A method for modeling lever force gradients in control systems based on system identification

Abstract

This invention discloses a control system stick force gradient modeling method based on system identification, which has significant advantages and potential for modeling and simulating control load systems in flight simulators. First, preliminary assumptions are made about the control system model based on the aircraft control system type, resulting in the control system model equations and the parameters required for modeling. Then, the static force characteristics of the control system are extracted and modeled using the "stick force-displacement" test. Next, dynamic characteristic curves during the stick release process are obtained using different "stick release" tests during takeoff, cruise, and landing phases. Using the static characteristic curves as a benchmark, system identification is performed using the dynamic characteristic curves to obtain the unknown parameters in the established control system model, thus obtaining a complete control system simulation model.

Description

Technical Field

[0001] This invention belongs to the field of aircraft control system design and flight simulation technology, specifically relating to a control system stick force gradient modeling method based on system identification. Background Technology

[0002] The control load system is an important component of advanced flight simulators. It provides pilots with realistic stick force feedback during simulated training. Since the control system is the system that pilots interact with most directly and use most frequently when flying an aircraft, its realism directly affects the pilot's control memory. In the objective evaluation standards for training simulators of all levels, in order to ensure the effectiveness of simulated training, there are specific tests for the realism of the control load system model, which mainly include the "stick force-displacement" test and the "stick release" test.

[0003] When conducting a "force-displacement" test, the tester needs to slowly move the control stick throughout its full stroke to obtain the "force-displacement" curve. Obviously, this type of test can only be conducted when the control stick is stationary on the ground.

[0004] The "stick release" test is primarily conducted to verify whether the stick's moment of inertia, friction, dynamic damping, aerodynamic hinge torque, and human-sensor force curves are consistent with the simulated object. During the test, the stick is typically moved to 25%–30% of its full travel and then released naturally, and the stick's oscillation characteristics under various forces are recorded. Since aerodynamic hinge torque and human-sensor force are usually related to aircraft dynamic pressure and aerodynamic surface deflection angles, the "stick release" test needs to be conducted while the stick is stationary on the ground and in different flight states (takeoff, cruise, landing) to obtain the stick force variation characteristics under different hinge torques and human-sensor force.

[0005] The common method for modeling the force gradient of a joystick is to model the mechanical links and proportional relationships of the control system, the centering spring, dead zone, friction, the force felt by the human sensor, and the torque of the pneumatic hinge, based on the principle of the control system. A complete joystick force gradient model is formed by using the control system design data and actual measurement and calculation of various system parameters.

[0006] However, the biggest problem with this type of modeling is that the parameters of the model are often difficult to determine accurately.

[0007] First, although the design parameters of each connection structure in the aircraft control stick system can be obtained, the actual state of the links deviates from the design parameters due to manufacturing, installation, and flight conditions. For hydraulically assisted aircraft, the application of hydraulic pressure affects the original force relationships between the links, and this effect is often difficult to derive from theoretical calculations. Modeling the stick force model solely based on theoretical parameters often results in a model that differs from the actual model.

[0008] Adding sensors to the control system to acquire some control system parameters during actual flight and calibrate the design parameters can solve the aforementioned problems to some extent. However, adding sensors is too expensive, and some parameters are difficult to measure simply by adding sensors. Furthermore, calibrating the design parameters requires the control system to be able to move throughout its entire range under different flight conditions, which is impossible to achieve during actual flight.

[0009] Building models based on the design principles of control systems often results in complex modeling, making it difficult to adjust the model when the control system changes, and leading to poor model robustness. Obtaining model parameters through system identification methods can effectively solve these problems. Summary of the Invention

[0010] To address the above problems, this invention proposes a method for modeling the lever force gradient of a control system based on system identification, comprising the following steps:

[0011] The present invention provides a method for modeling the lever force gradient of a control system based on system identification, comprising the following steps:

[0012] Step 1: Establish a control system model based on the characteristics of the simulated control system;

[0013] Step 2: Obtain the static characteristic data of the joystick;

[0014] Step 3: Obtain the dynamic characteristic data of the joystick;

[0015] Step 4: Establish the relationship between static force, motion damping, friction and mass characteristics of the joystick during its movement using the joystick release curve. Then, identify the unknown parameters in the control system model through system identification to obtain a complete joystick force gradient model.

[0016] The advantages of this invention are:

[0017] 1. The control system stick force gradient modeling method of this invention, based on system identification, only requires measuring a single "stick force-displacement" curve and a few "stick release" curves under different flight conditions to establish the model and identify system parameters. For commercially operating aircraft, the above two types of curves have already been collected during simulator flight tests; for aircraft still in the research and development flight test stage, the amount of data measured for the above curves is small, the measurement methods and tools are very mature, and the data collection is already part of the flight test subjects, so it will not incur additional data acquisition costs.

[0018] 2. The control system lever force gradient modeling method based on system identification of the present invention does not require the establishment of a complex control system principle model. By dividing the control system into static and dynamic characteristics, system identification can automatically merge modules with similar functional structures and output characteristics in the system and manage them through unified parameters, thereby simplifying the model and improving the system's computational efficiency and stability.

