Data-driven UAV actuator fault location method and system

Abstract

The present invention provides a data-driven method and system for locating the fault of an actuator of a UAV, which is applied to a rotary-wing UAV; the rotary-wing UAV includes multiple actuators; the method includes: obtaining the acceleration residual of the rotary-wing UAV based on an extended state observer; the acceleration residual includes the pitch angle acceleration residual and the roll angle acceleration residual; based on the acceleration residual, using a trained deep forest model to locate and identify the fault of the actuator of the rotary-wing UAV. The present invention alleviates the technical problem that the fault diagnosis method relying on the model in the prior art is not accurate enough.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle fault diagnosis, and in particular to a data-driven unmanned aerial vehicle actuator fault locating method and system. Background Art

[0002] Quad-rotor drone is a kind of unmanned aerial device widely used in various fields. It is easy to operate, easy to carry and flexible, so it is widely used in military and civilian fields.

[0003] Quadcopter drones mainly rely on providing control signals to four motors to control the rotation of the motors, thereby driving the propellers of the quadcopter drone to rotate, thereby generating lift to complete the given flight instructions. The motors of quadcopter drones are often in an environment of high vibration, high temperature and long-term work, which usually leads to motor failures such as decreased transmission efficiency and failure to stop rotating. In addition, due to various external environmental factors, the propellers will often wear out, break and other faults. The failure of the propellers will also be reflected in the torque of the motor, so the propeller failure can be attributed to the motor failure. When the motor of a quadcopter drone fails, it often leads to the inability to complete the specified task, and even causes the drone to lose control and cause personal or property losses. Therefore, it is very necessary to propose a method for locating the motor fault of a quadcopter drone. A good fault location method can effectively and timely find the fault location and provide reference information for the subsequent fault-tolerant control of quadcopter drones.

[0004] At present, there are two main fault diagnosis methods for quadrotor UAV actuators: analytical model-based methods and data-driven methods. For analytical model-based methods, it is necessary to establish a complex observer model for the UAV system. At the same time, it is difficult to accurately model interference, noise and errors, and the design requirements for observer parameters are very high. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a data-driven UAV actuator fault location method and system to alleviate the technical problem of low accuracy of model-dependent fault diagnosis methods in the prior art.

[0006] In the first aspect, an embodiment of the present invention provides a data-driven UAV actuator fault location method, which is applied to a rotorcraft UAV; the rotorcraft UAV includes multiple actuators; the method includes: obtaining the acceleration residual of the rotorcraft UAV based on an extended state observer; the acceleration residual includes a pitch angle acceleration residual and a roll angle acceleration residual; based on the acceleration residual, using a trained deep forest model to perform fault location and fault identification on the actuator of the rotorcraft UAV.

[0007] Furthermore, it also includes: constructing the extended state observer based on the multiple actuators; wherein the model of the extended state observer includes: Where β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are observed quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement.

[0008] Furthermore, it also includes: using a simulation method to obtain the acceleration residual data of the multiple actuators under different faults to obtain a training set; training a preset deep forest model based on the training set to obtain the trained deep forest model.

[0009] Furthermore, a simulation method is used to obtain the acceleration residual data of the multiple actuators under different faults, including: constructing a kinematic and dynamic model of the rotorcraft; constructing a fault model of the multiple actuators based on the kinematic and dynamic model; injecting different faults into different actuators at different times based on the fault model; and combining an extended state observer with an accelerometer to obtain the acceleration residual data of the multiple actuators under different faults through a simulation method.

[0010] Furthermore, the fault model includes: a speed constant deviation fault model and a speed gain loss model; wherein the speed constant deviation fault model includes: ω(t) is the speed of the actuator under normal conditions, ω f (t) is the actual speed of the actuator, △ is the deviation speed between the actual value and the expected value of the actuator speed, t f is the moment when the speed deviation fault occurs in the actuator; the speed gain loss model includes: ρ is the actuator speed gain loss factor.

[0011] In the second aspect, an embodiment of the present invention also provides a data-driven UAV actuator fault location system, which is applied to a rotorcraft UAV; the rotorcraft UAV includes multiple actuators; the system includes: an acquisition module and a positioning module; wherein the acquisition module is used to acquire the acceleration residual of the rotorcraft UAV based on an extended state observer; the acceleration residual includes a pitch angle acceleration residual and a roll angle acceleration residual; the positioning module is used to perform fault location and fault identification on the actuator of the rotorcraft UAV based on the acceleration residual using a trained deep forest model.

[0012] Furthermore, it also includes: a construction module, which is used to construct the extended state observer based on the multiple actuators; wherein the model of the extended state observer includes: Where β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are observed quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement.

