A control method of a rotary vane de-icing robot and the robot

Abstract

This invention relates to wind turbine de-icing, specifically to a control method and robot for a rotating blade de-icing robot. The control method includes: acquiring the centrifugal force vector, gravity vector, and airflow load vector acting on the robot, and then superimposing them to synthesize a comprehensive load vector; calculating the slippage trend and overturning moment center of the robot body relative to the working surface based on the spatial orientation of the comprehensive load vector; and performing dynamic adhesion compensation based on the slippage trend and overturning moment center: if a slippage trend is determined, the adsorption force of the adhesion units located in the region opposite to the slippage trend is directionally increased to generate a constraint torque that dynamically balances with the external dynamic load, thereby balancing the robot's adhesion on the working surface. This solution overcomes the limitation of traditional de-icing robots, which can only sense static gravity, by superimposing centrifugal force, gravity, and airflow load vectors, and dynamically adjusts the robot body's adhesion state by enhancing the adsorption force to generate a balancing torque.

Description

Technical Field

[0001] This invention relates to wind turbine de-icing, specifically to a control method and robot for a rotating blade de-icing robot. Background Technology

[0002] Current anti-icing and de-icing technologies are mainly divided into passive anti-icing and active de-icing. Passive anti-icing mainly relies on coating the blade surface with hydrophobic / superhydrophobic coatings, but its effectiveness is limited and its lifespan is short in extreme environments. Active de-icing mainly involves pre-embedding electrothermal films or hot air channels during the blade manufacturing stage. However, it is difficult to retrofit built-in de-icing systems into existing wind turbines, and long-term thermal stress may cause fatigue damage to the composite material structure of the blades.

[0003] Due to the aforementioned limitations, utilizing external operating platforms (such as robots or drones) for blade de-icing has become a hot research topic in the industry. However, existing external de-icing methods face high operation and maintenance costs: manual climbing or helicopter spraying is greatly limited by weather conditions and is highly dangerous; existing wall-climbing robots (which use negative pressure suction cups to climb on the blade surface) generally have a shutdown dependency, meaning that the entire de-icing operation cycle must be carried out with the wind turbine completely shut down and the hub locked. For ultra-large wind turbine units, prolonged shutdown and locking not only cause huge economic losses in power generation.

[0004] Therefore, the necessity of this design lies in developing a de-icing robot capable of performing tasks under conditions of online operation or low-speed rotation of the wind turbine, in order to minimize downtime and improve operational flexibility. However, traditional de-icing robots only need to consider the balance of gravity and constant adsorption force under static conditions. But when the robot is deployed on the blades in operation, its physical environment will change fundamentally. The robot will be in a dynamic, directional, composite force field consisting of gravity that periodically flips with phase, huge centrifugal force that fluctuates with rotational speed and position, and the intense turbulent airflow load generated during rotation.

[0005] Under the aforementioned harsh working conditions, traditional static adsorption control schemes for climbing robots cannot detect the instantaneous drift of the center of gravity, easily leading to insufficient adhesion or torque imbalance at specific angles of blade rotation. This can cause the robot to slip, overturn, or even detach from the rotating blade and be thrown off. Therefore, the present invention aims to solve the problem of designing a de-icing robot that can operate in complex rotating force fields and move steadily on blades. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a control method and a robot for a rotating blade de-icing robot, aiming to solve the problems that existing wall-climbing robots cannot maintain adhesion stability under wind turbine operation or low-speed rotation conditions, and are easily affected by gravity periodic overturning, centrifugal force fluctuations and strong turbulent load superposition interference, which can lead to slippage or overturning and falling off.

[0007] This invention provides a control method for a rotating blade de-icing robot, used to control the de-icing robot to walk on a rotating blade via an attachment unit formed by a single vacuum suction cup foot or a group of multiple vacuum suction cup feet. The control method includes:

[0008] The centrifugal force vector, gravity vector, and airflow load vector acting on the robot are obtained and superimposed to form a comprehensive load vector. Based on the spatial orientation of the comprehensive load vector, the sliding tendency and overturning moment center of the robot body relative to the working surface are calculated. Based on the sliding tendency and overturning moment center, dynamic adhesion compensation is performed: if a sliding tendency is determined, the adsorption force of the adhesion unit located in the region opposite to the sliding tendency is increased in a directional manner to generate a constraint torque that is dynamically balanced with the external dynamic load, so as to make the robot's adhesion on the working surface tend to be balanced.

[0009] This solution constructs a dynamic mechanical model mapping the working conditions of rotating blades by acquiring and superimposing centrifugal force vectors, gravity vectors, and airflow load vectors. This overcomes the technical deficiency of traditional de-icing robots, which can only sense static gravity and cannot cope with dynamic rotating loads. This multi-dimensional load perception and vector synthesis enables the control system to capture in real time the gravity direction reversal caused by phase angle changes, the transient change in centrifugal force caused by speed fluctuations, and the random airflow disturbances generated by complex wind fields. By directionally enhancing the adsorption force of the attachment units located in the region opposite to the slippage trend, a constraint torque is generated to balance the external dynamic load, thereby dynamically adjusting the attachment state of the robot body on the working surface. This dynamic compensation for force fluctuations helps the robot maintain operational stability when facing rotational phase changes, speed fluctuations, and environmental wind field interference, reduces the risk of slippage or detachment due to insufficient adhesion, and provides a guarantee for continuous climbing and de-icing under changing composite force fields.

[0010] In some embodiments, the underside of the robot body facing the work surface is configured as an omnidirectional convergent surface, which smoothly bulges inward from the peripheral edge of the body; the back side of the robot body away from the work surface is configured as a stratospheric surface.

[0011] During the step displacement of the drive attachment unit for climbing operations, a transient smoothing step based on fuselage aerodynamic clearance is also included, including acquiring the real-time clearance height between the fuselage bottom surface and the working surface, as well as the instantaneous rate of change of the clearance height.

[0012] Determine the execution phase of the current step displacement:

[0013] If it is determined that the attachment unit is in the lifting phase of detaching from the working surface, and the instantaneous increment of the real-time gap height is greater than the preset aerodynamic instability threshold, then the transient outward tilting moment of the fuselage is calculated according to the above-mentioned comprehensive load vector, and a collaborative compensation command is output to increase the adsorption force of the attachment unit on the fuselage side; in this way, the outer edge of the fuselage is lowered, limiting the excessive expansion of the Venturi effect flow channel.

[0014] If it is determined that the attachment unit is in the landing phase when it touches the working surface, and the negative instantaneous change rate of the gap height exceeds the preset stagnation limit, then an attitude depressurization command is output to drive the active joint of the attachment unit located at the reverse end of the fuselage to perform a preset slight lift, so as to build an exhaust channel to release the positive pressure of the bottom air cushion.

[0015] The robot acquires the real-time attachment pressure feedback from the attachment unit that performs the step displacement. If it determines that the pressure has reached a stable threshold, a reset command is output to restore the initial gap between the robot body and the working surface. The aforementioned execution phase refers to the time slice and posture characteristics of the robot attachment unit (mechanical leg) during the cycle of completing one step movement. Specifically, it is divided into the lifting phase (the process of the suction cup detaching from the surface and rising), the swinging phase (the process of moving forward in the air), and the landing phase (the process of the suction cup touching the surface and pressing down).

[0016] In the harsh operating environment of wind turbine blades, robots not only have to resist their own gravity and centrifugal force, but also constantly face strong turbulent wind fields. To address this, this embodiment employs a specific aerodynamic design in its fuselage structure: the back of the fuselage facing away from the working surface is designed as a horizontal surface to reduce external wind resistance; the belly of the fuselage facing the working surface is designed as an omnidirectional converging surface that smoothly bulges inward from the peripheral edges (i.e., a structure resembling an inverted shallow bowl or spherical crown). When natural wind blows from the side of the fuselage into the space between the belly and the blade, according to the continuity equation in fluid mechanics, the airflow is forced to accelerate within this converging channel, thereby generating a passive aerodynamic negative pressure (i.e., the Venturi effect) pointing towards the blade surface according to Bernoulli's principle, providing the robot with additional pressure towards the blade surface. However, when the robot performs alternating step displacements, a relatively dangerous transient aerodynamic instability trap can occur. Specifically, when the attachment unit (mechanical leg) momentarily lifts off the surface, the fuselage, due to the loss of support on one side, is prone to a slight outward tilt, resulting in a slight increase in the gap between the belly and the blade. Once the gap widens, the original Venturi acceleration channel is disrupted, and the aerodynamic negative pressure drops sharply. If a strong gust of wind strikes at this moment, the robot is highly susceptible to losing downforce and being instantly overturned. On the other hand, at the moment the attachment unit completes its step and heavily slams onto the blade surface, the fuselage is rapidly pulled back, and the abdominal gap is drastically compressed. At this time, if the air in the abdominal space has not had time to be expelled, it will be forcibly compressed, forming an instantaneous upward thrust (forming an air cushion with upward thrust). This upward positive pressure will strongly counteract the downward pressing action of the suction cup, causing the suction cup to consume more energy and time to complete the adhesion and sealing. To eliminate the transient instability caused by this lift and fall, this system introduces a transient smoothing control step based on the fuselage aerodynamic gap. First, the real-time gap height and its instantaneous rate of change are obtained in real time by ranging sensors (such as high-frequency laser sensors) distributed along the edge of the fuselage. The execution phase of the step is monitored, and targeted interventions are taken:

[0017] To address the outward tilt risk during the lift-off phase: When the lift-off phase is determined, monitor the instantaneous increment of the real-time gap height. If this increment exceeds the preset aerodynamic instability threshold (i.e., the critical height difference that would cause the loss of negative pressure on the fuselage beyond the safe limit, calculated through wind tunnel experiments or simulations), it indicates that the fuselage is tilting. At this point, the system immediately calls the integrated load vector to calculate the transient outward tilt moment that could potentially overturn the fuselage. Based on this, a collaborative compensation command is output to instantly increase the adsorption power of the fuselage's remaining side attachment units (i.e., the suction cup foot array on the side that is still firmly attached to the blades and has not participated in the stepping). By enhancing the pull on this side, the outer edge of the fuselage is pressed down again, thereby forcibly maintaining the narrow flow channel space required for the Venturi effect and preventing the negative pressure flow field from collapsing.

