Intelligent dry separation system

Abstract

The invention relates to an intelligent dry separation system. The system is used for raw coal sorting and comprises a material distribution system, the material distribution system comprises a material distribution device, materials are arranged on the material distribution device in a single-layer mode after entering the material distribution device, the materials comprise lump coal, gangue and sundries, and the materials fall into a storage bin after passing through the tail end of the material distribution device; the recognition system is used for collecting images of the materials on the material distribution device in real time through a camera, recognizing and positioning the materials in the images based on a machine learning model, and obtaining recognition information including material position information; the control system sends out an execution signal through the identification information; and the execution system comprises a pneumatic injection matrix, the pneumatic injection matrix comprises a plurality of electromagnetic valves, and the falling gangue and sundries are injected according to the execution signals, and / or the falling lump coal is injected, so that the gangue and the sundries fall into storage bins different from the lump coal. The intelligent dry separation system adopts AI visual recognition and has the advantages of being low in cost and high in safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mines; in particular, the present application relates to an intelligent dry separation system. BACKGROUND

[0002] In the process of coal production, the coal directly mined from the coal mining area without any processing is called raw coal. The raw coal contains gangue, anchor rods, iron blocks and other impurities, among which the gangue has the characteristics of large density and small calorific value, and contains a large amount of heavy metals. Mixing in the coal will reduce the combustion efficiency, and the combustion products will aggravate environmental pollution, so the gangue in the raw coal needs to be pre-separated.

[0003] In recent years, with the development of artificial intelligence technology, some artificial intelligence coal separation technologies have been developed, such as using X-ray transmission imaging to distinguish coal and gangue. However, this technology not only has high cost, but also has great safety problems. SUMMARY

[0004] Therefore, the present application provides an intelligent dry separation system, so as to solve or at least alleviate one or more of the above problems and other problems in the prior art.

[0005] In order to achieve the foregoing purpose, the first aspect of the present application provides an intelligent dry separation system, wherein the system is used for raw coal separation, and the system comprises: a material distribution system, which comprises a material distribution device, after the material enters the material distribution device, the material realizes single-layer arrangement on the material distribution device, the material includes lump coal, gangue and sundries, and the material falls into a storage chamber after passing through the end of the material distribution device; an identification system, which collects images of the material on the material distribution device in real time through a camera, identifies and locates the material in the images based on a machine learning model, and obtains identification information including the material position information; a control system, which sends an execution signal through the identification information; an execution system, which comprises a pneumatic blowing matrix, the pneumatic blowing matrix comprises a plurality of electromagnetic valves, and the execution system blows the falling gangue and sundries and / or blows the falling lump coal according to the execution signal, the blowing changes the falling trajectory of the blown material, so that the gangue and sundries fall into a different storage chamber from the lump coal; The identification system and the control system are in communication connection, and the control system and the execution system are in communication connection.

[0006] In the system as described above, optionally, the material distribution system further comprises a vibrating screen, the vibrating screen comprising a distributor, the material entering the vibrating screen through the material distribution device, the material being arranged in a single layer on the material distribution device by the distributor.

[0007] In the system as described above, optionally, the identification process of the identification system comprises: collecting images of the material on the material distribution device in real time; extracting features of the images by a convolutional neural network to obtain feature maps, and inputting the feature maps into RPN and DT algorithms; the RPN generating proposal regions based on the feature maps and passing the proposal regions to the DT algorithm; the DT algorithm classifying and regressing the proposal regions based on the feature maps and the proposal regions, outputting detection results, the classification being used to determine the category of the material in each proposal region, the category including lump coal, gangue and sundries, and the regression being used to adjust the coordinate position parameters of the proposal regions.

[0008] In the system as described above, optionally, the training process of the identification system comprises: collecting images of the material on the material distribution device, annotating at least one hundred thousand of the images to generate a training set, training a model containing at least one million parameters, deploying the trained model on a chip for real-time identification of the material on the material distribution device.

[0009] In the system as described above, optionally, the identification information further comprises shape, size and speed information of the material, and the control system sends the execution signal to the execution system by comprehensively calculating the identification information, the control system monitoring the running state of the system, calculating the coordination between parts of the system, and sending alarm information to the staff when an abnormal situation occurs.

[0010] In the system as described above, optionally, the execution system further comprises a controller, the controller obtaining the identification information and controlling the electromagnetic valve, the electromagnetic valve supporting alternate operation, the time accuracy being 1ms, the spatial accuracy of the jet gas flow being 5mm, the thrust being controlled by controlling the jet amount, so as to control the falling trajectory and falling position of the sprayed material, the pneumatic blowing matrix providing a thrust exceeding 1000N to push the material with a particle size in the range of 30mm-700mm.

