An autonomous inspection robot system for coal mine underground power transformation equipment

Abstract

The application discloses a kind of coal mine underground power transformation equipment autonomous inspection robot system, it is related to coal mine intelligent inspection technical field, comprising: the risk of comprehensive equipment operating state, historical defect and inspection interval is evaluated, and inspection sequence is dynamically adjusted. Multi-source data are fused and analyzed, and the consistency of observation quality evaluation evidence is combined to reduce false alarm and missed alarm caused by dust, reflection and environmental noise. Suspected anomaly is independently executed re-focusing, angle adjustment, distance change or switching mode and other re-inspection actions, to form abnormality discovery, review and disposal closed loop. Based on safety constraints, wait, detour, retreat, stop or return based on environmental anomalies. When communication is interrupted, the robot continues to perform limited inspection and safety disposal according to the local authorization strategy, saves complete evidence, and uploads after communication recovery. Inspection results update equipment health records for subsequent task scheduling, reduce invalid repeated inspection, and ensure stable operation of power supply system.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology in coal mines, and more specifically to an autonomous inspection robot system for underground power equipment in coal mines. Background Technology

[0002] Underground coal mine substations undertake crucial tasks such as power distribution, equipment supply, and safety protection. High- and low-voltage switchgear, transformers, feeder switches, protection devices, cable joints, and other electrical equipment operate long-term in environments with high humidity, high dust, vibration, and electromagnetic interference, making them prone to overheating of joints, insulation aging, partial discharge, instrument malfunctions, and mechanical failures. If these abnormalities are not detected in time, they may cause equipment shutdowns or even lead to underground safety accidents.

[0003] Currently, underground power equipment is mainly inspected manually on a regular basis. Manual inspection has drawbacks, including high labor intensity, limited inspection frequency, long exposure time to hazardous environments, and the influence of personnel experience on test results. Some mines have begun to use inspection robots, which collect information such as visible light, infrared radiation, sound, and gas concentration along pre-set routes and upload the results to a monitoring platform for analysis.

[0004] Existing inspection robots typically perform tasks according to fixed routes and cycles, making it difficult to dynamically adjust the inspection sequence based on equipment load, historical defects, real-time alarms, and abnormal development trends. When a suspected anomaly is detected, the robot usually only issues an alarm or uploads a single inspection result, and cannot autonomously change the inspection angle, distance, or inspection method for re-inspection based on the observation quality and consistency between different inspection information.

[0005] Furthermore, underground coal mines present challenges such as gas, dust, water accumulation, obstacles, narrow passages, and unstable communication. Existing surface substation inspection technologies are difficult to apply directly to the underground environment. When communication is interrupted or environmental conditions deteriorate, robots typically have no choice but to stop operating, lacking an autonomous degrade mechanism that balances equipment anomaly confirmation with the robot's own safety.

[0006] Therefore, how to propose an autonomous inspection robot system for underground power equipment in coal mines, realize the dynamic arrangement of inspection tasks based on equipment risks, perform reliable fusion of multimodal detection results, autonomously re-inspect after discovering suspected anomalies, and operate safely under environmental deterioration or communication limitations are problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides an autonomous inspection robot system for underground power equipment in coal mines, solving the problems of fixed task execution sequence, insufficient reliability of multimodal detection results, inability to autonomously re-inspect suspected anomalies, and difficulty in safe closed-loop operation when communication is interrupted or the underground environment deteriorates, in existing inspection methods. To achieve the above objectives, the present invention adopts the following technical solution: An autonomous inspection robot system for underground power equipment in coal mines includes: an explosion-proof inspection robot side and a ground and equipment side; The explosion-proof inspection robot includes: a risk-driven task orchestration module, a positioning, navigation and safety constraint module, a multimodal acquisition and quality evaluation module, an evidence consistency and anomaly diagnosis module, an active re-inspection and action decision module, and a local safety degradation and evidence storage module. The ground and equipment side includes: a ground monitoring and manual confirmation terminal, a power equipment operation data interface module, and a historical defect and inspection file module; The ground monitoring and manual confirmation terminal, the power equipment operation data interface module, and the historical defect and inspection archive module are all signal-connected to the risk-driven task orchestration module. The risk-driven task orchestration module is sequentially connected to the positioning, navigation and safety constraint module, the multimodal acquisition and quality evaluation module, the evidence consistency and anomaly diagnosis module, and the active re-inspection and action decision module, and is used for inspection planning and dynamic task updating based on inspection results. The evidence consistency and anomaly diagnosis module is connected to the ground monitoring and manual confirmation terminal signal through the local security degradation and evidence storage module, and is used to diagnose anomalies in the inspection results. The active re-inspection and action decision-making module is connected to the ground monitoring and manual confirmation terminal, and the positioning, navigation and safety constraint module, respectively, and triggers re-inspection feedback based on the diagnostic results.

