Curtain cleaning quadrotor unmanned aerial vehicle attitude instability fault detection and fault-tolerant control system

Abstract

The invention discloses a curtain cleaning four-rotor unmanned aerial vehicle attitude instability fault detection and fault-tolerant control system, which comprises a multi-source sensor fusion perception module, a recoil force dynamic modeling and identification module, an attitude instability fault feature extraction and classification module, and a hierarchical fault-tolerant control decision module. The unmanned aerial vehicle attitude, motor state and cleaning system parameters are collected in real time; the recoil force dynamic model is constructed based on Bernoulli equation and momentum theorem to generate disturbance observation values; multi-dimensional fault features are extracted and compared with the preset feature library to distinguish whether the attitude instability is caused by external strong disturbance or internal true fault; if it is external, the disturbance active compensation control mode is started, if it is internal, the degradation safety operation and controllable return mode is started, and if it is a sensor, the multi-source information fusion reconstruction mode is started. The invention effectively solves the problems that the curtain cleaning unmanned aerial vehicle is difficult to distinguish disturbance and fault in complex operation environment, lacks targeted fault-tolerant strategy, and significantly improves the operation safety and task robustness.

Description

Technical Field

[0001] This invention relates to the fields of unmanned aerial vehicle (UAV) control technology and building cleaning and maintenance technology, specifically to a fault-tolerant control system for attitude instability detection of a quadcopter UAV used for curtain wall cleaning. Background Technology

[0002] With the rapid development of the low-altitude economy, using drones to replace traditional "spider-men" for high-rise building curtain wall cleaning has become an important industry trend. Quadcopter drones, due to their compact structure, flexible operation, and vertical take-off and landing capabilities, have become the mainstream choice in the curtain wall cleaning field. However, in actual operations, these drones face severe technical challenges.

[0003] On the one hand, during cleaning operations, the continuous recoil force generated by the high-pressure water jet, combined with the interference of high-altitude gusts, can easily cause severe oscillations in the attitude of the drone. Existing research shows that recoil force can cause pitch / roll angle fluctuations to exceed ±5°, and hovering accuracy to plummet to below 0.5 meters. This not only seriously affects the cleaning effect but may also lead to safety accidents such as collisions and crashes.

[0004] On the other hand, under long-term, heavy-load operating conditions, critical components such as motors, ESCs, propellers, or inertial measurement units are prone to performance degradation or sudden failures. Existing commercial flight control systems lack dedicated fault diagnosis and fault-tolerance strategies for "clean operating conditions," making it impossible to effectively distinguish between recoil disturbances and genuine component failures, resulting in a high misjudgment rate. When a failure occurs, the system often directly triggers a return to base or crash, failing to complete the current task or conduct controlled emergency response while ensuring safety.

[0005] Existing technologies such as CN121232865A (a smart cleaning drone for high-rise building curtain wall maintenance and its cleaning method) and CN110239731B (a multi-functional drone system for suspended building exterior wall operations) both improve the drone from a structural perspective, but do not disclose specific fault diagnosis and fault tolerance methods for "cleaning conditions".

[0006] Therefore, developing a system and method that can accurately distinguish between external disturbances and internal faults, and can perform targeted fault-tolerant control according to the fault type, is of great significance for improving the operational safety and mission robustness of curtain wall cleaning drones. Summary of the Invention

[0007] 1. The technical problem to be solved:

[0008] To address the aforementioned technical problems, this invention provides a fault detection and fault-tolerant control system for attitude instability of a quadcopter drone used for curtain wall cleaning.

[0009] 2. Technical Solution:

[0010] A fault-tolerant control system for attitude instability detection of a quadcopter drone used for curtain wall cleaning includes a multi-source sensor fusion sensing module, a recoil force dynamic modeling and identification module, an attitude instability fault feature extraction and classification module, and a hierarchical fault-tolerant control decision module.

[0011] The multi-source sensor fusion sensing module includes an inertial measurement unit for acquiring the real-time attitude angle, angular velocity, and acceleration motor speed of the quadcopter drone; a visual sensor and a lidar for acquiring the relative distance to the curtain wall and obstacle avoidance information; and a pressure / flow sensor for monitoring the status of the cleaning system.

