Fractional power feedback and aggregation compensation cooperative fault-tolerant control method for tailless flying wing aircraft

Abstract

The present application belongs to the field of flight control and active fault-tolerant control technology, and relates to a tailless flying wing aircraft fault-tolerant control method based on fractional power feedback and aggregation compensation coordination. The present application uses angular velocity measurement and causal angular acceleration estimation to establish an incremental control interface; linear feedback and fractional power nonlinear fast feedback are constructed according to attitude error and angular velocity error; a three-axis aggregation adaptive compensation state is driven by using the prediction residual between the actual angular acceleration response and the integrated virtual angular acceleration at the last moment; the fractional power feedback term and the aggregation compensation term are jointly added to the virtual angular acceleration control quantity; then, the nominal control effectiveness matrix is used to generate the rudder surface command, and the rudder surface position and speed constraints are uniformly applied. The method coordinates fractional power nonlinear error feedback, three-axis aggregation adaptive compensation driven by residual error and incremental backstepping control.

Description

Technical Field

[0001] This invention belongs to the field of flight control and active fault-tolerant control technology, and relates to a fault-tolerant control method for tailless flying wing aircraft that combines fractional power feedback and aggregate compensation. It is an attitude fault-tolerant control method for tailless flying wing aircraft when the control surfaces suffer performance loss. This method coordinates fractional power nonlinear error feedback, residual-driven three-axis aggregate adaptive compensation, and incremental backstepping control. Background Technology

[0002] Tailless flying wing aircraft integrate the wing and fuselage into a single unit, typically eliminating conventional horizontal and vertical stabilizers. They are characterized by high aerodynamic efficiency, low drag, and low detectability. Due to the lack of a traditional tail, the control moments for roll, pitch, and yaw in tailless flying wing aircraft are generated by multiple trailing-edge control surfaces in coordination. Different control surfaces often participate in two or three moment channels simultaneously, thus the attitude control of tailless flying wing aircraft is characterized by multiple control surfaces, over-actuation, and channel coupling.

[0003] Control surface performance loss is a common type of failure. After this type of failure occurs, the control surface position feedback can still be largely consistent with the command, but the actual aerodynamic torque generated by the same command deflection is lower than the nominal aerodynamic torque under healthy conditions. If the failure is not compensated for in time in the early stages, tailless flying wing aircraft will experience increased attitude tracking error, angular velocity fluctuations, and control surface load redistribution. In cases of persistent failure, the control burden on the remaining healthy control surfaces may also increase, potentially triggering position or rate limits.

[0004] Existing nonlinear dynamic inverse, incremental nonlinear dynamic inverse, and incremental backstepping control methods can construct local control relationships using the angular acceleration and control increment at the current or previous sampling time. These methods typically combine the nominal reference torque and the local nominal control effectiveness matrix to form control corrections, eliminating the need for online identification and inverse solving of the complete nonlinear control mapping after a fault. However, when the control surface experiences performance loss, a deviation occurs between the nominal control effectiveness matrix and the actual control effectiveness. The fault's impact manifests as a lack of unknown aggregated control in the three-axis angular acceleration channels. While the controller can gradually recover attitude by increasing the error feedback using only conventional linear error feedback, it struggles to simultaneously address both rapid recovery after fault detection and compensation for unknown gaps during the sustained fault phase.

[0005] Fractional power nonlinear feedback introduces a power term less than one into the error feedback, maintaining a strong correction effect even when the error is small, thus reducing the transient trailing of conventional linear feedback. Applying this type of feedback to the attitude and angular velocity channels can enhance error correction capabilities during fault recovery, but its actual effectiveness is still affected by the severity of the fault, control parameters, flight conditions, and control surface constraints. Fractional power feedback essentially still generates control based on tracking error and does not directly compensate for the continuous loss of convergent control caused by decreased control surface effectiveness.

[0006] Adaptive fault-tolerant control methods can adjust the compensation amount online based on tracking error, prediction error, or observation residual, thus mitigating the impact of faults even without knowing their precise magnitude. Among existing adaptive methods, some require identifying the specific faulty control surface and its remaining effectiveness, while others rely on neural networks or complex nonlinear approximators. For multi-control-surface flying wing aircraft, surface-by-surface identification may be affected by channel coupling and the correlation of control derivatives; complex approximators also increase the burden of parameter selection and real-time computation. Furthermore, when using adaptive compensation alone, the establishment of the compensation state from its initial value to the formation of effective compensation requires a certain process, and the short-term attitude error after fault detection may still be relatively large.

[0007] Therefore, fractional power feedback and aggregate adaptive compensation address transient error correction and continuous control deviation compensation, respectively, and their shortcomings are complementary in terms of time: fractional power feedback is fast but cannot eliminate persistent control deficiencies; aggregate adaptive compensation can eliminate persistent deficiencies, but it takes time for its compensation state to establish from the initial value to form effective compensation. Therefore, using either one alone is insufficient to cover the complete process from fault detection to attitude recovery and maintenance. Furthermore, designing them as independent controllers and generating separate control surface commands would lead to unclear signal interfaces, command competition, and complex engineering implementation.

[0008] Therefore, the field needs a control surface fault-tolerant control method that can quickly correct attitude errors in a short period of time after fault detection, compensate for unknown aggregate control deficiencies during the continuous fault phase, and generate only one set of control surface commands throughout the entire process, without requiring faulty control surface location and single control surface performance identification. Summary of the Invention

[0009] To address the issues of insufficient attitude recovery speed, difficulty in online compensation of continuous control deviations, and reliance on specific fault location and performance identification in tailless flying wing aircraft after control surface performance loss, this invention provides a fault-tolerant control method for tailless flying wing aircraft that combines fractional power feedback and aggregate compensation. This method is applicable to tailless flying wing aircraft with multiple aerodynamic control surfaces and can also be used in other fixed-wing aircraft with redundant or multi-channel control surfaces.

[0010] This invention establishes an incremental control interface using angular velocity measurements and causal angular acceleration estimation; constructs linear feedback and fractional-power nonlinear fast feedback based on attitude and angular velocity errors; drives the three-axis aggregation adaptive compensation state using the predicted residual between the actual angular acceleration response and the previous time-series integrated virtual angular acceleration; incorporates the fractional-power feedback term and the aggregation compensation term into the virtual angular acceleration control quantity; and generates control surface commands using the nominal control efficiency matrix, uniformly applying control surface position and rate constraints. This process does not require knowledge of specific faulty control surfaces or the residual efficiency coefficients of each control surface. The overall technical effect stems from the coordinated action of rapid error correction and continuous deviation compensation within the same incremental backstepping closed loop, rather than the effect necessarily obtained by using any single module alone.

