Millimeter wave radar imaging type anti-collision monitoring device and method for roller face guard

Abstract

The invention relates to the technical field of coal mine fully-mechanized coal mining automation and safety monitoring, and discloses a roller side guard millimeter wave radar imaging type anti-collision monitoring device and method, and the device comprises a millimeter wave radar detection front end, a multi-dimensional signal processing host and a fully-mechanized coal mining equipment communication interface. The method comprises the following steps: firstly, filtering a coal rock background by using a millimeter wave reflection intensity threshold, and reserving a metal target point set; calculating the current spatial position of the roller by reading the real-time swing angle of the rocker arm of the coal mining machine and combining with mechanical kinematics parameters, and constructing a dynamic region of interest following the action of the rocker arm; and accurately segmenting the mixed metal point set into a roller point set and a face guard plate point set by utilizing the spatial constraint of the dynamic region, and calculating the minimum Euclidean distance between the roller point set and the face guard plate point set. Through double verification of physical materials and spatial motion, the problem that multiple targets made of the same material are difficult to distinguish in the underground high-dust environment is solved, and accurate monitoring and graded active protection of the distance between the roller and the face guard are achieved.

Description

Technical Field

[0001] This invention relates to the field of coal mine fully mechanized mining automation and safety monitoring technology, specifically to a millimeter-wave radar imaging anti-collision monitoring device and method for drum side guards. Background Technology

[0002] With the continuous improvement of automation and intelligence in fully mechanized coal mining faces, the safety of the coordinated operation of the coal mining machine and hydraulic supports, as two core pieces of equipment in fully mechanized mining operations, has become a key factor restricting production efficiency. During fully mechanized mining operations, the coal mining machine drum needs to be close to the sidewalls of the hydraulic supports for cutting operations to ensure coal recovery rate and maintain roof stability. However, due to the narrow working face space, high equipment operating speed, and complex coordination actions, the risk of collision between the drum and the sidewalls is extremely high. Once a collision occurs, it will not only cause damage to expensive equipment, but may also lead to production interruption or even safety accidents.

[0003] Currently, collision avoidance monitoring for underground equipment mainly uses optical sensors such as infrared, lidar, or industrial cameras. However, the cutting process in fully mechanized mining faces generates extremely high concentrations of coal dust, rock dust, and high-pressure water mist sprayed for dust suppression. This harsh visual environment can cause the lenses of optical sensors to be blocked or the light path to be scattered, thus blinding the monitoring system and making it impossible to provide stable distance measurement data during continuous production.

[0004] To overcome the problem of obstructed visibility, although some technologies have begun to utilize radar sensors with strong penetration capabilities, they still face significant technical bottlenecks in practical applications. Firstly, both the coal mining machine drum and the hydraulic support sidewalls are large metal components located very close to each other. In the radar echo signals, their signal characteristics are highly similar and often overlap, making it difficult for existing signal processing methods to accurately distinguish which part belongs to the drum and which to the sidewalls in the cluttered point cloud data. This challenge of identifying multiple targets of the same material at close range often prevents existing systems from calculating the exact distance between the two moving parts, leading to false alarms or missed detections. Furthermore, the underground coal face and rock strata generate significant background clutter interference, further reducing the accuracy of identification.

[0005] Existing anti-collision systems typically employ a fixed alarm zone setting method. However, the cutting height of a coal mining machine drum is adjusted by the swing of its rocker arm, and its spatial position changes dynamically in real time. Traditional fixed monitoring zones cannot adaptively adjust to follow the rocker arm's movements. When the rocker arm is raised or lowered significantly, the drum may move out of the monitoring zone, creating a blind spot, or mistakenly enter a non-dangerous area, triggering a false alarm. Current technology lacks a precise monitoring solution that combines the equipment's own mechanical motion status with environmental interference resistance detection capabilities, making it difficult to meet the efficient and safe active protection requirements of fully mechanized mining faces. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a millimeter-wave radar imaging anti-collision monitoring device and method for drum side guard plates. This solves the problem that existing monitoring methods are unable to accurately distinguish the spatial identities of the coal mining machine drum and the hydraulic support side guard plates under complex working conditions such as high dust cover and interference from similar metal materials at close range in fully mechanized mining faces, thus making it difficult to accurately determine the distance between the two and effectively prevent equipment collisions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides a millimeter-wave radar imaging anti-collision monitoring device for roller side guards.

[0008] The device constructs a monitoring architecture based on dual verification of physical characteristics and spatial logic. Among them, the millimeter-wave radar detection front end is deployed as a sensing unit above the hydraulic support. It is responsible for transmitting frequency-modulated continuous wave signals into the fully mechanized mining area. Utilizing the propagation characteristics of millimeter waves in the medium, it acquires three-dimensional raw point cloud data containing distance, angle, and reflection intensity information.

[0009] To achieve intelligent data processing, the device is equipped with a multi-dimensional signal processing host. This host integrates a signal preprocessing module, a dynamic RoI (Region of Interest) construction and classification module, and a minimum safe distance calculation module. Simultaneously, the device establishes a data path with the coal mining machine control system through the fully mechanized mining equipment communication interface, acquiring real-time data on the swing angle of the coal mining machine's rocker arm.

[0010] At the logic execution level, the multi-dimensional signal processing host first uses the signal preprocessing module to perform preliminary screening of the material dimension. Based on the difference in dielectric constant of different materials, a reflection intensity threshold is set, and background noise of coal walls and rocks with reflection intensity lower than the threshold is filtered out, retaining only the set of metal target points with high reflection characteristics.