[0008] Optionally, the risk-driven task orchestration module is used to create a risk profile for each piece of equipment or part to be inspected.

[0009] Optionally, the positioning, navigation and safety constraint module is used to maintain the roadway and substation map, equipment inspection pose set and restricted areas.

[0010] Optionally, the multimodal acquisition and quality evaluation module is used to control the pan-tilt unit, focal length, supplementary lighting, and detection distance to acquire visible light, infrared, sound, partial discharge, and environmental information from the target device.

[0011] Optionally, the evidence consistency and anomaly diagnosis module is used to map the detection results of different modalities to a unified device location and time window to determine whether the evidence supports each other.

[0012] Optionally, the proactive review and action decision module is used to select the next action based on the anomaly risk, evidence consistency, and observation quality.

[0013] Optionally, the local security degradation and evidence storage module is used to continue executing pre-authorized inspection and re-inspection actions when communication quality is insufficient, and to perform tamper-proof associated storage of original data, quality indicators, decision basis and action logs.

[0014] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an autonomous inspection robot system for underground power equipment in coal mines, which has the following beneficial effects: This invention comprehensively assesses equipment risk based on real-time operating status, historical defects, and inspection intervals, dynamically adjusting the inspection sequence to prioritize high-risk equipment. It integrates visible light, infrared, sound, partial discharge, and equipment operating data for analysis, combining this with observation quality evaluation evidence consistency to reduce false alarms and missed alarms caused by dust obstruction, reflections, and environmental noise. For suspected anomalies, it autonomously performs re-inspection actions such as refocusing, adjusting detection angles, changing detection distances, or switching detection modes, forming a closed loop for anomaly detection, verification, and handling. Environmental information such as gas, smoke, temperature, water accumulation, and obstacles are used as safety constraints, allowing the robot to autonomously perform actions such as waiting, detouring, retreating, parking, or returning to base when environmental anomalies occur. In the event of communication interruption, the robot can continue limited inspections and safety handling according to local authorized policies, preserving complete inspection evidence for upload after communication is restored. Inspection results can update equipment health records and be used for subsequent task scheduling, reducing unnecessary duplicate inspections, promptly detecting equipment deterioration trends, and ensuring the stable operation of underground power supply systems in coal mines. Attached Figure Description

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

[0016] Figure 1 This invention provides a structural framework diagram of an autonomous inspection robot system for underground power equipment in coal mines.

[0017] Figure 2 This is a side structural frame diagram of the explosion-proof inspection robot provided by the present invention.

[0018] Figure 3 The flowchart of risk-driven inspection and proactive re-inspection closed loop provided by this invention. Detailed Implementation

[0019] 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.

[0020] This invention discloses an autonomous inspection robot system for underground power equipment in coal mines, such as... Figure 1 As shown, it includes: the explosion-proof inspection robot side and the ground and equipment side; The explosion-proof inspection robot includes: a risk-driven task orchestration module, a positioning, navigation and safety constraint module, a multimodal acquisition and quality evaluation module, an evidence consistency and anomaly diagnosis module, an active re-inspection and action decision module, and a local safety degradation and evidence storage module. The ground and equipment side includes: a ground monitoring and manual confirmation terminal, a power equipment operation data interface module, and a historical defect and inspection file module; The ground monitoring and manual confirmation terminal, the power equipment operation data interface module, and the historical defect and inspection archive module are all signal-connected to the risk-driven task orchestration module. The risk-driven task orchestration module is sequentially connected to the positioning, navigation and safety constraint module, the multimodal acquisition and quality evaluation module, the evidence consistency and anomaly diagnosis module, and the active re-inspection and action decision module, and is used for inspection planning and dynamic task updating based on inspection results. The evidence consistency and anomaly diagnosis module is connected to the ground monitoring and manual confirmation terminal signal through the local security degradation and evidence storage module, and is used to diagnose anomalies in the inspection results. The active re-inspection and action decision-making module is connected to the ground monitoring and manual confirmation terminal, and the positioning, navigation and safety constraint module, respectively, and triggers re-inspection feedback based on the diagnostic results.

