Control moment gyro active-disturbance-rejection control method based on data-driven feed-forward compensation

By constructing an electromechanical friction dynamic model and using data-driven feedforward compensation technology, the problem of underutilization of dynamic model information in CMG frame speed control is solved, and higher control accuracy and anti-interference performance are achieved.

CN119002231APending Publication Date: 2024-11-22SICHUAN UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411078078.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing control torque gyro (CMG) frame speed control methods fail to fully combine dynamic model information, resulting in insufficient control accuracy and stability, especially when facing complex dynamics and multiple disturbances.

Method used

A feedforward compensation self-immunity control method based on data-driven is proposed. By constructing an electromechanical friction dynamic model, the model compensation amount is solved in real time using sensor data, and input it into the controller to achieve accurate self-immunity control.

Benefits of technology

It improves control accuracy and dynamic performance of the system, enhances the accuracy of estimation of unknown interference and the anti-interference ability of the system, and improves step response and adjustment time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119002231A_ABST
    Figure CN119002231A_ABST
Patent Text Reader

Abstract

The invention discloses a control moment gyroscope active-disturbance-rejection control method based on data-driven feed-forward compensation, and relates to the technical field of active-disturbance-rejection control, and the method comprises the following steps: S1, constructing an electromechanical friction dynamic model for a control moment gyroscope; s2, obtaining jerk of the control moment gyroscope according to the electromechanical friction dynamical model of the control moment gyroscope; s3, determining the compensation amount according to the jerk of the control moment gyroscope; and S4, determining active disturbance rejection control according to the compensation amount. According to the invention, under data-driven feed-forward compensation, the overshoot, the adjustment time, the anti-interference performance and the like of the system to the step response are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of active disturbance rejection control, and in particular to an active disturbance rejection control method of a control moment gyro based on data driven feedforward compensation. Background Art

[0002] Control moment gyro (CMG) is an indispensable core component in high-precision attitude control systems such as spacecraft and missiles. Because of its two independent drive axes and torque control capabilities, the control moment gyro can achieve precise three-axis attitude adjustment and is widely used in the aerospace field.

[0003] The high-speed rotor in the controlled torque gyro provides stable angular momentum, which rotates through the low-speed frame to change the direction of the angular momentum, so that the CMG outputs torque to the outside, thereby adjusting the attitude of satellites and other spacecraft. It can be seen that high-precision control of the rotation speed of the low-speed frame is an important link for CMG to achieve high-precision attitude control of satellites. At present, the control of the CMG frame speed is usually traditional control methods such as PID, and the design of the controller is not combined with the model. Nowadays, the field of dynamics is developing rapidly, and the direction of CMG modeling and analysis is continuously deepened, and great results have been achieved. Therefore, the model is combined with the design of the controller to propose a better controller algorithm to improve the accuracy and stability of the frame motion.

[0004] Active Disturbance Rejection Control (ADRC) is a control algorithm that is widely used in engineering. Its advantage is that it can achieve certain control effects without being based on an accurate mathematical model. However, if it is based on certain model information, it is beneficial to improve the final control effect of the controller. Some scholars use sensor measured information such as speed information and current information as model compensation in the design of the controller. Although this method uses model information to improve the control performance of the system, it does not deeply combine the established dynamic information. Therefore, it is necessary to propose an active disturbance rejection control method that combines dynamic model information to improve the control accuracy of the CMG framework.

[0005] In the field of robotics, a control method of dynamic feedforward compensation is often used. Specifically, the information collected by the sensor is used as input and input into the dynamic model or inverse dynamic model to obtain the acceleration, torque, current and other information that need to be compensated and feed them forward to the controller. This method has high hardware performance requirements for the controller, and requires the controller to solve the dynamic model in real time, and to achieve real-time and high-precision solution. Based on this, the self-disturbance rejection control algorithm combined with the dynamic model information proposed in the present invention also solves the model compensation through the dynamic model through the data measured by the sensor. Due to certain requirements for computational efficiency, a data-driven method is proposed to solve the model compensation, so as to achieve rapid solution of the model compensation in the control of CMG. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a control moment gyro anti-disturbance control method based on data driven feedforward compensation.

