Aviation aircraft control distribution method considering dynamic characteristics of execution mechanism

Abstract

The invention belongs to the field of over-driven aviation aircraft control distribution, and relates to an aviation aircraft control distribution method considering dynamic characteristics of an execution mechanism. According to the invention, the control instruction is effectively distributed to each execution mechanism by designing the control distribution strategy. The influence of response characteristics, control efficiency and physical constraints of the execution mechanism is comprehensively considered, a model predictive control algorithm thought is referred, a control allocation method is designed from dynamic characteristics of the execution mechanism of the aviation aircraft, and meanwhile cost function design is carried out according to task requirements to achieve optimal control. The overdrive aircraft control distributor is designed with dynamic control distribution as a starting point, the influence of the dynamic characteristics of the execution mechanisms on aircraft control is comprehensively considered, control instructions are effectively distributed to all the execution mechanisms, and the stability and maneuvering characteristics of an aircraft are improved.

Description

Technical Field

[0001] The present invention belongs to the field of control allocation for over-driven aircraft, and relates to a control allocation method for aircraft considering the dynamic characteristics of actuators. Background Art

[0002] Over-driven aircraft are equipped with multiple actuators that can be used for attitude and trajectory control, such as multiple control surfaces, thrust vector devices, etc. Compared with traditional aircraft, they have stronger maneuverability, higher control redundancy, and better fault tolerance. They can achieve different maneuvering actions through the coordinated cooperation of each actuator under normal circumstances, and can maintain good control ability by operating redundant actuators in the event of a failure of some actuators. However, this advantage also brings new challenges, such as problems of control surface redundancy, increased coupling degree, and enhanced nonlinear dynamic characteristics. To give full play to the advantages of over-driven aircraft, it is necessary to design a reasonable control allocation method to effectively allocate control commands to each control surface on the premise of meeting the preset index requirements, so that the aircraft can achieve the optimal control effect considering the dynamic characteristics of the actuators.

[0003] Considering that during the flight mission of an over-driven aircraft, since multiple actuators need to cooperate to achieve safe and stable completion of the desired maneuver, this places high requirements on the flight control strategy and system. If a control law is designed directly for each actuator separately, the control efficiency may decrease due to problems such as uneven use of actuators and mutual coupling interference of control commands. In contrast, the method based on the control allocation strategy can give full play to the advantage of actuator redundancy, enabling the aircraft to achieve the desired attitude and trajectory control. Based on the control allocation strategy, the control performance of the over-driven control system can be significantly improved, but a reasonable selection and design of the control allocation method are required. For example, the control allocation based on the pseudo-inverse method has a fast operation speed and high allocation efficiency, but does not consider the physical constraints of the actuators, and the allocation result may not be the optimal solution; the control allocation based on the series connection chain method is simple to implement in engineering and has a flexible allocation method, but cannot give full play to the maneuvering ability of multiple actuators. Therefore, it is necessary to design a control allocation method for aircraft that can fully consider the mission requirements and achieve a control allocation method for aircraft that can balance safety, stability, and the nonlinear dynamic characteristics of actuators.

[0004] The static control allocation method is premised on ignoring the dynamics of actuators, that is, the bandwidth of the actuators is much higher than that of the rigid aircraft. In fact, actuators have different dynamic characteristics, and the control efficiency is also different from the ideal situation. The dynamic characteristics of actuators, such as response time, rate limit, hysteresis, and friction, will all have an important impact on the control allocation results. Ignoring the dynamic performance of control allocation and actuators will have a serious impact on the performance of the control system. Therefore, when designing the control allocation method, it is necessary to comprehensively consider the dynamic model of actuators to improve the allocation accuracy and execution effect, thereby improving the overall control performance of the aircraft.

