An intelligent sorting system and method for soft and hard gangue in coal gangue

Abstract

The present application provides a system and method for intelligent sorting of soft and hard gangue in coal gangue, belonging to the field of coal gangue sorting; it solves the problem that the current coal gangue classification, utilization and disposal have a low degree of intelligence and are highly dependent on manual labor; the system includes an image acquisition and processing module: used to acquire dual-energy X-ray images of coal gangue and convert the images into data features; a soft and hard coal gangue evaluation label establishment module: used to establish coal gangue classification standards and evaluation labels based on point load test data of soft and hard coal gangue; an intelligent recognition and classification algorithm module: used to extract key features from data features to form feature vectors, and fuse the feature vectors with point load test data to form comprehensive feature vectors, use the comprehensive feature vectors as input features of the classification algorithm model, use the coal gangue evaluation labels as classification labels, train the classification algorithm model, and output classification recognition results; a control and sorting execution module; the present application is applied to the sorting of soft and hard coal gangue.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent sorting of coal gangue, and in particular to a system and method for intelligent sorting of soft and hard gangue in coal gangue. Background Art

[0002] Gangue is a solid waste generated during coal mining and washing. Its high production volume, wide variation in properties, and significant environmental hazards make its disposal a major challenge for the national coal industry. However, gangue actually has enormous potential for utilization, and its classification and tiered utilization and disposal are essential.

[0003] Current research on intelligent sorting technology primarily focuses on the identification and sorting of coal and gangue, leaving a significant technological gap in the field of fine-grained gangue identification and sorting. Traditional gangue sorting methods rely primarily on manual labor and simple mechanical equipment, resulting in low efficiency, poor sorting accuracy, resource waste, and a high reliance on manual experience. With the advancement of artificial intelligence and automation technologies, there is an urgent need for intelligent methods for fine-grained gangue identification and sorting to improve sorting efficiency and accuracy. Summary of the Invention

[0004] In order to solve the current problem of low intelligence level and high reliance on manual labor in the classification, utilization and disposal of coal gangue, this application proposes an intelligent sorting system and method for soft and hard gangue in coal gangue, which uses advanced artificial intelligence technology and automation equipment to achieve efficient and accurate soft and hard gangue sorting.

[0005] The technical solution adopted in this application is: an intelligent sorting system for soft and hard gangue in coal gangue, comprising:

[0006] Image acquisition and processing module: used to acquire dual-energy X-ray images of coal gangue and convert the images into data features;

[0007] Soft and hard coal gangue evaluation label establishment module: used to establish coal gangue classification standards and evaluation labels based on the point load test data of soft and hard coal gangue;

[0008] Intelligent recognition and classification algorithm module: used to extract key features from data features to form feature vectors, and fuse the feature vectors with point load test data to form comprehensive feature vectors. The comprehensive feature vectors are input into the classification algorithm model as input features, and the gangue evaluation labels are used as classification labels to train the classification algorithm model and output classification recognition results.

[0009] Control and sorting execution module: used to generate sorting instructions based on the classification and recognition results, and separate soft and hard coal gangue through the sorting execution device.

[0010] Furthermore, the image acquisition and processing module includes a dual-energy X-ray source, a ray signal detector and an image processing unit. The dual-energy X-ray source is used to generate high and low energy rays to perform penetrating scanning on coal gangue to obtain its internal structure information and material properties; the ray signal detector is used to receive and record X-ray signals passing through coal gangue to generate digital images; the image processing unit denoises, enhances and extracts features from the collected images, and obtains data features of X-ray images in high-energy and low-energy areas through theoretical calculations.

[0011] Furthermore, the soft and hard gangue evaluation label establishment module includes a strength testing device and a label database. The strength testing device is used to test the point load strength of the gangue and convert the point load test data of the gangue into the uniaxial compressive strength of the rock. The label database is established based on the uniaxial compressive strength of the gangue.