[0011] To achieve the above objectives, the technical solution of the present invention is as follows: A fault-tolerant control method for a tailless flying wing aircraft combining fractional power feedback and aggregate compensation is disclosed. Steps 3 to 6 are the core of this invention: Step 3 generates a fast-response error correction action; Step 4 obtains the basis for judging the degree of control action deficiency; Step 5 forms a control action to eliminate persistent deviations based on this basis; and Step 6 coordinates the above two types of actions into a single control quantity. Steps 1 and 2 provide mathematical representations of input signals and faults, and steps 7 and 8 provide the execution interface from virtual control quantity to actual control surface commands. The specific steps are as follows: Step 1: Obtain flight status and construct control variables; Step 1 is the starting point of the entire control flow and does not follow previous calculations. This step reads the attitude, angular velocity, and control surface position data from the tailless flying wing aircraft. First, the attitude command is converted into an angular velocity reference value by the attitude outer loop, and then the attitude error and angular velocity error are calculated separately. At the same time, the angular velocity data is processed through causal filtering and differential to obtain an angular acceleration estimate. The three quantities obtained in this step—attitude error, angular velocity error, and angular acceleration estimate—are sent to steps 3, 4, and 7, respectively, and serve as inputs for all subsequent calculations.

[0012] Obtain the attitude, angular velocity, control surface position, and flight status information of the tailless flying wing aircraft. Define the attitude output vector and the aircraft's angular velocity vector as shown in formula (1): (1) In the formula, This is the attitude output vector. The angular velocity vector of the machine system. For roll angle, The pitch angle, Sideslip angle, , , These are the aircraft's roll, pitch, and yaw angular velocities, respectively. The attitude command vector is shown in formula (2): (2) In the formula, This is the attitude command vector. , and These are the roll angle command, pitch angle command, and sideslip angle command.

[0013] The attitude error is constructed based on the attitude command and the current attitude output vector, and is used as the degree to which the current attitude deviates from the command. The attitude error is shown in formula (3): (3) In the formula, This is the attitude error vector.

[0014] Attitude angles and attitude errors are uniformly expressed in radians. (This is represented by the symbol ) indicates that the angular velocity reference quantity is generated from the attitude outer loop. The angular velocity reference value can be generated using linear attitude feedback, attitude feedback with feedforward, or attitude feedback with amplitude limiting. The angular velocity error is the degree to which the current angular velocity deviates from the reference value, and is defined as shown in formula (4): (4) In the formula, This is the angular velocity error vector. The angular velocity reference value is generated from the attitude outer ring.

[0015] Angular velocity and angular velocity error are uniformly adopted This indicates that, to ensure the causality of the digital control implementation, the angular acceleration estimate is obtained from the angular velocity measurement through causal filtering and differential processing. The angular acceleration estimation can use the filtered results available at the previous sampling time, or it can use other causal estimation methods that do not use future data. In the above measurements, attitude, angular velocity, and flight state information are used to construct control quantities in this step and subsequent steps; the control surface position measurement is used to confirm the actual response of the control surface and can be used as a reference value for the rate constraint in step 8.

[0016] Step 2: Establish an aggregated equivalent representation of the control surface effectiveness loss; Step 2 does not involve online calculations; its purpose is to determine the compensation target for subsequent compensation. This step first provides the rotational dynamics of the tailless flying wing aircraft around its center of mass, then expresses the control surface effectiveness loss as a multiplicative coefficient. Next, under the same command and flight conditions, the nominal aerodynamic torque corresponding to the healthy state is subtracted from the actual aerodynamic torque after the failure, yielding the missing aerodynamic torque. Finally, this is multiplied by the inverse of the rotational inertia matrix and negativeened, then converted to the angular acceleration channel, resulting in an equivalent convergent control loss across three axes. This demonstrates that regardless of which one or more control surfaces experience effectiveness loss, its impact ultimately converges into an unknown missing quantity in the three angular acceleration channels. Therefore, subsequent control only needs to compensate for these three quantities, without needing to identify the effectiveness coefficients of each control surface individually.

[0017] The rotational dynamics of a tailless flying wing aircraft about its center of mass is represented by equation (5): (5) In the formula, Here is the rotational inertia matrix. The system's angular acceleration vector. This represents the actual aerodynamic torque. Assume the tailless flying wing aircraft has... The command vector of the control surfaces of a tailless flying wing aircraft is shown in formula (6): (6) In the formula, For the control surface command vector, The total number of rudder surfaces, For vectors The The component, i.e., the first component The command deflection of each control surface .

[0018] The loss of control surface effectiveness is expressed by a multiplicative coefficient as shown in formula (7): (7) In the formula, The equivalent control surface deflection that actually produces aerodynamic effects. For the control surface effectiveness matrix, for Diagonal matrix of order 1 This is an operator that constructs a diagonal matrix using the elements within the parentheses as diagonal elements. For matrix The The diagonal element, i.e. the first... The efficiency coefficient of each control surface , Indicates the first Each rudder surface is healthy. This indicates that the control surface has experienced a partial loss of effectiveness.

[0019] Under the same control surface commands and flight conditions, let the nominal aerodynamic torque corresponding to the healthy state be... The loss of aerodynamic torque caused by the fault is shown in formula (8): (8) In the formula, The loss of aerodynamic torque is caused by a malfunction.

[0020] The aerodynamic torque loss caused by the fault is converted into an equivalent convergent control loss in the angular acceleration channel, as shown in Equation (9): (9) In the formula, The equivalent convergence control is missing in the angular acceleration channel.

[0021] Thus, the angular acceleration after the fault is expressed as shown in formula (10): (10) In the formula, This refers to the angular acceleration obtained under the same control commands based on the nominal control surface effectiveness. This invention incorporates the combined effects of multiple faulty rudder surfaces on the three-axis angular acceleration. It only compensates for this three-axis aggregation and does not... The efficiency coefficients in the data are identified online.

