Degradation-aware control method and system for large-diameter pipeline spraying robot mechanical arm

Abstract

The application provides a large-diameter pipeline spraying robot mechanical arm control method and system with degradation awareness. The method acquires an original control state of a spraying mechanical arm and an online monitoring state including ranging echo quality, and generates a synchronization flag based on the acquisition time stamps of the states; generates a basic end target linear velocity, a first matrix and a safe prediction distance based on the original control state; generates a process residual fingerprint based on the original control state, the online monitoring state and the synchronization flag, determines a fault and disturbance credibility, a degradation level, and generates a reconstructed control parameter accordingly; and determines a target joint angular velocity based on the basic end target linear velocity, the first matrix, the safe prediction distance and the reconstructed control parameter. Thus, the fault detection result can directly enter the mechanical arm motion control link.

Description

Technical Field

[0001] This application relates to the field of control technology for robotic arms of spraying robots, and in particular to a control method and system for a large-diameter pipe spraying robot robotic arm with degradation sensing. Background Technology

[0002] Large-diameter pipe interior spraying typically relies on mobile robots carrying multi-jointed spraying arms. When the robot moves inside the pipe, the distance measuring sensor is susceptible to glare, dust, and changes in pipe wall curvature. The joint servo loops may also exhibit feedback anomalies due to hose traction, load variations, or motor degradation. Existing robotic arm control methods often limit the distance or reduce speed after abrupt changes, making it difficult to distinguish between distance measurement faults, servo degradation, and actual pipe wall variations. This results in fault detection results not being promptly translated into robotic arm control quantities, affecting the stability of the distance between the nozzle and the pipe wall. Summary of the Invention

[0003] To address the shortcomings of existing technologies in large-diameter pipe spraying robots, which struggle to promptly convert fault detection results into robotic arm control quantities, thus affecting the stability of the distance between the nozzle and the pipe wall, this application provides a degradation-aware robotic arm control method and system for large-diameter pipe spraying robots. This method generates process residual fingerprints to determine the reliability and degradation level of faults and disturbances, and reconstructs the end-effector velocity, joint damping, and control mode accordingly, enabling fault detection results to directly participate in the robotic arm's motion control. Specifically, this application provides the following technical solutions: In a first aspect, this application provides a degradation-aware control method for a large-diameter pipe spraying robot arm, comprising: acquiring the original control state and online monitoring state of the spraying robot arm, and generating a synchronization flag based on the acquisition timestamps of each state; performing distance prediction and amplitude limiting control based on the original control state, generating a first matrix for establishing a mapping between end-effector linear velocity and joint angular velocity, a safe predicted distance, and a basic end-effector target linear velocity; generating a process residual fingerprint based on the original control state, the online monitoring state, and the synchronization flag, wherein the process residual fingerprint includes at least distance residual, echo quality residual, kinematic closure residual, servo residual, load residual, pressure disturbance residual, and the synchronization flag; determining the fault and disturbance confidence level and degradation level based on the process residual fingerprint; generating reconstructed control parameters based on the fault and disturbance confidence level and the degradation level; determining the target joint angular velocity based on the basic end-effector target linear velocity, the first matrix, the safe predicted distance, and the reconstructed control parameters, and controlling the movement of the spraying robot arm based on the target joint angular velocity.

[0004] Optionally, before acquiring the original control state and online monitoring state of the spraying robot arm, the method further includes: performing low-speed baseline operation within the large-diameter pipe; collecting ranging readings, ranging echo quality, joint feedback state, drive load state, and nozzle pressure state during the baseline operation; and determining a threshold set, baseline load state, and baseline pressure state for generating the process residual fingerprint based on the ranging readings, ranging echo quality, joint feedback state, drive load state, and nozzle pressure state.

[0005] Optionally, the process residual fingerprint includes distance residual, echo quality residual, kinematic closure residual, servo residual, and load residual, and also includes pressure disturbance residual and synchronization flag; the distance residual is obtained based on the current actual distance and the predicted distance generated by the distance prediction limiting control; the echo quality residual is obtained based on the ranging echo quality; the kinematic closure residual is obtained based on the current actual distance, historical actual distance, first matrix, and joint speed command of the previous cycle; the servo residual is obtained based on the joint speed command of the previous cycle and joint feedback speed; the load residual is obtained based on the drive load state and the reference load state; the pressure disturbance residual is obtained based on the nozzle pressure state and the reference pressure state; the synchronization flag is used to characterize whether each monitoring state meets the synchronization conditions of the same control cycle.

[0006] Optionally, determining the reliability of faults and disturbances and the degradation level based on the process residual fingerprint includes: performing window statistics on the process residual fingerprints over multiple consecutive control cycles; determining the reliability of ranging link faults, pipe wall real disturbances, and hose disturbances based on the window statistics results; determining the reliability of joint servo link degradation based on the window statistics results; and determining the degradation level based on the reliability interval of each reliability level.

[0007] Optionally, the reconfiguration control parameters include ranging weight, tangential velocity coefficient, joint space damping length, and control mode; the ranging weight is used to adjust the influence of the ranging closed loop on the normal control quantity; the tangential velocity coefficient is used to adjust the spraying propulsion speed; the joint space damping length is used to construct the joint space damping matrix; the control mode includes at least a normal mode, a guard mode, a reconfiguration mode, and a safety hold mode.

[0008] Secondly, this application provides a distance prediction and control method for a large-diameter pipe spraying robot arm, comprising: acquiring the current joint angle and current actual distance of the spraying robot arm, and reading the historical actual distance, a first delay time, a target distance, a target tangential velocity, and a pre-calibrated pipe center axis; determining the current end position of the spraying robot arm and a first matrix for establishing a mapping relationship between the end linear velocity and the joint angular velocity based on the current joint angle; determining the predicted distance based on the current actual distance, the historical actual distance, and the first delay time; determining a first displacement boundary based on the maximum withstand acceleration of the end of the spraying robot arm and the first delay time, and limiting the predicted distance according to the first displacement boundary to generate a safe predicted distance; generating the end target linear velocity of the spraying robot arm based on the safe predicted distance, the target distance, and the target tangential velocity; determining the target joint angular velocity for driving the joint movement of the spraying robot arm based on the first matrix and the end target linear velocity, and controlling the movement of the spraying robot arm according to the target joint angular velocity.

[0009] Optionally, generating the safe predicted distance includes: determining the current rate of change of the actual distance based on the current actual distance and the historical actual distance; obtaining the predicted distance based on the current rate of change, the first delay time, and the current actual distance; determining the first displacement boundary based on the maximum withstand acceleration of the end effector of the spraying robot and the first delay time; and limiting the predicted distance based on the first displacement boundary to obtain the safe predicted distance.

[0010] Optionally, generating the terminal target linear velocity includes: determining the normal distance error based on the safe predicted distance and the target distance; determining the normal compensation velocity based on the normal distance error; determining the normal unit vector and the tangential unit vector based on the current terminal position and the pre-calibrated pipeline centerline; and obtaining the terminal target linear velocity according to the normal compensation velocity, the normal unit vector, the target tangential velocity, and the tangential unit vector.

[0011] Thirdly, this application provides a degradation-aware control system for a large-diameter pipe painting robot arm, including a state perception module, a distance prediction control module, a residual fingerprint generation module, a reliability determination module, a control reconstruction module, and a joint control module. The state perception module acquires the original control state and online monitoring state of the painting robot arm and generates a synchronization flag based on the timestamps of each state acquisition. The distance prediction control module executes the distance prediction control method and provides the joint control module with a first matrix, a safe predicted distance, and a basic end-effector target linear velocity. The residual fingerprint generation module generates a process residual fingerprint based on the original control state, the online monitoring state, and the synchronization flag. The reliability determination module determines the reliability of faults and disturbances and the degradation level based on the process residual fingerprint. The control reconstruction module generates reconstruction control parameters based on the reliability of faults and disturbances and the degradation level. The joint control module determines the target joint angular velocity based on the basic end-effector target linear velocity, the first matrix, the safe predicted distance, and the reconstruction control parameters, and controls the movement of the painting robot arm.

[0012] The beneficial effects of this application are as follows: By combining distance residual, echo quality residual, kinematic closure residual, servo residual, load residual, pressure disturbance residual, and synchronization state into a process residual fingerprint, this application can distinguish between ranging link anomalies, joint servo degradation, hose disturbances, and actual pipe wall changes; by mapping fault and disturbance confidence to ranging weight, tangential velocity coefficient, joint space damping length, and control mode, this application can enable fault detection results to enter the robotic arm motion control link; by retaining the distance prediction limiting control path, this application can maintain the original distance control mechanism under normal operating conditions and reconstruct its output under degraded operating conditions. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the degradation-sensing robotic arm control method provided in an embodiment of this application.