[0012] Furthermore, the intelligent recognition and classification algorithm module includes a feature data extraction unit and a classification model training unit. The feature data extraction unit extracts key features from the image to form a feature vector. The classification model training unit adopts a machine learning algorithm and uses the pre-trained deep transfer learning model ResNet to fuse the data features of the dual-energy X-ray image and the point load test data to form a comprehensive feature vector as the input feature, and uses the soft and hard coal gangue evaluation labels as the classification labels. A classification algorithm model is established for recognition and sorting, and then the model is trained through machine learning to output the classification recognition results.

[0013] Furthermore, the classification algorithm model adopts a random forest classification model based on the PSO particle swarm optimization algorithm.

[0014] Furthermore, the control and sorting execution module includes a control system and a sorting execution device. The control system generates a sorting instruction according to the classification and recognition result and sends it to the sorting execution device.

[0015] Furthermore, the sorting execution device includes a gas pushing device and a high-pressure air pump. The gas pushing device separates the soft and hard coal gangue by controlling the gas pushing intensity.

[0016] Furthermore, the calculation principle of the classification algorithm model for identification and classification is as follows:

[0017] X d =f R (I m θ R );

[0018] X c =[X h ,X d ];

[0019] X f =[Xc ,P];

[0020]

[0021] Where: X d The high-level features output by the ResNet learning model, f R Learning model for ResNet, I m is the image of gangue obtained by dual-energy X-ray, θ R is the model parameter of ResNet deep transfer learning, X c X is the feature after the two features are concatenated. h To calculate the extracted dual-energy X-ray image features, X f is the comprehensive characteristic vector, P is the point load test data, is the classification result, f s is the random forest model, θ s is the model parameter of the random forest model, y is the true label, L is the loss function, is the hyperparameter obtained after the PSO particle swarm optimization algorithm.

[0022] Furthermore, the calculation formula of data features is as follows:

[0023]

[0024]

[0025] T = μρh;

[0026]

[0027] Where: I l0 is the grayscale mean of the empty conveyor belt image in the low-energy region where X-rays do not transmit the gangue, I l is the grayscale mean of the image after low-energy X-ray transmission through coal gangue; I h0 is the grayscale mean of the image of the empty conveyor belt in the high-energy region where X-rays do not transmit the gangue, I h is the grayscale mean of the image after high-energy X-ray transmission through coal gangue, μ l is the mass attenuation coefficient of coal gangue in low energy zone; μ h is the mass attenuation coefficient of the gangue in the high-energy zone, h is the effective thickness of the transmitted gangue, ρ is the density of the transmitted gangue, R is the material property value, and T is the characteristic data including the thickness and density properties of the gangue.

[0028] A method for intelligently sorting soft and hard gangue in coal gangue, using the intelligent sorting system for soft and hard gangue in coal gangue, comprises the following steps:

[0029] Step 1: Collect and load dual-energy X-ray image data and point load test data;

[0030] Step 2: Select the pre-trained deep transfer learning model ResNet, fuse the feature items of dual-energy X-ray image data and point load test data to form a comprehensive feature vector, and use the comprehensive feature vector as the input feature;

[0031] Step 3: Establish evaluation criteria through point load strength testing and calculation to obtain classification labels for the classification algorithm model;

[0032] Step 4: Training the classification algorithm model: Use dual-energy X-ray image data and point load test data to train the classification algorithm model;

[0033] Step 5: Input the newly acquired dual-energy X-ray image data into the trained classification algorithm model, output the classification results, and the control system generates sorting instructions based on the classification results. The sorting execution device realizes the accurate sorting of soft and hard gangue according to the sorting instructions.

[0034] The beneficial effects of this application compared to the existing technology are as follows: Based on the differences in the compressive strength of coal gangue and its utilization methods, this application proposes a method for intelligently identifying and separating soft and hard gangue in coal gangue by combining dual-energy X-ray technology with machine learning artificial intelligence algorithms. The advantages are:

[0035] (1) High sorting efficiency. It automatically distinguishes between soft and hard coal gangue through intelligent recognition algorithms, reducing manual intervention and improving sorting speed.

[0036] (2) High sorting accuracy, using dual-energy X-ray technology combined with advanced machine learning algorithms to improve the accuracy of recognition and classification.

[0037] (3) High resource utilization rate, precise separation of soft and hard coal gangue, and maximization of the utilization efficiency of coal gangue resources.