[0005] Yu Kaiming and Dong Jiuxiang proposed an optimized control allocation method based on the weighted pseudo-inverse method in "An Optimized Control Allocation Method for Aircraft under Faults Based on the Weighted Pseudo-Inverse Method" (CN202410091967.9). First, a weighted pseudo-inverse control allocation model for the aircraft was established to solve the expression of the control vector after control allocation; an aircraft fault model was established and the constant matrix was solved; an improved positive definite symmetric weight matrix was constructed; then the constant matrix was substituted into the expression of the control vector after the fault, and the improved positive definite symmetric weight matrix was used to replace the positive definite symmetric weight matrix to obtain the final control vector after control allocation, thereby realizing the control allocation under the fault of the aircraft actuator and improving the safety and reliability of the flight control of the aircraft. Although the weighted pseudo-inverse method can improve the control allocation efficiency, it does not consider the dynamic characteristics and physical constraints of the actuators during the allocation process, which may have a certain impact on the performance of the control system under the influence of the actual actuator characteristics.

[0006] Wang Yin and Tan Jiyun proposed an optimal allocation method for the control structure of the scaled reallocation pseudo-inverse method in "Flight Control of Compound Wing UAVs Based on Reallocation Pseudo-Inverse" (see "Control Theory and Applications", 2024, 41(10): 1765-1773). Through scaling, it is ensured that the solution can be searched near the control quantity generated by the pseudo-inverse method, so that the realized pseudo-control is as consistent as possible with the direction of the desired control, solving the problem of large angular deviation of the control vector generated by the reallocation pseudo-inverse method. However, when the expectation is unreachable, the average error will increase sharply, and at the same time, the non-linearity of the control effectiveness of the control surface is not comprehensively considered, which may cause a large allocation error.

[0007] Liu Liang, Li Kangwei, Zhang Da, etc. in the "Control Method for the Transition Section of a Vertical / Short Takeoff and Landing Aircraft Based on Intelligent Control Allocation" (CN202211615355.2) calculated the actual control quantity according to the acceleration control instruction, designed an optimization objective function, and then used an improved grey wolf optimization algorithm to solve the control quantity. An initialization strategy based on the good point set was adopted to avoid the problem that the individuals are concentrated at the local optimum caused by the common random initialization, improving the search efficiency. At the same time, the global and local search capabilities were adjusted based on the non-linear change strategy of the LeCunTanh activation function of the neural network, obtaining better optimization results. However, the intelligent control allocation method has a large amount of calculation, a long online search time, and high requirements for the computer, so there are certain limitations in current engineering applications. Summary of the Invention

[0008] To solve the above problems, the present invention proposes a control allocation method for an aircraft considering the dynamic characteristics of the actuators, constructs an overactuated aircraft and actuator models, and predicts the future output of the aircraft system according to the current state of the model and the future control quantity, realizing a control allocation strategy that fully considers the dynamic characteristics of the actuators; designs the performance function according to indexes such as the minimum allocation error accuracy and the minimum rudder deflection, calculates the optimal control sequence in the control interval, and realizes the optimal control considering the actuator constraints; selects the first item of the calculated optimal control sequence as the expected rudder deflection at the current moment, and realizes rolling optimization through repeated online optimization, achieving the effect of "predicting multiple steps and taking one step".

[0009] The technical solution of the present invention is as follows:

[0010] A control allocation method for an aircraft considering the dynamic characteristics of the actuators, comprising the following steps:

[0011] Step (1) Establish an overactuated aircraft model

[0012] To clarify the flight state of the overactuated aircraft, the aircraft dynamics and kinematics models are expressed in the following way:

[0013]