[0019] 3. This invention, based on system identification, provides a method for modeling lever force gradients in control systems. This method separates the system model from the parameters. The system model can be built from a principle model based on the characteristics of the system to be simulated, while parameter identification can be achieved through a separate toolchain. Both are universal. When the simulated object changes, there is no need for extensive modifications and adjustments to the system model and parameters. Simply select a suitable principle model based on the new system characteristics, and then use the parameter identification toolchain to obtain the system parameters, thus quickly obtaining the system model.

[0020] 4. The control system lever force gradient modeling method based on system identification of the present invention can quickly update the model using newly measured "lever force-displacement" data and "lever release" data, thereby achieving rapid calibration and correction of the model. Attached Figure Description

[0021] Figure 1 This is a flowchart of the lever force gradient modeling method for control systems based on system identification, as described in this invention.

[0022] Figure 2 This is a schematic diagram of the rod force-displacement measurement curve and the interpolation table obtained after data fitting.

[0023] Figure 3 The joystick release measurement curve and the velocity and acceleration time history curves are shown.

[0024] Figure 4 To identify the simulation results of the model. Detailed Implementation

[0025] The present invention will now be described in further detail with reference to the accompanying drawings.

[0026] This invention relates to a method for modeling the lever force gradient of a control system based on system identification, such as... Figure 1 As shown, it includes the following steps:

[0027] Step 1: Determine the preliminary model structure of the flight control system.

[0028] Currently, there are several flight control systems, including mechanical control, hydraulically assisted control, and fly-by-wire control. The factors affecting the stick force gradient differ for different types of control systems.

[0029] The following explanation uses a hydraulic power-assisted control system as an example:

[0030] The lever force gradient of a hydraulic power steering system is generally affected by several factors, including the centering spring, friction, the force felt by the human sensor system, and motion damping.

[0031] The centering spring is typically a mechanical spring structure, and its force is related to the angle θ between the joystick and the neutral position.

[0032] Fspring = f(θ)

[0033] The sensory force of the human sensory system is artificially programmed and is typically a function of the aircraft dynamic pressure Qc and the control stick deflection angle θ.

[0034] Ffeel=f(θ,Qc)

[0035] Friction is usually sliding friction, which is opposite to the direction of motion v, and is represented as:

[0036] Ffriction = sign(v)

[0037] Velocity damping is usually proportional to velocity, expressed as:

[0038] Fdamp = pv

[0039] According to theoretical mechanics, the movement of the joystick will generate an inertial force, which can be expressed as:

[0040] Finertia = ma

[0041] The lever force gradient principle model of the control system is the sum of the above forces.

[0042] Where p is the velocity damping coefficient; m is the equivalent joystick mass; and a is the joystick acceleration.

[0043] Meanwhile, based on whether each parameter changes with changes in the external environment such as joystick movement / flight status, the flight control system parameters are divided into static parameters and dynamic parameters, which are used to obtain the static and dynamic features in steps 2 and 3, respectively.

[0044] Step 2: Obtain the static characteristic data of the joystick;

[0045] The static characteristic data of the joystick are obtained by conducting a "force-displacement" test on the joystick, recording the time history data of joystick displacement, force, and control surface deflection angle during the test, and further obtaining the data through data fitting to form a force-displacement interpolation table.

[0046] The "stick force-displacement" curve is a static curve that includes information from the centering spring, the force felt by the human sensor system, and friction. Testing the "stick force-displacement" curve requires the pilot to slowly move the control stick throughout its full travel at approximately 1 degree per second. During this movement, the acceleration of the control stick and the damping caused by speed are almost negligible; that is, the resulting curve is a static curve that includes information from the centering spring, the force felt by the human sensor system, and friction. For most civil aircraft, their "stick force-displacement" curves have already been determined and can be obtained from aircraft suppliers or through flight test measurements.

[0047] Due to friction, the "force-displacement" curve is typically a hysteresis curve. Because the measurement data contains noise and interference, it is necessary to fit the "force-displacement" measurement data using a smoothed spline fitting method before data fitting. This involves replacing each point on the curve with the average of several neighboring points, thus smoothing the data and ultimately obtaining a "force-displacement" interpolation table. This interpolation table can then be used to describe the static characteristic model, as shown in the results. Figure 2 As shown.

[0048] Step 3: Obtain the dynamic characteristic data of the joystick.

[0049] The dynamic characteristics of the control stick are obtained through a "control stick release" test. The test requires the pilot to move the control stick to 25%–30% of its full travel during takeoff, cruise, and landing, and then release it naturally, recording its oscillation characteristics. Therefore, the data includes not only static curves but also velocity damping caused by the control stick's movement and inertial forces caused by the control stick's mass. Velocity damping can be obtained by multiplying the control stick's velocity by the velocity damping coefficient p; inertial forces are obtained by multiplying the control stick's acceleration by its mass.

[0050] During the "joystick release" curve test, record the test time, joystick displacement, joystick force, control surface deflection angle, joystick speed, and acceleration time history data. Accurate recording of speed and acceleration can improve the model recognition accuracy. If a dedicated sensor is not used to record the speed and acceleration information of the joystick movement, it is not necessary to record them. Approximate joystick speed and acceleration information can be obtained by washing out / differentiating the joystick position.