[0012] The dynamic modeling and identification module for recoil force acquires data from pressure / flow sensors and calculates the expected recoil force based on Bernoulli's equation and momentum theorem, taking into account the real-time pump pressure and nozzle diameter parameters. The dynamic response characteristics of the sensor generate real-time recoil force disturbance observations. The attitude instability fault feature extraction and classification module receives data from the multi-source sensor fusion sensing module and the recoil force dynamic modeling and identification module, calculates attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay characteristics, and compares them with the preset disturbance fault feature library to distinguish whether the current attitude instability is caused by external strong disturbance or by a real fault of internal actuator or sensor.

[0013] The hierarchical fault-tolerant control decision module executes the corresponding fault-tolerant control strategy based on the fault classification results of the attitude instability fault feature extraction and classification module; the fault classification includes external strong disturbance faults, internal true faults, and sensor true faults.

[0014] Furthermore, the generation of real-time recoil force disturbance observations in the recoil force dynamic modeling and identification module specifically includes the following process:

[0015] Substituting the real-time nozzle inlet pressure, nozzle outlet pressure, and nozzle outlet diameter detected by the pressure / flow sensor into the following formula for steady-state recoil force... The formula yields the steady-state recoil force:

[0016] Where d2 is the nozzle outlet diameter; P is the nozzle inlet pressure; P2 is the nozzle outlet pressure;

[0017] Substituting the steady-state recoil force into the dynamic recoil force in the following equation... The calculation formula yields the real-time recoil force disturbance observation value:

[0018] Where T is a preset transition time constant.

[0019] Furthermore, the preset disturbance fault feature library categorizes disturbance fault features into external strong disturbance features, internal true fault features, and sensor true fault features based on fault classification, thereby constructing parameter ranges for external strong disturbance features, internal true fault features, and sensor true fault features. The external strong disturbance feature parameter range includes a first parameter set of recoil force-attitude response delay, which is the range of time it takes for the magnitude of the recoil force to react with the attitude angle, angular velocity, acceleration, and motor speed when a fault occurs. The internal true fault feature parameter range includes a second parameter set of parameters where the deviation between the single-axis motor speed command and the actual feedback exceeds a preset threshold. The sensor true fault feature parameter range includes the range between the attitude angle error, angular velocity change rate, and motor speed deviation features when a fault occurs and the preset corresponding feature threshold. When the collected data falls within the data range of the disturbance fault features in the disturbance fault feature library, the fault type can be determined.

[0020] Furthermore, the hierarchical fault-tolerant control decision module includes:

[0021] The disturbance active compensation control submodule is activated when a strong external disturbance is detected. Based on backstep control and Lyapunov stability theory, it uses the observed value of the recoil force disturbance as a feedforward compensation term to input the attitude controller and correct the motor speed distribution in real time.

[0022] The downgraded safe operation and controllable return submodule is activated when an internal genuine fault is detected. It reconstructs the control torque through a nonlinear control distribution algorithm, sends instructions to the cleaning system to reduce or cut off the recoil force source, and plans a safe return path.

[0023] The multi-source information fusion and reconstruction submodule is activated when a genuine sensor fault is determined. It uses visual SLAM, lidar point cloud data, and motor speed feedback to perform state estimation through extended Kalman filtering, temporarily replacing the data from the failed sensor.

[0024] Furthermore, its control method includes the following steps:

[0025] Step 1: Real-time acquisition of quadcopter drone attitude data, motor status data, and cleaning system pressure / flow data through a multi-source sensor fusion sensing module;

[0026] Step 2: Using the dynamic modeling and identification module for recoil force, the expected recoil force is calculated based on Bernoulli's equation and the momentum theorem, according to the real-time pressure at the nozzle inlet, the real-time pressure at the nozzle outlet, and the nozzle diameter parameters. Based on its dynamic response characteristics, real-time recoil force disturbance observations are generated;

[0027] Step 3: Through the attitude instability fault feature extraction and classification module, calculate the attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay characteristics, and compare them with the preset disturbance fault feature library to distinguish whether the current attitude instability is caused by strong external disturbance or by a real fault of internal actuator or sensor.

[0028] Step 4: Execute the corresponding fault-tolerant control strategy based on the fault classification results through the hierarchical fault-tolerant control decision module.

