Boiler heating surface inspection unmanned aerial vehicle displacement closed loop and anti-shake control method

Abstract

The present application provides a kind of boiler heating surface inspection unmanned aerial vehicle displacement closed loop and anti-shake control method, belong to unmanned aerial vehicle flight control and industrial equipment automation operation technology field, it can at least partially solve the problem of insufficient position stepping accuracy of existing technology when unmanned aerial vehicle is inspected in restricted space, dynamics shaking caused by motion instruction step and detection load coordination misalignment.This application includes: based on positioning data to construct physical displacement feedback closed loop, monitor positioning signal quality and switch between multi-modal to maintain displacement feedback continuity;Under the constraint of jerk, seven-segment smoothing processing is performed on motion instruction to eliminate step impact, and lateral deviation correction is superimposed to maintain straight path;Physical displacement increment is used as operation state jump criterion to drive inspection tempo, and detection load sampling is triggered when displacement reaches preset threshold.The present application realizes centimeter-level step control and dynamics anti-shake of boiler heating surface inspection.

Description

Technical Field

[0001] This invention belongs to the field of UAV flight control and industrial equipment automated operation and maintenance technology, specifically relating to a closed-loop displacement and anti-shake control method for a boiler heating surface inspection UAV. Background Technology

[0002] The heating surface tubes of boilers in thermal power plants are prone to thinning, cracking, and deformation due to long-term exposure to high-temperature corrosion and complex flue gas environments, making them a major cause of unplanned unit shutdowns. Current inspection methods relying on manual scaffolding or wall-climbing robots generally suffer from limitations such as long preparation times, high safety risks, and poor obstacle-crossing ability on ash-covered and coked surfaces. While UAV-based inspection solutions have emerged as a new alternative due to their flexibility, they still face significant challenges in inspections within confined spaces.

[0003] When hovering over a furnace, drones are typical aerial floating platforms, lacking physical support and highly susceptible to turbulent airflow. Existing industrial drones mostly employ time-based open-loop control logic, which, in the presence of random position drift, makes it difficult to guarantee the absolute physical accuracy of the inspection trajectory. This often leads to missed inspections or duplicate coverage during pipe inspections, failing to meet the requirements for centimeter-level precision operations.

[0004] Operations in confined spaces place extremely high demands on dynamic stability. When the automated system issues virtual joystick commands, a sudden jump in amplitude can cause violent fluctuations in rotor thrust, leading to inertial jitter and overshoot of the fuselage. This instability is amplified during near-wall operations, not only causing blurred visual imaging but also disrupting the stable coupling between the probe and the tube wall. Furthermore, the lack of a physical position-level timing coordination mechanism between the flight control system and the mission payload makes it difficult to achieve tight physical and spatial alignment of the "displacement-trigger-sampling" closed-loop process.

[0005] Therefore, how to design a UAV control method that can overcome the instability of floating platforms, eliminate the step impact of motion commands, and realize physical displacement triggering is a technical problem that urgently needs to be solved in the field of boiler heating surface inspection. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method for closed-loop displacement and anti-shake control of a boiler heating surface inspection drone.

[0007] To achieve the above objectives, the present invention provides a closed-loop displacement and anti-shake control method for a boiler heating surface inspection drone, comprising: A physical displacement feedback closed loop is constructed based on real-time acquired UAV 3D spatial positioning data. The standard deviation of the positioning solution and its change slope in the positioning data are monitored. The loop switches between normal mode, degraded mode, shadow tracking mode and convergence mode according to the standard deviation of the positioning solution and the change slope, so as to maintain the continuity of displacement feedback in the physical displacement feedback closed loop. Under jerk constraints, the target displacement is smoothed to generate a continuous velocity command curve. During the execution of the velocity command curve, the lateral deviation of the UAV's current position relative to the preset flight path is calculated based on the vector projection algorithm. A corrected velocity opposite to the direction of the lateral deviation is generated and superimposed on the velocity command curve to maintain a straight inspection track. Using the actual displacement increment in the physical displacement feedback closed loop as the criterion for switching operating states, the UAV is driven to perform cyclic operations of translation, dwelling and stepping along the boiler heating surface tube bank. When the actual displacement increment reaches a preset displacement threshold, a trigger signal is sent to the external detection load to achieve alignment between the detection sampling and the spatial position.

[0008] Furthermore, the method for switching between the normal mode, the degraded mode, the shadow tracking mode, and the convergence mode is as follows: When the instantaneous value of the standard deviation of the positioning solution is lower than the preset standard deviation threshold and the slope of change is lower than the preset slope threshold, the normal mode is maintained, and the positioning data is directly used to construct the physical displacement feedback closed loop. When the instantaneous value of the standard deviation of the positioning solution increases or the slope of change exceeds the preset slope threshold but the positioning signal is not completely lost, switch to the degraded mode, reduce the flight speed of the UAV and continue to use the positioning data; When the positioning signal is completely lost, switch to the shadow tracking mode and use the historical velocity vector buffered before the loss to calculate the virtual position in order to maintain the logical continuity of displacement feedback. When the positioning signal is recovered from the lost state, the system switches to the convergence mode and constructs a Bézier curve to guide the UAV to smoothly transition from the virtual position to the recovered real position.

[0009] Furthermore, the method for calculating the virtual position in the shadow tracking mode is as follows: set up The location signal was lost at the time of loss, and the position and velocity at the time of loss were respectively... and ,exist At that moment, the virtual location Estimate using the following formula: ; in, Indicates at time The calculated virtual location, This indicates the actual location at the moment the positioning signal was lost. The velocity vector representing the moment the positioning signal was lost. Indicates the current moment.

[0010] Furthermore, the Bézier curves in the convergence mode are constructed as follows: Let the virtual location be The restored true location is The Bézier curve path Defined by four control points, parameters : ; in, Starting point End point; intermediate control point , ; This represents the velocity vector before the positioning signal was lost. This represents the target velocity vector after convergence. and This is the preset trajectory shape adjustment coefficient.