[0022] Step 3: Construct a fractional power nonlinear fast feedback; Step 3, following the attitude error and angular velocity error obtained in Step 1, generates a fast-responding fractional power feedback term in the integrated virtual angular acceleration control quantity. This step first defines a fractional power nonlinear function, then applies this function to the angular velocity error and attitude error respectively. Together with the derivatives of the linear angular velocity feedback and the angular velocity reference quantity, it forms the fast fractional power feedback part, resulting in a three-axis virtual angular acceleration metric, which is then sent to Step 6. Its function is as follows: conventional linear feedback reduces its correction force proportionally when the error is small, causing slow error convergence and transient tailing; if the first power of the error is changed to a power less than one, this term becomes relatively larger when the error is small, thus maintaining a strong correction effect in the small error range.

[0023] The fractional power nonlinear function is defined as shown in formula (11): (11) In the formula, It is a fractional power nonlinear function. For any scalar error component, For symbolic functions, It is a fractional power exponent. When the independent variable is a vector, equation (11) calculates each component of the vector separately, and the results are arranged in the original order to form a vector of the same dimension. The fractional power fast feedback part based on attitude error and angular velocity error is constructed as shown in equation (12): (12) In the formula, For rapid feedback of virtual angular acceleration components in fractional powers, The bounded derivative or difference estimate of the angular velocity reference quantity. The linear feedback gain matrix for angular velocity. The fractional power feedback gain matrix for angular velocity error. The three terms are the fractional power feedback gain matrices for attitude error, and are all positive definite diagonal matrices. The linear angular velocity feedback is used to provide basic damping, and the fractional power terms of angular velocity error and attitude error are used to improve the correction strength in the small error interval after a fault, thereby reducing transient tailing.

[0024] The fractional power fast feedback section does not require direct use of fault type, faulty control surface, or efficiency coefficient. The output of the fractional power fast feedback section is a three-axis virtual angular acceleration metric, which has the same physical dimension and control interface as the aggregated adaptive compensation quantity obtained in step 5. This invention uses the fractional power fast feedback section to enhance the error correction capability after fault detection, but does not claim that the error reaches zero within a strictly finite time.

[0025] This invention also allows for different fractional power exponents to be used for the angular velocity error term and the attitude error term. In this case, equation (11) is calculated separately according to their respective exponents, without changing the control flow of this invention.

[0026] Step 4: Construct the triaxial convergent prediction residuals; Step 4, following the angular acceleration estimate obtained in Step 1 and the combined virtual angular acceleration control quantity formed in Step 6 at the previous sampling time, provides a basis for judging the degree of control action deficiency. This step subtracts the angular acceleration required by the controller at the previous sampling time from the actual angular acceleration performed by the tailless flying wing aircraft at that time, obtaining the three-axis aggregated prediction residual, which is then sent to Step 5. Its purpose is: this residual reflects the difference between the controller's expected response and the actual response; the greater the deficiency, the larger the residual; this residual is obtained entirely from measurements that have already occurred, without using future data.

[0027] remember The discrete sampling time number. For the first The comprehensive virtual angular acceleration control quantity formed by the coordinated fractional power fast feedback and aggregate adaptive compensation at each sampling time is given in step 6. The complete expression is given in step 6. The triaxial aggregate prediction residual is constructed using the comprehensive virtual angular acceleration at the previous time and the currently available angular acceleration estimate as shown in formula (13): (13) In the formula, For the first The three-axis aggregated prediction residuals at each sampling time point This is an estimate of the angular acceleration at the previous sampling time. This is the combined virtual angular acceleration control quantity formed at the previous sampling time. The prediction residual reflects the difference between the actual angular acceleration response and the controller's expected response, which includes persistent control deviations caused by control surface performance losses, and may also include measurement noise, angular acceleration estimation errors, filtering and differential errors, and local incremental model errors.

[0028] To reduce the impact of noise and outliers on adaptive updates, one can... Amplitude limiting or low-pass processing can be applied. This processing only needs to maintain causality and does not require the use of real angular acceleration from the tailless flying wing aircraft simulation model, nor does it require the use of real fault torque.

[0029] Step 5: Construct residual-driven aggregation adaptive compensation with projection and updated limiting; Step 5, following the three-axis aggregated prediction residuals obtained in Step 4, generates an aggregated compensation term in the integrated virtual angular acceleration control quantity used to eliminate persistent deviations. This step first constructs an unconstrained adaptive update quantity from the prediction residuals, then applies a rate limit to this update quantity, and uses a projection operator to keep the update result within a preset bounded set. The three-axis aggregated compensation state is obtained by updating step-by-step and sent to Step 6. Its function is to continuously accumulate the angular acceleration deficiency reflected by the prediction residuals, so that the controller can make additional angular acceleration demands in subsequent commands based on this deficiency, thereby offsetting the control action generated by the reduced control surface.

[0030] set up This represents a triaxial aggregation compensation state. An unconstrained adaptive update amount is constructed based on the predicted residuals. As shown in formula (14): (14) In the formula, To adaptively update the gain matrix, To predict the residual feedback gain matrix, For the first The three-axis aggregation compensation state at each sampling time. Let be the leakage gain matrix. The predicted residual is used to drive the aggregate compensation state to update in the direction of reducing the persistent response bias, as shown in equation (14). The term is used to adjust the response ratio between the predicted residual and the aggregated compensation state and to suppress its continuous growth, in equation (14). The term causes the aggregate compensation state to decay proportionally, which is used to prevent the aggregate compensation state from accumulating to the boundary when the prediction residual deviates from zero for a long time.

[0031] A rate limit is imposed on the update quantity as shown in Equation (15): (15) In the formula, Let be the upper bound vector of the positive update rate for the three angular acceleration channels. This is the angular acceleration channel number. For vectors The The component, i.e., the first component The upper limit of the update rate for each channel. , ; To restrict input by component Saturation operators within, This is the adaptive update amount after amplitude limiting.

[0032] The aggregation compensation state is kept within a pre-defined bounded set by using the projection operator. Inside: (16) In the formula, For the first The three-axis aggregation compensation state at each sampling time. For a predefined bounded closed set, The sampling period is Represents a predefined bounded set The projection. The projection boundaries and update limits are determined based on the allowable angular acceleration range, control surface capabilities, and sensor noise levels of the tailless flying wing aircraft. Preset bounded set. The specific form is not limited, and the method of taking the form in this invention is shown in equation (23).

[0033] Upon receiving a fault detection signal, the aggregate adaptive update shown in formula (16) is initiated; when the aggregate adaptive update is not initiated, the process is maintained. This is a preset initial state. The fault detection signal is only used to determine whether the aggregate compensation state should be updated, and does not require the identification of the faulty control surface or its remaining effectiveness.