[0029] Furthermore, in step four, when a strong external disturbance is detected, the active disturbance compensation control mode is activated: based on backstep control and Lyapunov stability theory, the observed value of the recoil force disturbance is used as a feedforward compensation term input to the attitude controller to correct the motor speed distribution and nozzle angle in real time, so that the attitude angle fluctuation is controlled within ±0.5° and the hovering accuracy is restored to ±0.1m.

[0030] Furthermore, in step four, when a genuine failure of the internal actuator is determined, a degraded safety operation and controllable return mode is initiated: the control torque is reconstructed through a nonlinear control allocation algorithm, and the remaining healthy motors are activated to maintain the basic attitude stability of the fuselage; at the same time, instructions are sent to the cleaning system to reduce the water pump output pressure to below the preset safety value, and a safe descent and return path is planned.

[0031] Furthermore, in step four, when a genuine sensor failure is determined, a multi-source information fusion reconstruction mode is initiated: using visual SLAM, lidar point cloud data, and motor speed feedback, state estimation is performed through extended Kalman filtering to temporarily replace the failed sensor data and maintain stable flight for a short period of time to perform emergency recovery.

[0032] Furthermore, the quadcopter UAV adopts a six-axis, twelve-motor architecture; the inertial measurement unit consists of BMI088 and ADIS16470, which form a high-performance sensor array with a gyroscope range of ±2000° / s and zero-bias stability of 8° / h.

[0033] 3. Beneficial effects:

[0034] (1) The present invention discloses a fault detection and fault-tolerant control system for attitude instability of a quadcopter UAV for curtain wall cleaning. It decouples external operational disturbances from internal component faults, effectively avoids misjudgment or missed judgment caused by the inability of traditional flight control systems to distinguish between disturbances and faults, and significantly reduces the false alarm rate.

[0035] (2) The present invention discloses a fault detection and fault-tolerant control system for attitude instability of a quadcopter UAV for curtain wall cleaning. It adopts active feedforward compensation control for strong external disturbances. Based on backstep control and Lyapunov stability theory, the observed value of the recoil force disturbance is used as the feedforward compensation term input to the attitude controller, so that the UAV has the ability to operate stably under harsh conditions such as wind of level 6-7 and high pressure recoil. The attitude angle fluctuation is reduced by an order of magnitude, and the operation efficiency and coverage are greatly improved.

[0036] (3) The present invention discloses a fault detection and fault-tolerant control system for attitude instability of a quadcopter UAV for curtain wall cleaning. It adopts a layered fault-tolerant strategy for internal true faults, reconstructs the control torque through a nonlinear control allocation algorithm, prioritizes flight safety and avoids catastrophic accidents; at the same time, it cuts off the source of danger by degrading the cleaning operation mode, wins valuable time for emergency response, and significantly improves the mission robustness of the UAV. Attached Figure Description

[0037] Figure 1 This is a block diagram of the attitude instability fault detection and fault-tolerant control system for the curtain wall cleaning quadcopter drone of the present invention;

[0038] Figure 2 This is a control block diagram of the hierarchical fault-tolerant control decision module in this invention;

[0039] Figure 3 This is a flowchart illustrating the control method implemented by the control system of the present invention. Detailed Implementation

[0040] The present invention will now be described in detail with reference to the accompanying drawings.

[0041] As attached Figure 1 To be continued Figure 3 As shown, a fault detection and fault-tolerant control system for attitude instability of a quadcopter drone used for curtain wall cleaning includes a multi-source sensor fusion sensing module, a recoil force dynamic modeling and identification module, an attitude instability fault feature extraction and classification module, and a hierarchical fault-tolerant control decision module.

[0042] The multi-source sensor fusion sensing module includes an inertial measurement unit for acquiring the real-time attitude angle, angular velocity, and acceleration motor speed of the quadcopter drone; a visual sensor and a lidar for acquiring the relative distance to the curtain wall and obstacle avoidance information; and a pressure / flow sensor for monitoring the status of the cleaning system.

[0043] The dynamic modeling and identification module for recoil force acquires data from pressure / flow sensors and calculates the expected recoil force based on Bernoulli's equation and momentum theorem, taking into account the real-time pump pressure and nozzle diameter parameters. The dynamic response characteristics of the sensor generate real-time recoil force disturbance observations. The attitude instability fault feature extraction and classification module receives data from the multi-source sensor fusion sensing module and the recoil force dynamic modeling and identification module, calculates attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay characteristics, and compares them with the preset disturbance fault feature library to distinguish whether the current attitude instability is caused by external strong disturbance or by a real fault of internal actuator or sensor.