[0018] Regarding the risks during the landing phase, when the descent phase is determined, the negative instantaneous rate of change of the gap height (i.e., the speed at which the fuselage slams down and compresses the air) is monitored. If this rate exceeds a preset stagnation limit (i.e., the critical rate at which the air exhaust speed cannot keep up with the fuselage's downward pressure, leading to fluid stagnation and localized high pressure), it indicates that positive pressure is being generated at the bottom of the fuselage, hindering adsorption. At this point, the system outputs an attitude depressurization command, driving the active joints of the attachment units (such as the mechanical legs at the robot's tail) located at the reverse end of the fuselage's movement to lift slightly upwards by a few millimeters. This tiny and brief tiptoeing motion cleverly creates an open exhaust channel at the tail of the fuselage, guiding the rapidly compressed air to escape smoothly and instantly dismantling the air cushion drag at the bottom of the fuselage. Here, the stagnation limit refers to the critical value of the descent speed at which the bottom of the fuselage presses against the blade surface during landing. If the descent is too fast and exceeds this limit, the air at the bottom will not have enough time to disperse, resulting in fluid stagnation and thus creating a positive air cushion that pushes the fuselage upwards. Once the attachment unit performing the step reports that its actual attachment pressure has reached a stable sealing threshold, it proves that the landing is absolutely safe. The system then outputs a reset command, causing the raised tail-end mechanical leg to fall back, restoring the initial optimal aerodynamic clearance so that the robot body is completely parallel to the working surface, preparing for the next step. Through this transient smoothing control, the aerodynamic advantages of the shape can be deeply coupled with the robot's gait control. At the moment of leg lifting, the active pressurization of the left-side suction cup locks the aerodynamic clearance in a feedforward manner, avoiding the risk of being overturned by the wind due to the weakening of the Venturi effect; at the moment of leg landing, the attitude linkage of the tail-side mechanical leg actively builds a pressure relief channel, eliminating the suction resistance caused by the positive pressure of the air cushion. This control strategy, which resolves aerodynamic transient conflicts at the root of mechanical motion, not only significantly shortens the time required for suction cup pressurization and reduces energy consumption, but also gives the robot better continuous climbing stability under extreme conditions of strong turbulent wind fields and high centrifugal forces.

[0019] In another aspect, the present invention provides a rotary blade de-icing robot for executing the above-mentioned control method. The robot includes a body, a negative pressure attachment matrix, and a central processing unit. The back of the body away from the working surface is a stratospheric surface, and its belly facing the working surface is an omnidirectional convergent surface. The omnidirectional convergent surface smoothly bulges inward from the peripheral edge of the body, and the spatial extreme point of the bulge is located at the geometric center of the belly of the body, so that the body, under the airflow intrusion of any horizontal phase, jointly defines a negative pressure channel with the working surface (to convert the multi-directional turbulent wind field into aerodynamic downforce pointing towards the working surface). The negative pressure attachment matrix is ​​symmetrically mounted on the two wing sides of the body. The negative pressure attachment matrix includes multiple walking units, each of which includes an active joint and a vacuum suction cup disposed at the end of the active joint. The central processing unit is built into the body or disposed outside the body and is communicatively connected to each walking unit and a multi-dimensional sensor network. It is used to acquire environmental data to synthesize a comprehensive load vector and calculate the slip trend and overturning moment center based on the spatial orientation of the comprehensive load vector, and to control the negative pressure attachment matrix to perform dynamic attachment compensation. The robot employs an omnidirectional convergent configuration, converting multidirectional wind fields into aerodynamic downforce through negative pressure channels, thereby enhancing grip with the airflow. Combined with a symmetrically distributed negative pressure adhesion matrix (including active joints and vacuum suction cups), the robot can support and move on complex surfaces. The central processing unit calculates loads, slippage trends, and overturning moments in real time, dynamically adjusting the adhesion power. This aerodynamic layout, combined with redundant structures, improves adhesion reliability under centrifugal force and changing wind fields, reducing the risk of instability. Attached Figure Description

[0020] Figure 1 This is a side view diagram illustrating the de-icing robot structure under lateral airflow in the implementation method.

[0021] Figure 2 This is a schematic diagram illustrating the structure of the de-icing robot in the implementation method;

[0022] Figure 3 This is a schematic diagram illustrating the operation of the de-icing robot with a supply base station in the embodiment.

[0023] Figure 4 for Figure 3 A magnified view of a section at point A in the middle;

[0024] Figure 5 This is a schematic diagram illustrating the architecture of the de-icing robot in the base station-free solution of the implementation method;

[0025] Figure 6 This is a schematic diagram illustrating the architecture of a de-icing robot with a base station solution in the implementation method.

[0026] 1-Fuse; 110-Omnidirectional converging surface; 2-Walking unit; 3-Ultrasonic de-icer; 4-Electrically heated de-icing probe; 5-Pneumatic injection assembly; 510-Main nozzle; 520-Secondary nozzle; 6-Laser rangefinder sensor; 7-Vision sensor; 8-Umbilical cable; 9-Supply base station. Detailed Implementation

[0027] Example 1: A control method for a rotating blade de-icing robot, used to control the de-icing robot to walk on a rotating blade via an attachment unit formed by a single vacuum suction cup foot or multiple vacuum suction cup foot groups. The control method includes:

[0028] The centrifugal force vector, gravity vector, and airflow load vector acting on the robot are obtained and superimposed to form a composite load vector. Based on the spatial orientation of the composite load vector, the slippage trend and overturning moment center of the robot body 1 relative to the working surface are calculated. Based on the slippage trend and overturning moment center, dynamic attachment compensation is performed.

[0029] If a slippage tendency is determined, the adsorption force of the attachment unit located in the region opposite to the slippage tendency is increased in a directional manner to generate a constraint torque that is dynamically balanced with the external dynamic load, so as to make the robot's attachment on the working surface tend to be balanced.

[0030] In a specific embodiment, this solution integrates a sensor network and an embedded control system to calculate the force model of the robot on the rotating blade in real time. The calculation model can be performed according to the following scheme:

[0031] Combination Figure 3 and Figure 4 First, the robot uses its built-in six-axis inertial measurement unit (IMU), real-time communication with the wind turbine's main controller to obtain rotational speed parameters, and pressure sensor arrays on the chassis to calculate the physical vectors acting on the robot in real time, thereby completing the acquisition and synthesis of dynamic load vectors.

[0032] (1) Solve for the centrifugal force vector c

[0033] The robot obtains its radial coordinate r on the blade and its mass m through a hub distance encoder or heading calculation. Combined with the current angular velocity ω of the wind turbine, the magnitude of the centrifugal force (its vector direction is from the center of the hub to the tip of the blade) is calculated.

[0034]

[0035] (2) Gravity vector Solution:

[0036] The IMU monitors in real time the Euler angles (pitch θ, roll ϕ, yaw ψ) of the fuselage 1 coordinate system relative to the absolute Earth coordinate system (G-frame), and projects the gravitational constant g onto the fuselage 1 coordinate system:

[0037]

[0038] (3) Airflow load vector w Solution:

[0039] The composite field v of the induced wind speed generated by the blade rotation and the ambient wind speed is obtained by using air pressure sensors around the chassis. r Combined with the aerodynamic drag coefficient C of fuselage 1 d Air mass density ρ and windward area A in the robot's operating environment:

[0040]

[0041] (4) Composite load vector The composite load is synthesized through a vector superposition algorithm:

[0042]

[0043] Then, the calculation steps for slip tendency and overturning moment center are performed. The following steps can be followed during implementation:

[0044] The central processing unit, based on the spatial relationship between the fuselage chassis 1 and the blade working surfaces, will... Decomposed into positive pressure F n (Perpendicular to the blade) and shear force F τ (Parallel to the blade).

[0045] (1) Determination of slippage trend:

[0046] Calculate the current static friction constraint:

[0047]

[0048] Where μ is the coefficient of friction, P i Let F be the negative pressure value of the i-th suction cup, and S be the area of ​​the suction cup. τ > If so, it is determined that there is a slippage tendency, and the slippage direction is F. τ The positive aspect.

[0049] (2) Center of overturning moment O t Positioning:

[0050] Based on the geometric coordinates (x) of each suction cup foot i ,y i), calculate the resultant moment generated by the combined load on the support edge. The overturning moment center O. t Defined as composite load vector The point where the line of action intersects the chassis plane. If this point exceeds the geometric convex hull formed by the current support feet, a risk of rollover is deemed present.

[0051] Then, once a risk of slippage or overturning is determined, the controller executes a compensation strategy based on proportional-integral-derivative or fuzzy control.

[0052] (1) Targeted enhancement logic:

[0053] The controller identifies the suction cup assembly S located in the region opposite to the slippage trend (i.e., the side where the force is counteracted). c .

[0054] For example, if the combined force is directed towards the trailing edge (rear) of the blade, it will enhance the negative pressure of the 12 suction cup feet (out of a total of 24) near the leading edge (front).

[0055] (2) Adsorption power adjustment algorithm:

[0056] Let the target compensation torque be M. t Then the negative pressure increment ΔP of each controlled suction cup is... i The calculation is as follows:

[0057]

[0058] Where e(t) is the deviation between the current resultant torque and the safe torque threshold;

[0059] , , These are all PID (Proportional-Integral-Derivative) gain parameters, referring to the proportional, integral, and derivative coefficients, respectively, as in this embodiment. This determines the strength of the enhanced adsorption force when a slippage tendency is detected, and The system predicts and mitigates instantaneous impacts in advance based on the rate of change of load. In actual operation, each of the 24 suction cups is equipped with an independent electronic proportional control valve. When the centrifugal force increases with the blade rotation speed, the central processing unit increases the reference negative pressure of all suction cups according to the above algorithm; when the robot is in the horizontal phase of the blade, causing gravity to generate lateral sliding force, the processing unit only increases the duty cycle of the vacuum pump of the array foot located in the opposite direction of the lateral sliding, generating a resisting torque through asymmetrical pressure distribution, so that the robot always keeps in close contact with the blade surface.

[0060] Based on the constructed force model, this method can be implemented as follows:

[0061] S1. Real-time multi-source load data acquisition is performed, including the control system entering a high-frequency data sampling state after the robot is deployed on the rotating blade and begins operation. The system obtains the real-time angular velocity ω of the wind turbine through a communication link established with the wind turbine's main control unit; simultaneously, it uses a built-in positioning module (such as an encoder or laser rangefinder 6) to determine the radial radius r of the robot's center distance from the hub. At the same time, the built-in inertial measurement unit (IMU) obtains the three-axis attitude angles (pitch angle θ and roll angle ϕ) of the fuselage 1 relative to the horizontal plane. Furthermore, wind speed sensing modules distributed around the fuselage 1 measure the ambient wind speed in real time and calculate the relative wind speed v acting on the surface of the fuselage 1 by combining this with the blade rotation speed. r .

