Image-pressure-vibration information fusion inversion coal caving state test platform and method
Through the fusion method of image-pressure-vibration information, combined with multi-head attention mechanism and multi-layer long and short-term memory network, the accuracy of coal gangue identification in the existing technology is solved, and real-time coal gangue identification is achieved The reversal of the release state of the top coal improves the accuracy of automatic control.
Patent Information
- Application Number
- CN202411560199.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-02-14
AI Technical Summary
The existing coal gangue identification methods are difficult to adapt in real time under complex working conditions, and a single method cannot effectively obtain coal gangue characteristic information, resulting in low accuracy of inversion of the top coal release state.
Using the test method and platform for inversion of coal release status of image-pressure-vibration information, coal gangue images are collected through high-definition cameras, pressure sensors obtain the bearing pressure of the mask beam, and magnetic vibration sensors collect the vibration signal of the tail beam, combining the attention of multi-heads The mechanism is characterized by fusion with multi-layer long and short-term memory networks to obtain the mixing rate and coal gangue release quality in real time.
Real-time coal gangue identification and inversion of the top coal release status under complex working conditions is achieved, the accuracy of the gangue mixing rate and the quality of coal gangue release is improved, and the closing timing of the coal discharge port can be automatically controlled.
Smart Images
Figure CN119442152A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of gangue identification and automatic coal caving control in mining engineering technology, and in particular to an image-pressure-vibration information fusion inversion coal caving state test platform and method. Background Art
[0002] Comprehensive top coal caving is the main method of mining thick coal seams. Obtaining the top coal release status is of great significance to improving the level of intelligent mining in comprehensive top coal caving. The indicators for measuring the top coal release status include top coal release rate and mixed gangue rate. Improving the top coal release rate and reducing the mixed gangue rate are the core goals of comprehensive top coal caving. The most core technology is to obtain the real-time mixed gangue rate, that is, coal gangue recognition technology. At present, the methods of coal gangue recognition are mainly divided into two categories: image recognition technology and non-image recognition technology, including visible light image recognition, infrared image recognition, gamma ray, vibration signal, etc. The above recognition methods have limitations to varying degrees. For example, visible light image recognition has high requirements for image clarity and is seriously affected by dust and light; the gamma ray receiver is too expensive and the cost is too high; the vibration wave method is easily affected by on-site noise and the properties of coal gangue are different.
[0003] The fully mechanized caving face is in a complex and harsh working environment. A single coal gangue identification method cannot adapt to various working conditions in real time. Considering obtaining coal gangue feature information at the same position and time, a coal gangue identification method with multi-source information fusion is established. Through research, it is found that the top coal release quality has a certain relationship with the bearing pressure of the support shield beam, based on which the top coal release amount can be obtained; visible light image recognition has a high recognition accuracy, and the coal and gangue images have obvious distinguishing features, which can be distinguished by deep learning; vibration signal recognition has good environmental applicability, and the signals generated by the collision of coal and gangue with the support tail beam and rear scraper also have significant signal characteristics. Extract image and vibration signal features, and perform feature fusion through multi-head attention mechanism and multi-layer long short-term memory network. Based on this, the mixed gangue rate can be obtained in real time. The total coal gangue release amount and the proportion of coal gangue can be obtained by fusing the pressure signal of the support shield beam with the vibration signal of the support tail beam, rear scraper, and coal gangue image above the rear scraper, so as to realize the inversion of the top coal release state.
[0004] At present, the research on collaborative identification of coal gangue and inversion of top coal dropping status by multi-source information fusion has not been applied in actual production and can only be carried out through similar simulation. At present, the platform for indoor similar simulation is small, and the support advancement and the start and stop of the coal dropping port are all manually operated. The experimental data has large deviation and low accuracy. Coal gangue particles are usually made of barite and other similar materials. The vibration signal generated by their dropping is quite different from the real coal gangue vibration signal, and cannot fully reflect the real top coal dropping status.
[0005] It should be noted that the information disclosed in the background technology section of the present invention is only intended to increase the understanding of the overall background of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes prior art already known to those skilled in the art. Summary of the invention
[0006] In view of the deficiencies in the above-mentioned prior art, the present invention proposes an image-pressure-vibration information fusion inversion coal caving state test method, and adopts an image-pressure-vibration information fusion inversion coal caving state test platform;
[0007] The test platform includes a coal gangue particle container body, a hydraulic support system, a rear scraper conveyor system, a bearing base, an image acquisition system and a control system; the coal gangue particle container body is a rectangular structure with upper and lower openings; the bearing base is a rectangular frame structure, and coal gangue baffles are arranged on the middle and upper parts of the front side, the rear side and the right side, and the lower part of the container body is connected to the upper part of the bearing base; a hydraulic support system is arranged inside the bearing base, the top surface of the hydraulic support system is located at the connection between the container body and the bearing base, the right side of the bottom of the hydraulic support system is connected to the rear scraper conveyor system through a connecting cylinder, and the bottom of the hydraulic support system and the bottom of the rear scraper conveyor system can be slidably connected On the bearing base, the rear scraper conveyor system is located below the gangue baffle, the rear scraper conveyor system is located below the shield beam, tail beam and retractable plug of the hydraulic support system, and the range of movement of the shield beam, tail beam and retractable plug of the hydraulic support system is set within the range of the gangue baffle; the hydraulic support system is moved by traction of the traction cylinder, the shield beam of the hydraulic support system is fixed with a pressure sensor on the outside, and the inner side of the tail beam of the hydraulic support system is fixed with a magnetic vibration sensor; the image acquisition system is located above the rear scraper conveyor system for collecting images of released gangue particles; the electronic control system is used to control the operation of each component of the test platform and collect and store data;
[0008] The test method comprises the following steps:
[0009] S101: Prepare the test platform;
[0010] S102: A coal caving test is conducted using a test platform. During the test, a high-definition camera is used to collect images of coal gangue particles during the coal caving process, a pressure sensor is used to obtain the bearing pressure of the shield beam during the coal caving process, and a magnetic vibration sensor is used to collect vibration signals generated by the tail beam being impacted by coal gangue particles;
[0011] The gangue particle images are made into a gangue particle image dataset according to the format of the PASCAL VOC dataset, and the gangue particle images are manually annotated using the CVAT image annotation software. The annotated gangue particle images are exported to the PASCAL VOC 1.1 format as a gangue particle image dataset.
