Rail-mounted teleoperated robotic system for wind turbines and control method thereof

Abstract

The application provides a track type remote operation robot system for a wind turbine generator and a control method thereof, comprising a walking track laid along the inside of a machine cabin and a tower drum of the wind turbine generator, a track robot body moving along the track, a six-degree-of-freedom mechanical arm installed on the robot body, a terminal quick-change operation device, a multi-source visual perception and positioning module, a terminal six-dimensional force / torque perception module, a remote operation terminal and a safety management and control platform. The residual vibration amplitude at the end of the mechanical arm is controlled to be sub-millimeter level through a three-stage series hybrid anti-vibration system; a hybrid impedance / admittance force-position control strategy is adopted to realize soft contact operation; a multi-level permission control system and a double-check mechanism based on dynamic trust score are constructed; and a shared autonomous control theory is introduced to realize dynamic allocation of human and machine permissions. The defects that the existing track inspection robot can only perceive alarms but cannot physically operate are solved, the efficiency of emergency disposal of the wind turbine generator is significantly improved, and the frequency and safety risk of personnel boarding are reduced.

Description

Technical Field

[0001] This invention relates to the field of wind power operation and maintenance robot technology, specifically to a track-mounted remote operation robot system for wind turbine generators and its control method. Background Technology

[0002] As a crucial component of renewable energy, wind power has seen its share in my country's energy structure steadily increase in recent years. With onshore wind power expanding into desert and Gobi regions and offshore wind power extending into deep-sea areas, the single-unit capacity of wind turbines is continuously increasing, with nacelles and tower heights generally exceeding 100 meters, and deep-sea turbines reaching distances of over 50 kilometers from the shore. Simultaneously, the operating environment of wind turbines is becoming increasingly harsh: desert turbines face challenges such as sandstorms, large temperature differences, and dust; while deep-sea turbines are affected by severe sea conditions such as salt spray, typhoons, and high waves.

[0003] In terms of unit operation and maintenance, the current common practice is to conduct regular manual on-board inspections and handle faults on-site. When a unit experiences a fault alarm or requires manual intervention (such as resetting, pressing buttons, switching circuit breakers, tightening wiring, or replacing modules), maintenance personnel must drive or travel by boat from their base to the unit's location and climb to the engine room or tower platform to perform the operation. For remote units, a single round trip can take several hours or even a whole day; for deep-sea units, the effective operating window is extremely limited due to tides and waves. This "mandatory on-site" operation and maintenance method has prominent problems such as delayed emergency response, high personnel safety risks, low operation and maintenance efficiency, and poor accessibility for remote units.

[0004] Existing inspection robots or systems are generally classified as "inspection" robots, with core functions of sensing, detection, and alarm. They lack the ability to physically operate equipment inside wind turbine units. Specifically, they cannot press buttons, reset switches, or close or open circuit breakers; they cannot tighten wiring terminals inside control cabinets; and they cannot replace control modules or other components. In other words, existing track-mounted robots can only "see" and "report," not "move" or "do." When a fault requiring actual operation occurs in the unit, even if the robot is already at the location, maintenance personnel must still arrive to handle the situation, failing to fundamentally solve the core issues of remote emergency response and reducing the frequency of boarding. Furthermore, existing solutions do not address access control, hierarchical verification, and manual supervision mechanisms for remote operation, failing to meet the power industry's management requirements for safe production and prevention of misoperation. Therefore, there is an urgent need for a robot system and its control method that can move along tracks inside wind turbine units and remotely perform various physical tasks. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a track-mounted remote operation robot system and its control method for wind turbine units. This invention solves the technical defects of existing track inspection robots that can only "sensitize alarms" but cannot "physically operate," enabling remote contact operation of equipment inside the wind turbine unit control cabinet, while also possessing a complete safety management mechanism.

[0006] According to one objective of the present invention, the present invention provides a track-mounted remote-operated robot system for wind turbine generators, comprising: The travel track is laid inside the wind turbine nacelle and on each platform of the tower. The track robot body moves along the walking track, and the track robot body includes a mobile chassis and an on-board real-time controller; A six-degree-of-freedom robotic arm is mounted on the mobile chassis, and the end of the six-degree-of-freedom robotic arm is equipped with an electro-pneumatic integrated standard quick-change interface. Multiple end effectors are connected to the six-degree-of-freedom robotic arm via the quick-change interface. Each end effector includes at least one of a single-finger elastic pressing head, a two-finger parallel gripper, a constant torque electric screwdriver module, and a four-finger module gripping fixture. The multi-source visual perception and positioning module includes a color camera mounted on the end of the robotic arm and a depth camera mounted on a mobile chassis. The six-dimensional force / torque sensing module at the end of the robotic arm is integrated between the end flange and the quick-change interface for real-time detection of end contact force and torque. The remote operation terminal, deployed in the central control center, is used to send operation commands and receive robot status and video data; The security management platform is used to realize user identity authentication, hierarchical access control, operation approval and full-process traceability management; The track robot body also includes a three-level series hybrid vibration damping system, which includes a floating suspension walking vibration reduction system, a multi-directional damping vibration reduction structure between the chassis and the mounting frame, and an active vibration suppression system based on acceleration feedback.

[0007] Furthermore, in the aforementioned three-stage series hybrid vibration damping system: The floating suspension travel vibration reduction system includes: each travel wheel set is connected to the chassis frame through a suspension floating arm, one end of the floating arm is hinged to the chassis via a pivot, and the other end is fixed to the chassis through a rubber-steel composite elastic damping component. The output end of the rotary drive motor and the travel wheel are connected by a double diaphragm flexible coupling to transmit the drive torque. The multi-directional damping vibration reduction structure includes three types of vibration reduction components arranged in parallel: cross-link elastic sliding structure, arc-shaped buffer rod sliding structure, and sliding rod-mounting cylinder mating structure; The active vibration suppression system includes: four piezoelectric stacked actuators symmetrically arranged between the mounting frame and the chassis, a triaxial MEMS accelerometer installed at the end flange of the robotic arm, and an adaptive active vibration control strategy based on the filter-x least mean square algorithm running on the vehicle-mounted real-time controller.

[0008] Furthermore, the onboard real-time controller runs an end-effector force-position control algorithm based on an impedance / admittance hybrid strategy, the algorithm comprising: The joint space impedance control layer models the robotic arm as a second-order dynamic system of mass-spring-damped system, based on the dynamic equations. Adjusting the desired inertia matrix Md and the damping matrix B d and stiffness matrix K d This enables the robotic arm to exhibit the desired compliant characteristics in response to contact forces; The Cartesian space admittance control layer uses a six-dimensional force sensor to measure force F during the precision force tracking stage. ext As input, and with the end-effector pose correction ΔX as output, according to the transfer relationship Adjust the desired trajectory in real time; The mission state machine automatically switches control modes according to the operation phase: pure position control is used during the aerial motion phase; when it enters the range of 5mm from the target, it switches to impedance control mode; and after entering the precision force tracking phase, it switches to admittance control mode.