[0021] Furthermore, the risk-driven task orchestration module is used to create a risk profile for each device or part to be inspected. The risk profile integrates the device's real-time operating volume, protection alarms, historical defects, time since the last effective inspection, and defect development trends. Based on the risk profile, an initial task priority is generated, and during the inspection process, new abnormal evidence, environmental conditions, and remaining battery power are received, and unexecuted tasks are reordered in real time.

[0022] Specifically, the information used to establish risk profiles includes real-time equipment operation volume, protection alarms, historical defects, time since the last effective inspection, and defect development trends.

[0023] The initial task priority of each equipment or part to be inspected is determined based on the risk profile; during the inspection process, newly discovered abnormal evidence, environmental status and remaining robot power are received, and the priority of unexecuted tasks is updated and rearranged in real time accordingly, so that newly added high-risk equipment or related equipment can be inspected first.

[0024] Furthermore, the positioning, navigation, and safety constraint module is used to maintain maps of tunnels and substations, equipment inspection pose sets, and restricted areas. It not only plans the shortest path but also uses gas concentration, smoke, temperature, water accumulation, obstacles, slope, communication blind spots, and remaining battery power as constraints. When any safety indicator reaches the limit, the corresponding action is prohibited, and the system chooses to wait, detour, retreat, stop, or return to base.

[0025] Furthermore, the multimodal acquisition and quality evaluation module is used to control the gimbal, focal length, supplementary lighting, and detection distance to acquire visible light, infrared, sound, partial discharge, and environmental information from the target device. Each acquisition simultaneously records the robot pose, device identification, and timestamp, and calculates the observation quality based on sharpness, occlusion rate, signal-to-noise ratio, ranging stability, and target area integrity. Low-quality observations are not directly used as a definitive fault conclusion.

[0026] Furthermore, the evidence consistency and anomaly diagnosis module is used to map detection results from different modalities to a unified device location and time window to determine whether the evidence supports each other. For example, infrared temperature rise at the connection point and load current rise support each other, and abnormal sound and enhanced partial discharge support each other; if only the visible light instrument identifies an anomaly but the image quality is low, it is judged to be pending verification rather than directly triggering an alarm.

[0027] Specifically, firstly, the detection results of different modalities are mapped to the same equipment location and the same time window, and the observation quality and anomaly probability of each modality are determined. Then, the anomaly probabilities of each modality are weighted according to the observation quality to obtain the observation quality-weighted anomaly probability, and the consistency of multimodal evidence is determined based on the differences between the detection results of different modalities. Finally, the anomaly confidence level is calculated by combining the observation quality-weighted anomaly probability with the evidence consistency, and the detection results are divided into three states based on preset thresholds: anomaly confirmed, pending review, and no anomaly determined at this time. When the observation quality or evidence consistency is insufficient, the anomaly is not directly confirmed, but rather an active re-inspection is triggered.

[0028] Furthermore, the proactive re-inspection and action decision module is used to select the next action based on anomaly risk, evidence consistency, and observation quality. Optional actions include refocusing and reshooting, changing the gimbal angle, changing the detection distance, extending sound sampling, switching partial discharge detection, waiting for the load to stabilize before retesting, comparing with adjacent related devices, and creating a manual review task. After re-inspection, the anomaly confidence level is recalculated until the conditions for confirmation, exclusion, or safe exit are met.

