Anti-collision early warning method for coal mining machine and hydraulic support

Abstract

The invention belongs to the technical field of coal mining machine collision prevention, and discloses a coal mining machine and hydraulic support collision prevention early warning method which comprises the steps that a plurality of recognition structures are installed at the position, to be recognized, of a hydraulic support, the recognition structures comprise a plurality of corner reflectors, and the position, to be recognized, of the hydraulic support comprises a face guard or a top beam; echo signal data of the to-be-recognized part are obtained through a radar; inputting the echo signal data into a contour recognition model for contour recognition, and outputting contour data of the to-be-recognized part; wherein the contour recognition model is constructed based on a convolutional neural network; the distance between the coal mining machine and the hydraulic support is judged based on the contour data of the to-be-recognized part, and when the distance between the coal mining machine and the hydraulic support is smaller than a preset collision threshold value, a collision early warning and protection strategy is sent out. According to the technical scheme, the contour of the side protection plate / top beam structure of the hydraulic support can be rapidly and correctly recognized, and then anti-collision early warning of the coal mining machine and the hydraulic support is achieved.

Description

Technical Field

[0001] This invention belongs to the field of anti-collision technology for coal mining machines, and in particular relates to an anti-collision early warning method between a coal mining machine and a hydraulic support. Background Technology

[0002] In fully mechanized coal mining faces, the harsh environment (filled with coal dust, water mist, etc.) reduces the operator's accurate perception of the dynamic distance between the coal mining machine and the hydraulic supports, thus increasing the probability of collisions. Currently, to reduce the probability of collisions, anti-collision radar is typically installed on the coal mining machine. When a collision is imminent, it issues a collision warning or provides protection, thus preventing the collision. In collisions, the coal mining machine often collidees with the hydraulic support's sidewall / top beam. Therefore, accurate detection of the hydraulic support's sidewall / top beam outline is crucial for effectively preventing collisions. Currently, lidar or 4D millimeter-wave radar technology is commonly used to identify the outline of the hydraulic support's sidewall / top beam, but these methods have the following drawbacks: (1) Identification of hydraulic support sidewall / top beam contours based on lidar technology: Hydraulic support sidewalls and top beams typically have regular geometric shapes (such as planes and edges). After lidar scanning acquires point cloud data, specific point cloud segmentation, feature extraction, and model fitting algorithms are used to calculate their contours. However, this method severely attenuates the laser signal when dealing with dense coal dust, water mist, or other similar conditions underground, causing the acquisition of hydraulic support sidewall / top beam point cloud data to fail, thus making it impossible to identify the contours of the hydraulic support sidewall / top beams.

[0003] (2) Contour of hydraulic support side plate / top beam based on 4D millimeter-wave radar technology: 4D millimeter-wave radar can obtain the distance, azimuth, velocity and height information of the target by transmitting frequency modulated continuous wave (FMCW) and receiving signals reflected back from components such as hydraulic support side plate and top beam. This allows for the acquisition of point cloud data. By performing clustering and tracking on these point cloud data, the contours of components such as side plate and top beam can be calculated. However, when there is coal dust coverage, water film, corrosion or angle problems on the surface of the hydraulic support side plate / top beam, it will cause diffuse or weak reflection of millimeter waves, resulting in sparse, flickering or unstable point clouds received by the radar, which in turn leads to low accuracy of 4D millimeter-wave identification of hydraulic support side plate / top beam contours.

[0004] Based on the current state of the technology, there is an urgent need to develop an accurate, reliable and stable anti-collision early warning method for coal mining machines and hydraulic supports, so as to quickly and accurately identify the outline of the hydraulic support side plate / top beam structure. Summary of the Invention

[0005] The purpose of this invention is to provide a method for preventing collisions between a coal mining machine and a hydraulic support, so as to solve the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides a method for preventing collisions between a coal mining machine and a hydraulic support, comprising: S1: Install several identification structures at the identification location of the hydraulic support. The identification structure includes several corner reflectors. The identification location includes a side plate or a top beam. Each corner reflector is arranged sequentially along the outline of the side plate / top beam. S2: Acquire echo signal data of the part to be identified via radar; S3: Input the echo signal data into the contour recognition model for contour recognition, and output the contour data of the part to be recognized; wherein, the contour recognition model is constructed based on a convolutional neural network; Based on the contour data of the part to be identified, the distance between the coal mining machine and the hydraulic support is determined. When the distance between the coal mining machine and the hydraulic support is less than a preset collision threshold, a collision warning and protection strategy are issued.