[0038] (4) High degree of automation reduces dependence on manual operation, reducing labor costs and labor intensity.

[0039] (5) Strong adaptability: the system can operate stably under different working conditions, and a sample database training model is established for different coal mines / coal preparation plants, which is suitable for a variety of coal gangue types and mining environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present application will be further described below with reference to the accompanying drawings:

[0041] Figure 1 A schematic diagram of the system structure provided in an embodiment of the present application;

[0042] Figure 2A flow chart of the method provided in the embodiment of the present application;

[0043] Figure 3 A three-dimensional schematic diagram of the system provided in an embodiment of the present application;

[0044] In the figure: 1 is the gangue feeding port, 2 is the belt, 3 is the point load tester, 4 is the dual-energy X-ray source, 5 is the radiation signal detector, 6 is the junction box, 7 is the computer workstation, 8 is the gas pushing device, 9 is the first gangue bin, 10 is the second gangue bin, and 11 is the high-pressure air pump. DETAILED DESCRIPTION

[0045] like Figures 1 to 3 As shown, the present application provides an intelligent sorting system for soft and hard gangue in coal gangue, establishes a coal gangue quality classification system based on compressive strength differences; combines dual-energy X-ray technology with machine learning algorithms to establish a classification algorithm model; applies artificial intelligence technology and automated equipment in the field of coal gangue classification and sorting, solving the problems existing in traditional manual sorting.

[0046] like Figure 1 As shown, the sorting system provided in the embodiment of the present application includes an image acquisition and processing module, a soft and hard coal gangue evaluation label establishment module, an intelligent recognition and classification algorithm module, and a control and sorting execution module, wherein the image acquisition and processing module includes hardware devices such as a dual-energy X-ray source and a ray signal detector, and software modules such as an image processing unit. The dual-energy X-ray source is used to generate high and low energy rays to perform penetrating scanning on the coal gangue to obtain its internal structure information and material properties; the ray signal detector is used to receive and record the X-ray signal passing through the coal gangue to generate a digital image; the image processing unit performs denoising, enhancement, and feature extraction on the collected image, and obtains I through theoretical calculation. l , I h , R, T l 、T h The above dual-energy X-ray image features serve as important input features for the subsequent classification algorithm model.

[0047] The calculation formula of data features is as follows:

[0048]

[0049]

[0050] T = μρh;

[0051]

[0052] Where: I l0 is the grayscale mean of the empty conveyor belt image in the low-energy region where X-rays do not transmit the gangue, Il is the grayscale mean of the image after low-energy X-ray transmission through coal gangue; I h0 is the grayscale mean of the image of the empty conveyor belt in the high-energy region where X-rays do not transmit the gangue, I h is the grayscale mean of the image after high-energy X-ray transmission through coal gangue, and I l0 , I l with I h0 , I h Substituting these two sets of quantities into the Lambert-Beer formula, we can obtain μ l is the mass attenuation coefficient of coal gangue in low energy zone; μ h is the mass attenuation coefficient of the gangue in the high-energy zone, h is the effective thickness of the transmitted gangue, ρ is the density of the transmitted gangue, R is the material property value, and T is the characteristic data including the thickness and density properties of the gangue. The T value retains the influence of the thickness h and density ρ of the gangue. Using the T value as a classification feature can take into account the influence of the thickness and density of the gangue and reduce the impact of the thickness effect.

[0053] The module for establishing labels for evaluating and labeling soft and hard gangue includes hardware equipment such as a strength testing device and software modules such as a label database for soft and hard gangue. The strength testing device is used to perform strength tests on gangue, which serves as a key basis for classifying soft and hard gangue. By measuring the point load strength of gangue samples from coal mines and coal preparation plants and calculating and converting the rock's uniaxial compressive strength, the soft and hard gangue classification standards are established with reference to GB / T50218-2014, "Engineering Rock Mass Classification Standard." By compiling sample data and establishing a label database for storing characteristic data of different gangue types, this database serves as a reference for training and identifying soft and hard gangue.