[0011] Furthermore, the smoothing process is executed as follows: Meeting the preset maximum speed Preset maximum acceleration and preset maximum jerk Under the constraints, the target displacement is decomposed into seven stages: uniform acceleration stage, uniform acceleration stage, uniform deceleration stage, uniform speed stage, uniform deceleration stage, uniform deceleration stage, and uniform acceleration stage. During the uniform acceleration phase, the acceleration is constant. acceleration according to It grows linearly with time, and the speed follows With a second increase over time, the displacement follows It increased three times over time; During the uniform acceleration phase, the jerk is zero, and the acceleration remains constant. speed according to It grows linearly over time, where The duration of the uniform acceleration phase; During the uniform deceleration and acceleration phase, the jerk remains constant. acceleration according to It decreases linearly to zero over time, where This is the end time of the uniform acceleration phase.

[0012] Furthermore, the calculation and correction method for the lateral deviation is as follows: Let the preset route start from the starting point and the end point The trajectory direction vector is defined as: ; From the starting point to the current position of the drone The vector is The projection of the vector onto the trajectory direction is ; The lateral deviation vector is ; The correction speed is ,in The preset proportional gain coefficient; the final speed command issued. The main speed output by the speed command curve With the corrected speed Superposition: ; Furthermore, the method for determining whether the actual displacement increment reaches the preset displacement threshold is as follows: For the starting point The trigger criterion for the initial translation operation, namely the actual displacement increment, is: ; in, Indicates the drone at a certain time The current real-time location, This indicates the preset displacement threshold; when the trigger criterion is met, a dwell command is issued and the trigger signal is sent to the external detection payload via a network protocol.

[0013] Furthermore, the execution method of the cyclical operation is as follows: After the translation operation is completed and the device remains stationary, the external detection payload receives the trigger signal to perform detection sampling; After the detection and sampling are completed, a vertical step displacement is performed. The step displacement is monitored in a closed loop by the elevation information in the positioning data to ensure that the physical spacing of each step displacement is consistent with the spacing of the boiler heating surface tube bank. After completing the stepping displacement, a new round of translation operation is started in the opposite direction to form a sweeping inspection track covering the boiler heating surface tube bank.

[0014] Furthermore, it also includes a safety limit logic judgment step: After each cycle of operation is completed, verify the value of the cumulative descent depth and the standard deviation of the localization solution; When the cumulative descent depth exceeds the preset depth red line or the standard deviation of the positioning solution continues to deteriorate beyond the preset safety threshold, the automated operation process is terminated, and the drone is forced to enter a zero-position hovering state.

[0015] Furthermore, it also includes environmental perception and modal initialization steps: After the drone arrives at the starting area of ​​the boiler heating surface, wait for the positioning data to stabilize, initialize the zero displacement point, and obtain a snapshot of the current position as the reference point for the physical displacement feedback closed loop. A status warning signal is sent to the external detection payload to establish a communication link.

[0016] The beneficial effects of this invention are as follows: This invention combines high-frequency feedback of positioning data with a lateral deviation correction algorithm, which not only controls the step displacement error within the centimeter level, but also effectively solves the problem of lateral drift caused by crosswinds or flow field interference during the translation of UAVs, ensuring the consistency between the inspection path and the preset baseline.

[0017] This invention performs a seven-segment smoothing process under acceleration constraints, replacing the traditional linear or step commands. It eliminates the inertial impact caused by sudden acceleration changes from the bottom layer, significantly reducing the attitude overshoot of the fuselage during frequent start-stop operations, and improving the imaging quality and data signal-to-noise ratio of the airborne detection payload.

[0018] The multi-modal fault-tolerant switching mechanism introduced in this invention solves the problem of positioning signal jumps caused by metal shielding in confined spaces. By maintaining position prediction after the positioning signal is lost through the shadow tracking mode, and by using Bézier curve convergence technology to achieve smooth calibration of the positioning reference after the signal is recovered, the flight safety risks caused by positioning jumps are effectively avoided.

[0019] This invention constructs a multi-level fault protection system from the algorithm layer to the task layer by real-time monitoring of the positioning signal quality slope and absolute depth limit, ensuring the safety of airborne equipment in confined spaces. Attached Figure Description

[0020] Figure 1 This is a hardware and software logic architecture diagram of the flight control system involved in the displacement closed-loop and anti-shake control method of the boiler heating surface inspection UAV according to an embodiment of the present invention. Figure 2 This is a mode switching logic diagram of a multimodal fault-tolerant state machine according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the collaborative working principle of S-shaped trajectory planning and track maintenance under higher-order dynamic constraints according to an embodiment of the present invention. Figure 4This is a flowchart of the displacement closed-loop and anti-shake control method of a boiler heating surface inspection drone according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates this application. It should be understood that the specific embodiments described in this specification are merely for explaining this application and are not intended to limit it.

[0022] The boiler heating surface inspection UAV displacement closed-loop and anti-shake control method of the present invention is implemented based on the UAV flight control system. (See also...) Figure 1 The flight control system adopts a layered integrated architecture, which includes three core components: the UAV flight platform, the RTK high-precision positioning system, and the airborne control algorithm module integrated into the UAV's onboard microprocessor.

[0023] The drone flight platform serves as the execution carrier, employing an industrial-grade drone that supports a virtual joystick control protocol, enabling it to execute three-dimensional velocity vector commands within a confined space. The flight control system receives smooth velocity commands from the onboard control algorithm module and executes control outputs at a frequency of 10Hz through four channels: pitch, roll, yaw, and throttle.

[0024] The RTK high-precision positioning system comprises two components: a ground reference station and an airborne rover station. The ground reference station is erected at a known coordinate point outside the boiler, while the airborne rover station is integrated onto an unmanned aerial vehicle (UAV). Utilizing carrier phase differential technology, this positioning system provides centimeter-level real-time 3D geographic coordinates and positioning solution status information for the entire control link. The positioning solution status includes three types: FIXED solution, FLOAT solution, and single-point solution. Simultaneously, it outputs the standard deviation (Std) of the positioning solution for signal quality assessment.

[0025] The airborne control algorithm module is the core software-defined component of this invention, running on the UAV's onboard microprocessor. This module listens to the high-frequency positioning data stream output by the RTK positioning system via a data bus, processes it using a multimodal algorithm, and then sends the smoothed control commands to the flight control system. The airborne control algorithm module comprises three mutually cooperating functional sub-modules: a multimodal positioning reliability monitoring sub-module, a high-order dynamic trajectory planning sub-module, and a displacement closed-loop task scheduling sub-module.