[0051] The accuracy of the calculated velocity and acceleration data was then verified. The velocity and acceleration data calculated using the aforementioned method were used to reconstruct the joystick position data, and the results were compared with the original measured joystick position data. The results are as follows: Figure 3 As shown.

[0052] Step 4: Identify joystick system parameters.

[0053] When a person holds the joystick, they apply an artificial force Fman to it. According to Newtonian mechanics, the following equation can be obtained:

[0054] Fspring+Ffeel+Ffriction+Fdamp+Finertia=Fman

[0055] When the "joystick release" test is performed, the manual force in the above formula becomes 0, and the formula can be rewritten as:

[0056] Fspring+Ffeel=-sign(v)-pv-ma

[0057] The left side of the equation represents the measured static characteristic curve, while the right side contains two unknown parameters, p and m. These unknown parameters can be identified using parameter identification methods such as least squares or maximum likelihood. After obtaining the model parameters, substituting them into the original equation yields the complete rod force gradient model. An example identification result is shown below. Figure 4 As shown.

[0058] In this invention, the force acting on the control system equals the static force plus the dynamic force. When the joystick is stationary at a certain position, it is only subject to a static force, the magnitude of which can be obtained by interpolating the static characteristic curve (obtained through fitting the "joystick force-displacement" curve) and the joystick position. When the joystick moves, it is simultaneously subject to both static and dynamic forces. The acceleration and velocity can be obtained from the "joystick release" curve, and the mass, velocity damping, and other coefficients can be identified through system identification, thereby obtaining the dynamic force.

[0059] By utilizing the stick force model obtained through the invention, the real-time motion acceleration, velocity, and stick force output of the control stick can be calculated based on any control stick displacement and input, thereby realizing control stick force feedback in flight simulators.

[0060] The model obtained by this method has a simple structure, high computational efficiency, and does not depend on the acquisition of the original design parameters of the control system during modeling. Furthermore, when using this method to model the stick forces of mechanical and fly-by-wire flight control systems, the only difference from the hydraulic power-assisted control system is the model structure; the remaining modeling steps are completely identical. This method has good versatility and significant advantages and potential in the field of flight simulator control system modeling and simulation.

Claims

1. A method for modeling the lever force gradient of a control system based on system identification, characterized in that: The specific steps are as follows: Step 1: Establish a control system model based on the characteristics of the simulated control system; The parameters in the joystick system are divided into static parameters and dynamic parameters based on whether they change with the external environment. These are used to obtain the static and dynamic features in steps 2 and 3, respectively. Step 2: Obtain the static characteristic data of the joystick; The static characteristic data of the joystick are obtained by performing a joystick force-displacement test on the joystick, recording the joystick displacement, joystick force, and control surface deflection angle time history data during the test, obtaining the static characteristic data of the joystick through data fitting, and forming a joystick force-displacement interpolation table. Step 3: Obtain the dynamic characteristic data of the joystick; The test requirements for the joystick release curve are as follows: during takeoff, cruise, and landing phases, the pilot moves the joystick to 25%–30% of its full travel and then releases it naturally, recording its oscillation characteristics. The test data includes not only the static curve but also the velocity damping caused by the joystick movement and the inertial force caused by the joystick's mass. The velocity damping is obtained by multiplying the joystick velocity by the velocity damping coefficient p; the inertial force is obtained by multiplying the joystick acceleration by its mass. During the joystick release curve test, the test time, joystick displacement, joystick force, control surface deflection angle, joystick velocity, and acceleration time history data are recorded. Step 4: Establish the relationship between static force, motion damping, friction and mass characteristics of the joystick during its movement using the joystick release curve. Then, identify the unknown parameters in the control system model through system identification to obtain a complete joystick force gradient model. When a person holds the joystick, applying an artificial force Fman to the joystick results in: （1） in, These are the constant spring force, the force perceived by the human sensory system, the frictional force, the velocity damping force, and the inertial force, respectively. Friction is usually sliding friction, which is opposite to the direction of motion. Conversely, it is expressed as: Velocity damping is usually proportional to velocity, expressed as: According to theoretical mechanics, the movement of the joystick will generate an inertial force, which can be expressed as: The lever force gradient principle model of the control system is the sum of the above forces; Where p is the velocity damping coefficient; m is the equivalent joystick mass; and a is the joystick acceleration. When the joystick release test is performed, the manual force in the above formula becomes 0, and the formula is rewritten as: （2） In the formula, p is the velocity damping coefficient; m is the equivalent joystick mass; and a is the joystick acceleration. Indicates the direction of friction. In terms of direction of motion; The left side of equation (2) is the measured static characteristic curve, while the right side contains two unknown parameters, p and m. The unknown parameters are confirmed by the least squares or maximum likelihood parameter identification method. After obtaining the model parameters, the parameters are substituted into formula (2) to obtain the complete rod force gradient model.

2. The method for modeling the lever force gradient of a control system based on system identification as described in claim 1, characterized in that: The data fitting method used is smooth spline fitting.

Images