[0062] S2. The central processing unit, which performs spatial synthesis of the integrated load vector, inputs the collected data into a preset dynamic algorithm model. The system first calculates the centrifugal force vector pointing towards the blade tip. The gravity vector that is always vertically downward And the airflow load vector distributed tangentially along the windward surface of fuselage 1 Subsequently, the system performs vector superposition operations in the three-dimensional coordinate system of fuselage 1 to synthesize the comprehensive load vector. This resultant force vector can reflect in real time the overall mechanical trend of the robot under the combined action of the current rotation phase, current rotation speed and current wind force. (This completes the mechanical modeling; once completed, subsequent processes only require inputting variables.)

[0063] S3. Determine slippage trend and overturning risk; the system will synthesize the comprehensive load vector. It is decomposed into a normal pressure component pressing against the blade surface and a tangential shear force component distributed along the blade surface.

[0064] (1) Perform slip determination, that is, the system compares the tangential shear force with the maximum static friction force that can be provided by all units currently in the adsorption state (such as vacuum suction cup feet) in real time. Once the shear force approaches or exceeds the static friction threshold, the system immediately determines that there is a slippage tendency and locks the specific location of the slippage.

[0065] (2) Perform overturning determination, i.e., calculate the resultant force vector of the system. The point where the line of action of the robot intersects with the bottom of the fuselage 1 is the center of the overturning moment. If this center deviates from the geometric center of the fuselage 1 and approaches the support boundary of the edge of the fuselage 1, it is determined that the robot is at risk of tilting or overturning to that side.

[0066] S4. Execute directional attachment compensation; once a slippage or overturning tendency is detected, the control system does not blindly increase power globally, but instead executes a directional compensation strategy: based on the reverse spatial direction of the slippage tendency or overturning moment, the system automatically identifies a specific attachment unit group located in the force-counteracting area among the 24 suction cup feet array. The control system sends a pressurization command to the actuator (such as a vacuum proportional control valve) corresponding to that attachment unit group, and rapidly increases the negative pressure value P of the suction cups in that area through a PID control algorithm. i The enhanced negative pressure generates a stronger constraint force pointing towards the blade surface on the side of the body 1 where the forces are unbalanced, thus creating a constraint torque that is dynamically balanced with the external dynamic load. For example, when centrifugal force attempts to fling the robot toward the blade tip, the system directionally enhances the suction cup power near the hub side, increasing the gripping force on that side to counteract the flinging tendency.

[0067] S5. Maintain a dynamic closed loop for adhesion balance; during compensation, the system continuously monitors the actual pressure values ​​reported by each adhesion unit. If determined... Returning to the preset stable envelope range indicates that the robot has reached a new attachment equilibrium state, and the system will maintain the current compensation power. If the load fluctuates further, the system will repeat the above steps to ensure that the robot's adsorption power is always allocated and fed forward according to the evolution of the external load throughout the entire cycle of the blade's 360-degree rotation.

[0068] Based on the above scheme, specifically, while maintaining attachment balance, the robot is guided to perform path cleaning and cooperative stepping cycles. If exposed substrate is detected on the travel path and the current support group's attachment pressure change value is less than a preset threshold (pressure stable), the non-support group in the attachment unit is driven to perform step displacement for climbing operations under a changing composite force field. In specific implementation, the path cleaning and cooperative stepping cycle of this scheme is a control process that deeply couples de-icing operations with displacement logic. Before executing the displacement command, the robot first performs path de-icing, cleaning, and sweeping of the area within the preset step width using a composite de-icing device (e.g., integrating an ultrasonic transducer, an electrothermal probe, and a high-pressure nozzle) installed at the front end of the travel direction of the robot body 1. For example, the robot first uses ultrasonic high-frequency oscillation to break the freezing interface between the ice layer and the blade substrate, then the electrothermal probe cuts into the ice layer to generate thermal stress cracks, and finally the compressed air nozzle blows away the loose ice. A vision sensor 7 (camera) or an infrared spectral detection module is installed at the front end of the chassis of the robot body 1 for real-time scanning of the cleaned area. The system identifies the exposed state of the blade composite material (substrate) through image feature operators or reflectivity differences. For identifying ice layers or ice surfaces, a combination of a long-wave infrared thermal imaging camera (LWIR) and a high-definition visible light camera can be used. If the pixel grayscale value or spectral characteristics within the target area match the substrate template, and the flatness change rate is within a preset threshold, the substrate is determined to be exposed. Before driving the non-support group (i.e., the suction cup feet about to step) off the ground, the control system must ensure that the current support group in the attached state has sufficient attachment redundancy. Then, pressure stability calculations are performed; the system reads the pressure sensor data P of the N suction cup feet in the current support group. i The criterion for determining whether the current support level is stable is:

[0069]

[0070] in, This is a safety reserve factor (typically taken as 1.5~2.0). Only when this inequality holds and the pressure fluctuation rate... The system only issues a stepping command when the pressure approaches zero. The robot's 24 suction cup feet on both sides are logically divided into support and swing groups. The robot body 1 controls the active joints of the suction cup feet via a multi-axis servo driver. Once the system confirms the path has been cleaned and the support group's adhesion pressure is stable, the electromagnetic pressure relief valve activates, causing the suction cups of the swing group to detach from the blades. The servo motor drives the swing group forward along the travel direction, taking a preset step (e.g., 100mm-300mm) and landing on the cleaned exposed substrate area. After the swing group lands, the corresponding vacuum generator activates, quickly establishing negative pressure. When the feedback pressure P of the swing group... nOnce the preset working value is reached, the original support group transforms into a swing group, and the original swing group transforms into a support group, completing a full cooperative stepping cycle, which is repeated sequentially. During the climbing process, the robot is subjected to periodically changing gravity and centrifugal force, and the system dynamically adjusts the phase of the step.

[0071] If the centrifugal force vector at this time In the event of a sudden increase (such as a wind turbine speed increase), the system will automatically reduce the centrifugal height (i.e., ground clearance) of the oscillating assembly and shorten the step cycle to reduce the support time of a single assembly. If the lateral wind load... This causes the robot body 1 to yaw. The servo mechanism asymmetrically adjusts the landing coordinates of the suction cup feet on the left and right sides, correcting the trajectory through the difference in physical step distance. Based on the above loop logic, this solution enables the robot to achieve stable displacement propulsion on the rotating blades rhythmically, ensuring that each fulcrum lands on a reliable substrate, thus reducing the risk of slippage caused by blindly walking on ice.

[0072] Example 2: In this example, the belly of the robot body 1 facing the work surface is set as an omnidirectional convergent surface 110, which smoothly bulges inward from the periphery of the body 1; the back of the robot body 1 away from the work surface is set as a stratospheric surface; during the climbing operation, the robot body 1 also includes a transient smoothing step based on the aerodynamic clearance of the body 1, including obtaining the real-time clearance height between the bottom surface of the body 1 and the work surface, as well as the instantaneous rate of change of the clearance height.

[0073] Determine the execution phase of the current step displacement:

[0074] If it is determined that the attachment unit is in the lifting phase of detaching from the working surface, and the instantaneous increment of the real-time gap height is greater than the preset aerodynamic instability threshold, then the transient outward tilting moment of fuselage 1 is calculated based on the comprehensive load vector, and a collaborative compensation command is output to increase the adsorption force of the attachment unit on the side of fuselage 1; in this way, the outer edge of fuselage 1 is lowered, limiting the excessive expansion of the Venturi effect flow channel.

[0075] If it is determined that the attachment unit is in the landing phase of touching the working surface, and the negative instantaneous change rate of the gap height exceeds the preset stagnation limit, then an attitude depressurization command is output to drive the active joint of the attachment unit located at the reverse end of the fuselage 1 to perform a preset slight lift; in this way, an exhaust channel is constructed to release the positive pressure of the bottom air cushion.

[0076] The system acquires the real-time adhesion pressure fed back by the adhesion unit that performs the step displacement. If it determines that the pressure has reached the stability threshold, it outputs a reset command to restore the initial gap between fuselage 1 and the working surface. Here, the aerodynamic instability threshold refers to the critical value of the instantaneous increase in the bottom surface gap generated when fuselage 1 raises its legs. Once this threshold is exceeded, the Venturi convergence channel in the belly of fuselage 1 will be destroyed, resulting in a cliff-like decrease in passive aerodynamic negative pressure.

[0077] During implementation, high-frequency industrial-grade laser macro sensors are concealed at the four corner edges of the robot's chassis 1. Throughout the robot's stride-climbing cycle, these four sensors provide real-time feedback to the central processing unit (CPU) on the height of the four corners of chassis 1 from the blade surface, thereby constructing a model of the transient pitch and roll angles of chassis 1 in space. Simultaneously, the CPU performs real-time differential calculations on the height data to obtain the instantaneous rate of change (velocity) of the clearance height. If one side of chassis 1 is raised, the height increases, and the rate of change is positive (increment); if chassis 1 is lowered, the height decreases, and the rate of change is negative (compression). Thus, combined with the aforementioned comprehensive load vector, the system captures minute aerodynamic attitude distortions caused by gait switching. For example, consider the robot's right mechanical leg needing to lift and stride forward:

[0078] When the right mechanical leg begins to exert force and detach from the blade surface (lift phase), the right edge of fuselage 1 will often naturally tilt upward by a few millimeters due to the instantaneous disappearance of the right support force and the offset of the support fulcrum of the left-hand leg.

[0079] When the laser sensor detects that the gap increment on the right edge instantly exceeds the preset aerodynamic instability threshold (e.g., set to 3mm), it means that the Venturi acceleration channel at the bottom of fuselage 1 is leaking rapidly due to the widening space, and the aerodynamic grip is about to plummet. The central processing unit responds within 5ms, extracting the transient outward tilting torque synthesized by natural wind and centrifugal force. Instead of taking deceleration actions, the system directly issues a compensation command to the left-side attachment unit (the left-side suction cup array adsorbed on the blades at this time). The electronic proportional valve on the left vacuum branch instantly opens fully, raising the negative pressure value of the left suction cup from the normal -60kPa to -85kPa. This generates a huge downward pulling force on the left, forcibly flattening and lowering the raised edge of fuselage 1 on the right side, thus forcibly locking the gap height of the abdomen within an extremely narrow optimal flow channel range (e.g., 10-15mm). This maintains the passive aerodynamic downforce without relying on additional actions, resisting the risk of being overturned by the airflow.

[0080] When the right mechanical leg completes its forward swing and prepares to step down for landing (landing phase):

[0081] The laser sensor detected that the overall clearance height of fuselage 1 is decreasing rapidly. If the calculated negative instantaneous rate of change exceeds the stagnation limit (e.g., descent speed > 0.5 m / s), it indicates that the abdomen is pressing down rapidly, and the internal air cannot escape in time due to the obstruction of the outriggers around the edges, forming an instantaneous positive pressure air cushion.