[0026] The multimodal positioning reliability monitoring submodule serves as the entry point for RTK data, responsible for real-time signal quality analysis and providing stable and continuous displacement feedback to subsequent submodules based on preset state machine logic. This submodule integrates an RTK quality evaluation model and a four-modal fault-tolerant state machine, assessing signal quality by analyzing the instantaneous value of the positioning solution's standard deviation and its slope within a time window.

[0027] The high-order dynamics trajectory planning submodule receives commands from the upper-level task, combines them with displacement feedback information, generates smooth virtual joystick commands limited by jerk, and sends them to the UAV flight platform. This submodule integrates two core algorithm components: an S-curve generator and a trajectory correction model.

[0028] The displacement closed-loop task scheduling submodule drives the operation cycle based on the physical displacement increment fed back from the positioning data, and sends asynchronous trigger commands to external detection payloads via network protocols. External detection payloads, representing independent detection devices such as UT probes, receive trigger commands from this submodule via HTTP or TCP network interfaces, enabling coordination between detection actions and the UAV's spatial position.

[0029] To provide a consistent description of the coordinate system involved in this invention, the world coordinate system W is defined as a fixed inertial reference system, using the Northeast-Eastern Sky (ENU) coordinate system with its z-axis pointing vertically upwards. The position of the UAV... and speed Represented in the world coordinate system. The body coordinate system B is defined as a non-inertial frame fixed to the UAV's center of gravity, with its x-axis pointing forward, y-axis pointing to the right, and z-axis pointing downward, conforming to the standard definition of an aircraft. The UAV's attitude is determined by the rotation matrix from the world coordinate system to the body coordinate system. or unit quaternion It is represented as follows. Its basic kinematic equations are expressed as: ; ; in, The vector of the body's angular velocity. This represents quaternion multiplication. The RTK high-precision positioning module directly provides the position in the world coordinate system. .

[0030] Example 1 This embodiment uses the inspection of the heating surface tube bank of a 660MW supercritical boiler in a thermal power plant as an application scenario. The water-cooled wall tube banks of this boiler are arranged around the furnace, with a spacing of 80mm between tube banks and a horizontal extension length of approximately 1.0m for each row. The inspection task requires a UAV carrying a UT ultrasonic thickness gauge to scan the surface of the tube banks row by row, collecting wall thickness data to assess the degree of wall thinning. The boiler's internal space is narrow, with shielding from metal components and interference from hot airflow, placing extremely high demands on the UAV's position control accuracy and dynamic stability.

[0031] The flight control system is configured as follows: The UAV flight platform is a quadcopter industrial UAV supporting virtual joystick control protocol, with a body weight of 2.5kg, a payload capacity of 1.0kg, and a flight speed range of 0 to 2.0m / s. The ground reference station of the RTK high-precision positioning system is set up at a known coordinate point 30m from the boiler inlet. The airborne rover station is integrated into the UAV carrier via top-mounted mounting, and the RTK positioning data update frequency is set to 10Hz. The airborne control algorithm module runs on an ARM-based embedded microprocessor with a main frequency of 1.8GHz and 4GB of memory. The external detection payload is a UT ultrasonic thickness gauge probe, which establishes a communication link with the airborne control algorithm module via an HTTP interface.

[0032] See Figures 1 to 4 The inspection control method in this embodiment includes the following steps: Step S1: Environmental perception and modal initialization.

[0033] After the UAV arrives at the starting area inside the furnace, the onboard control algorithm module activates the RTK positioning system and waits for the positioning solution to stabilize into a fixed solution. The initial value of the standard deviation Std of the positioning solution should be less than 0.02m, and the slope of change should be less than 0.005m / s. Once these conditions are met, the algorithm module takes a snapshot of the current position and records this position as the reference point for the physical displacement feedback closed loop, i.e., the zero displacement point. Simultaneously, the algorithm module sends an HTTP status warning signal to the UT ultrasonic thickness gauge probe, establishing a two-way communication link and confirming the readiness of the probe load.

[0034] In this embodiment, the coordinates of the reference point obtained after initialization are: The corresponding world coordinate system ENU coordinates are The RTK positioning system reports an initial positioning solution with a standard deviation of Std = 0.012m and a FIXED solution status. The multimodal fault-tolerant state machine is initialized to normal mode.

[0035] Step S2: Monitor the standard deviation of the positioning solution and its slope of change of the positioning data, and switch between multiple modes.

[0036] See Figure 2 The multimodal positioning reliability monitoring submodule in the airborne control algorithm module receives the positioning data stream output by the RTK positioning system in real time and extracts three pieces of information from it: three-dimensional spatial coordinates, the standard deviation of the positioning solution (Std), and the positioning solution status. This submodule synchronously monitors the instantaneous value and the slope of change of the standard deviation of the positioning solution (Std) and dynamically switches between the four modes accordingly.

[0037] The slope of change is calculated as follows: over a length of... Within a time window, linear regression is performed on the continuously collected Std numerical sequence, and the slope of the fitted line is the slope of change. In this embodiment, the time window... Set to 2.0 seconds, corresponding to 20 RTK data samples.

[0038] The operating logic of the normal mode (NORMAL) is as follows: when the instantaneous value of the standard deviation (Std) of the localization solution is lower than the preset standard deviation threshold of 0.03m, and the slope of change is lower than the preset slope threshold of 0.01m / s, the system maintains the normal mode. In the normal mode, the system directly uses RTK real-time localization data to construct a physical displacement feedback closed loop, and the flight speed is executed according to the mission setting value.

[0039] The DEGRADED mode operates as follows: When the instantaneous value of the localization solution standard deviation (Std) increases by more than 0.03m but less than 0.08m, or the slope of change exceeds the preset slope threshold of 0.01m / s but the localization signal is not completely lost, the system switches from normal mode to degraded mode. In degraded mode, the system reduces the UAV's flight speed from the current set value to 50% of its original speed, while continuing to use the current RTK localization data. When the localization solution standard deviation falls back below the preset standard deviation threshold and the slope of change returns to normal, the system returns from degraded mode to normal mode.