[0011] Furthermore, the dynamic ROI construction and classification module performs fine-tuning of the spatial dimension. Based on the real-time acquired rocker arm swing angle data and the inherent mechanical geometry of the coal mining machine, this module reconstructs the real-time spatial position of the drum in the radar coordinate system and generates a dynamic region of interest (ROI) that follows the drum's movement. By determining whether each point in the metal target point set falls within this dynamic region, the system can accurately segment the mixed metal point cloud into a set of moving target points belonging to the drum and a set of stationary / slow-moving target points belonging to the side guard plate.

[0012] Finally, the minimum safe distance calculation module calculates the minimum Euclidean distance between the two sets of points after segmentation, and generates corresponding early warning or control commands according to the preset safety strategy.

[0013] Preferably, to ensure the coverage accuracy of the dynamic region, the dynamic ROI construction and classification module uses a kinematic transformation algorithm to construct the dynamic region of interest. Specifically, the system uses the radar installation position as the origin, and combines the spatial offset vector of the rocker arm hinge point, the rocker arm length vector, and the rotation matrix generated by the real-time swing angle to calculate the precise three-dimensional coordinates of the drum's geometric center at the current moment. Subsequently, using this geometric center as the sphere's center and the drum's physical radius superimposed with a preset safety redundancy boundary (used to compensate for mechanical vibration and measurement errors) as the radius, a spherical or polyhedral three-dimensional spatial envelope is constructed as the dynamic region of interest.

[0014] Preferably, the point cloud segmentation logic implemented by the dynamic ROI construction and classification module is as follows: the system performs spatial position verification on each retained metal target point. Points whose coordinates fall within the dynamic region of interest are determined to be reflection points on the surface of the coal mining machine drum; points whose coordinates are located outside the dynamic region of interest and whose spatial orientation conforms to the distribution characteristics of the hydraulic support sidewall plate (e.g., located on a specific side of the radar field of view) are determined to be reflection points of the sidewall plate.

[0015] Preferably, the millimeter-wave radar detection front-end operates in the 77GHz to 81GHz frequency band to utilize the short wavelength and high resolution characteristics of this band to capture subtle distance changes. The reflection intensity threshold is configured as the signal intensity boundary value that distinguishes between coal and rock media and metallic media, thereby eliminating non-metallic interference at the data source.

[0016] In one specific embodiment, the device also integrates an alarm execution unit and constructs a hierarchical safety protection mechanism. The minimum safe distance calculation module is internally configured with a primary warning threshold and a secondary interlocking threshold. When the monitored minimum Euclidean distance shortens to between the primary warning threshold and the secondary interlocking threshold, the system drives the alarm execution unit to issue an audible and visual warning signal; once the distance further shortens and falls below the secondary interlocking threshold, the system immediately issues a speed limit or stop command to the coal mining machine through the fully mechanized mining equipment communication interface, and simultaneously issues a forced retraction command to the hydraulic support for the side guard plate, thereby achieving active interlocking protection.

[0017] A second aspect of this invention provides a millimeter-wave radar imaging-based collision avoidance monitoring method for roller side guards. This method, applied to the aforementioned device, mainly includes the following logical steps: S1. Data Acquisition and Mapping: Control the millimeter-wave radar detection front end to scan the working surface and collect echo signals. Since the original radar data is in polar coordinate format, this step uses a coordinate transformation algorithm to map the range, azimuth, and elevation angle data to a Cartesian coordinate system with the radar as the origin, forming spatial point cloud data.

[0018] S2. Initial screening of material features: Traverse the spatial point cloud data and extract the reflection intensity features of each target point. Utilize the physical characteristics of coal and rock—high absorption rate and low reflectivity of electromagnetic waves—to filter out invalid points with reflection intensity below a preset threshold, generating a set of metal target points containing only equipment information.

[0019] S3. Kinematic Model Calculation: Real-time reading of the swing angle sensor data of the coal mining machine's rocker arm, combined with the mechanical dimensional parameters of each component of the coal mining machine, using rigid body kinematic equations to calculate the theoretical spatial position of the drum in the radar coordinate system, and generating a dynamic region of interest spatial threshold that follows the drum's movement in real time.

[0020] S4. Spatial Feature Matching and Classification: Using the generated dynamic region of interest spatial threshold as a spatial screen, the metal target point set is traversed and matched. Points falling within the spatial threshold are marked as roller point clouds, and points outside the threshold but conforming to the support position constraints are marked as side panel point clouds, thereby completing the identification of multiple targets of the same material.

[0021] S5. Distance Calculation and Control: The minimum distance algorithm between point sets is used to calculate the minimum geometric distance between the roller point cloud and the side plate point cloud. This distance is then compared with a safety threshold to trigger the corresponding hierarchical early warning control strategy.

[0022] Preferably, in step S1, the coordinate transformation is achieved through trigonometric function projection relationships to ensure that the spatial position sensed by the radar maintains geometric consistency with the mechanical coordinate system of the coal mining machine.

[0023] Preferably, in step S3, the process of calculating the spatial position of the roller is performed by vector chain calculation: first, the fixed offset vector of the rocker arm hinge point relative to the radar center is determined, and then a rotation transformation matrix is ​​constructed based on the real-time swing angle. The rocker arm length vector is rotated and transformed and then superimposed with the offset vector to obtain the real-time three-dimensional coordinates of the roller's geometric center.