[0012] The vibration signal data is used to calculate the optimal variational modal decomposition parameters through an iterative process using a particle swarm algorithm. The initial vibration signal is modally decomposed to obtain a series of modal functions. The correlation coefficient between each modal function and the original signal is calculated. The irrelevant component modal functions are filtered out through a threshold and the signal is reconstructed. The eigenvalues of the remaining modal function components are calculated to obtain a vibration signal data set.
[0013] S103: Establish the functional relationship between the bearing pressure of the shield beam and the mass of the released body
[0014] Record several groups of shield beam bearing pressure data and corresponding released body mass data; obtain the functional relationship L(x) between the shield beam bearing pressure and the released body mass by fitting; substitute the functional relationship between the shield beam bearing pressure and the released body mass based on the relationship S(t) between the shield beam bearing pressure data and time, and obtain the functional equation L(S(t)) between the released body mass and time;
[0015] S104: Training coal gangue image-mixed gangue ratio detection calculation model
[0016] U-Net is used as the coal gangue image-mixed gangue rate detection calculation model. The U-Net model structure includes an encoder, a jump connection and a decoder. It is divided into a training set, a validation set and a test set for training, validation and testing. The error between the true value and the predicted value of the coal gangue image data set is compared, and the training parameters are continuously adjusted. When the set total error is reached, the coal gangue image-mixed gangue rate detection calculation model training is considered to be completed.
[0017] S105: Training vibration signal-mixing ratio recognition network model
[0018] The impact vibration simulation was carried out using coal gangue particles with different gangue ratios. Energy, energy moment, kurtosis, waveform factor and matrix singular value were selected as feature vectors, and one-dimensional convolutional network, ReLU activation function and maximum pooling layer were used for feature extraction. The vibration signal, its feature vector and simulated gangue ratio were divided into training set, validation set and test set, and the deep learning neural network was introduced for training, validation and testing. The error between the true value and the predicted value of the vibration signal data set was compared, and the training parameters were continuously adjusted. When the set total error was reached, the training of the vibration signal-gangue ratio recognition network model was considered to be completed.
[0019] S106: Establishing a coal gangue recognition model based on image-vibration information fusion
[0020] Select several groups of gangue images and vibration signals collected at the same time and position, make a fusion data set and extract the image features and vibration signal features in each group of data; use a multi-head attention mechanism and a multi-layer long short-term memory network to fuse the image features with the vibration signal features; select the training parameters of the gangue image-mixed gangue rate detection calculation model and the vibration signal-mixed gangue rate recognition network model when the total errors reach the set requirements as the training parameter values for training; compare the true value and predicted value errors of the fusion data set, and continuously adjust the training parameters. When the set total error is reached, it is considered that the training of the gangue recognition model fused with image-vibration information is completed;
[0021] The real-time gangue ratio γ is predicted by using the trained image-vibration information fusion coal gangue recognition model, and the real-time predicted gangue ratio variation function γ(t) is obtained;
[0022] S107: Fuse image-pressure-vibration information to obtain real-time coal gangue discharge mass and mixed gangue ratio, invert the total discharge mass and total mixed gangue ratio of each step; calculate the real-time mixed gangue ratio of the total discharge volume based on the real-time predicted mixed gangue ratio change function over time and the real-time discharge volume mass, and the calculation formula is as follows:
[0023]
[0024] Where: t is time, unit is s; R mix is the real-time mixed waste rate of the total discharged body; γ(t) is the predicted mixed waste rate changing with time, L(S(t)) is the function of the discharged body mass and time, and L′(S(t)) is the derivative of the function of the discharged body mass and time.
[0025] Preferably, the coal gangue particle container body comprises a restraint frame and a restraint plate, the restraint plate forms a rectangular parallelepiped structure with upper and lower openings, and the restraint frame is fixed to the outer periphery of the restraint plate to act as a rib to reinforce and fix the restraint plate.
[0026] Preferably, the bearing base is composed of a supporting frame forming a rectangular frame main structure, the right side, front side and rear side of the rectangular frame main structure are all aligned with and connected to the coal gangue particle container body, and the left side is longer than the coal gangue particle container body; the bottom of the interior of the rectangular frame main structure is provided with a guide rail groove arranged along the length direction, and one end of the traction cylinder is fixed to the left side of the rectangular frame main structure; a collecting device is provided for collecting the coal gangue particles transported by the rear scraper conveyor system.