[0040] The operating logic of the shadow tracking mode is as follows: when the RTK positioning signal is completely lost, that is, when the positioning solution state jumps from the FIXED solution or FLOAT solution to the unsolvable state, the system switches from the normal mode or degraded mode to the shadow tracking mode. At the instant the signal is lost, the system caches the current position. and speed During the shadow tracking mode operation, the system uses the cached velocity vector to integrate the position and calculate the virtual position: ; in, Indicates at time The calculated virtual location, This indicates the actual location at the moment the positioning signal was lost. The velocity vector representing the moment the positioning signal was lost. Indicates the time when the signal was lost. This indicates the current moment. This virtual position serves as the input to the physical displacement feedback loop, maintaining the normal operation of subsequent sub-modules. In this embodiment, the maximum duration of the shadow tracking mode is set to 5.0 seconds; exceeding this duration will forcibly terminate the automated workflow.

[0041] The operating logic of the convergence mode (CONVERGENCE) is as follows: when the RTK positioning signal recovers from the lost state, i.e., the positioning solution state changes back to a FIXED or FLOAT solution, the system switches from the shadow tracking mode to the convergence mode. At this time, the virtual position calculated by the shadow tracking mode... With the newly acquired real location There are discrepancies between them. To avoid sudden position jumps causing severe fuselage shaking, the system uses a third-order Bézier curve for a smooth transition. (Bézier curve path) Defined by four control points, parameters : ; in, This is the starting point, i.e., the virtual position at the end of the shadow tracking mode; The endpoint is the actual location after signal recovery. Intermediate control points are set as follows: ; ; in, This represents the velocity vector before the signal was lost. This represents the target velocity vector after convergence. and This is a preset trajectory shape adjustment coefficient. In this embodiment, Set to 0.3, Set to 0.1, Set to zero vector. The duration of the convergence process is set to 1.0 second, parameters... The system linearly changes from 0 to 1. After convergence, the system determines whether to regress to the normal mode or the degraded mode based on the standard deviation of the recovered localized solution.

[0042] In the actual operation of this embodiment, the RTK signal was briefly blocked at the 37-second mark when the UAV approached the metal support beam of the furnace. The system switched from the normal mode to the shadow tracking mode via the degraded mode, and the signal was restored after about 2.3 seconds. The system completed the position calibration within 1.0 second through the convergence mode. The UAV's attitude remained stable throughout the process, and no position jumps occurred.

[0043] Step S3: Perform a seven-segment smoothing process on the target displacement under jerk constraints to generate a velocity command curve.

[0044] See Figure 3 On the left, the high-order dynamics trajectory planning submodule in the airborne control algorithm module receives a task command to translate horizontally by 1.0m. This submodule performs a seven-segment smoothing process on the target displacement of 1.0m under jerk constraints, while meeting the preset maximum speed. Preset maximum acceleration and preset maximum jerk Generate a continuous speed command curve under the constraints.

[0045] In this embodiment, a maximum speed is preset. Set to 0.5 m / s, preset maximum acceleration. Set to 0.3 m / s², preset maximum jerk. Set to 0.5 m / s³.

[0046] The seven-stage smoothing process decomposes the target displacement into the following seven stages: The time range of the uniform acceleration phase is ,in Seconds. During this phase, the jerk is constant. m / s³, acceleration according to It grows linearly with time, and the speed follows With a second increase over time, the displacement follows It increased three times over time. At a time of seconds, the acceleration reaches m / s², speed reaches m / s.

[0047] The time range of the uniform acceleration phase is During this phase, the jerk is zero, and the acceleration remains constant. m / s², velocity according to It increases linearly over time. This phase continues until the speed approaches the preset maximum speed. Subtract the difference in velocity increment during the uniform deceleration acceleration phase.

[0048] The time range of the uniform deceleration acceleration phase is During this phase, the jerk is constant. m / s³, acceleration according to It decreases linearly to zero over time. At the end of this phase, the speed smoothly reaches the preset maximum speed. m / s.

[0049] During the uniform velocity phase, both jerk and acceleration are zero, and the velocity remains constant. m / s, displacement increases linearly.

[0050] The subsequent deceleration process includes a uniform deceleration-acceleration phase, a uniform deceleration phase, and a uniform acceleration-acceleration-deceleration phase, the mathematical form of which is symmetrical to the acceleration phase. By solving the above piecewise function, the system generates a smooth speed command curve. The curve remains continuous across all four dimensions: jerk, acceleration, velocity, and displacement. This velocity command curve is sent to the virtual joystick interface of the UAV flight control system.

[0051] In this embodiment, the total time to complete a 1.0m horizontal translation is approximately 3.2 seconds, and the maximum rate of change of acceleration on the velocity command curve never exceeds [a certain value]. m / s³, eliminating the inertial impact caused by traditional step commands.

[0052] Step S4: Calculate the lateral deviation and add the corrected speed during the execution of the speed command curve.

[0053] See Figure 3 On the right, as the speed command curve drives the UAV to move horizontally, the trajectory correction model of the high-order dynamics trajectory planning submodule operates synchronously. This model calculates the lateral deviation of the UAV's current position relative to the preset route in real time based on a vector projection algorithm, and generates a corrected velocity in the opposite direction to the lateral deviation, which is then superimposed on the speed command curve to ensure that the UAV always moves along a straight trajectory.

[0054] The preset route starts from the origin. and the end point Definition. In this embodiment, the starting point is the starting coordinate of the current inspection row. m, the endpoint is the target coordinates after horizontal translation of 1.0m. m.

[0055] The trajectory direction vector is calculated as follows: ; During each control cycle, the system obtains the current position of the UAV from the RTK positioning data. Calculate the vector from the starting point to the current position. The projection of this vector onto the track direction is: ; Lateral deviation vector for and The difference: ; The correction speed is generated according to the following formula: ; in This is a preset proportional gain coefficient. In this embodiment, Set to 2.0. The final speed command sent to the flight control system is the main speed output through a seven-segment smoothing process. With correction speed Superposition: ; In the actual operation of this embodiment, when the drone is in At a certain second, due to interference from the lateral airflow inside the furnace, a lateral shift occurs, and the current position is... When m, the lateral deviation vector is m, corrected speed is m / s. After this corrected velocity is superimposed on the main velocity, the system corrects the lateral deviation to within 0.002m within approximately 0.5 seconds.

