A tunnel rock burst inspection method based on a gecko-like robot

Abstract

The application discloses a kind of based on imitative gecko robot's roadway rock burst inspection method.The imitative gecko robot is connected with communication transducing base station by radio frequency signal when carrying out rock burst inspection operation in the roadway where rock burst has occurred, and control command, feedback information and surrounding rock measurement data are exchanged;It is transmitted to integrated control console by Ethernet, and is shown in display in multimedia form after being calculated and analyzed by central processing unit.The inspection method includes continuously emitting radio frequency signal to surrounding rock wall surface and receiving echo signal, analyzing surrounding rock vibration energy in combination with electromagnetic wave Doppler effect, recording inspection wall surface state using stereo vision camera, calculating fracture area and depth when suspected fracture area is found, and using particle ball clustering algorithm to dynamically cluster analyze multidimensional characteristics, to realize the judgment of whether rock burst has recurrence risk.The application can comprehensively consider the precursor phenomenon and inducing factor of rock burst, realize unmanned, large-scale inspection of hard rock mine roadway rock burst disaster, and has the advantages of simple and efficient, high degree of automation.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology for rockburst disasters in hard rock mines, and in particular provides a method for rockburst inspection in roadways based on a gecko-inspired robot. Background Technology

[0002] As hard rock mining progresses to deeper levels, the stress level within the tunnels continuously increases, leading to frequent rockburst disasters. Especially beyond a certain depth, rockburst risk becomes a core challenge that mine design and production must address. Rockbursts are often highly destructive, posing a significant threat to the safety of workers and equipment. Rockburst disasters have become one of the key issues restricting the safe and efficient development and utilization of deep mineral resources. Currently, common methods for monitoring rockbursts in high-stress tunnels of hard rock mines include on-site observation, drill cuttings analysis, microgravity analysis, infrared thermography, and microseismic monitoring. These methods achieve the purpose of rockburst monitoring and early warning by directly or indirectly measuring the macroscopic deformation, stress concentration state, expansion phenomenon, and temperature changes of the surrounding rock. However, in actual operation, existing monitoring methods require significant manual intervention, exposing both workers and monitoring equipment to high-risk environments. Furthermore, existing monitoring methods only focus on key locations such as the surrounding rock of newly excavated tunnels or the working face, failing to conduct large-scale inspections of the entire tunnel and lacking the ability to monitor time-delayed rockbursts.

[0003] Therefore, how to solve the problem of rockburst monitoring in hard rock mines, especially when a rockburst has already occurred and personnel are afraid to enter the mine, and how to reduce the risk to workers, is an urgent problem that needs to be solved. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, this invention provides a method for rockburst inspection in tunnels based on a gecko-inspired robot. The robot conducts inspections of areas where rockbursts have occurred, monitoring and analyzing the vibration energy of the surrounding rock face based on the Doppler effect of electromagnetic waves. A stereoscopic vision camera collects and analyzes the fracture range and depth of the surrounding rock surface in real time. By combining the vibration energy index, fracture range, and depth index of the surrounding rock in the area, an assessment is made to determine whether there is a risk of recurrence of rockbursts. If a rockburst risk is identified, personnel access should be restricted until the risk is eliminated. This invention enables automated operation through remote control and automatic patrol, reducing manual intervention and expanding the scale of inspections, which is of great significance for the safe and efficient development of deep mineral resources.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for rockburst inspection in tunnels based on a gecko-inspired robot, characterized in that the rockburst inspection implementation steps include: S1, Deploy the inspection system: The inspection system includes... mA gecko-like robot n The system comprises one communication power base station and one integrated control console. The gecko-like robot is functionally divided into four parts: a perception module, a control module, a motion module, and an auxiliary module. These are deployed on the head, body, four legs, and tail of the gecko-like robot, respectively. The perception module consists of a stereo vision camera, an ultrasonic sensor, a radio frequency transceiver, and a radio frequency processor. The stereo vision camera is used to collect information on the surrounding rock and road conditions in the tunnel; the ultrasonic sensor is used for obstacle identification and distance measurement; the radio frequency transceiver is used to achieve non-contact measurement and wireless communication; and the radio frequency processor is used to amplify the power of the radio frequency signal and convert the radio frequency frequency. The integrated control console is deployed into the communication network according to a suitable topology, and the integrated control console and the gecko-like robot are then activated. S2, Set the inspection path; The inspection path is mainly set on the path of the area where rock bursts have occurred. Place the gecko-like robot on the wall at the starting point of the tunnel to be tested. Input the motion command or set the cruise path through the controller to control the gecko-like robot to move along the tunnel; Set and record the inspection speed of the gecko-like robot and the camera observation direction so that the central processor can calculate. Deploy 1 gecko-like robot for each path. S3 collects inspection data; the gecko-like robot's head sensing module continuously emits electromagnetic waves towards the surrounding rock face and receives the echoes. The echo signal data is modulated by a radio frequency processor, uploaded wirelessly to a communication transducer base station, and then uploaded to the integrated control console via an Ethernet link. The integrated control console analyzes and calculates the echo signals from each sensor using a central processing unit, determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves; it also records the condition of the inspected rock face using a stereo vision camera, and when a suspected fracture area is found, it calculates the area and maximum depth of the fracture area; and constructs a three-dimensional feature vector. (Vibrational energy, fracture area, and depth) and location tags were used to perform granular-ball clustering (GBCT) on the data. Cluster analysis is performed on the combined datasets to determine the likelihood of rockburst recurrence, and the results are output to the display in a multimedia format.