[0014] Figure 2 This is a schematic diagram of the data flow from the periodic state packet to the process residual fingerprint provided in an embodiment of this application.

[0015] Figure 3 This is a schematic diagram illustrating the mapping between fault reliability and degradation level provided in the embodiments of this application.

[0016] Figure 4 The control block diagram for solving the end-effector velocity and joint velocity using the control reconfiguration parameters provided in the embodiments of this application is shown.

[0017] Figure 5 A flowchart of the distance prediction and amplitude limiting control path provided in the embodiments of this application.

[0018] Figure 6This is a schematic diagram of the control system module architecture provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] The technical solutions in the embodiments of this application are described below with reference to the accompanying drawings. The described embodiments are used to illustrate the technical implementation of this application and do not constitute a limitation on the scope of protection. Unless otherwise stated, the modules in this embodiment can be implemented by program tasks in the onboard computing unit, or by a combination of controller, driver and sensor interface.

[0022] The large-diameter pipe spraying robot provided in this embodiment includes a mobile chassis, a spraying robotic arm, a nozzle, and a spraying material supply unit. It also includes a non-contact ranging sensor, joint actuators, joint encoders, and an onboard computing unit. The onboard computing unit executes the robotic arm control program, the joint actuators execute target joint angular velocity commands, and the non-contact ranging sensor acquires the current actual distance between the nozzle and the pipe wall. To enable the fault detection results to directly influence the robotic arm control, this embodiment adds process residual fingerprint generation, fault and disturbance confidence determination, and control reconstruction processing to the original distance prediction limiting control path. This processing does not change the input and output of the original distance prediction limiting control path. Instead, after outputting the first matrix, safe predicted distance, and basic end-target linear velocity from the original path, it reconstructs the end-target velocity, joint space damping length, and control mode based on the fault and disturbance confidence.

[0023] like Figure 1 As shown, this method includes the following steps: S100: Obtain the original control state and online monitoring state of the spraying robot arm, and generate a synchronization flag.

[0024] like Figure 2As shown, this stage packages the control quantities required for the original distance prediction and limiting control path with the online monitoring quantities required for fault identification into the same control cycle. The original control state includes the current joint angle, current actual distance, historical actual distance, first delay time, target distance, and target tangential velocity. The online monitoring state includes the ranging echo quality, joint feedback velocity, previous cycle joint velocity command, drive load state, and nozzle pressure state, as well as the acquisition timestamps corresponding to each state. The state perception module generates a synchronization flag based on the acquisition timestamps.

[0025] Before executing S100, a baseline window establishment step can be performed first. After the robot enters the pipe section to be sprayed, the spraying robotic arm equipped with a non-contact ranging sensor is controlled to perform a circumferential scan of at least three axial sections. Each section can collect 120 ranging points, forming a total of 360 spatial points across the three sections. The motion calculation module converts each ranging point into a point cloud on the inner wall of the pipe based on the joint angles and ranging values. The onboard computing unit performs cylindrical surface fitting on the point cloud to obtain the pipe's central axis and radius, thereby establishing a spatial transformation from the robot's base coordinate system to the pipe coordinate system.

[0026] Cylindrical surface fitting can be achieved using the least squares method. Let the first... The point cloud points are A point on the axis of the cylinder is The unit vector of the axis is , radius is The error between the point and the cylindrical surface is... By minimizing ,get , and This coordinate calibration process provides a unified coordinate basis for subsequent normal unit vectors, tangential unit vectors, and kinematic closure residuals.

[0027] Once the baseline window is established, the robot performs a short-distance trial run along the inner wall of the pipe at low speed. This low-speed trial run can last from 10 to 30 seconds, and the control cycle can be [missing information]. to Within this window, the state-aware module acquires ranging readings, ranging echo quality, joint feedback speed, drive load status, and nozzle pressure status. Simultaneously, the motion calculation module generates the reference end position, normal unit vector, and first matrix. The onboard calculation unit determines the threshold set based on residual samples from normal operation. Furthermore, the statistical benchmarks of the joint drive loads within the benchmark window are used as the benchmark load state, and the statistical benchmarks of the nozzle pressure within the benchmark window are used as the benchmark pressure state. For example, the distance residual threshold... Take the 99th percentile of the absolute value of the distance prediction residual, and add the ranging resolution margin; kinematic closure residual threshold. Take the 99th percentile of the absolute value of the distance change closure error and add the ranging resolution margin; joint servo residual threshold. Take the first The 99th percentile of the absolute value of the joint velocity tracking residuals, plus the driver velocity resolution margin; load residual threshold. Take the first The 99th percentile of the absolute value of the joint drive current residuals, plus the current detection resolution margin; pressure disturbance threshold. Take the 99th percentile of the absolute value of the nozzle pressure deviation and add the pressure sampling resolution margin. Continuous loss threshold. Determined based on the allowable continuous failure time of ranging and the control cycle.

[0028] In one embodiment, the robot is in diameter A baseline window was established within the pipeline. A total of 360 point cloud points were collected from three axial sections, and the radius was obtained through fitting. The axial unit vector is The low-speed reference window lasts for 20 seconds, and the control cycle is... This yields 4000 samples. If the distance from the 99th percentile of the absolute value of the predicted residual is... The ranging resolution margin is Then the distance residual threshold is If the 99th percentile of the absolute value of the kinematic closed residual is ,but If the 99th percentile of the absolute value of the velocity tracking residual of the first joint is... The driver speed resolution margin is Then the servo residual threshold of the first joint is If the 99th percentile of the nozzle pressure deviation is The pressure sampling resolution is ,but If the allowable continuous ranging failure time is ,but .

[0029] S100 executes cyclically during formal painting control. The status sensing module reads the current joint angle via the joint encoder. The current actual distance is read through a non-contact ranging sensor. and ranging echo quality And read the historical actual distance from the circular cache. The onboard computing unit retains the first delay time. Target distance Target tangential velocity These variables are consistent with the existing distance prediction and amplitude limiting control path, ensuring that the original path is unaffected by the new fault identification process.

[0030] Meanwhile, the state perception module reads the current joint feedback velocity from the joint actuator. and drive current Read the joint speed command from the control buffer in the previous cycle. Read the nozzle pressure status from the spraying supply unit. If the nozzle pressure status is unavailable, the spray on / off status can be used as an alternative online monitoring status. The onboard computing unit writes a timestamp for each status channel to obtain the current cycle time. The ranging echo quality It is a dimensionless quantity, and its range of values ​​is 1. .

[0031] To avoid misinterpreting sampling delays from different interfaces as physical faults, the S100 also generates a synchronization flag. Let the acquisition times for ranging, joint, drive, and spraying states be respectively... , , and .when When, the corresponding state is determined to be synchronized; when When this happens, the corresponding state is marked as timeout. Synchronization tolerance. Half of the control cycle can be taken. If the control cycle is... ,but .

[0032] A periodic state packet can be represented as .in, This is a synchronization flag vector. The state packet is written to a circular buffer, which holds at least the most recent... One cycle. The value can be between 20 and 100. If the control cycle is... and The window then covers 100 milliseconds of robotic arm state changes.

[0033] In one embodiment, the current joint angle of a certain control cycle is The current actual distance is The actual historical distance is The ranging echo quality is The joint feedback speed is The joint speed command for the previous cycle was The driving current is The nozzle pressure is If the ranging, joint, drive, and spraying conditions all meet the requirements... Synchronization tolerance, then The periodic status packet is sent to the residual fingerprint generation module.

[0034] In terms of data acquisition, the state awareness module can be triggered uniformly by the main control task, or it can be implemented by each interface sampling independently and the main control task aggregating data according to timestamps. For the first approach, the main control task sends sampling trigger signals to the ranging interface, joint encoder interface, and driver interface at the beginning of each control cycle. After sampling, each interface writes the data to a shared buffer. For the second approach, each interface continuously writes data to the buffer according to its own sampling frequency, and the main control task reads the data closest to that center moment at the center of the cycle. Regardless of the approach used, the corresponding timestamp is retained in the state packet, allowing subsequent synchronization flags to be recalculated.

[0035] When a certain online monitoring status becomes unavailable in the current cycle, the status awareness module does not directly replace the status with a zero value. Instead, it uses the status value from the previous reliable cycle and sets the corresponding synchronization flag to anomaly. This is because a zero value might be mistaken for a real speed or current change, while the synchronization flag clearly indicates that the reliability of the status has been compromised. For example, if the driver feedback speed is delayed by one cycle due to bus congestion, the status packet can still carry the feedback speed from the previous cycle, but the synchronization flag indicates that this value should not be used as normal feedback in the low-risk fault reliability determination.