[0024] Preferably, in step S4, the dynamic region of interest spatial threshold is defined as a geometric space whose center coordinates translate in real time as the rocker arm swings. Its spatial dimensions are constrained by the physical dimensions of the drum and a preset mechanical vibration tolerance to ensure that the drum body can still be stably enveloped under the condition of severe vibration of the coal mining machine.

[0025] Preferably, in step S5, the minimum distance is calculated using a traversal algorithm, which calculates the Euclidean distance between any point in the roller point cloud set and any point in the side panel point cloud set, and selects the minimum value as the current real-time distance to reflect the proximity of the most dangerous point.

[0026] This invention provides a millimeter-wave radar imaging-based anti-collision monitoring device and method for roller side guards. It has the following beneficial effects: 1. This invention introduces a kinematic model of a coal mining machine to construct a dynamic region of interest that is updated in real time with the swing angle of the rocker arm. By utilizing the difference between the mechanical structure constraint on the spatial position of the drum and the relatively independent position of the side guard plate, the invention accurately achieves the identification of the drum point cloud and the side guard plate point cloud among a group of similar metal targets detected by radar. This effectively solves the problem that existing technologies rely on a single sensor to make it difficult to distinguish targets of the same material at close range.

[0027] 2. This invention employs a 77GHz-81GHz band millimeter-wave radar combined with a reflectivity threshold screening mechanism. By utilizing the strong penetrating power of millimeter waves and the difference in dielectric constant between coal and metal, it effectively eliminates interference signals from high-concentration dust, water mist, and coal wall background in the mine. Combined with spatial threshold filtering, it further reduces false echoes from other non-critical metal components of the machine body, ensuring the purity and accuracy of monitoring data under harsh working conditions.

[0028] 3. This invention is based on the minimum Euclidean distance calculation from point cloud to point cloud and a hierarchical interlocking control strategy. The system can respond to minute distance changes between the roller and the side guard plate in milliseconds, and the monitoring area is adjusted in real time with the rocker arm movement, eliminating blind spots that may occur when the rocker arm is raised significantly in traditional fixed monitoring areas; through a dual protection mechanism of primary audible and visual alarm and secondary equipment linkage, reliable active obstacle avoidance is achieved while ensuring continuous production. Attached Figure Description

[0029] Figure 1This is a flowchart of the signal processing logic of the multidimensional signal processing host of the present invention; Figure 2 This is a schematic diagram illustrating the dynamic ROI construction and target classification logic of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] This invention provides a millimeter-wave radar imaging anti-collision monitoring device for drum side guard plates. This device is applied to fully mechanized mining faces and is configured to monitor the spatial positional relationship between the cutting drum of the coal mining machine and the side guard plates of the hydraulic support in real time.

[0032] The millimeter-wave radar imaging collision avoidance monitoring device for drum side guards mainly includes: a millimeter-wave radar detection front end, a multi-dimensional signal processing host, a communication interface for the fully mechanized mining equipment, and an alarm execution unit. The millimeter-wave radar detection front end is connected to the multi-dimensional signal processing host via a high-speed data cable. The multi-dimensional signal processing host is electrically connected to both the communication interface for the fully mechanized mining equipment and the alarm execution unit.

[0033] The millimeter-wave radar detection front end is rigidly mounted on the body of the coal mining machine. Specifically, it is installed on the side of the machine facing the coal face, avoiding the rotating hinge structure of the machine's rocker arm, and maintaining a stationary state relative to the machine body. The detection field of view of the millimeter-wave radar front end covers the working area of ​​the coal mining machine drum and the extension / retraction area of ​​the adjacent hydraulic support sidewall plate.

[0034] Considering the high-intensity vibration environment during coal mining machine operation, the millimeter-wave radar detection front end is fixed to the coal mining machine body via a vibration-damping mounting bracket. The vibration-damping mounting bracket adopts a sandwich structure of metal, rubber, and metal, with the middle layer filled with nitrile rubber vibration damping pads with a damping coefficient greater than 0.5.

[0035] In addition, the radome surface of the millimeter-wave radar detection front end is coated with a nano-hydrophobic coating. This nano-hydrophobic coating is configured to prevent water films generated by downhole spray dust suppression from adhering to the radar surface, thereby reducing the attenuation and phase distortion of the millimeter-wave signal caused by the water film and ensuring detection stability in long-term high-humidity environments.

[0036] The millimeter-wave radar detection front-end is configured as a frequency-modulated continuous wave radar sensor, with its operating center frequency set in the 77GHz to 81GHz band. The millimeter-wave radar detection front-end is used to transmit linear frequency-modulated electromagnetic wave signals towards the target area and receive the echo signals reflected by the target. The front-end integrates an analog-to-digital converter unit to convert the echo signals into digital intermediate frequency signals and output raw point cloud data containing target range, azimuth, elevation angle, and reflection intensity.

[0037] The multidimensional signal processing host is the core computing unit of this device, and its hardware carrier adopts an embedded computing platform that includes FPGA (Field Programmable Gate Array) and DSP (Digital Signal Processor). Internally, the multidimensional signal processing host is logically configured with a signal preprocessing module, a dynamic ROI construction and classification module, and a minimum safe distance calculation module.

[0038] The signal preprocessing module is configured to receive raw point cloud data and perform binarization filtering on the data according to a preset reflection intensity threshold. This module is configured to remove non-metallic interference points with reflection intensity below the threshold and retain the set of metallic target points with reflection intensity higher than or equal to the threshold.