[0034] Step 6: Coordinate rapid feedback of fractional powers and aggregate adaptive compensation; Step 6, which connects the fractional power fast feedback part obtained in Step 3 and the three-axis aggregated compensation state obtained in Step 5, is a key step in this invention. This step directly adds the fractional power fast feedback part and the three-axis aggregated compensation state at the virtual angular acceleration layer to obtain the comprehensive virtual angular acceleration control quantity. Since both outputs are three-axis angular acceleration metrics, having the same physical dimensions and the same downstream interface, they can be merged into a single control quantity without needing to generate separate control surface commands and then attempt to merge them. This control quantity is sent to Step 7 for torque conversion and returned to Step 4 at the next sampling time to participate in the construction of the prediction residual.

[0035] The fractional power fast feedback part and the aggregate compensation state are coordinated in the virtual angular acceleration layer to obtain the comprehensive virtual angular acceleration control quantity as shown in formula (17): (17) In the formula, For the first Bounded derivative or difference estimate of the angular velocity reference at each sampling time. For the first angular velocity error at each sampling time, For the first The attitude error at each sampling time. This is the aggregate compensation gain matrix. In equation (17)... That is, the fractional power fast feedback part defined in equation (12) In the The values ​​taken at each sampling time; Since it is a aggregation compensation term, equation (17) can also be written as , For the rapid feedback section of fractional powers In the The values ​​at each sampling time point. Fractional power fast feedback and aggregate adaptive compensation are both entered into equation (17) through the three-axis virtual angular acceleration interface to form a single virtual control command, avoiding multiple independent fault-tolerant controllers from generating competing rudder surface commands.

[0036] Before the convergent compensation state is fully established, fractional power feedback enhances the transient correction effect based on attitude error and angular velocity error; as... Based on progressive updates of the predicted residuals, aggregate adaptive compensation is used to reduce residual control deviations caused by persistent faults. Fractional power fast feedback and aggregate adaptive compensation address different control requirements and coordinate their influence on closed-loop control through a unified virtual angular acceleration interface.

[0037] Step 7: Generate torque commands based on incremental dynamics; Step 7 takes the integrated virtual angular acceleration control quantity obtained in Step 6 and the angular acceleration estimate obtained in Step 1, and converts the angular acceleration control quantity into the torque required by the tailless flying wing aircraft. This step first constructs the current measured torque from the current angular velocity and angular acceleration estimate, then uses this measured torque as a reference, adds a correction amount proportional to the angular acceleration increment, and obtains the incremental torque command, which is sent to Step 8. Its purpose is to use the measured torque already achieved by the tailless flying wing aircraft as a reference, calculating only the additional torque increment needed to achieve the desired angular acceleration, thereby avoiding solving for the complete nonlinear aerodynamic relationship after the fault at each sampling time.

[0038] The current measured torque is estimated and constructed based on the current angular velocity and causal angular acceleration as shown in formula (18): (18) In the formula, For the first The measured torque at each sampling time, For the first The angular velocity of the machine system at each sampling time. This is the angular acceleration estimate at the previous sampling time.

[0039] The incremental torque command is shown in formula (19): (19) In the formula, For the first Incremental torque command at each sampling time, The incremental correction coefficient is used to adjust the magnitude of the incremental torque correction. Equation (19) uses the measured torque already achieved by the tailless flying wing aircraft as a reference, and applies only a correction amount proportional to the angular acceleration increment. The form of the angular acceleration increment is consistent with the existing incremental nonlinear dynamic inverse and incremental backstepping methods. It only requires the nominal reference torque and the local nominal control performance matrix, and does not require online identification and inverse solution of the complete nonlinear control mapping after the fault. This invention does not regard this item as its improvement over the existing incremental methods. The improvement lies in the coordination of the two types of control actions described in steps 3 to 6. The measured torque feedback reflects the control action already achieved by the tailless flying wing aircraft, while the aggregated compensation state is used to correct the residual control deviation that persists in the prediction residual. The two work in coordination through the same virtual control quantity and the same control allocation interface.

[0040] Step 8: Use the nominal control effectiveness matrix to complete the control surface allocation and apply constraints; Step 8 receives the incremental torque command from Step 7 and obtains the final command sent to the control surfaces. This step uses the generalized inverse of the nominal control effectiveness matrix under healthy conditions to allocate the torque command to candidate commands for each control surface, and then applies position and velocity constraints to the candidate commands. Its purpose is to implement the torque requirements of the three channels onto specific control surfaces and ensure that the issued commands do not exceed the control surface's capabilities. Since fault compensation has already been completed in Steps 5 and 6, the incremental torque command already contains the torque requirements corresponding to the aggregated compensation state, and the allocation process will automatically distribute it to all control surfaces; therefore, it is not necessary to reconstruct the matrix after the fault.

[0041] set up To balance the rudder surface vector, To determine the reference aerodynamic moment for trimming the control surfaces under the current flight conditions. Given the nominal local control performance matrix, the unconstrained control surface commands are as shown in formula (20): (20) In the formula, For the first Unconstrained control surface candidate commands at each sampling time. This represents the Moore–Penrose generalized inverse. and It can be obtained from the nominal aerodynamic model of a tailless flying wing aircraft, aerodynamic database, or a predetermined local linearization relationship; The present invention provides a nominal control performance matrix under healthy conditions, eliminating the need to reconstruct the control performance matrix using fault performance coefficients after a fault.

[0042] remember For the first The control surface command is finally sent to the control surface after applying position and rate constraints at each sampling time. and These are the lower limit vector and upper limit vector of each control surface position, respectively. This is a vector composed of the maximum permissible positive velocities of each control surface. The control surface command issued at the previous sampling time is used. Control surface position and velocity constraints are applied to the unconstrained control surface candidate commands obtained from equation (20) as shown in equation (21): (twenty one) The constrained control surface command is sent to the control surface, and steps 1 to 8 are repeated at the next sampling time, thus forming a closed-loop fault-tolerant control.

[0043] The beneficial effects of this invention are: (1) This invention coordinates fractional power nonlinear feedback and residual-driven aggregate adaptive compensation within the same incremental backstepping virtual control layer, making them complementary in time: in the short period after fault detection, fractional power feedback provides a strong error correction effect; after the aggregate compensation state is established, aggregate adaptive compensation eliminates the persistent control deficiency. Together, they cover the complete process from fault detection to attitude recovery and maintenance. In specific example verification, the method of this invention achieves the lowest values ​​in all six root mean square error indices across the two evaluation stages (fault recovery stage and persistent fault stage) and the three attitude channels (roll, pitch, and sideslip), while outperforming methods containing only fractional power feedback, methods containing only aggregate adaptive compensation, and the nonlinear dynamic inverse method used as a conventional engineering reference. This embodiment also shows that the two single-component methods have their own advantages and disadvantages in different channels and stages, and are not completely dominant to each other. The method containing only fractional power feedback is not superior to the nonlinear dynamic inverse method in the roll and pitch channels. Therefore, the above performance improvement comes from the coordination of the two types of control actions in the same closed loop, rather than the effect necessarily obtained by using any one component module alone.