[0036] For multi-DOF robotic arms, the joint angles, joint feedback velocities, and drive load states in the state packet are all arranged in the same joint order. This joint order is written into the parameter area during system initialization, and subsequent residual fingerprint generation, fault confidence calculation, and damping matrix construction all reference this order. By fixing the joint order, the degradation confidence of the first joint can be prevented from being incorrectly applied to the damping length of the second joint. If the robot uses a three-DOF painting robotic arm, the joint state vector contains three components; if a six-DOF robotic arm is used, the joint state vector contains six components, but the state packet structure and subsequent processing rules remain consistent.

[0037] In one alternative implementation, the nozzle pressure status does not necessarily have to be provided by an additional pressure sensor. If the spray supply unit already has feedback on pump current, valve opening, or spray on / off status, these feedbacks can also be used as substitutes for the nozzle pressure status. The purpose of these substitutes is not to measure precise pressure, but rather to reflect whether there are sudden changes in supply, blockage, or drag on the nozzle output side. Thus, this application enables online monitoring of the status using the robot's existing control interface without adding dedicated fault sensors.

[0038] In another example, the control cycle is ten milliseconds, the ranging interface sampling time is two milliseconds after the start of the current cycle, the drive feedback interface sampling time is three milliseconds after the start of the current cycle, and the paint supply interface sampling time is twelve milliseconds after the start of the current cycle. If the synchronization tolerance is five milliseconds, the ranging state and drive feedback state are written into the current cycle synchronization state, and the paint supply state is marked as a timeout state. Subsequently, S200 can still use the previous trusted paint supply state, but S300 will use this timeout flag as the basis for distinguishing between data loss and physical anomalies when establishing the trustworthiness of the ranging link or servo link.

[0039] S200, based on the original control state, the online monitoring state, and the synchronization flag, generate a process residual fingerprint.

[0040] like Figure 2 As shown, the residual fingerprint generation module reads the periodic status packet. Threshold set The system also includes the baseline prediction values ​​output by the distance prediction control module. This module does not directly output alarm results; instead, it organizes the deviations from different sources into a process residual fingerprint. The process residual fingerprint includes at least distance residuals, echo quality residuals, kinematic closure residuals, servo residuals, and load residuals, as well as pressure disturbance residuals and synchronization flags. The purpose of this fingerprint is to preserve the structural differences of the anomaly sources, enabling subsequent steps to distinguish between ranging link failures, joint servo link degradation, hose disturbances, and actual pipe wall geometry changes.

[0041] First, the residual fingerprint generation module generates the predicted distance based on the original distance prediction limiting control path. In one implementation, the predicted distance can be based on the current actual distance. Historical actual distance and the first delay time Obtained. Distance residual is To ensure that residuals from different units can be included in the same confidence level calculation, the distance residuals are normalized to... .in, The distance residual threshold established before S100. It is a dimensionless quantity, and its range of values ​​is 1. .

[0042] Secondly, the residual fingerprint generation module generates an echo quality residual based on the ranging echo quality, and then generates a kinematic closed residual based on the first matrix. The echo quality residual can be expressed as... The kinematic closed residual can be expressed as: ,in Determined based on the first matrix, the pipe normal unit vector, the joint velocity command from the previous cycle, and the control cycle. The normalized kinematic closure residual is... .

[0043] Next, the residual fingerprint generation module obtains the servo residual based on the joint speed command and the joint feedback speed. The servo residual of each joint is The normalized servo residual is .in, For the first The servo residual threshold determined for each joint within the reference window. If a joint experiences motor response lag, reducer jamming, or abnormal drive loop feedback, the servo residual of that joint will continue to increase over multiple cycles.

[0044] Next, the residual fingerprint generation module obtains the load residual based on the drive load state and the baseline load state. The load residual of each joint is .in, For the first Current drive current of each joint This is the average current at this joint within the reference window. The normalized load residual is... The pressure disturbance residual is After normalization, it becomes When the traction force of the spray hose increases, the joint resistance increases, or the nozzle approaches the hose wall, the load residual or pressure disturbance residual will usually increase.

[0045] The synchronization flag is generated by S100 based on the acquisition timestamps of each state. If a sensor or drive feedback state fails to arrive within the synchronization tolerance, the corresponding synchronization flag is set to 1. The synchronization flag does not represent the magnitude of a physical quantity, but rather whether the reliable basis of the signal in the current control cycle has been compromised. Writing the synchronization flag into the process residual fingerprint can prevent the controller from misinterpreting communication timeouts as actual pipe wall mutations.

[0046] The process residual fingerprint can be represented as .in, This represents the number of joints in the painting robot arm. Each element in this vector has a clearly defined data source and production step. The residual fingerprint generation module also generates a continuous anomaly count vector. and will Write to a sliding window, the window length is One control cycle.

[0047] In one embodiment, the current actual distance is The original distance prediction path yields the predicted distance. Then the distance residual is .like ,but If the predicted normal displacement obtained from the first matrix and the velocity command of the previous cycle is... The measured distance change is Then the kinematic closed residual is The corresponding normalized value is approximately 0.029. The joint velocity command for the first joint in the previous cycle was... The current feedback speed is The servo residual of the first joint is .like ,but The joint velocity command for the third joint in the previous cycle was... The current feedback speed is ,like ,but .

[0048] Continuing with this embodiment, if the average reference current of the three joints is The current drive current is The corresponding load threshold is Then the normalized load residuals of the first, second, and third joints are 1, 0.25, and 1, respectively. If the reference pressure is... The current pressure is ,and The normalized value of the pressure disturbance is approximately 0.571. The residual fingerprint of the process shows that the individual ranging residuals did not reach a serious anomaly, but the first and third joints showed persistent anomalies on the servo or load side, indicating a slight source of disturbance in the nozzle pressure.

[0049] In another embodiment, if the distance residual suddenly increases in a single cycle, but the servo residual, load residual, and synchronization flag are all normal, the residual fingerprint generation module will output a combination of high distance residual and low servo residual. This combination in S300 is more likely to be due to transient noise in the ranging link than joint degradation. If the distance residual increases while the synchronization flag indicates ranging data timeout, S300 will increase the ranging link failure confidence level. If the distance residual increases while the servo residual and load residual also increase, S300 will increase the joint servo link degradation confidence level or hose disturbance confidence level.

[0050] In the generation of process residual fingerprints, the residuals are not required to have the same physical units. Distance residuals can be on the order of millimeters, servo residuals on the order of radians per second, and load residuals on the order of amperes. After normalization using a threshold set, each residual is converted into a dimensionless index. This normalization process allows the confidence determination module to compare the intensity of anomalies from different sources without directly mixing current and distance dimensions. This process satisfies the requirement of dimensional consistency.

[0051] To avoid misjudgments caused by a single spike, the residual fingerprint generation module can also generate a persistence flag for each normalized residual. The persistence flag indicates whether the residual has been continuously higher than the corresponding low-risk threshold for several recent periods. For example, if the distance to the normalized residual reaches 1 within one period but then immediately returns to below 0.2, the persistence flag is not set; if the distance to the normalized residual is higher than 0.6 for several consecutive periods, the persistence flag is set. The persistence flag can be used as an additional element in the process residual fingerprint or calculated from historical fingerprints during S300 window statistics.

[0052] The process residual fingerprint can also include directional consistency information. For real pipe wall curvature changes, the current actual distance change is usually consistent with the movement trend of the robotic arm's end effector relative to the pipe's central axis; for distance sensor reflections or dust obstruction, distance readings may change abruptly, but joint kinematic predictions do not support an end effector displacement of the same amplitude. Therefore, the residual fingerprint generation module can obtain the kinematic closure residual based on the current end effector position, the normal unit vector, and the end effector position of the previous cycle. This residual is used to help determine whether distance measurement anomalies are caused by real geometric changes.

[0053] In one example, when the robot passes near a pipe weld, the actual distance changes for three consecutive cycles, while the servo residual and load residual remain low, and the kinematic closure residual is in the same direction as the distance change. In this case, the process residual fingerprint is closer to the actual pipe wall disturbance pattern. As another example, paint droplet reflection occurs near the robot's nozzle, and the actual distance abruptly changes in a single cycle. The kinematic closure residual does not support this abrupt change, and the ranging synchronization flag indicates a delay in the ranging data for that cycle. In this case, the process residual fingerprint is closer to a ranging link failure mode.