[0014] In the formula: u, v, and w are the velocity components on the x, y, and z axes in the airflow coordinate system respectively; p, q, and r are the roll angular velocity, pitch angular velocity, and yaw angular velocity in the body axis system respectively; g is the gravitational acceleration; φ, θ, and ψ are the roll angle, pitch angle, and yaw angle respectively; are the aerodynamic force components on the x, y, and z axes in the body coordinate system respectively; I x 、I y 、I z respectively represent the moments of inertia of the aircraft body about the x, y, and z axes in the body coordinate system, I xzrepresents the product of inertia of the entire cross-section of the airframe with respect to the x-axis and z-axis of the airframe coordinate system; M and N represent the rolling moment, pitching moment, and yaw moment respectively; the superscript "·" represents the derivative; V is the airspeed in the velocity coordinate system; α is the angle of attack; β is the sideslip angle. T is the maximum thrust of the aircraft engine; m is the mass of the aircraft; D, Y, and L are the drag, side force, and lift in the velocity coordinate system respectively; g1, g2, and g3 are the components of the gravitational acceleration, where the expressions of D, Y, L, and g1, g2, g3 are as shown in the following formula:

[0015]

[0016] The expressions of L, M, and N are:

[0017]

[0018] In the formula: ρ is the atmospheric density; S, b, are the characteristic area of the airframe, the wingspan length, and the mean aerodynamic chord length respectively; C L , C M , C N are the dimensionless rolling moment coefficient, dimensionless pitching moment coefficient, and dimensionless yaw moment coefficient respectively.

[0019] Step (2) Establish the actuator model of the aircraft

[0020] In order to comprehensively consider the influence of the dynamic characteristics of the aircraft actuator on the performance of the control system, the aircraft servo mechanism is selected as the actuator, and a second-order system with amplitude and rate limitations is selected as the dynamic model of the servo mechanism, without uncertainty interference, and its form is as follows:

[0021]

[0022] In the formula: s is the Laplace operator; ω n is the natural frequency; ζ is the damping ratio; δ c , δ real are the desired control surface deflection command and the actual control surface deflection respectively. By reasonably designing the natural frequency and damping ratio, a reference model that meets the response characteristics of the actual servo mechanism is obtained.

[0023] Based on the above dynamic model of the servo mechanism and combined with the control surface efficiency matrix of the aircraft, the aircraft system can be described in the form of state space equations, as shown in the following formula:

[0024]

[0025] In the formula: x δ = [δ 1real δ 2real δ3real …] T is the actual deflection sequence of multiple control surfaces, δ 1real , δ 2real , δ 3real are the actual deflections of different control surfaces respectively; u δ = [δ 1c δ 2c δ 3c …] T is the ideal deflection sequence of multiple control surfaces, δ 1c , δ 2c , δ 3c are the expected deflections of different control surfaces respectively; are the rolling moment coefficient, pitching moment coefficient, and yaw moment coefficient actually output by the system respectively; A δ , B δ , C δ are the system state matrix, control matrix, and output matrix respectively, where the variables represented by capital bold symbols are matrices, and the variables represented by lowercase bold symbols are vectors.

[0026] The actuator model of the aircraft constructed in step (2) can output the actual actuator actuation value. After the actual actuator actuation value is input into the overactuated aircraft model constructed in step (1), it can change the aircraft's control force / moment, thereby realizing the adjustment of the aircraft's attitude.

[0027] Optimal control design of aircraft considering control surface response characteristics in step (3)

[0028] Based on the servo model and the current control surface deflection and future expected control surface deflection of the aircraft, the future control moment as the system output can be predicted. The prediction model construction process is as follows.

[0029] In step (2), the construction of the aircraft system state space equation is completed. Since the future output needs to be predicted based on the dynamic characteristics of the aircraft servo link, the continuous state space equation needs to be discretized. The discretized state space equation is:

[0030]

[0031] In the formula: are the actual deflection sequence of the system control surface, the ideal deflection sequence of multiple control surfaces, and the actual output sequence of the system at time k respectively; are the discretized system state matrix, control matrix, and output matrix.