[0007] The technical solution of the present invention is: a control moment gyro self-disturbance rejection control method based on data-driven feedforward compensation comprises the following steps:

[0008] S1. Construct an electromechanical friction dynamics model for the control torque gyro;

[0009] S2. According to the electromechanical friction dynamics model of the control moment gyroscope, the acceleration of the control moment gyroscope is obtained;

[0010] S3, determining a compensation amount according to the jerk of the control moment gyro;

[0011] S4. Determine the anti-disturbance control according to the compensation amount.

[0012] Furthermore, in S1, the expression of the electromechanical friction dynamics model is:

[0013]

[0014] Where a is the displacement of 12 degrees of freedom, is the velocity of 12 degrees of freedom, is the acceleration of 12 degrees of freedom, [M] is the mass matrix, [C] is the damping matrix, [G] is the gyro matrix, [K] is the stiffness matrix, and F(t) is the input electromagnetic torque and other interference torques.

[0015] Furthermore, S2 includes the following sub-steps:

[0016] S21. Construct the equation of motion based on the electromechanical friction dynamics model of the control torque gyro;

[0017] S22, determining total interference;

[0018] S23. Determine the jerk of the control torque gyro based on the motion equation and the total disturbance.

[0019] Furthermore, in S21, the expression of the motion equation is:

[0020]

[0021] In the formula, is the jerk of the control moment gyro, b 0 is the system gain, u q is the current, m 0 is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration, and h is the derivative of the external unknown disturbance.

[0022] Furthermore, in S22, the total interference C cal The calculation formula is:

[0023] C cal =m o +q 1 +q 2 ;

[0024] In the formula, m o is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration.

[0025] Furthermore, in S23, the acceleration of the control moment gyro The calculation formula is:

[0026]

[0027] Where b 0 is the system gain, u q is the current, C cal is the total disturbance, and h is the derivative of the external unknown disturbance.

[0028] Furthermore, in S3, the acceleration of the control moment gyro is input into the input layer, the hidden layer and the output layer connected in sequence to obtain the compensation amount.

[0029] Further, S4 includes the following sub-steps:

[0030] S41, determining an expansion state observation according to the acceleration and compensation amount of the control moment gyro;

[0031] S42. Determine the active disturbance rejection control according to the expansion state observation.

[0032] Further, in S41, the extended state observation includes a derivative of the velocity estimate, a derivative of the acceleration estimate, and a derivative of the unknown disturbance estimate;

[0033] where the derivative of the velocity estimate is The calculation formula is:

[0034]

[0035] In the formula, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, l 1 is the first gain parameter of the observer, is the actual speed;

[0036] Derivative of the acceleration estimate The calculation formula is:

[0037]

[0038] Where b 0 is the system gain, u q is the current, z 3 is the estimated value of the unknown interference, l 2 is the second gain parameter of the observer, C pre is the compensation amount;

[0039] Derivative of the unknown disturbance estimate The calculation formula is:

[0040]

[0041] In the formula, l 3 is the third gain parameter of the observer.

[0042] Furthermore, in S42, the calculation formula of the active disturbance rejection control u is:

[0043]

[0044]

[0045] In the formula, u 0 is the output of the feedback controller, C pre is the compensation amount, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, z 3 is the estimated value of the unknown interference, b 0 is the system gain, K p is the first gain of the controller, Kd is the second gain of the controller, is the target speed.