[0054] For servo link degradation, the process residual fingerprint typically shows an increase in both servo residual and load residual. If the joint speed feedback is consistently lower than the speed command of the previous cycle, and the drive current increases relative to the reference load, it indicates that the joint may be unable to follow the command due to hose traction, reducer resistance, or motor degradation. If only the load residual increases while the servo residual is not high, it may be due to short-term external disturbances or changes in the spray hose attitude. This difference will translate into different joint degradation confidence levels in S300.

[0055] Through the above processing, the process residual fingerprint is not a single threshold comparison result, but a structured state containing the source of the anomaly, the persistence of the anomaly, and a reliable basis for synchronization. Subsequently, S300 can form the confidence level of the fault and disturbance based on this structured state, and S400 can generate control reconfiguration parameters based on the confidence level of the fault and disturbance, thereby establishing a deterministic data link between fault detection and robotic arm control.

[0056] S300, based on the process residual fingerprint, determines the confidence level of the fault and disturbance, and the degradation level.

[0057] like Figure 3 As shown, the credibility determination module reads the most recent... The process residual fingerprints for each cycle are used to generate confidence levels for ranging link faults, pipe wall disturbances, hose disturbances, and joint servo link degradation. Fault and disturbance confidence levels are not diagnostic labels for individual display but rather inputs to subsequent control reconfiguration modules. Each confidence level is limited to... Within the range, the larger the value, the more credible the corresponding fault or disturbance source.

[0058] The reliability of a ranging link failure is jointly determined by the distance residual, echo quality residual, kinematic closure residual, and ranging synchronization flag. Let the cumulative ranging anomaly within the window be... .in, This is the ranging synchronization flag; it is 0 when the ranging data meets the synchronization conditions and 1 when the ranging data times out. The reliability of the ranging link failure is... .in, This is a dimensionless sensitivity coefficient, which can be taken from 1.2 to 2.5. This mapping ensures that slight distance deviations in a single cycle will not directly trigger high confidence, while continuous deviations or synchronization anomalies will accumulate.

[0059] No. The reliability of servo link degradation at each joint is determined jointly by the servo residual and the load residual. Let... . No. The reliability of joint degeneration is: Here, 0.6 corresponds to the speed tracking residual weight, and 0.4 corresponds to the load residual weight. Both are dimensionless weights. If the drive current better reflects joint degradation in a certain scenario, the weights can be adjusted to 0.5 and 0.5, but the sum of the two weights remains 1.

[0060] The reliability of the actual pipe wall disturbance is jointly determined by the distance residual, kinematic closure residual, echo quality residual, and ranging synchronization indicator. When the distance residual increases, and the echo quality is reliable, the kinematic closure residual is low, and ranging synchronization is normal, the anomaly more closely matches the actual pipe wall geometry. Let the cumulative amount of the actual pipe wall disturbance within the window be... The reliability of the actual pipe wall disturbance is: .in, The dimensionless sensitivity coefficient can be taken as 1.2 to 2.5. This formula increases the reliability of the true pipe wall disturbance when the distance residual increases and the ranging link itself remains reliable; when the echo quality is poor, the kinematic closure is not supported, or the ranging synchronization is abnormal, the reliability of the true pipe wall disturbance is suppressed.

[0061] The reliability of the hose disturbance is jointly determined by the pressure disturbance residual, the average load residual, and the average servo residual. Let the average load residual be... The average servo residual is The cumulative amount of hose disturbance within the window is... The reliability of the hose disturbance is... .in, The dimensionless sensitivity coefficient can be taken as 1.2 to 2.5. The fault and disturbance confidence vector can be expressed as... Its dimensions are .

[0062] In one embodiment, the mean of the distance normalized residual over the most recent 20 periods is 0.74, the mean of the kinematic closure normalized residual is 0.10, the mean of the echo quality normalized residual is 0.05, and the ranging synchronization flag is 0 for all of them. .when hour, This corresponds to the confidence level of the actual pipe wall disturbance entering the reconstruction mode range. In another embodiment, the normalized mean of the pressure disturbance is 0.57, the average load residual is 0.75, and the average servo residual is 0.30. .when hour, The corresponding hose disturbance confidence level enters the protection mode range.

[0063] Degradation level Obtained from the deterministic gradation mapping table. If the confidence levels of all faults and disturbances are less than 0.35, then... This is the normal mode. If at least one confidence level is not less than 0.35 and less than 0.60, then... In guardian mode. If at least one confidence level is not less than 0.60 and less than 0.85, then... For reconstruction mode. If at least one confidence level is not less than 0.85, or ranging synchronization is continuously lost for more than [number missing] hours... One cycle, then Maintain mode for safety. It is determined by the allowable continuous failure time and control cycle of the baseline window phase.

[0064] In one embodiment, window length The average normalized servo residual of the first joint over the last 20 periods is 0.82, and the average normalized load residual of the first joint is 0.75. .when hour, The confidence level is in the range of 0.60 to 0.85, therefore the degradation level is at least in the reconstruction mode.

[0065] In the same embodiment, if the mean distance residual is 0.28, the mean echo quality residual is 0, the mean kinematic closure residual is 0.03, and there is no distance synchronization timeout, then .when hour, This value did not enter the guardian mode range. Since the confidence level of the first joint degradation is higher than that of the ranging link failure, the controller primarily attributes the anomaly to joint servo link degradation rather than directly rejecting the ranging data. Therefore, the S400 reconstructs the first joint damping length and tangential velocity coefficient.

[0066] In another embodiment, if the ranging synchronization flag times out for 5 consecutive cycles, and Even if the distance residual value has not yet increased, the degradation level enters a safety hold mode. This can handle scenarios where invalid data is caused by packet loss in the ranging link or reflection. The safety hold mode is not a simple alarm, but directly overrides normal spray propulsion control in S400 and S500.

[0067] The aforementioned confidence level and grade mapping are all deterministic calculations. Each input comes from the process residual fingerprint generated in S200, and each threshold comes from the baseline window prior to S100. Those skilled in the art can reproduce this fault confidence level calculation process based on the formulas, window lengths, threshold generation rules, and grade intervals provided in this paper.

[0068] The trustworthiness determination module can also use hysteresis to handle degradation level transitions. Entry and exit thresholds can be set for daemon mode, refactoring mode, and security maintenance mode, respectively. The entry threshold is higher than the exit threshold. For example, the entry threshold for refactoring mode can be 0.6, and the exit threshold can be 0.52. When the joint degradation trustworthiness increases from 0.4 to 0.65, the system enters refactoring mode; when the trustworthiness subsequently drops to 0.55, the system remains in refactoring mode; only when the trustworthiness falls below 0.52 and persists for several cycles does the system exit refactoring mode. This hysteresis rule avoids frequent switching of control modes near the boundaries.

[0069] Fault and disturbance confidence levels can also be output along with fault or disturbance source tags. Fault or disturbance source tags include ranging link, joint servo link, spray material supply link, and actual pipe wall disturbance. These tags are not manually entered but automatically determined by combinations of residuals in the process residual fingerprint. When the ranging residual is high, the synchronization flag is abnormal, and the servo residual is low, the source tag is biased towards the ranging link; when both the servo residual and load residual are high, the source tag is biased towards the joint servo link; when the distance residual and kinematic closure residual are in the same direction, the source tag is biased towards the actual pipe wall disturbance.

[0070] In a specific state transition example, the initial state is normal mode. If the first joint degradation confidence level is below 0.35 for twenty consecutive cycles, the controller remains in normal mode. Subsequently, if the first joint degradation confidence level rises to 0.48 within thirty cycles, the system enters guard mode, and S400 reduces the tangential velocity coefficient. If the confidence level continues to rise to 0.72, the system enters reconfiguration mode, and S400 increases the damping length of the first joint. If the confidence level further rises to 0.88, or if the ranging link synchronization is continuously lost for a set number of cycles, the system enters safety hold mode, and S500 executes a hold or pullback action.

[0071] In another example, the distance residual reaches 1 in a single cycle, but then remains below 0.2 for the next four cycles. The ranging synchronization flag is normal, and the servo and load residuals are also normal. The ranging link failure confidence level after windowing statistics will not exceed the guardian mode entry threshold. Therefore, the system will not immediately enter safe hold mode due to a single ranging spike. Conversely, if the distance residual is above 0.8 for several consecutive cycles, accompanied by ranging synchronization timeouts, the ranging link failure confidence level will rise rapidly, and the degradation level will enter either refactoring mode or safe hold mode.

[0072] The aforementioned degradation level conversion rules do not rely on a trained model or require online learning. Ordinary technicians can reproduce the degradation level output simply by considering the window length, entry threshold, exit threshold, and duration. This deterministic design ensures that the fault reliability reflects the equipment degradation trend and is fully supported by the disclosed information in the patent text.