[0006] Furthermore, the method for determining the vibration energy, area, and maximum depth of the suspected fracture region in step S3 is as follows: The gecko-like robot uses a stereoscopic vision camera mounted on its head to continuously acquire images and perform three-dimensional scanning of the surrounding rock surface in the tunnel. Based on image recognition technology, it compares the current surrounding rock image with the most recently recorded corresponding position and shape image to identify newly formed, irregular fracture zones on the rock wall. When a suspected fracture zone is identified, the following calculation steps are performed: Based on the boundary coordinates of the fracture zone, the system queries the vibration energy of all measurement points within a 0.5-meter range extending outward along its normal direction, and takes the maximum value as the regional vibration energy associated with the rockburst crater. Reference points are selected at the outer boundary of the fractured area. These reference points form a two-dimensional point set and constitute a closed boundary surrounding the fractured area. The area of ​​the region enclosed by this boundary is calculated using the integral method, and this area value is taken as the fractured area of ​​the rock mass.

[0007] The depth of the fractured area is measured using an ultrasonic sensor, and the maximum value is taken as the maximum fracture depth.

[0008] Furthermore, in step S3, the specific steps for analyzing the area and maximum depth of the fractured region and the three-dimensional feature vector of vibration energy using a particle clustering algorithm to determine the possibility of rockburst recurrence are as follows: ① During the algorithm initialization phase, the parameter set is read, and the data is preprocessed using the Z-score standardization method. For any feature in the data, the Z-score standardization formula is: , in, , These represent the original data size and the standardized size, respectively. This represents the mean of the corresponding feature parameters in the dataset. is the standard deviation of the corresponding feature parameters in the dataset.

[0009] ② Define an initial sphere, whose characteristics include: Particle size: The number of feature vectors contained within a particle.

[0010] The center of the grain: the mean of all feature vectors within the grain. Its three-dimensional coordinates are determined by calculating the arithmetic mean of the three feature parameters of all data points within the grain. c A、 c D、 c E The three-dimensional coordinates correspond to the arithmetic mean of the three characteristic parameters.

[0011] , , , In the formula, , , The first i The arithmetic mean of the area, depth, and vibrational energy of all data points within a single sphere; , , They are respectively the first in the granules j Area, depth, and vibration energy of each data point; This represents the number of data points within the particle.

[0012] Sphere radius: The maximum Euclidean distance from all feature vectors within the sphere to the center. After determining the center of the sphere, the Euclidean distance from each data point within the sphere to the center is calculated. The maximum value among all these distances is then taken as the radius of the sphere, representing the maximum spatial extent of the sphere. The Euclidean distance is calculated as follows: , In the formula: l、c These represent a specific data point and its corresponding particle center; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E The values ​​are the sizes of the three feature parameters.

[0013] During initialization, the set of normalized feature vectors collected by the gecko-like robot is regarded as the initial sphere, and its center is calculated. c With radius r .

[0014] ③ Perform pellet splitting based on quality assessment, and periodically analyze each pellet. The evaluation is conducted, and its quality evaluation function is: , In the formula: GB i Indicates the first i Each ball; This represents the number of data points within the particle. c i The center of the particle; x j For a data point inside the sphere, A splitting threshold is set. If the particle quality assessment result is greater than the threshold, it indicates that the data distribution inside the particle is loose and may contain multiple patterns. A splitting operation is required, splitting it into two sub-particles along the dimension with the largest variance in the feature space. The splitting threshold can be fine-tuned to avoid generating too few particles; the smaller the threshold, the higher the particle quality, the more particle splits, and the longer the algorithm runs.

[0015] ④ Calculate the similarity between any two spheres, using a metric that considers the geometry of the spheres: , In the formula: c i , c j The center of the two particles to be measured; r i , r j Let be the radii of two particles to be measured.

[0016] S The closer the similarity is to 1, the more similar the two spheres are or the more their spatial positions overlap. A merging threshold is set; if the similarity exceeds the threshold, they are merged into a new sphere. Similar to the splitting threshold, the merging threshold can be fine-tuned; the larger the threshold, the higher the sphere quality and the fewer spheres are merged.

[0017] ⑤ Recursively execute steps ③ and ④ until all pellets meet the quality requirements, thus dividing the initial inspection data into different clusters; according to the rockburst mechanism, the greater the vibration energy and the larger the fracture scale, the higher the risk of recurrence. Therefore, using the three indicators of fracture area, fracture depth, and vibration energy, the safety thresholds for each are determined based on historical rockburst records. By comparing the deviations of the indicator data of each cluster from the corresponding thresholds, and weighting and summing the deviations according to preset weights, a comprehensive risk score is obtained: , In the formula, k These represent different indicators (rupture area, rupture depth, vibration energy); The weights for different indicator characteristics are determined by expert evaluation. For different indicators, For the first i The first cluster k The central value of each indicator characteristic.

[0018] The risk of rockburst recurrence is assessed and graded based on the comprehensive risk score. When the comprehensive risk score is less than or equal to 0, it is determined that there is no risk of rockburst. When the comprehensive risk score is greater than 0, it is classified into low, medium, or high risk levels according to the score. The risk score range corresponding to different risk levels can be determined based on expert evaluation results.

[0019] ⑥ The system continuously collects new monitoring data, determines the cluster category to which it belongs, and thus determines the risk level of rockburst recurrence. When processing new monitoring data, the system calculates the Euclidean distance between the new data collected by the gecko-like robot and the centers of all existing grains, assigning it to the nearest grain. Based on the grain category determined by the new data, the corresponding rockburst risk is quickly identified. The Euclidean distance calculation method is as follows: , In the formula: l For a certain data point, c i The center of the grain; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E Let d be the size of the three feature parameters. Calculate all Euclidean distances and divide the new data into d. i The particle with the smallest value.

[0020] During system operation, as new data accumulates to a certain amount, the system automatically re-evaluates the quality of the particles and triggers necessary splitting and merging, thereby achieving online adaptive evolution of the discrimination model.

