ore is obtained

[0145] Further, the network structures of the prediction network and the target network are the same, and both adopt a BP neural network as the basic network structure

[0146] The BP neural network is composed of an input layer, a hidden layer, and an output layer

[0147] The number of neurons in the input layer is equal to the number of internal defect state sets of the ore

[0148] The number of neurons in the output layer is equal to the number of actions in the intelligent ore dressing machine parameter action set

[0149] By storing the expert experience sequence into the expert experience pool and training the model using the expert experience pool, the expert experience pool serves as the initial transfer learning weight of the multi-agent reinforcement learning model, which can improve the training efficiency of the model. The expert experience sequence contains the information of the initial internal defect state of the ore and the action of the intelligent beneficiation machine parameters, which helps the model better understand the relationship between the internal defect state of the ore and the intelligent beneficiation machine parameters, and improves the generalization ability of the model. The multi-agent reinforcement learning model is used to obtain the internal defect grade separation parameter model of the ore, which realizes the optimization of the intelligent beneficiation machine parameters for different internal defect states of the ore. Through the training model, the best parameter setting suitable for the current internal defect state of the ore can be obtained, which improves the efficiency and accuracy of ore separation and realizes the optimization of X-ray intelligent beneficiation machine parameters, thereby providing beneficial contributions to the intelligentization of the ore internal defect separation system.

[0150] S5, conveying the ore to be screened to the X-ray intelligent beneficiation machine, based on the internal defect characterization model of the ore and the internal defect grade separation parameter optimization model of the ore, and separating different internal defect grade ores.

[0151] In a preferred embodiment, the X-ray intelligent beneficiation machine is started, the ore to be screened is sent into the X-ray binocular imaging device area through the conveying belt, and the internal part of the ore is scanned with high precision by using the X-ray three-dimensional imaging technology. The imaging data is transmitted to the data processing unit in real time, the internal defect characterization model of the ore quickly analyzes the data, and identifies the internal defect type and internal defect grade of the ore. The internal defect grade separation parameter optimization model of the ore dynamically adjusts the separation parameters of the beneficiation machine according to the analysis results. Finally, different defect grade ores are collected into designated containers, and the whole separation process is completed. Through this series of steps, the accurate identification and efficient separation of the internal defects of the ore are realized, the X-ray intelligent beneficiation machine parameters for processing different internal defects of the ore are dynamically and adaptively adjusted, the parameter setting is customized and updated, and the separation efficiency and accuracy of different internal defect grades of the ore are improved, thereby improving the resource utilization rate and reducing the energy consumption and cost in the beneficiation process.

[0152] Embodiment 2

[0153] Figure 4 It is an X-ray three-dimensional imaging and defect quantitative characterization system for internal defects of an ore in an X-ray intelligent beneficiation machine provided by an embodiment of the present application. The system as a whole can be divided into:

[0154] The ore image processing module is used to obtain the three-dimensional image of the ore by using the X-ray binocular imaging device; the three-dimensional image of the ore is processed to obtain the internal defect data set of the ore;

[0155] Ore defect characterization module: used to input 3D images of ore into the ore internal defect characterization model to obtain ore internal defect characterization data and classify the ore internal defect levels; the ore internal defect characterization model is trained from the ore internal defect dataset.

[0156] Mining machine parameter acquisition module: used to acquire the current X-ray intelligent mineral processing machine setting parameters, and to construct an expert experience sequence using the internal defect characterization data of the ore and the current X-ray intelligent mineral processing machine setting parameters;

[0157] Mining machine parameter optimization module: used to store expert experience sequences into the expert experience pool and train a multi-agent reinforcement learning model to obtain the ore internal defect level sorting parameter model;

[0158] Intelligent ore sorting module: It is used to transport the ore to be sorted to the X-ray intelligent ore beneficiation machine, and sort the ore into different grades of internal defects based on the ore internal defect characterization model and the ore internal defect level sorting parameter optimization model.