[0044] The hierarchical fault-tolerant control decision module executes the corresponding fault-tolerant control strategy based on the fault classification results of the attitude instability fault feature extraction and classification module; the fault classification includes external strong disturbance faults, internal true faults, and sensor true faults.

[0045] Furthermore, the generation of real-time recoil force disturbance observations in the recoil force dynamic modeling and identification module specifically includes the following process:

[0046] Substituting the real-time nozzle inlet pressure, nozzle outlet pressure, and nozzle outlet diameter detected by the pressure / flow sensor into the following formula for steady-state recoil force... The formula yields the steady-state recoil force:

[0047] Where d2 is the nozzle outlet diameter; P is the nozzle inlet pressure; P2 is the nozzle outlet pressure;

[0048] Substituting the steady-state recoil force into the dynamic recoil force in the following equation... The calculation formula yields the real-time recoil force disturbance observation value:

[0049] Where T is a preset transition time constant.

[0050] Furthermore, the preset disturbance fault feature library categorizes disturbance fault features into external strong disturbance features, internal true fault features, and sensor true fault features based on fault classification, thereby constructing parameter ranges for external strong disturbance features, internal true fault features, and sensor true fault features. The external strong disturbance feature parameter range includes a first parameter set of recoil force-attitude response delay, which is the range of time it takes for the magnitude of the recoil force to react with the attitude angle, angular velocity, acceleration, and motor speed when a fault occurs. The internal true fault feature parameter range includes a second parameter set of parameters where the deviation between the single-axis motor speed command and the actual feedback exceeds a preset threshold. The sensor true fault feature parameter range includes the range between the attitude angle error, angular velocity change rate, and motor speed deviation features when a fault occurs and the preset corresponding feature threshold. When the collected data falls within the data range of the disturbance fault features in the disturbance fault feature library, the fault type can be determined.

[0051] Furthermore, the hierarchical fault-tolerant control decision module includes:

[0052] The disturbance active compensation control submodule is activated when a strong external disturbance is detected. Based on backstep control and Lyapunov stability theory, it uses the observed value of the recoil force disturbance as a feedforward compensation term to input the attitude controller and correct the motor speed distribution in real time.

[0053] The downgraded safe operation and controllable return submodule is activated when an internal genuine fault is detected. It reconstructs the control torque through a nonlinear control distribution algorithm, sends instructions to the cleaning system to reduce or cut off the recoil force source, and plans a safe return path.

[0054] The multi-source information fusion and reconstruction submodule is activated when a genuine sensor fault is determined. It uses visual SLAM, lidar point cloud data, and motor speed feedback to perform state estimation through extended Kalman filtering, temporarily replacing the data from the failed sensor.

[0055] Furthermore, its control method includes the following steps:

[0056] Step 1: Real-time acquisition of quadcopter drone attitude data, motor status data, and cleaning system pressure / flow data through a multi-source sensor fusion sensing module;

[0057] Step 2: Using the dynamic modeling and identification module for recoil force, the expected recoil force is calculated based on Bernoulli's equation and the momentum theorem, according to the real-time pressure at the nozzle inlet, the real-time pressure at the nozzle outlet, and the nozzle diameter parameters. Based on its dynamic response characteristics, real-time recoil force disturbance observations are generated;

[0058] Step 3: Through the attitude instability fault feature extraction and classification module, calculate the attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay characteristics, and compare them with the preset disturbance fault feature library to distinguish whether the current attitude instability is caused by strong external disturbance or by a real fault of internal actuator or sensor.

[0059] Step 4: Execute the corresponding fault-tolerant control strategy based on the fault classification results through the hierarchical fault-tolerant control decision module.

[0060] Furthermore, in step four, when a strong external disturbance is detected, the active disturbance compensation control mode is activated: based on backstep control and Lyapunov stability theory, the observed value of the recoil force disturbance is used as a feedforward compensation term input to the attitude controller to correct the motor speed distribution and nozzle angle in real time, so that the attitude angle fluctuation is controlled within ±0.5° and the hovering accuracy is restored to ±0.1m.