[0021] Furthermore, in step S3, the method for determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves is as follows: The vibration of the surrounding rock wall can be considered as a simple harmonic motion perpendicular to the rock wall. Based on the Doppler frequency shift of the electromagnetic waves caused by the vibration of the surrounding rock wall, the vibration velocity of the wall can be calculated. Suppose a transceiver transmits a sine wave with a starting point A, an ending point B, a spatial length D, and a frequency of . The target moves at a radial velocity. Flying towards the transceiver, since the target is in motion, the time required from point B to point A after contact with the target is [not specified]. : , When point A contacts the target, the distance of point B relative to the target is the length of the reflected sine wave. The formula for its calculation is: , Since the frequency of a wave is inversely proportional to its wavelength, then: , in, Echo frequency, Where is the incident frequency, At the speed of light, The radial velocity between the radio frequency transceiver and the target.

[0022] Based on the set incident frequency and received echo frequency data, calculation can be performed. Value, of which Calculated by the following formula: , In the formula, The vibration velocity of the surrounding rock wall. The angle is the complementary angle between the surrounding rock wall and the observation direction, where the observation direction is the direction of the line connecting the robot and the measured point on the surrounding rock.

[0023] The vibration velocity of the surrounding rock wall can be deduced from the above formula. Since the measurement is point-to-point, and vibration energy represents energy per unit space, the formula for calculating the maximum energy density of vibration is as follows: , in, For rock density, This represents the maximum vibration velocity of the surrounding rock wall.

[0024] Furthermore, the threshold determination method for the three types of feature values ​​is as follows: Collect sufficient historical data on rockbursts that have occurred in the mine, including fracture area, fracture depth and corresponding rockburst level. Take the minimum fracture area and minimum depth values ​​corresponding to the historical data of minor rockbursts as the safe thresholds for fracture area and fracture depth.

[0025] The mine is divided into multiple sections, the length of which can be determined based on engineering geological conditions. Preferably, areas with similar engineering geological conditions are considered as one section. Monitoring points are set at equal intervals within each section, and the vibration energy at each monitoring point is monitored 24 hours a day using vibration monitoring equipment. If the vibration energy at each monitoring point remains stable and without fluctuation over a long period of time, the vibration energy at each monitoring point during that period is statistically considered as the vibration energy under stable conditions. The long period of time refers to at least approximately 24 consecutive hours of monitoring to cover a complete daily production cycle, ensuring that the vibration energy data accurately reflects the natural stable state of the surrounding rock.

[0026] The vibration energy data of each monitoring point in the steady state within each interval are sorted in ascending order, and the value at which the vibration energy values ​​of 95% of the monitoring points are low is taken as the safety threshold of that interval.

[0027] Furthermore, the motion module of the inspection system consists of four moving legs, symmetrically deployed on both sides of the body and electrically driven. Each moving leg comprises a base, a base motor, a thigh, a joint motor, a calf, a telescopic air pump, a foot, and a palm suction cup. The base motor has two degrees of freedom, enabling the thigh to rotate 120° to the side of the body and swing 120° perpendicular to the horizontal plane of the body. The joint motor has one degree of freedom, enabling the calf to swing 60° within the plane of the thigh and calf. The telescopic air pump is installed in a groove at the end of the calf and connected to the foot via a ball joint. The palm suction cup is installed on the bottom of the foot. The air chamber of the telescopic air pump is connected to the air chamber of the palm suction cup via a flexible hose. When the telescopic air pump draws in air, the contact surface between the palm suction cup and the wall contracts inward, achieving an adsorption effect. When the air pump delivers air, the contact surface between the palm suction cup and the wall expands outward, thereby releasing the adsorption effect and allowing the foot to lift off the ground.

[0028] Furthermore, the auxiliary module of the inspection system includes an inertial measurement unit, an IoT locator, and a coordination unit. The inertial measurement unit integrates an accelerometer, a gyroscope, and an angular velocity sensor to realize the gecko-like robot's posture perception and motion recognition, and to achieve overall coordinated movement of the auxiliary module. The IoT locator is used to realize robot positioning. The coordination unit consists of a servo motor and a tail linkage, which can realize the tilting and shifting of the tail's center of gravity, and achieve balanced and coordinated movement in complex terrain.

[0029] Furthermore, the control module of the inspection system consists of a battery, a data storage device, a system-on-a-chip (SoC), a ribbon cable slot, a motherboard, and a magnetic charging module. The battery provides power to the entire system, the data storage device is used to cache data and control commands, the SoC is used for integrated computing and processing, the ribbon cable slot is used to connect the ribbon cables of the other three modules, the magnetic charging module is used to connect the charger of the communication power conversion base station, and the motherboard is used to integrate the above-mentioned components in the control module.

[0030] Furthermore, when data is scarce in the early stages of system deployment, the inspection robot can inspect data from multiple mines with similar geological conditions, rather than just one. After accumulating a certain amount of data over a certain period of time, cluster analysis can then be performed.

[0031] Beneficial effects The beneficial effects of this invention are as follows: (1) This invention is designed for robotic inspection of areas where rockbursts have occurred. It determines the vibration energy level of the surrounding rock wall through the Doppler effect analysis of electromagnetic waves, collects and analyzes the fracture range and depth of the surrounding rock surface in real time through a stereo vision camera, and uses a particle clustering algorithm to perform dynamic clustering analysis on multidimensional features, thereby achieving the purpose of judging the recurrence of rockbursts. This method can integrate the precursor phenomena and inducing factors of rockbursts, and has the advantages of being simple, efficient and highly automated. It can realize unmanned, large-scale inspection of rockburst disaster recurrence in hard rock mine roadways, reduce personnel risks, and is of great significance to the safe and efficient development of deep mineral resources.