[0029] Specifically, the next action refers to the subsequent re-inspection or handling action selected by the system based on the current detection result, including refocusing and reshooting, changing the gimbal angle, changing the detection distance, extending the sound sampling time, switching the partial discharge detection mode, waiting for the equipment load to stabilize before retesting, going to adjacent related equipment for comparative testing, and creating a manual review task.

[0030] When the anomaly confidence level is in the pending review range, the evidence consistency is lower than the preset threshold, or the observation quality of the key modality is lower than the preset threshold, the system triggers an active re-examination. After the re-examination is completed, the newly collected data is returned to the observation quality evaluation and anomaly diagnosis stage to recalculate the evidence consistency and anomaly confidence level until the conditions for anomaly confirmation, anomaly exclusion, or safe exit are met.

[0031] Furthermore, the local security degradation and evidence storage module is used to continue executing pre-authorized inspection and re-inspection actions when communication quality is insufficient, and to perform tamper-proof associated storage of original data, quality indicators, decision-making basis, and action logs. For high-risk anomalies, a local audio-visual alert is triggered and the system moves to a safe location; after communication is restored, the evidence chain is uploaded according to event priority.

[0032] Specifically, the tamper-proof associated storage is implemented using event identifiers, timestamps, data digests, and sequential association. The system generates a unique event identifier for each inspection task or abnormal event, and associates and stores raw inspection data, observation quality indicators, anomaly diagnosis results, decision-making basis, robot pose, and action logs according to the same event identifier. A timestamp and data digest are generated for each data record, and each subsequent record stores the data digest of the previous record, forming evidence records associated in chronological order. During data reading, uploading, or manual review, the system recalculates the data digest and performs a consistency check with the stored digest; if the check result is inconsistent, the corresponding data is marked as an abnormal record. After communication is restored, the original data, data digest, and their associated index are uploaded according to event priority.

[0033] Specifically, the tamper-proof associated storage is used to generate a unique event identifier for each inspection task or abnormal event, and to associate and store the original detection data, observation quality indicators, abnormal diagnosis results, decision basis, robot pose and action log according to the event identifier; to generate timestamps and data digests for each data record, and to ensure that the subsequent data record saves the data digest of the previous data record to form a sequentially associated evidence record; when data is read or uploaded, the data digest is recalculated and consistency verification is performed to identify whether the data has changed.

[0034] In a specific implementation, a risk-driven inspection and proactive re-inspection closed-loop process for underground coal mine power equipment is described as follows: Figure 3As shown, it includes: S1: The basic data acquisition system first obtains three types of basic information: the real-time operating status of the equipment under inspection, historical operation and maintenance records, and the status of the surrounding environment, as the original input for risk assessment.

[0035] S2: Risk assessment and dynamic task scheduling rely on multi-source data collected by S1 to complete the quantitative calculation of equipment risk level, and dynamically and automatically arrange the priority and execution plan of inspection tasks according to the level of risk.

[0036] Specifically, the overall risk of the equipment is calculated according to formula (1): ; in, The abnormal operation weight, ranging from 0.3 to 0.6, is used to indicate the degree of importance attached to real-time operation volume and protection alarms; The abnormal weight is set to 0.2 to 0.5, which indicates the degree of importance attached to historical defects and degradation trends. The weight for overdue inspections ranges from 0.1 to 0.4, and is used to indicate the importance attached to the time elapsed since the last valid inspection. The priority combination weights are non-negative; they are configured according to the on-site objectives to adjust the impact of risk, anomaly credibility, execution cost, and environmental hazard, respectively. The environmental hazard level is 0 to 1; entry is prohibited when the threshold is reached. It is formed by a combination of factors such as gas, smoke, temperature, and water accumulation. For observation quality, the value is set to 0-1; values ​​below 0.5 are given priority for re-inspection and are used for evaluation by mode. This is the quality-weighted anomaly probability, ranging from 0 to 1, used to indicate that it is calculated jointly by the anomaly probabilities of each mode and the observation quality. The minimum retention coefficient for consistency, ranging from 0.2 to 0.6, is used to prevent a single, highly credible piece of evidence from being completely suppressed. For anomaly confidence level, example: below 0.4 excludes, 0.4-0.75 re-inspects, above 0.75 confirms. The threshold can be configured according to the device type. To prevent the constant from being zero, a small value greater than 0 is used to prevent the total observation mass from being zero; This is a dynamic priority setting, with values ​​assigned in descending order. It is only used for sorting when safety constraints allow. Index of equipment or equipment parts to be inspected. , For observing modal index, , For equipment or parts The overall risk, with a value between 0 and 1. For equipment The abnormality level of the operation, with a value of 0 to 1. For equipment Historical defect score, ranging from 0 to 1. For equipment The degree of overdue inspection, ranging from 0 to 1. For modality For equipment The given anomaly probability ranges from 0 to 1. For equipment Multimodal evidence consistency, with a value ranging from 0 to 1. For equipment The anomaly confidence level, ranging from 0 to 1, From robots to equipment The normalized execution cost takes a value of 0 to 1.