[0039] The dynamic ROI construction and classification module is configured to perform spatial segmentation of targets. This module receives real-time swing angle data of the coal mining machine's rocker arm from the communication interface of the fully mechanized mining equipment, and combines it with the coal mining machine's mechanical structure parameters (including rocker arm length, radar installation offset, and drum diameter) stored in the host computer. In a three-dimensional coordinate system with the radar installation center as the origin, it calculates the geometric center coordinates of the drum in real time.

[0040] The dynamic ROI construction and classification module is further configured to generate a dynamic region of interest that spatially translates with the rocker arm movement, based on the calculated coordinates of the drum's geometric center. This module performs spatial location matching between data points in the metal target point set and the dynamic region of interest: points falling into the region are marked as drum point sets, and points falling outside the region but conforming to the spatial distribution characteristics of the side guard plate are marked as side guard plate point sets.

[0041] The minimum safe distance calculation module is configured to calculate the minimum Euclidean distance between the roller point set and the side guard plate point set. This module internally stores a primary warning threshold and a secondary locking threshold, which are used to compare the calculated minimum Euclidean distance with the thresholds to generate corresponding status control signals.

[0042] The communication interface of the fully mechanized mining equipment integrates a CAN bus controller and an industrial Ethernet interface. This interface is configured to connect to the coal mining machine's body control system to periodically read the angle sensor values ​​of the coal mining machine's rocker arm and transmit these values ​​to the multi-dimensional signal processing host. Simultaneously, this interface is configured to receive control commands from the multi-dimensional signal processing host and forward these commands to the coal mining machine control system to execute speed limiting or shutdown operations, or forward them via the fully mechanized mining face network to the hydraulic support electro-hydraulic control system to execute the retraction of the side guard plates.

[0043] The alarm execution unit includes an explosion-proof audible and visual alarm installed on the coal mining machine. This alarm execution unit receives drive signals from the multi-dimensional signal processing host and issues an audible and visual warning when the minimum Euclidean distance is less than the first-level warning threshold.

[0044] In one embodiment of the present invention, the multi-dimensional signal processing host establishes a radar Cartesian coordinate system. The origin of this coordinate system is defined at the phase center of the millimeter-wave radar detection front end. The X-axis of this coordinate system is parallel to the longitudinal axis of the coal mining machine body, the Y-axis is perpendicular to the coal wall direction, and the Z-axis is perpendicular to the bottom plate direction. All spatial position calculations regarding the drum and side guards are completed within this coordinate system, ensuring the spatial reference consistency of the monitoring data.

[0045] The multi-dimensional signal processing host is also equipped with a parameter calibration working mode. The system enters this working mode after the initial installation or position adjustment of the device. The operator adjusts the coal mining machine rocker arm to the horizontal zero position and places a standard corner reflector with high reflectivity at the geometric center of the drum.

[0046] The millimeter-wave radar detection front end detects the corner reflector and obtains its measured coordinates in the radar coordinate system. The multi-dimensional signal processing host compares these measured coordinates with the theoretically calculated coordinates of the roller center, calculates the installation error vector, and stores this installation error vector as a correction compensation parameter in non-volatile memory. In subsequent real-time monitoring, the dynamic ROI construction and classification module automatically calls this correction compensation parameter to make corrections, thereby eliminating the impact of mechanical installation tolerances on monitoring accuracy.

[0047] In one embodiment of the present invention, the swing angle data of the rocker arm of the coal mining machine is obtained by an angle encoder or tilt sensor installed at the rocker arm hinge shaft. The mechanical structural parameters stored in the multi-dimensional signal processing host include: the three-dimensional offset vector of the rocker arm hinge point relative to the origin of the radar coordinate system, the physical length of the rocker arm, and the physical radius of the drum. These parameters are calibrated and stored in the memory during the device installation and commissioning phase.

[0048] See attached document Figure 1The system logic architecture of this invention mainly consists of three consecutive processing stages: radar raw data acquisition, spatial coordinate transformation, and background filtering based on reflectivity.

[0049] The multi-dimensional signal processing host first controls the millimeter-wave radar detection front-end to transmit a frequency-modulated continuous wave (FMCH) signal. The frequency of this FMCH signal is linearly modulated over time. When the transmitted signal encounters a target and is reflected back to the receiving antenna, an echo signal is generated. The radar front-end mixes the transmitted signal and the echo signal using a mixer to obtain a beat signal. By performing analog-to-digital conversion and fast Fourier transform on the beat signal, the peak frequency in the spectrum is extracted. This peak frequency is proportional to the radial distance to the target. Simultaneously, using the phase difference principle of the antenna array, the azimuth and elevation angles of the target are calculated. Thus, the system generates a set of raw measurement data in each sampling period. This raw measurement data, in polar coordinates, represents the spatial distribution and signal strength of all detection points within the field of view.

[0050] After acquiring the raw measurement data, the multidimensional signal processing host executes spatial coordinate transformation logic. Since the raw data is based on the spherical coordinate system of the radar sensor, while subsequent kinematic calculations of the coal mining machine are based on the Cartesian coordinate system, spatial mapping is necessary. The system constructs a three-dimensional Cartesian coordinate system with the phase center of the millimeter-wave radar detection front end as the origin. It is assumed that the radar's line of sight is the Y-axis (pointing towards the coal face), the horizontal axis perpendicular to the line of sight is the X-axis (along the direction of the coal mining machine body), and the axis perpendicular to the floor is the Z-axis.