[0082] To prevent the suction cup from being blocked by the positive pressure and unable to adhere to the blade, the central processing unit immediately issues an attitude depressurization command to the attachment unit (i.e., the trailing mechanical leg at the tail of the fuselage 1) located at the reverse end of the fuselage 1. After receiving the command, the active servo joint of the tail mechanical leg performs a slight retraction, causing the tail fuselage 1 to rise slightly by a small compensation height (e.g., 4-12mm).

[0083] This brief upward movement physically creates a wide exhaust vent at the rear of the robot body 1. The rapidly compressed air instantly bursts out through this least-resistance rear passage, instantly dissolving the positive pressure of the air cushion at the bottom of the robot body 1. Without upward resistance, the right suction cup foot, which is descending, lands smoothly on the blade surface, reducing the time required for vacuum sealing. Once the pressure sensor inside the right suction cup foot confirms that the negative pressure has reached a stable sealing threshold (e.g., −50 kPa), the danger period for this step has passed. The central processing unit then outputs a reset command, and the active joints of the rear robotic leg extend and reset. The robot body 1 returns to its initial narrow gap state, aligned front to back and parallel to the blade surface, and the entire robot enters the next climbing cycle under stable pneumatic protection.

[0084] Example 3, based on the above examples, further includes calculating the predicted load window within a preset displacement step based on the evolution trend of the instantaneous amplitude and time rate of change of the comprehensive load vector in the stability state space. This includes: if the predicted load window is within a preset stable envelope interval, then the step displacement and de-icing operation are synchronized in phase to maintain continuous robot operation; if the predicted load window touches the instability boundary, then while increasing the adsorption power, the operating parameters of the de-icing device are adjusted to smooth the dynamic load fluctuations of the robot, and the triggering of the step displacement is delayed until the predicted load window returns to the stable envelope interval. The aforementioned stability state space is a multi-dimensional mathematical coordinate system constructed internally by the central processing unit, whose coordinate axes represent the forces (gravity, centrifugal force, aerodynamic force) and overturning moment of the robot, respectively, used to digitally map the real-time dynamic mechanical state of the robot. The aforementioned stable envelope interval refers to a safe area defined in the stability state space. For example, this range is calculated in real time based on the total adsorption capacity limit of the 24 vacuum suction cups and the friction coefficient of the blade surface. When the load falls within this range, there will be no slippage or overturning from a physical perspective. The aforementioned instability boundary refers to the outermost critical line of the stable envelope range. When the predicted load window touches or crosses this boundary, it indicates that in the step to be taken, the external destructive force (such as centrifugal force and strong wind) will be equal to or greater than the robot's maximum adhesion resistance, and the system is in a critical state on the verge of detachment. The aforementioned phase synchronization refers to aligning the pulse jet frequency / ultrasonic oscillation period of the de-icing device with the step displacement frequency of the robot's mechanical leg on the time axis. When the predicted window is stable, the step displacement and de-icing operation are synchronized to improve efficiency; when unstable, the de-icing parameters are adjusted to smooth load fluctuations, and combined with adsorption power enhancement and step delay triggering, a multi-parameter coupled active intervention is formed. The combination of predictive control and parameter feedback suppresses sudden load instability under dynamic rotation conditions, providing a flexible adjustment means for directional load environments.

[0085] In practice, the following plan can be followed:

[0086] First, a stable state space is constructed, and the predicted load window is calculated. The robot's central processing unit can have a built-in predictive model based on Kalman filtering or a state observer. The system monitors the comprehensive load vector in real time. The instantaneous amplitude and its derivative with respect to time. (i.e., load change rate). The system constructs a multi-dimensional stability state space with force and torque as coordinate axes, and pre-defines a stability envelope interval, which is the geometric boundary calculated based on the robot's current adsorption capacity, friction coefficient, and support center of gravity. It predicts the displacement step period T to be executed in the upcoming operation. s Within, the evolution trajectory of the load vector. Predicted load value. It can be represented as:

[0087]

[0088] Wherein, the value of Δt is within a single step cycle; For predicting the load vector (estimated future resultant force);

[0089] The current moment is used as the reference time point; For predicting stride time (look-ahead window), it refers to the time span required for a robot to complete a preset displacement stride (e.g., from starting to landing). This represents the current composite load vector; The load change rate represents how fast the combined load changes over time (i.e., the slope). As the load evolution increment, the cumulative load change caused by the change in blade rotation position in the future time interval Δt is estimated by integrating the load change rate within the prediction window. The load acceleration (the second derivative of the load) represents the rate of change of the load itself, i.e., the severity of load fluctuations. This is a nonlinear correction term used to adjust the prediction curve to include the acceleration effects of load variations. This term increases significantly when the wind field environment is extremely unstable, thus expanding the prediction load window. The system can more quickly reach the instability boundary, triggering the robot's safety fuse mechanism. This algorithm allows the system to predict whether the robot will reach the instability boundary during displacement due to a step change in blade speed or gust of wind, even before it takes a step. In efficiency mode, the robot operates in phase synchronization, meaning that when the system determines the predicted load window... When the robot remains within a stable envelope range, it enters a high-efficiency continuous operation mode. The central processing unit controls the stepping servo mechanism and the three-in-one de-icing module to perform phase synchronization. That is, as the suction cup foot steps forward, the ultrasonic transducer and high-pressure nozzle at the front end are simultaneously activated. The system aligns the pulse frequency of the de-icing device with the frequency of the stepping displacement, allowing the robot to clean while moving, utilizing the inertia of the body and continuous airflow to increase the de-icing area per unit time. At this time, the robot exhibits smooth and uninterrupted climbing dynamics. In safe mode, i.e., if the predicted load window... When the evolution trend reaches or exceeds the instability boundary (e.g., the robot rotates to the phase point where gravity and centrifugal force are at their maximum superposition), the system immediately triggers load intervention: the system instantaneously increases the overall vacuum of the 24 suction cup foot array via an electronic proportional valve. To smooth out the dynamic load fluctuations of the robot body 1, the system automatically reduces the physical output of the de-icing device. For example, it reduces the jet pressure of the high-pressure nozzle via a pressure regulating valve or adjusts the amplitude of the ultrasonic drive power supply to reduce the disturbance of the reaction force generated by the de-icing operation itself on the stability of the robot body 1. Secondly, the current stepping command is forcibly suspended via a command counter, keeping the robot in a static, gripping posture with all feet adsorbed. At this time, the system continuously monitors the predicted load window. The dynamics. Until the algorithm detects that the load vector has returned to the safe envelope and its rate of change... After the pressure level stabilizes and the system maintains a level that lasts for at least one detection cycle (e.g., 50ms-100ms), the displacement restriction is lifted and the step displacement is retried.

[0090] In this design, high-frequency pressure sensors and displacement feedback encoders can be distributed throughout the robot body 1. When the prediction window indicates a risk, the system sends a millisecond-level power compensation command to the chassis drive module via the CAN communication bus. This adjustment method enables the robot to adaptively adjust its operating rhythm according to the rapidly changing force state of the rotating blades, achieving robust climbing under harsh working conditions.

[0091] The steps to smooth out dynamic load fluctuations in the robot by adjusting the operating parameters of the de-icing device include:

[0092] A1. If a lifting displacement of the robot body away from the working surface is detected, or the predicted load window touches the instability boundary, the load intervention mode is triggered, and step A2 is executed. Here, the predicted load window refers to the set of maximum and minimum load vectors that the robot may experience within a preset displacement step time period in the future (i.e., the predicted extreme value range of the force) based on the comprehensive load vector at the current moment, combined with its first derivative (rate of change of time) and second derivative (load acceleration), calculated by the integral prediction algorithm.

[0093] A2. Perform asymmetric adjustment, including controlling the airflow used to perform de-icing and cleaning operations to switch from a first path pointing to the work area to a second path sprayed along the bottom surface of the fuselage, so as to accelerate the airflow tangentially at the bottom of the fuselage and generate negative pressure suction pointing to the work surface; while switching to the second path, the thermal energy input power in the de-icing and cleaning operation can be reduced to concentrate the kinetic energy of the air to strengthen the negative pressure suction.

[0094] A3. During asymmetric adjustment, the current displacement command is forcibly interrupted. If the overall load vector is determined to return to the stable envelope range and maintain a preset period, the airflow direction input of the first path is restored, and the displacement interruption is released. In this system, the displacement interruption refers to the control layer instantly suspending / clearing all current and subsequent gait displacement commands, forcing all servo joints to lock, and commanding all suction cup feet to instantly open the maximum negative pressure for in-situ locking. Monitoring the body lifting or load status, after identifying instability risks, the airflow is switched to tangential jetting at the bottom of the body, using fluid acceleration to generate aerodynamic negative pressure to provide auxiliary downpressure compensation. The forced interruption of displacement commands allows the robot to maintain its attachment posture when subjected to abnormal forces, and uses the return of the load vector to the stable range as the criterion for resuming operations. This mechanism converts de-icing air energy into aerodynamic downpressure, combined with motion state logic locking, to suppress the tendency of the body to deviate from the working surface, improving attachment safety under extreme loads.

[0095] In a specific embodiment, this solution combines dynamic attitude monitoring of fuselage 1 with aerodynamic energy redistribution logic to achieve an aerodynamic downforce reinforcement mechanism for extreme operating conditions. The following details a reference implementation of the intervention mode in this embodiment:

[0096] During the A1 phase, the main purpose is to improve displacement monitoring and instability warning. During the robot's operation, the system monitors the distance h between the bottom surface of the chassis 1 and the working surface of the blade in real time through laser rangefinders 6 or ultrasonic probes deployed at the four corners of the chassis 1.

[0097] If the sensor detects a change rate Δh of h exceeding a preset threshold (e.g., 5 mm), or the predicted load window in the aforementioned steps... Upon instantaneously crossing the safety boundary of the stable envelope interval, the central processing unit immediately issues a load intervention signal and interrupts the current de-icing task sequence.

[0098] During stage A2, the airflow path is switched and aerodynamic downforce is generated. At this time, the system, for example, controls a three-way solenoid directional valve installed in the aerodynamic circuit to perform an instantaneous switch of the airflow output path. The compressed air is switched from the first path (de-icing path) that was originally pointing to the surface of the ice layer in front to the second path (pressurization path) that flows to the gap between the fuselage 1 underside and the blade surface.