[0027] Preferably, the hydraulic support system includes a hydraulic support main support body, and the hydraulic main support body serves as both a base and a top beam; the left side of the hydraulic support main support body is connected to the other end of the traction cylinder; a hydraulic support slot pin is provided at the lower part of the hydraulic support main support body, and the hydraulic support slot pin is slidably connected to the guide rail slot; the upper left end of the shield beam is rotatably connected to the top of the right side of the hydraulic support main support body and can be fixed at a set angle, the lower right end of the shield beam is rotatably connected to the upper left end of the tail beam, and the rotation angle is controlled by the tail beam hydraulic cylinder, and the retractable plug plate is connected to the internal groove of the tail beam through the plug plate hydraulic cylinder so that the retractable plug plate can be retracted relative to the tail beam; the front and rear widths of the hydraulic support system are consistent with the front and rear widths of the internal space of the coal gangue particle container and the bearing base, and the left side of the hydraulic support main support body reaches or slightly exceeds the coal gangue particle container, and the retractable plug plate can at least reach the coal gangue baffle on the right side of the bearing base in the extended state.
[0028] Preferably, the rear scraper conveyor system includes a scraper, a conveying chain, a conveying gear, a connecting cylinder, a middle trough, and a conveyor slot pin; the scraper is driven by the conveying chain to scrape away the coal gangue particles placed above the middle trough, and the conveying gear is located at the front and rear ends of the conveying chain to drive the conveying chain to rotate in a circle; the hydraulic support system is connected to the scraper conveyor system through a connecting cylinder, and a conveyor slot pin is provided at the bottom of the rear scraper conveyor system, and the conveyor slot pin is slidably connected to the guide rail slot.
[0029] Preferably, the image acquisition system comprises a high-definition camera and a support frame, wherein the high-definition camera can acquire images of discharged coal gangue particles, and the support frame fixes the high-definition camera above the rear scraper conveyor system.
[0030] Preferably, in step S104, the encoder extracts feature maps of four different scales through four sub-modules with the same structure, each sub-module consists of two 3×3 convolution kernels, a ReLU activation function and a 2×2 maximum pooling layer; the jump connection is responsible for passing the four feature maps to the decoder; the decoder is symmetrical in structure to the encoder, and supplements the detailed features of the image by integrating the shallow feature maps output by the corresponding encoding layer, and restores the resolution of the feature maps layer by layer through upsampling to generate the segmentation result.
[0031] Preferably, in steps S104-S106, a stochastic gradient descent optimizer is used; the training strategy is Dropout regularization, and class label smoothing is used; and the evaluation indicators of the calculation model are accuracy, precision, recall rate, and F1 value.
[0032] Preferably, when the total discharged volume real-time mixed waste rate R mix When the critical value is reached, the coal discharge port is closed and coal discharge is stopped.
[0033] Beneficial effects of the present invention:
[0034] 1. The test method of the present invention proposes to collect coal gangue images, shield beam bearing pressure, and tail beam vibration signals during coal discharge, establish a functional relationship between the shield beam bearing pressure and the mass of the discharge body, establish an image-vibration information fusion coal gangue recognition model, fuse the image, pressure, and vibration information to obtain the real-time mass proportion of gangue in the total discharge body (the real-time mixed gangue rate of the total discharge body), and control the closing timing of the coal discharge port according to the real-time mixed gangue rate requirement of the total discharge body, which can solve the problem of the limitation of a single coal gangue recognition method under complex working conditions.
[0035] 2. In the test platform of the present invention, real gangue particles are selected for coal placement, the collected gangue images and vibration signals have good authenticity, the bracket moving and scraper conveyor coal transportation are fully electro-hydraulic automatic control, and the test operation is convenient. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention. In the accompanying drawings:
[0037] Figure 1 It is a front view of the overall structure of the test platform of the present invention;
[0038] Figure 2 It is a side view of the overall structure of the test platform of the present invention;
[0039] Figure 3a It is a schematic diagram of the overall structure of the coal gangue particle container in the test platform of the present invention;
[0040] Figure 3b It is a schematic diagram of the overall structure of the hydraulic support system in the test platform of the present invention;
[0041] Figure 3c It is a schematic diagram of the local structure of the hydraulic support system in the test platform of the present invention;
[0042] Figure 3d It is a schematic diagram of the overall structure of the rear scraper conveyor system in the test platform of the present invention;
[0043] Figure 3e It is a schematic diagram of the overall structure of the bearing base in the test platform of the present invention;
[0044] Figure 4 It is a schematic diagram of the principle of fusion of image features and vibration signal features in the test method of the present invention;
[0045] Figure 5 It is a flow chart of vibration signal decomposition in the test method of the present invention;
[0046] The following are the descriptions of the reference numerals:
[0047] 1. Coal gangue particle container body; 2. Hydraulic support system; 3. Rear scraper conveyor system; 4. Carrying base; 5. Image acquisition system; 101. Constraint frame; 102. Constraint plate; 103. Coal gangue particles; 201. Hydraulic support main support body; 202. Hydraulic support slot pin; 203. Protective beam; 204. Tail beam; 205. Retractable plug plate; 206. Pressure sensor; 207. Magnetic vibration sensor ; 208, tail beam hydraulic cylinder; 209, plug plate hydraulic cylinder; 301, scraper; 302, transmission chain; 303 transmission gear; 304, connecting cylinder; 305, middle groove; 306, conveyor slot pin; 401, support frame; 402, guide rail groove; 403, traction cylinder; 404, collection device; 405, traction motor; 406, coal gangue baffle; 501, high-definition camera; 502, support frame. DETAILED DESCRIPTION
[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0049] In the description of the present invention, it is necessary to understand that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.