[0013] Furthermore, it also includes: a training module, which is used to: use a simulation method to obtain the acceleration residual data of the multiple actuators under different faults to obtain a training set; train a preset deep forest model based on the training set to obtain the trained deep forest model.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute the method described in the first aspect.

[0016] The present invention provides a data-driven UAV actuator fault location method and system, which combines an extended state observer with a deep forest model, and can quickly complete the actuator fault identification and fault location based on the combined acceleration residual of the roll angle and the pitch angle. The design of the extended state observer is independent of the UAV model, which reduces the complexity and improves the fault location accuracy, alleviating the technical problem of low accuracy of the model-dependent fault diagnosis method in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a data-driven method for locating a fault of an actuator of a UAV provided by an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of the structure of a quad-rotor drone provided by an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of a data-driven UAV actuator fault location system provided by an embodiment of the present invention;

[0021] Figure 4 A schematic diagram of another data-driven UAV actuator fault location system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] Embodiment 1:

[0024] Figure 1 The flowchart of a data-driven method for locating a fault of an actuator of a UAV according to an embodiment of the present invention is applied to a rotary-wing UAV; wherein the rotary-wing UAV includes a plurality of actuators. For example, when the rotary-wing UAV is a quad-rotor UAV, the actuators are four evenly distributed rotors. Figure 1 As shown, the method specifically comprises the following steps:

[0025] Step S102, obtaining the acceleration residual of the rotor UAV based on the extended state observer; the acceleration residual includes the pitch angle acceleration residual and the roll angle acceleration residual.

[0026] Step S104: Based on the acceleration residual, the trained deep forest model is used to perform fault location and fault identification on the actuator of the rotor UAV.

[0027] The present invention provides a data-driven UAV actuator fault location method, which combines an extended state observer with a deep forest model. It can quickly and simultaneously complete the actuator fault identification and fault location based on the combined acceleration residual of the roll angle and the pitch angle. The design of the extended state observer does not depend on the UAV model, which reduces the complexity and improves the fault location accuracy, alleviating the technical problem of low accuracy of the model-dependent fault diagnosis method in the prior art.

[0028] Optionally, the method provided by the embodiment of the present invention further includes: constructing an extended state observer based on multiple actuators; wherein the model of the extended state observer includes:

[0029]

[0030] Where β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are the observed quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement.

[0031] Optionally, in an embodiment of the present invention, taking a quad-rotor drone as an example, when one of the rotors of the quad-rotor drone fails, the measured values ​​of the accelerometers in the pitch and roll directions will be affected first. Since the yaw and roll of the quad-rotor drone are symmetrical and independent, the angular accelerations in the pitch and roll directions are observed by an extended state observer respectively. Therefore, the accelerometer measurement value residual obtained by the extended state observer in the embodiment of the present invention includes the pitch angular acceleration residual and the roll angular acceleration residual.

[0032] Optionally, in this embodiment of the present invention, the bandwidth ω of the system b =100, the estimation of b0 is calculated based on the control law transfer function of the autonomous disturbance rejection control quadrotor drone, b0 = 13.9237, when t → ∞, a1 → θ, It is the residual between the extended state observer and the accelerometer measurement.

[0033] It can be seen that in the embodiment of the present invention, the design of the extended state observer does not rely too much on the quadrotor drone. It only needs to solve the gain coefficient, and there is no need to find the best and optimize. The bandwidth of the system can be determined in combination with the performance of the quadrotor drone. Compared with other observers, the complexity of the design is greatly reduced, and the parameters are also much less.

[0034] Optionally, the method provided in the embodiment of the present invention further includes a step of training a deep forest model, specifically including:

[0035] Step S21, using a simulation method to obtain acceleration residual data of multiple actuators under different faults to obtain a training set. Specifically, it includes the following steps:

[0036] Construct kinematic and dynamic models of rotary-wing UAVs;

[0037] Based on kinematic and dynamic models, fault models of multiple actuators are constructed;

[0038] Based on the fault model, different faults are injected into different actuators at different times;

[0039] Combining the extended state observer with the accelerometer, the acceleration residual data of multiple actuators under different faults are obtained through simulation.

[0040] Step S22: training the preset deep forest model based on the training set to obtain a trained deep forest model.

[0041] Optionally, the embodiment of the present invention takes a quad-rotor drone with X-shaped distribution as an example to construct a model, wherein the drone has a uniform mass distribution and geometric symmetry, and the motors of the four rotors are distributed as follows: Figure 2 As shown, b x b y b z b is the coordinate system of the quadrotor drone, o b Indicates the center of mass and geometric center of the quadrotor drone, x b Represents the roll axis of the quadrotor drone, y b Represents the pitch axis of the quadcopter, z b represents the yaw axis of the quadrotor drone, φ, θ, Ψ represent the roll angle, pitch angle, and yaw angle respectively, f1, f2, f3, and f4 represent the lift generated by motors 1, 2, 3, and 4 respectively, and also represent the position of each motor, which is located at a 45° angle between the roll axis and the pitch axis.