[0007] Optionally, the corner reflector consists of three mutually perpendicular metal plates.

[0008] Optionally, the radar is installed in the cutting section of the coal mining machine.

[0009] Optionally, the radar is a 4D millimeter-wave radar.

[0010] Optionally, the training process of the contour recognition model specifically includes: Acquiring training data: Real-time radar acquisition of echo data from the side panels / top beams; separate and mark corner reflectors and surrounding objects in the echo data; standardize the marked data; apply moving target display and background subtraction algorithms to suppress stationary clutter; apply clustering and multi-target tracking algorithms to remove interference from random false points to obtain preprocessed echo data. An initial contour recognition model is constructed based on a convolutional neural network. The preprocessed echo data is input into the initial contour recognition model for contour recognition. The model is then trained based on the target loss function to obtain a trained contour recognition model. The contour recognition task in step S3 is then performed based on the trained contour recognition model.

[0011] Optionally, the target loss function includes the Focal Loss loss function and the Smooth L1 Loss loss function.

[0012] Optionally, after the model is deployed, misjudgment cases are continuously recorded, and the contour recognition model is fine-tuned based on the misjudgment cases to improve the model's recognition accuracy.

[0013] The technical effects of this invention are as follows: This invention greatly improves the reliability and stability of point cloud data extraction by 4D millimeter-wave radar by installing a special identification structure on the outline edge of the hydraulic support side plate / top beam structure, and enhances the radar's perception capability under complex and harsh environmental conditions.

[0014] This invention can significantly improve the accuracy and precision of fitting the contour of hydraulic support side plate / top beam structure, greatly reduce the complexity and computational cost of the algorithm, thereby achieving fast and error-free identification of the contour of hydraulic support side plate / top beam structure. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of a special identification structure in an embodiment of the present invention; Figure 2 This is a schematic diagram showing the installation position of the special identification structure of the hydraulic support side plate in an embodiment of the present invention; Figure 3 This is a schematic diagram showing the installation position of the special identification structure of the hydraulic support top beam in an embodiment of the present invention; Figure 4 This is a diagram showing the positional relationship between the profile of the hydraulic support side plate / top beam detected by 4D millimeter-wave radar in an embodiment of the present invention. Figure 5 This is a flowchart of the model training algorithm in an embodiment of the present invention; Figure 6 This is a flowchart illustrating the implementation of an embodiment of the present invention. Detailed Implementation

[0017] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as a limitation of the present invention, but rather as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0018] It should be understood that the terminology used in this invention is merely for describing particular embodiments and is not intended to limit the invention. Furthermore, with respect to numerical ranges in this invention, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Every smaller range between any stated value or intermediate value within a stated range, and any other stated value or intermediate value within said range, is also included in this invention. The upper and lower limits of these smaller ranges may be independently included or excluded from the range.

[0019] Various modifications and variations can be made to the specific embodiments described in this specification without departing from the scope or spirit of the invention, as will be apparent to those skilled in the art. Other embodiments derived from this specification will also be obvious to those skilled in the art. This application specification and embodiments are merely exemplary.

[0020] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] like Figure 1 - Figure 6 As shown, this embodiment provides a method for collision avoidance early warning between a coal mining machine and a hydraulic support, comprising: installing several identification structures at the identification location of the hydraulic support, each identification structure including several corner reflectors; the identification location including a side guard plate or a top beam; the corner reflectors being arranged sequentially along the contour edge of the side guard plate / top beam; acquiring echo signal data of the identification location using radar; inputting the echo signal data into a contour recognition model for contour recognition, and outputting the contour data of the identification location; wherein the contour recognition model is constructed based on a convolutional neural network; determining the distance between the coal mining machine and the hydraulic support based on the contour data of the identification location; and issuing a collision warning and protection strategy when the distance between the coal mining machine and the hydraulic support is less than a preset collision threshold. The technical solution described in this invention is capable of...

[0023] In this embodiment, a set of special identification structures is installed along the outline of the hydraulic support side plate / top beam structure to concentrate the radar wave energy scattered by its outline and reflect it back in a directional manner. This makes the echo signal intensity several orders of magnitude higher than that of other metal structures around the hydraulic support side plate / top beam, thereby obtaining a set of easily distinguishable echo signal targets. The point cloud formed by this set of echo signal targets can accurately delineate the outline of the hydraulic support side plate / top beam within a blurry and sparse point cloud cluster.