[0050] Embodiment 1
[0051] like Figure 1-3eAs shown, the image-pressure-vibration information fusion inversion coal discharge state test platform of the present invention includes a coal gangue particle container body 1, a hydraulic support system 2, a rear scraper conveyor system 3, a bearing base 4, an image acquisition system 5 and a control system (not shown in the figure); the coal gangue particle container body 1 is a rectangular structure with upper and lower openings; the bearing base 4 is a rectangular frame structure, and a coal gangue baffle 406 is arranged on the middle and upper parts of the front side, the rear side and the right side, and the lower part of the container body 1 is connected to the upper part of the bearing base 4; a hydraulic support system 2 is arranged inside the bearing base 4, and the top surface of the hydraulic support system 2 is located at the connection between the container body 1 and the bearing base 4, and the bottom right side of the hydraulic support system 2 is connected to the rear scraper conveyor system 3 through a connecting cylinder 304, and the bottom of the hydraulic support system 2 and the bottom of the rear scraper conveyor system 3 are slidable The hydraulic support system 2 is movably connected to the bearing base 4, the rear scraper conveyor system 3 is located below the coal gangue baffle 406, and the rear scraper conveyor system 3 is located below the shielding beam 203, the tail beam 204 and the retractable plug plate 205 of the hydraulic support system 2. The range of movement of the shielding beam 203, the tail beam 204 and the retractable plug plate 205 of the hydraulic support system 2 is set within the range of the coal gangue baffle 406; the hydraulic support system 2 is towed and moved by the traction cylinder 403, the shielding beam 203 of the hydraulic support system 2 is fixed with a pressure sensor 206 on the outside, and the inner side of the tail beam 204 of the hydraulic support system 2 is fixed with a magnetic vibration sensor 207; the image acquisition system 5 is located above the rear scraper conveyor system 3 and is used to collect images of the discharged coal gangue particles; the electronic control system is used to control the operation of each component of the test platform and to collect and store data.
[0052] like Figure 1-3a As shown, the coal gangue particle container body 1 includes a constraint frame 101 and a constraint plate 102; the constraint plate 102 forms a rectangular structure with upper and lower openings, and the constraint frame 101 is fixed to the outer periphery of the constraint plate to act as a rib, reinforcing and fixing the constraint plate 102, which can effectively constrain the expansion force of the coal gangue particles 103 to the surroundings; the coal gangue particle container body 1 is used to hold the coal gangue particles 103 during the test; the coal gangue particles 103 can be selected from real coal gangue particles of different properties and different sizes according to the test requirements; the constraint plate 102 is made of a transparent acrylic plate.
[0053] like Figure 1-2 , Figure 3eAs shown, the bearing base 4 includes a supporting frame 401, a guide rail groove 402, a traction cylinder 403, a collecting device 404, a traction motor 405 and a coal gangue baffle 406; the bearing base is composed of a supporting frame 401 to form a rectangular frame main structure, the right side, the front side and the rear side of the rectangular frame main structure are aligned with and connected to the coal gangue particle container body 1, and the left side is longer than the coal gangue particle container body 1; the bottom of the inner part of the rectangular frame main structure is provided with The guide rail groove 402, one end of the traction cylinder 403 is fixed to the left side of the rectangular frame main structure, the traction cylinder 403 is connected to the traction motor 405, and the traction motor 405 provides telescopic power for the traction cylinder 403; the collecting device 404 is used to collect the coal gangue particles 103 transported by the rear scraper conveyor system 3; a coal gangue baffle 406 is set in the middle and upper part of the front side, rear side and right side of the rectangular frame main structure, and the coal gangue baffle 406 is made of a transparent acrylic plate.