[0046] The beneficial effects of the present invention are:

[0047] (1) The present invention uses data-driven technology in the extended state observer of the active disturbance rejection control to obtain the model compensation according to the sensor value and then through the pre-trained neural network model, which helps to improve the accuracy of the observer's estimation of unknown disturbances;

[0048] (2) The present invention also obtains feedforward compensation in the controller through data driving, and provides the compensation to the controller to improve the control accuracy of the system;

[0049] (3) Under data-driven feedforward compensation, the present invention improves the overshoot, adjustment time and anti-interference performance of the system in response to step changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Flow chart of the control torque gyro anti-disturbance control method based on data-driven feedforward compensation. DETAILED DESCRIPTION

[0051] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0052] like Figure 1 As shown, the present invention provides a control moment gyro self-disturbance rejection control method based on data-driven feedforward compensation, comprising the following steps:

[0053] S1. Construct an electromechanical friction dynamics model for the control torque gyro;

[0054] S2. According to the electromechanical friction dynamics model of the control moment gyroscope, the acceleration of the control moment gyroscope is obtained;

[0055] S3, determining a compensation amount according to the jerk of the control moment gyro;

[0056] S4. Determine the anti-disturbance control according to the compensation amount.

[0057] The present invention fully considers the tribological characteristics, electromechanical coupling effect and gyroscopic torque interference in the CMG electromechanical system, and establishes a flexible bearing model under grease lubrication, a flexible rod model, a permanent magnet synchronous motor model and a CMG dynamics model, and then couples and combines them to establish an electromechanical friction dynamics model of the CMG system.

[0058] According to the established CMG electromechanical friction dynamics model, the input of the model is the frame velocity, frame acceleration and electromagnetic torque, and the output of the model is the interference to the frame in the CMG system (model compensation). Through a large amount of simulation data, the equipment well network training model is used to accurately map the input and output, and the output value corresponding to the input is quickly obtained in a data-driven manner.

[0059] Finally, an auto-disturbance rejection control algorithm based on data-driven feedforward compensation is proposed, and the model compensation is used to feedforward the controller to improve the control accuracy of the controller and enhance the dynamic performance of the system.

[0060] In the embodiment of the present invention, in S1, the expression of the electromechanical friction dynamics model is:

[0061]

[0062] Where a is the displacement of 12 degrees of freedom, is the velocity of 12 degrees of freedom, is the acceleration of 12 degrees of freedom, [M] is the mass matrix, [C] is the damping matrix, [G] is the gyro matrix, [K] is the stiffness matrix, and F(t) is the input electromagnetic torque and other interference torques.

[0063] The CMG system contains two main components, the flywheel and the frame. The main flexible component is the bearing that supports the rotation. The bearing is lubricated with grease. The stiffness and damping parameters of the bearing under grease lubrication are studied by using the Reynolds equation. At the same time, a permanent magnet synchronous motor model is established. Considering the parameters such as the current, voltage and magnetic flux of the motor, a model is established to drive the CMG frame to rotate. Considering the effect of the gyroscopic torque in the CMG system, a CMG dynamic model containing a frame with 6 degrees of freedom and a rotor with 6 degrees of freedom is established. The above models are combined to obtain a CMG electromechanical friction dynamic model with 12 degrees of freedom.

[0064] In this embodiment of the present invention, S2 includes the following sub-steps:

[0065] S21. Construct the equation of motion based on the electromechanical friction dynamics model of the control torque gyro;

[0066] S22, determining total interference;

[0067] S23. Determine the jerk of the control torque gyro based on the motion equation and the total disturbance.

[0068] For the total interference C calThe solution usually takes the speed information, acceleration information and current information collected by the sensor as input and inputs them into the electromechanical friction dynamics model. This places high requirements on the hardware performance of the controller. The controller needs to solve the problem in real time and put the solution results into each step of the controller.

[0069] In the embodiment of the present invention, in S21, the expression of the motion equation is:

[0070]

[0071] In the formula, is the jerk of the control moment gyro, b 0 is the system gain, u q is the current, m 0 is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration, and h is the derivative of the external unknown disturbance.