[0009] Furthermore, the on-board real-time controller also runs an algorithm for real-time identification of operating states and adaptive parameter adjustment based on force feedback curves: Data from a six-dimensional force sensor was acquired at a frequency of 1kHz, and a force-time response curve was constructed. By detecting in real time whether the first derivative of dF / dt exceeds a preset threshold, the three-stage characteristics of the force-time response curve are identified: the non-contact stage, the elastic compression stage, and the bottoming-out step stage. Based on the recognition results, determine whether the button is triggered, the circuit breaker is in place, or the screw is tightened, and automatically trigger a retraction or stop command. An online adaptive control law based on Lyapunov stability theory is introduced to simultaneously estimate the contact environment stiffness k online. e and contact position x e This allows the actual contact force to gradually converge to the desired value.

[0010] Furthermore, the multi-source visual perception and localization module runs an intelligent operation control algorithm based on visual servoing and trajectory optimization: The camera intrinsic parameters were calibrated using Zhang Zhengyou's chessboard calibration method, and the hand-eye calibration was performed using the AX=XB equation solution method based on random sampling consistency. Deploy a lightweight object detection and instance segmentation algorithm based on Mask DINO, and output pixel-level segmentation masks and class confidence scores for each target device in color images; By using a mask to clip the depth point cloud, local geometric descriptors are extracted through a fast point feature histogram. The generalized iterative nearest point algorithm is used to accurately register the point cloud with the pre-constructed CAD template, and the six-DOF pose of the target device is output. It adopts an image-based visual servo control framework and integrates the RRT-Connect real-time obstacle avoidance path planning module.

[0011] Furthermore, the security management platform includes a hierarchical permission model based on roles and dynamic trust scoring: A role-based access control mechanism is adopted, which divides users into three role levels: inspector, operator, and supervisor / approver. Each user is bound to a unique SM2 public-private key pair digital identity certificate, which, combined with facial liveness detection, fingerprint recognition dual biometric verification, and dynamic SMS OTP verification code, constitutes a three-factor authentication mechanism. Based on the hazard level of the operation task, the real-time operating status of the unit, and the real-time behavioral characteristics of the operator, the confidence level of operation authorization is calculated in real time through a weighted scoring model. When the confidence level is lower than the threshold, the level of authorization requirement is automatically increased.

[0012] Furthermore, the security management platform also includes: The dual-verification mechanism requires that for operations such as resetting the main control system, opening and closing the circuit breaker, and replacing the control module, the operator must submit an application, and the supervisor must verify the equipment status through real-time video and authorize it with an SM2 digital signature before the operation instructions can be executed. The operation tickets are managed electronically, recording the entire process including the operation initiation timestamp, operator ID, approver digital signature, operation task type, target equipment number, video clips of key steps, six-dimensional force / torque time series curve, and operation result status. All records are written into a blockchain-based tamper-proof evidence database after being hashed and digested using SM3.

[0013] Furthermore, the system also includes a dynamic human-machine permission allocation mechanism based on the theory of shared autonomous control: Machine-led mode: In standardized operation tasks, the robotic arm executes autonomously according to the pre-programmed path, and the on-board controller completes precise alignment and contact operations through visual servo and force control closed loop. The operator is only responsible for macroscopic confirmation and process monitoring. Human-machine shared control mode: When the field of view is obstructed, the target recognition confidence is lower than the threshold, or the force sensor detects abnormal fluctuations, the system automatically transfers the control of the end part of the freedom to the remote operator, while the machine end simultaneously retains collision detection and contact force limiting protection; Automatic permission revocation mechanism: When an operator is detected to be unresponsive for an extended period or to have a persistent operational deviation, control is automatically reclaimed to the machine system and a safety rollback action is executed, while an alarm notification is triggered.

[0014] Furthermore, the system also includes an operational knowledge base and a self-optimization mechanism for standardized operating procedures: After each operation is completed, the visual recognition results, six-dimensional force feedback curve, operation success / failure judgment flag and control parameter group are written into the operation knowledge database; The knowledge base backend regularly runs a Bayesian-optimized hyperparameter search to automatically optimize the standardized operating procedure control parameters for different device types and operating scenarios. The parameter optimization results are synchronized to all online robot nodes using an incremental update method.

[0015] According to another objective of the present invention, the present invention provides a control method for the above-described track-mounted remote-operated robot system for wind turbine generators, comprising the following steps: S1. The track robot moves along the walking track to the target work position and activates the three-level series hybrid vibration damping system; S2, the multi-source visual perception and localization module performs instance segmentation and six-degree-of-freedom pose estimation on the target device, and outputs the precise pose of the target device; S3. Automatically select the corresponding end effector according to the type of operation task, and the vehicle controller plans the movement trajectory of the robotic arm. S4. During the aerial movement phase of the robotic arm, pure position control is used. When it enters the range of 5mm from the target, it switches to impedance control mode to achieve compliant contact. S5. After entering the precision force tracking stage, switch to admittance control mode to achieve precise force control based on feedback from a six-dimensional force sensor. S6. Based on the force feedback curve, the operation status is identified in real time, and the retraction action is automatically executed after the operation is completed. S7. The entire operation is subject to the permission verification and supervision of the security management platform. Critical operations can only be executed after being reviewed and authorized by two people. S8. After the operation is completed, the data of this operation is written to the knowledge base, and the Bayesian optimization parameters are tuned regularly in the background.

[0016] This invention upgrades the function of the internal track-based mobile platform of a wind turbine from "perception-alarm" to "perception-operation," fundamentally solving the technical deficiency of existing inspection robots that can only "see" but not "do." By integrating a six-degree-of-freedom robotic arm and a replaceable end effector, it can remotely perform various physical operations such as button pressing, circuit breaker opening and closing, wiring tightening, and module replacement, reducing emergency fault response time from days to minutes. A single operation can save 4 to 24 hours of maintenance round-trip time, significantly reducing unplanned downtime losses. It also significantly reduces the frequency of maintenance personnel climbing the tower, fundamentally reducing the safety risks of operations at high altitudes and in harsh sea conditions, providing an efficient and safe remote maintenance solution for wind turbines in remote areas such as deep-sea areas and deserts. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the architecture of the track-type remote-operated robot system according to an embodiment of the present invention; Figure 2 This is another structural schematic diagram of the track-type remote-operated robot system according to an embodiment of the present invention; Figure 3 This is a block diagram of the impedance / admittance hybrid force-potential control algorithm architecture according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating the operation control process based on visual servoing and force control according to an embodiment of the present invention. Figure 5 This is a block diagram of the access control system of the security management platform according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the state machine for switching the human-machine shared control mode according to an embodiment of the present invention; Figure 7 This is a flowchart of the control method for a track-type remote-operated robot according to an embodiment of the present invention.