[0056] Step S5: Use the actual displacement increment as the criterion for the operation state transition and trigger the detection load sampling.

[0057] The displacement closed-loop task scheduling submodule in the airborne control algorithm module continuously monitors the actual displacement increment in the physical displacement feedback closed loop during the translation operation. This submodule uses the actual displacement increment as the sole criterion for job state transitions, rather than the traditional time-based timing method.

[0058] For the starting point The trigger criterion for the initial translation operation, based on the actual displacement increment, is: ; in, Indicates the drone at a certain time The current real-time location, This represents a preset displacement threshold. In this embodiment, Set to 1.0m.

[0059] When the trigger criterion is met, the displacement closed-loop task scheduling submodule immediately performs two operations: First, it sends a dwell command to the flight control system. This dwell command is also processed through a seven-segment smoothing process to generate a deceleration curve, so that the UAV can smoothly decelerate to zero speed and hover. Second, it sends a trigger signal to the UT ultrasonic thickness measurement probe through an HTTP POST network request, instructing the probe to start wall thickness measurement sampling.

[0060] In this embodiment, when the drone is When the actual displacement increment reaches 1.0m per second, the system issues a hovering command. The UAV decelerates and enters a stable hovering state within approximately 0.8 seconds. Simultaneously, the system sends a trigger frame to the UT probe via HTTP POST. Upon receiving the trigger signal, the probe completes a wall thickness data acquisition within 200 milliseconds. The acquisition result is returned to the onboard control algorithm module via HTTP response and recorded in the flight log.

[0061] Step S6: Perform vertical step displacement and cyclic operation.

[0062] See Figure 3After the UT probe completes its detection and sampling, the displacement closed-loop task scheduling submodule automatically initiates a vertical step displacement command. This step displacement also undergoes a seven-segment smoothing process under jerk constraints to generate a velocity command curve, and is subject to closed-loop monitoring of elevation information from the RTK positioning data.

[0063] In this embodiment, the vertical step distance is set to 0.08m, corresponding to the spacing between the boiler water-cooled wall tubes. The preset maximum speed is set to 0.1m / s, the preset maximum acceleration is set to 0.1m / s², and the preset maximum jerk is set to 0.2m / s³. The trigger criterion for the step displacement is the same as that for the horizontal translation, that is, when the actual vertical displacement increment reaches 0.08m, the system issues a dwell command.

[0064] After the vertical step is completed, the system starts a new round of translation in the opposite direction of the horizontal direction. The starting point of the new round is the position after the current vertical step is completed, and the ending point is the target coordinates after translating 1.0m in the opposite direction. This "horizontal translation-stable dwell-vertical step" cycle constitutes a sweeping inspection track covering the boiler's heating surface tube bank.

[0065] In this embodiment, a single vertical step takes approximately 1.8 seconds, and the step displacement error is controlled within 0.003m. Completing a full cycle of horizontal translation plus vertical step takes approximately 5.8 seconds, including 3.2 seconds of horizontal translation, 1.0 second for dwell and detection sampling, and 1.6 seconds for vertical step.

[0066] Step S7: Safety limit logic judgment.

[0067] After each cycle of operation is completed, the airborne control algorithm module automatically performs safety limit logic checks. The checks include two dimensions: cumulative descent depth verification and positioning quality verification.

[0068] The method for verifying the cumulative descent depth is as follows: the system calculates the cumulative vertical descent distance of the UAV since the start of the operation. When this distance exceeds the preset depth red line, the system terminates the automated operation process. In this embodiment, the preset depth red line is set to 3.0m, corresponding to the inspection depth of approximately 37 rows of pipes.

[0069] The method for determining positioning quality is as follows: the system checks the statistical characteristics of the standard deviation of the RTK positioning solution in the most recent cycle. When the mean of the standard deviation of the positioning solution continuously exceeds a preset safety threshold, the system determines that the positioning quality has deteriorated and terminates the automated operation process. In this embodiment, the preset safety threshold is set to 0.05m.

[0070] When the safety limit logic determines that the termination condition has been triggered, the system immediately shuts down the automated operation process, forcing the drone into a zero-position hovering state. In the zero-position hovering state, the drone maintains its current position and waits for the operator to manually take over or remotely issue a return-to-home command.

[0071] In this embodiment, the UAV performed 35 cycles of inspection, covering the wall thickness detection of 35 rows of pipes. The total operation time was approximately 203 seconds. The average displacement error for horizontal translation was 0.008m, and the average displacement error for vertical stepping was 0.003m. Throughout the process, the multimodal fault-tolerant state machine triggered degradation mode switching 4 times and shadow tracking mode switching once. All switching processes were completed within the safety limits, and no flight safety incidents occurred.

[0072] Example 2 This embodiment uses the inspection of the heating surface of a 300MW circulating fluidized bed boiler as an application scenario. Compared with the supercritical boiler in Embodiment 1, the furnace space of the circulating fluidized bed boiler is smaller, and the inner wall has a large area of ​​refractory material lining. The metal structural components are also more densely packed, resulting in a higher probability of RTK signal shielding. The tube bank spacing is 60mm, and the horizontal extension length of a single tube bank is approximately 0.8m.

[0073] The flight control system is configured as follows: The UAV flight platform is a small quadcopter industrial UAV with a body weight of 1.8 kg and a payload capacity of 0.6 kg. The RTK high-precision positioning system is configured the same as in Example 1, with the RTK positioning data update frequency set to 10 Hz. The airborne control algorithm module runs on the same ARM architecture embedded microprocessor. The external detection payload is a miniaturized UT ultrasonic thickness gauge probe.

[0074] See Figures 1 to 4 The inspection control method in this embodiment includes the following steps.

[0075] Step S1: Environmental perception and modal initialization.