[0037] Provided that safety constraints allow execution, the dynamic priority of the task is calculated according to formula (2): ; Among them, risk weight And the confidence weight of the detected anomalies Increase task priority when the priority is increased; increase execution cost weight. and environmental risk weight Priority decreases as environmental hazard weight increases. When the preset safety threshold is exceeded, the sorting will no longer rely solely on equation (2), but will directly prohibit entry into the corresponding area.

[0038] Specifically, environmental risk is represented by a normalized value of 0 to 1, and an environmental early warning threshold is set. Environmental stopping threshold .

[0039] when At that time, the robot performed its inspection tasks normally; when At this time, the robot is restricted from approaching the target area and only performs long-distance observation, low-speed passage or non-contact re-inspection; when When the robot stops entering the target area, it will execute actions such as waiting, retreating, parking, or returning to its starting point. The threshold can be adjusted according to mine safety regulations, the robot's explosion-proof rating, water wading capability, traversing gradient, and obstacle avoidance capability.

[0040] S3: Environmental safety constraint judgment (first-level branch node) verifies whether the current environmental conditions and safety rules allow the inspection task to be executed: If the judgment result is negative: proceed to stage S4 and perform actions such as waiting, rerouting, obstacle avoidance, or returning directly to base. The mission will not be executed for the time being. The determination is: proceed downwards to the S5 on-site inspection operation.

[0041] S5: The inspection vehicle, which is used for close-range operations and multimodal data acquisition, navigates to the designated position of the target equipment and simultaneously acquires multi-dimensional modal perception data such as images and sensors during the arrival process.

[0042] S6: Quantitative evaluation of the credibility of observation data involves quality verification of various modal observation data collected, as well as calculation of the consistency of multi-source evidence and the credibility of anomalous features to quantify the validity of the data.

[0043] Specifically, the observation quality-weighted anomaly probability is calculated according to equation (3): (3); The consistency of multimodal evidence is calculated according to equation (4): ; Where M represents the number of observation modes participating in the anomaly judgment of the same inspection target; m represents the observation mode number; Let be the observation quality weight of the i-th inspection target in the m-th observation mode; Let be the anomaly probability obtained for the i-th inspection target in the m-th observation mode; The weighted anomaly probability of the observation quality for the i-th inspection target; To prevent extremely small positive numbers with a denominator of zero; Let be the multimodal evidence consistency coefficient for the i-th inspection target.

[0044] The confidence level of anomalies is calculated according to formula (5): ; in, To prevent positive numbers with a denominator of zero, The minimum retention coefficient for consistency. When the quality of all observations is below the minimum quality threshold, the system does not use equation (5) to output confirmation of the anomaly, but instead creates a re-inspection task.

[0045] Anomaly confidence level uses a low threshold and high threshold Perform a classification judgment, among which Take 0.45, Take 0.75. When ≥ When, it is determined to be an abnormality; when ≤ < When, it enters the pending review period and triggers active re-inspection; when < At this time, it is not considered abnormal.

[0046] Furthermore, when Located in the area awaiting review, or Below the consistency threshold, or critical modalities When the quality falls below a certain threshold, an active re-inspection is triggered. The re-inspection action is selected based on the principles of improving the expected observation quality, strengthening mutually independent evidence, and meeting environmental safety constraints.