[0024] Figure 1This is a schematic diagram of a special identification structure installed along the outline of the hydraulic support's sidewall / top beam structure. The device consists of three mutually perpendicular metal plates. To avoid interfering with the normal operation of the hydraulic support's sidewall / top beam, a suitable small-sized special identification structure should be selected. Figure 2 - Figure 3 As shown, this illustrates the special identification structure for the hydraulic support's side guard plate / top beam (hydraulic supports are divided into those with and without side guard plates; the side guard plate is detected when it's present, and the top beam is detected when it's absent). A set of special identification structures is evenly installed along the outline of the hydraulic support's side guard plate / top beam structure. For example... Figure 4 As shown, this is a 4D millimeter-wave radar detection of the positional relationship of the hydraulic support side plate / top beam contour. The radar is installed at the cutting part of the coal mining machine, so that the radar can rotate in sync with the cutting part. Due to the special recognition structure, the radar wave energy scattered from the edge of the hydraulic support side plate / top beam contour is concentrated and reflected back in a directional manner, making the radar echo signal very pure and strong, providing a more reliable data basis for the accurate positioning of the hydraulic support side plate / top beam contour.

[0025] This embodiment is feasible. Based on the installation of a special identification structure on the outline of the hydraulic support sidewall / top beam structure under different working conditions in different fully mechanized mining faces, a set of training algorithms for the coal mining machine cutting section and the outline of the hydraulic support sidewall / top beam structure is established. The training algorithm first performs data acquisition and labeling, signal data preprocessing, then uses a convolutional neural network model to extract features, and then performs loss function algorithm training, so that the coal mining machine anti-collision millimeter-wave radar can "learn" to accurately identify the outline of the hydraulic support sidewall / top beam in complex underground conditions.

[0026] Under different working conditions, although the special recognition structure can enhance the contour recognition of the hydraulic support sidewall / top beam structure, interference still exists. Furthermore, extracting this contour feature is also a problem. Therefore, a dedicated model training algorithm is needed to process the contour of the hydraulic support sidewall / top beam structure with the added special recognition structure. The model training algorithm process is as follows: Figure 5 As shown, the specific steps include: (1) Data acquisition and labeling of 4D millimeter-wave radar for coal mining machine: 4D millimeter-wave radar acquires echo data of hydraulic support sidewall / top beam in underground fully mechanized mining face in real time, and marks the special identification structure of hydraulic support sidewall / top beam and its surrounding objects separately. (2) Preprocessing the collected data: Standardize the collected data of special identification structures of hydraulic support side plate / top beam and other surrounding objects, apply moving target display and background subtraction algorithms to suppress static clutter, and apply clustering and multi-target tracking algorithms to remove interference from random false points; (3) Use of convolutional neural network models: A convolutional neural network is used to process data on the special identification structures of hydraulic support sidewalls / top beams and other surrounding objects in space, thereby extracting the features of the relevant objects. Specific steps include: A convolutional neural network is used to perform feature voxelization on the point cloud data of special identification structures such as hydraulic support side plates / top beams and other surrounding objects in space. Specifically, this includes: defining the voxel space, point cloud allocation and voxel filtering, in-voxel feature engineering, and constructing sparse tensors.

[0027] Hierarchical feature learning in convolutional neural networks, with specific algorithm steps: The three-level structure of hierarchical feature learning: Level 1 low-level features - local pattern perception (identifying local pattern edges / corners / movements), Level 2 mid-level features - component combination (combining into object components such as side guard plate edges / top beam planes), and Level 3 high-level features - global object and context perception (synthesizing the complete object, the entire hydraulic support). Multi-scale feature fusion: Upsampling of deep high-level feature maps, and fusing the sampled features with mid-level or low-level feature maps from the corresponding layers of the backbone network; thereby extracting the features of relevant objects.

[0028] (4) Model setup and training strategy: Task definition: The task type is 3D keypoint detection; Objective: Instead of detecting the entire side panel / top beam, precisely detect the center or corner points of special identification structures (such as corner reflectors or marker blocks) installed on it; use the spatial coordinates of these key points to calculate the pose, angle, and extension / retraction of the profile. Output: 3D coordinates (x, y, z) of K key points, where K is the number of identified structures (e.g., one at each end of the side panel and one at the center of the top beam, for a total of 3). Model input: Multi-channel bird's-eye view image, with channels including maximum reflection intensity (highlighting markers), point density (highlighting shapes), average height, and average velocity; Backbone network: HRNet (High Resolution Network) was chosen because its core advantage lies in parallel processing of multi-resolution features and repeated fusion, which can always maintain high resolution representation, which is crucial for pixel-level accurate localization.