[0054] The calculation and conversion process of rock uniaxial compressive strength is as follows:

[0055]

[0056] I s(50) =FI s ;

[0057]

[0058] Where: I s is the uncorrected point load strength index, in MPa; P is the failure load, in N; D e is the equivalent core diameter, in mm; w is the width of the minimum cross section through the two loading points, in mm; D is the distance between the loading points, in mm; F ​​is the correction factor; m is the correction index, which is determined by the empirical value of similar rocks and is generally taken as 0.45; Rc is the uniaxial compressive strength of the rock, which is the measured rock point load strength index Is(50) The conversion value is in MPa.

[0059] The intelligent recognition and classification algorithm module includes a feature data extraction unit and a classification model training unit. The feature data extraction unit extracts key features from the digitized image to form a feature vector. The classification model training unit adopts a machine learning algorithm and uses the pre-trained deep transfer learning model ResNet to transform the pre-processed dual-energy X-ray image feature I l , I h , R, T l 、T h The comprehensive feature vector formed by the fusion of the point load test data features is used as the input feature, and the soft and hard coal gangue evaluation label Rc is used as the classification label. A classification algorithm model is established for identification and sorting, and then the model is trained through machine learning to output the classification and recognition results.

[0060] The classification algorithm is a random forest classification model based on the PSO particle swarm optimization algorithm, trained using dual-energy X-ray image data and point load test data. The trained model can be used to analyze real-time images to identify and classify soft and hard coal gangue.

[0061] The principles of classification algorithm model recognition and classification are as follows:

[0062] X d =f R (I m θ R );

[0063] X c =[X h ,X d ];

[0064] X f =[X c ,P];

[0065]

[0066] Where: X d The high-level features of the image extracted by the convolutional layer when the ResNet learning model processes the image, including the edge, texture, shape and structure of the image. The purpose of extracting these high-level features is to more accurately identify and classify soft and hard coal gangue. R Learning model for ResNet, I m is the image of gangue obtained by dual-energy X-ray, θ R is the model parameter of ResNet deep transfer learning, X c X is the feature after the two features are concatenated. hThe image processing unit performs denoising, enhancement and feature extraction on the collected digital image, and calculates the extracted I l , I h , R, T l 、T h These dual-energy X-ray image features, X f is the comprehensive characteristic vector, P is the point load test data, is the classification result, f s is the random forest model, θ s is the model parameter of the random forest model, y is the true label, L is the loss function, is the hyperparameter obtained after the PSO particle swarm optimization algorithm.

[0067] The control and sorting execution module includes a control system and a sorting execution device. The control system generates sorting instructions based on the classification and recognition results and sends them to the sorting execution device. The sorting execution device accurately separates the identified soft and hard coal gangue.

[0068] like Figure 2 As shown, the embodiment of the present application also proposes an intelligent sorting method for soft and hard gangue in coal gangue. Based on the above-mentioned sorting system, its main implementation steps are as follows:

[0069] Step 1: Collect and load dual-energy X-ray image data and point load test data;

[0070] Step 2: Select the pre-trained deep transfer learning model ResNet, fuse the feature items of dual-energy X-ray image data and point load test data to form a comprehensive feature vector, and use the comprehensive feature vector as the input feature;

[0071] Step 3: Establish evaluation criteria through point load strength testing and calculation to obtain classification labels for the classification algorithm model;

[0072] Step 4: Training the classification algorithm model: Use dual-energy X-ray image data and point load test data to train the classification algorithm model;

[0073] Step 5: Input the newly acquired dual-energy X-ray image data into the trained classification algorithm model, output the classification results, and the control system generates sorting instructions based on the classification results. The sorting execution device realizes the precise sorting of soft and hard coal gangue according to the sorting instructions.