[0073] S400, based on the fault and disturbance confidence level and the degradation level, generate reconstruction control parameters.

[0074] like Figure 4 As shown, the control reconfiguration module converts fault and disturbance confidence levels and degradation levels into robotic arm control parameters. These reconfigured control parameters include ranging weights, tangential velocity coefficients, joint space damping lengths, and control modes. The purpose of this step is to integrate fault detection results into the control link, ensuring that the detection results do not remain merely at the alarm or recording level.

[0075] Ranging weights are used to adjust the influence of the ranging closed loop on the normal control input. Ranging weights can be expressed as... .in, To assess the reliability of ranging link failures, Weights are retained for the lowest distance measurement. A value between 0.15 and 0.30 is acceptable. The reason for setting a minimum retention weight is that even if there is an anomaly in the ranging link, a portion of the ranging feedback can still be retained within a safe range to prevent the controller from completely losing normal distance awareness.

[0076] The tangential velocity coefficient is used to adjust the spray advance speed. Let... The tangential velocity coefficient, representing the maximum value among the fault and disturbance confidence levels, can be expressed as: .in, The velocity reduction coefficient can be taken as 0.5 to 1.0. When the reliability of the fault or the reliability of the hose disturbance increases, the tangential velocity coefficient decreases, and the advance speed of the nozzle along the axial or circumferential direction of the pipe decreases accordingly, leaving more control margin for the adjustment of the normal distance.

[0077] The joint space damping length is used to adjust the joint velocity contribution in the solution of the target joint angular velocity. The joint damping length can be expressed as... .in, Based on the basic damping length, This is the damping amplification factor. For the first The reliability of servo link degradation for each joint. In this expression... and All of them are dimensionless quantities, therefore and They have the same dimensions. When the first matrix is ​​a linear velocity Jacobian matrix, the diagonal elements of the joint space damping matrix are taken as the square of the corresponding damping length, i.e., the values ​​in the subsequent S500. satisfy .thus, and They have the same dimensions. As the damping length increases, the velocity contribution of the degenerated joint in the pseudo-inverse solution of joint space damping decreases.

[0078] The control mode is determined by the degradation level. In normal mode, the control reconstruction module allows the baseline target linear velocity from the original distance prediction slack path output to directly participate in the solution. In guard mode, the control reconstruction module reduces the tangential velocity while retaining the ranging closed loop. In reconstruction mode, the control reconstruction module enables joint space damping reconstruction and limits the velocity contribution of degraded joints. In safety hold mode, the control reconstruction module outputs a safety hold flag, and the S500 subsequently freezes or retracts the nozzle, ceasing normal spraying propulsion.

[0079] In one embodiment, S300 output , The confidence scores for the remaining joints were all below 0.35. ,but .Pick ,but If the original target's tangential velocity is The reconstructed tangential propulsion velocity is approximately .

[0080] Continuing with this embodiment, if the basic damping length Damping amplification factor Then the damping length of the first joint is The remaining joints, due to their lower reliability due to degradation, maintain damping lengths close to the basic damping lengths. Therefore, when calculating the target joint angular velocity, the S500 reduces the velocity abrupt change in the first joint, while utilizing other healthy joints to adjust the nozzle attitude and distance.

[0081] In another embodiment, if Furthermore, if the ranging synchronization flag times out continuously, the degradation level enters a safe hold mode. The control reconfiguration module sets the control mode to safe hold and outputs a tangential velocity coefficient of 0. At this time, the S500 no longer continues spraying based on the current ranging reading, but instead performs a hold or retraction action based on the most recent reliable pipe coordinates and normal unit vector.

[0082] The control reconfiguration module can also limit the rate of change of parameters when outputting reconfigured control parameters. Although the ranging weight, tangential velocity coefficient, and joint space damping length are determined by the fault confidence level, directly updating them according to the instantaneous confidence level value in each cycle may cause jitter in the robot arm's control input. Therefore, the control reconfiguration module can use a first-order smoothing method to update the reconfigured control parameters. Based on the control parameters of the previous cycle, the current cycle is only allowed to approach the target parameter within a limited range. This range limit can be determined based on the control cycle and the robot arm's allowable acceleration.

[0083] In guardian mode, the control reconfiguration module primarily reduces the tangential velocity coefficient while maintaining the continuity of normal distance control. This mode is suitable for situations where the fault confidence level has just entered the abnormal range but has not yet reached the point where joint reconfiguration is required. By reducing the tangential velocity, the nozzle's advance distance along the pipe direction is reduced, allowing more time for normal distance adjustment and preventing minor anomalies from being amplified by rapid movement.

[0084] In reconfiguration mode, the control reconfiguration module simultaneously adjusts the ranging weights and joint space damping lengths. If the ranging link failure confidence level is high, the ranging weights are reduced, and the controller relies more on the reliable and safe predicted distance buffer and kinematic prediction results. If the degradation confidence level of a certain joint is high, the damping length corresponding to that joint is increased, so that the target joint angular velocity solved by S500 reduces the velocity contribution of that joint. In this way, the controller does not simply stop the robotic arm, but reallocates the motion task within the available degrees of freedom.

[0085] In safety hold mode, the output of the control reconfiguration module has the highest priority. Even if the original distance prediction limiting control path can still generate the basic end target linear velocity, safety hold mode will override this basic output. The control reconfiguration module sets the tangential velocity coefficient to zero and outputs a hold or retraction flag to the S500. The hold action is suitable for situations where the nozzle distance is within the safe range but the ranging link is temporarily unavailable; the retraction action is suitable for situations where the nozzle distance is close to the pipe wall or the reliability of joint servo degradation is high.

[0086] In one embodiment, the ranging link failure confidence level is 0.7, the first joint degradation confidence level is 0.4, and the remaining confidence levels are below 0.35. The control reconstruction module reduces the ranging weights to near the minimum retention weights and decreases the tangential velocity coefficient, but does not significantly increase the first joint damping length. In another embodiment, the ranging link failure confidence level is 0.3, and the first joint degradation confidence level is 0.75. In this embodiment, the control reconstruction module primarily increases the first joint damping length and moderately decreases the tangential velocity coefficient. The output differences between the two embodiments reflect the differentiated control of fault sources in this application.

[0087] After the reconstructed control parameters are generated, the control reconstruction module writes the parameter package to the real-time cache. The parameter package includes the control cycle number, ranging weight, tangential velocity coefficient, damping length of each joint, control mode, and fault source label. The S500 reads parameter packages with the same cycle number to avoid cross-cycle parameter mismatch. If the cycle number of the parameter package read by the S500 is inconsistent with the current basic end target linear velocity cycle number, the joint control module enters guard mode and requests resynchronization for the next cycle.

[0088] S500 determines the target joint angular velocity based on the basic end target linear velocity, the first matrix, the safe predicted distance, and the reconstructed control parameters, and controls the movement of the spraying robot arm according to the target joint angular velocity.

[0089] like Figure 4 As shown, the joint control module receives the basic end-target linear velocity and safe predicted distance from the original distance prediction amplitude-limited path output, as well as the first matrix output from the motion calculation module and the reconstructed control parameters output from the S400. The basic end-target linear velocity includes normal control components and tangential control components. The first matrix is ​​used to establish the mapping relationship between the end-target velocity and the joint angular velocity. The reconstructed control parameters are used to change the weight of the normal control quantity, the magnitude of the tangential velocity, the joint space damping length, and the control mode.

[0090] In normal mode, the joint control module can use the basic end-target linear velocity as the reconstructed end-target linear velocity. In guard mode and reconstructed mode, the joint control module processes the basic end-target linear velocity based on the ranging weight and tangential velocity coefficient. Let the basic end-target linear velocity be... The normal unit vector of the pipeline is The normal component of the ranging is The tangential component is .in, The components of the target linear velocity at the base end along the unit vector normal to the pipe. It is the propulsion component along the tangential unit vector in the linear velocity of the base terminal target.

[0091] The joint control module also maintains a trusted and secure predicted distance cache. This is based on the current period's ranging synchronization flag. A value of 0 indicates a reliability of the ranging link failure. Below the trusted update threshold If the echo quality residual is lower than the echo confidence threshold, the joint control module will predict the current safe distance. The current pipeline normal unit vector and the current period number are written to the trusted secure prediction distance cache. The trusted update threshold and the echo trusted threshold can both be set to 0.35. If the current period does not meet the above update conditions, the cache retains the secure prediction distance from the previous trusted period and records the cache age. .when Not greater than the cache validity period At that time, the joint control module generates a reliable predicted distance based on the safe predicted distance in the cache and the kinematic predicted displacement from the cache period to the current period. ;when Greater than When this happens, the joint control module no longer uses the cache and puts the control mode into a safe hold mode. Five to 20 control cycles can be selected.