[0032] (2) The gecko-like robot provided by this invention has functions such as remote control, automatic cruise, and automatic charging. It can replace manual labor in monitoring operations and does not occupy the roadway track, thus having certain economic and convenience advantages. Moreover, the robot perceives the road conditions through a stereo vision camera and ultrasonic sensors, and detects and warns of surrounding rock damage through radio frequency echoes. When it is about to be impacted by rocks, it can automatically perform a crouching action and stay close to the moving wall to avoid disaster, thus having high reliability. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of an embodiment of the tunnel rockburst inspection device based on a gecko-like robot provided by the present invention. Figure 2 A schematic diagram of the moving leg structure of the device provided by the present invention; Figure 3 A schematic diagram illustrating the foot-adhesion principle of the device provided by this invention; Figure 4 A schematic diagram of the coordination unit controlling the tail balance of the device provided in this invention; Figure 5 A system schematic diagram of the device is provided for this invention; The meanings of the symbols marked in the figure are as follows: 1-Stereo vision camera; 2-Ultrasonic sensor; 3-RF transceiver; 4-RF processor; 5-Battery; 6-Data storage; 7-System-on-a-chip; 8-Cable slot; 9-Motherboard; 10-Motion leg; 11-Magnetic charging module; 12-Inertial measurement unit; 13-IoT locator; 14-Coordination unit; 101-Base; 102-Base motor; 103-Thigh; 104-Joint motor; 105-Lower leg; 106-Telescopic air pump; 107-Foot; 108-Palm suction cup; 141-Servo motor; 142-Tail linkage. Detailed Implementation

[0034] The technical solutions of the present invention will be systematically described below with reference to the embodiments and accompanying drawings.

[0035] like Figure 1-5 As shown, a method for rockburst inspection in tunnels based on a gecko-inspired robot includes the following steps: S1, Deploy the inspection system: The inspection system includes... m A gecko-like robot n The system comprises a communication power base station and an integrated control console. The gecko-like robot is functionally divided into four parts: a perception module, a control module, a motion module, and an auxiliary module, deployed respectively at the head, body, four legs, and tail of the robot. The perception module consists of a stereo vision camera 1, an ultrasonic sensor 2, a radio frequency transceiver 3, and a radio frequency processor 4. The stereo vision camera 1 is used to collect information on the surrounding rock and road conditions in the tunnel; the ultrasonic sensor 2 is used for obstacle recognition and distance measurement; the radio frequency transceiver 3 is used to achieve non-contact measurement and wireless communication; and the radio frequency processor 4 is used to amplify the power of the radio frequency signal and convert the radio frequency frequency. The integrated control console is deployed into the communication network according to a suitable topology, and the integrated control console and the gecko-like robot are then activated. m、n The quantity depends on the actual conditions underground in the mine.

[0036] S2, Set the inspection path: The inspection path is mainly set on the path in the area where rock bursts have occurred. Place the gecko-like robot on the wall at the starting point of the tunnel to be tested. Input the motion command or set the cruise path through the controller to control the gecko-like robot to move along the tunnel. Set and record the inspection speed of the gecko robot and the camera observation direction so that the central processor can calculate. Deploy one gecko-like robot for each path. S3, Data Collection: The gecko-like robot's head sensing module continuously emits electromagnetic waves towards the surrounding rock face and receives the echoes. The echo signal data is modulated by a radio frequency processor, uploaded wirelessly to a communication transducer base station, and then uploaded to the integrated control console via an Ethernet link. The integrated control console analyzes and calculates the echo signals from each sensor using a central processing unit, determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves; it also records the condition of the inspected rock face using a stereo vision camera, calculating the area and maximum depth of any suspected fracture area when it is detected; and constructs a three-dimensional feature vector. (Vibrational energy, fracture area, and depth) and location tags were used to perform granular-ball clustering (GBCT) on the data. Cluster analysis is performed on the combined datasets to determine the likelihood of rockburst recurrence, and the results are output to the display in a multimedia format.

[0037] Furthermore, the method for determining the vibration energy, area, and maximum depth of the suspected fracture region in step S3 is as follows: The gecko-like robot uses a stereoscopic vision camera mounted on its head to continuously acquire images and perform three-dimensional scanning of the surrounding rock surface in the tunnel. Based on image recognition technology, it compares the current surrounding rock image with the most recently recorded corresponding position and shape image to identify newly formed, irregular fracture zones on the rock wall. When a suspected fracture zone is identified, the following calculation steps are performed: Based on the boundary coordinates of the fracture zone, the system queries the vibration energy of all measurement points within a 0.5-meter range extending outward along its normal direction, and takes the maximum value as the regional vibration energy associated with the rockburst crater. Reference points are selected at the outer boundary of the fractured area. These reference points form a two-dimensional point set and constitute a closed boundary surrounding the fractured area. The area of ​​the region enclosed by this boundary is calculated using the integral method, and this area value is taken as the fractured area of ​​the rock mass.

[0038] The depth of the fractured area is measured using an ultrasonic sensor, and the maximum value is taken as the maximum fracture depth.

[0039] Based on spatiotemporal information, the system dynamically updates the surface condition of the surrounding rock through continuous inspection. During each inspection, the current scan result is compared with the most recently recorded image to identify newly added fracture areas. When abnormal fracture propagation is detected, the system performs the aforementioned calculation steps to calculate the corresponding rockburst risk, triggers the corresponding rockburst risk warning, and coordinates the implementation of personnel access restrictions and safety protection measures. Simultaneously, the status is updated for the next round of monitoring.

[0040] Furthermore, in step S3, the specific steps for analyzing the area and maximum depth of the fractured region and the three-dimensional feature vector of vibration energy using a particle clustering algorithm to determine the possibility of rockburst recurrence are as follows: ① During the algorithm initialization phase, the parameter set is read, and the data is preprocessed using the Z-score standardization method. For any feature in the data, the Z-score standardization formula is: , in, , These represent the original data size and the standardized size, respectively. This represents the mean of the corresponding feature parameters in the dataset. is the standard deviation of the corresponding feature parameters in the dataset.