[0074] like Figure 3As shown, this embodiment specifically provides a structure of a feasible sorting system, including a gangue conveying device using a belt 2 for conveying, a dual-energy X-ray source 4 is installed above the belt 2 of the gangue conveying device, and a radiation signal detector 5 is provided below the belt 2 at the position corresponding to the dual-energy X-ray source 4, a sorting execution device and a first gangue bin 9 and a second gangue bin 10 are provided at the discharge port of the gangue conveying device, the sorting execution device includes a high-pressure air pump 11 and a gas pushing device 8, the gas pushing device 8 is located at the discharge port of the gangue conveying device, and the gas pushing device 8 is connected to the high-pressure air pump 11; a computer workstation 7 and a point load tester 3 are also provided next to the gangue conveying device, and the computer workstation 7 is respectively connected to the dual-energy X-ray source 4, the radiation signal detector 5, the point load tester 3 and the gas pushing device 8 through the junction box 6. When in use, the system is set at the gangue discharge port of the gangue sorting device. The gangue coming out of the gangue sorting device directly falls into the gangue inlet 1 of the gangue conveying device. After being transported by the belt 2, X-ray images of the gangue are collected in real time. The gangue is classified according to the trained classification algorithm model. Finally, the soft and hard gangue are blown into the first gangue bin 9 and the second gangue bin 10 respectively according to the gas pushing device 8, thereby realizing the accurate distinction between soft and hard gangue.

[0075] The high-pressure air pump 11 and the gas pusher 8, in accordance with the control system's instructions, precisely separate the identified soft and hard gangue by varying the gas push intensity, placing them into the corresponding gangue bins. For example, for soft gangue with a classification label of 0, the gas push intensity is controlled to be relatively weak (the specific gas pressure can be determined based on the measured point load test data for the soft and hard gangue), pushing it into the gangue bin about 1 meter away. For hard gangue with a classification label of 1, the gas push intensity is controlled to be relatively strong, pushing it into the gangue bin about 5 meters away, thereby achieving precise sorting of soft and hard gangue.

[0076] This application also has the following advantages or innovations:

[0077] (1) Based on dual-energy X-ray technology, combined with density indicators and machine learning algorithms, it is applied to the classification and sorting of coal gangue, achieving high-precision identification and sorting of soft and hard coal gangue.

[0078] (2) Based on the differences in the compressive strength of coal gangue and point load strength tests, a reliable coal gangue classification standard and evaluation label system is established.

[0079] (3) Innovation in multi-source data fusion - Previous sorting methods relied on only a single type of data and were unable to comprehensively and accurately distinguish between soft and hard gangue. This application achieved a multi-dimensional analysis of gangue characteristics by fusing two different types of data, dual-energy X-ray image data and point load test data, significantly improving the accuracy and reliability of sorting. This multi-source data fusion method provides a new technical approach for the field of gangue sorting, effectively solving the limitations of traditional methods when faced with complex gangue characteristics.

[0080] (4) Innovation in the precision of the sorting execution device - highlighting the innovation of accurately separating soft and hard coal gangue by the difference in gas pushing intensity in the control and sorting execution module. This module can not only accurately classify soft and hard coal gangue according to the recognition results, but also accurately push coal gangue of different hardness by precisely controlling the gas pushing intensity of the high-pressure air pump and the gas pushing device. This gas pushing mechanism based on different intensities ensures that soft and hard coal gangue can be accurately separated and fall into the corresponding coal gangue bin, effectively avoiding the problem of coal gangue mixed loading caused by inaccurate sorting execution, and improving the purity and quality of sorting.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent sorting system for soft and hard gangue in coal gangue, characterized by: include: Image acquisition and processing module: used to acquire dual-energy X-ray images of coal gangue and convert the images into data features; Soft and hard coal gangue evaluation label establishment module: used to establish coal gangue classification standards and evaluation labels based on the point load test data of soft and hard coal gangue; The soft and hard gangue evaluation label establishment module includes a strength testing device and a label database. The strength testing device is used to test the point load strength of the gangue and convert the point load test data of the gangue into the uniaxial compressive strength of the rock. The label database is established based on the uniaxial compressive strength of the gangue. Intelligent recognition and classification algorithm module: used to extract key features from data features to form feature vectors, and fuse the feature vectors with point load test data to form a comprehensive feature vector, input the comprehensive feature vector as input feature into the classification algorithm model, use the coal gangue evaluation label as the classification label, train the classification algorithm model, and output the classification recognition result; the intelligent recognition and classification algorithm module includes a feature data extraction unit and a classification model training unit. The feature data extraction unit extracts key features from the image to form a feature vector. The classification model training unit adopts a machine learning algorithm and uses the pre-trained deep transfer learning model ResNet to fuse the data features of the dual-energy X-ray image and the point load test data to form a comprehensive feature vector as input feature, use the soft and hard coal gangue evaluation label as the classification label, establish a classification algorithm model for recognition and sorting, and then use machine learning to train the model and output the classification recognition result; The calculation principle of the classification algorithm model for recognition and classification is as follows: X d =f R (I m ;θ R ); X c =[X h ,X d ]; X f =[X c ,P]; Where: X d The high-level features output by the ResNet learning model, f R Learning model for ResNet, I m is the image of gangue obtained by dual-energy X-ray, θ R is the model parameter of ResNet deep transfer learning, X c X is the feature after the two features are concatenated. h To calculate the extracted dual-energy X-ray image features, X f is the comprehensive characteristic vector, P is the point load test data, is the classification result, f s is the random forest model, θ s is the model parameter of the random forest model, y is the true label, L is the loss function, is the hyperparameter obtained after the PSO particle swarm optimization algorithm; Control and sorting execution module: used to generate sorting instructions based on the classification and recognition results, and separate soft and hard coal gangue through the sorting execution device.