[0099] This jet of air directed towards the underside of fuselage 1 not only cleans the fuselage and its feet, but also produces an additional effect: when the high-pressure airflow is tangentially ejected along the omnidirectional converging surface 110 at the bottom of fuselage 1, the airflow is forced to accelerate within the narrow flow channel at the bottom of fuselage 1. According to Bernoulli's equation, the faster the flow velocity, the lower the pressure, thus generating a negative pressure suction force pointing towards the blade surface at the bottom of fuselage 1. The aerodynamic downforce estimation formula used here is as follows:

[0100]

[0101] The pneumatic pressure increment represents the resultant force generated by the pressure difference between the inside and outside of the fuselage 1, which is perpendicular to the working surface of the blades, when the compressed air is switched to be injected along the bottom surface of the fuselage 1. The external atmospheric pressure; The density of the jet fluid (compressed air); The overall flow velocity within the underside channel of fuselage 1; The relative wind speed outside the body 1 refers to the combined speed of the ambient wind speed outside the robot body 1 without jet interference and the wind speed induced by the rotation of the blades; The speed of the continuous high-speed airflow ejected from the flat secondary nozzle 520 on the underside of the fuselage 1; β is the active coupling coefficient, an empirical constant characterizing the kinetic energy exchange efficiency between the active jet flow and the passive flow field; β is the passive geometric acceleration coefficient (a constant determined by the physical structure of the omnidirectional convergent surface 110 of the fuselage 1 belly, for example, it can be set as a spherical cap-shaped fuselage 1 belly, β is the ratio of the inlet area of ​​the fuselage 1 belly region to the narrowest area of ​​the throat, even if the active nozzle is not working, β>1, the belly flow velocity) It will also be passively higher than the external wind speed. A e It is the effective projected area of ​​the robot's omnidirectional convergent surface (spherical crown-shaped belly) in the direction perpendicular to the working surface of the blade.

[0102] As can be seen from the above scheme, since the underside of fuselage 1 is omnidirectionally converging spherical cap, the ambient wind field entering from any phase will converge at the extreme point of the spherical cap. At this time, even if the continuous high-speed fluid ejected from the active nozzle can accelerate the air when it is in the direction of the natural wind, even if there is an angle between the active jet direction and the natural wind direction, the high-speed jet will also drive the surrounding turbulent air to converge towards the throat due to the entrainment phenomenon, transforming the disordered turbulence into an ordered acceleration flow field, which still brings a boost to the grip of fuselage 1.

[0103] Example 4: The above control method further includes a dynamic protection step based on the coupling of thermal load and aerodynamic pressure.

[0104] The system acquires the ambient temperature and the real-time attachment status of the robot body 1, and adjusts the distribution path of the thermo-pressed fluid based on the judgment results. If the robot body 1 is in a stable operating range, the thermo-pressed fluid is guided according to the first distribution ratio: a portion of the fluid is maintained to flow to the frame and attachment unit to perform anti-icing heat maintenance operation, and the remaining fluid is guided to the operating path to perform de-icing and cleaning operation. If the predicted load window touches the instability boundary, the system switches to the second distribution ratio to guide the full amount of thermo-pressed fluid to the frame and attachment unit, using the heat energy carried by the full amount of fluid to forcibly suppress the icing of the robot body 1, and using the aerodynamic negative pressure generated by the fluid flowing along the frame surface to offset the lift load, so as to maintain the robot's attachment reliability under the variable-direction composite force field.

[0105] In implementation, a combined heat and gas production module can be integrated inside the robot body 1, consisting of a high-pressure micro air compressor and a high-power electric heating resistor array. Compressed air generated by the air compressor flows through the heating array and is converted into a thermo-pressurized fluid with stable pressure (e.g., 0.4MPa-0.8MPa) and controllable temperature (e.g., 50℃-80℃). In the main pneumatic circuit of body 1, a three-way proportional control valve with millisecond-level response speed is installed to adjust the fluid flow ratio η between the de-icing operation path and the protection path of body 1 in real time, according to instructions from the central processing unit.

[0106] When the system monitors the predicted load window When the system is within a stable envelope range and the ambient temperature is below a set threshold (e.g., 0°C), it executes the first distribution ratio mode. In this mode, the three-way proportional control valve guides approximately 20%-30% of the thermo-pressurized fluid to the internal flow channels of the frame. This fluid, through sealed microchannels arranged inside the leg joints and active support rods, maintains the surface temperature of the traveling mechanism above the freezing point via heat conduction, preventing motion lag caused by grease freezing or icing in mechanical clearances. At this time, the remaining 70%-80% of the fluid is directed to the working nozzle at the front of the traveling mechanism and, in conjunction with the control algorithm, switches to a high-frequency pulse jet mode. The instantaneous momentum impact force generated by the pulsed airflow is used to peel off the ice layer, thereby maximizing the propulsion efficiency of the de-icing operation while maintaining the basic thermal metabolism of the fuselage. Once the predicted load window... Upon reaching the instability boundary, or if the sensor detects abnormal lifting displacement of the fuselage 1, the system immediately switches to the second distribution ratio mode to perform protective intervention. At this time, the three-way proportional regulating valve cuts off the air path to the front-end working nozzle, directing 100% of the thermopressurized fluid to the frame and attachment unit. While performing path redistribution, the system automatically switches the fluid output mode from high-frequency pulse to constant pressure continuous spraying to ensure the establishment of a stable physical field around the fuselage 1.

[0107] In this emergency protective state, the full volume of thermo-pressurized fluid overflows outward through linear guide slots or micro-pore arrays pre-installed on the sides of the legs and the edges of the suction cup feet. This process generates a significant coupled protective effect: First, the high thermal entropy carried by the full volume of fluid forms a thermal barrier at the contact interface between the suction cup feet and the blade surface, forcibly suppressing the growth of frost and ensuring that the sealing redundancy of the vacuum suction cup is not diminished by interface icing; second, the continuously flowing airflow rapidly sweeps across the specific curvature (guide groove) of the frame surface, generating a low-pressure zone on the frame surface according to Bernoulli's principle. This dynamic adjustment logic based on the coupling of thermal load and airflow pressure allows the robot to no longer passively endure environmental disturbances, but rather flexibly convert its own thermo-pressurized energy into an anti-icing thermal barrier and an aerodynamic anchoring mechanism to prevent takeoff. In critical moments such as sudden increases in fan speed, drastic changes in centrifugal force, or encounters with sudden strong gusts, this mechanism sacrifices de-icing progress to ensure the absolute adhesion reliability of the fuselage, significantly improving the robot's survivability and environmental adaptability under extreme changing composite force fields.

[0108] Example 5: Based on the above scheme, this method further includes the following for calculating the predicted load window within the preset displacement step:

[0109] The system collects projection data of the gravity vector from the inertial measurement unit built into the turbine hull, or receives rotor position data from the wind turbine main control system, and calculates and generates an instantaneous rotational phase angle sequence for the wind turbine hub. Substituting this instantaneous rotational phase angle sequence into the gravity vector, it calculates the normal component of the gravity vector relative to the working surface within a preset displacement step. If the direction of the normal component deviates from the working surface and its amplitude exceeds a preset safety threshold, the predicted load window is determined to have reached the instability boundary, a displacement meltdown command is output, and the static adsorption state of the current attachment unit is maintained. If the direction of the normal component points towards the working surface, the predicted load window is determined to be within the stable envelope region, the target adsorption negative pressure of the attachment unit is lowered according to the amplitude of the normal component, and a stepping command is output to drive the non-support group to perform a step displacement. The aforementioned instantaneous rotational phase angle is the real-time geometric angle between the longitudinal central axis of the wind turbine blade and the absolute gravity direction when the blade makes a 360-degree circular motion around the hub. For example, the highest point of the blade pointing upwards is the 0-degree phase (overweight compression phase), and the lowest point pointing downwards is the 180-degree phase (weightlessness tearing phase).

[0110] In implementation, the central processing unit can establish a sequence mapping of the instantaneous rotational phase angle of the wind turbine hub through two paths: The first is endogenous sensing, which utilizes a built-in six-axis inertial measurement unit (IMU) to monitor the real-time projection changes of the gravitational acceleration vector on the three axes of the turbine's 1-axis coordinate system. Through quaternion calculation or complementary filtering algorithms, the rotational trajectory of the gravity vector is inversely reconstructed into the real-time azimuth angle Φ of the wind turbine rotor. The second is exogenous communication, which uses the industrial Ethernet interface of the base station (such as EtherCAT or CANopen) to read the encoder values ​​output by the wind turbine's main control system in real time, obtaining absolute blade phase data. The two paths can serve as redundant checks to ensure the accuracy of phase data during wind turbine yaw or shutdown switching processes.

[0111] After acquiring the phase angle sequence, the system substitutes Φ into a preset mechanical model to calculate the normal projection component of the gravity vector relative to the working surface of the blade within the next displacement step cycle. When the robot rotates to the phase interval where the blade faces the ground, gravity manifests as a peeling force away from the working surface. If the system calculates that the amplitude of the normal projection component is about to exceed the safe attachment margin, such as 80% (i.e., reaching the instability boundary), the central processing unit immediately issues a displacement fuse command, forcibly interrupting the drive current of the walking motor and controlling the solenoid valves of each attachment unit to switch to the maximum negative pressure state (e.g., −85kPa), allowing the robot to use its full feet to utilize maximum grip force to resist the tearing force of gravity. Conversely, when the robot rotates to the phase interval where the blade points to the sky, gravity manifests as a positive pressure pressing against the working surface. At this time, the system determines that the predicted load window has returned to the stable envelope interval and performs energy efficiency optimization and acceleration stepping. The central processing unit automatically reduces the target negative pressure threshold of the suction cup (e.g., from -70 kPa to -40 kPa) via a proportional pressure control valve based on the natural clamping force provided by the current normal projection component, significantly reducing the operating energy consumption of the vacuum pump while ensuring physical safety. Simultaneously, the system outputs high-speed stepping commands, controlling the walking unit 2 to move rapidly within this phase by increasing its step frequency or stride. This implementation method enables the robot to achieve optimal allocation of operational efficiency while ensuring absolute safety in dynamic rotational attachment.

[0112] Example 6: Based on the above scheme, the control method further includes a risk avoidance step:

[0113] B1. During the robot's movement, the ranging feedback sequence of multiple spatial poles distributed around the robot body (the outermost extreme points of the robot chassis on the two-dimensional or three-dimensional geometric projection (such as the four outer corners of the left front, right front, left rear, and right rear), which are the locations of the ranging sensors and are used to represent the physical contour boundary of the robot body) relative to the working surface is collected in real time to map the geometric boundary topological field of the local movement area.

[0114] B2. Dynamically analyze the topological field of the geometric boundary to extract the curvature change profile representing the edge of the working surface (representing the leading and trailing edges of the wind turbine blades (crossing this line means falling off the blade)) and the structural interference envelope surface representing the root rotation hub (the hub entity at the root of the wind turbine (touching this surface means a rigid collision)).