[0159] Example 3

[0160] Figure 5 This is a schematic diagram of an electronic device structure provided in one embodiment of this application. Figure 5 As shown, according to another aspect of this application, an electronic device 100 is also provided. The electronic device 100 may include one or more processors and one or more memories. The memory stores computer-readable code, which, when executed by the one or more processors, can perform the X-ray three-dimensional imaging and quantitative characterization method for internal defects of ore in an X-ray intelligent mineral processing machine as described above.

[0161] The method or apparatus according to the embodiments of this application can also be used by means of Figure 5 The architecture of the electronic device shown is used to implement this. For example... Figure 5As shown, the electronic device 100 can include a bus 101, one or more CPUs 102, a read-only memory (ROM) 103, a random access memory (RAM) 104, a communication port connected to a network 105, an input / output component 106, a hard disk 107, etc. The storage device in the electronic device 100, such as the ROM 103 or the hard disk 107, can store the X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in the X-ray intelligent ore dressing machine provided in the present application. The X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in the X-ray intelligent ore dressing machine may, for example, include the following steps: acquiring a three-dimensional image of the ore by using an X-ray binocular imaging device; performing feature processing on the three-dimensional image of the ore to obtain an internal defect data set of the ore; inputting the three-dimensional image of the ore into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and divide the internal defect grade of the ore; the internal defect characterization model of the ore is obtained by training the internal defect data set of the ore; acquiring current X-ray intelligent ore dressing machine setting parameters, constructing an expert experience sequence by using the internal defect characterization data of the ore and the current X-ray intelligent ore dressing machine setting parameters; storing the expert experience sequence into an expert experience pool, training a multi-agent reinforcement learning model to obtain an internal defect grade separation parameter model of the ore; conveying the ore to be screened to the X-ray intelligent ore dressing machine, and based on the internal defect characterization model of the ore and the internal defect grade separation parameter optimization model, the ore with different internal defect grades is separated.

[0162] Further, the electronic device 100 can further include a user interface 108. Of course, Figure 5 The architecture shown is only exemplary, and when implementing different devices, according to actual needs, some of the components in the electronic device shown can be omitted Figure 6 One or more components in the electronic device shown.

[0163] Embodiment 4

[0164] Figure 6 is a computer readable storage medium structure provided by an embodiment of the present application. As ​ As shown, it is a computer readable storage medium 200 according to an embodiment of the present application. The computer readable storage medium 200 stores computer readable instructions. When the computer readable instructions are run by a processor, the X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in the X-ray intelligent ore dressing machine according to the embodiment of the present application described with reference to the above figures can be executed. The computer readable storage medium 200 includes but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and cache memory, etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0165] In addition, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions executable by a processor to perform instructions corresponding to the method steps provided by the present application, which perform the above-mentioned functions defined in the method of the present application when the computer program is executed by a central processing unit (CPU).

[0166] The method and device, apparatus of the present application can be implemented in many ways. For example, the method and device, apparatus of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, firmware. The above-mentioned order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the above specifically described order, unless otherwise specifically described. In addition, in some embodiments, the present application can also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present application. Thus, the present application also covers the recording medium storing the program for executing the method according to the present application.

[0167] In addition, the part of the above technical solutions provided in the embodiments of the present application that is consistent with the implementation principle of the corresponding technical solutions in the prior art is not described in detail to avoid excessive repetition.

[0168] The specific embodiments described above further illustrate the purposes, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0169] The above-mentioned preset parameters or preset thresholds are set by a person skilled in the art according to actual conditions or obtained by a large amount of data simulation.