[0051] The system utilizes a trigonometric projection algorithm to map each target point from the polar coordinate system to a three-dimensional Cartesian coordinate system. For any point in the original data... The spatial location transformation calculation for each target point is performed based on the following mathematical relationship: ; In the formula, Indicates the first The coordinates of each target point in a three-dimensional Cartesian coordinate system; Indicates the first The radial distance of each target point relative to the radar origin; Indicates the first The horizontal azimuth angle of a target point relative to the radar line of sight; Indicates the first The elevation angle of a target point relative to the radar horizontal plane.

[0052] After coordinate transformation, the signal preprocessing module performs background clutter filtering based on material properties. In the fully mechanized mining face environment, radar echoes not only contain signals from the coal mining machine drum and hydraulic support sidewalls, but also a large amount of clutter signals from the coal face, roof strata, and airborne coal dust. Utilizing the physical differences between metallic and coal / rock materials in terms of dielectric constant and electromagnetic wave reflectivity (metals exhibit total reflection of millimeter waves, while coal / rock materials partially absorb and transmit millimeter waves), the system sets a specific reflection intensity threshold.

[0053] The processing procedure specifically includes: the signal preprocessing module traverses all spatial point cloud data after coordinate transformation and reads the reflection intensity value of each target point. The system then compares this reflection intensity value with a preset reflection intensity threshold.

[0054] If the reflection intensity value of a target point is less than the reflection intensity threshold, the system determines that the point is non-metallic background noise (such as coal wall or dust) and removes it from the dataset, excluding it from subsequent calculations.

[0055] If the reflection intensity value of a target point is greater than or equal to the reflection intensity threshold, the system determines that the point is a potential metal device target and stores it in the metal target point set cache. Through this step, the system eliminates more than 90% of environmental background interference at the data source, ensuring that subsequent kinematic analysis is only performed on key metal device components, thereby significantly reducing computing power consumption and improving the robustness of recognition.

[0056] See attached document Figure 2 This embodiment details how the dynamic ROI construction and classification module in the multidimensional signal processing host utilizes the principles of mechanical kinematics to construct a following spatial threshold in the radar coordinate system and accurately divide the metal target point set into the roller point set and the side plate point set.

[0057] The dynamic ROI construction and classification module first performs real-time calculation of the drum's geometric center. Since the millimeter-wave radar detection front end is rigidly fixed to the coal mining machine body, and the cutting drum is installed at the end of an adjustable-height rocker arm, the relative spatial position between the two changes dynamically with the rocker arm's swing angle. The system pre-configures the key mechanical dimensional parameters of the coal mining machine in its memory. These parameters are obtained and solidified through high-precision measurements during the equipment's factory delivery or installation and commissioning phases. They mainly include: the spatial offset vector of the rocker arm's hinge axis center relative to the radar phase center, the physical length of the rocker arm, and the reference direction vector when the rocker arm is at zero position.

[0058] Within each signal processing cycle, the system reads the current rocker arm swing angle value through the communication interface of the fully mechanized mining equipment. Based on the principles of rigid body kinematics, the system establishes a vector transfer chain starting from the origin of the radar coordinate system, passing through the rocker arm hinge point, and finally reaching the geometric center of the drum. This calculation process does not rely on radar detection data but is purely based on the geometric deduction of the mechanical structure, and therefore is unaffected by dust or electromagnetic interference. The formula for calculating the real-time three-dimensional coordinates of the drum's geometric center is as follows: ; in, express The three-dimensional coordinate vector of the drum's geometric center in the radar Cartesian coordinate system, calculated at each moment; This represents the fixed position vector of the center of the rocker arm hinge point relative to the origin of the radar coordinate system. This represents the length vector of the rocker arm, whose modulus is equal to the straight-line distance from the rocker arm hinge axis to the roller axis, and whose direction is defined as the reference direction when the rocker arm is at a zero-degree angle; express Real-time swing angle data of the rocker arm is constantly acquired through sensors; Indicates based on swing angle data A three-dimensional rotational transformation matrix is ​​constructed to describe the rotational motion of the rocker arm about its hinge axis.

[0059] After obtaining the precise coordinates of the drum's geometric center, the dynamic ROI construction and classification module constructs a dynamic region of interest (ROI). This dynamic ROI is defined as a spherical space centered on the calculated drum's geometric center (or a cylindrical envelope space defined according to the drum's shape). To compensate for the severe mechanical vibrations during coal mining operations, the minute displacements caused by the rocker arm pin clearance, and the systematic errors in radar ranging, the physical dimensions of this spatial region are superimposed with a safety redundancy boundary on top of the actual drum dimensions. The mathematical description of the dynamic ROI is defined as a set of points satisfying the following spatial distance constraints: ; in, express A dynamically constructed set of regions of interest at any given time; Represents the coordinates of any point in space within the radar coordinate system; Indicates the physical radius of the cutting drum; This represents a preset safety redundancy boundary constant, which is typically set to 3 to 5 times the radar range resolution. This represents the Euclidean norm, i.e., spatial distance.

[0060] Based on the constructed dynamic ROI, the system performs a classification operation on the metal target point set. Since the preprocessing module has filtered out the non-metallic background, the current input point set mainly consists of reflection points from the roller surface, reflection points from the side guard plate surface, and a small number of reflection points from other metal parts of the machine body. The classification logic employs a strategy combining inclusion and exclusion checks.

[0061] The system iterates through each target point in the current input point set, first calculating the distance from that point to the geometric center of the drum at the current moment. If this distance is less than or equal to the set radius of the dynamic ROI, the system determines that the point belongs to the coal mining machine drum, marks it, and adds it to the drum point set.