[0011] In the system as described above, optionally, the execution system uses multiple electromagnetic valves to spray the material to be sprayed as the target, controls the thrust on the target by controlling the spray airflow of the multiple electromagnetic valves, thereby controlling the falling trajectory of the target, and in combination with the control of the spray time, avoids or reduces the interference on the material outside the target, and the spray time control accuracy is 1 ms.

[0012] In the system as described above, optionally, the system further comprises: The shell is closed at the inlet and outlet, and the system comprises two layers of shells, and fireproof and soundproof cotton is arranged between the two layers of shells, An industrial dust collector is used to remove dust in the system, A camera automatic cleaning device comprises a windshield arranged outside the camera lens, and the windshield automatically cleans the lens.

[0013] In the system as described above, optionally, the system further comprises a video monitoring device, which is used for remote viewing of the working condition of the system for sorting raw coal by workers, and the video monitoring device communicates with the machine room and the general control room and supports communication with a handheld device.

[0014] In the system as described above, optionally, the running speed of the distributing device is set to be in the range of 0.5-3.0 m / s, and the material flow of the distributing device is set to be in the range of 50-400 t / h.

[0015] The intelligent dry separation system has the advantages of low cost and high safety. BRIEF DESCRIPTION OF DRAWINGS

[0016] The disclosure of the present application will be more apparent with reference to the accompanying drawings. It should be understood that these drawings are only for illustrative purposes, and are not intended to limit the scope of protection of the present application. In the drawings: Figure 1 A schematic diagram of an embodiment of the intelligent dry separation system of the present application; Figure 2 A schematic diagram of the main device and external supporting device of an embodiment of the intelligent dry separation system of the present application; Figure 3 A schematic block diagram of the recognition system of an embodiment of the intelligent dry separation system of the present application; Figure 4 A schematic diagram of the relationship between the recognition accuracy of the recognition system of some embodiments of the intelligent dry separation system of the present application and the number of training pictures; Figure 5 A control system schematic diagram of an embodiment of the intelligent dry separation system of the present application. DETAILED DESCRIPTION

[0017] The structure, composition, features and advantages of the intelligent dry separation system of the present application will be described below in an exemplary manner with reference to the accompanying drawings and specific embodiments, however, all the descriptions shall not be used to form any limitation on the present application.

[0018] In addition, for any single technical feature described or implied in the embodiments mentioned herein, or any single technical feature shown or implied in the drawings, the present application still allows any combination or deletion to be continued between these technical features (or their equivalents) without any technical obstacles, so it should be considered that more embodiments according to the present application are also within the scope of the description herein.

[0019] It should also be noted that the terms "left side", "right side", "exterior" and the like indicate the orientation or positional relationship based on the orientation or positional relationship of the components of the intelligent dry separation system shown in the drawings, which are only for the convenience of describing the present disclosure and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present disclosure.

[0020] In addition, the terms "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implying the number of technical features indicated. Therefore, the features defined with "first" and "second" can explicitly or implicitly include at least one of the features.

[0021] Figure 1 Schematic diagram of an embodiment of the intelligent dry separation system of the present application.

[0022] Figure 1 The raw coal bin, crusher, vibrating screen, distributor, industrial lighting device, high-definition camera, intelligent identification control module, pneumatic separation equipment, and clean coal and block gangue after separation are shown, Figure 1 The distribution device for transporting materials and two storage chambers are also included.

[0023] As Figure 1 shown, in this embodiment, the system includes a distribution system, an identification system, a control system and an execution system, wherein the distribution system includes a vibrating screen and a distributor, the identification system and the control system are integrated in an intelligent identification control module, and the execution system includes pneumatic separation equipment. The system also includes an industrial lighting device, a high-definition camera and two storage chambers for temporarily storing the block coal (clean coal) and gangue (block gangue) after separation, respectively. The system also includes dust removal, air supply and electric control auxiliary systems, which are not shown. Figure 1 Figure 1

[0024] ​​The embodiment utilizes the difference in apparent characteristics of coal and gangue, uses a high-performance computing system, a high-precision identification sensor, and a DT algorithm to collect and identify image information of the gangue and lump coal on the material distribution device in real time, extracts, models, stores, and learns the characteristics of lump coal and gangue based on machine learning, neural network, and deep learning technology to realize identification and classification of lump coal and gangue, and transmits the processed information such as shape, size, position, and speed of the lump coal and gangue to the control system through a data interface within milliseconds, and then uses an execution system to quickly and accurately separate.