[0042] The mathematical model of the quadrotor drone consists of dynamic equations and kinematic equations, and its six-degree-of-freedom model is described by the following equations:

[0043]

[0044] Where x, y and z represent the displacement of the center of mass of the quadrotor drone in the earth coordinate system. r represents the moment of inertia of the motor rotor. m, g represent the mass and gravity acceleration of the quadrotor drone, respectively. U1 represents the lift of the quadrotor drone. U2, U3 and U4 represent the generated rolling moment, pitching moment and yaw moment, respectively. x , D y and D z Respectively represent the air resistance coefficient of the three axes of the quadcopter. ω1, ω2, ω3 and ω4 represent the rotation speed of the four rotors. x , J y and J z They represent the moment of inertia along the three axes of the quadrotor drone system, and the moment of inertia matrix is ​​defined as follows:

[0045]

[0046] And J x =J y .

[0047] In the embodiment of the present invention, the actuator of the quad-rotor drone mainly involves the speed constant deviation fault and the speed gain loss fault. Each actuator of the quad-rotor drone will have these two faults. The fault model includes: the speed constant deviation fault model and the speed gain loss model; wherein,

[0048] The speed constant deviation fault model includes:

[0049]

[0050] ω(t) is the speed of the actuator under normal conditions, ω f (t) is the actual speed of the actuator, △ is the deviation between the actual value and the expected value of the actuator speed, t f This is the moment when the actuator has a speed deviation fault.

[0051] The speed gain loss model includes:

[0052]

[0053] ρ is the actuator speed gain loss factor.

[0054] like Figure 2 As shown in the figure, for a quad-rotor drone, when motor 1 fails, the pitch angle decreases and the roll angle increases; when motor 2 fails, the pitch angle increases and the roll angle decreases; when motor 3 fails, the pitch angle decreases and the roll angle decreases; when motor 4 fails, the pitch angle increases and the roll angle increases. When the four motors fail, the pitch and roll angle values ​​will change in different combinations, so the combination of pitch and roll angle signals can be used as a mapping of motor failure.

[0055] Since the quadrotor drone is a highly maneuverable system, during the flight, the pitch angle and roll angle will change and adjust in real time according to the control instructions. Therefore, it is not advisable to directly measure their combined changes. Therefore, an extended state observer is introduced to extract the residual of the angle signal measurement value and the estimated state value as a fault signal. Only when the motor fails, the residual of the measured value and the estimated value will show the motor fault characteristics, and will not change with the change of the control instruction.

[0056] The method provided by the embodiment of the present invention can extract the acceleration residual signals of the pitch angle and the roll angle through the extended state observer as the fault data for actuator fault location. It is not necessary to perform mathematical calculations on the residual signals of the pitch angle and the roll angle. It is only necessary to correspond the time series information of the residual signal as a group of data, and add the corresponding fault label to each group of fault data collected, and inject each group of data and the corresponding label into the deep forest model for training. Through the internal training and fitting of the deep forest, a classification model for actuator fault location of the quad-rotor UAV is obtained. Finally, it is only necessary to inject the collected residual signal into the classification model to obtain the position of the actuator where the fault occurs and the specific fault type.

[0057] The advantage of this process is that there is no need to manually fit and analyze the residual signals of the roll angle and pitch angle. At the same time, due to the advantages of deep forests such as few hyperparameters, low performance requirements for computing resources, and good generalization ability, there is no need to perform complex parameter adjustment on the deep forest, and good accuracy in actuator fault location can be obtained based on the residual data.

[0058] Embodiment 2:

[0059] Figure 3 1 is a schematic diagram of a data-driven UAV actuator fault location system according to an embodiment of the present invention, the system is applied to a rotary-wing UAV; the rotary-wing UAV includes multiple actuators. Figure 3 As shown, the system includes: an acquisition module 10 and a positioning module 20.

[0060] Specifically, the acquisition module 10 is used to acquire the acceleration residual of the rotor UAV based on the extended state observer; the acceleration residual includes the pitch angle acceleration residual and the roll angle acceleration residual.

[0061] The positioning module 20 is used to perform fault location and fault identification on the actuator of the rotary-wing UAV based on the acceleration residual and using the trained deep forest model.

[0062] The present invention provides a data-driven UAV actuator fault location system, which combines an extended state observer with a deep forest model. It can quickly and simultaneously complete the actuator fault identification and fault location based on the combined acceleration residual of the roll angle and the pitch angle. The design of the extended state observer does not depend on the UAV model, which reduces the complexity and improves the fault location accuracy, alleviating the technical problem of low accuracy of the model-dependent fault diagnosis method in the prior art.