[0044] (2) This invention only requires a detection signal indicating that a fault has occurred. It does not require the fault diagnosis module to provide the identity, location, or remaining efficiency coefficient of the faulty control surface, nor does it reconstruct the control efficiency matrix after the fault occurs. Therefore, this invention can be used in the control stage where the fault has been detected but detailed fault information is not yet available. It can also be used as an independent fault-tolerant control scheme without configuring a per-control-surface efficiency identification module, thereby avoiding the difficulties of per-control-surface identification under conditions of multi-control-surface coupling and control derivative correlation.

[0045] (3) The rapid correction effect in this invention is provided by fractional power nonlinear feedback of attitude error and angular velocity error. Since the power used is less than one, this term still maintains a strong correction strength relative to linear feedback when the error is small, thereby reducing transient tailing when only linear feedback is used.

[0046] (4) The continuous deviation compensation function in this invention is provided by the three-axis convergence compensation state. This state updates online based on the angular acceleration prediction residual, and merges the combined effect of one or more control surface performance losses into the convergence amount on the three angular acceleration channels for compensation, thus eliminating the need to identify each faulty control surface individually.

[0047] (5) The present invention enables the above two types of control actions to generate control surface commands through the same virtual angular acceleration interface, the same incremental torque interface and the same control allocation link, avoiding interface conflicts and command competition caused by multiple independent fault-tolerant controllers outputting control surface commands respectively, which facilitates engineering implementation.

[0048] (6) The projection, update limit, and control surface position and rate constraints in this invention can limit the amplitude of the aggregation compensation state and control surface commands, so that the control quantity is always kept within the preset control surface capability range, which is convenient to implement in the digital flight control system. Attached Figure Description

[0049] Figure 1 This is the overall control structure diagram of the fractional power feedback and aggregate adaptive compensation collaborative fault-tolerant control method of the present invention; Figure 2 These are comparison diagrams of roll angle, pitch angle and sideslip angle responses of different control methods under the condition of loss of control surface effectiveness. Among them, (a) is a comparison diagram of roll angle response, (b) is a comparison diagram of pitch angle response, and (c) is a comparison diagram of sideslip angle response. Figure 3 These are comparison diagrams of roll, pitch, and sideslip errors in the recovery phase after a fault occurs between the fractional power feedback incremental backstepping method and the method of this invention. Among them, (a) is a comparison diagram of roll error, (b) is a comparison diagram of pitch error, and (c) is a comparison diagram of sideslip error. Figure 4 These are comparison diagrams of the offline fault equivalent quantity and the three-axis converged compensation state of the method of the present invention, wherein (a) is a comparison diagram of the roll channel offline fault equivalent quantity and the converged compensation state, (b) is a comparison diagram of the pitch channel offline fault equivalent quantity and the converged compensation state, and (c) is a comparison diagram of the yaw channel offline fault equivalent quantity and the converged compensation state. Figure 5 These are comparison charts of attitude root mean square error for different control methods during the fault recovery phase and the continuous fault phase. (a) is a comparison chart of attitude root mean square error during the fault recovery phase from 31 to 35 seconds, and (b) is a comparison chart of attitude root mean square error during the continuous fault phase from 31 to 60 seconds. Detailed Implementation

[0050] The embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0051] This embodiment uses an ICE tailless flying wing nonlinear aircraft model. The tailless flying wing aircraft achieves attitude control through six trailing-edge aerodynamic control surfaces: left elevon ELE_L, right elevon ELE_R, left pitch flap PF_L, right pitch flap PF_R, left all-moving wingtip AMT_L, and right all-moving wingtip AMT_R. These six control surfaces simultaneously participate in roll, pitch, and yaw moment control, forming a multi-input, three-moment-channel overacting system.

[0052] The overall control structure of the present invention is as follows: Figure 1As shown. The flight state measurement module provides attitude angle, angular velocity, and angular acceleration estimates to the attitude outer loop, fractional power feedback module, and incremental control module. The attitude outer loop generates angular velocity reference quantities based on attitude commands; the fractional power feedback module generates fast feedback terms based on attitude and angular velocity errors; the residual-driven aggregation adaptive module updates the three-axis aggregation compensation state based on angular acceleration prediction residuals; the two types of control actions are coordinated in the virtual angular acceleration control quantity; subsequently, six control surface commands are generated through the incremental torque interface, nominal control effectiveness matrix, and control surface constraints. Step 2 is the aggregated equivalent representation of control surface effectiveness loss, which belongs to the mathematical definition of the compensation object and does not correspond to... Figure 1 An independent component within it.

[0053] In this embodiment, the initial simulation state and control surface constraints are shown in Table 1.

[0054] Table 1 Initial flight state and control surface constraints of the embodiment

[0055] At each sampling time, attitude angle and angular velocity measurements are first read. The angular velocity measurements are then differentially analyzed after causal filtering, and the differential result is filtered again to obtain an angular acceleration estimate usable from the previous sampling period. This angular acceleration estimate is calculated from the measured signals and does not use the actual angular acceleration output from the tailless flying wing aircraft model.

[0056] The attitude outer loop generates an angular velocity reference value based on roll, pitch, and sideslip commands. The sideslip command is set to zero to maintain coordinated flight. The attitude error and angular velocity error are then calculated according to equations (3) and (4), and the fractional power fast feedback part is calculated according to equation (12).

[0057] This embodiment uses the same power exponent for both attitude error and angular velocity error. To avoid numerical discontinuities near zero error, Calculations are performed on a component-by-component basis, and numerical protection intervals can be set for extremely small errors in the digital implementation. Fractional power feedback requires no switching of the controller structure and is continuously calculated using the same expression during both normal and faulty flight phases. This embodiment utilizes this structure to enhance error correction after fault detection, but does not present it as a strictly finite-time convergence proof.