[0025] The process flow of the embodiment includes: the material enters the belt conveyor distribution device through the feeding device, the material is arranged in a single layer on the distribution device, the high-speed camera above the distribution device collects images in real time, the identification system identifies whether the material is lump coal or gangue, and the control system transmits the position information to the electric control system, the electric control system opens the electromagnetic valve at the corresponding position according to the position of the executed target to complete the injection. The gangue and lump coal fall into different storage chambers, and the separation of the material is completed. The storage chamber can be replaced by a chute.

[0026] In the embodiment, the material to be separated (raw coal) is provided to the system through a raw coal bin and a crusher. The raw coal bin is a facility for storing and transporting unprocessed raw coal, which can include some auxiliary functions such as preliminary screening and drying of coal. The crusher is responsible for breaking large pieces of raw coal into smaller particles. The raw coal bin and the crusher work together to provide a stable coal flow from the raw coal bin and break the coal into the appropriate size for the system, improving the separation efficiency and accuracy.

[0027] In the embodiment, the raw coal is screened by a vibrating screen to remove coal particles smaller than 30 mm in diameter, and the raw coal on the distribution device (such as a transport belt) is evenly laid out by a distributor. In the embodiment, the vibrating screen has a screen hole diameter of 30 mm to pre-screen coal particles smaller than 30 mm in diameter; in other embodiments, the vibrating screen hole diameter can be adjusted according to actual needs. In the embodiment, the distribution device is inclined as shown in the distribution device, in other alternative embodiments, the distribution device can be horizontal, or have a different inclination angle than the present embodiment. Figure 1

[0028] The embodiment uses a high-definition camera and an industrial lighting device to stably provide high-definition video or image data for the identification system. An industrial ultra-high-resolution and ultra-high-speed camera is used to obtain visual information of the raw coal on the distribution device for training the model of the identification system or real-time identification, which is beneficial to improve the identification accuracy and speed, and thus improve the amount of raw coal separated per unit time. The industrial lighting device provides a stable and suitable lighting environment for camera shooting, which is beneficial to improve the identification accuracy and stability.

[0029] ​Further, in order to ensure the image clarity, the system can further comprise an industrial camera automatic cleaning device: a professional windshield is arranged outside the camera lens, which can automatically clean the lens and further ensure the image source clarity.

[0030] In this embodiment, the intelligent identification control module integrates the identification system and the control system. The identification system has high-precision visual identification capability and identifies lump coal and gangue based on a deep learning model and accurately locates. The control system comprehensively calculates position information, size information, weight information, etc., and controls the spraying operation of the execution system.

[0031] Specifically, the above-mentioned identification system identifies based on a pre-trained model and can intelligently and autonomously learn in the identification work, continuously improves the identification accuracy for the independent coal quality of each coal mine, and as the amount of processed data increases, the visual identification accuracy will be higher and higher. In this embodiment, when training the model, photos of lump coal and gangue are manually labeled to establish a training data set with a total of 100,000 photos or more, and in other optional embodiments, the number of labeled photos reaches the order of magnitude of millions; then, the model containing millions of parameters is trained by using the large-scale training data set to obtain a lump-gangue separation expert system. After training, the expert system is deployed on a chip for rapid logical judgment and outputs identification information. In an optional embodiment, the expert system is deployed on an NVIDIA high-speed chip, and the maximum identification speed can reach 5000 pieces / second.

[0032] Due to the fast identification speed, the identification task can be completed in a short movement distance, so the overall length of the device using the system can be greatly shortened, and the device is small in size and easy to build. For example, in this embodiment, the overall length of the device using the system is 7 meters, and after removing the feeding chute and the discharging chute, the main body length only needs to be 4.5 meters.

[0033] The control system of this embodiment adopts a backboard type design, is partitioned and modularized, has a reasonable layout, is easy to operate, and is easy to maintain. The control system gives the subsequent execution system an execution signal by comprehensively calculating the position information of the identified material on the distribution device and the transportation speed of the distribution device. The control system also generally monitors the running state of the whole intelligent dry separation system, timely calculates and adjusts the organic and efficient cooperation of each subsystem, and sends alarm information to the operation and maintenance personnel in case of abnormal situation.