[0115] B3. Real-time calculation of the spatial approximation rate of the spatial pole relative curvature change profile and structural interference envelope surface of the fuselage periphery;

[0116] B4. If any spatial approach rate is determined to exceed the preset safety threshold, a forced risk avoidance intervention is triggered. In the comprehensive load vector, a virtual repulsive force vector opposite to the boundary crossing trend is dynamically injected. The robot's trajectory is guided to converge towards the safe longitudinal axis by asymmetrically adjusting the step displacement and attachment timing of the attachment units distributed on both sides of the fuselage. The aforementioned virtual repulsive force vector is the core mathematical variable in the artificial potential field control algorithm. It is a penalized algorithmic thrust calculated by the central processing unit; the closer to the edge, the larger the algorithmic value, used to forcibly push the robot's trajectory from the edge back to the safe center of the blade during vector synthesis. This scheme constructs the geometric boundary topological field of the local area through ranging feedback, identifies curvature abrupt changes and structural interference envelope surfaces at the edge of the working surface, and calculates the approach rate in real time to predict boundary crossing risks. When a risk is detected, the system injects a virtual repulsive force into the load vector and adjusts the displacement and timing of the attachment units to guide the robot's trajectory towards the safe axis. This method of transforming boundary constraints into dynamic control reduces the risk of robot falls and collisions in complex contours.

[0117] In a specific embodiment, this solution employs the aforementioned approach to configure an end-effector for the safe operation of the robot on a high-altitude rotating blade. This step transforms complex geometric boundary constraints into intuitive motion control quantities by constructing a virtual potential field logic.

[0118] Specifically, a sensor array consisting of multiple sets of high-frequency TOF laser rangefinders 6 or ultrasonic sensors can be mounted around the chassis 1 of the robot (i.e., the spatial poles). The sensors are respectively positioned on the front, rear, and outermost walking foot support rods on both sides of the chassis 1. During the robot's climbing process, the sensor array synchronously collects distance data perpendicular to the working surface at each pole at a frequency of no less than 100Hz. The central processing unit combines these multi-point distance feedbacks into a local geometric topological field matrix through time series analysis. This matrix can reflect in real time the flatness, tilt, and distance distribution of each edge relative to the physical boundary of the blades below fuselage 1. The system executes feature extraction operators by analyzing the topological field matrix in real time.

[0119] Because the leading and trailing edges of wind turbine blades have significant curvature changes, the distance measurement value at a certain extreme point on one side... A momentary step change occurs (such as from a contact state to a range overflow), or a range gradient occurs. When a pre-set mutation threshold is exceeded, the system marks it as a physical edge of the blade. As the robot approaches the blade root, sensors detect structural features that transition from a planar to a vertical plane, and the system uses this information to establish the rotational pivot interference envelope. The system calculates the projection of the robot's center-of-mass velocity vector onto the normal directions of each boundary, obtaining the spatial approximation rate. :

[0120]

[0121] in, For travel speed, The angle between the direction of travel and the boundary normal. This represents the remaining safety clearance. When... Exceeding the safety threshold indicates that the robot is at risk of falling or colliding.

[0122] Once a risk is identified, the system no longer relies on manual intervention but activates an automatic risk avoidance mechanism. The central processing unit generates a virtual repulsive force vector whose magnitude is proportional to the approximation rate and whose direction deviates from the boundary. This vector is superimposed on the composite load vector in real time. The corrected resultant force is obtained. = + .

[0123] The controller adjusts the asymmetrical gait parameters of the 24 suction cup feet on the left and right sides according to the corrected resultant force direction. If the robot deviates to the left edge, the system automatically increases the step displacement amplitude of the left suction cup foot and decreases the step length on the right side. The step length difference is used to generate a corrected yaw angle. By delaying the lifting timing of the suction cup foot on the risk side (near the edge), the attachment time ratio on that side is increased, thereby generating a torque pointing towards the longitudinal axis of the blade center.

[0124] This risk avoidance logic transforms defined boundaries into calculable thrust, enabling the robot to achieve self-aware tracking based solely on local topological information around the fuselage, without requiring a global map. In environments with rotating blades exhibiting variable cross-sections and complex twist angles, this mechanism guides the robot to consistently perform de-icing within a safe area at the blade's center, effectively mitigating falls caused by positioning errors or strong wind deflection.

[0125] Example 7, referring to Figures 1 to 6A rotating blade de-icing robot is used to execute the aforementioned control method. The robot includes a body 1, a negative pressure attachment matrix, and a central processing unit. The back of the body 1 away from the working surface is a stratospheric surface, and its belly facing the working surface is an omnidirectional convergent surface 110. The omnidirectional convergent surface 110 smoothly bulges inward from the peripheral edge of the body 1, and the spatial extreme point of the bulge is located at the geometric center of the belly of the body 1. This is used to ensure that the body 1, under any horizontal phase airflow intrusion, defines a negative pressure channel together with the working surface, so as to convert the multi-directional turbulent wind field into aerodynamic downforce pointing towards the working surface. The negative pressure attachment matrix is ​​symmetrically assembled on the two wing sides of the body 1. The negative pressure attachment matrix includes multiple walking units 2. Multiple walking units 2 include active joints and vacuum suction cups located at the ends of the active joints. In a specific arrangement, the walking units 2 on one side are arranged in the following topology along the longitudinal axis of the fuselage 1: an independent single-pole pilot foot at the front end, ten array feet arranged linearly in the middle section, and an independent single-pole drag foot at the rear end. A total of twenty-four independently controlled vacuum suction cups on both sides jointly construct a multi-point redundant gripping topology. The central processing unit is built into the fuselage 1 or located outside the fuselage 1 and is connected to each walking unit 2 and the multi-dimensional sensor network. It is used to acquire environmental data to synthesize a comprehensive load vector and calculate the slip trend and overturning moment center based on the spatial orientation of the comprehensive load vector. It is used to control the negative pressure adhesion matrix to perform dynamic adhesion compensation.

[0126] The robot's fuselage 1 is integrally molded from lightweight, high-strength composite materials (such as carbon fiber or fiberglass), exhibiting a unique asymmetric aerodynamic configuration. The back of fuselage 1 is designed as a purely planar stratospheric surface. This design not only facilitates the placement of large-area flexible solar panels but also effectively reduces the drag from the backflow in high-altitude, strong winds. The underside of fuselage 1 features a spherical, omnidirectionally converging surface 110 that smoothly bulges inward. The geometric feature of this surface is that its highest point (the extreme point of the bulge) is precisely located at the center of the underside, creating a three-dimensional Venturi channel that is wider on the outside and narrower on the inside between the edge of fuselage 1 and the blade surface. When natural wind or induced wind generated by blade rotation cuts into the underside of fuselage 1 from the horizontal phase periphery (whether front, rear, left, or right), the airflow is forcibly accelerated as it passes the throat of the underside (i.e., the central extreme point). According to Bernoulli's equation, this velocity surge generates a stable negative pressure zone, thus producing passive aerodynamic downforce directed towards the blade surface. Experiments have shown that this configuration ensures that the higher the ambient wind speed and linear velocity, the stronger the downward grip force generated by the fuselage 1, which automatically offsets part of the centrifugal overturning tendency caused by high-speed rotation from the physical structure level.

[0127] The robot's body 1 is symmetrically equipped with a total of 24 independently controlled walking units 2 on both sides. To achieve efficient displacement and stable attachment on long-span wind turbine blades, 12 units are configured on each side: a pilot leg (one at the front) installed at the very front of the body 1, equipped with a high-stroke active joint, primarily used for detecting the working path ahead and serving as the first exploratory fulcrum after de-icing operations; and 10 array legs (ten in the middle section) arranged in a compact linear configuration, serving as the robot's core gripping module. These 20 array legs (on both sides) form a highly redundant support plane, ensuring sufficient attachment margin even if some suction cups experience momentary air leakage under extreme load fluctuations. A trailing leg (one at the rear) is located at the tail of the body 1, balancing the body 1's center of gravity and assisting in trajectory correction during the walking sequence. Each walking unit 2's active joint is driven by a high-torque servo motor or proportional hydraulic cylinder, and its end is equipped with a high-tear-strength silicone vacuum suction cup. Each suction cup branch is independently connected to an electronic proportional vacuum valve and a miniature pressure sensor. This multi-point redundant topology allows the central processing unit to perform independent negative pressure adjustment for each foot, achieving dynamic allocation of adsorption power on demand. The central processing unit can be built into the sealed chamber of the robot body 1 or deployed externally via umbilical cable 8, depending on actual operation and maintenance needs. The unit is connected to a multi-dimensional sensor network via a bus, including a built-in six-axis inertial measurement unit (IMU), four laser obstacle avoidance sensors mounted on the edge of the robot body 1, and 24 pressure feedback modules distributed within the suction cups. The central processing unit acquires the robot's three-axis attitude, revolution speed, and ambient wind direction in real time, and synthesizes a comprehensive load vector according to a preset algorithm. Once a slippage trend or a shift in the overturning center is calculated, the processor immediately sends a compensation command to the actuator on the corresponding side, increasing the negative pressure value of the suction cups in that area to generate a constraint torque on the opposite side of the robot body 1.

[0128] The forward-moving front end of the fuselage 1 is equipped with an ultrasonic de-icer 3, an electrothermal de-icing probe 4 (e.g., a single-headed electric heating tube), and a pneumatic jet assembly 5. The air inlet of the pneumatic jet assembly 5 is connected to a high-pressure air source, and its exhaust end is connected to the fluid inlet of a multi-way reversing valve (three-way proportional reversing valve). The multi-way reversing valve has a first outlet and a second outlet. The first outlet pipe is connected to the main nozzle 510 facing the working surface in front of the ultrasonic transducer array. The second outlet pipe is connected to the auxiliary nozzle 520. The opening of the auxiliary nozzle 520 faces the omnidirectional converging curved surface 110 (e.g., a spherical crown surface) on the belly of the fuselage 1, and the auxiliary nozzle 520 is a horizontally flat opening (three can be arranged, two on the side and one tangentially facing the belly). The jetting direction of the auxiliary nozzle 520 is parallel to the plane of the back panel of the fuselage. The ultrasonic, electrothermal, and pneumatic components integrated at the front end of the fuselage 1 form a multi-physical energy field coupled end-effector, facilitating targeted treatment of the icing layer. Fluid path guidance via a multi-way reversing valve allows the high-pressure air source to switch output between the main and auxiliary nozzles 520 as needed. The auxiliary nozzle 520 employs a transversely flat orifice structure, pointing towards the underside of the fuselage 1 at a specific tangential angle. This geometric arrangement promotes the formation of a uniformly diffused jet of fluid below the omnidirectional converging surface 110. This jet not only helps maintain the cleanliness of the adhesion unit interface but also utilizes the acceleration effect of the fluid within the narrow flow channel to generate aerodynamic pressurization, providing additional physical support for the robot's stability when facing centrifugal force and changing wind fields. This combination of pipeline layout and nozzle shape design balances path cleaning efficiency with adhesion redundancy during the operation of the fuselage 1.