[0061] Furthermore, in step four, when a genuine failure of the internal actuator is determined, a degraded safety operation and controllable return mode is initiated: the control torque is reconstructed through a nonlinear control distribution algorithm, and the remaining healthy motors are activated to maintain the basic attitude stability of the fuselage (usually preset to: pitch / roll angle ≤ 5°); at the same time, a command is sent to the cleaning system to reduce the water pump output pressure to below the preset safety value, and a safe descent and return path is planned.

[0062] Furthermore, in step four, when a genuine sensor failure is determined, a multi-source information fusion reconstruction mode is initiated: using visual SLAM, lidar point cloud data, and motor speed feedback, state estimation is performed through extended Kalman filtering to temporarily replace the failed sensor data and maintain stable flight for a short period of time to perform emergency recovery.

[0063] Furthermore, the quadcopter UAV adopts a six-axis, twelve-motor architecture; the inertial measurement unit consists of BMI088 and ADIS16470, which form a high-performance sensor array with a gyroscope range of ±2000° / s and zero-bias stability of 8° / h.

[0064] Example 1: Active Compensation Control under Strong External Disturbances

[0065] This embodiment utilizes the present invention and verifies it in a glass curtain wall cleaning task of a super high-rise commercial complex. The drone adopts a "six-axis, twelve-motor" architecture, equipped with a BMI088 and ADIS16470 combined inertial measurement unit, as well as pressure / flow sensors.

[0066] After the drone enters hovering operation mode, the pressure / flow sensor detects that the water pump has started, and the recoil force dynamic modeling module uses the formula... The steady-state recoil force was calculated to be approximately 200 N, and its dynamic response time was estimated. Simultaneously, the attitude sensor detected a pitch angle fluctuation of ±3.2° due to the superimposed force of a Force 7 gust.

[0067] The attitude instability fault feature extraction module compares the current attitude angle error, angular velocity change rate, and motor speed deviation with the expected disturbance output by the recoil force modeling module. The calculations show that the motor speed response is highly correlated with the expected recoil force value, and there is no abnormal saturation of the single-axis motor speed. Based on this, the fault classification module determines it to be an "external strong disturbance."

[0068] The hierarchical fault-tolerant control decision module then activates the "disturbance active compensation control mode." The core controller, based on a backstepping control framework, processes the recoil force observations... As a feedforward, it is superimposed on the attitude control law in real time to dynamically adjust the speed distribution of the six-axis twelve-motor system. At the same time, based on the distance error with the curtain wall fed back by the lidar, the nozzle angle is finely adjusted to balance the direction of the recoil force vector with the force on the fuselage.

[0069] Ultimately, despite the continuous strong winds and high pressure backlash, the drone's attitude angle fluctuations were stably controlled within ±0.4°, with a hovering accuracy of ±0.08m and a daily cleaning efficiency of 8000㎡, ensuring the continuity and coverage of the cleaning operation.

[0070] Example 2: Fault-tolerant control under true failure of internal actuators

[0071] This embodiment utilizes the present invention and verifies it in a simulated curtain wall cleaning operation scenario. During normal operation of the drone, the flight control and monitoring module detected that the deviation between the speed command and the actual feedback of motor No. 1 was consistently greater than 15%, accompanied by the drone rapidly tilting towards the direction of that motor.

[0072] The attitude instability fault feature extraction module compared the healthy motor status with the observed recoil force interference values ​​and found that other motors responded normally to the recoil force compensation at normal speeds, while motor No. 1 showed obvious failure characteristics. The fault classification module determined it to be a "true fault of the internal actuator".

[0073] The system immediately switched to "degraded safe operation and controlled return mode." First, the control allocation algorithm was restructured. Based on the achievable torque set of the quadcopter's "X" configuration under single-axis failure, the system calculated in real time the rotational speed commands of the remaining three healthy motors to maintain the fuselage's basic level (pitch / roll angle ≤ 5°). Simultaneously, commands were sent to the cleaning system to reduce the water pump output pressure from 8MPa to 2MPa and shut down the cleaning agent spray, significantly reducing the recoil force source. The flight controller then planned a safe, backward-and-upward descent return path and controlled the drone to return to the takeoff point at a speed of 2 m / s.