[0032] According to the discretized state space equation, the future output can be predicted based on the current state quantity, control quantity, and future control quantity. Its form is as follows:

[0033]

[0034] Where: N p is the prediction time domain interval; N c is the control time domain interval; respectively represent the actual deflection sequence of the system rudder surface at the (k + 1)-th, (k + 2)-th, and (k + N)-th p moments; respectively represent the desired deflection sequence of the system rudder surface at the k-th, (k + 1)-th, (k + 2)-th, and (k + N)-th p (k - 1)-th moments. Rearranging the above equation gives:

[0035] X = MX0 + CU (9)

[0036] Wherein, the definitions of each symbol are as follows:

[0037]

[0038] Combined with the system state space equation, the predicted system output equation is:

[0039]

[0040] The above derivation of the prediction model can predict the future output of the aircraft through the current state and future control quantities of the aircraft, laying a foundation for the subsequent design of the cost function for the overactuated aircraft mission requirements. During the flight of the overactuated aircraft, it is necessary to generate the required torque through the coordinated deflection of multiple rudder surfaces to complete relevant maneuvering actions. Therefore, when designing the cost function, minimizing the error between the system desired output and the predicted output is the primary goal. Secondly, in order to improve the flight stability of the aircraft, reduce the control load, and extend the service life of the rudder surface, a control quantity term is introduced into the cost function to minimize the rudder surface deflection during the maneuvering process. For the above strategy, a cost function in the following form is designed:

[0041]

[0042] Where: Q e is the system output weight matrix; R e is the control input weight matrix, and the definitions are as follows:

[0043]

[0044] Where: Q is a diagonal weight matrix with the same dimension as the system output; R is a diagonal weight matrix with the same dimension as the system control input.

[0045] Selecting an appropriate solution method to solve the optimal problem shown in the above formula (12) can obtain an optimal control sequence that minimizes the error between the expected output and the predicted output of the system and minimizes the rudder deflection. Since the predicted model is difficult to perfectly describe the real system, at time k, the first item of the calculated optimal control sequence is selected, that is, the first item u of the optimal control sequence U in formula (10). δk As the expected rudder deflection at the current moment, and then the control interval is shifted one moment to the right and the optimal problem is solved again. Repeat the above steps to achieve the effect of "predicting N p steps and moving one step".

[0046] In step (3), after calculating the expected rudder deflection of the system at the current moment, it is input into the actuator model established in step (2). The actuator model dynamically processes the expected rudder deflection, and then it is used as the actual rudder deflection to further act on the overactuated aircraft model.

[0047] Advantages of the present invention:

[0048] 1. Considering the problems of actuator redundancy, increased coupling degree, and enhanced nonlinear dynamic characteristics caused by the increase in the number of actuators of the overactuated aircraft, the control allocation strategy is designed to effectively distribute the control instructions to each actuator to ensure the stability and maneuverability of the aircraft.

[0049] 2. Considering the response characteristics of the actuator and the influence of the control efficiency comprehensively, based on the idea of the model predictive control algorithm, the control allocation method is designed starting from the dynamic characteristics of the aircraft actuator, fully considering the influence of the actuator dynamic characteristics on the control allocation, and improving the control accuracy and dynamic response performance of the aircraft.

[0050] 3. Considering the physical constraints of the actuator, such as the rudder deflection position, the rudder deflection rate, etc., the control instructions are generated starting from the physical constraints of the rudder, and the control instructions are allocated to each control rudder surface with the optimal goal. At the same time, when a fault occurs in the actuator, the control reconfiguration can be realized without re-designing the flight control law, improving the safety and maneuverability of the aircraft. Description of the Drawings

[0051] Figure 1 is the cascade control structure diagram of the overactuated aircraft flight control system;

[0052] Figure 2 is the allocation flow chart based on the model predictive control strategy;

[0053] Figure 3 is the unit step response diagram of the aircraft actuator link;

[0054] Figure 4It is a schematic diagram of the model prediction rolling optimization process;

[0055] Figure 5 It is a simulation curve graph of the roll moment coefficient command tracking under the natural frequency of 6Hz.