[0054] like Figure 1-2 , Figure 3b-3cAs shown, the hydraulic support system 2 includes a hydraulic support main support body 201, a hydraulic support slot pin 202, a shield beam 203, a tail beam 204, a retractable plug plate 205, a pressure sensor 206, a magnetic vibration sensor 207, a tail beam hydraulic cylinder 208, and a plug plate hydraulic cylinder 209; considering that the pressure of the hydraulic support support column is not studied, the hydraulic main support body 201 in the present invention plays the role of both the traditional hydraulic support base and the hydraulic support top beam, that is, the lower part of the hydraulic main support body 201 in the present invention The function is equivalent to that of the top beam of a traditional hydraulic support base. The middle part is not connected by a hydraulic cylinder but is directly connected and fixed as a whole. The left side of the main support body 201 of the hydraulic support is connected to the other end of the traction cylinder 403. The lower part of the main support body 201 of the hydraulic support is provided with a hydraulic support slot pin 202, and the hydraulic support slot pin 202 is slidably connected to the guide rail groove 402. The upper left end of the shielding beam 203 is rotatably connected to the top of the right side of the main support body 201 of the hydraulic support and can be fixed at a set angle. The lower right end of the shield beam 203 is rotatably connected to the upper left end of the tail beam 204, and the rotation angle is controlled by the tail beam hydraulic cylinder 208. The retractable plug plate 205 is connected to the inner groove of the tail beam 204 through the plug plate hydraulic cylinder 209 so that the retractable plug plate 205 can be retracted relative to the tail beam 204; a pressure sensor 206 is fixed to the outside of the shield beam 203, and a magnetic vibration sensor 207 is fixed to the inner side of the tail beam 204; the front and rear width of the hydraulic support system 2 is the coal gangue particle collection The front and rear widths of the internal spaces of the container body 1 and the bearing base 4 are consistent, and the left and right lengths of the hydraulic support system 2 are not less than the left and right lengths of the coal gangue particle container body 1 (specifically: the left side of the hydraulic support main support body 201 reaches or slightly exceeds the coal gangue particle container body 1, and the retractable plug plate 205 can at least reach the coal gangue baffle 406 on the right side of the bearing base 4 in the extended state); the hydraulic support system 2 has a similarity ratio of 1:6 to the real hydraulic support, and the height is 0.6m; the pressure sensor 206 uses a thin film surface pressure sensor.
[0055] like Figure 1-2 , Figure 3dAs shown, the rear scraper conveyor system 3 includes a scraper 301, a transmission chain 302, a transmission gear 303, a connecting cylinder 304, a middle groove 305, and a conveyor slot pin 306; the scraper 301 drives the transmission chain 302 to scrape away the coal gangue particles placed on the top of the middle groove 305, and the transmission gear 303 is located at the front and rear ends of the transmission chain 302 to drive the transmission chain 302 to rotate in a circle; the main structure and working mechanism of the scraper conveyor are exactly the same as those used in actual engineering projects, and its structure is not repeated here. The similarity ratio between the scraper conveyor system 3 and the real scraper conveyor is 1:6; in addition, the hydraulic support system 2 is connected to the scraper conveyor system 3 through the connecting cylinder 304, and a conveyor slot pin 306 is provided at the lower part of the rear scraper conveyor system 3, and the conveyor slot pin 306 is slidably connected to the guide rail groove 402.
[0056] The image acquisition system 5 includes a high-definition camera 501 and a support frame 502. The high-definition camera 501 can acquire images of released coal gangue particles. The support frame 502 fixes the high-definition camera 501 above the rear scraper conveyor system 3; preferably, the frame rate of the high-definition camera 501 is higher than 200 frames / s.
[0057] During the test, the shield beam 203 is adjusted to a suitable angle, the tail beam 204 is adjusted to a suitable angle, and the retractable plug plate 205 is opened. At this time, the retractable plug plate 205 abuts against the right side coal gangue baffle 406; coal gangue particles 103 are put into the coal gangue particle container body 1 and the shield beam 203, the tail beam 204, and the coal gangue baffle 406 directly above the retractable plug plate 205; then the traction cylinder 403 drives the hydraulic support system 2 and the rear scraper conveyor system 3 to move forward according to the set step distance, retracts the retractable plug plate 205, and the coal gangue particles 1 03 is released to the rear scraper conveyor system 3 for transportation, and the retractable plug plate 205 is extended to stop releasing coal; the pressure signal collected by the pressure sensor 206, the vibration signal collected by the magnetic vibration sensor 207, and the coal gangue particle image collected by the high-definition camera 501 during the coal releasing process are recorded, and the releasing time and the released body mass of this step are recorded; a coal releasing test is completed; the traction cylinder 403 drives the hydraulic support system 2 and the rear scraper conveyor system 3 to continue to move forward according to the set step distance, and repeats the above steps to release coal and collect data, and so on.
[0058] Embodiment 2
[0059] like Figure 4-5 As shown, the present invention also proposes an experimental method for inverting coal caving state by fusing image-pressure-vibration information, comprising the following steps:
[0060] S101: Install a high-definition camera 501 above the rear scraper conveyor system 3, fix a pressure sensor 206 on the outside of the shield beam 203 of the hydraulic support system 2, and fix a magnetic vibration sensor 207 on the inside of the tail beam 204 of the hydraulic support system 2;
[0061] S102: A coal caving test is performed using a test platform. During the test, a high-definition camera 501 is used to collect images of coal gangue particles during the coal caving process, a pressure sensor 206 is used to obtain the bearing pressure of the shield beam 203 during the coal caving process, and a magnetic vibration sensor 207 is used to collect vibration signals generated by the tail beam being impacted by coal gangue particles;
[0062] Specifically: the coal gangue particle images during the test were obtained, and a coal gangue particle image dataset was prepared according to the format of the PASCAL VOC (Visual Object Classes) dataset. The coal gangue particle images were manually annotated using the CVAT image annotation software, and the annotated coal gangue particle images were exported to the PASCAL VOC 1.1 format as a coal gangue particle image dataset.
[0063] The vibration signal data during the test is obtained, and the optimal variational mode decomposition (VMD) parameters [K, α] are calculated through an iterative process using the particle swarm (PSO) algorithm. The initial vibration signal is modally decomposed to obtain a series of modal functions (IMFs). The correlation coefficient between each modal function and the original signal is calculated. The irrelevant component modal functions are filtered out through a threshold and the signal is reconstructed. The eigenvalues of the remaining modal function components are calculated to obtain a vibration signal data set.