[0072] In the embodiment of the present invention, in S22, the total interference C cal The calculation formula is:

[0073] C cal =m o +q 1 +q 2 ;

[0074] In the formula, m o is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration.

[0075] In the embodiment of the present invention, in S23, the acceleration of the control moment gyro The calculation formula is:

[0076]

[0077] Where b 0 is the system gain, u q is the current, C cal is the total disturbance, and h is the derivative of the external unknown disturbance.

[0078] In the embodiment of the present invention, in S3, the acceleration of the control moment gyro is input into the input layer, the hidden layer and the output layer connected in sequence to obtain the compensation amount.

[0079] In S3, the speed, acceleration of the frame and the electromagnetic torque of the motor collected by the sensor are used as key input parameters, and these parameters are mapped to the solution results of the dynamic model to achieve one-to-one correspondence, thereby achieving real-time and efficient model compensation and realizing data-driven model compensation.

[0080] Through the established electromechanical friction dynamics model, different working conditions of the model were set, and a large number of simulation experiments were carried out to systematically collect data on input parameters and dynamic model compensation. These data not only cover a wide range of operating conditions, but also take into account various possible system disturbances, ensuring the diversity and representativeness of the training data. Using the rich data set collected, a deep neural network model was designed and trained. The model uses advanced nonlinear activation functions and optimization algorithms to learn the complex mapping relationship between input parameters and output results. The network structure and parameters are optimized through repeated training and verification to ensure the robustness and accuracy of the model under different operating conditions. The trained neural network model is embedded in the controller to form a data-driven feedforward compensation mechanism. This mechanism can use real-time data collected by sensors to quickly map and predict model compensation. Let the model compensation solved by dynamics be C cal , the model compensation amount predicted by data-driven method is C pre .

[0081] Specifically, the PyTorch deep learning framework was used to train the model. The input data set was adjusted and optimized using the maximum and minimum normalization process. Each network consists of an input layer, a hidden layer, and an output layer. Each hidden layer consists of a fully connected (FC) layer with a rectified linear unit (ReLU) activation function. The number of nodes contained in each hidden layer is 64. The output layer consists of an fc layer without an activation function. Minimizing the mean square error (mse) in the Adam solver was used as the loss function. During the training process, the learning rate was set to 0.00005. The batchsize used in the study was 10, because the small batch training strategy is easy to converge for this study. In addition, the entire training cycle was 60,000, and the final mean square error converged to 0.000001.

[0082] In this embodiment of the present invention, S4 includes the following sub-steps:

[0083] S41, determining an expansion state observation according to the acceleration and compensation amount of the control moment gyro;

[0084] S42. Determine the active disturbance rejection control according to the expansion state observation.

[0085] In the embodiment of the present invention, in S41, the extended state observation includes a derivative of the velocity estimate, a derivative of the acceleration estimate, and a derivative of the unknown interference estimate;

[0086] where the derivative of the velocity estimate is The calculation formula is:

[0087]

[0088] In the formula, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, l 1 is the first gain parameter of the observer, is the actual speed;

[0089] Derivative of the acceleration estimate The calculation formula is:

[0090]

[0091] Where b 0 is the system gain, u q is the current, z 3 is the estimated value of the unknown interference, l 2 is the second gain parameter of the observer, C pre is the compensation amount;

[0092] Derivative of the unknown disturbance estimate The calculation formula is:

[0093]

[0094] In the formula, l 3 is the third gain parameter of the observer.

[0095] In the embodiment of the present invention, in S42, the calculation formula of the active disturbance rejection control u is:

[0096]

[0097]

[0098] In the formula, u 0 is the output of the feedback controller, C pre is the compensation amount, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, z 3 is the estimated value of the unknown interference, b 0 is the system gain, K p is the first gain of the controller, K d is the second gain of the controller, is the target speed.