[0076] After the UAV arrives at the starting area inside the circulating fluidized bed boiler furnace, the onboard control algorithm module activates the RTK positioning system. Due to the confined furnace space and dense metal components, the initial settling time of the RTK signal is longer than in Example 1. The system waits until the standard deviation of the positioning solution (Std) drops below 0.025m and the slope of change is below 0.008m / s before acquiring a snapshot of the current position as the zero point of displacement. Simultaneously, the algorithm module sends an HTTP status warning signal to the UT probe to establish a communication link.

[0077] In this embodiment, the RTK signal wait time from activation to stabilization is 15 seconds, and the initial localization solution standard deviation is Std = 0.018m. The multimodal fault-tolerant state machine is initialized to normal mode.

[0078] Step S2: Monitor the standard deviation of the positioning solution and its slope of change of the positioning data, and switch between multiple modes.

[0079] The monitoring logic of the multimodal positioning reliability monitoring submodule is the same as in Example 1, but considering the special environment of the circulating fluidized bed boiler, some threshold parameters are adjusted as follows: the preset standard deviation threshold is adjusted to 0.035m, the preset slope threshold is adjusted to 0.012m / s, and the maximum duration of the shadow tracking mode is adjusted to 3.0 seconds.

[0080] The slope of change is calculated in the same way as in Example 1, and linear regression fitting is performed on the Std numerical sequence within a time window of 2.0 seconds.

[0081] The operating logic of the normal mode is as follows: when the instantaneous value of the standard deviation of the localization solution Std is less than 0.035m and the slope of change is less than 0.012m / s, the system maintains the normal mode and directly uses RTK real-time localization data to construct a physical displacement feedback closed loop.

[0082] The DEGRADED mode operates as follows: when the instantaneous value of the standard deviation of the positioning solution (Std) increases by more than 0.035m but not more than 0.08m, or the slope of change exceeds 0.012m / s but the positioning signal is not completely lost, the system switches to the degraded mode. The flight speed is reduced to 40% of the original speed.

[0083] The operating logic of the SHADOW tracking mode is as follows: when the RTK positioning signal is completely lost, the system switches to the SHADOW tracking mode. The virtual position calculation formula is the same as in Example 1: ; In this embodiment, due to the dense metal components in the furnace, the system triggered the shadow tracking mode 3 times in 35 cycles, with each duration being 1.8 seconds, 2.1 seconds, and 1.5 seconds, respectively.

[0084] The operating logic of the convergence mode CONVERGENCE is as follows: when the RTK signal recovers, the system constructs a Bézier curve to guide the UAV through a smooth transition. The expression for the Bézier curve path is the same as in Example 1: ; In this embodiment, the trajectory shape adjustment coefficient Set to 0.2, Set it to 0.05, and the convergence duration to 0.8 seconds.

[0085] Step S3: Perform a seven-segment smoothing process on the target displacement under jerk constraints to generate a velocity command curve.

[0086] The high-order dynamics trajectory planning submodule receives a task command for a horizontal translation of 0.8m. Considering the limited furnace space in a circulating fluidized bed boiler, the seven-segment smoothing parameters in this embodiment are adjusted as follows: preset maximum speed. Set to 0.3 m / s², preset maximum acceleration. Set to 0.2 m / s², preset maximum jerk. Set to 0.4 m / s³.

[0087] The duration of the uniform acceleration phase is Seconds. At a time of seconds, the acceleration reaches m / s², speed reaches m / s.

[0088] During the uniform acceleration phase, the jerk is zero, and the acceleration remains constant. m / s², velocity according to It increases linearly with time. During the uniform deceleration and acceleration phase, the jerk remains constant. m / s³, acceleration according to The velocity decreases linearly to zero over time. The calculation logic for the subsequent constant velocity phase and deceleration process is the same as in Example 1. The total time to complete the 0.8m horizontal translation is approximately 3.8 seconds.

[0089] Step S4: Calculate the lateral deviation and add the corrected speed during the execution of the speed command curve.

[0090] The operational logic of the trajectory correction model is consistent with that of Example 1. The preset route starts from the origin. and the end point The horizontal distance between the two is defined as 0.8m. Proportional gain coefficient. The value is set to 2.5, which is an improvement over Example 1 to accommodate smaller working environments.

[0091] The calculation formulas for the lateral deviation vector, correction velocity, and final velocity command are the same as in Example 1: ; ; ; In this embodiment, the maximum lateral deviation during the horizontal translation process is 0.018m, which is controlled within 0.005m after correction.

[0092] Step S5: Use the actual displacement increment as the criterion for the operation state transition and trigger the detection load sampling.

[0093] The triggering criteria for the displacement closed-loop task scheduling submodule are the same as in Example 1: ; In this embodiment, The setting is 0.8m. When the trigger criterion is met, the system issues a dwell command and sends a trigger signal to the UT probe via HTTP POST. The probe completes wall thickness data acquisition upon receiving the trigger signal.

[0094] Step S6: Perform vertical step displacement and cyclic operation.

[0095] The vertical stepping distance is set to 0.06m, corresponding to the tube bank spacing of a circulating fluidized bed boiler. The preset maximum speed in the stepping direction is set to 0.08m / s, the preset maximum acceleration is set to 0.08m / s², and the preset maximum jerk is set to 0.15m / s³.

[0096] After the vertical stepping is completed, the system starts a new round of translational operation in the opposite direction, forming a sweeping inspection track covering the boiler's heating surface tube bank. A single vertical stepping takes about 1.6 seconds, and the stepping displacement error is controlled within 0.004m.

[0097] Step S7: Safety limit logic judgment.

[0098] After each cycle of operation is completed, the system performs a safety limit check. In this embodiment, the preset depth red line is set to 2.0m, and the preset safety threshold is set to 0.06m.

[0099] In this embodiment, the UAV performed 32 cycles of inspection, covering the wall thickness detection of 32 rows of pipes. The total operation time was approximately 192 seconds. The average displacement error for horizontal translation was 0.010m, and the average displacement error for vertical stepping was 0.004m. The multimodal fault-tolerant state machine triggered degradation mode switching 6 times and shadow tracking mode switching 3 times.