[0041] ② Define an initial sphere, whose characteristics include: Particle size: The number of feature vectors contained within a particle.

[0042] The center of the grain: the mean of all feature vectors within the grain. Its three-dimensional coordinates are determined by calculating the arithmetic mean of the three feature parameters of all data points within the grain. cA、 c D、 c E The three-dimensional coordinates correspond to the arithmetic mean of the three characteristic parameters.

[0043] , , , In the formula, , , The first i The arithmetic mean of the area, depth, and vibrational energy of all data points within a single sphere; , , They are respectively the first in the granules j Area, depth, and vibration energy of each data point; This represents the number of data points within the particle.

[0044] Sphere radius: The maximum Euclidean distance from all feature vectors within the sphere to the center. After determining the center of the sphere, the Euclidean distance from each data point within the sphere to the center is calculated. The maximum value among all these distances is then taken as the radius of the sphere, representing the maximum spatial extent of the sphere. The Euclidean distance is calculated as follows: , In the formula: l、c These represent a specific data point and its corresponding particle center; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E The values ​​are the sizes of the three feature parameters.

[0045] During initialization, the set of normalized feature vectors collected by the gecko-like robot is regarded as the initial sphere, and its center is calculated. c With radius r .

[0046] ③ Perform pellet splitting based on quality assessment. Periodically split each pellet. The evaluation is conducted, and its quality evaluation function is: , In the formula: GBi Indicates the first i Each ball; This represents the number of data points within the particle. c i The center of the particle; x j For a data point inside the sphere, Set a splitting threshold (e.g., initially set to 0.5). If the particle quality assessment result is greater than the threshold, it indicates that the data distribution within the particle is loose and may contain multiple patterns. A splitting operation needs to be performed, splitting it into two sub-particles along the dimension with the largest variance in the feature space. The splitting threshold can be fine-tuned to avoid generating too few particles (e.g., only 2-3). The smaller the threshold, the higher the particle quality, the more particle splits, and the longer the algorithm runs.

[0047] ④ Calculate the similarity between any two spheres, using a metric that considers the geometry of the spheres: , In the formula: c i , c j The center of the two particles to be measured; r i , r j Let be the radii of two particles to be measured.

[0048] S The closer the similarity is to 1, the more similar the two spheres are or the more their spatial positions overlap. A merging threshold is set (e.g., initially 0.9). If the similarity exceeds the threshold, the spheres are merged into a new sphere. Similar to the splitting threshold, the merging threshold can be fine-tuned; a higher threshold results in higher sphere quality and fewer spheres being merged.

[0049] ⑤ Recursively execute steps ③ and ④ until all pellets meet the quality requirements, thus dividing the initial inspection data into different clusters; according to the rockburst mechanism, the greater the vibration energy and the larger the fracture scale, the higher the risk of recurrence. Therefore, using the three indicators of fracture area, fracture depth, and vibration energy, the safety thresholds for each are determined based on historical rockburst records. By comparing the deviations of the indicator data of each cluster from the corresponding thresholds, and weighting and summing the deviations according to preset weights, a comprehensive risk score is obtained: , In the formula, k These represent different indicators (rupture area, rupture depth, vibration energy); The weights for different indicator characteristics are determined by expert evaluation. For different indicators, For the first i The first clusterk The central value of each indicator characteristic.

[0050] The risk of rockburst recurrence is graded and evaluated based on the magnitude of the comprehensive risk score. When the comprehensive risk score is less than or equal to 0, it is determined that there is no risk of rockburst. When the comprehensive risk score is greater than 0, it is divided into low risk, medium risk, or high risk levels according to the score. The risk score range corresponding to different risk levels can be determined based on the expert evaluation results.

[0051] ⑥ The system continuously collects new monitoring data (including secondary monitoring of new areas and already identified areas, distinguished by spatiotemporal labels), determines the cluster category to which it belongs, and thus determines the risk level of rockburst recurrence. When processing new monitoring data, the system calculates the Euclidean distance between the new data collected by the gecko-like robot and the centers of all existing grains, assigning it to the nearest grain. Based on the grain category defined by the new data, the corresponding rockburst risk is quickly identified. The Euclidean distance calculation method is as follows: , In the formula: l For a certain data point, c i The center of the grain; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E Let d be the size of the three feature parameters. Calculate all Euclidean distances and divide the new data into d. i The particle with the smallest value.

[0052] During system operation, as new data accumulates to a certain amount, the system automatically re-evaluates the quality of the particles and triggers necessary splitting and merging, thereby achieving online adaptive evolution of the discrimination model.

[0053] Furthermore, in step S3, the method for determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves is as follows: Vibrations generated by internal rock fractures, blasting operations, and excavation operations are the main triggering factors for rockburst disasters. When vibrations propagate to the surrounding rock wall, they will cause a Doppler frequency shift in the echo of electromagnetic waves. The greater the shift in the echo frequency, the higher the impact frequency and the greater the amplitude, that is, the stronger the vibration energy.

[0054] The vibration of the surrounding rock wall can be considered as a simple harmonic motion perpendicular to the rock wall. Based on the Doppler frequency shift of the electromagnetic waves caused by the vibration of the surrounding rock wall, the vibration velocity of the wall can be calculated. Suppose a transceiver transmits a sine wave with a starting point A, an ending point B, a spatial length D, and a frequency of . The target moves at a radial velocity. Flying towards the transceiver, since the target is in motion, the time required from point B to point A after contact with the target is [not specified]. : , When point A contacts the target, the distance of point B relative to the target is the length of the reflected sine wave. The formula for its calculation is: , Since the frequency of a wave is inversely proportional to its wavelength, then: , in, Echo frequency, Where is the incident frequency, At the speed of light, The radial velocity between the radio frequency transceiver and the target.