[0099] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A control torque gyro anti-disturbance control method based on data-driven feedforward compensation, It is characterized in that The following steps are involved: S1. Construct an electromechanical friction dynamics model for the control torque gyro; S2. According to the electromechanical friction dynamics model of the control moment gyroscope, the acceleration of the control moment gyroscope is obtained; S3, determining a compensation amount according to the jerk of the control moment gyro; S4. Determine the anti-disturbance control according to the compensation amount.

2. The control moment gyro anti-disturbance control method based on data-driven feedforward compensation according to claim 1, It is characterized in that In S1, the expression of the electromechanical friction dynamics model is: Where a is the displacement of 12 degrees of freedom, is the velocity of 12 degrees of freedom, is the acceleration of 12 degrees of freedom, [M] is the mass matrix, [C] is the damping matrix, [G] is the gyro matrix, [K] is the stiffness matrix, and F(t) is the input electromagnetic torque and other interference torques.

3. The control moment gyro active disturbance rejection control method based on data driven feedforward compensation according to claim 1, It is characterized in that The S2 comprises the following sub-steps: S21. Construct the equation of motion based on the electromechanical friction dynamics model of the control torque gyro; S22, determining total interference; S23. Determine the jerk of the control torque gyro based on the motion equation and the total disturbance.

4. The control moment gyro anti-disturbance control method based on data-driven feedforward compensation according to claim 3, It is characterized in that In S21, the expression of the motion equation is: In the formula, is the jerk of the control moment gyro, b 0 is the system gain, u q is the current, m 0 is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration, and h is the derivative of the external unknown disturbance.

5. The control moment gyro anti-disturbance control method based on data-driven feedforward compensation according to claim 3, It is characterized in that In S22, the total interference C cal The calculation formula is: C cal =m o +q 1 +q 2 ; In the formula, m o is the effect of motor properties on jerk, q 1 is the effect of its own flexibility on the acceleration, q 2 is the influence of other degrees of freedom on the acceleration.

6. The control moment gyro anti-disturbance control method based on data-driven feedforward compensation according to claim 3, It is characterized in that In S23, the acceleration of the control moment gyro The calculation formula is: Where b 0 is the system gain, u q is the current, C cal is the total disturbance, and h is the derivative of the external unknown disturbance.

7. The control moment gyro active disturbance rejection control method based on data driven feedforward compensation according to claim 1, It is characterized in that In S3, the acceleration of the control moment gyro is input into the input layer, hidden layer and output layer connected in sequence to obtain the compensation amount.

8. The control moment gyro active disturbance rejection control method based on data driven feedforward compensation according to claim 1, It is characterized in that The S4 comprises the following sub-steps: S41, determining an expansion state observation according to the acceleration and compensation amount of the control moment gyro; S42. Determine the active disturbance rejection control according to the expansion state observation.

9. The control moment gyro active disturbance rejection control method based on data driven feedforward compensation according to claim 8, It is characterized in that In S41, the extended state observation includes a derivative of the velocity estimate, a derivative of the acceleration estimate, and a derivative of the unknown disturbance estimate; Wherein, the derivative of the velocity estimate The calculation formula is: In the formula, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, l 1 is the first gain parameter of the observer, is the actual speed; The derivative of the acceleration estimate The calculation formula is: Where b 0 is the system gain, u q is the current, z 3 is the estimated value of the unknown interference, l 2 is the second gain parameter of the observer, C pre is the compensation amount; The derivative of the unknown disturbance estimate The calculation formula is: In the formula, l 3 is the third gain parameter of the observer.

10. The control moment gyro active disturbance rejection control method based on data driven feedforward compensation according to claim 8, It is characterized in that In S42, the calculation formula of the active disturbance rejection control u is: In the formula, u 0 is the output of the feedback controller, C pre is the compensation amount, z 1 is the estimated value of velocity, z 2 is the estimated value of acceleration, z 3 is the estimated value of the unknown interference, b 0 is the system gain, K p is the first gain of the controller, K d is the second gain of the controller, is the target speed.