[0063] Figure 4 FIG. 1 is a schematic diagram of another data-driven UAV actuator fault location system according to an embodiment of the present invention. Figure 4As shown, the system further includes: a construction module 30, which is used to construct an extended state observer based on multiple actuators; wherein the model of the extended state observer includes:

[0064]

[0065] Where β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are the observed quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement.

[0066] Alternatively, if Figure 4 As shown, the system further includes: a training module 40, for:

[0067] The acceleration residual data of multiple actuators under different faults are obtained by simulation method to obtain a training set. Specifically, a kinematic and dynamic model of the rotorcraft is constructed. Based on the kinematic and dynamic model, a fault model of multiple actuators is constructed. Based on the fault model, different faults are injected into different actuators at different times. Combined with the extended state observer and the accelerometer, the acceleration residual data of multiple actuators under different faults are obtained by simulation method.

[0068] Among them, the fault model includes: speed constant deviation fault model and speed gain loss model; the speed constant deviation fault model includes: ω(t) is the speed of the actuator under normal conditions, ω f (t) is the actual speed of the actuator, △ is the deviation between the actual value and the expected value of the actuator speed, t f It is the moment when the speed deviation fault occurs in the actuator;

[0069] The speed gain loss model includes: ρ is the actuator speed gain loss factor.

[0070] The preset deep forest model is trained based on the training set to obtain a trained deep forest model.

[0071] An embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method in the first embodiment when executing the computer program.

[0072] The embodiment of the present invention further provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code enables the processor to execute the method in the above-mentioned embodiment 1.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data-driven method for locating faults of UAV actuators, characterized in that: Applied to rotary wing UAVs; The rotary-wing UAV includes a plurality of actuators; the method includes: Based on the multiple actuators, an extended state observer is constructed; wherein the model of the extended state observer includes: Among them, β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are observation quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement; Acquiring the acceleration residual of the rotor UAV based on an extended state observer; the acceleration residual includes a pitch angle acceleration residual and a roll angle acceleration residual; Based on the acceleration residual, the trained deep forest model is used to perform fault location and fault identification on the actuator of the rotor UAV.

2. The method according to claim 1, characterized in that Also includes: Using a simulation method to obtain acceleration residual data of the plurality of actuators under different faults to obtain a training set; The preset deep forest model is trained based on the training set to obtain the trained deep forest model.

3. The method according to claim 2, characterized in that The acceleration residual data of the plurality of actuators under different faults are obtained by using a simulation method, including: Constructing kinematic and dynamic models of the rotary-wing UAV; Based on the kinematic and dynamic models, construct fault models of the plurality of actuators; Based on the fault model, injecting different faults into different actuators at different times; The extended state observer and the accelerometer are combined to obtain the acceleration residual data of the multiple actuators under different faults through a simulation method.

4. The method according to claim 3, characterized in that The fault model includes: a speed constant deviation fault model and a speed gain loss model; wherein, The speed constant deviation fault model includes: ω(t) is the speed of the actuator under normal conditions, ω f (t) is the actual speed of the actuator, △ is the deviation speed between the actual value and the expected value of the actuator speed, t f It is the moment when the speed deviation fault occurs in the actuator; The speed gain loss model includes: ρ is the actuator speed gain loss factor.

5. A data-driven UAV actuator fault location system, characterized in that: Applied to a rotary wing UAV; the rotary wing UAV includes a plurality of actuators; the system includes: an acquisition module and a positioning module; wherein, The acquisition module is used to acquire the acceleration residual of the rotor UAV based on the extended state observer; the acceleration residual includes the pitch angle acceleration residual and the roll angle acceleration residual; The positioning module is used to perform fault location and fault identification on the actuator of the rotary-wing UAV based on the acceleration residual using a trained deep forest model; A construction module is used to construct the extended state observer based on the multiple actuators; wherein the model of the extended state observer includes: Among them, β1, β2 and β3 are parameters related to the bandwidth of the system, b0 is the gain coefficient of the extended state observer, a1, a2 and a3 are observation quantities, U3 is the pitch moment, θ is the angle measurement value, is the accelerometer measurement value, e θ is the angle measurement residual, is the residual of the accelerometer measurement.

6. The system according to claim 5, characterized in that Also includes: Training modules for: Using a simulation method to obtain acceleration residual data of the plurality of actuators under different faults to obtain a training set; The preset deep forest model is trained based on the training set to obtain the trained deep forest model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

8. A computer-readable medium having a non-volatile program code executable by a processor, characterized in that: The program code enables the processor to execute the method according to any one of claims 1 to 4.

Images