[0064] S103: Establish the functional relationship between the bearing pressure of the shield beam and the mass of the released body
[0065] Record several groups of shield beam bearing pressure data and corresponding discharge body mass data within a coal discharge step; fit the function equation of shield beam bearing pressure and discharge body mass, select multiple fitting function forms, and select the corresponding function form when the correlation coefficient R is satisfied. 2 The one with the largest correlation coefficient among the fitting function equations with a value greater than 0.95 is selected as the functional relationship L(x) between the bearing pressure of the shield beam and the mass of the released body;
[0066] Based on the relationship between the bearing pressure data of the shield beam and time S(t), the functional relationship between the bearing pressure of the shield beam and the mass of the released body is substituted into the functional relationship between the bearing pressure of the shield beam and the mass of the released body, and the functional equation L(S(t)) between the mass of the released body and time is obtained;
[0067] S104: Training coal gangue image-mixed gangue ratio detection calculation model
[0068] U-Net is used as the coal gangue image-mixed gangue rate detection calculation model. The U-Net model structure consists of an encoder (Encoding), a skip connection (Skip Connection) and a decoder (Decoding); the encoder extracts four feature maps of different scales through four sub-modules with the same structure, each of which consists of two 3×3 convolution kernels, a ReLU activation function and a 2×2 maximum pooling layer; the skip connection is responsible for passing these four feature maps to the decoder; the decoder has a symmetrical structure with the encoder, and supplements the image's detailed features by integrating the shallow feature maps output by the corresponding encoding layer, and restores the resolution of the feature maps layer by layer through upsampling to generate segmentation results; the coal gangue particle images collected by the high-definition camera 501 are manually annotated using CVAT image annotation software, and are divided into training set, validation set and test set according to 6:2:2, and the loss function is the cross entropy loss function (Cross Entropy Loss), in the initial training parameters, the training batch size is 4, the optimizer type is selected as stochastic gradient descent (SGD), the momentum parameter is 0.9, and the weight decay parameter is 0.0005; the training strategy is Dropout regularization, class label smoothing is adopted, and the smoothing parameter is 0.1; the evaluation indicators of the model are accuracy (Acc), precision (Pre), recall rate (R), and F1 value (F1-Score); the true value and predicted value errors of the coal gangue image data set are compared, and the training parameters are continuously adjusted. When the total error is <5%, it is considered that the coal gangue image-mixed gangue rate detection calculation model training is completed; the basis for the calculation of the true value is the actual segmentation label of the image after manual annotation;
[0069] S105: Training vibration signal-mixing ratio recognition network model
[0070] The impact vibration simulation was carried out using coal gangue particles with mixing ratios of 0%, 5%, 10%, 15%, 20%, 25%, and 30% (mass ratio). Energy, energy moment, kurtosis, waveform factor, and matrix singular value were selected as feature vectors, and a one-dimensional convolutional network (1D-Conv), ReLU activation function, and maximum pooling layer (Max Pooling) is used for feature extraction; the vibration signal, its feature vector and simulated contamination rate are divided into training set, validation set and test set according to 6:2:2, and imported into deep learning neural network. In the training parameters, the training batch size is 4, the optimizer type is selected as stochastic gradient descent (SGD), the momentum parameter is 0.9, and the weight decay parameter is 0.0005; the training strategy is Dropout regularization, class label smoothing is adopted, and the smoothing parameter is 0.1; the evaluation indicators of the model are accuracy (Acc), precision (Pre), recall rate (R), and F1 value (F1-Score); the error between the true value and the predicted value of the vibration signal data set is compared, and the training parameters are continuously adjusted. When the total error is <5%, the vibration signal-contamination rate recognition network model training is considered to be completed;
[0071] S106: Establishing a coal gangue recognition model based on image-vibration information fusion
[0072] Select several groups of gangue images and vibration signals collected at the same time and position, make a fusion data set and extract the image features and vibration signal features in each group of data; use multi-head attention mechanism (MA) and multi-layer long short-term memory network (ML-LSTM) to fuse the image features with the vibration signal features; select the training parameters when the total error of the gangue image-mixed gangue rate detection calculation model and the vibration signal-mixed gangue rate recognition network model is less than 5% as the training parameter value for this training. Among the training parameters, the training batch size is 4, the optimizer type is selected as stochastic gradient descent (SGD), the momentum parameter is 0.9, and the weight decay parameter is 0.0005; the training strategy is Dropout regularization, class label smoothing is used, and the smoothing parameter is 0.1; the evaluation indicators of the model are accuracy (Acc), precision (Pre), recall rate (R), and F1 value (F1-Score); compare the true value and predicted value errors of the fusion data set, and continuously adjust the training parameters. When the total error is less than 5%, it is considered that the training of the gangue recognition model fused with image-vibration information is completed;
[0073] The real-time gangue ratio γ is predicted by using the trained image-vibration information fusion coal gangue recognition model, and the real-time predicted gangue ratio variation function γ(t) is obtained;
[0074] S107: Fuse image-pressure-vibration information to obtain real-time coal gangue discharge mass and mixed gangue ratio, invert the total discharge mass and total mixed gangue ratio of each step; calculate the mass proportion of gangue in the total discharged coal gangue body (real-time mixed gangue ratio of the total discharged body) based on the real-time predicted mixed gangue ratio change function over time and the real-time discharged body mass, and the calculation formula is as follows:
[0075]
[0076] Where: t is time, unit is s; R mix is the real-time mixed waste rate of the total discharged body; γ(t) is the function of the predicted mixed waste rate changing with time, L(S(t)) is the function of the discharged body mass and time; L′(S(t)) is the derivative of the function of the discharged body mass and time;
[0077] When the total discharge volume real-time mixing ratio R mix When the critical value is reached, the coal discharge port is closed and coal discharge is stopped (this formula can be used to guide engineering practice to determine the real-time comprehensive mixed gangue rate when discharging coal at each coal discharge step, and determine the time to stop coal discharge based on the set total mixed gangue rate).