[0019] In the diagram: 1. Walking track; 2. Tracked robot body; 3. Mobile chassis; 4. Six-DOF robotic arm; 5. End effector quick-change interface; 6. End effector; 7. Eye-in-hand color camera; 8. Eye-to-hand depth camera; 9. Six-dimensional force / torque sensor; 10. Walking wheel; 11. Rubber-steel composite elastic damping component; 12. Double diaphragm flexible coupling; 13. First guide rod; 14. First slider; 15. Mounting bracket; 16. Second guide rod; 17. Second slider; 18. Connecting rod; 19. Support base; 20. Arc-shaped buffer rod; 21. Rubber damping pad; 22. Sliding rod; 23. Mounting cylinder; 24. Disc-shaped elastic component. Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0021] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0022] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] Example 1 like Figure 1 and Figure 2As shown, this embodiment of the invention provides a track-type remote operation robot system for wind turbine generators, which mainly includes a walking track 1, a track robot body 2, a six-degree-of-freedom robotic arm 4, an end effector quick-change interface 5, multiple sets of end effectors 6, a multi-source vision perception and positioning module, an end effector six-dimensional force / torque perception module, a remote operation terminal, and a safety management and control platform.

[0024] The travel track 1 uses I-beam shaped aluminum alloy profiles and is laid in front of the control cabinet inside the wind turbine nacelle and on each platform of the tower. The total length of the track is customized according to the nacelle layout, generally 5 to 15 meters. The track surface is anodized, which provides corrosion resistance and wear resistance.

[0025] The track-mounted robot body 2 includes a mobile chassis 3 and an onboard real-time controller. The mobile chassis 3 engages with the walking track 1 via four sets of wheels. Each set of wheels includes two support wheels and one drive wheel. The drive wheel is driven by a DC brushless servo motor, with a maximum walking speed of 0.5 m / s and a positioning accuracy of ±5 mm. The onboard real-time controller uses an industrial-grade embedded computer equipped with an Intel Core i7 processor, pre-installed with the Ubuntu 22.04 LTS operating system and ROS 2 Humble distribution, and runs real-time control algorithms and visual inference tasks.

[0026] The six-degree-of-freedom robotic arm 4 adopts a serial articulated structure and is mounted on the mounting frame of the mobile chassis 3. Its end-effector working space radius is 850mm. Each joint is driven by a high-precision hollow harmonic reducer paired with a brushless servo motor. An absolute magnetic encoder achieves closed-loop positioning of each joint, with a joint repeatability accuracy of ±0.03°. The robotic arm has a folding storage compartment; when folded in non-operational mode, its total height is 350mm. During operation, it unfolds as needed, fully covering the control cabinet panel area and terminal block area. The end effector of the robotic arm is equipped with an electro-pneumatic integrated standard quick-change interface 5, integrating 12 electrical signal channels and 4 pneumatic channels. The tool switching time is 12 seconds, and automatic tool identification is achieved through an RFID chip.

[0027] The end effector 6 includes: Single-finger elastic press head: used for light touch of buttons, reset switches, etc. The head is made of silicone material and has a built-in pressure spring; Two-finger parallel gripper: used for rotary switch handles and small air switch levers, with a maximum clamping force of 50N; Torque-controlled electric screwdriver module: used for tightening control cabinet wiring terminals, torque range 0.5~3.0N. m, accuracy ±5%; Four-finger module gripper: Used to control the insertion, removal and replacement of modules, and has a guiding and positioning function.

[0028] The multi-source vision perception and positioning module includes an Eye-in-hand color camera 7 (1920×1080 resolution, 30fps) mounted on the end flange of the robotic arm and an Eye-to-hand depth camera 8 (IntelRealSense D455, depth ranging range 0.3~10m) mounted on the mobile chassis.

[0029] The end effector's six-dimensional force / torque sensing module employs a silicon strain beam six-dimensional force / torque sensor 9, integrated between the robotic arm's end effector flange and quick-change interface 5. The sensor can simultaneously detect forces in three orthogonal directions (Fx, Fy, Fz) and torques in three directions (Tx, Ty, Tz), with a measurement range of ±200N / ±20N. m, detection accuracy ±0.1N / ±0.01N m, sampling frequency 1kHz. The sensor has a built-in temperature compensation circuit and cross-axis crosstalk cancellation algorithm (crosstalk error <2%), and transmits data to the vehicle controller in real time via EtherCAT high-speed bus, with an end-to-end delay of no more than 2ms.

[0030] The remote control terminal is deployed in the central control center, adopts a B / S architecture, and provides a real-time display interface for multiple video streams (RGB + depth + panorama), overlaid with an augmented reality operation guidance layer. Operators send operation commands via a six-DOF force feedback handle or mouse drag and drop, and the total round-trip delay of control commands from the central control center to the robot end effector is controlled within 200ms.

[0031] The security management platform is deployed on the central control center server to achieve user authentication, hierarchical access control, operation approval, and full-process traceability management.

[0032] The fundamental innovation of this invention lies in upgrading the functional positioning of the internal track-moving platform of the wind turbine from "perception-alarm" to "perception-operation". It proposes an operation robot system that integrates a walking track, a track robot body, a six-degree-of-freedom robotic arm, an end-effector quick-change operation device, a multi-source vision perception and positioning module, an end-effector six-dimensional force / torque perception module, a remote operation terminal, and a safety management and control platform.

[0033] This invention uses a track-based mobile platform as a carrier and integrates a six-degree-of-freedom robotic arm with end-effector quick-change capability. This enables the system to perform remote contact-type physical operations on devices such as buttons, reset switches, air switch levers, wiring terminals, and control modules inside the wind turbine control cabinet, fundamentally solving the technical deficiency of existing track robots that can only "see" but not "do".

[0034] Example 2 This embodiment, based on Embodiment 1, adopts a three-stage series hybrid vibration damping system, such as... Figure 1 and Figure 2As shown, the three-stage series hybrid vibration damping system in this embodiment is designed to address the continuous wideband low-frequency vibration (dominant frequency 0.3~5Hz) of the wind turbine nacelle under operating conditions and the impact excitation of the track joint. Specifically, it includes: Level 1: Floating suspension travel damping system (passive damping) Each of the 10 sets of traveling wheels is connected to the chassis frame of the mobile chassis 3 via a suspension floating arm: one end of the floating arm is rotatably hinged to the mobile chassis 3 via a pivot, and the other end is fixed to the mobile chassis 3 via a rubber-steel composite elastic damping element 11; the output end of the rotary drive motor and the traveling wheel 10 are connected by a double diaphragm flexible coupling 12 to transmit drive torque, ensuring uninterrupted power transmission during the floating of the wheel set. When there are minor unevennesses on the track surface or when the track joint generates a step impact, the traveling wheel 10 can float up and down around the pivot. The elastic damping element dissipates the impact energy through compression-tension cycles, and the high-frequency vibration component (>10Hz) is attenuated by 70% before being transmitted to the chassis, while ensuring continuous contact between the wheel set and the track, eliminating the instantaneous slippage phenomenon under rigid connection conditions.