[0055] Based on the set incident frequency and received echo frequency data, calculation can be performed. Value, of which Calculated by the following formula: , In the formula, The vibration velocity of the surrounding rock wall. The angle is the complementary angle between the surrounding rock wall and the observation direction, where the observation direction is the direction of the line connecting the robot and the measured point on the surrounding rock.

[0056] The vibration velocity of the surrounding rock wall can be deduced from the above formula. Since the measurement is point-to-point, and vibration energy represents energy per unit space, the formula for calculating the maximum energy density of vibration is as follows: , in, For rock density, This represents the maximum vibration velocity of the surrounding rock wall.

[0057] Furthermore, the threshold determination method for the three types of feature values ​​is as follows: Collect sufficient historical data on rockbursts that have occurred in the mine, including fracture area, fracture depth and corresponding rockburst level. Take the minimum fracture area and minimum depth values ​​corresponding to the historical data of minor rockbursts as the safe thresholds for fracture area and fracture depth.

[0058] Here, the statistical lower bound of minor rockburst samples is used as the safety threshold, which essentially reflects the smallest identifiable scale of damage caused by a rockburst. Changes below this threshold mostly correspond to non-rockburst micro-damage (such as localized spalling) rather than true rockburst events.

[0059] The mine is divided into multiple sections, the length of which can be determined based on engineering geological conditions. Preferably, areas with similar engineering geological conditions are considered as one section. Monitoring points are set at equal intervals within each section, and the vibration energy at each monitoring point is monitored 24 hours a day using vibration monitoring equipment. If the vibration energy at each monitoring point remains stable and without fluctuation over a long period of time, the vibration energy at each monitoring point during that period is statistically considered as the vibration energy under steady-state conditions.

[0060] The long-term duration refers to a monitoring period of at least 24 consecutive hours to cover a complete production day cycle, ensuring that the vibration energy data can accurately reflect the natural stability of the surrounding rock.

[0061] The vibration energy data of each monitoring point in the steady state within each interval are sorted in ascending order, and the value at which the vibration energy values ​​of 95% of the monitoring points are low is taken as the safety threshold of that interval.

[0062] Furthermore, the motion module of the inspection system consists of four moving legs 10, symmetrically deployed on both sides of the body and electrically driven. Each moving leg 10 comprises a base 101, a base motor 102, a thigh 103, a joint motor 104, a lower leg 105, a telescopic air pump 106, a foot 107, and a palm suction cup 108. The base motor 102 has two degrees of freedom, enabling the thigh 103 to rotate 120° on the side of the body and swing 120° perpendicular to the horizontal plane of the body. The joint motor 104 has one degree of freedom, enabling the lower leg 105 to swing 60° within the plane of the thigh and lower leg. The telescopic air pump 106 is installed in the groove at the end of the lower leg 105 and connected to the foot 107 via a ball joint. The palm suction cup 108 is installed on the bottom surface of the foot 107. The air chamber of the telescopic air pump 106 is connected to the air chamber of the palm suction cup 108 via a hose. When the telescopic air pump 106 draws in air, the contact surface between the palm suction cup 108 and the wall contracts inward to achieve the adsorption effect. When the air pump delivers air, the contact surface between the palm suction cup 108 and the wall expands outward to release the adsorption effect, allowing the foot to be lifted into the air.

[0063] Furthermore, the auxiliary module of the inspection system includes an inertial measurement unit 12, an IoT locator 13, and a coordination unit 14. The inertial measurement unit integrates an accelerometer, a gyroscope, and an angular velocity sensor to realize the attitude perception and action recognition of the gecko-like robot, and to achieve the overall coordinated movement of the auxiliary module. The IoT locator is used to realize the robot's positioning. The coordination unit consists of a servo motor 141 and a tail link 142, which can realize the tilting and transfer of the center of gravity of the tail, and achieve balanced and coordinated movement in complex terrain.

[0064] Furthermore, the control module of the inspection system consists of a battery 5, a data storage device 6, a system-on-a-chip 7, a ribbon cable slot 8, a motherboard 9, and a magnetic charging module 11. The battery 5 provides power to the entire device, the data storage device 6 is used to cache data and control commands, the system-on-a-chip 7 is used for integrated computing, the ribbon cable slot 8 is used to connect the ribbon cables of the other three modules, and the magnetic charging module 11 is used to connect the charger of the communication power conversion base station. The motherboard 9 is used to integrate the above-mentioned components in the control module.

[0065] Furthermore, when data is scarce in the early stages of system deployment, the inspection robot can inspect data from multiple mines with similar geological conditions, rather than just one. After accumulating a certain amount of data over a certain period of time, cluster analysis can then be performed.

[0066] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for roadway rockburst inspection based on a gecko-inspired robot, characterized in that, The steps for rockburst inspection include: S1, Deploy the inspection system: The inspection system includes... m A gecko-like robot n The system comprises one communication power base station and one integrated control console. The gecko-like robot is functionally divided into four parts: a perception module, a control module, a motion module, and an auxiliary module. These are deployed on the head, body, four legs, and tail of the gecko-like robot, respectively. The perception module consists of a stereo vision camera, an ultrasonic sensor, a radio frequency transceiver, and a radio frequency processor. The stereo vision camera is used to collect information on the surrounding rock and road conditions in the tunnel; the ultrasonic sensor is used for obstacle identification and distance measurement; the radio frequency transceiver is used to achieve non-contact measurement and wireless communication; and the radio frequency processor is used to amplify the power of the radio frequency signal and convert the radio frequency frequency. The integrated control console is deployed into the communication network according to a suitable topology, and the integrated control console and the gecko-like robot are then activated. S2, Set the inspection path: The inspection path is mainly set on the path of the area where rock burst has occurred. Place the gecko robot on the wall at the starting point of the tunnel to be tested. Input the motion command or set the cruise path through the controller to control the gecko robot to move along the tunnel. Set and record the inspection speed of the gecko robot and the camera observation direction so that the central processor can calculate. Deploy 1 gecko robot for each path. S3, Data Collection: The gecko-like robot's head sensing module continuously emits electromagnetic waves towards the surrounding rock face and receives the echoes. The echo signal data is modulated by a radio frequency processor, uploaded wirelessly to a communication transducer base station, and then uploaded to the integrated control console via an Ethernet link. The integrated control console analyzes and calculates the echo signals from each sensor using a central processing unit, determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves; it also records the condition of the inspected rock face using a stereo vision camera, calculating the area and maximum depth of any suspected fracture area when it is detected; and constructs a three-dimensional feature vector. (Vibrational energy, fracture area, and depth) and location tags were used to perform granular-ball clustering (GBCT) on the data. Cluster analysis is performed on the combined datasets to determine the likelihood of rockburst recurrence, and the results are output to the display in a multimedia format.