[0056] Figure 6 It is a simulation curve graph of the pitch moment coefficient command tracking under the natural frequency of 6Hz.

[0057] Figure 7 It is a simulation curve graph of the yaw moment coefficient command tracking under the natural frequency of 6Hz.

[0058] Figure 8 It is a simulation curve graph of the rudder deflection under the natural frequency of 6Hz.

[0059] Figure 9 It is a simulation curve graph of the command tracking under different dynamic characteristics of the actuator. Specific implementation manners

[0060] The following further illustrates the specific implementation manners of the present invention in combination with the accompanying drawings and technical solutions.

[0061] The present invention adopts a cascade control structure, reasonably distributes virtual commands through a control effectiveness mapping function, and then realizes the control of an aircraft through an actuator, and its form is as Figure 1 shown; the overall control architecture of the present invention for an overactuated aircraft considering the dynamic characteristics of the actuator is as Figure 2 shown, and its implementation steps are as follows:

[0062] The first step: Establish an overactuated aircraft model; establish a dynamic model of the overactuated aircraft, clarify the form of the control moment, and lay a foundation for the subsequent design of the control allocation method and the application of the control allocator.

[0063] The second step: Establish an aircraft actuator model; considering the actual rudder deflection response characteristics, design a servo model of the aircraft using a second-order oscillation system, and construct the state space equation of the servo link according to the designed transfer function to describe the dynamic response characteristics of the servo under the action of the control input, which helps to further study the influence of the rudder response characteristics on the flight control system and makes the flight control process closer to real engineering applications.

[0064] Step 3: Construct a prediction model for the control moment of an aircraft; based on the discretized state-space equation, using the system state variables, actual control surface deflections, and control inputs at the current moment, and combining with the sequence of future desired control surface deflections, a prediction model is constructed to predict the future system control moment. Meanwhile, a cost function that includes minimizing both the system output error and the control surface deflection is designed, where the system output error is measured by a weight matrix and the control surface deflection is constrained by the control input weight; a rolling optimization strategy is adopted, and the first term of the optimal control sequence is selected as the current control input to advance the control interval and achieve optimal control that meets the preset index requirements.

[0065] The specific steps are as follows:

[0066] Step 1: Establish an overactuated aircraft model

[0067] For the subsequent design of the control allocation method and simulation requirements, the aircraft model is expressed in the following form:

[0068]

[0069] In the formula: The rolling moment, pitching moment, and yaw moment are respectively, and their expressions can be fitted as:

[0070]

[0071] In the formula: ρ is the atmospheric density; S is the reference area of the wing; c is the mean aerodynamic chord length of the aircraft; C M 、C N are the dimensionless rolling moment coefficient, dimensionless pitching moment coefficient, and dimensionless yaw moment coefficient respectively.

[0072] Step 2: Establish an aircraft actuator model

[0073] To comprehensively consider the influence of the dynamic characteristics of the aircraft actuator on the performance of the control system, the aircraft servo mechanism is selected as the actuator, and a second-order system with amplitude and rate limitations is selected as the dynamic model of the servo mechanism, without uncertainty interference, and its form is as follows:

[0074]

[0075] In the formula: s is the Laplace operator; ω n is the natural frequency; ζ is the damping ratio; δ c , δ real are the desired control surface deflection command and the actual control surface deflection respectively. By reasonably designing the natural frequency and damping ratio, a reference model that meets the response characteristics of the actual servo mechanism is obtained, where as the natural frequency ω nWith the increase of , the system response time decreases and the response amplitude increases; with the decrease of the natural frequency, there will be an obvious lag between the output signal and the input signal. If the influence of this characteristic is not considered in the distribution, it will have a significant impact on the distribution accuracy. The unit step response diagram is as shown in Figure 3 shown.