[0078] It should be noted that "coal caving" is a professional term in top coal mining. When caving, generally coal is caving in the early stage, and what is caving afterwards is a mixture of coal and gangue, collectively referred to as coal caving.
[0079] The above description of the disclosed embodiments is provided in a progressive manner to enable those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. The test method of image-pressure-vibration information fusion inversion of coal caving state, using the image-pressure-vibration information fusion inversion coal caving state test platform; The test platform includes a coal gangue particle container body, a hydraulic support system, a rear scraper conveyor system, a bearing base, an image acquisition system and a control system; the coal gangue particle container body is a rectangular structure with upper and lower openings; the bearing base is a rectangular frame structure, and coal gangue baffles are arranged on the middle and upper parts of the front side, the rear side and the right side, and the lower part of the container body is connected to the upper part of the bearing base; a hydraulic support system is arranged inside the bearing base, the top surface of the hydraulic support system is located at the connection between the container body and the bearing base, the right side of the bottom of the hydraulic support system is connected to the rear scraper conveyor system through a connecting cylinder, and the bottom of the hydraulic support system and the bottom of the rear scraper conveyor system can be slidably connected On the bearing base, the rear scraper conveyor system is located below the gangue baffle, the rear scraper conveyor system is located below the shield beam, tail beam and retractable plug of the hydraulic support system, and the range of movement of the shield beam, tail beam and retractable plug of the hydraulic support system is set within the range of the gangue baffle; the hydraulic support system is moved by traction of the traction cylinder, the shield beam of the hydraulic support system is fixed with a pressure sensor on the outside, and the inner side of the tail beam of the hydraulic support system is fixed with a magnetic vibration sensor; the image acquisition system is located above the rear scraper conveyor system for collecting images of released gangue particles; the electronic control system is used to control the operation of each component of the test platform and collect and store data; It is characterized in that The test method comprises the following steps: S101: Prepare the test platform; S102: A coal caving test is conducted using a test platform. During the test, a high-definition camera is used to collect images of coal gangue particles during the coal caving process, a pressure sensor is used to obtain the bearing pressure of the shield beam during the coal caving process, and a magnetic vibration sensor is used to collect vibration signals generated by the tail beam being impacted by coal gangue particles; The gangue particle images are made into a gangue particle image dataset according to the format of the PASCAL VOC dataset, and the gangue particle images are manually annotated using the CVAT image annotation software. The annotated gangue particle images are exported into the PASCAL VOC1.1 format as a gangue particle image dataset. The vibration signal data is used to calculate the optimal variational modal decomposition parameters through an iterative process using a particle swarm algorithm. The initial vibration signal is modally decomposed to obtain a series of modal functions. The correlation coefficient between each modal function and the original signal is calculated. The irrelevant component modal functions are filtered out through a threshold and the signal is reconstructed. The eigenvalues of the remaining modal function components are calculated to obtain a vibration signal data set. S103: Establish the functional relationship between the bearing pressure of the shield beam and the mass of the released body Record several groups of shield beam bearing pressure data and corresponding released body mass data; obtain the functional relationship L(x) between the shield beam bearing pressure and the released body mass by fitting; substitute the functional relationship between the shield beam bearing pressure and the released body mass based on the relationship S(t) between the shield beam bearing pressure data and time, and obtain the functional equation L(S(t)) between the released body mass and time; S104: Training coal gangue image-mixed gangue ratio detection calculation model U-Net is used as the coal gangue image-mixed gangue rate detection calculation model. The U-Net model structure includes an encoder, a jump connection and a decoder. It is divided into a training set, a validation set and a test set for training, validation and testing. The error between the true value and the predicted value of the coal gangue image data set is compared, and the training parameters are continuously adjusted. When the set total error is reached, the coal gangue image-mixed gangue rate detection calculation model training is considered to be completed. S105: Training vibration signal-mixing rate recognition network model The impact vibration simulation was carried out using coal gangue particles with different gangue ratios. Energy, energy moment, kurtosis, waveform factor and matrix singular value were selected as feature vectors, and one-dimensional convolutional network, ReLU activation function and maximum pooling layer were used for feature extraction. The vibration signal, its feature vector and simulated gangue ratio were divided into training set, validation set and test set, and the deep learning neural network was introduced for training, validation and testing. The error between the true value and the predicted value of the vibration signal data set was compared, and the training parameters were continuously adjusted. When the set total error was reached, the training of the vibration signal-gangue ratio recognition network model was considered to be completed. S106: Establishing a coal gangue recognition model based on image-vibration information fusion Select several groups of gangue images and vibration signals collected at the same time and position, make a fusion data set and extract the image features and vibration signal features in each group of data; use a multi-head attention mechanism and a multi-layer long short-term memory network to fuse the image features with the vibration signal features; select the training parameters of the gangue image-mixed gangue rate detection calculation model and the vibration signal-mixed gangue rate recognition network model when the total errors reach the set requirements as the training parameter values for training; compare the true value and predicted value errors of the fusion data set, and continuously adjust the training parameters. When the set total error is reached, it is considered that the training of the gangue recognition model fused with image-vibration information is completed; The real-time gangue ratio γ is predicted by using the trained image-vibration information fusion coal gangue recognition model, and the real-time predicted gangue ratio variation function γ(t) is obtained; S107: Fuse image-pressure-vibration information to obtain real-time coal gangue discharge mass and mixed gangue ratio, invert the total discharge mass and total mixed gangue ratio of each step; calculate the real-time mixed gangue ratio of the total discharge volume based on the real-time predicted mixed gangue ratio change function over time and the real-time discharge volume mass, and the calculation formula is as follows: Where: t is time, unit is s; R mix is the real-time mixed waste rate of the total discharged body; γ(t) is the predicted mixed waste rate changing with time, L(S(t)) is the function of the discharged body mass and time, and L′(S(t)) is the derivative of the function of the discharged body mass and time.