[0062] For points not falling within the dynamic ROI space, the system further performs secondary discrimination based on the spatial orientation constraints of the radar field of view. Since the hydraulic support sidewalls are always located above or directly in front of the coal mining machine drum in the physical structure of the longwall mining face (depending on the radar installation angle), the system presets a spatial orientation filter mask. If a point falling outside the dynamic ROI simultaneously satisfies this spatial orientation constraint, the system determines that the point belongs to the hydraulic support sidewall, marks it, and adds it to the sidewall point set.

[0063] Through the aforementioned spatial dynamic filtering based on the kinematic model, the system can achieve stable separation of the point clouds of the roller and the side guard plate under the working conditions of two types of targets with identical materials, extremely close distances, and a large amount of dust interference, providing an accurate independent target source for subsequent distance measurement.

[0064] This embodiment details how a multi-dimensional signal processing host assesses collision risk based on segmented point cloud data and executes a tiered active protection strategy.

[0065] The minimum safe distance calculation module receives two independent point cloud sets output from the preceding dynamic ROI construction and classification module: a point set representing the drum of the coal mining machine cutting section and a point set representing the side guard plate of the hydraulic support. Since the target surface detected by radar is a discrete, irregular three-dimensional point cloud, and the relative attitude of the drum and side guard plate changes in real time with the swing of the rocker arm and the movement of the support, collision risk cannot be simply characterized by the centroid distance. Therefore, this embodiment uses the minimum Euclidean distance algorithm between point sets to quantify the physical distance.

[0066] The system performs either a full traversal or a distance search based on spatial indexing. Specifically, the system traverses every spatial point in the roller point set and every spatial point in the side panel point set, calculating all possible point pairs. The Euclidean distance between them is calculated, and the minimum value is selected from all the calculation results as the current real-time minimum safe distance. The mathematical expression of this calculation logic is as follows: ; In the formula, This indicates the closest physical distance between the roller surface and the side guard plate surface at the current sampling moment; Represents any three-dimensional coordinate point in the roller point set; Represents any three-dimensional coordinate point in the set of points on the side panel; Point With point The geometric straight-line distance between them.

[0067] In a preferred embodiment, to reduce the time complexity of minimum Euclidean distance calculation, the minimum safe distance calculation module is configured with spatial index acceleration logic. Before performing point-to-point traversal calculation, the system first constructs a spatial index structure for the large set of guardrail points based on the KD-Tree algorithm or octree algorithm.

[0068] The specific accelerated computation process is as follows: The system traverses each query point in the roller point set and performs a nearest neighbor search in the side panel point set using a spatial index structure. During the search, the system utilizes a tree-structured bounding box pruning strategy to quickly eliminate spatial branches that cannot contain the nearest point, thereby reducing computational complexity and ensuring that the real-time response requirement of less than 50ms can still be met even when the point cloud density is high.

[0069] Calculated The system then proceeds to the hierarchical comparison logic. The minimum safe distance calculation module has two preset levels of safety threshold parameters: a primary warning threshold and a secondary locking threshold. These two thresholds are set according to the principle that the primary warning threshold > the secondary locking threshold.

[0070] In a specific application example, the first-level warning threshold is set to 500 mm to delineate the warning buffer zone; the second-level locking threshold is set to mm to delineate the absolute danger zone.

[0071] System basis The comparison result with the above threshold generates a corresponding state control signal and triggers the following hierarchical response mechanism: when When the Level 1 warning threshold is reached, the system determines that the current operation is safe. At this time, the multi-dimensional signal processing host maintains a silent monitoring mode, does not send intervention commands, and renders the current distance value in real time as a green icon on the local display terminal.

[0072] When the secondary locking threshold < When the threshold for Level 1 warning is reached, the system determines that it is in a critical approach state and triggers a Level 1 warning response. The multi-dimensional signal processing host outputs a high-level drive signal to the alarm execution unit through the I / O port, which drives the machine body's audible and visual alarm to emit a flashing yellow light and intermittent buzzing signal, prompting the coal mining machine operator to pay attention and reduce the traction speed.

[0073] when When the threshold for secondary interlocking is ≤, the system determines that a collision is imminent and immediately triggers the secondary interlocking response. The multi-dimensional signal processing host performs bidirectional concurrent control through the communication interface of the fully mechanized mining equipment: on the one hand, it sends high-priority CAN bus commands to the coal mining machine control system. These CAN bus commands contain function codes for stopping traction or emergency stop of the rocker arm, forcing the coal mining machine to immediately stop its tendency to move towards the side guard plate. On the other hand, it sends a side guard plate retraction action command to the electro-hydraulic control system of the hydraulic support. This action command forces the controller of the target support to drive the solenoid valve to actuate, quickly folding and retracting the side guard plate, thereby actively increasing the spatial distance with the drum.

[0074] During the execution of the level 2 interlock response, the system continues to... The calculation and monitoring. Only when the calculated... Only after the distance exceeds the secondary lockout threshold and a certain time lag (e.g., 500ms debounce time) is maintained will the system release the lockout command, allowing the equipment to resume normal operation, thus forming a closed-loop safety reset logic. Through this hard real-time distance calculation and bidirectional control, the system blocks the path of physical collisions within a millisecond timescale.

[0075] This embodiment describes in detail the complete closed-loop process of the aforementioned monitoring device performing collision avoidance monitoring at the fully mechanized mining face, using a time-series logic approach. This process is a periodic, real-time processing task, with each monitoring cycle typically controlled within milliseconds to meet the control response requirements of rapidly moving equipment in industrial settings.