[0076] Based on the above dynamic model of the servo link and combined with the aileron efficiency matrix of the aircraft, the aircraft system can be described in the form of state space equations as follows:

[0077]

[0078] In the formula: A δ , B δ , C δ are the system state matrix, control matrix, and output matrix respectively. For a certain aileron δ i in its state space equation and can be expressed in the following form, where ω ni is the natural frequency of the aileron δ i , and ζ i is the damping ratio of the aileron δ i .

[0079]

[0080] Step 3: Optimal control design of the aircraft considering the aileron response characteristics

[0081] During the maneuvering process of the aircraft, the attitude of the aircraft is mainly adjusted by changing the rolling moment, pitching moment, and yaw moment. Therefore, this control allocation problem can be described as:

[0082] v = Bu (6)

[0083] In the formula, v is the desired virtual control command, B is the control efficiency matrix, and u is the control input quantity.

[0084] The desired virtual control command can be expressed as:

[0085]

[0086] In the formula, v L is the virtual rolling moment coefficient command; v M is the virtual pitching moment coefficient command; v N is the virtual yaw moment coefficient command.

[0087] Set the research object to have ailerons δ a , elevators δ e , rudders δ r and flaps δlef Four actuators, so the control input is:

[0088] u = [δ a δ e δ r δ lef T (8)

[0089] The control efficiency matrix form is:

[0090]

[0091] Wherein, are respectively the derivatives of the rolling moment coefficient with respect to the aileron δ a , elevator δ e , rudder δ r and flap δ lef ; are respectively the derivatives of the pitching moment coefficient with respect to the aileron δ a , elevator δ e , rudder δ r and flap δ lef ; are respectively the derivatives of the yaw moment coefficient with respect to the aileron δ a , elevator δ e , rudder δ r and flap δ lef .

[0092] The model predictive rolling optimization process is as Figure 4 shown.

[0093] ① Given the desired virtual control command v(k|k) calculated by the controller at time k. The rolling moment, pitching moment and yaw moment generated by the aircraft at time k can be measured by sensors, and then the actual control command v real ;

[0094] ​②Optimal control is carried out based on the control interval; under the condition that the desired deflection of the servo at time k is u(k|k) as the system input, the actual flap deflection at time k, δ(k|k), can be obtained by inputting the desired servo deflection into the servo link. Combining with the control efficiency matrix, the control force / moment coefficient generated by the actual deflection of the servo at time k+1 can be predicted as Bδ(k|k). Under the condition that the desired deflection of the servo at time k+1 is u(k+1|k) as the system input, similarly, by inputting the desired servo deflection into the servo link, the actual flap deflection at time k+1, δ(k+1|k), can be obtained, and then the control force / moment coefficient generated by the actual deflection of the servo at time k+2 can be predicted as Bδ(k+1|k). Such predictions are carried out in a cycle. Assuming that the model predictive control allocates and predicts the future N steps, under the condition that the desired deflection of the servo at time k+N-1 is u(k+N-1|k), the control force / moment coefficient generated by the actual deflection of the servo at time k+N is Bδ(k+N-1|k). Combining the above analysis, it can be seen that the selection of u(k|k), u(k+1|k), …, u(k+N-1|k) in these N-step predictions is an optimization problem. By considering the minimization of the control allocation error, the allocation accuracy can be improved; secondly, considering the minimization of the flap deflection, the losses of each actuator can be reduced. Therefore, the cost function is determined as:

[0095]

[0096] In the formula, J is the cost function; δ(k+i|k) is the actual flap deflection at the i-th step in the future at time k; u(k+i|k) is the desired flap deflection at the i-th step in the future at time k; δ(k+N|k) is the actual flap deflection at the N-th step in the future at time k; Q r 、R r 、F r are the process error weight matrix, the control variable weight matrix, and the terminal error weight matrix respectively.