[0092] The joint control module is based on the reliable predicted distance. Distance to target Generate historical security prediction normal components The historical security prediction normal component can be expressed as: .in, Gain is controlled by normal distance. This represents the upper limit of the normal compensation velocity. The reconstructed end-target linear velocity is... When the reliability of the ranging fault increases, As the historical security prediction normal component decreases, its proportion in normal control increases.

[0093] In reconfiguration mode, the joint control module uses a pseudo-inverse of joint space damping to solve for the target joint angular velocity. Let the first matrix be... The joint space damping matrix is The target joint angular velocity is One implementation method is... Joint space damping matrix It is a diagonal matrix. When When the linear velocity Jacobian matrix is ​​unnormalized, The The diagonal elements are ;when When dimensionless normalization is performed beforehand, The The diagonal elements are the normalized damping coefficients. Joints with higher degradation confidence have larger damping lengths and lower velocity contributions.

[0094] In safety hold mode, the joint control module does not continue normal spraying propulsion. If the reliable safety prediction distance buffer shows that the nozzle distance is still within the safe range, the joint control module outputs a hold posture command to maintain the nozzle's current safe position. If the reliable safety prediction distance buffer shows that the nozzle distance is close to the pipe wall, or the buffer has exceeded the effective cycle count, the joint control module generates a retraction velocity along the pipe normal unit vector and converts this retraction velocity into the target joint angular velocity. The retraction velocity does not exceed the safe velocity. For example, it is advisable to .

[0095] In one embodiment, the ranging normal component in the linear velocity of the base end target is: The tangential control component is If the historical normal component formed by the trusted security prediction distance cache and kinematic prediction results is... S400 output , Therefore, the reconstructed terminal target linear velocity is This speed retains normal distance adjustment and reduces tangential propulsion speed.

[0096] Continuing with this embodiment, the damping length of the first joint is... Increase to The damping lengths of the second and third joints remain close. The joint control module substitutes the joint space damping matrix into the damping pseudo-inverse solution to obtain the target joint angular velocity. If the speed limit of each joint is... If the target joint angular velocity meets the limit requirement, the joint drive module receives the instruction and controls the painting robot arm to continue moving.

[0097] In another embodiment, if the degradation level is security preservation mode, and the nearest trusted distance in the trusted security prediction distance cache is... The target distance is This indicates that the nozzle has become too close to the pipe wall. The joint control module generates a retraction velocity along the normal unit vector. The corresponding joint angular velocity is then calculated using the first matrix. The spraying supply unit can simultaneously reduce the spraying flow rate or pause spraying, and resume normal spraying once the ranging link is synchronized and the degradation level is below the safety maintenance mode.

[0098] S600, Distance Prediction Limiting Control Path.

[0099] like Figure 5 As shown, the distance prediction control module executes the existing distance prediction limiting control method. This method serves as a secondary independent element in this application and also provides the S500 with the basic end-target linear velocity. This path maintains its original control logic and is not deleted or changed due to the addition of degradation sensing processing.

[0100] The distance prediction and control module first acquires the current joint angles and the current actual distance of the painting robot arm. The current joint angles are obtained from feedback by the joint encoder or joint actuator. The current actual distance is obtained by a non-contact ranging sensor. The motion calculation module performs forward kinematics calculations based on the current joint angles to determine the current end effector position and the first matrix of the painting robot arm. The first matrix can be the Jacobian matrix under the current posture, or it can be an equivalent matrix that can map the linear velocity of the end effector target to the joint angular velocity.

[0101] Subsequently, the distance prediction control module determines the predicted distance based on the current actual distance, the historical actual distance, and the first delay time. Let the control period be... The current actual distance is The actual historical distance is The first delay time is The current rate of change of the actual distance. The predicted distance is The first delay time may include the sampling delay of the ranging sensor, the processing delay of the controller, and the drive response delay. To prevent the predicted distance from deviating excessively from the actual distance due to sensor glitch or abnormal reflection, the distance prediction control module limits the predicted distance based on the first displacement boundary to generate a safe predicted distance. Let the maximum withstand acceleration at the end of the spraying robot arm be... The first displacement boundary is The safe predicted distance is... ,in, Indicates will Limit to the lower limit and upper limit between.

[0102] After generating the safe predicted distance, the distance prediction control module determines the normal distance error based on the safe predicted distance and the target distance. Let the target distance be... The normal distance error is The normal compensation speed is .in, Gain is controlled by normal distance. This represents the upper limit of the normal compensation velocity. The motion calculation module determines the normal unit vector at the current end position based on the current end position and the pre-calibrated pipe centerline. and tangential unit vector The final target linear velocity is The joint calculation module then determines the target joint angular velocity based on the first matrix and the end-effector linear velocity.

[0103] In one embodiment, the current actual distance is The actual historical distance is The control cycle is The first delay time is The target distance is The target tangential velocity is The maximum acceleration that the end can withstand is The current rate of change of the actual distance. The predicted distance is The first displacement boundary is Therefore, the safe prediction distance is limited to [value]. The normal distance error is... .like And if the normal compensation speed has not reached the upper limit, then the normal compensation speed is The controller synthesizes the terminal target linear velocity based on the normal compensation velocity and the tangential velocity components.

[0104] In normal mode, the distance prediction limiting control path can directly drive the painting robot arm. In the new degradation sensing scheme, the first matrix, safe predicted distance, and end-target linear velocity output by this path are used as basic control variables. The end-target linear velocity output by the distance prediction control module is recorded as the basic end-target linear velocity in the degradation sensing main scheme. The S500 reconstructs this basic end-target linear velocity based on the fault and disturbance confidence level and degradation level. Thus, the original scheme is retained as a complete secondary priority, while also becoming the basic control path of the new main scheme.

[0105] System Implementation Examples like Figure 6 As shown, this embodiment also provides a degradation-aware robotic arm control system for large-diameter pipe painting robots. The system includes a state perception module, a residual fingerprint generation module, a reliability determination module, and a control reconstruction module, as well as a joint control module and a distance prediction control module.

[0106] The state awareness module acquires the original control state and online monitoring state of the painting robot arm, and generates a synchronization flag based on the acquisition timestamps of each state. The original control state includes the current joint angle, current actual distance, historical actual distance, first delay time, target distance, and target tangential velocity. The online monitoring state includes the ranging echo quality, joint feedback speed, and joint speed command from the previous cycle, as well as the drive load state, nozzle pressure state, and corresponding acquisition timestamps. The state awareness module can be implemented through the joint encoder interface, ranging interface, driver interface, and painting material supply interface.

[0107] The residual fingerprint generation module generates process residual fingerprints based on the original control state, online monitoring state, and synchronization flags. This module includes distance residual subunits, echo residual subunits, kinematic closure residual subunits, and servo residual subunits, as well as load residual subunits, pressure residual subunits, and synchronization flag subunits. The distance residual subunit calls the predicted distance output by the distance prediction control module. The echo residual subunit reads the ranging echo quality. The kinematic closure residual subunit reads the first matrix and the normal unit vector. The servo residual subunit compares the joint speed command and the joint feedback speed. The load residual subunit compares the current drive load state with the reference load state. The pressure residual subunit compares the nozzle pressure state with the reference pressure state. The synchronization flag subunit records whether each monitoring state meets the synchronization condition for the same control cycle.

[0108] The reliability determination module is used to determine the reliability and degradation level of faults and disturbances based on process residual fingerprints. This module includes a window statistics subunit, a reliability mapping subunit, and a level mapping subunit. The window statistics subunit stores the most recent... The process residual fingerprint for each cycle. The reliability mapping subunit determines the reliability of ranging link failure, pipe wall real disturbance, and hose disturbance, and determines the reliability of joint servo link degradation. The level mapping subunit outputs normal mode, guard mode, reconfiguration mode, or safety maintenance mode based on the reliability interval.

[0109] The control reconfiguration module generates reconfigured control parameters based on fault and disturbance confidence levels and degradation levels. This module includes a ranging weight subunit, a tangential velocity coefficient subunit, a joint space damping length subunit, and a control mode subunit. The ranging weight subunit adjusts the influence of the ranging closed loop on the normal control quantity. The tangential velocity coefficient subunit adjusts the spraying propulsion speed. The joint space damping length subunit adjusts the contribution of degraded joints in the joint velocity solution. The control mode subunit outputs a hold or retraction flag in safe hold mode.