[0034] Further, the system supports remote control, and the staff in the machine room, the main control room and other places can remotely view the system operation and sorting effect and control the system without going to the site to supervise and view. Further, to fully ensure the operation safety and the safety of the inspection personnel and facilitate the collaborative management of various departments, in addition to the machine room and the main control room being able to remotely monitor the system operation, the system also supports handheld device monitoring, and the system operation state can be viewed at any time and anywhere, facilitating the collaborative work between various departments and groups.

[0035] In this embodiment, the execution system for pneumatic sorting includes a pneumatic blowing matrix composed of a plurality of high-precision electromagnetic valves. Exemplarily, as shown in Figure 1 The execution system blows the lump coal to change its falling trajectory, so that it falls into the first storage chamber on the left, while the falling trajectory of the gangue is unchanged, and the gangue falls into the second storage chamber on the right. According to different embodiments, only the gangue can be blown, or the lump coal and the gangue can be blown differently. In any case, the lump coal and the gangue need to be separated by blowing, so that the lump coal and the gangue fall into different storage chambers to achieve the sorting effect.

[0036] Since the gangue content in the raw coal is low, the execution system can set the blowing target only to the gangue. This setting can reduce the air consumption, effectively reduce the working frequency of the execution system and the air supply system providing air for the execution system, reduce the overall operation energy consumption, reduce the operation and maintenance workload of the staff, and greatly save costs.

[0037] The execution system can also include a controller that obtains the identification information and controls the electromagnetic valves in the pneumatic blowing matrix. The pneumatic blowing matrix can control the time interval of the jet, with a precision of 1 ms. By precisely controlling the airflow of the pneumatic blowing matrix, the gangue or lump coal with a size of 30-700 mm can be separated.

[0038] The execution system has the advantages of large thrust, accurate jet and high yield. The execution system has a densely arranged high-pressure pneumatic blowing matrix that provides a large thrust, with a maximum of 700 mm in size. The execution system also has a high-speed electromagnetic valve control system that supports alternating work, with a time precision of 1 ms, accurately controlling the jet amount and thus the thrust, and further controlling the accurate falling point of the lump coal or gangue. The high-speed pneumatic blowing matrix can support 400 tons / hour of raw coal sorting.

[0039] The pneumatic blowing matrix of the execution system also has the advantages of long service life and small interval. The pneumatic blowing matrix is an integrated pneumatic matrix with a service life of ≥15 years. The spatial precision of the jet airflow is 5 mm, which can accurately blow the lump coal or gangue in the adjacent space, and the smallest lump coal or gangue that can be separated is 30 mm.

[0040] Further, the execution system can use multiple electromagnetic valves to spray the same target, which refers to the material that needs to be sprayed. By controlling the cooperation between the multiple electromagnetic valves, the total thrust received by the target is controlled, so that the falling trajectory of the target is more accurately controlled. The direction, diameter, and amount of the spray gas flow can be adjusted according to the control signal. Using multiple gas flows to spray the same target, combined with the control of the spray time of each electromagnetic valve, can also avoid or reduce the interference to other materials outside the sprayed target, and the control accuracy of the spray time can be accurate to 1 ms.

[0041] Some key parameters of the intelligent dry separation system in this embodiment and the corresponding system indicators are shown in Table 1.

[0042] As shown in Table 1, in this embodiment, in the system indicators, the gangue content in the coal is ≤3%, and the coal content in the gangue is ≤3%, that is, the mass proportion of gangue in the lump coal after separation by the system is not more than 3%, and the mass proportion of coal in the gangue after separation is not more than 3%; the system supports a selection lump size of 30-700mm, and supports a lump coal density without limitation, that is, the system can separate materials with a particle size in the range of 30-700mm, and has no limitation on the density of the lump coal entering the system, and is suitable for different coal qualities; the system requires only one ordinary worker; the system can support 50-400t / h of raw coal separation; the service life of the system can reach 15 years, and the core components are guaranteed, and the maintenance cost is 5-20 million per year; the equipment using the system has a small land area, the width range is 1.0m-3.0m, the total length is 7m, and the main part of the inlet chute and the outlet chute is 4.5m; the belt speed is 0.5-3.0m / s, that is, the running speed of the material distribution device in the system can be set in the range of 0.5-3.0m / s. According to different embodiments, the system can adjust the support lump size, maintenance cost, width, and belt speed according to the implementation requirements. From the above key parameters, it can be seen that the system has high screening accuracy, high speed, wide range of lump coal volume, and wide range of coal quality; and compared with the heavy medium coal separator, jigging machine and other coal separation equipment, the system has lower coal production cost; in addition, the system can freely switch between the "coal in gangue" or "gangue in coal" mode to adapt to different needs.