[0170] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore sorter, characterized in that, The method comprises the following steps: S1, obtaining a three-dimensional image of the ore by using an X-ray binocular imaging device; performing feature processing on the three-dimensional image of the ore to obtain an internal defect data set of the ore; S2, inputting the three-dimensional image of the ore into an internal defect representation model of the ore to obtain internal defect representation data of the ore and divide the internal defect grade of the ore; the internal defect representation model of the ore is obtained by training the internal defect data set of the ore; S3, obtaining current X-ray intelligent ore dressing machine setting parameters, and constructing an expert experience sequence by using the internal defect representation data of the ore and the current X-ray intelligent ore dressing machine setting parameters; S4, storing the expert experience sequence into an expert experience pool, and training a multi-agent reinforcement learning model to obtain an internal defect grade separation parameter model of the ore; S5, conveying the ore to be screened to an X-ray intelligent ore dressing machine, and separating different internal defect grade ores based on the internal defect representation model of the ore and the internal defect grade separation parameter optimization model of the ore; The internal defect representation data of the ore includes a defect classification result and an internal defect volume ratio, the defect classification result includes two classification results of having defects and having no defects, and the internal defect volume ratio refers to a ratio of all internal defect volumes in the ore to a volume of the ore; The internal defect grade of the ore is divided according to the internal defect volume ratio; The method for dividing the internal defect grade of the ore is: a first threshold X1 and a second threshold X2 are set; if the internal defect volume ratio is greater than and equal to the first threshold X1, the internal defect grade of the ore is A grade; if the internal defect volume ratio is greater than the second threshold X2 and less than the first threshold X1, the internal defect grade of the ore is B grade; if the internal defect volume ratio is less than and equal to the second threshold X2, the internal defect grade of the ore is C grade; The current X-ray intelligent ore dressing machine setting parameters include a feeding belt speed, a machine body belt speed, a vibration amplitude, a vibration frequency, a vibration direction angle, and a nozzle injection pressure; The step of training the multi-agent reinforcement learning model to obtain the internal defect grade separation parameter model of the ore is: Step S31: constructing an expert experience sequence by using the ore internal defect characterization data and the current X-ray intelligent ore dressing machine setting parameters ; The is a set of internal defect states of the ore, is a set of intelligent beneficiation machine parameter actions, is a set of sorting accuracy reward values, is a set of internal defect states of the ore at the next moment. Step S32: store the expert experience sequence to the expert experience pool, and complete the initialization of the expert experience pool. Step S33: training the multi-agent reinforcement learning model by using the expert experience pool to obtain the internal defect grade separation parameter model of the ore; The step of training the multi-agent reinforcement learning model to obtain the internal defect grade separation parameter model of the ore is: Step S41: constructing a prediction network and a target network; Step S42: constructing N agents based on the prediction network, the target network and the experience pool, wherein are variables and ; Step S43: initializing the experience pool as the expert experience pool; the capacity of the experience pool is M; Step S44: Use random prediction network weights Initialize the prediction network using random target network weights Initialize the target network, initialize the parameter state set , batch n and step m; Step S45: Random batch n obtains expert experience sequence from experience pool Training is performed, wherein ; A set of internal ore defect states is predicted as an input to the network and a set of intelligent beneficiation machine parameter actions , an output prediction value function and a set of intelligent beneficiation machine parameter actions for the next time ; a set of internal defect states of the ore at the next time under the target network input , output target value function , The calculation formula is: ； wherein is the operating effect reward value of the set of intelligent beneficiation machine parameter actions at the set of ore internal defect states , is the discount factor and , is the maximum predicted value function obtained by selecting the set of intelligent beneficiation machine parameter actions at the set of ore internal defect states at the next time instant Step S46: passing the weight parameters of the prediction network through m steps copying the weight parameters to the target network ; Step S47: Calculate the joint prediction value function of the multi-agent The calculation formula is: ； Computing a joint objective value function for multiple agents , the formula is: ； In the formula, The functions f and g are calculated by a BP neural network model. The multi-agent loss function is updated as The calculation formula is: ； Step S48: minimizing the MSE loss function using gradient descent method , updating the network model parameters and until convergence, obtaining the internal defect grade separation parameter model of the ore The network structures of the prediction network and the target network are the same, and both adopt a BP neural network as a basic network structure; the BP neural network comprises an input layer, a hidden layer, and an output layer; The number of neurons of the input layer is equal to the number of internal defect state sets of the ore; The number of neurons of the output layer is equal to the number of actions in the intelligent ore dressing machine parameter action set.