[0035] Level 2: Multi-directional damping vibration reduction structure between chassis and mounting frame (passive vibration reduction) Three types of shock absorption components are arranged in parallel between the chassis frame and the robotic arm mounting frame: 1. Cross-link elastic sliding structure: A first guide rod 13 is provided inside the first fixed shell on one side of the mobile chassis 3, and two first sliding blocks 14 that can slide freely along the guide rod are installed on it; a second guide rod 16 is provided inside the second fixed shell on one side of the mounting bracket 15, and two second sliding blocks 17 are installed on it; two connecting rods 18 are arranged crosswise and hinged at the intersection point, with one end of each connecting rod 18 hinged to the first sliding block 14 and the other end hinged to the second sliding block 17; a polyurethane elastic element is sleeved on each guide rod, with both ends abutting the end face of the sliding block and the inner wall of the fixed shell, respectively. When the chassis is subjected to vibration excitation in any direction, the cross-link linkage drives the sliding blocks to slide back and forth along the guide rod, and the elastic element converts the vibration mechanical energy into heat energy dissipation in the compression-tension cycle.

[0036] 2. Arc-shaped buffer rod sliding structure: Arc-shaped sliding grooves are opened on the support seats 19 on both sides of the movable chassis 3. Arc-shaped buffer rods 20 extend from the mounting bracket 15 and slide within the arc-shaped grooves. Guide pins are provided on the buffer rods to engage with the guide grooves of the support seats. Rubber damping pads 21 are laid at the bottom of the grooves. Rotational vibration causes relative displacement of the buffer rods within the arc-shaped grooves, and the rubber damping pads 21 dissipate torsional vibration energy through shear deformation.

[0037] 3. Sliding rod-mounting cylinder mating structure: One end of the sliding rod 22 is rotatably mounted on the movable chassis, and the mounting cylinder 23 is rotatably mounted on the mounting frame 15. The upper end of the sliding rod 22 slides into the mounting cylinder 23, and a disc-shaped elastic element 24 is provided between the bottom wall of the cylinder and the end of the rod. This structure provides stroke limit under vertical large amplitude conditions to prevent hard collision between the mounting frame and the chassis.

[0038] After two-stage passive vibration reduction treatment, the high-frequency vibration amplitude transmitted from the track to the robotic arm mounting base is reduced by 88%, and the residual vibration amplitude at the end of the robotic arm is controlled within ±0.25mm.

[0039] Level 3: Active vibration suppression system based on acceleration feedback (active vibration reduction) Four piezoelectric stacked actuators (PZTs) are symmetrically arranged between the mounting frame and the chassis, and a triaxial MEMS accelerometer (sampling frequency 10kHz) is installed at the end flange of the robotic arm. The system employs an adaptive active noise control strategy based on the Filtered-x Least Mean Square (FxLMS) algorithm: the onboard real-time controller continuously acquires the end-effector acceleration signal, predicts vibration propagation characteristics through an identified secondary path filtering model, updates the FxLMS weight vector online, and drives the PZT actuators to output an anti-phase compensation force, thereby achieving real-time cancellation of broadband residual vibration. The active vibration damping layer can further compress the end-effector vibration amplitude to ≤0.08mm during the static alignment phase.

[0040] Example 3: Impedance / Admittance Hybrid Force-Position Control Algorithm like Figure 3 As shown, the end-effector force-position control algorithm of this invention adopts a hierarchical architecture that combines impedance control and admittance control: Joint spatial impedance control layer A joint spatial impedance control loop is built into the servo driver of each joint of the robotic arm, and the robotic arm is modeled as a second-order dynamic system of mass-spring-damped system. The desired inertia matrix M is adjusted. d Damping matrix B d and stiffness matrix K d This allows the robotic arm to exhibit the desired compliant characteristics in response to contact forces. The core dynamic equation of the impedance controller is: ; Where X is the actual pose of the end effector, X d For the desired pose, F ext This is the external contact force / torque vector measured by a six-dimensional force sensor. Through offline identification and on-site self-tuning, parameter sets (M) are pre-configured for different operational tasks such as button pressing, button triggering, and wiring tightening. d B d K dThe robotic arm uses pure position control during its aerial movement; when it enters a range of 5 mm from the target device, it switches to impedance control mode to achieve compliant contact and prevent collisions or impacts from damaging the equipment or tools.

[0041] Cartesian space admittance control layer During operations requiring precise force tracking, the system switches to Cartesian space admittance control mode. Admittance control is based on the measured force F from a six-dimensional force sensor. ext As input, and with the end-effector pose correction ΔX as output, the desired trajectory is adjusted in real time based on force feedback. The transmission relationship is as follows: ; In precision assembly scenarios, force control accuracy can reach the level of 0.1 N.

[0042] The mode switching logic driven by the mission state machine is as follows: pure position control is used during the aerial motion phase; when the distance between the end effector and the target is less than 5 mm, it switches to impedance control to achieve compliant contact; after entering the precision force tracking phase, it switches to admittance control; each phase transitions smoothly to eliminate force abrupt changes during switching.

[0043] Example 4: Real-time identification of operating state and adaptive parameter adjustment based on force feedback curve like Figure 4 As shown, to address the varying operational force requirements of different devices within the control cabinet (e.g., a light touch button trigger force of approximately 2...), Air switch lever push force 8-10 The tightening torque for the wiring terminals is 1.0 to 2.0. This invention designs a real-time operation state identification algorithm based on force feedback time curve.

[0044] Taking button pressing as an example: When the robotic arm performs the pressing action, the onboard controller collects force data in the Z direction from the six-dimensional force sensor at a frequency of 1 kHz, constructing a force-time response curve. This curve exhibits three-stage characteristics: the non-contact stage ( ), elastic compression stage ( (linear increase with displacement), bottoming-out step phase () After mutation (Approaching saturation). The system detects in real time... If the first derivative exceeds a preset threshold, the system determines the "button triggered" state and then instructs the robotic arm to perform a retraction action to avoid excessive pressing that could damage the button mechanism.

[0045] Taking the terminal tightening operation as an example: When the electric screwdriver module is working, the system monitors in real time. Directional torque feedback occurs when the torque reaches a preset target value (e.g., 1.2). When the screw is tightened, a stop signal is immediately output, and the tightening torque value, tightening angle and operation timestamp are recorded to form a complete operation data chain for each screw, which can be used for compliance traceability in the future.

[0046] To address the problem of discrepancies between theoretical control force and actual contact force caused by unknown stiffness in the actual environment, this invention introduces an online adaptive parameter estimation method based on gradient descent: the onboard controller synchronously estimates the stiffness of the contact environment online during operation. and contact position A dynamic equation for force tracking error is constructed, and an adaptive control law is designed based on Lyapunov stability theory to make the actual contact force gradually converge to the desired force value, thus ensuring the repeatability of the operation.