[0043] Figure 2 The figure is a schematic diagram of the main equipment and external supporting equipment of an embodiment of the intelligent dry separation system of the present application.

[0044] Figure 2The diagram shows the outer casing of the AIS (Automatic Identification System) intelligent dry separation system, including a lump coal conveyor belt, a gangue conveyor belt, a vibrating screen, a temporary storage bin for undersize fines, an air compressor, a high-pressure air tank, a condenser dryer, and a three-stage filtration device. The lump coal conveyor belt leads to the lump coal bin, and the gangue conveyor belt leads to the gangue bin. The AIS intelligent dry separation system is the first officially put into production in China to use an intelligent separation system based on visual recognition technology followed by pneumatic separation.

[0045] like Figure 2 As shown, in this embodiment, the intelligent dry separation system ( Figure 2 The AIS intelligent dry separation system includes a lump coal conveyor belt and a gangue conveyor belt. The lump coal conveyor belt connects a first storage chamber and a lump coal bin, while the gangue conveyor belt connects a second storage chamber and a gangue bin. After separation, the lump coal and gangue fall into the first and second storage chambers respectively. Subsequently, the lump coal enters the lump coal bin via the lump coal conveyor belt, and the gangue enters the gangue bin via the gangue conveyor belt. In this embodiment, the storage chambers can be in the form of chutes to facilitate the discharge of the separated materials onto the conveyor belts. Some embodiments choose to send the lump coal to the original crusher via the chutes for crushing.

[0046] like Figure 2 As shown, the system is also equipped with external auxiliary equipment, including a temporary storage bin for undersize coal. Before sorting, the raw coal is first screened by a vibrating screen to remove undersized fines, and the temporary storage bin is used to temporarily store the undersize coal that has been screened out by the vibrating screen. In this embodiment, the diameter of the vibrating screen aperture is 50mm. In other embodiments, the aperture of the vibrating screen can be adjusted to adjust the lower limit of the particle size of the material to be sorted.

[0047] like Figure 2 As shown, the external supporting equipment of this system also includes an air supply system to provide treated air to the pneumatic jet matrix. Exemplarily, the air supply system includes an air compressor, a high-pressure air tank, a condenser dryer, and a three-stage filtration system. The air compressor compresses the air, then the high-pressure air tank stores the air and stabilizes the pressure. Next, the condenser dryer removes moisture from the air, and then the three-stage filtration system further purifies the air, removing impurities. Finally, the treated, clean, dry, and pressure-stable compressed air is provided to the actuator system for the jetting operation of the pneumatic jet matrix. The air supply system ensures the high efficiency and precision of the pneumatic jet matrix's jetting operation. The following equipment selection can be used as an example: an air-cooled screw air compressor with a displacement ≥ 45.5 m³ / h is selected. 3 / min; gas storage tank volume 10m³ 3 Pressure 1.0 MPa; condenser dryer includes three-stage filtration, processing capacity 48 Nm³. 3 / min, working pressure 0.8MPa.

[0048] like Figure 2The system adopts a closed shell. The system shell is designed to reduce dust and noise. On the one hand, the original coal rolling may produce dust, and the system pneumatic blowing action also produces dust, so except for the inlet and outlet, the rest adopts a closed structure, and further, the system is also equipped with an industrial dust collector not shown, which can effectively remove dust in the closed structure to meet environmental protection requirements, and a filter cartridge dust collector with reverse dust removal function can be selected; on the other hand, in order to reduce the noise generated by the falling, collision of the original coal and the pneumatic blowing matrix blowing, the system adopts a closed structure, and fireproof soundproof cotton is used between the two layers of closed shells. The shell design can further protect the equipment, thereby reducing the daily maintenance workload of the system equipment.

[0049] Figure 3 A schematic block diagram of the identification system of an embodiment of the intelligent dry separation system of the present application.

[0050] Figure 3 The identification process of the identification system in this embodiment is shown, including: an industrial camera collects data in real time to obtain real-time video data of the material conveying belt; Figure 3 a convolution layer (conv layer) extracts features from the real-time video data of the conveying belt to obtain a feature map; Figure 3 an RPN (Region Proposal Network) outputs a proposal region based on the feature map; a DT (Decision Tree) algorithm outputs classification (including distinguishing foreground and background) and coordinate position based on the feature map and the proposal region, and generates a detection result. Among them, the real-time video data includes the material to be separated on the material conveying device.