[0047] Minimum observation quality threshold Take 0.60. When the observation quality of the key observation mode... Below Or the observation quality of all observation modes is lower than that of the observed modes. When an anomaly is confirmed, the system does not directly use equation (5) to confirm the anomaly, but instead generates an active re-inspection task to reacquire observation data that meets the quality requirements.

[0048] The interval to be reviewed is set as follows ≤ < ,in It is 0.45. The threshold for consistency of evidence is 0.75. Set to 0.65; Key mode observation quality threshold Take 0.60. When Located in the area awaiting review, or < ,or < At that time, an active re-inspection is triggered.

[0049] Environmental safety constraints are determined using the environmental hazard level 'e', ​​and restrictions are set to limit proximity to thresholds. The threshold value is 0.60, at which point entry is stopped. It is 0.80. When e < At that time, the robot is allowed to perform close-range re-inspection; when ≤e< When e ≥ 1, the robot does not perform close-range re-inspection, but instead observes from a distance, moves at low speed, or waits in a safe location; When the time comes, the robot stops entering the target area and either returns to its home position or docks safely.

[0050] S7: Evidence Completeness Threshold Judgment (Secondary Branch Node) This judgment determines whether the current evidence is sufficient, whether the conclusion of the fault is clear, and whether data quality, evidence consistency, and anomaly credibility all meet the preset thresholds. If the result is negative: proceed to S8 to initiate the proactive re-inspection strategy; The determination result is: proceed to the final confirmation stage of S9 anomaly.

[0051] S8: The active re-inspection and re-collection mechanism system automatically selects a re-inspection strategy and re-collects data by adjusting the observation angle, detection distance, carrier pose, and detection sensor mode. After the re-inspection is completed, the new data is sent back to S6 to conduct a re-evaluation of the observation quality and credibility, forming a small loop iteration until the evidence meets the standards.

[0052] S9: Final Confirmation of Equipment Anomaly (Level 3 Branch Node): Determine whether there is a substantial anomaly in the equipment based on compliance evidence. The determination result is: Execute S10 exception handling; If the result is negative: proceed directly to S11 to update the archive and schedule subsequent tasks.

[0053] S10: Abnormal closed-loop handling and evidence retention. Local security handling operations are carried out for confirmed equipment abnormalities, alarm push is triggered simultaneously, and the entire evidence chain is completely retained to complete the abnormal event closure; after the handling is completed, the data is imported into S11.

[0054] S11: File update and task rescheduling uniformly update the equipment health status file, and at the same time, re-sort and reschedule the remaining unfinished inspection tasks based on the results of this inspection. The process can be reverted to initiate a new round of risk-driven inspections, achieving an overall closed loop.

[0055] Furthermore, the S9 anomaly determination rule is as follows: The system categorizes inspection results based on observation quality, evidence consistency, and anomaly credibility. When observation quality or evidence consistency is insufficient, the system is deemed to require re-inspection and initiates an active re-inspection process. When the anomaly credibility reaches the confirmation threshold, local alarms are triggered, evidence is stored, events are uploaded, and tasks are rescheduled. When the anomaly credibility is below the exclusion threshold, the system updates the equipment health record and continues inspection. When the system is in a suspected anomaly zone, re-inspection actions such as re-taking, changing angles, changing distances, or switching detection modes are performed, provided safety constraints are met. If environmental or power conditions do not meet the re-inspection requirements, the system retreats, stops, or returns to base, recording the reason for non-compliance.

[0056] Example 1: Routine Autonomous Inspection

[0057] The robot retrieves operational data and historical records of the equipment at its docking point. A certain power supply switch, due to increased load and nearing its inspection cycle limit, poses a higher overall risk than ordinary equipment; therefore, the system advances its task. After reaching the target position, the robot collects instrument images, infrared images, and sound. No abnormalities are detected in the instrument identification or infrared temperature rise; the observation quality and evidence consistency meet the requirements. The system updates the equipment health record and continues with the next task.