2. The intelligent separation system for soft and hard gangue in coal gangue according to claim 1 is characterized by: The image acquisition and processing module includes a dual-energy X-ray source, a radiation signal detector, and an image processing unit. The dual-energy X-ray source is used to generate high- and low-energy rays to perform penetrating scanning on coal gangue to obtain its internal structure information and material properties. The X-ray signal detector is used to receive and record X-ray signals passing through coal gangue and generate digital images; the image processing unit denoises, enhances and extracts features from the collected images, and obtains the data features of X-ray images in high-energy and low-energy areas through theoretical calculations.

3. The intelligent separation system for soft and hard gangue in coal gangue according to claim 1 is characterized by: The classification algorithm model adopts the random forest classification model based on the PSO particle swarm optimization algorithm.

4. The intelligent separation system for soft and hard gangue in coal gangue according to claim 1, characterized in that: The control and sorting execution module includes a control system and a sorting execution device. The control system generates a sorting instruction according to the classification and recognition results and sends it to the sorting execution device.

5. The intelligent separation system for soft and hard gangue in coal gangue according to claim 4, characterized in that: The sorting execution device includes a gas pushing device and a high-pressure air pump. The gas pushing device separates the soft and hard coal gangue by controlling the gas pushing intensity.

6. The intelligent separation system for soft and hard gangue in coal gangue according to claim 1, characterized in that: The calculation formula of data features is as follows: T = μρh; Where: I l0 is the grayscale mean of the empty conveyor belt image in the low-energy region where X-rays do not transmit the gangue, I l is the grayscale mean of the image after low-energy X-ray transmission through coal gangue; I h0 is the grayscale mean of the image of the empty conveyor belt in the high-energy region where X-rays do not transmit the gangue, I h is the grayscale mean of the image after high-energy X-ray transmission through coal gangue, μ l is the mass attenuation coefficient of coal gangue in low energy zone; μ h is the mass attenuation coefficient of the gangue in the high-energy zone, h is the effective thickness of the transmitted gangue, ρ is the density of the transmitted gangue, R is the material property value, and T is the characteristic data including the thickness and density properties of the gangue.

7. An intelligent separation method for soft and hard gangue in coal gangue, characterized by: The intelligent separation system for soft and hard gangue in coal gangue according to any one of claims 1 to 6 comprises the following steps: Step 1: Collect and load dual-energy X-ray image data and point load test data; Step 2: Select the pre-trained deep transfer learning model ResNet, fuse the feature items of dual-energy X-ray image data and point load test data to form a comprehensive feature vector, and use the comprehensive feature vector as the input feature; Step 3: Establish evaluation criteria through point load strength testing and calculation to obtain classification labels for the classification algorithm model; Step 4: Training the classification algorithm model: Use dual-energy X-ray image data and point load test data to train the classification algorithm model; Step 5: Input the newly acquired dual-energy X-ray image data into the trained classification algorithm model, output the classification results, and the control system generates sorting instructions based on the classification results. The sorting execution device realizes the accurate sorting of soft and hard gangue according to the sorting instructions.

Images