2. The method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore concentrator according to claim 1, characterized in that, The X-ray binocular imaging device comprises one X-ray source and two flat panel detector arrays.

3. The method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore concentrator according to claim 2, characterized in that, The step of performing feature processing on the three-dimensional image of the ore to obtain the internal defect data set of the ore is: Step S11: performing normalization, image denoising, and voxel data enhancement on the three-dimensional image of the ore; Step S12: edge positioning is performed on the internal defects of the ore by using an edge detection algorithm to obtain edge data of the defects; Step S13: the region growing algorithm is used to segment the defect region to obtain geometric data of the defects; Step S14: the internal defects of the processed three-dimensional image of the ore are labeled to obtain an internal defect data set of the ore; The ore is a particle size ore obtained after crushing and vibrating screening of raw ore; the raw ore is a concentrate containing only one kind of ore after pre-selection and waste rejection.

4. The method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore concentrator according to claim 3, characterized in that, The step of inputting the three-dimensional image of the ore into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and dividing the internal defect grade of the ore is: Step S21: the internal defect data set of the ore is divided into a training set and a validation set, and is input into a 3D-CNN network model for training to obtain an internal defect characterization model of the ore; The training set and the validation set are in a ratio of 7:3; the 3D-CNN is composed of an input layer, a convolution layer, a pooling layer, a full connection layer, and an output layer, and the output layer outputs internal defect characterization data of the ore; Step S22: inputting the three-dimensional image of the ore into the internal defect characterization model of the ore to obtain internal defect characterization data of the ore.

5. The method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore concentrator according to claim 4, characterized in that, The current X-ray intelligent ore dressing machine setting parameters are obtained, and the internal defect characterization data of the ore and the current X-ray intelligent ore dressing machine setting parameters are used to construct an expert experience sequence, wherein: The expert experience sequence includes an internal defect state set, an intelligent ore dressing machine parameter action set, a reward value set, and a next state set; The internal defect state set includes defect classification result data and internal defect volume ratio data; The intelligent ore dressing machine parameter action set includes feeding belt speed data, machine body belt speed data, vibration amplitude data, vibration frequency data, vibration direction angle data, and nozzle injection pressure data.

6. A system for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore sorter, based on the method for X-ray three-dimensional imaging and quantitative characterization of internal defects of ores in an X-ray intelligent ore sorter according to any one of claims 1-5, characterized in that, The system comprises: An ore image processing module: used for obtaining a three-dimensional image of an ore by using an X-ray binocular imaging device; and performing feature processing on the three-dimensional image of the ore to obtain an internal defect data set of the ore; An ore defect characterization module: used for inputting the three-dimensional image of the ore into an internal defect characterization model of the ore to obtain internal defect characterization data of the ore and divide the internal defect grade of the ore; the internal defect characterization model of the ore is trained from the internal defect data set of the ore; An ore machine parameter acquisition module: used for obtaining current X-ray intelligent ore dressing machine setting parameters, and constructing an expert experience sequence by using the internal defect characterization data of the ore and the current X-ray intelligent ore dressing machine setting parameters; An ore machine parameter optimization module: used for storing the expert experience sequence into an expert experience pool, and training a multi-agent reinforcement learning model to obtain an internal defect grade separation parameter model of the ore; An ore intelligent separation module: used for conveying a to-be-screened ore to an X-ray intelligent ore dressing machine, and separating different internal defect grade ores based on the internal defect characterization model of the ore and the internal defect grade separation parameter optimization model of the ore.

7. An electronic device, comprising: The computer program product comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the processor executes the program to implement the steps in the X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in an X-ray intelligent ore dressing machine according to any one of claims 1-5.

8. A readable storage medium, characterized by, The readable storage medium stores a computer program, and the computer program is suitable for being loaded by the processor to execute the steps in the X-ray three-dimensional imaging and quantitative characterization method of internal defects of ores in an X-ray intelligent ore dressing machine according to any one of claims 1-5.

Images