[0074] The aircraft maintained a stable posture throughout the entire process, without any loss of control, rollover, or fall, and successfully achieved safe recovery in a fault condition.

[0075] Example 3: Multi-source information fusion and reconstruction under true sensor faults

[0076] The system and method described in this invention were verified in a simulated environment. During the flight of the UAV, the attitude angle data output by the IMU module exhibited abnormal jumps, showing significant deviations from the observations of visual SLAM and lidar.

[0077] The attitude instability fault feature extraction module detected that the fusion residual between IMU data and visual / LiDAR data exceeded a preset threshold, and the motor speed feedback did not match the expected attitude response. The fault classification module determined it to be a "genuine sensor fault".

[0078] The system immediately activated the "multi-source information fusion reconstruction mode." The extended Kalman filter observer used the attitude information calculated by visual SLAM and the relative pose information obtained by matching the lidar point cloud as the primary state estimation source, combined with the dynamic model based on motor speed feedback, to temporarily replace the failed IMU data. The system maintained stable flight for approximately 60 seconds in reconstruction mode, successfully executed the emergency recovery procedure, and returned to the takeoff point.

[0079] Comparative Example 1

[0080] The curtain wall cleaning operation is performed using a traditional quadcopter drone flight control system, excluding the recoil dynamic modeling and identification module and the hierarchical fault-tolerant control decision module of this invention. The drone is equipped with a conventional IMU sensor and uses a traditional PID control algorithm. During the cleaning operation, the water pump pressure is 8MPa, and the nozzle diameter is the same as in Example 1. The operation includes the following steps: (1) real-time acquisition of drone attitude data through the IMU sensor; (2) the flight control system directly calculates the motor speed compensation based on the attitude error; (3) curtain wall cleaning is performed in a wind disturbance environment of level 6-7. The remaining conditions are the same as in Example 1, and will not be repeated here.

[0081] Comparative Example 2

[0082] A system with only fault detection functionality but no hierarchical fault-tolerant control is used for curtain wall cleaning. This system includes the recoil force dynamic modeling and identification module and the attitude instability fault feature extraction and classification module of this invention, capable of identifying external disturbances and internal faults, but lacking a hierarchical fault-tolerant control decision module. When attitude instability is detected, the system only issues an alarm signal, and attitude control is still performed by a traditional PID controller, without active feedforward compensation or control allocation reconfiguration functions. The remaining conditions are the same as in Example 1, and will not be repeated here.

[0083] Result detection

[0084] The fault detection and fault-tolerant control effects achieved in Examples 1-3 and Comparative Examples 1-2 were evaluated, and the results are shown in Table 1.

[0085] Table 1. Evaluation of the Fault Detection and Fault-Tolerant Control Effects of the Invention and Comparative Examples

[0086]

[0087] Comparative analysis

[0088] Comparative Example 1 did not operate according to the specific implementation method of this invention and did not include the recoil force dynamic modeling module and the hierarchical fault-tolerant control module. Experimental results show that under strong external disturbances, traditional PID control cannot effectively suppress the coupling effect of recoil force and wind disturbance, the attitude angle fluctuation exceeds ±8°, the hovering accuracy drops significantly, resulting in a cleaning coverage rate of less than 70%; when an internal actuator failure occurs, the system cannot identify the fault type and directly loses control and crashes, with an equipment damage rate as high as 28.3%, significantly higher than that of the embodiment of this invention.

[0089] Comparative Example 2 includes the fault identification module of this invention, which can accurately distinguish between external disturbances and internal faults. However, due to the lack of a hierarchical fault-tolerant control module, it cannot execute targeted fault-tolerant strategies. Under external disturbances, it still uses traditional PID control, resulting in attitude angle fluctuations exceeding ±6.5° and a task completion rate of only 42.6%. When an internal fault occurs, although the fault can be identified, it cannot reconstruct the control torque, ultimately leading to loss of control and crash, with a damage rate of 22.1%.

[0090] Based on the above performance comparison, it can be seen that the quadcopter UAV attitude instability fault detection and fault-tolerant control system and method proposed in this invention for curtain wall cleaning operations has the following significant advantages compared with the comparative example that does not contain the key module of this invention:

[0091] (1) High fault identification accuracy: By introducing a dynamic model of recoil force, external disturbances and internal faults can be effectively distinguished. The identification accuracy is more than 20% higher than that of traditional systems, which is comparable to systems that only detect faults without fault tolerance.