2. The test method according to claim 1, characterized in that The coal gangue particle container body comprises a restraint frame and a restraint plate. The restraint plate forms a rectangular parallelepiped structure with upper and lower openings. The restraint frame is fixed on the outer periphery of the restraint plate and acts as a rib to reinforce and fix the restraint plate.
3. The test method according to claim 3, characterized in that: The bearing base is composed of a supporting frame to form a rectangular frame main structure, the right side, front side and rear side of the rectangular frame main structure are all aligned with and connected to the coal gangue particle container body, and the left side is longer than the coal gangue particle container body; the bottom of the interior of the rectangular frame main structure is provided with a guide rail groove arranged along the length direction, and one end of the traction cylinder is fixed to the left side of the rectangular frame main structure; a collecting device is provided to collect the coal gangue particles transported by the rear scraper conveyor system.
4. The test method according to claim 3, characterized in that The hydraulic support system includes a hydraulic support main support body, and the hydraulic main support body serves as both a base and a top beam; the left side of the hydraulic support main support body is connected to the other end of the traction cylinder; a hydraulic support slot pin is provided at the lower part of the hydraulic support main support body, and the hydraulic support slot pin is slidably connected to the guide rail slot; the upper left end of the shielding beam is rotatably connected to the top of the right side of the hydraulic support main support body and can be fixed at a set angle, the lower right end of the shielding beam is rotatably connected to the upper left end of the tail beam, and the rotation angle is controlled by the tail beam hydraulic cylinder, and the retractable plug plate is connected to the internal groove of the tail beam through the plug plate hydraulic cylinder so that the retractable plug plate can be retracted relative to the tail beam; the front and rear widths of the hydraulic support system are consistent with the front and rear widths of the internal space of the coal gangue particle container body and the bearing base, and the left side of the hydraulic support main support body reaches or slightly exceeds the coal gangue particle container body, and the retractable plug plate can at least reach the coal gangue baffle on the right side of the bearing base in the extended state.
5. The test method according to claim 4, characterized in that The rear scraper conveyor system includes a scraper, a transmission chain, a transmission gear, a connecting cylinder, a middle trough, and a conveyor slot pin; the scraper is driven by the transmission chain to scrape away the coal gangue particles placed above the middle trough, and the transmission gear is located at the front and rear ends of the transmission chain to drive the transmission chain to rotate in a circle; the hydraulic support system is connected to the scraper conveyor system through a connecting cylinder, and a conveyor slot pin is provided at the bottom of the rear scraper conveyor system, and the conveyor slot pin is slidably connected to the guide rail slot.
6. The test method according to any one of claims 1 to 5, characterized in that: The image acquisition system includes a high-definition camera and a support frame. The high-definition camera can acquire images of discharged coal gangue particles. The support frame fixes the high-definition camera above the rear scraper conveyor system.
7. The test method according to claim 6, characterized in that: In step S104, the encoder extracts feature maps of four different scales through four submodules with the same structure, each of which consists of two 3×3 convolution kernels, a ReLU activation function and a 2×2 maximum pooling layer; The skip connection is responsible for passing these four feature maps to the decoder; The decoder has a symmetrical structure with the encoder. It supplements the detailed features of the image by integrating the shallow feature maps output by the corresponding encoding layer, and restores the resolution of the feature maps layer by layer through upsampling to generate the segmentation result.
8. The test method according to claim 7, characterized in that: In steps S104-S106, a stochastic gradient descent optimizer is used; the training strategy is Dropout regularization, and class label smoothing is used; the evaluation indicators of the calculation model are accuracy, precision, recall rate, and F1 value.
9. The test method according to claim 1, characterized in that: When the total discharge volume real-time mixing ratio R mix When the critical value is reached, the coal discharge port is closed and coal discharge is stopped.