[0058] At sampling time Based on the comprehensive virtual angular acceleration at the previous sampling time Calculate the predicted residuals using currently available angular acceleration estimates. Before the fault detection signal is valid, the aggregate compensation state remains in its initial state; after the detection signal is valid, it is updated according to equations (14) to (16). The compensation state in this embodiment includes three components: roll, pitch, and yaw, which are used to form a three-axis convergent compensation effect. It does not represent the effectiveness coefficient of any individual control surface, nor is it required to be strictly equal to the offline calculated actual fault equivalent quantity.

[0059] Projected boundaries are used to limit The absolute value of the update rate limit is used to prevent measurement noise or instantaneous prediction residuals from causing abrupt changes in the aggregate compensation state. If the continuous control deviation changes slowly in practical applications, a smaller update rate can be used; if a faster aggregate compensation establishment speed is required, the adaptive gain and update limit can be appropriately increased, but the selection should be based on the control surface capabilities.

[0060] Fractional power fast feedback and aggregate adaptive compensation form a combined virtual angular acceleration according to equation (17). In this expression, the fractional power term is directly generated by attitude and angular velocity errors, while the aggregate compensation term is generated by prediction residuals; neither is separately assigned to the control surfaces. The controller obtains the required torque using equations (18) and (19), and then uses the nominal control performance matrix. Calculate candidate commands for control surfaces.

[0061] All six control surface candidate commands are uniformly constrained by position and rate. For the ELE and PF control surfaces, the position constraint is [-30°, 30°]; for the AMT control surface, the position constraint is [0°, 60°]; the rate limit for all control surfaces is 60° / s. This embodiment does not delete columns in the effectiveness matrix based on the fault location, nor does it utilize actual fault effectiveness correction. .

[0062] In this embodiment, the attitude outer loop uses scalar gain. Generate angular velocity reference values ​​for each channel; This represents the feedback gain from the attitude error to the angular velocity reference. To ensure that the scalar parameters in the table are consistent with the aforementioned matrix form, we take: (twenty two) In the formula, It is a third-order identity matrix; For angular velocity linear feedback gain scalar, The scalar of the feedback gain is the fractional power of the angular velocity error. The attitude error fractional-power feedback gain scalar. To adaptively update the gain scalar, To predict the residual feedback gain scalar, For aggregation compensation gain scalar, Let the leakage gain be a scalar. This is a scalar representing the upper limit of the update rate.

[0063] (twenty three) In the formula, For the predefined bounded set of aggregate compensation states, For set Any three-dimensional vector in, For vectors The One portion, This serves as the unified boundary for the aggregation compensation state of each axis.

[0064] The main control parameters used in this embodiment are shown in Table 2. The parameters in the table are only one possible option and do not constitute a limitation on the parameter range of the present invention.

[0065] Table 2 Main control parameters of the method of the present invention

[0066] The following is This represents the continuous time from the start of the simulation. To verify the method of this invention, in... At 30s, the right elevator (ELE_R) and right pitch flap (PF_R) simultaneously experience multiplicative performance loss. The performance coefficients of both faulty control surfaces decrease from 1.0 to 0.5, meaning the remaining performance is 50%. A smooth transition of 0.5s is used for the fault to avoid unnecessary ideal step shocks in the numerical simulation. At 31 seconds, a fault detection signal is given and aggregated adaptive compensation state update is initiated. The controller only receives the detection event, not the faulty control surface name or the actual performance coefficient. The simulation continues until 60 seconds.

[0067] To verify the roles of different components, four control methods were compared: the first was nonlinear dynamic inverse control (NDI), serving as a reference for conventional engineering control; the second was a fractional power feedback incremental backstepping method, which included fractional power fast feedback but not aggregate adaptive compensation; the third was an aggregate adaptive compensation incremental backstepping method, which included residual-driven aggregate compensation but not attitude and angular velocity fractional power feedback; and the fourth was the method of this invention. All four methods used the same attitude command, fault time, fault severity, sampling period, and measurement noise sequence.

[0068] Figure 2 The roll, pitch, and sideslip responses for four control methods within the range of 25 to 60 seconds are presented, with the fault time at 30 seconds marked in the figure. Figure 2 It is evident that after the control surface effectiveness decreases, the roll and pitch tracking errors of the nonlinear dynamic inverse method and the fractional power feedback incremental backstepping method are relatively large; the aggregate adaptive compensation incremental backstepping method reduces the impact of continuous faults; the method of this invention can quickly reduce roll and pitch tracking errors after fault detection and enabling aggregate adaptive updates, while maintaining a small sideslip angle.

[0069] Figure 3 The attitude error of the fractional-power feedback incremental backstepping method and the method of this invention is further presented in the 30-40s range. Both methods employ the same fractional-power feedback structure; the main difference lies in the addition of triaxial convergence compensation driven by the prediction residual in this invention. Figure 3 As can be seen, after fault detection in 31 seconds and enabling aggregated adaptive update, the method of the present invention has small recovery errors in the roll, pitch and sideslip channels, indicating that the coordination of fractional power fast feedback and aggregated adaptive compensation in the unified virtual control layer can improve the recovery process after detection.

[0070] Figure 4 A comparison between the offline calculated triaxial fault equivalents and the aggregated compensation state is presented. To facilitate offline verification, the missing fault torque stored in the simulation data is... Convert to angular acceleration equivalent quantity and compare with Compare within the same units. Fault equivalent quantities are used only for plotting and results analysis, and are not involved in controller calculations. Figure 4 It is evident that the aggregated compensation state is gradually established after fault detection, and it is consistent with the offline fault equivalent in terms of main change direction, trend and magnitude, indicating that the prediction residual can provide effective information for the formation of the aggregated compensation effect; however, this invention does not claim that the two values ​​are equal or that the aggregated compensation state accurately converges to the true fault equivalent.

[0071] Figure 5 The root mean square error (RMSE) was used to quantitatively compare different methods. The fault recovery phase was measured from 31 to 35 seconds, and the sustained fault phase from 31 to 60 seconds. The RMSE of each method is shown in Tables 3 and 4. The RMSEs in Tables 3 and 4 are rounded to four decimal places. The percentage reduction given later is calculated based on full-precision data and may differ slightly from the results calculated directly using the four decimal places in the tables.