[0097] For the upper and lower bounds of the constraints of the optimization problem u and ΔT is the control step size, taking the comprehensive quantity of the flap position constraint and the deflection rate constraint, that is:

[0098]

[0099] In the formula, u min and u max represent the lower and upper bounds of the flap deflection of the servo link respectively; and represent the lower and upper bounds of the flap deflection rate of the servo link respectively; ΔT is the unit simulation time.

[0100] Selecting an appropriate solution method to solve the above optimal problem can obtain a control sequence U = [u(k|k) u(k+1|k) … u(k+N-1|k)] of the future N-step optimal control that satisfies the minimization of the control allocation error and the minimization of the rudder deflection.

[0101] Since it is difficult for the predicted model to perfectly describe the real system, the first item u(k|k) of the control sequence of the optimal control calculated at time k is selected as the expected rudder deflection at the current time. Then, the control interval is shifted one time step to the right and the optimal problem is solved again. By repeating the above steps, the effect of "predicting N steps and taking one step" is achieved to complete the prediction.

[0102] Based on the designed control allocation strategy, simulation experiments are carried out for verification: First, the aircraft maintains level flight at an altitude of 5000 m and a speed of 200 m / s. Then, it tracks the dimensionless aerodynamic coefficient pseudo-control commands on the roll, pitch, and yaw axes generated by the upper controller to achieve relevant flight maneuvers. The flight simulation frequency is set to 50 Hz.

[0103] Table 1 shows the aircraft data and state data:

[0104] Table 1 Simulation parameters

[0105]

[0106] The simulation results when the natural frequency of the actuator is 6 Hz are as Figures 5 - 9 shown. Based on the aircraft control allocation method considering the dynamic characteristics of the actuator proposed in the present invention for allocation, it can achieve good tracking of the virtual control commands generated by the upper controller. Taking the root mean square error of the three-axis command tracking as the performance index, as can be seen from Figures 5 - 7 , when the actuator dynamics characteristics have a relatively small bandwidth, the root mean square errors of the virtual roll moment coefficient, pitch moment coefficient, and yaw moment coefficient based on the model predictive control allocation method considering the actuator dynamics characteristics are 0.0013, 0.0058, and 0.0014 respectively, showing good allocation accuracy; as can be seen from Figure 8 the rudder deflection curve, the method proposed in the present invention can make the actual rudder deflection of the aircraft not only operate within the physical constraints of the actuator, but also the actuator deflection is smoother, which is beneficial to extending the service life of the actuator; as can be seen from Figure 9 , as the natural frequency of the actuator decreases, the method proposed in the present invention can still maintain good allocation accuracy, and the allocation of the virtual control commands can fully consider the dynamic characteristics of the actuator.