[0100] Example 3 This embodiment uses the inspection of the heating surface of a large gas turbine waste heat boiler as an application scenario. The furnace space of the waste heat boiler is more open than the previous two embodiments, and the tube arrangement is relatively regular. However, there is a strong updraft caused by the high-temperature waste heat flow, which poses a greater challenge to the dynamic stability of the UAV. The tube spacing is 100mm, and the horizontal extension length of a single tube row is approximately 1.5m.

[0101] The flight control system is configured as follows: The UAV flight platform is a high-payload hexacopter industrial UAV with a body weight of 4.2 kg and a payload capacity of 2.0 kg. The RTK positioning data update frequency is set to 20 Hz to provide higher temporal resolution position feedback. The external detection payload is a multi-channel UT ultrasonic thickness gauge array.

[0102] See Figures 1 to 4The inspection control method in this embodiment includes the following steps: Step S1: Environmental perception and modal initialization.

[0103] After the drone arrives at the initial working area inside the waste heat boiler furnace, the onboard control algorithm module activates the RTK positioning system. Due to the open space of the furnace and minimal metal shielding, the RTK signal stabilizes quickly. The system waits for the standard deviation of the positioning solution (Std) to drop below 0.015m before acquiring the zero displacement point. The algorithm module then establishes a communication link by sending an HTTP status warning signal to the multi-channel UT probe array.

[0104] In this embodiment, the initial standard deviation of the localization solution is Std = 0.010m. The multimodal fault-tolerant state machine is initialized to normal mode.

[0105] Step S2: Monitor the standard deviation of the positioning solution and its slope of change of the positioning data, and switch between multiple modes.

[0106] The monitoring logic of the multimodal positioning reliability monitoring submodule is consistent with that of the aforementioned embodiment. Considering the environmental characteristics of the waste heat boiler, the preset standard deviation threshold is set to 0.025m, the preset slope threshold is set to 0.008m / s, and the maximum duration of the shadow tracking mode is set to 8.0 seconds.

[0107] The switching logic between normal mode, degraded mode, shadow tracking mode, and convergence mode is the same as in Example 1. The virtual position estimation formula is: ; The path expression for a Bézier curve is: ; In this embodiment, the trajectory shape adjustment coefficient Set to 0.4, Set it to 0.15, and the convergence duration to 1.2 seconds.

[0108] Step S3: Perform a seven-segment smoothing process on the target displacement under jerk constraints to generate a velocity command curve.

[0109] The advanced dynamics trajectory planning submodule receives a task command to translate horizontally by 1.5m. Due to the relatively open space of the waste heat boiler, the seven-segment smoothing parameters in this embodiment are set as follows: preset maximum speed. The preset maximum acceleration is 0.8 m / s². The preset maximum jerk is 0.4 m / s². It is 0.6 m / s³.

[0110] The duration of the uniform acceleration phase is Seconds. At that moment, the acceleration reached m / s², speed reaches m / s.

[0111] During the uniform acceleration phase, the jerk is zero, and the acceleration remains constant. m / s², velocity according to It increases linearly with time. During the uniform deceleration and acceleration phase, the jerk remains constant. m / s³, acceleration according to The speed decreases linearly to zero over time. The calculation logic for subsequent stages is the same as in the aforementioned embodiment. The total time to complete a 1.5m horizontal translation is approximately 3.5 seconds. In strong updraft conditions, the seven-stage smoothing process effectively suppresses attitude overshoot caused by sudden acceleration changes, keeping the maximum pitch angle change during start-up and shutdown within 2 degrees.

[0112] Step S4: Calculate the lateral deviation and add the corrected speed during the execution of the speed command curve.

[0113] The operational logic of the trajectory correction model is consistent with the aforementioned embodiment. The preset horizontal distance of the flight path is 1.5m, and the proportional gain coefficient... Set to 1.5. The calculation formulas for the lateral deviation vector, corrected speed, and final speed command are the same as in the previous embodiment: ; ; ; In this embodiment, due to the influence of strong updrafts, the lateral deviation during horizontal translation is larger than that in the previous embodiment, with the maximum lateral deviation reaching 0.025m. After correction by the trajectory correction model, it is controlled within 0.008m.

[0114] Step S5: Use the actual displacement increment as the criterion for the operation state transition and trigger the detection load sampling.

[0115] The trigger criterion for the displacement closed-loop task scheduling submodule is: ; In this embodiment, The distance is set to 1.5m. When the trigger criterion is met, the system issues a dwell command and sends a trigger signal to the multi-channel UT probe array via HTTP POST. Due to the use of a multi-channel probe array, a single sampling can simultaneously cover 3 measurement points, improving sampling efficiency by 3 times compared to a single-channel probe.

[0116] Step S6: Perform vertical step displacement and cyclic operation.

[0117] The vertical step distance is set to 0.10m, corresponding to the tube spacing of the waste heat boiler. The preset maximum speed in the stepping direction is set to 0.15m / s, the preset maximum acceleration is set to 0.12m / s², and the preset maximum jerk is set to 0.25m / s³.

[0118] Each vertical step takes approximately 1.5 seconds, with the step displacement error controlled within 0.005m. The system then initiates a new round of translational operations in the opposite direction, forming a sweeping inspection track covering the boiler's heating surface tube banks.

[0119] Step S7: Safety limit logic judgment.

[0120] In this embodiment, the preset depth red line is set to 5.0m, and the preset safety threshold is set to 0.04m.

[0121] During the entire inspection operation, the UAV performed 48 cycles, covering the wall thickness detection of 48 rows of pipes. The total operation time was approximately 288 seconds. The average displacement error for horizontal translation was 0.012m, and the average displacement error for vertical stepping was 0.005m. The multimodal fault-tolerant state machine triggered a degradation mode switch twice, but did not trigger the shadow tracking mode.

[0122] In summary, the embodiments of the present invention have at least the following technical effects: This invention constructs a physical displacement feedback closed loop based on positioning data and uses the actual displacement increment instead of time as the criterion for switching operating states, achieving centimeter-level step displacement control accuracy in various boiler types and operating environments.

[0123] This invention performs a seven-segment smoothing process under jerk constraints, eliminating the inertial impact caused by step commands, and can effectively suppress fuselage attitude overshoot even in strong airflow interference environments.