[0058] Example 2: Active re-inspection of suspected joint overheating

[0059] The robot detected a temperature rise at a cable joint for the first time from the side of the channel, but the infrared target area was partially obstructed, leading to... Low, abnormal credibility The system is currently in the verification phase. It moves to a safer position and adjusts the gimbal angle for retesting, while simultaneously reading the corresponding loop load. After the retest, the temperature rise and load increase mutually support each other, improving the consistency of the evidence, and the anomaly confidence exceeds the confirmation threshold. The robot generates a high-priority event, records the original thermal image, the retested thermal image, operational data, pose, and decision-making process, and reorders the inspection tasks of adjacent related devices.

[0060] Example 3: Communication Interruption and Environmental Deterioration

[0061] When the robot detects a decrease in communication quality during its inspection, it switches to local authorization mode and saves the data. Subsequently, if the environmental hazard level in the target area reaches the threshold restricting approach, the robot cancels the close-range re-inspection, retreats to a safe position, and continues to monitor the environmental status. If the hazard level continues to rise, it returns to base; if the environment recovers and the remaining battery power allows, it rejoins the task queue. After communication is restored, the system prioritizes uploading the abnormal event and the reason for the incomplete task.

[0062] Example 4: The robot mobile platform can be wheeled, tracked, or rail-mounted; the positioning method can be laser, vision, inertial, odometer, wireless positioning, or a combination thereof; partial discharge detection can use ultrasonic, transient ground voltage, or other detection methods suitable for the target equipment. Weights and thresholds can be configured based on equipment type, mine safety regulations, and historical data. For example... Figure 2 As shown, for example: (1) Underlying carrier: Explosion-proof mobile chassis 101 The robot's basic walking support mechanism is equipped with two sets of drive wheels on the left and right sides, serving as the mounting base for all upper modules: It receives movement commands from the explosion-proof control box 102 to complete the movement; The left side is connected to the explosion-proof power supply assembly 108 for power supply; The right side is equipped with a communication component 107 to enable data interaction.

[0063] (2) Explosion-proof control box 102 (edge ​​computing, evidence storage unit)

[0064] The core of the whole machine's decision-making and data processing is located in the center above the chassis, and undertakes local computing power tasks at the edge: Local execution functions include: risk orchestration, evidence consistency calculation, proactive review decision-making, and security degradation control. Sending commands downwards to drive the explosion-proof mobile chassis; It connects to the upper gimbal, the left-side environmental detection component, and the right-side positioning and obstacle avoidance component respectively, and aggregates all perception data to make calculations and decisions.

[0065] (3) Upper-level execution acquisition mechanism

[0066] 1. Gimbal 103

[0067] Installed on the top of the explosion-proof control box, it supports pitch and rotation dual-degree-of-freedom movements. The top of the device carries a multi-modal detection component, and the detection head can be aligned with the measuring points of the power equipment at multiple angles through attitude adjustment.

[0068] 2. Multimodal detection component 104

[0069] Mounted on the top of the gimbal, the detection head is aimed at the power equipment and simultaneously collects multi-dimensional data on the target points of the equipment; the detection modes include: visible light imaging, infrared thermal imaging, sound pickup, and partial discharge detection.

[0070] (4) Side sensing module

[0071] 1. Left side: Environmental monitoring component 105

[0072] Connect to the explosion-proof control box to collect on-site environmental parameters. Detection targets include: combustible / toxic gases, smoke, and ambient temperature and humidity.

[0073] 2. Right side: Positioning and obstacle avoidance component 106

[0074] It is connected to an explosion-proof control box and adopts a fusion solution of lidar, vision camera and inertial navigation; its function is to generate robot inspection poses and autonomously complete the determination of restricted areas, obstacle detour and collision avoidance constraint control.

[0075] (5) Auxiliary support components

[0076] 1. Right side: Communication component 107

[0077] Mounted on the right side of an explosion-proof mobile chassis, it features wireless communication transmission and local data caching capabilities during network outages, ensuring that inspection data is not lost when uploaded externally or when the network is interrupted.

[0078] 2. Left side: Explosion-proof power supply assembly 108

[0079] Mounted on the left side of the explosion-proof mobile chassis, it provides explosion-proof safety power supply for the entire machine, and also features real-time monitoring of remaining power and autonomous recharging.

[0080] (6) All functional components are electrically and signal connected through explosion-proof cables and intrinsically safe communication buses, which complies with the electrical safety specifications for explosion-proof scenarios; the appearance of each module and the actual installation position can be flexibly adjusted according to the working conditions of the substation site.