[0047] This invention proposes a real-time operation status identification algorithm based on the feedback time curve of a six-dimensional force sensor. By real-time differential detection of the three-stage characteristics (non-contact segment, elastic compression segment, and bottoming-out step segment) of the force-time response curve, it determines the completion status of operations such as button triggering, circuit breaker positioning, and screw tightening, and automatically triggers a reversal or stop command accordingly to prevent excessive operation from damaging the equipment.

[0048] Furthermore, to address the problem of force control deviation caused by unknown contact environment stiffness parameters, this invention introduces an online adaptive control law based on Lyapunov stability theory. The onboard controller synchronously estimates the environmental stiffness k online during operation. e and contact position x e A dynamic equation for force tracking error is constructed, and control parameters are adaptively adjusted to gradually converge the actual contact force to the desired value, ensuring the repeatability of operation and the accuracy of force control.

[0049] Example 5: Security Management Platform and Access Control - Multi-level Access Control and Operation Traceability System Based on Dynamic Trust Scoring and National Cryptographic Algorithms like Figure 5 As shown, the security management platform of this invention adopts a hierarchical permission model based on roles and dynamic trust scoring: 1) Hierarchical permission model based on role and dynamic trust scoring The system security management platform employs a role-based access control (RBAC) mechanism, dividing all logged-in users into three basic role levels: Inspector (can only view video streams and system status data), Operator (can initiate operation requests and execute low-risk operations), and Supervisor / Approver (can review operation requests, authorize high-risk operations, and has the highest authority to suspend operations). Each user is bound to a unique SM2 public-private key pair digital identity certificate, which, combined with facial liveness detection (using a Silent-Face lightweight anti-spoofing model) and fingerprint recognition dual biometric verification, as well as dynamic SMS OTP verification codes, constitutes a three-factor authentication mechanism to prevent certificate theft and identity forgery.

[0050] Building upon static RBAC, this invention introduces a dynamic permission adjustment mechanism based on real-time risk scoring. The system calculates the operation authorization confidence level in real time using a weighted scoring model based on the following three types of inputs: Operational task hazard levels (three categories: low risk – button / reset operation; medium risk – circuit breaker opening and closing operation; high risk – wiring tightening and control module replacement); Real-time operating status of the unit (whether it is generating electricity, grid connection status, real-time wind speed level, nacelle vibration intensity); Operator real-time behavioral characteristics (response speed of the current operation, historical error rate, continuous online duration, and human-computer interaction anomaly detection results).

[0051] When the authorization confidence level is lower than the preset threshold, the system automatically raises the permission requirement by one level; under extreme operating conditions (such as when the unit is connected to the grid and the wind speed exceeds the rated value), the high-risk operation interface is directly locked to prohibit remote physical intervention and prevent grid disturbances or unit damage accidents caused by operational errors.

[0052] 2) Dual-person review system and electronic management of operation tickets For critical equipment operations (main control system reset, circuit breaker opening and closing, control module replacement), the system enforces a dual-person verification mechanism: after the operator submits an operation request, the system pushes real-time video footage and equipment status snapshots to the supervisor's terminal; the supervisor verifies the current status of the target equipment through real-time video (such as confirming that the circuit breaker is indeed in the disconnected position and that the terminal block number matches the operation ticket), and only after authorization through independent SM2 digital signature can the operation instructions be issued to the robot execution unit.

[0053] The system records the following data throughout the entire operation for post-operation traceability: operation initiation timestamp (accurate to milliseconds), operator digital identity ID, approver digital signature, operation task type, target equipment asset number, video clips of key steps (H.264 compressed, stored for at least 90 days), six-dimensional force / torque time series curve of the robotic arm end effector, and final operation result status. All records are hashed using SM3 and written to a blockchain-based tamper-proof evidence database, forming an unforgeable chain of operational evidence, meeting the compliance requirements of the Power Industry Safety Production Misoperation Prevention Management Standard (DL / T 1212).

[0054] Example 6: Human-Machine Shared Control Mechanism like Figure 6 As shown, this invention establishes a dynamic switching mechanism for three control modes based on shared autonomous control theory: Machine-dominated mode In standardized operation tasks (such as "pressing the reset button"), the system defaults to machine-led mode: the robotic arm executes autonomously according to the pre-programmed operation path, and the vehicle controller automatically completes the precision alignment and contact operation through visual servo closed loop and force control closed loop. The operator is only responsible for macro-level task start confirmation and execution process monitoring, and does not intervene in end-point control.

[0055] Human-machine shared control mode This invention introduces the concept of Shared Autonomy into remote operation processes, establishes a dynamic allocation mechanism for human and machine permissions, and achieves an organic integration of human cognitive advantages and machine computational execution advantages.

[0056] In standardized operation tasks (such as the "press the reset button" sequence), the system defaults to machine-led mode: the robotic arm executes autonomously according to the pre-programmed operation path, and the vehicle controller automatically completes the precision alignment and contact operation through visual servo closed loop and force control closed loop. The operator is only responsible for macro-level task start confirmation and execution process monitoring, and does not intervene in end-point control.

[0057] When encountering non-preset scenarios during operation (temporary obstruction of the field of view, target recognition confidence level below the threshold, abnormal fluctuations detected by the force sensor), the system automatically switches to human-machine shared control mode: transferring some degrees of freedom control of the end-effector's pose to the remote operator, who can then manually fine-tune the posture using a six-degree-of-freedom force feedback control handle, while retaining collision detection and contact force limiting protection on the machine side (force limits can be configured according to the operation type) to prevent damage to the equipment due to operational errors.

[0058] Automatic permission revocation mechanism: When the system detects prolonged operator inactivity (timeout threshold 30 s configurable) or persistent operational deviations (such as the robotic arm's end effector continuously approaching a non-target device under visual guidance), permission is automatically revoked back to the machine system, and a safe retraction action is executed. The robotic arm retracts to its initial storage position, and an alarm notification is triggered. This dynamic permission allocation framework dynamically determines human-machine permission boundaries based on real-time risk assessment of the environment and task, ensuring effective collaborative control in unexpected and complex situations.

[0059] Example 7: Intelligent Operation Control Based on Visual Servo and Trajectory Optimization like Figure 7 As shown, this invention constructs a triple closed-loop architecture of "visual positioning - trajectory planning - force control execution" in the field of robotic arm visual servoing and trajectory planning, realizing fully automated operation from macroscopic target recognition and positioning to microscopic compliant contact.

[0060] 1) Hybrid vision system and hand-eye calibration A color camera with a resolution of at least 1920×1080 (Eye-in-hand) is installed on the end effector flange of the robotic arm, while an Intel RealSense D455 depth camera (Eye-to-hand) is fixedly mounted on the chassis, forming a hand-eye hybrid vision system. The camera's intrinsic parameters are calibrated using Zhang Zhengyou's checkerboard calibration method, and hand-eye calibration is performed using the AX=XB equation solution method based on Random Sample Consensus (RANSAC). This yields a precise transformation matrix between the end effector coordinate system and the camera coordinate system, achieving millimeter-level pose transformation positioning accuracy.