[0051] Specifically, in this embodiment, the RPN generates a series of candidate proposal regions based on the feature map and outputs them to the DT algorithm. These proposal regions are potential regions that may contain targets, including lump coal and gangue to be separated. The DT algorithm classifies and regresses each proposal region based on the feature map and the proposal region output by the RPN, and finally obtains a detection result. In classification, it is judged whether each proposal region belongs to the foreground (contains a target) or the background (does not contain a target), and if it belongs to the foreground, it is judged whether the target belongs to lump coal or gangue; in regression, the coordinate position of the proposal region is accurately regressed to obtain a more accurate target position.

[0052] The identification system of this embodiment can identify a wide range of particle sizes, and for super large gangue and small gangue in the same picture, the algorithm can also accurately identify. Moreover, the identification system can accurately identify gangue and lump coal for the stacking of lump coal and gangue, and can cope with the scene of a large amount of material to be separated.

[0053] Further, the above-mentioned targets can also include coal-like materials and sundries, wherein the coal-like materials include coal with gangue and lump coal containing iron elements, and the sundries include wood blocks, ironware and paste fillers.

[0054] The coal with gangue has a density between that of coal and gangue, and it is difficult to be processed by traditional shallow groove and jigging separation process, and it is also difficult to be recognized by the ray recognition method due to the mixing of coal elements and gangue elements. However, the coal with gangue can be easily visually recognized from the surface, so the system has a natural advantage in recognizing the coal with gangue.

[0055] As for various sundries such as wood blocks and ironware, the magnetic metals such as ironware can be removed by an iron remover, but the non-magnetic sundries such as wood blocks are difficult to be recognized and removed by traditional methods. However, the sundries such as wood blocks have obvious visual features, so they can be easily recognized in the machine learning algorithm. After recognition, the sundries can be sprayed and blown according to requirements, and separated from the lump coal or gangue.

[0056] Specifically, when the DT algorithm classifies the targets, according to different embodiments, it can only determine whether the target is lump coal, or further classify the gangue, coal-like materials and sundries which are not lump coal, so as to formulate a targeted spraying strategy based on more detailed target categories. In subsequent separation, according to different embodiments, the lump coal can be planned to fall into a first storage chamber, and other targets such as gangue, coal-like materials and sundries fall into a second storage chamber; or the storage chambers can be increased, so that the lump coal falls into a first storage chamber, and the gangue, coal-like materials and sundries fall into different storage chambers.

[0057] Further, the system can also effectively recognize the lump coal and gangue with moisture and water. Although the lump coal and gangue with moisture and water have similar luster on the surface, they are completely different in surface texture features and line structure. The artificial intelligence method can capture various features and details of the pictures, and then comprehensively distinguish the lump coal and gangue. For example, the surface texture of the lump coal with moisture is obviously crystalline, while the surface texture of the gangue with moisture is rough and the edges are sharp.

[0058] This embodiment utilizes the difference in apparent features of coal and gangue, adopts a high-performance computing system, a high-precision recognition sensor and a DT algorithm, collects and recognizes the image information of the gangue and lump coal on the distribution device in real time, extracts, models, stores and learns the features of the lump coal and gangue based on machine learning, neural network and deep learning technology, realizes the recognition and classification of the lump coal and gangue, and transmits the processed information such as shape, size, position and speed of the lump coal and gangue to the control system through a data interface within milliseconds.

[0059] Further, the recognition system can intelligently and autonomously learn in the recognition work, continuously improve the recognition accuracy for the independent coal quality of each coal mine, and the visual recognition accuracy will be higher and higher with the increase of the processing data amount, and the recognition system can also self-check faults.

[0060] Figure 4 This diagram illustrates the relationship between the recognition accuracy and the number of training images in some embodiments of the intelligent dry selection system of the present invention.

[0061] For example, when training the model, the recognition system first establishes a training dataset of 100,000 or more photos of lump coal and gangue manually labeled. In an optional embodiment, the number of labeled photos reaches the millions. Then, a model containing millions of parameters is trained using the large-scale training dataset to obtain a coal and gangue sorting expert system. After training, the expert system is deployed on a chip for rapid logical judgment and outputs recognition information. In an optional embodiment, the expert system is deployed on an NVIDIA high-speed chip, with a maximum recognition speed of 5000 pieces / second.

[0062] Figure 4 The left side shows the recognition rate improvement curve of one embodiment, which illustrates the relationship between recognition accuracy (vertical axis) and the number of training images (horizontal axis); Figure 4 The right side shows a comparison of the recognition rate improvement curves of different recognition systems in the two embodiments. The curves of different colors show the relationship between the recognition accuracy (vertical axis) and the number of training images (horizontal axis) of different recognition systems.