[0110] The joint control module determines the target joint angular velocity based on the baseline end-effector linear velocity, the first matrix, the safe predicted distance, and reconstructed control parameters, and controls the spraying robot arm's movement accordingly. The joint control module includes a velocity decomposition subunit, a historical normal fusion subunit, and a velocity envelope subunit, as well as a joint space damping pseudo-inverse solver subunit and a safety holding subunit. The velocity decomposition subunit processes normal and tangential velocities. The historical normal fusion subunit combines the reliable safe predicted distance cache and kinematic prediction results with the current safe predicted distance to form a normal substitution control variable. The joint space damping pseudo-inverse solver subunit converts the end-effector linear velocity into joint angular velocity. The safety holding subunit overrides the normal propulsion output when the ranging link failure confidence level is too high or the degradation level is in safety holding mode.

[0111] The distance prediction control module is used to execute the distance prediction limiting control method. This module may include a status reading subunit, a motion calculation subunit, and a predicted distance subunit, and may also include a displacement boundary limiting subunit, an end-effector velocity generation subunit, and a joint velocity generation subunit. This module provides the joint control module with a first matrix, a normal unit vector, a safe predicted distance, and a basic end-effector target linear velocity; it can also independently output the original target joint angular velocity.

[0112] The modules described above can be integrated within the same onboard computing unit, or they can be distributed among the onboard computing unit, microcontroller, and driver internal control tasks. Modules can exchange status packets, process residual fingerprints, and fault confidence vectors via shared memory, message queues, or real-time buses, and can also exchange reconfiguration control parameters. Module division is used to describe functional logic and does not limit the physical hardware boundaries.

[0113] Before outputting the target joint angular velocity, the joint control module can also perform joint velocity limiting and acceleration limiting. Velocity limiting ensures that the target joint angular velocity of each joint does not exceed the allowable range of the actuator. Acceleration limiting restricts the change in target joint angular velocity between adjacent cycles. This range of change is determined by the range of joint space change corresponding to the maximum acceleration that the robotic arm end effector can withstand. If the target joint angular velocity obtained by the damped pseudo-inverse solution exceeds the limit, the joint control module first scales the velocity vector proportionally, and then decides whether to reduce the tangential velocity coefficient or enter a safety holding mode based on the control mode.

[0114] In guardian mode, if the speed limit is triggered continuously, it indicates that the current tangential speed is still too high. The joint control module can then send a speed limit flag back to the control reconfiguration module. The control reconfiguration module further reduces the tangential speed coefficient in the next cycle. This feedback is not a new fault diagnosis model, but rather a deterministic correction of the reconfigured control parameters to the control output saturation state. Through this correction, the system can improve control stability under degraded conditions without changing the main control flow.

[0115] In reconfiguration mode, if a degraded joint still exhibits high servo residuals after its damping is increased, the joint control module can limit the upper speed limit of that joint in subsequent cycles. For example, when the first joint has a high degradation confidence level, its upper speed limit can be reduced from the normal upper limit to 60% of the normal upper limit. This limit, in conjunction with the damping pseudo-inverse, allows the robotic arm to prioritize other joints to complete feasible movements at the nozzle tip. If the robotic arm's degrees of freedom are insufficient to maintain the spraying distance under this limit, the control mode enters a safety hold state.

[0116] In the safety-maintaining mode, the retraction action generated by the joint control module can be divided into two stages. The first stage stops tangential advance and maintains the nozzle attitude to prevent the spray trajectory from continuing to expand into unknown areas. The second stage retracts the nozzle along the normal unit vector in the nearest reliable pipe coordinate system, bringing the distance between the nozzle and the pipe wall back to a safe range. If the ranging link regains synchronization, and the reliability of the fault and disturbance remains below the exit threshold, the system can revert from the safety-maintaining mode to the guard mode and then gradually resume normal spray advance.

[0117] The distance prediction limiting control path in this application serves two purposes. First, it acts as a separate, secondary method to fully implement the control of the painting robot arm. This control is based on the current joint angle, current actual distance, historical actual distance, and a first delay time, as well as a first displacement boundary, a safe predicted distance, and the end-effector linear velocity. Second, it provides the basic end-effector linear velocity in the main method. The newly added degradation perception scheme does not change the calculation order of this path; instead, it reconstructs the weights, velocity, and damping when the output enters the joint control module. Therefore, under normal operating conditions, the robot can achieve the same distance prediction limiting effect as the original scheme, while under degradation conditions, it can achieve fault source differentiation and control quantity reconstruction effects.

[0118] For different types of large-diameter pipelines, the target distance for the distance prediction and limiting control path can be set according to the spraying process. For example, the target distance can be set to 120 mm when spraying anti-corrosion coating, and to 90 mm when spraying repair coating. The specific value of the target distance does not affect the fault reliability and control reconfiguration logic of this application. As long as the distance prediction control module can output the basic end target linear velocity, S500 can process it according to the reconfiguration control parameters.

[0119] In the system implementation, the residual fingerprint generation module, the confidence determination module, and the control reconfiguration module can run in the real-time thread of the onboard computing unit. The distance prediction control module and the joint control module can run in the same real-time thread, or they can run separately in the onboard computing unit and the microcontroller. When the modules are distributed across different processors, the periodic state packets, process residual fingerprints, and reconfiguration control parameters can be transmitted via a real-time message queue. Each message carries a control period number to ensure consistency between the data preamble and the control period.

[0120] If the robot uses multiple nozzles, the state perception module can establish corresponding ranging and pressure states for each nozzle. The residual fingerprint generation module can generate multiple distance residual components according to the nozzle number. The control reconstruction module selects the corresponding ranging weight based on the nozzle number currently being sprayed and uniformly reconstructs the joint space damping length of the shared robotic arm. Thus, this application can be applied to single-nozzle spraying robots as well as multi-nozzle alternating spraying robots.

[0121] If the robot uses different types of non-contact ranging sensors, such as laser ranging sensors, ultrasonic ranging sensors, or structured light ranging sensors, the distance residual generation rules in S200 still apply. The differences between different sensors are mainly reflected in the distance residual thresholds and synchronization flag conditions obtained in the reference window. By re-establishing the threshold set before S100, the system can adapt to different ranging hardware without changing the process residual fingerprint, fault and disturbance confidence, and reconstructed control parameters as defined in the claims.

[0122] In maintenance scenarios, project personnel can read the historical records of fault and disturbance reliability to determine whether a certain joint or ranging link frequently enters guardian mode or refactoring mode. This historical record is not a necessary condition for the establishment of the control loop in this application, but it can serve as a maintenance reference. After maintenance is completed, the robot re-executes the baseline window establishment and refreshes the threshold set to ensure that subsequent reliability calculations match the new hardware state.

[0123] During fault recovery, the system does not directly jump back to normal mode from safety hold mode. The reliability determination module first confirms that the reliability of the ranging link fault and the reliability of the joint servo link degradation are both continuously below the corresponding exit threshold, while simultaneously confirming that the synchronization flag is continuously normal. The control reconfiguration module then enters guard mode and gradually increases the tangential speed coefficient over multiple control cycles. If any reliability increases again during the recovery process, the system stops recovery and returns to reconfiguration mode or safety hold mode. This recovery process prevents the nozzle from suddenly accelerating due to a brief fault resolution.

[0124] During continuous spraying operations, the robot may need to traverse pipe bends, welds, flange joints, and areas of localized corrosion. For bends and weld areas, the reliability of actual pipe wall disturbances may increase, but the ranging synchronization flag and servo residuals may not necessarily be abnormal. In this case, the controller can reduce the tangential velocity while maintaining the ranging closed loop. For flange joints or areas of localized corrosion, the distance residual, load residual, and servo residual may all increase simultaneously. In this case, the controller enters a reconfiguration mode, increasing the damping of relevant joints and reducing the tangential velocity to prevent the nozzle from rapidly approaching the pipe wall.

[0125] In anomalies caused by spray hose traction, the distance measurement data may remain reliable, but the load residual of a certain joint continues to increase. Traditional distance limiting methods struggle to identify such anomalies because the nozzle-to-wall distance has not yet undergone abrupt changes. This application establishes the joint servo link degradation reliability through load residuals and servo residuals, enabling the system to adjust the joint speed contribution before distance loss of control. Thus, fault detection is not triggered only after a spray distance anomaly occurs, but rather enters a guard or refactoring mode as soon as signs of degradation appear in the joint control link.

[0126] In cases where the ranging sensor is affected by paint atomization and reflection, the distance residual may suddenly increase, but the drive load and joint feedback do not support a corresponding amplitude of real movement of the nozzle. At this point, the reliability of the ranging link failure increases, the control reconfiguration module reduces the ranging weight, and uses a reliable and safe predicted distance buffer and kinematic prediction results to maintain short-term control. If ranging synchronization is restored and the residual decreases, the system gradually restores the ranging weight; if ranging synchronization is continuously lost, the system enters a safe hold state and performs a rollback.