[0129] Perform multimodal collaborative de-icing intervention steps in chronological order, including:

[0130] The ultrasonic de-icer 3 outputs directional high-frequency oscillating waves to the ice layer in front of the travel path to break the initial freezing interface between the ice and the working surface; the electrothermal de-icing probe 4 injects transient concentrated heat flow into the ice area affected by the oscillating waves, and uses the thermal expansion stress generated by the local temperature step to induce the deterioration and micro-cracks in the internal structure of the ice to assist in de-icing; a pulsed high-pressure airflow is released from the main nozzle 510 along the first path and pointing forward, using fluid kinetic energy to peel the deteriorated and loosened ice off the working surface.

[0131] During implementation, the front end of the composite de-icing module is equipped with an ultrasonic transducer array (such as a piezoelectric ceramic transducer), which generates directional high-frequency oscillation waves with a frequency that can be set between 20-40kHz. As the robot advances towards the ice layer, the ultrasonic mechanical waves generated by the transducer act directly on the contact interface between the ice layer and the blade composite material (substrate) through the waveguide rod at the front end of the robot body 1. This high-frequency, low-amplitude mechanical energy can effectively sever the hydrogen bonds between the ice crystals and the substrate surface, placing the ice layer in a loose, decoupled state, thus providing the physical prerequisite for subsequent peeling. The module is equipped with four sets of electrothermal de-icing probes (or a high-intensity infrared radiation array). The central processing unit controls the drive power supply to inject a transient high current into the probe sets, generating a high-density heat flow field in a very short time. The probes cut into the decoupled ice layer or adhere closely to its surface, injecting heat energy into the ice body and creating a severe temperature gradient in the depth direction of the ice layer. Due to the significant difference between the linear expansion coefficient of ice and the blade substrate, this local temperature jump will induce huge transient thermal stress. Under the action of this stress, the originally hard ice layer will rapidly generate a network of microcracks from the inside out, transforming it from a dense whole into a loose granular or sheet-like structure, reducing the mechanical strength of the ice layer. The end of the composite de-icing module is equipped with a pulsed high-pressure pneumatic nozzle, which is connected to the high-pressure air source of the system through a high-speed pulse valve. After ultrasonic decoupling and thermal degradation are completed, the system releases a pulsed high-pressure airflow (with an adjustable pressure range of 0.5-1 MPa) pointing forward along the first path. The high-energy pulsed jet directly scours the weakened ice surface, completely peeling off the broken ice blocks that have lost their adhesion from the working surface and blowing them away from the blade rotation path.

[0132] Furthermore, after a single peeling action is completed, the jet direction of the high-pressure airflow is dynamically changed, causing it to perform a surface-to-surface sweep along the gap between the underside of the robot body 1 and the working surface. The sweeping flow field is used to blow away the attachment unit to peel off the derived frost, and simultaneously, a transient hydrodynamic negative pressure is generated in the gap, which serves as a pre-tightening adhesion force for entering the next travel cycle. In this way, after a single peeling action, the jet direction of the airflow is dynamically changed, causing it to perform a surface-to-surface sweep along the gap between the underside of the robot body 1 and the working surface. This allows the high-speed fluid to blow away the derived frost on the surface of the attachment unit and its joints, reducing the risk of motion lag caused by icing in the walking mechanism under low-temperature conditions. At the same time, the transient hydrodynamic negative pressure generated by the sweeping flow field in the narrow gap at the bottom of the robot body 1 provides the robot body 1 with an additional load pointing towards the working surface, serving as a pre-tightening adhesion force for entering the next travel cycle. This helps to maintain the stability of the attachment state at the moment of step switching. This logic of recovering and utilizing pneumatic energy for interface purification and active pressurization improves the dynamic attachment redundancy of the robot during cyclic operations.

[0133] This rotary blade de-icing robot also includes a power and media supply assembly. This design can be implemented in two ways: one is that the power and media supply assembly is built into the body 1. This assembly includes an energy storage battery pack, a micro vacuum pump, and a micro air compressor. The pumping end of the micro vacuum pump is connected to each vacuum suction cup via pipelines, and the exhaust end of the micro air compressor serves as a high-pressure air source connected to the pneumatic injection assembly 5. Using either a built-in integrated or pipeline-connected power and media supply assembly allows the robot to adapt to diverse operational environments. The layout of the built-in battery pack and micro pump helps eliminate the additional resistance generated by external pipelines in the rotating flow field, improving the robot's movement flexibility along its operating path. Furthermore, the design of introducing external base station energy and air sources through a composite connection port and umbilical cable 8 provides long-term operational capability while reducing the load on the body 1, which helps reduce centrifugal disturbances caused by its own weight. This switchable supply architecture optimizes the configuration relationship between the power source and the mass of the body 1 according to the specific blade size and de-icing conditions, enhancing the robot's environmental adaptability and task redundancy under complex force fields. A high-pressure buffer tank is connected in series between the micro air compressor and the pneumatic injection assembly 5 to smooth airflow pulsations; a vacuum stabilizing tank is connected in series between the micro vacuum pump and the vacuum suction cup to store negative pressure redundancy. Furthermore, the power and media supply components can also integrate a heat exchange structure. A heat exchange channel is fixedly laid inside the body 1. The heat-absorbing end of this channel covers the outer side of the micro air compressor housing and the electric heating array, while its heat-releasing end is physically connected to the insulation layer of the energy storage battery pack and the active joints of each walking unit 2, used to conduct waste heat generated during system operation to the core components to perform anti-icing heat maintenance. Specifically, when the power and medium supply components adopt an integrated configuration built into the fuselage 1, at the structural level of the fuselage 1, the fuselage 1, which serves as a carrier, is symmetrically equipped with a 24-way vacuum suction cup foot array on both sides, and an ultrasonic de-icing device 3, an electrothermal de-icing probe 4, and a pneumatic jet assembly 5 are fixedly arranged at its front end; while in the safety cabin inside the fuselage 1, the energy storage battery pack, the central processing unit, the micro air compressor, and the micro vacuum pump are centrally anchored.

[0134] The power output of the energy storage battery pack is connected to the central processing unit via internal wiring harnesses, and branches outwards to the ultrasonic de-icer 3, the electrothermal de-icing probe 4, the micro air compressor, the micro vacuum pump, and the servo valve assembly controlling each vacuum suction cup. Simultaneously, the waste heat generated by the micro air compressor and electrothermal devices during operation is directed back through a pre-designed heat-conducting channel within the machine body 1, wrapping around the outer shell connected to the energy storage battery pack and the mechanical joints of each vacuum suction cup foot, utilizing heat conduction to achieve anti-icing and heat maintenance operation under low-temperature conditions. The exhaust end of the micro air compressor is first connected to a buffer tank to smooth airflow pulsations, and the high-pressure output port of the buffer tank is then connected to the fluid inlet of a three-way proportional reversing valve. This reversing valve is controlled to branch into two paths: the first outlet pipe connects to the main nozzle 510 facing the working surface, and the second outlet pipe connects to the auxiliary nozzle 520 facing the omnidirectional converging curved surface 110 on the underside of the machine body 1. The pumping end of the miniature vacuum pump (buffered by a vacuum pressure stabilizing tank) is directly connected to the vacuum manifold of the chassis 1. The manifold further branches, each connecting to a 24-way vacuum suction cup foot array on both sides, providing gripping negative pressure for the attachment unit. The input end of the aforementioned central processing unit is connected to the sensor array (including an inertial measurement unit, a ranging sensor, and a pressure sensor) on the periphery of the chassis 1 via a communication bus. Its output control end is connected to and controls the switching coil of the three-way proportional reversing valve, the power trigger end of the ultrasonic and electrothermal de-icing probe 4, and the negative pressure regulating valve of the vacuum suction cup foot array via electrical signal wiring harnesses.

[0135] Combination Figure 6Another type is a power and medium supply assembly, which includes a composite connection port located at the rear of the fuselage 1 and an umbilical cable 8 connected to the composite connection port. The umbilical cable 8 integrates a power supply cable, a negative pressure suction hose connected to each vacuum suction cup, and a high-pressure air supply hose as a high-pressure air source. The power and medium supply assembly also includes a supply base station 9; the supply base station 9 is externally equipped with a mounting connector, which is used to fix the supply base station 9 to the rotating hub surface of the wind turbine. The end of the umbilical cable 8 away from the composite connection port of the fuselage 1 is connected to the supply base station 9, and the wiring trajectory of the umbilical cable 8 extends along the spanwise direction of the rotating blades, so that the supply base station 9 and the fuselage 1 are in the same rotational coordinate system. When the wind turbine performs unidirectional continuous revolution or experiences a sudden change in speed, the supply base station 9 maintains zero relative angular velocity with the fuselage 1. This ensures that the umbilical cable 8 connecting the two only experiences longitudinal bending and dragging as it extends along the blade span. This allows the heavy power and medium to be supplied to the peripheral device of the base station 9, transforming the robot's main working body on the blade into a lightweight end effector configuration. This peripheral device structure significantly reduces the physical mass of the fuselage 1 itself, thereby directly reducing the centrifugal force and dynamic inertial load generated when the robot rotates with the blade. This greatly alleviates the adhesion pressure and overturning / falling risk of the walking unit 2 under the combined force field of changing direction. At the same time, the peripheral base station removes the spatial volume constraints inside the fuselage 1, allowing for the configuration of larger capacity energy storage units and higher power air source equipment, providing continuous medium support for long-term, high-intensity de-icing operations.