[0076] Step S1: Radar data acquisition and spatial mapping.

[0077] At the start of the monitoring cycle, the multi-dimensional signal processing host sends a trigger command to the millimeter-wave radar detection front end. The radar front end transmits a frequency-modulated continuous wave signal to scan the area of ​​the coal mining machine drum and hydraulic support, and acquires echo signals. The digital signal processor inside the radar front end performs FFT transformation and constant false alarm rate (CFAR) detection on the echo signals, and outputs raw point cloud data. Since the raw radar detection data is based on a polar coordinate system (including radial distance, azimuth, elevation angle, and reflection intensity), the multi-dimensional signal processing host immediately calls the coordinate transformation algorithm. This step uses trigonometric function projection relationships to map the coordinates of all detection points from the radar spherical coordinate system to a three-dimensional Cartesian coordinate system with the radar phase center as the origin, forming spatial point cloud data with a unified measurement benchmark.

[0078] Step S2: Material feature screening based on reflectivity.

[0079] To extract effective targets from the complex downhole environment, the system performs an initial screening based on material properties. The signal preprocessing module iterates through each point in the spatial point cloud data generated in step S1, extracting its reflection intensity value. The system then compares this reflection intensity value point-by-point with a preset reflection intensity threshold.

[0080] The logical judgment is as follows: if the reflection intensity value is less than the reflection intensity threshold, the system determines that the point is a low reflectivity clutter generated by coal wall, rock or suspended coal dust, and discards it directly; If the reflection intensity value is greater than or equal to the reflection intensity threshold, the system determines that the point is a metallic reflection point, retains it, and adds it to the set of metallic target points. This step ensures that the data input to subsequent algorithms only contains information about key equipment such as the coal mining machine and hydraulic supports, significantly reducing the amount of computation.

[0081] Step S3: Kinematic modeling and dynamic ROI generation.

[0082] While processing point cloud data, the system reads the current rocker arm swing angle data from the coal mining machine control system via the fully mechanized mining equipment communication interface. The dynamic ROI construction and classification module retrieves mechanical structure parameters (rocker arm length, installation offset, etc.) from memory and uses rigid body kinematics equations to calculate the three-dimensional coordinates of the drum's geometric center in the radar coordinate system in real time. Subsequently, the system constructs a dynamic region of interest in the shape of a sphere or polyhedron in a virtual three-dimensional space, using this three-dimensional coordinate as the center and the sum of the drum's physical radius and the safety redundancy boundary as the radius. The spatial position of this dynamic region of interest is updated in real time with each swing of the rocker arm, always maintaining its envelope over the physical entity of the drum.

[0083] Step S4: Spatial feature matching and point cloud classification.

[0084] The system uses the dynamic ROI constructed in step S3 as a spatial filter to perform refined classification of the metal target point set generated in step S2. The system traverses each point in the metal target point set, calculating its Euclidean distance to the three-dimensional coordinates of the drum's geometric center. Points with a distance less than the radius of the dynamic ROI are marked as drum point clouds and included in the metal target point set; points with a distance greater than the radius of the dynamic ROI and whose spatial coordinates are located on a specific side of the radar field of view (i.e., the side where the hydraulic support is located) are marked as sidewall point clouds and included in the sidewall point set. Through this process, the system successfully separates two independent moving entities from the metal point cloud of the same material.

[0085] Step S5: Distance calculation and hierarchical active protection.

[0086] After point cloud classification, the minimum safe distance calculation module performs a collision risk assessment. The system uses a point-to-point traversal algorithm to calculate the minimum Euclidean distance between all points in the metal target point set and all points in the sidewall point set. The system will Logically compare the results with the system's preset first-level warning threshold and second-level blocking threshold: like The first-level warning threshold is reached, and the system maintains normal monitoring status and enters the next processing cycle; If the secondary locking threshold is < If the value is ≤ Level 1 warning threshold, the system determines that there is a potential risk and controls the alarm execution unit to issue an audible and visual alarm signal to alert the operator. like If the value is less than or equal to the secondary interlocking threshold, the system determines that a collision is imminent and immediately sends a stop or speed limit command to the coal mining machine via the communication interface, and sends a side guard plate retraction command to the hydraulic support, until the calculated value is reached. Return to a safe range.

[0087] After completing step S5, the system automatically resets all temporary variables and immediately jumps back to step S1 to start the next cycle of sampling and processing, thereby achieving continuous, all-time, blind-spot-free monitoring of collision prevention for the fully mechanized mining face equipment.

Claims

1. A millimeter-wave radar imaging anti-collision monitoring device for roller side guards, characterized in that, include: The millimeter-wave radar detection front end is mounted on a hydraulic support and is used to transmit frequency-modulated continuous wave signals to the target area and receive echoes, outputting raw point cloud data including range, azimuth, elevation angle and reflection intensity. A multi-dimensional signal processing host, which is connected to the millimeter-wave radar detection front end, is equipped with a signal preprocessing module, a dynamic ROI construction and classification module, and a minimum safe distance calculation module; The communication interface of the fully mechanized mining equipment is used to acquire real-time swing angle data of the rocker arm of the coal mining machine and send control commands; The multidimensional signal processing host executes the following logic: The signal preprocessing module filters the original point cloud data according to the reflection intensity threshold, removing non-metallic interference points with a reflection intensity lower than the reflection intensity threshold, and retaining the set of metallic target points; The dynamic ROI construction and classification module calculates the spatial position of the drum in real time under the radar coordinate system based on the real-time swing angle data and the mechanical structure parameters of the coal mining machine, constructs a dynamic region of interest, and divides the metal target point set into the drum point set and the side plate point set. The minimum safe distance calculation module calculates the minimum Euclidean distance between the roller point set and the side guard plate point set, and generates an early warning or control command based on the preset safe distance threshold.