[0081] (7) The three types of perception data, namely, power supply of explosion-proof power components, chassis bearing and walking, multi-modal equipment detection, environmental perception, and positioning and obstacle avoidance, are integrated into the edge computing unit of the explosion-proof control box. Risk assessment, path planning, and re-inspection decision are completed locally. Data is uploaded through the communication component. At the same time, the pan-tilt unit adjusts the detection angle according to the instructions to realize fully automatic safety inspection in explosion-proof scenarios such as high-risk substations.

[0082] This invention provides an autonomous inspection robot system for underground power equipment in coal mines, which solves the problems of fixed task execution sequence, insufficient reliability of multimodal detection results, inability to autonomously re-inspect suspected anomalies, and difficulty in safe closed-loop operation when communication is interrupted or the underground environment deteriorates in existing inspection methods.

[0083] This invention dynamically assesses equipment risks and schedules inspection tasks by comprehensively considering the real-time operating status of the equipment, historical defects, and the timeliness of inspections. The credibility of anomalies is calculated by combining observation quality and consistency of multimodal evidence. Suspected anomalies are re-examined by autonomously choosing to re-take photos, change the detection angle, adjust the detection distance, or switch the detection mode. Simultaneously, the robot's movement and detection actions are restricted by the safety constraints of the downhole environment, and local safety degradation, anomaly handling, and evidence preservation are performed when communication is limited.

[0084] This improves the timeliness, accuracy, and autonomy of underground power equipment inspections, reduces false alarms, missed alarms, and the risks associated with personnel conducting underground inspections, and ensures the safe and stable operation of the coal mine power supply system.

[0085] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0086] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An autonomous inspection robot system for underground power equipment in coal mines, characterized in that, include: Explosion-proof inspection robot side and ground and equipment side; The explosion-proof inspection robot includes: a risk-driven task orchestration module, a positioning, navigation and safety constraint module, a multimodal acquisition and quality evaluation module, an evidence consistency and anomaly diagnosis module, an active re-inspection and action decision module, and a local safety degradation and evidence storage module. The ground and equipment side includes: a ground monitoring and manual confirmation terminal, a power equipment operation data interface module, and a historical defect and inspection file module; The ground monitoring and manual confirmation terminal, the power equipment operation data interface module, and the historical defect and inspection archive module are all signal-connected to the risk-driven task orchestration module. The risk-driven task orchestration module is sequentially connected to the positioning, navigation and safety constraint module, the multimodal acquisition and quality evaluation module, the evidence consistency and anomaly diagnosis module, and the active re-inspection and action decision module, and is used for inspection planning and dynamic task updating based on inspection results. The evidence consistency and anomaly diagnosis module is connected to the ground monitoring and manual confirmation terminal signal through the local security degradation and evidence storage module, and is used to diagnose anomalies in the inspection results. The active re-inspection and action decision-making module is connected to the ground monitoring and manual confirmation terminal, and the positioning, navigation and safety constraint module, respectively, and triggers re-inspection feedback based on the diagnostic results.

2. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The risk-driven task orchestration module is used to create a risk profile for each piece of equipment or part to be inspected.

3. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The positioning, navigation and safety constraint module is used to maintain maps of tunnels and substations, equipment inspection pose sets, and restricted areas.

4. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The multimodal acquisition and quality evaluation module is used to control the pan-tilt unit, focal length, supplementary lighting, and detection distance, and to acquire visible light, infrared, sound, partial discharge, and environmental information from the target device.

5. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The evidence consistency and anomaly diagnosis module is used to map the detection results of different modalities to a unified device location and time window to determine whether the evidence supports each other.

6. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The proactive review and action decision module is used to select the next action based on the risk of anomalies, consistency of evidence, and observation quality.

7. The autonomous inspection robot system for underground power equipment in coal mines according to claim 1, characterized in that, The local security degradation and evidence storage module is used to continue executing pre-authorized inspection and re-inspection actions when communication quality is insufficient, and to perform tamper-proof associated storage of original data, quality indicators, decision basis and action logs.

Images