Claims

1. A method for allocating control of an aircraft considering the dynamic characteristics of actuators, Step 1: Build an Overdriven Aircraft Model The aircraft model is expressed as follows: Where: M, N are rolling moment, pitching moment, and yaw moment respectively, and their expressions are fitted as follows: Where: ρ is the atmospheric density; S is the reference area of ​​the wing; is the average aerodynamic chord length of the aircraft; C M , C N They are dimensionless rolling moment coefficient, dimensionless pitching moment coefficient and dimensionless yaw moment coefficient respectively; Step 2: Establish the actuator model of the aircraft The aircraft servo link is selected as the actuator, and the second-order system with amplitude and rate limits is selected as the servo link dynamic model, which is shown as follows: Where: s is the Laplace operator; ω n is the natural frequency; ζ is the damping ratio; δ c ,δ real They are the expected rudder surface deflection command and the actual rudder surface deflection respectively; Based on the above-mentioned dynamic model of the steering gear link and combined with the aircraft control surface efficiency matrix, the aircraft system is described in the form of a state space equation, as shown in the following formula: Where: A δ , B δ , C δ They are the system state matrix, control matrix, and output matrix respectively. For a certain control surface δ i In its state space equation and It can be expressed as follows, where ω ni is the rudder surface δ i The natural frequency, ζ i is the rudder surface δ i Damping ratio; Step 3: Optimal control design of aircraft considering control surface response characteristics During the maneuvering process, the aircraft attitude is adjusted by changing the rolling moment, pitch moment and yaw moment. Therefore, the control allocation problem can be described as: v=Bu (6) In the formula, v is the expected virtual control command, B is the control efficiency matrix, and u is the control input; The expected virtual control instructions are expressed as: In the formula, is the virtual rolling moment coefficient instruction; v M is the virtual pitch moment coefficient instruction; v N is the virtual yaw moment coefficient instruction; Set the research object to have aileron δ a 、Elevator δ e 、Rudder δ r and flap δ lef There are four actuators, so the control input is: u=[δ a d e d r d lef ] T (8) The control efficiency matrix is ​​in the form of: in, are the rolling moment coefficients about the aileron δ a 、Elevator δ e 、Rudder δ r and flap δ lef The derivative of are the pitch moment coefficients about the aileron δ a 、Elevator δ e 、Rudder δ r and flap δ lef The derivative of are the yaw moment coefficients about the aileron δ a 、Elevator δ e 、Rudder δ r and flap δ lef The derivative of The model prediction rolling optimization process is as follows: ① The expected virtual control command calculated by the controller at time k is known to be v(k|k); the rolling moment, pitch moment and yaw moment generated by the aircraft at time k are measured by sensors, and then the actual control command v generated by the deflection of the control surface is calculated real ; ② Optimize control based on the control interval; when the expected deflection of the servo as the system input at time k is u(k|k), the expected deflection of the servo is input into the servo link to obtain the actual deflection of the rudder surface at time k(k|k), and the control efficiency matrix is ​​combined to predict that the control force / torque coefficient generated by the actual deflection of the servo at time k+1 is Bδ(k|k). When the expected deflection of the servo as the system input at time k+1 is u(k+1|k), similarly, the expected deflection of the servo is input into the servo link to obtain the actual deflection of the rudder surface at time k+1 δ(k+1|k), and then the control force / torque coefficient generated by the actual deflection of the servo at time k+2 is predicted. The moment coefficient is Bδ(k+1|k), and the prediction is repeated in this way. Assuming that the model predictive control distribution predicts the next N steps, when the expected deflection of the servo at time k+N-1 is u(k+N-1|k), the control force / torque coefficient generated by the actual deflection of the servo at time k+N is Bδ(k+N-1|k); in this N-step prediction, the selection of u(k|k), u(k+1|k),…, u(k+N-1|k) is an optimization problem. By considering the minimization of the control distribution error, the distribution accuracy can be improved; secondly, considering the minimization of the deflection of the rudder surface, the loss of each actuator can be reduced, so the cost function is determined as: Where J is the cost function; δ(k+i|k) is the actual deflection of the rudder at the i-th step in the future at time k; u(k+i|k) is the expected deflection of the rudder at the i-th step in the future at time k; δ(k+N|k) is the actual deflection of the rudder at the N-th step in the future at time k; Q r , R r 、F r They are process error weight matrix, control quantity weight matrix and terminal error weight matrix respectively; For the optimization problem, the upper and lower bounds u and ΔT is the control step length, and the comprehensive quantity of the control surface position constraint and the deflection rate constraint is taken as: In the formula, u min and u max They represent the lower and upper limits of the deflection of the servo link respectively; and They represent the lower and upper bounds of the steering gear link deflection rate respectively; ΔT is the unit simulation time; The above optimal problem is solved to obtain the control sequence U=[u(k|k)u(k+1|k)…u(k+N-1|k)] of the future N-step optimal control that satisfies the minimization of the control allocation error and the minimization of the rudder surface deflection; At time k, the first item u(k|k) of the optimal control sequence is selected as the expected rudder deflection at the current moment, and then the control interval is moved to the right for one moment to solve the optimal problem again. The above steps are repeated to complete the prediction.

Images