[0061] 2) Target device instance segmentation and six-DOF pose estimation To address the compact layout and diverse types of equipment (buttons, indicator lights, knobs, circuit breakers, terminal blocks, control module slots, etc.) within wind turbine control cabinets, this invention deploys a lightweight target detection and instance segmentation algorithm based on Mask DINO (an instance segmentation framework incorporating a DINO converter backbone network). This algorithm outputs pixel-level segmentation masks and class confidence scores for each target device in a color image. The core reason for choosing Mask DINO lies in its ability to balance detection accuracy (approximately 2.5% higher AP than YOLOv8-seg) and inference speed (TensorRT accelerated backend frame rate ≥25 FPS) on the COCO benchmark, while also exhibiting superior detection performance for small-sized devices.

[0062] In the pose estimation stage, the depth point cloud is clipped using a mask, and local geometric descriptors are extracted from the point cloud using Fast Point Feature Histograms (FPFH) based on normal vector consistency constraints. These descriptors serve as the initial registration input. Subsequently, the Generalized Iterative Closest Point (G-ICP) algorithm is used to perform precise registration with the pre-constructed CAD template point cloud, outputting the target device's six-DOF pose (position coordinates). and attitude angle To address the issue of visual similarity or pose degradation in some devices, a multi-view matching mechanism is introduced: when the confidence level of single-view recognition is below a threshold, the camera at the end of the robotic arm is controlled to move slightly around the target to collect supplementary viewpoint point clouds, and the multi-view estimation results are weighted and fused to improve pose accuracy.

[0063] 3) Dynamic obstacle avoidance trajectory planning based on image visual servoing Within the confined space of an aircraft control cabinet, the movement path of a robotic arm may be interfered with by static obstacles such as wiring harnesses, pipes, and adjacent equipment. This invention employs an image-based visual servoing (IBVS) framework and integrates a real-time obstacle avoidance path planning module.

[0064] The specific implementation steps are as follows: In the traditional IBVS controller, a distance-adaptive gain scheduling mechanism is introduced. When the robotic arm approaches the target, the controller gain is increased to accelerate convergence. During the obstacle avoidance phase, the gain is reduced to improve trajectory smoothness. The two-stage gain is smoothly transitioned through the Sigmoid function to eliminate error abruptness during switching.

[0065] The system determines whether to enter the obstacle avoidance zone based on the real-time Euclidean distance between the robotic arm end effector and the obstacle (the safe distance threshold is configurable, with a default of 50 mm). Once in the obstacle avoidance zone, the system uses the Rapid Exploration Random Tree-Connect (RRT-Connect) algorithm to quickly plan a collision-free escape path in the configuration space. It then combines field of view (FoV) constraints and end effector kinematic constraints to construct a multi-objective optimization model and select the feasible path with the best operational continuity.

[0066] The obstacle avoidance path direction vector is projected onto the image feature space to generate dynamic guiding feature points that are linked with real-time feature points, so that the visual servoing task can maintain a closed loop during obstacle avoidance and visual tracking will not be interrupted due to obstacle avoidance.

[0067] The overall control system adopts a ROS 2+MoveIt 2 architecture: the decision and planning nodes maintain a finite state machine, managing the task state sequence of "orbit positioning—robot approach—end-end alignment—force control operation—safe withdrawal," calling parameterized trajectory templates and combining visual pose to generate end-end path points. Inverse kinematics and collision checks are performed through the MoveIt 2 interface, outputting safe joint trajectories. The real-time control node runs on the lower-level RT-Linux real-time system (PREEMPT-RT patch, 2 ms timing period), executing joint-space PID+feedforward control, synchronously reading six-dimensional force sensor data to run an impedance / admittance controller, and superimposing pose corrections onto the desired trajectory to achieve compliant control. The three-level asynchronous feedback loops—visual loop (25–30 Hz), planning loop (10–30 Hz), and force control loop (≥500 Hz)—are effectively integrated through a dual-channel mechanism of ROS 2 real-time topic and shared memory.

[0068] 4) Remote assisted operation and augmented reality operation guidance The remote operation terminal is deployed in the central control center, providing a real-time display interface for multiple video streams (RGB + depth + panorama), overlaid with an augmented reality (AR) operation guidance layer. The operation interface marks the outline of the target device and operation points with colored highlighted borders, and overlays virtual pressing areas and expected end-effector motion trajectories on the image for targets requiring buttons or knobs, assisting the remote operator in quickly locating the target. After the robotic arm approaches the target device, the system automatically retrieves the device's standardized operating procedure (SOP) and executes the subsequent action sequence in semi-automatic mode; the operator only needs to confirm the start with one button, or manually fine-tune the end-effector posture using a six-degree-of-freedom control handle in non-preset scenarios. In critical steps such as opening and closing circuit breakers and module replacement, a secondary confirmation pop-up is forcibly triggered, and the operation process is locked until the approver remotely digitally signs off.

[0069] This invention constructs an operation knowledge base and a continuous self-optimization mechanism for Standard Operating Procedures (SOPs): After each operation, the visual recognition results, six-dimensional force feedback curves, operation success / failure judgment flags, and control parameter sets (impedance stiffness values, admittance parameters, and trajectory planning constraints) are written into the operation knowledge database; the background periodically runs a hyperparameter search based on Bayesian optimization to automatically optimize the SOP control parameters for different equipment types and operation scenarios, and synchronizes them to all online robot nodes in an incremental update manner, thereby realizing cross-unit knowledge reuse and continuous improvement of operation success rate.

[0070] Example 8: Communication Architecture and Cloud Collaboration The system communication network is divided into three layers: vehicle-mounted local area network, on-site transmission network, and central control center wide area network.

[0071] The vehicle-mounted local area network uses EtherCAT real-time industrial bus to connect the joint servo drivers, six-dimensional force sensors, PZT actuators and vision cameras to achieve deterministic real-time data interaction with a period of 2 ms and end-to-end delay jitter of less than ±100μs. The on-site transmission network transmits robot telemetry data and multiple video streams to the edge computing node of the wind farm booster station via industrial fiber optic (main link, bandwidth ≥1 Gbps) or 5G SA standalone wireless link (backup link, air interface latency <10 ms). The edge node completes visual inference acceleration (deployment of TensorRT quantization model) and video transcoding (H.265, 720P@30fps, bandwidth about 2 Mbps / channel), reducing the backbone network transmission bandwidth requirements. The central control center's wide area network is interconnected with the edge nodes of each wind farm via dedicated power fiber optic lines, supporting centralized monitoring and concurrent remote operation of multiple units. The central control center deploys a remote operation terminal application platform (B / S architecture). After logging in via an industrial console or ruggedized tablet, operators can view multi-view video feeds from the robot in real time and send operation commands via a six-degree-of-freedom force feedback handle or mouse drag-and-drop. The total round-trip time (RTT) of control commands from the central control center to the robot's end effector is controlled within 200 ms, meeting the ergonomic requirements of remote real-time operation.