[0127] In cases of joint actuator overheating or feedback loss anomalies, the joint servo residual may continue to rise with increasing drive temperature or load. If the actuator can provide temperature status, this temperature status can be used as an implementation of the drive load status; if the actuator cannot provide temperature status, the load residual and servo residual can be formed using only current and feedback speed. Neither implementation alters the main flow from process residual fingerprint to fault confidence to control reconfiguration parameters.

[0128] During multi-shift operations, the system can re-establish the baseline window before each entry into a new pipe section, or refresh the baseline residual set within the same pipe section according to a fixed work length. During refresh, existing control logic is not deleted; instead, the threshold set is updated to match the thresholds with the current spraying material, pipe surface condition, and sensor status. If the baseline residual is found to be higher than the maintenance threshold during the refresh process, the system can prompt maintenance personnel to check the distance sensor, nozzle mounting base, or joint actuator; however, this prompt does not affect the validity of the robotic arm control method as defined in the claims.

[0129] The above embodiments illustrate that the main solution of this application can cover various operating conditions such as ranging link failure, joint servo degradation, hose traction disturbance, and real pipe wall disturbance. Each operating condition enters the robotic arm control link through the same process residual fingerprint, the same fault and disturbance confidence level, and the same reconstructed control parameters; therefore, a single technical concept is maintained between different embodiments. The original distance prediction and amplitude limiting control path is still fully retained in the normal mode and secondary independent methods, while the newly added degradation perception scheme provides fault source differentiation and control quantity reconstruction capabilities on this basis.

[0130] In this recovery embodiment, all mode switching is based on the control cycle number, thus maintaining the same timing as the aforementioned state packet, residual fingerprint, and reconstructed control parameters.

[0131] This application can also be implemented as a computer program product. A computer program is stored in a computer-readable storage medium, which, when executed by a processor, can implement any of the method steps in this embodiment. The storage medium may include a read-only memory, a random access memory, a mobile storage medium, a solid-state memory, or other media capable of storing programs.

[0132] The above embodiments are used to illustrate the technical solutions of this application, and are not intended to limit the scope of protection of this application. Those skilled in the art can replace or adjust the specific implementation methods, threshold value ranges, window lengths and interface forms of each module without departing from the technical concept of this application, and such replacements or adjustments should all fall within the scope of protection of this application.

[0133] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0134] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for controlling a robotic arm of a large-diameter pipe spraying robot with degradation sensing, characterized in that, include: The original control state and online monitoring state of the painting robot arm are obtained, and a synchronization flag is generated based on the timestamp of each state acquisition. Based on the original control state, distance prediction limiting control is performed to generate a first matrix, a safe predicted distance, and a basic end target linear velocity for establishing the mapping between end linear velocity and joint angular velocity. Based on the original control state, the online monitoring state, and the synchronization flag, a process residual fingerprint is generated, including distance residual, echo quality residual, kinematic closure residual, servo residual, load residual, pressure disturbance residual, and synchronization flag. Based on the process residual fingerprint, the fault, disturbance confidence level, and degradation level are determined, and reconstruction control parameters are generated accordingly. Based on the basic end target linear velocity, the first matrix, the safe predicted distance, and the reconstructed control parameters, the target joint angular velocity is determined, and the movement of the spraying robot arm is controlled according to the target joint angular velocity.

2. The method according to claim 1, characterized in that, Before obtaining the original control state and online monitoring state of the painting robot arm, the process also includes: Low-speed baseline operation is performed within the large-diameter pipeline. Collect ranging readings, ranging echo quality, joint feedback status, drive load status, and nozzle pressure status during the baseline operation period; Based on the ranging readings, the ranging echo quality, the joint feedback state, the drive load state, and the nozzle pressure state, a threshold set, a reference load state, and a reference pressure state are determined for generating the process residual fingerprint.

3. The method according to claim 2, characterized in that, The process residual fingerprint includes distance residual, echo quality residual, kinematic closure residual, servo residual, and load residual, as well as pressure disturbance residual and synchronization flag; The distance residual is obtained based on the current actual distance and the predicted distance generated by the distance prediction limiting control; The echo quality residual is obtained based on the ranging echo quality. The kinematic closure residual is obtained based on the current actual distance, historical actual distance, the first matrix, and the joint velocity command of the previous cycle; The servo residual is obtained based on the joint speed command and joint feedback speed of the previous cycle; The load residual is obtained based on the driving load state and the reference load state; The pressure disturbance residual is obtained based on the nozzle pressure state and the reference pressure state; The synchronization flag is used to indicate whether each monitoring state meets the synchronization conditions of the same control cycle.

4. The method according to claim 3, characterized in that, Based on the residual fingerprint of the process, the confidence level of the fault and disturbance, and the degradation level are determined, including: Window statistics are performed on the process residual fingerprints over multiple consecutive control cycles; Based on the window statistics results, the reliability of ranging link failure, pipe wall real disturbance, and hose disturbance are determined respectively. The reliability of joint servo link degradation is determined based on the window statistics results. The degradation level is determined based on the confidence interval in which each confidence level falls.

5. The method according to claim 4, characterized in that, The reconstructed control parameters include ranging weight, tangential velocity coefficient, joint space damping length, and control mode; The ranging weight is used to adjust the influence of the ranging closed loop on the normal control quantity; The tangential velocity coefficient is used to adjust the spraying propulsion speed; The joint space damping length is used to construct the joint space damping matrix to adjust the joint velocity contribution in the target joint angular velocity solution. The control modes include at least normal mode, guardian mode, refactoring mode, and security maintenance mode.

6. A method for predicting and controlling the distance of a robotic arm in a large-diameter pipe spraying robot, characterized in that, Applications include large-diameter pipe spraying robots equipped with spraying robotic arms, including: The current joint angle and current actual distance of the spraying robot arm are obtained, and the historical actual distance, first delay time, target distance, target tangential velocity and pre-calibrated pipe center axis are read. Based on the current joint angle, determine the current end position of the painting robot arm and a first matrix for establishing the mapping relationship between the end linear velocity and the joint angular velocity; The predicted distance is determined based on the current actual distance, the historical actual distance, and the first delay time; Based on the maximum withstand acceleration at the end of the spraying robot arm and the first delay time, a first displacement boundary is determined, and the predicted distance is limited according to the first displacement boundary to generate a safe predicted distance. Based on the predicted safety distance, the target distance, and the target tangential velocity, the end target linear velocity of the spraying robot arm is generated; Based on the first matrix and the target linear velocity of the end effector, the target joint angular velocity for driving the joint movement of the spraying robot arm is determined, and the movement of the spraying robot arm is controlled according to the target joint angular velocity.

7. The method according to claim 6, characterized in that, Generating the safe predicted distance includes: Based on the current actual distance and the historical actual distance, determine the current rate of change of the actual distance; The predicted distance is obtained based on the current rate of change, the first delay time, and the current actual distance; The first displacement boundary is determined based on the maximum withstand acceleration at the end of the spraying robotic arm and the first delay time. Based on the first displacement boundary, the predicted distance is limited to obtain the safe predicted distance.

8. The method according to claim 6, characterized in that, Generating the terminal target linear velocity includes: Based on the predicted safety distance and the target distance, the normal distance error is determined; Based on the normal distance error, determine the normal compensation speed; Based on the current end position and the pre-calibrated pipeline centerline, determine the normal unit vector and the tangential unit vector; The terminal target linear velocity is obtained based on the normal compensation velocity, the normal unit vector, the target tangential velocity, and the tangential unit vector.

9. A control system for a large-diameter pipe spraying robot arm with degradation sensing, characterized in that, Applications include large-diameter pipe spraying robots equipped with spraying robotic arms; The control system includes: The state perception module is used to acquire the original control state and online monitoring state of the painting robot arm, and generate a synchronization flag based on the timestamp of each state acquisition. A distance prediction control module is configured to execute the distance prediction control method according to any one of claims 6 to 8, and to provide the joint control module with a first matrix, a safe predicted distance, and a basic end target linear velocity; The residual fingerprint generation module is used to generate a residual fingerprint based on the original control state, the online monitoring state, and the synchronization flag. A credibility determination module is used to determine the credibility of faults and disturbances and the degradation level based on the process residual fingerprint; The control reconfiguration module is used to generate reconfiguration control parameters based on the fault and disturbance confidence level and the degradation level. The joint control module is used to determine the target joint angular velocity based on the basic end target linear velocity, the first matrix, the safe prediction distance, and the reconstructed control parameters, and to control the movement of the spraying robot arm.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program; When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 8.

Images