[0124] This invention uses a multimodal fault-tolerant state machine to dynamically switch between four modes: normal, degraded, trail tracking, and convergence, ensuring the continuity and reliability of displacement feedback in environments where positioning signals are shielded or undergo abrupt changes.

[0125] This invention constructs a multi-level fault fuse protection system through safety limit logic discrimination, ensuring the safe operation of airborne equipment in confined spaces.

[0126] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A closed-loop displacement and anti-shake control method for a boiler heating surface inspection drone, characterized in that, include: A physical displacement feedback closed loop is constructed based on real-time acquired UAV 3D spatial positioning data, so that the standard deviation of the positioning solution and its slope of change in the positioning data switch between multiple preset fault-tolerant modes to maintain the continuity of displacement feedback in the physical displacement feedback closed loop. Under jerk constraints, the target displacement is smoothed to generate a continuous velocity command curve. During the execution of the velocity command curve, the lateral deviation of the UAV's current position relative to the preset flight path is calculated based on the vector projection algorithm. A corrected velocity opposite to the direction of the lateral deviation is generated and superimposed on the velocity command curve to maintain a straight inspection track. Using the actual displacement increment in the physical displacement feedback closed loop as the criterion for switching operating states, the UAV is driven to perform cyclic operations of translation, dwelling and stepping along the boiler heating surface tube bank. When the actual displacement increment reaches a preset displacement threshold, a trigger signal is sent to the external detection load to achieve alignment between the detection sampling and the spatial position.

2. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 1, characterized in that, The preset fault-tolerant modes include normal mode, degradation mode, shadow tracking mode, and convergence mode; the method for switching between the preset fault-tolerant modes is as follows: When the instantaneous value of the standard deviation of the positioning solution is lower than the preset standard deviation threshold and the slope of change is lower than the preset slope threshold, the normal mode is maintained, and the positioning data is directly used to construct the physical displacement feedback closed loop. When the instantaneous value of the standard deviation of the positioning solution increases or the slope of change exceeds the preset slope threshold but the positioning signal is not completely lost, switch to the degraded mode, reduce the flight speed of the UAV and continue to use the positioning data; When the positioning signal is completely lost, switch to the shadow tracking mode and use the historical velocity vector buffered before the loss to calculate the virtual position in order to maintain the logical continuity of displacement feedback. When the positioning signal is recovered from the lost state, the system switches to the convergence mode and constructs a Bézier curve to guide the UAV to smoothly transition from the virtual position to the recovered real position.

3. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 2, characterized in that, The method for calculating the virtual position in the shadow tracking mode is as follows: set up The location signal was lost at the time of loss, and the position and velocity at the time of loss were respectively... and ,exist At that moment, the virtual location Estimate using the following formula: ； in, Indicates at time The calculated virtual location, This indicates the actual location at the moment the positioning signal was lost. The velocity vector representing the moment the positioning signal was lost. Indicates the current moment.

4. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 3, characterized in that, The Bézier curve in the convergence mode is constructed as follows: Let the virtual location be The restored true location is The Bézier curve path Defined by four control points, parameters : ； in, Starting point End point; intermediate control point , ; This represents the velocity vector before the positioning signal was lost. This represents the target velocity vector after convergence. and This is the preset trajectory shape adjustment coefficient.

5. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 1, characterized in that, The smoothing process is executed as follows: Meeting the preset maximum speed Preset maximum acceleration and preset maximum jerk Under the constraints, the target displacement is decomposed into seven stages: uniform acceleration stage, uniform acceleration stage, uniform deceleration stage, uniform speed stage, uniform deceleration stage, uniform deceleration stage, and uniform acceleration stage. During the uniform acceleration phase, the acceleration is constant. acceleration according to It grows linearly with time, and the speed follows With a second increase over time, the displacement follows It increased three times over time; During the uniform acceleration phase, the jerk is zero, and the acceleration remains constant. speed according to It grows linearly over time, where The duration of the uniform acceleration phase; During the uniform deceleration and acceleration phase, the jerk remains constant. acceleration according to It decreases linearly to zero over time, where This is the end time of the uniform acceleration phase.

6. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 1, characterized in that, The calculation and correction method for the lateral deviation is as follows: Let the preset route start from the starting point and the end point Definition: Track direction vector for: ； From the starting point to the current position of the drone The vector is The projection of the vector onto the trajectory direction is ; The lateral deviation vector is ; The correction speed is ,in The preset proportional gain coefficient; the final speed command issued. The main speed output by the speed command curve With the corrected speed Superposition: 。 7. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 1, characterized in that, The method for determining whether the actual displacement increment reaches the preset displacement threshold is as follows: For starting point The trigger criterion for the initial translation operation, namely the actual displacement increment, is: ； in, Indicates the drone at a certain time The current real-time location, This indicates the preset displacement threshold; when the trigger criterion is met, a dwell command is issued and the trigger signal is sent to the external detection payload via a network protocol.

8. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 7, characterized in that, The execution method of the cyclical operation is as follows: After the translation operation is completed and the device remains stationary, the external detection payload receives the trigger signal to perform detection sampling; After the detection and sampling are completed, a vertical step displacement is performed. The step displacement is monitored in a closed loop by the elevation information in the positioning data to ensure that the physical spacing of each step displacement is consistent with the spacing of the boiler heating surface tube bank. After completing the stepping displacement, a new round of translation operation is started in the opposite direction to form a sweeping inspection track covering the boiler heating surface tube bank.

9. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAVs according to claim 1, characterized in that, It also includes a safety limit logic judgment step: After each cycle of operation is completed, verify the value of the cumulative descent depth and the standard deviation of the localization solution; When the cumulative descent depth exceeds the preset depth red line or the standard deviation of the positioning solution continues to deteriorate beyond the preset safety threshold, the automated operation process is terminated, and the drone is forced to enter a zero-position hovering state.

10. The displacement closed-loop and anti-shake control method for boiler heating surface inspection UAV according to any one of claims 1 to 9, characterized in that, It also includes environmental perception and modal initialization steps: After the drone arrives at the starting area of ​​the boiler heating surface, wait for the positioning data to stabilize, initialize the zero displacement point, and obtain a snapshot of the current position as the reference point for the physical displacement feedback closed loop. A status warning signal is sent to the external detection payload to establish a communication link.

Images