[0072] Example 9: Control Method Flow like Figure 7 As shown, the track-mounted remote-operated robot control method of the present invention includes the following steps: S1, Track positioning and vibration reduction start-up The track robot body 2 moves along the walking track 1 to the target working position. After arriving at the position, it activates the three-level series hybrid vibration damping system and waits for the end vibration to stabilize.

[0073] S2, Visual localization and pose estimation The multi-source visual perception and localization module is activated: the eye-to-hand depth camera acquires the global scene, and the Mask DINO algorithm detects the target device; the eye-in-hand color camera acquires local close-ups, performs instance segmentation and six-DOF pose estimation, and outputs the precise pose (X, Y, Z, R) of the target device. x ,R y ,R z ).

[0074] S3, Tool Selection and Trajectory Planning The corresponding end effector 6 is automatically selected according to the type of operation task. The vehicle controller plans the motion trajectory of the robotic arm based on MoveIt 2 and performs collision detection and inverse kinematics solution.

[0075] S4, Aerial Movement and Compassionate Contact During the aerial movement phase, the robotic arm uses pure position control to quickly approach the target; when the distance between the end effector and the target is less than 5mm, it switches to impedance control mode to achieve compliant contact.

[0076] S5, Precision Force Control Execution After entering the precision force tracking stage, the system switches to admittance control mode, achieving precise force control based on feedback from a six-dimensional force sensor. • Button press: Control the contact force in the Z direction to 5N, hold for 0.5s and then release; • Opening and closing of the circuit breaker: Control the thrust in the Z direction to 15N to complete the lever action; • Wiring tightening: Control torque 1.2N m stops once the set value is reached.

[0077] S6. Operation Status Identification and Rollback Based on the force feedback curve, the operation status is identified in real time, and the robot arm automatically performs a retraction action after the operation is completed, returning to a safe position.

[0078] S7. Security Control and Approval The entire operation is subject to authorization verification and supervision by the security management platform. Critical operations (such as switching on / off circuit breakers and module replacements) can only be executed after being reviewed and authorized by two supervisors with digital signatures.

[0079] S8, Knowledge Base Update and Self-Optimization After the operation is completed, the operation data (visual recognition results, force feedback curve, success / failure flags, control parameters) is written to the operation knowledge database. The knowledge base backend runs a Bayesian optimization hyperparameter search once a week to automatically adjust the SOP control parameters for different device types and operation scenarios, and synchronizes them to all online robot nodes in an incremental update manner.

[0080] This invention constructs an operational knowledge base and a continuous self-optimization mechanism for standardized operating procedures (SOPs), continuously improving system adaptability through closed-loop feedback of operational data. After each operation, the system writes the visual recognition results, force feedback curves, success / failure judgment flags, and operational parameter sets (impedance stiffness values, admittance parameters, trajectory planning constraints) into the operational knowledge database. The knowledge base backend periodically runs a hyperparameter search based on Bayesian optimization to automatically fine-tune the SOP parameters for different equipment types and operational scenarios, and synchronizes the parameter optimization results to all online robot nodes in an incremental update manner, achieving cross-unit knowledge reuse and continuous improvement in success rate.

[0081] Compared with the prior art, the present invention has the following beneficial effects: This invention enables remote physical operation capabilities. By integrating a six-degree-of-freedom robotic arm and replaceable end effectors onto a track-moving platform, and combining a force-position hybrid control algorithm and a vision servo system, it achieves remote contact operation of various devices such as buttons, reset switches, air switches, terminal blocks, and control modules within the wind turbine control cabinet, filling a gap in existing technology.

[0082] This invention significantly reduces the frequency of boarding and safety risks by using a track-mounted robot to replace personnel in climbing towers to perform routine operations and emergency responses. A single operation can save 4 to 24 hours of travel and climbing time, fundamentally reducing the operational risks for maintenance personnel in adverse weather and high-altitude environments.

[0083] This invention improves the timeliness of emergency fault handling. In emergency situations, maintenance personnel can remotely intervene in real time from the central control center, reducing fault response time from days / hours to minutes, effectively shortening unplanned downtime.

[0084] This invention features multi-stage vibration reduction to ensure operational accuracy. Through a three-stage series anti-vibration design of floating suspension, damping cross linkage structure, and active vibration suppression, the residual vibration amplitude of the robotic arm end in the machine vibration environment is controlled to the sub-millimeter level, providing a stable platform for precision operation.

[0085] This invention ensures safety and controllability throughout the entire process. Based on a hierarchical permission model with role and risk scoring, a dual-person review mechanism, dynamic allocation of shared control permissions, and electronic traceability records of the entire operation process, it makes the remote operation process comply with the safety production management standards of the power industry.

[0086] This invention features operational knowledge self-optimization. Through closed-loop feedback of operational data and knowledge base construction, it achieves continuous optimization of standard operating procedure parameters and cross-unit knowledge reuse, enabling the system's operational success rate and adaptability to continuously improve over time.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A track-mounted remote-operated robot system for wind turbine generators, characterized in that, include: The travel track is laid inside the wind turbine nacelle and on each platform of the tower. The track robot body moves along the walking track, and the track robot body includes a mobile chassis and an on-board real-time controller; A six-degree-of-freedom robotic arm is mounted on the mobile chassis, and the end of the six-degree-of-freedom robotic arm is equipped with an electro-pneumatic integrated standard quick-change interface. Multiple end effectors are connected to the six-degree-of-freedom robotic arm via the quick-change interface. Each end effector includes at least one of a single-finger elastic pressing head, a two-finger parallel gripper, a constant torque electric screwdriver module, and a four-finger module gripping fixture. The multi-source visual perception and positioning module includes a color camera mounted on the end of the robotic arm and a depth camera mounted on a mobile chassis. The six-dimensional force / torque sensing module at the end of the robotic arm is integrated between the end flange and the quick-change interface for real-time detection of end contact force and torque. The remote operation terminal, deployed in the central control center, is used to send operation commands and receive robot status and video data; The security management platform is used to realize user identity authentication, hierarchical access control, operation approval and full-process traceability management; The track robot body also includes a three-level series hybrid vibration damping system, which includes a floating suspension walking vibration reduction system, a multi-directional damping vibration reduction structure between the chassis and the mounting frame, and an active vibration suppression system based on acceleration feedback.