[0063] pass Figure 4 It can be seen that the more training images a system has, the higher its recognition accuracy. In the early stages, the accuracy increases rapidly as the number of images increases. However, once the number of images exceeds a critical point (around 250,000-300,000 images), the accuracy decreases with increasing data volume, with a theoretical upper limit of approximately 99.95%.

[0064] In an optional embodiment, the recognition system includes a DT algorithm model, which uses millions of parameters to describe an image. As the number of training images increases, the upper limit of recognition accuracy continuously increases, with a theoretical upper limit of 99.95%. In practice, a model with an accuracy of 99.2% has been achieved.

[0065] Figure 5 This is a schematic diagram of the control system of an embodiment of the intelligent dry separation system of the present invention.

[0066] like Figure 5 As shown, the control system in this embodiment (i.e. Figure 5 The intelligent management and control system has a power supply module and an input control interface. The power supply module provides stable power to the control system, and the input control interface is used for the control system to receive signals and instructions.

[0067] And, as Figure 5As shown, the control system includes a position information calculation module, a size information processing module, a weight information processing module, and an air valve control module. This control system comprehensively calculates and processes the material position and size information obtained from the identification system. It can also analyze and calculate the material weight information based on the size and material type, and can combine this information with the conveying speed of the fabric distribution device to control the solenoid valves of the pneumatic jet matrix. Figure 5 The system is controlled by an air valve. In some embodiments, the actuator includes a controller, in which the control system can control the actuator to perform the blowing operation by transmitting a signal to the controller.

[0068] like Figure 5 As shown, the control system also includes a status display, a display module, and a programming interface. The status display and display module are used to display system status, system operation, sorting results, etc. The programming interface is used for developers to interact with the control system, allowing developers to input control commands, operate hardware devices, dynamically adjust control logic, etc., and to realize communication between external programs and the control system. This makes the control system more flexible and programmable, facilitating monitoring, debugging, and functional expansion.

[0069] Furthermore, such as Figure 5 As shown, the control system also includes a communication interface and a wireless communication module. This control system supports remote control, allowing staff in the computer room, central control room, and other locations to remotely view the system's operation and sorting results, and operate the control system without on-site supervision. Furthermore, to fully ensure operational safety and the safety of inspection personnel, and to facilitate collaborative management among departments, in addition to remote monitoring of system operation from the computer room and central control room, the control system also supports handheld device monitoring, enabling users to view the system's operating status anytime, anywhere, facilitating collaboration between departments and teams.

[0070] The control system in this embodiment adopts a backplane design, with modular partitioning, a reasonable layout, simple operation, and easy maintenance. The control system also monitors the overall operating status of the entire intelligent dry separation system, promptly calculates and adjusts the efficient coordination of each subsystem, and sends alarm information to operation and maintenance personnel in case of abnormalities.

[0071] Some embodiments of the present invention have been applied in numerous coal mines and coal washing plants, and have gained market recognition. Some embodiments are used to replace manual sorting in coal mines, while others are used to supplement or replace existing coal washing plant sorting systems.

[0072] Some embodiments of the present invention have one or more of the following beneficial effects: (1)Sorting precision is high: the existing technology based on ray recognition mode depends on the setting of threshold value, and cannot realize intelligent learning; while the visual recognition rate of some embodiments of the present application will be higher and higher with the increase of the amount of processed data, and the sorting precision of some already put into production equipment reaches less than 1% of the coal rate in gangue; (2) High recognition efficiency: in the existing technology based on ray recognition mode, X-ray scans the material based on linear array, and only recognizes once; while some embodiments of the present application use machine vision based on surface array recognition, can process 5000 pieces of material per second, and the same material will be tracked and recognized multiple times, so the recognition efficiency is much higher than that of the linear array recognition mode of X-ray; (3) High recognition rate of sundries: various sundries can be effectively recognized, including sleepers, plastic bags, anchor rods and various sundries, and underground paste filling materials; (4) High recognition rate of coal-like materials: coal-like materials, including coal gangue and coal containing iron elements, can be effectively recognized; (5) Recognize wet coal: has a special algorithm model for wet coal to identify wet coal; (6) Wide sorting range: can recognize and spray 30mm to 700mm materials; (7) Small equipment volume, no need to build new plant, low project construction investment, less land use, short construction period; (8) Low operating cost and low complexity, ordinary workers can maintain; (9) Low maintenance cost, less wear and tear, most of the parts are standard products, and the cost is low; (10) Safe, no radiation, no pollution, no water consumption, energy saving and environmental protection, and significant social benefits; (11) Support local and remote synchronous monitoring, ensure operation safety, patrol personnel safety, and facilitate collaborative management of various departments.