2. The method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, The method for determining the vibration energy, area, and maximum depth of the suspected fracture region in step S3 is as follows: The gecko-like robot uses a stereoscopic vision camera mounted on its head to continuously acquire images and perform three-dimensional scanning of the surrounding rock surface in the tunnel. Based on image recognition technology, it compares the current surrounding rock image with the most recently recorded corresponding position and shape image to identify newly formed, irregular fracture zones on the rock wall. When a suspected fracture zone is identified, the following calculation steps are performed: Based on the boundary coordinates of the fracture zone, the system queries the vibration energy of all measurement points within a 0.5-meter range extending outward along its normal direction, and takes the maximum value as the regional vibration energy associated with the rockburst crater. Reference points are selected at the outer boundary of the fracture zone. These reference points form a two-dimensional point set and constitute a closed boundary surrounding the fracture zone. The area of ​​the region enclosed by this boundary is calculated using the integral method, and this area value is taken as the fracture area of ​​the rock mass. The depth of the fractured area is measured using an ultrasonic sensor, and the maximum value is taken as the maximum fracture depth.

3. The method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, In step S3, the specific steps for analyzing the area and maximum depth of the fractured region and the three-dimensional feature vector of vibration energy using the particle clustering algorithm to determine the possibility of rockburst recurrence are as follows: ① During the algorithm initialization phase, the parameter set is read, and the data is preprocessed using the Z-score standardization method. For any feature in the data, the Z-score standardization formula is: ， in, , These are the original data size and the standardized size, respectively. This represents the mean of the corresponding feature parameters in the dataset. is the standard deviation of the corresponding feature parameters in the dataset. ② Define an initial sphere, whose characteristics include: Particle size: The number of feature vectors contained within a particle; The center of the grain: the mean of all feature vectors within the grain. Its three-dimensional coordinates are determined by calculating the arithmetic mean of the three feature parameters of all data points within the grain. c A、 c D、 c E The three-dimensional coordinates correspond to the arithmetic mean of three characteristic parameters; ， ， ， In the formula, , , The first i The arithmetic mean of the area, depth, and vibrational energy of all data points within a single sphere; , , They are respectively the first in the granules j Area, depth, and vibration energy of each data point; This represents the number of data points within the particle. Sphere radius: The maximum Euclidean distance from all feature vectors within the sphere to the center. After determining the center of the sphere, the Euclidean distance from each data point within the sphere to the center is calculated. The maximum value among all these distances is then taken as the radius of the sphere, representing the maximum spatial extent of the sphere. The Euclidean distance is calculated as follows: ， In the formula: l 、c These represent a specific data point and its corresponding particle center; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E The values ​​are the sizes of the three feature parameters. During initialization, the set of all normalized feature vectors collected by the gecko-like robot is regarded as the initial sphere, and its center is calculated. c With radius r ; ③ Perform pellet splitting based on quality assessment, and periodically analyze each pellet. The evaluation is conducted, and its quality evaluation function is: ， In the formula: GB i Indicates the first i There are 10 spheres; n is the number of data points within a sphere. c i The center of the particle; l j For a given data point inside the sphere; Set a splitting threshold. If the particle quality assessment result is greater than the threshold, it indicates that the data distribution inside the particle is loose and may contain multiple patterns. A splitting operation needs to be performed to split it into two sub-particles along the dimension with the largest variance in the feature space. ④ Calculate the similarity between any two spheres, using a metric that considers the sphere's geometry: ， In the formula: c i , c j The center of the two particles to be measured; r i , r j Let be the radii of two particles to be measured. S The closer the similarity is to 1, the more similar the two balls are or the more their spatial positions overlap. A merging threshold is set. If the similarity exceeds the threshold, the balls are merged into a new ball. ⑤ Recursively execute steps ③ and ④ until all pellets cease splitting and merging, thus dividing the initial inspection data into different clusters; according to the rockburst mechanism, the greater the vibration energy and the larger the fracture scale, the higher the risk of recurrence; therefore, using the three indicators of fracture area, fracture depth, and vibration energy, and based on historical rockburst records, their respective safety thresholds are determined. By comparing the deviations of the indicator data of each cluster from the corresponding thresholds, and weighting and summing the deviations according to preset weights, a comprehensive risk score is obtained: ， In the formula, k These represent different indicators (rupture area, rupture depth, vibration energy); The weights for different indicator characteristics are determined by expert evaluation. For different indicators, For the first i The first cluster k The central value of each indicator characteristic; The risk of rockburst recurrence is graded and evaluated based on the comprehensive risk score. When the comprehensive risk score is less than or equal to 0, it is determined that there is no risk of rockburst. When the comprehensive risk score is greater than 0, it is divided into low risk, medium risk or high risk level according to the score. The risk score range corresponding to different risk levels can be determined based on the expert evaluation results. ⑥ The system continuously collects new monitoring data, determines the cluster category to which it belongs, and thus determines the risk level of rockburst recurrence. When processing new monitoring data, it calculates the Euclidean distance between the new data collected by the gecko-like robot and the centers of all existing grains, classifying it into the nearest grain. Based on the grain category classified by the new data, it quickly identifies the corresponding rockburst risk. The Euclidean distance calculation method is as follows: ， In the formula: l For a certain data point, c i The center of the grain; c A、 c D、 c E The coordinates of the center of the particle are the three-dimensional coordinates, representing the fracture area, depth, and center value of the vibrational energy, respectively. A 、 D 、 E Given the three feature parameters, calculate all Euclidean distances and divide the new data into d. i The particle with the smallest value; During system operation, as new data accumulates to a certain amount, the system automatically re-evaluates the quality of the particles and triggers necessary splitting and merging, thereby achieving online adaptive evolution of the discrimination model.

4. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, In step S3, the method for determining the vibration energy of the surrounding rock surface based on the Doppler effect of electromagnetic waves is as follows: The vibration of the surrounding rock wall can be considered as a simple harmonic motion perpendicular to the rock wall. Based on the Doppler frequency shift of the electromagnetic waves caused by the vibration of the surrounding rock wall, the vibration velocity of the wall can be calculated. Suppose a transceiver transmits a sine wave with a starting point A, an ending point B, a spatial length D, and a frequency of . The target has a radial velocity Flying towards the transceiver, since the target is in motion, the time required from point B to point A after contact with the target is [not specified]. : ， When point A contacts the target, the distance of point B relative to the target is the length of the reflected sine wave. The formula for its calculation is: ， Since the frequency of a wave is inversely proportional to its wavelength, then: ， in, Echo frequency, Where is the incident frequency, At the speed of light, The radial velocity between the radio frequency transceiver and the target. Based on the set incident frequency and received echo frequency data, calculation can be performed. Value, of which Calculated by the following formula: ， in, The vibration velocity of the surrounding rock wall. The angle between the surrounding rock wall and the observation direction is the complementary angle, and the observation direction is the direction of the line connecting the robot and the measured point of the surrounding rock. The vibration velocity of the surrounding rock wall can be deduced from the above formula. Since the measurement is point-to-point, and vibration energy represents energy per unit space, the formula for calculating the maximum energy density of vibration is as follows: , in, For rock density, This represents the maximum vibration velocity of the surrounding rock wall.

5. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, The threshold determination method for the three types of feature values ​​is as follows: Collect sufficient historical data on rockbursts that have occurred in the mine, including fracture area, fracture depth and corresponding rockburst level. Take the minimum fracture area and minimum depth values ​​corresponding to the historical data of minor rockbursts as the safe thresholds for fracture area and fracture depth. The mine is divided into multiple sections, and the length of each section can be determined according to the engineering geological conditions. Preferably, areas with similar engineering geological conditions are considered as one section. Monitoring points are set at equal intervals within each section, and the vibration energy of each monitoring point is monitored by a 24-hour vibration monitoring device. If the vibration energy of each monitoring point remains stable and without fluctuation over a long period of time, the vibration energy of each monitoring point during that period is taken as the vibration energy under a stable state for statistical purposes. The long-term duration is a monitoring period of at least 24 consecutive hours to cover a complete production day cycle, so as to ensure that the vibration energy data can truly reflect the natural stability of the surrounding rock. The vibration energy data of each monitoring point in the steady state within each interval are sorted in ascending order, and the value at which the vibration energy values ​​of 95% of the monitoring points are low is taken as the safety threshold of that interval.

6. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, The motion module consists of four movable legs, symmetrically arranged on both sides of the body, and is electrically driven. Each movable leg comprises a base, a base motor, a thigh, a joint motor, a calf, a telescopic air pump, a foot, and a palm suction cup. The base motor has two degrees of freedom, enabling the thigh to rotate 120° to the side of the body and swing 120° perpendicular to the horizontal plane of the body. The joint motor has one degree of freedom, enabling the calf to swing 60° within the plane of the thigh and calf. The telescopic air pump is installed in a groove at the end of the calf and connected to the foot via a ball joint. The palm suction cup is installed on the bottom of the foot. The air chamber of the telescopic air pump is connected to the air chamber of the palm suction cup via a hose. When the telescopic air pump inhales, the contact surface between the palm suction cup and the wall contracts inward to achieve an adsorption effect. When the air pump delivers air, the contact surface between the palm suction cup and the wall expands outward, thereby releasing the adsorption effect and allowing the foot to lift off the ground.

7. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, The auxiliary module includes an inertial measurement unit, an IoT locator, and a coordination unit. The inertial measurement unit integrates an accelerometer, a gyroscope, and an angular velocity sensor to achieve posture perception and motion recognition for the gecko-like robot, enabling coordinated movement of the entire auxiliary module. The IoT locator is used for robot positioning. The coordination unit consists of a servo motor and a tail link, which can tilt the tail to transfer the center of gravity, achieving balanced and coordinated movement in complex terrain.

8. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, The control module consists of a battery, a data storage device, a system-on-a-chip (SoC), a ribbon cable slot, a motherboard, and a magnetic charging module. The battery provides power to the entire device, the data storage device is used to cache data and control commands, the SoC is used for integrated computing and processing, the ribbon cable slot is used to connect the ribbon cables of the other three modules, the magnetic charging module is used to connect the charger of the communication power transducer base station, and the motherboard is used to integrate the above-mentioned components in the control module.

9. A method for roadway rockburst inspection based on a gecko-inspired robot according to claim 1, characterized in that, In the early stages of system deployment, when data is scarce, the inspection robot can inspect data from multiple mines with similar geological conditions, rather than just one. After accumulating a certain amount of data over a certain period of time, cluster analysis can then be performed.

Images