2. The track-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The three-stage series hybrid vibration damping system: The floating suspension travel vibration reduction system includes: each travel wheel set is connected to the chassis frame through a suspension floating arm, one end of the floating arm is hinged to the chassis via a pivot, and the other end is fixed to the chassis through a rubber-steel composite elastic damping component. The output end of the rotary drive motor and the travel wheel are connected by a double diaphragm flexible coupling to transmit the drive torque. The multi-directional damping vibration reduction structure includes three types of vibration reduction components arranged in parallel: cross-link elastic sliding structure, arc-shaped buffer rod sliding structure, and sliding rod-mounting cylinder mating structure; The active vibration suppression system includes: four piezoelectric stacked actuators symmetrically arranged between the mounting frame and the chassis, a triaxial MEMS accelerometer installed at the end flange of the robotic arm, and an adaptive active vibration control strategy based on the filter-x least mean square algorithm running on the vehicle-mounted real-time controller.

3. The track-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The on-board real-time controller runs an end-effector force-position control algorithm based on an impedance / admittance hybrid strategy, the algorithm including: The joint space impedance control layer models the robotic arm as a second-order dynamic system of mass-spring-damped system, based on the dynamic equations. Adjusting the desired inertia matrix M d Damping matrix B d and stiffness matrix K d This enables the robotic arm to exhibit the desired compliant characteristics in response to contact forces; The Cartesian space admittance control layer uses a six-dimensional force sensor to measure force F during the precision force tracking stage. ext As input, and with the end-effector pose correction ΔX as output, according to the transfer relationship Adjust the desired trajectory in real time; The mission state machine automatically switches control modes according to the operation phase: pure position control is used during the aerial motion phase; when it enters the range of 5mm from the target, it switches to impedance control mode; and after entering the precision force tracking phase, it switches to admittance control mode.

4. The track-mounted remote-operated robot system for wind turbine generators according to claim 3, characterized in that, The on-board real-time controller also runs an algorithm for real-time identification of operating states and adaptive parameter adjustment based on force feedback curves. Data from a six-dimensional force sensor was acquired at a frequency of 1kHz, and a force-time response curve was constructed. By detecting in real time whether the first derivative of dF / dt exceeds a preset threshold, the three-stage characteristics of the force-time response curve are identified: the non-contact stage, the elastic compression stage, and the bottoming-out step stage. Based on the recognition results, determine whether the button is triggered, the circuit breaker is in place, or the screw is tightened, and automatically trigger a retraction or stop command. An online adaptive control law based on Lyapunov stability theory is introduced to simultaneously estimate the contact environment stiffness k online. e and contact position x e This allows the actual contact force to gradually converge to the desired value.

5. The track-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The multi-source visual perception and localization module runs an intelligent operation control algorithm based on visual servoing and trajectory optimization: The camera intrinsic parameters were calibrated using Zhang Zhengyou's chessboard calibration method, and the hand-eye calibration was performed using the AX=XB equation solution method based on random sampling consistency. Deploy a lightweight object detection and instance segmentation algorithm based on Mask DINO, and output pixel-level segmentation masks and class confidence scores for each target device in color images; By using a mask to clip the depth point cloud, local geometric descriptors are extracted through a fast point feature histogram. The generalized iterative nearest point algorithm is used to accurately register the point cloud with the pre-constructed CAD template, and the six-DOF pose of the target device is output. It adopts an image-based visual servo control framework and integrates the RRT-Connect real-time obstacle avoidance path planning module.

6. The track-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The security management platform includes a hierarchical permission model based on roles and dynamic trust scoring: A role-based access control mechanism is adopted, which divides users into three role levels: inspector, operator, and supervisor / approver. Each user is bound to a unique SM2 public-private key pair digital identity certificate, which, combined with facial liveness detection, fingerprint recognition dual biometric verification, and dynamic SMS OTP verification code, constitutes a three-factor authentication mechanism. Based on the hazard level of the operation task, the real-time operating status of the unit, and the real-time behavioral characteristics of the operator, the confidence level of operation authorization is calculated in real time through a weighted scoring model. When the confidence level is lower than the threshold, the level of authorization requirement is automatically increased.

7. The rail-mounted remote-operated robot system for wind turbine generators according to claim 6, characterized in that, The security management platform also includes: The dual-verification mechanism requires that for operations such as resetting the main control system, opening and closing the circuit breaker, and replacing the control module, the operator must submit an application, and the supervisor must verify the equipment status through real-time video and authorize it with an SM2 digital signature before the operation instructions can be executed. The operation tickets are managed electronically, recording the entire process including the operation initiation timestamp, operator ID, approver digital signature, operation task type, target equipment number, video clips of key steps, six-dimensional force / torque time series curve, and operation result status. All records are written into a blockchain-based tamper-proof evidence database after being hashed and digested using SM3.

8. The rail-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The system also includes a dynamic human-machine permission allocation mechanism based on the shared autonomous control theory: Machine-led mode: In standardized operation tasks, the robotic arm executes autonomously according to the pre-programmed path, and the on-board controller completes precise alignment and contact operations through visual servo and force control closed loop. The operator is only responsible for macroscopic confirmation and process monitoring. Human-machine shared control mode: When the field of view is obstructed, the target recognition confidence is lower than the threshold, or the force sensor detects abnormal fluctuations, the system automatically transfers the control of the end part of the freedom to the remote operator, while the machine end simultaneously retains collision detection and contact force limiting protection; Automatic permission revocation mechanism: When an operator is detected to be unresponsive for an extended period or to have a persistent operational deviation, control is automatically reclaimed to the machine system and a safety rollback action is executed, while an alarm notification is triggered.

9. The rail-mounted remote-operated robot system for wind turbine generators according to claim 1, characterized in that, The system also includes an operational knowledge base and a self-optimization mechanism for standardized operating procedures. After each operation is completed, the visual recognition results, six-dimensional force feedback curve, operation success / failure judgment flag and control parameter group are written into the operation knowledge database; The knowledge base backend regularly runs a Bayesian-optimized hyperparameter search to automatically optimize the standardized operating procedure control parameters for different device types and operating scenarios. The parameter optimization results are synchronized to all online robot nodes using an incremental update method.

10. The control method for a track-mounted remote-operated robot system for wind turbine generators according to any one of claims 1-9, characterized in that, Includes the following steps: S1. The track robot moves along the walking track to the target work position and activates the three-level series hybrid vibration damping system; S2, the multi-source visual perception and localization module performs instance segmentation and six-degree-of-freedom pose estimation on the target device, and outputs the precise pose of the target device; S3. Automatically select the corresponding end effector according to the type of operation task, and the vehicle controller plans the movement trajectory of the robotic arm. S4. During the aerial movement phase of the robotic arm, pure position control is used. When it enters the range of 5mm from the target, it switches to impedance control mode to achieve compliant contact. S5. After entering the precision force tracking stage, switch to admittance control mode to achieve precise force control based on feedback from a six-dimensional force sensor. S6. Based on the force feedback curve, the operation status is identified in real time, and the retraction action is automatically executed after the operation is completed. S7. The entire operation is subject to the permission verification and supervision of the security management platform. Critical operations can only be executed after being reviewed and authorized by two people. S8. After the operation is completed, the data of this operation is written to the knowledge base, and the Bayesian optimization parameters are tuned regularly in the background.

Images