[0073] The technical scope of the present application is not limited to the contents in the above description, and those skilled in the art can make various modifications and changes to the above embodiments without departing from the technical idea of the present application, and these modifications and changes should be within the scope of the present application.

Claims

1. An intelligent dry sorting system, characterized in that, The system is used for raw coal separation, and the system includes: A material distribution system, comprising a material distribution device, wherein the material enters the material distribution device and is arranged in a single layer on the material distribution device, the material comprising lump coal, gangue and debris, and the material falls into a storage chamber after passing the end of the material distribution device; The identification system acquires images of the material on the fabric device in real time using a camera, identifies and locates the material in the images based on a machine learning model, and obtains identification information including the location information of the material. The control system issues an execution signal based on the identification information; An execution system includes a pneumatic jetting matrix, which includes multiple solenoid valves. The execution system, according to the execution signal, jettisons falling gangue and debris, and / or jettisons falling lump coal. The jetting alters the falling trajectory of the jettisoned material, causing the gangue and debris to fall into a storage compartment different from that of the lump coal. The identification system is communicatively connected to the control system, and the control system is communicatively connected to the execution system.

2. The system as described in claim 1, characterized in that, The fabric feeding system also includes a vibrating screen, which includes a fabric feeder. The material enters the vibrating screen through the fabric feeding device, and the fabric feeder arranges the material in a single layer on the fabric feeding device.

3. The system as described in claim 1, characterized in that, The identification process of the identification system includes: Real-time acquisition of images of the material on the fabric-making device; Feature maps are obtained by extracting features from the image using a convolutional neural network, and the feature maps are then input into the RPN and DT algorithms. The RPN generates proposed regions based on the feature map and passes them to the DT algorithm; The DT algorithm classifies and regresses the suggested regions based on the feature map and the suggested regions, and outputs detection results. The classification is used to determine the category of materials in each suggested region, and the categories include lump coal, gangue and debris. The regression is used to adjust the coordinate position parameters of the suggested regions.

4. The system as described in claim 1, characterized in that, The training process of the recognition system includes: Acquire images of the material on the fabric assembly. At least 100,000 of the images were labeled to generate a training set. Training a model with at least one million parameters. The trained model is deployed on a chip for real-time identification of the material on the fabric device.

5. The system as described in claim 1, characterized in that, The identification information also includes the shape, size, and speed information of the material. The control system calculates the identification information and sends the execution signal to the execution system. The control system monitors the operating status of the system, calculates and adjusts the coordination between the various parts of the system, and sends alarm information to the staff when an abnormal situation occurs.

6. The system as described in claim 1, characterized in that, The execution system also includes a controller, which receives the identification information and controls the solenoid valve. The solenoid valve supports alternating operation with a timing accuracy of 1ms and a jet airflow spatial accuracy of 5mm. By controlling the jet volume, the thrust is controlled, thereby controlling the trajectory and drop position of the sprayed material. The pneumatic jet matrix provides a thrust of over 1000N to propel materials with a particle size in the range of 30mm-700mm.

7. The system as described in claim 1, characterized in that, The execution system takes the material to be sprayed as the target and uses multiple solenoid valves to spray the same target. By controlling the airflow of multiple solenoid valves, the thrust on the target is controlled, thereby controlling the target's falling trajectory. In addition, by controlling the spraying time, interference with materials other than the target is avoided or reduced. The spraying time control accuracy is 1ms.

8. The system as described in claim 1, characterized in that, The system also includes: The outer shell, except for the inlet and outlet, is a closed structure. The system includes two outer shells, with fireproof and sound-insulating cotton installed between the two outer shells. Industrial dust collectors are used to remove dust from the system. An automatic camera cleaning device includes a wiper mounted outside the camera lens, which automatically cleans the lens.

9. The system as described in claim 1, characterized in that, The system also includes video surveillance, which is used by staff to remotely view the system's coal sorting operations. The video surveillance communicates with the computer room and the central control room, and also supports communication with handheld devices.

10. The system as claimed in claim 1, characterized in that, The operating speed of the fabric spreading device is set in the range of 0.5-3.0 m / s, and the material flow rate of the fabric spreading device is set in the range of 50-400 t / h.

Images