[0029] Detection head: A simple 1x1 convolutional layer is used to convert the output of HRNet (high-resolution network) into a heatmap with K channels. The peak position of each heatmap corresponds to a key point; Model training strategy: By using the Focal Loss loss function, adjusting the modulation factor and weight factor, focusing on feature samples that are difficult to distinguish, the model's performance in complex scenes is improved. The Smooth L1 Loss loss function is used to regress the position and distance of the detected object. It is not sensitive to outliers and the training is more stable.

[0030] (5) Model deployment: After the trained model is deployed, suspicious misjudgment cases can be continuously recorded (confirmed by the operator). These new data are used regularly to fine-tune the model so that it can continuously adapt to new working surfaces and environmental changes, thereby improving the accuracy of the hydraulic support side plate / top beam structure outline.

[0031] In summary, this embodiment installs a set of special identification structures (corner reflectors, consisting of three mutually perpendicular metal plates) for 4D millimeter-wave radar along the outline of the hydraulic support side plate / top beam. This device has an extremely high radar cross-section (RCS) and is attitude-insensitive, creating stable and strong reflection points, thereby improving the radar's perception effect and reliability at specific key locations. In sparse point clouds, algorithms struggle to identify the outline of the hydraulic support side plate / top beam within a blurry cluster of points. When a set of special identification structures appears stably according to a preset geometry (e.g., a rectangle), they provide strong anchor points for the identification algorithm. The algorithm can easily and accurately locate and identify the outline of the hydraulic support side plate / top beam by recognizing these "lighthouses." This significantly reduces the difficulty of extracting information from sparse point clouds in complex and noisy environments.

[0032] Meanwhile, this embodiment has developed a dedicated training algorithm for the contour model of hydraulic support side plate / top beam based on a special recognition structure for different working conditions. After training with this model training algorithm, the detection accuracy of the contour of hydraulic support side plate / top beam can be improved, thereby realizing accurate anti-collision warning between the coal mining machine and the hydraulic support.

[0033] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A coal mining machine and hydraulic support anti-collision early warning method, characterized in that, The method comprises the following steps: S1: installing a plurality of identification structures at a to-be-identified part of a hydraulic support, the identification structures comprising a plurality of corner reflectors, the to-be-identified part comprising a shield or a roof beam, and each of the corner reflectors being arranged along the contour edge of the shield / roof beam in sequence; S2: acquiring echo signal data of the to-be-identified part by using a radar; S3: inputting the echo signal data into a contour identification model to perform contour identification, and outputting contour data of the to-be-identified part; wherein the contour identification model is constructed based on a convolutional neural network; Based on the contour data of the to-be-identified part, the distance between the coal mining machine and the hydraulic support is determined, and when the distance between the coal mining machine and the hydraulic support is less than a preset collision threshold, a collision warning and a protection strategy are issued.

2. The coal mining machine and hydraulic support anti-collision early warning method according to claim 1, characterized in that, The corner reflector is composed of three mutually perpendicular metal plates.

3. The method for preventing collision between a coal mining machine and a hydraulic support according to claim 1, characterized in that, The radar is installed on the cutting part of the coal mining machine.

4. The coal mining machine and hydraulic support anti-collision early warning method according to claim 1, characterized in that, The radar is a 4D millimeter wave radar.

5. The coal mining machine and hydraulic support anti-collision early warning method according to claim 1, characterized in that, The training process of the contour identification model comprises the following steps: Obtaining training data: collecting echo data of the shield / roof beam in real time by using the radar, marking each corner reflector and the surrounding objects in the echo data, performing standardization processing on the marked data, applying a moving target display and background subtraction algorithm to suppress static clutter, applying a clustering and multi-target tracking algorithm to remove the interference of random false points, and obtaining preprocessed echo data; Constructing an initial contour identification model based on a convolutional neural network, inputting the preprocessed echo data into the initial contour identification model to perform contour identification, training based on a target loss function, obtaining a trained contour identification model, and performing the contour identification task in step S3 based on the trained contour identification model.

6. The coal mining machine and hydraulic support anti-collision early warning method according to claim 5, characterized in that, The target loss function comprises a Focal Loss loss function and a Smooth L1 Loss loss function.

7. The coal mining machine and hydraulic support anti-collision early warning method according to claim 5, characterized in that, After the model is deployed, misjudgment cases are continuously recorded, the contour identification model is fine-tuned based on the misjudgment cases, and the identification accuracy of the model is improved.

Images