[0092] (2) Excellent fault-tolerant control: Targeted fault-tolerant strategies are implemented for different types of faults. High-precision and stable operation can be maintained under strong external disturbances, and flight safety can be guaranteed under internal true faults. It is significantly better than the comparison scheme with no fault tolerance or only detection function.

[0093] (3) Strong task robustness: The task completion rate under various fault scenarios is significantly higher than that of the comparison, which reflects the comprehensive advantages of the "detection + fault tolerance" integrated architecture of the present invention.

[0094] Based on the above performance comparison, it can be seen that the quadcopter UAV attitude instability fault detection and fault-tolerant control system and method proposed in this invention for curtain wall cleaning operations has the following significant advantages:

[0095] (1) High fault identification accuracy: By introducing a dynamic model of recoil force, external disturbances and internal faults can be effectively distinguished, and the identification accuracy is improved by more than 20% compared with traditional systems.

[0096] (2) Excellent fault-tolerant control: Targeted fault-tolerant strategies are implemented for different types of faults. High-precision and stable operation can be maintained under strong external disturbances, and flight safety can be guaranteed under internal true faults. The equipment damage rate is reduced to 0%.

[0097] (3) Strong task robustness: The task completion rate under various failure scenarios is significantly higher than that of traditional systems, which improves the reliability and economy of UAV curtain wall cleaning operations.

[0098] This invention can meet the long-term operation requirements of curtain wall cleaning drones in harsh marine environments and has broad development prospects in smart city building operation and maintenance and low-altitude economic industry applications.

[0099] Although the present invention has been disclosed above with reference to preferred embodiments, these are not intended to limit the present invention. Any person skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be defined by the scope of the claims of this application.

Claims

1. A fault detection and fault-tolerant control system for attitude instability of a quadcopter drone used for curtain wall cleaning, characterized in that: It includes a multi-source sensor fusion sensing module, a recoil force dynamic modeling and identification module, an attitude instability fault feature extraction and classification module, and a hierarchical fault-tolerant control decision module; The multi-source sensor fusion sensing module includes an inertial measurement unit for acquiring the real-time attitude angle, angular velocity, and acceleration motor speed of the quadcopter drone; a visual sensor and a lidar for acquiring the relative distance to the curtain wall and obstacle avoidance information; and a pressure / flow sensor for monitoring the status of the cleaning system. The dynamic modeling and identification module for recoil force acquires data from pressure / flow sensors and calculates the expected recoil force based on Bernoulli's equation and momentum theorem, taking into account the real-time pump pressure and nozzle diameter parameters. Based on its dynamic response characteristics, real-time recoil force disturbance observations are generated; The attitude instability fault feature extraction and classification module receives data from the multi-source sensor fusion perception module and the recoil force dynamic modeling and identification module, calculates attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay features, and compares them with a preset disturbance fault feature library to distinguish whether the current attitude instability is caused by external strong disturbance or by a real fault of internal actuator or sensor. The hierarchical fault-tolerant control decision module executes the corresponding fault-tolerant control strategy based on the fault classification results of the attitude instability fault feature extraction and classification module. The fault classification includes external strong disturbance faults, internal true faults, and sensor true faults.

2. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 1, characterized in that: The recoil dynamic modeling and identification module specifically generates real-time recoil interference observations. Includes the following processes: Substituting the real-time nozzle inlet pressure, nozzle outlet pressure, and nozzle outlet diameter detected by the pressure / flow sensor into the following formula for steady-state recoil force... The formula yields the steady-state recoil force: Where d2 is the nozzle outlet diameter; P is the nozzle inlet pressure; P2 is the nozzle outlet pressure; Substituting the steady-state recoil force into the dynamic recoil force in the following equation... The calculation formula yields the real-time recoil force disturbance observation value: Where T is the preset transition time constant.

3. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 1, characterized in that: The preset disturbance fault feature library classifies disturbance fault features into external strong disturbance features, internal true fault features and sensor true fault features according to the fault classification, and then constructs the parameter range of external strong disturbance features, internal true fault features and sensor true fault features. The external strong disturbance characteristic parameter range includes a first parameter set of recoil force-attitude response delay; the recoil force-attitude response delay is the range of time during which the magnitude of the recoil force and the attitude angle, angular velocity, acceleration, and motor speed react when a fault occurs; the internal true fault characteristic parameter range includes a second parameter set of deviations between the single-axis motor speed command and the actual feedback exceeding a preset threshold; the characteristics of the sensor true fault characteristic parameter range include the size range between the attitude angle error, angular velocity change rate, and motor speed deviation characteristics when a fault occurs and the preset corresponding characteristic thresholds; when the collected data falls within the data range of the disturbance fault characteristic library, the fault type can be determined.

4. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 1, characterized in that: The hierarchical fault-tolerant control decision module includes: a disturbance active compensation control submodule, which is activated when a strong external disturbance is detected. Based on backstep control and Lyapunov stability theory, it uses the observed recoil force disturbance as a feedforward compensation term to input the attitude controller and corrects the motor speed distribution in real time; a degraded safe operation and controllable return submodule, which is activated when a genuine internal fault is detected. It reconstructs the control torque through a nonlinear control distribution algorithm, sends instructions to the cleaning system to reduce or cut off the recoil force source, and plans a safe return path; and a multi-source information fusion reconstruction submodule, which is activated when a genuine sensor fault is detected. It uses visual SLAM, lidar point cloud data, and motor speed feedback to perform state estimation through extended Kalman filtering, temporarily replacing the failed sensor data.

5. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 4, characterized in that: Its control method includes the following steps: Step 1: Real-time acquisition of quadcopter drone attitude data, motor status data, and cleaning system pressure / flow data through a multi-source sensor fusion sensing module; Step 2: Using the dynamic modeling and identification module for recoil force, the expected recoil force is calculated based on Bernoulli's equation and the momentum theorem, according to the real-time pressure at the nozzle inlet, the real-time pressure at the nozzle outlet, and the nozzle diameter parameters. Based on its dynamic response characteristics, real-time recoil force disturbance observations are generated; Step 3: Through the attitude instability fault feature extraction and classification module, calculate the attitude angle error, angular velocity change rate, motor speed deviation and recoil force-attitude response delay characteristics, and compare them with the preset disturbance fault feature library to distinguish whether the current attitude instability is caused by strong external disturbance or by a real fault of internal actuator or sensor. Step 4: Execute the corresponding fault-tolerant control strategy based on the fault classification results through the hierarchical fault-tolerant control decision module.

6. The attitude instability fault detection and fault-tolerant control system for a quadcopter drone used for curtain wall cleaning according to claim 5, characterized in that: In step four, when a strong external disturbance is detected, the active disturbance compensation control mode is activated: based on backstep control and Lyapunov stability theory, the observed value of the recoil force disturbance is used as a feedforward compensation term input to the attitude controller to correct the motor speed distribution and nozzle angle in real time, so that the attitude angle fluctuation is controlled within ±0.5° and the hovering accuracy is restored to ±0.1m.

7. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 6, characterized in that: In step four, when a genuine failure of the internal actuator is determined, a degraded safety operation and controllable return mode is initiated: the control torque is reconstructed through a nonlinear control distribution algorithm, the remaining healthy motors are activated to maintain the basic attitude stability of the fuselage; at the same time, instructions are sent to the cleaning system to reduce the water pump output pressure to below the preset safety value, and a safe descent and return path is planned.

8. The curtain wall cleaning quadcopter drone attitude instability fault detection and fault-tolerant control system according to claim 6, characterized in that: In step four, when a genuine sensor failure is determined, the multi-source information fusion reconstruction mode is activated: using visual SLAM, lidar point cloud data and motor speed feedback, state estimation is performed through extended Kalman filtering to temporarily replace the failed sensor data and maintain stable flight for a short period of time to perform emergency recovery.

9. The attitude instability fault detection and fault-tolerant control system for a quadcopter drone used for curtain wall cleaning according to claim 1, characterized in that: The quadcopter drone adopts a six-axis, twelve-motor architecture; the inertial measurement unit consists of BMI088 and ADIS16470, which form a high-performance sensor array with a gyroscope range of ±2000° / s and zero-bias stability of 8° / h.

Images