2. The millimeter-wave radar imaging anti-collision monitoring device for roller side guards according to claim 1, characterized in that, The specific method by which the dynamic ROI construction and classification module constructs dynamic regions of interest is as follows: Based on the radar installation location, the spatial coordinate offset of the rocker arm hinge point is determined. Combined with the rocker arm length and the real-time swing angle data, the spatial coordinates of the drum geometric center at the current moment are calculated using kinematic transformation relationships. Using the geometric center of the roller as a reference, and based on the physical radius of the roller and superimposed with a safety redundancy boundary, a three-dimensional spatial range enclosing the roller is constructed as the dynamic region of interest.

3. The millimeter-wave radar imaging anti-collision monitoring device for roller side guards according to claim 2, characterized in that, The specific logic for segmenting the metal target point set by the dynamic ROI construction and classification module is as follows: Spatial location determination is performed on each target point in the set of metal target points; Target points falling within the dynamic region of interest are classified into the roller point set; Target points that fall outside the dynamic region of interest and are located on one side of the radar field of view are classified as the sidewall point set.

4. The millimeter-wave radar imaging anti-collision monitoring device for roller side guards according to claim 1, characterized in that, The millimeter-wave radar detection front-end operates in the frequency band of 77GHz to 81GHz; the reflection intensity threshold is set as the signal intensity critical value that distinguishes between coal and rock materials and metal materials, and only target data with metal reflection characteristics are retained.

5. The millimeter-wave radar imaging anti-collision monitoring device for roller side guards according to claim 1, characterized in that, It also includes an alarm execution unit, which is used for: The minimum safe distance calculation module is set with a first-level early warning threshold and a second-level locking threshold, wherein the first-level early warning threshold is greater than the second-level locking threshold. When the minimum Euclidean distance is between the first-level warning threshold and the second-level locking threshold, the alarm execution unit triggers an audible and visual alarm. When the minimum Euclidean distance is less than the secondary interlocking threshold, a speed limit command is sent to the coal mining machine control system through the fully mechanized mining equipment communication interface, and a side guard plate retraction command is sent to the hydraulic support control system.

6. A millimeter-wave radar imaging-based collision avoidance monitoring method for roller side guard plates, characterized in that, The millimeter-wave radar imaging collision avoidance monitoring device for roller side guards as described in any one of claims 1-5 includes the following steps: S1. Control the millimeter-wave radar detection front end to collect echo signals and convert the raw point cloud data in polar coordinate format into spatial point cloud data in Cartesian coordinate system; S2. Iterate through the reflection intensity of each target point in the spatial point cloud data, filter out points with a reflection intensity less than a preset threshold, and generate a set of metal target points; S3. Real-time reading of the rocker arm swing angle of the coal mining machine, combined with the kinematic parameters of the coal mining machine to calculate the current spatial position of the drum, and generate a dynamic region of interest spatial threshold; S4. Use the dynamic region of interest spatial threshold to perform spatial matching on the metal target point set, mark the points falling within the threshold as roller point cloud, and mark the points falling outside the threshold as side panel point cloud. S5. Calculate the minimum distance between the roller point cloud and the side guard plate point cloud, and perform hierarchical early warning control based on the calculation result.

7. The millimeter-wave radar imaging collision avoidance monitoring method for roller side guards according to claim 6, characterized in that, In step S1, converting the original point cloud data in polar coordinates to spatial point cloud data in Cartesian coordinates specifically includes: Based on the radial distance obtained by radar detection, and combined with the trigonometric projection relationship of azimuth and elevation angles, each target point is mapped to a three-dimensional rectangular coordinate system with the radar installation center as the origin.

8. The millimeter-wave radar imaging collision avoidance monitoring method for roller side guards according to claim 6, characterized in that, In step S3, calculating the current spatial position of the roller specifically includes: Obtain the fixed offset vector of the rocker arm hinge point relative to the radar center; Construct a rotation transformation matrix based on the real-time swing angle data; The vector corresponding to the length of the rocker arm is processed by the rotation transformation matrix and then superimposed with the fixed offset vector to determine the real-time three-dimensional coordinates of the geometric center of the roller.

9. The millimeter-wave radar imaging collision avoidance monitoring method for roller side guards according to claim 6, characterized in that, In step S4, the logic for generating the dynamic region of interest spatial threshold is as follows: Define a spherical or polyhedral spatial region. The center coordinates of the spherical or polyhedral spatial region change in real time following the geometric center of the roller calculated by S3. The spatial dimensions of the spherical or polyhedral spatial region are jointly determined by the physical dimension parameters of the roller and the preset mechanical vibration tolerance.

10. The millimeter-wave radar imaging collision avoidance monitoring method for roller side guards according to claim 6, characterized in that, The method for calculating the minimum distance described in S5 specifically includes: Traverse every point in the point cloud of the roller and every point in the point cloud of the side guard plate, and calculate the geometric distance between all point pairs; The minimum geometric distance among all point pairs is selected as the real-time distance between the roller and the side guard plate.

Images