[0136] During deployment, within the base station 9, components such as the energy storage battery pack, air compressor, vacuum pump, and base station control unit are centrally fixed. The power output lines of the energy storage battery pack, the high-pressure exhaust pipe of the air compressor (via a buffer tank), and the negative pressure extraction pipe of the vacuum pump (via a vacuum pressure stabilizing tank) converge within the base station and are statically connected to the junction interface at the front end of the base station. Simultaneously, the waste heat generated by the air compressor circulates within the base station through heat exchange pipes and is then connected to the energy storage battery pack. The base station control unit is electrically connected to the start / stop control terminals of the air compressor and vacuum pump, and a reserved signal port for external communication is also connected to this junction interface. The umbilical cable 8 is connected at its starting end and rigidly locked to the junction interface of the base station. The power supply cable, high-pressure air supply hose, negative pressure extraction hose, and communication signal line (such as RS485 or CAN bus) integrated inside the umbilical cable 8 pass sequentially through a set of elastic tensioning wheel mechanisms when exiting the base station cavity, forming a physical buffer for retraction and extension. Subsequently, the umbilical cable 8, in terms of its length configuration, can extend along the span of the wind turbine blades, with its end furthest from the base station connected to the composite connection port at the tail of the robot body 1. Inside the robot body 1, the high-pressure air supply hose leading from the connection port connects to the fluid inlet of a three-way proportional reversing valve. The two outlets of the reversing valve are respectively connected to the main nozzle 510 of the front de-icing module and the auxiliary nozzle 520 of the abdomen. The negative pressure suction hose leading from the connection port connects to the vacuum manifold of the chassis of the body 1, and then distributes to the vacuum suction cup arrays on the left and right sides. The power supply cable in the connection port is physically connected to the power management motherboard inside the body 1. The motherboard discharges externally to the ultrasonic de-icer 3, the electrothermal de-icing probe 4, the suction cup foot servo motor, and the central processing unit. The robot's central processing unit (e.g., an STM32H743 as the main brain) is connected not only to the peripheral sensor array and various actuators, valves, and injection components via the internal bus of the robot body 1, but its communication port is also connected to a composite connector port, and then to the base station control unit (e.g., an STM32F405 as the slave brain) via a communication cable within the umbilical cable 8. In this distributed dual-brain architecture, the base station control unit receives remote commands from the central processing unit, connects downwards, and controls the take-up and release motors of the tensioning wheel mechanism.

Claims

1. A control method for a rotating blade de-icing robot, used to control the de-icing robot to walk on a rotating blade via an attachment unit formed by a single vacuum suction cup foot or a group of multiple vacuum suction cup feet, characterized in that, The control method includes: The centrifugal force vector, gravity vector, and airflow load vector acting on the robot are obtained and then superimposed to synthesize a comprehensive load vector. Based on the spatial orientation of the comprehensive load vector, the slippage trend of the fuselage relative to the working surface and the center of the overturning moment are calculated. Based on the slippage trend and overturning moment center, perform dynamic adhesion compensation: If the slippage trend is determined to exist, the adsorption force of the attachment unit located in the region opposite to the slippage trend is increased in a directional manner to generate a constraint torque that is dynamically balanced with the external dynamic load, so as to make the robot's attachment on the working surface tend to be balanced.

2. The control method for a rotary blade de-icing robot according to claim 1, characterized in that, The underside of the robot body facing the work surface is set as an omnidirectional convergent surface, which smoothly bulges inward from the peripheral edge of the body; the back side of the robot body away from the work surface is set as a stratospheric surface. During the climbing operation, the drive attachment unit performs a step displacement, and a transient smoothing step based on the fuselage aerodynamic clearance is also included, including obtaining the real-time clearance height between the fuselage bottom surface and the working surface, and the instantaneous rate of change of the clearance height. Determine the execution phase of the current step displacement: If it is determined that the attachment unit is in the lifting phase of detaching from the working surface, and the instantaneous increment of the real-time gap height is greater than the preset aerodynamic instability threshold, then the transient outward tilting moment of the fuselage is calculated according to the comprehensive load vector, and a collaborative compensation command is output to increase the adsorption force of the attachment unit on the fuselage side. If it is determined that the attachment unit is in the landing phase when it touches the working surface, and the negative instantaneous change rate of the gap height exceeds the preset stagnation limit, then an attitude decompression command is output to drive the active joint of the attachment unit located at the reverse end of the fuselage to perform a preset slight lift. The real-time adhesion pressure fed back by the adhesion unit that performs the step displacement is obtained. If it is determined that the pressure has reached the stable threshold, a reset command is output to restore the initial gap between the machine body and the working surface.

3. The control method for a rotary blade de-icing robot according to claim 1, characterized in that, It also includes calculating the predicted load window within a preset displacement step based on the evolution trend of the instantaneous amplitude and time rate of change of the comprehensive load vector in the stability state space, including: If the predicted load window is within the preset stable envelope range, the step displacement is synchronized with the phase of the de-icing operation to maintain the continuous operation of the robot. If the predicted load window touches the instability boundary, while increasing the adsorption power, the operating parameters of the de-icing device are adjusted to smooth the dynamic load fluctuation of the robot, and the triggering of the step displacement is delayed until the predicted load window returns to the stable envelope range.

4. The control method for a rotary blade de-icing robot according to claim 3, characterized in that, The step of adjusting the operating parameters of the de-icing device to smooth out fluctuations in the robot's dynamic load includes: A1. If a lifting displacement of the fuselage away from the working surface is detected, or the predicted load window touches the instability boundary, the load intervention mode is triggered and step A2 is executed. A2. Perform asymmetric adjustment, including controlling the airflow used to perform de-icing and cleaning operations to switch from a first path pointing to the work area to a second path sprayed along the bottom of the fuselage, in order to accelerate the airflow tangentially at the bottom of the fuselage; A3. During the execution of the asymmetric adjustment, the current displacement command is forcibly interrupted; if it is determined that the comprehensive load vector returns to the stable envelope interval and maintains a preset period, the airflow direction input of the first path is restored.

5. The control method for a rotary blade de-icing robot according to claim 3, characterized in that, It also includes dynamic protection measures based on the coupling of thermal load and aerodynamic pressure: The system acquires the ambient temperature and the real-time adhesion status of the machine body, and adjusts the distribution path of the thermo-pressurized fluid based on the judgment results. If the machine body is in a stable operating range, the thermo-pressurized fluid is guided according to the first distribution ratio: Maintain a portion of the fluid flow to the frame and attachment unit to perform anti-icing and heat maintenance operations, and direct the remaining fluid to the work path to perform de-icing and cleaning operations; If the predicted load window reaches the instability boundary, the system switches to the second allocation ratio to guide the full amount of the thermo-pressed fluid to the frame and attachment unit.

6. The control method for a rotary blade de-icing robot according to claim 3, characterized in that, The calculation of the predicted load window within the preset displacement step also includes: Collect the projection data of the gravity vector from the inertial measurement unit built into the fuselage, or receive the rotor position data from the main control system of the wind turbine, and calculate and generate the instantaneous rotation phase angle sequence of the wind turbine hub. Substitute the instantaneous rotation phase angle sequence into the gravity vector to calculate the normal component of the gravity vector relative to the working surface within the preset displacement step. If the direction of the normal component deviates from the working surface and the amplitude exceeds the preset safety threshold, it is determined that the predicted load window has reached the instability boundary, a displacement melting command is output, and the static adsorption state of the current attachment unit is maintained. If the direction of the normal component points to the working surface, it is determined that the predicted load window is in the stable envelope range. The target adsorption negative pressure of the attachment unit is reduced according to the amplitude of the normal component, and a stepping command is output to drive the non-support group to perform step displacement.

7. The control method for a rotary blade de-icing robot according to claim 1, characterized in that, It also includes risk avoidance steps: B1. During the robot's movement, the ranging feedback sequence of multiple spatial poles distributed around the robot body relative to the working surface is collected in real time to map the geometric boundary topological field of the local movement area; B2. Dynamically analyze the topological field of the geometric boundary to extract the curvature abrupt change profile representing the edge of the working surface, and the structural interference envelope representing the root rotation hub; B3. Real-time calculation of the spatial approximation rate of the spatial poles of the fuselage periphery relative to the curvature change profile and the structural interference envelope surface; B4. If any of the aforementioned spatial approximation rates is determined to exceed the preset safety threshold, a forced risk avoidance intervention is triggered; in the aforementioned comprehensive load vector, a virtual repulsive force vector opposite to the out-of-bounds trend is dynamically injected, and the robot's trajectory is guided to converge toward a safe longitudinal axis by asymmetrically adjusting the step displacement and attachment timing of the attachment units distributed on both sides of the body.

8. A rotary blade de-icing robot, used to perform the control method as described in any one of claims 1 to 7, characterized in that, The robot includes: The fuselage has a horizontal flow surface on its back side away from the working surface, and an omnidirectional convergent surface on its belly side facing the working surface. The omnidirectional convergent surface smoothly bulges inward from the peripheral edge of the fuselage, and the spatial extreme point of the bulge is located at the geometric center of the belly side of the fuselage. This is to ensure that the fuselage, together with the working surface, defines a negative pressure flow channel under any horizontal phase airflow intrusion. The negative pressure adhesion matrix is ​​symmetrically assembled on the two wing sides of the fuselage. The negative pressure adhesion matrix includes multiple walking units, each of which includes an active joint and a vacuum suction cup disposed at the end of the active joint. The central processing unit, either built into the fuselage or located outside the fuselage, is communicatively connected to each of the walking units and the multi-dimensional sensor network. It is used to acquire environmental data to synthesize the comprehensive load vector, and to calculate the slip trend and overturning moment center based on the spatial orientation of the comprehensive load vector. It is used to control the negative pressure adhesion matrix to perform dynamic adhesion compensation.

9. A rotary blade de-icing robot according to claim 8, characterized in that, The front end of the fuselage is equipped with an ultrasonic de-icer, an electrothermal de-icing probe, and a pneumatic jet assembly. The air inlet of the pneumatic injection assembly is connected to a high-pressure air source, and its exhaust end is connected to a fluid inlet with a multi-way reversing valve. The multi-way reversing valve has a first outlet and a second outlet, and the first outlet pipeline is connected to the main nozzle facing the working surface in front of the ultrasonic transducer array. The second outlet pipe is connected to the secondary nozzle, the opening of which faces the omnidirectional converging curved surface of the fuselage belly, and the secondary nozzle is a horizontally flat opening; the spray direction of the secondary nozzle is parallel to the plane of the fuselage back panel.

10. A rotary blade de-icing robot according to claim 9, characterized in that, It also includes power and media supply components; The power and medium supply assembly is built into the body. This assembly includes an energy storage battery pack, a micro vacuum pump, and a micro air compressor. The suction end of the micro vacuum pump is connected to each of the vacuum suction cups via a pipeline, and the exhaust end of the micro air compressor serves as the high-pressure air source connected to the pneumatic injection assembly; or The power and medium supply assembly includes a composite connection port located at the rear of the fuselage and an umbilical cable connected to the composite connection port. The umbilical cable integrates a power supply cable, a negative pressure suction hose connected to each of the vacuum suction cups, and a high-pressure air supply hose serving as the high-pressure air source.

Images