[0072] Table 3: Root mean square error of attitude during the fault recovery phase from 31 to 35 seconds

[0073] Table 4: Root mean square error of attitude during the 31-60s continuous fault phase

[0074] According to Table 3, the method of this invention reduces the root mean square errors of roll angle, pitch angle, and sideslip angle by approximately 52.08%, 48.18%, and 31.67%, respectively, within 31 to 35 seconds compared to the fractional power feedback incremental backstepping method; and by approximately 38.56%, 33.66%, and 43.68%, respectively, compared to the aggregation adaptive compensation incremental backstepping method. Within this window, the root mean square errors of roll and pitch for the fractional power feedback incremental backstepping method are 0.6203° and 0.5291°, respectively, which are not superior to NDI's 0.5320° and 0.4818°, indicating that fractional power feedback alone does not necessarily lead to an improvement in the performance of all channels. Among them, the sideslip index of the fractional power feedback incremental backstepping method is better than that of the aggregate adaptive compensation incremental backstepping method, while the latter has better roll and pitch indices, indicating that the two single component methods do not completely dominate each other. The method of this invention achieves the lowest root mean square error in all three channels, further indicating that the performance improvement comes from the synergy of the two types of control actions in a unified closed loop, rather than the effect necessarily obtained by using any component module alone.

[0075] According to Table 4, the method of the present invention reduces the root mean square errors of roll angle, pitch angle, and sideslip angle by approximately 70.74%, 68.62%, and 20.71% respectively within 31 to 60 seconds compared to the fractional power feedback incremental backstepping method; and by approximately 48.12%, 46.98%, and 8.36% respectively compared to the aggregated adaptive compensation incremental backstepping method. Combining Tables 3 and 4, it can be seen that the method of the present invention achieves the lowest root mean square error in all six indicators across the two evaluation stages and three attitude channels, indicating that it can simultaneously improve the recovery process after fault detection and the attitude tracking performance under conditions of continuous performance loss.

[0076] As can be seen from Tables 3 and 4, the technical effect demonstrated in this embodiment does not stem from the individual effect of either fractional power feedback or aggregate adaptive compensation. The two individual methods have their own advantages and disadvantages in different channels and stages, and are not entirely mutually dominant. In fact, the fractional power feedback incremental backstepping method, when used alone, is not even superior to the conventional nonlinear dynamic inverse method in the roll and pitch channels. Only when the two are coordinated at the same virtual angular acceleration layer do they simultaneously achieve optimal results across all six indicators. This aligns with the technical effect described in the invention's description, namely, the coordinated effect of rapid error correction and continuous deviation compensation is the reason for the aforementioned performance of this invention.

[0077] In this embodiment, the controller's aggregate compensation state, control surface deflection angle, and control surface rate all remain bounded, with no numerical divergence, non-numerical quantities, or infinite values. Although the fault causes some control surface rate constraints to frequently participate in the control, the control commands remain within the preset control surface capability range through unified control surface position and rate limits.

[0078] In addition to the above-described embodiment with a 50% performance loss due to dual control surfaces, this invention can also be applied to single control surface performance loss, combined performance loss of multiple different control surfaces, and performance degradation that changes slowly over time. For different tailless flying wing aircraft, the fractional power feedback gain, adaptive gain, projection boundary, update limit, and incremental correction coefficients can be adjusted according to inertia, control performance, allowable attitude error, and control surface capabilities without changing the basic control flow of this invention.

[0079] The control allocation in this invention is not limited to the Moore–Penrose generalized inverse form. While maintaining the core relationship of "coordination between fractional power fast feedback and aggregated adaptive compensation at the virtual angular acceleration layer," weighted generalized inverse, constrained least squares, or other allocation methods utilizing the nominal control performance matrix can be employed. Angular acceleration estimation can also be obtained from other causal estimation methods provided by a filter differential, state observer, or inertial measurement unit.

[0080] The above embodiments are merely preferred embodiments of the present invention. Any equivalent substitutions made within the framework of the present invention for the fractional power feedback function form, predictive residual construction, adaptive projection method, incremental torque interface, control allocation form, or rudder surface constraint method shall fall within the protection scope of the present invention.

Claims

1. A fault-tolerant control method for a tailless flying wing aircraft that combines fractional power feedback and aggregate compensation, characterized in that, The specific steps are as follows: Step 1: Obtain flight status and construct control variables; The attitude, angular velocity, and control surface position data of the tailless flying wing aircraft are read. First, the attitude command is converted into an angular velocity reference value by the attitude outer loop, and then the attitude error and angular velocity error are calculated separately. At the same time, the angular velocity data is processed by causal filtering and differential to obtain the angular acceleration estimate. Step 2: Establish an aggregated equivalent representation of the control surface effectiveness loss; First, the rotational dynamics of the tailless flying wing aircraft around its center of mass are given. Then, the control surface efficiency loss is expressed as a multiplicative coefficient. Then, under the same command and the same flight conditions, the nominal aerodynamic torque corresponding to the healthy state is subtracted from the actual aerodynamic torque after the failure to obtain the aerodynamic torque loss. Finally, the inverse of the rotational inertia matrix is ​​multiplied on the left and the negative sign is taken, and it is converted to the angular acceleration channel to obtain an equivalent convergent control loss in three axes. Step 3: Construct a fractional power nonlinear fast feedback; First, a fractional power nonlinear function is defined. Then, this function is applied to the angular velocity error and the attitude error respectively. Together with the derivatives of the linear angular velocity feedback and the angular velocity reference, it forms a fractional power fast feedback part, resulting in a three-axis virtual angular acceleration metric. Step 4: Construct the triaxial convergent prediction residuals; Subtract the angular acceleration required by the controller at the previous sampling moment from the angular acceleration actually performed by the tailless flying wing aircraft at that moment to obtain the three-axis aggregated prediction residual. Step 5: Construct residual-driven aggregation adaptive compensation with projection and updated limiting; First, an unconstrained adaptive update quantity is constructed from the prediction residual. Then, a rate limit is applied to the update quantity, and the update result is kept within a preset bounded set by the projection operator. The three-axis aggregated compensation state is obtained by updating step by step. Step 6: Coordinate rapid feedback of fractional powers and aggregate adaptive compensation; The fractional power fast feedback part and the three-axis aggregate compensation state are directly added at the virtual angular acceleration layer to obtain the comprehensive virtual angular acceleration control quantity; Step 7: Generate torque commands based on incremental dynamics; First, the current measured torque is estimated and constructed from the current angular velocity and angular acceleration. Then, based on this measured torque, a correction amount proportional to the increment of angular acceleration is added to obtain the incremental torque command. Step 8: Use the nominal control effectiveness matrix to complete the control surface allocation and apply constraints; Using the generalized inverse of the nominal control effectiveness matrix under healthy conditions, the torque command is assigned as a candidate command for each control surface, and then position and rate constraints are applied to the candidate commands.

2. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 1 is as follows: Obtain the attitude, angular velocity, control surface position, and flight status information of the tailless flying wing aircraft; define the attitude output vector and the aircraft's angular velocity vector as shown in formula (1): (1) In the formula, This is the attitude output vector. The angular velocity vector of the machine system. For roll angle, The pitch angle, Sideslip angle, , , These are the aircraft's roll, pitch, and yaw angular velocities, respectively. The attitude command vector is shown in formula (2): (2) In the formula, This is the attitude command vector. , and These are roll angle command, pitch angle command, and sideslip angle command; The attitude error is constructed based on the attitude command and the current attitude output vector, and is used as the degree to which the current attitude deviates from the command. The attitude error is shown in formula (3): (3) In the formula, This is the attitude error vector; Angular velocity error is the degree to which the current angular velocity deviates from the reference value. The definition of angular velocity error is shown in formula (4): (4) In the formula, This is the angular velocity error vector. The angular velocity reference value is generated from the attitude outer ring; Angular acceleration estimates are obtained from angular velocity measurements through causal filtering and differential processing. .

3. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 2 is as follows: The rotational dynamics of a tailless flying wing aircraft about its center of mass is represented by equation (5): (5) In the formula, Here is the rotational inertia matrix. The system's angular acceleration vector. The actual aerodynamic torque; assuming the tailless flying wing aircraft has The command vector of the control surfaces of a tailless flying wing aircraft is shown in formula (6): (6) In the formula, For the control surface command vector, The total number of rudder surfaces, For vectors The The component, i.e., the first component The command deflection of each control surface ; The loss of control surface effectiveness is expressed by a multiplicative coefficient as shown in formula (7): (7) In the formula, The equivalent control surface deflection that actually produces aerodynamic effects. For the control surface effectiveness matrix, for Diagonal matrix of order 1 This is an operator that constructs a diagonal matrix using the elements within the parentheses as diagonal elements. For matrix The The diagonal element, i.e. the first... The efficiency coefficient of each control surface , Indicates the first Each rudder surface is healthy. This indicates that the control surface has suffered some performance loss; Under the same control surface commands and flight conditions, let the nominal aerodynamic torque corresponding to the healthy state be... The loss of aerodynamic torque caused by the fault is shown in formula (8): (8) In the formula, The loss of aerodynamic torque is caused by a malfunction; The aerodynamic torque loss caused by the fault is converted into an equivalent convergent control loss in the angular acceleration channel, as shown in Equation (9): (9) In the formula, The equivalent convergence control in the angular acceleration channel is missing; Thus, the angular acceleration after the fault is expressed as shown in formula (10): (10) In the formula, This refers to the angular acceleration obtained under the same control command based on the nominal control surface effectiveness.

4. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 3 is as follows: The fractional power nonlinear function is defined as shown in formula (11): (11) In the formula, It is a fractional power nonlinear function. For any scalar error component, For symbolic functions, It is a fractional power exponent. The fractional-power fast feedback part, constructed based on attitude error and angular velocity error, is shown in formula (12): (12) In the formula, For rapid feedback of virtual angular acceleration components in fractional powers, The bounded derivative or difference estimate of the angular velocity reference quantity. The linear feedback gain matrix for angular velocity. The fractional power feedback gain matrix for angular velocity error. Let be the attitude error fractional power feedback gain matrix, and all three are positive definite diagonal matrices.

5. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 4 is as follows: remember The discrete sampling time number. For the first The integrated virtual angular acceleration control quantity at each sampling time is formed by the coordination of fractional power fast feedback and aggregate adaptive compensation; the three-axis aggregate prediction residual is constructed using the integrated virtual angular acceleration at the previous time and the currently obtained angular acceleration estimate as shown in formula (13): (13) In the formula, For the first The three-axis aggregated prediction residuals at each sampling time point This is an estimate of the angular acceleration at the previous sampling time. This is the comprehensive virtual angular acceleration control quantity formed at the previous sampling time.

6. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 5 is as follows: set up This represents the triaxial aggregation compensation state; an unconstrained adaptive update value is constructed based on the predicted residuals. As shown in formula (14): (14) In the formula, To adaptively update the gain matrix, To predict the residual feedback gain matrix, For the first The three-axis aggregation compensation state at each sampling time. The leakage gain matrix; A rate limit is imposed on the update quantity as shown in Equation (15): (15) In the formula, Let be the upper bound vector of the positive update rate for the three angular acceleration channels. This is the angular acceleration channel number. For vectors The The component, i.e., the first component The upper limit of the update rate for each channel. , ; To restrict input by component Saturation operators within, This is the adaptive update amount after amplitude limiting; The aggregation compensation state is kept within a pre-defined bounded set by using the projection operator. Inside: (16) In the formula, For the first The three-axis aggregation compensation state at each sampling time. For a pre-defined bounded closed set, The sampling period is Represents the predefined bounded set The projection; Upon receiving a fault detection signal, the aggregated adaptive update shown in formula (16) is initiated; when the aggregated adaptive update is not initiated, the process is maintained. This is the preset initial state.

7. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 6 is as follows: The fractional power fast feedback part and the aggregate compensation state are coordinated in the virtual angular acceleration layer to obtain the comprehensive virtual angular acceleration control quantity as shown in formula (17): (17) In the formula, For the first Bounded derivative or difference estimate of the angular velocity reference at each sampling time. For the first angular velocity error at each sampling time, For the first The attitude error at each sampling time. The aggregated compensation gain matrix; in equation (17) That is, the fractional power fast feedback part defined in equation (12) In the The values ​​taken at each sampling time; Since it is a aggregation compensation term, equation (17) is denoted as , For the rapid feedback section of fractional powers In the The value taken at each sampling time.

8. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 7 is as follows: The current measured torque is estimated and constructed based on the current angular velocity and causal angular acceleration as shown in formula (18): (18) In the formula, For the first The measured torque at each sampling time, For the first The angular velocity of the machine system at each sampling time. This is an estimate of the angular acceleration at the previous sampling time. The incremental torque command is shown in formula (19): (19) In the formula, For the first Incremental torque command at each sampling time. This is the incremental correction coefficient.

9. The fault-tolerant control method for a tailless flying wing aircraft with fractional power feedback and aggregate compensation synergy as described in claim 1, characterized in that, Step 8 is as follows: set up To balance the rudder surface vector, To determine the reference aerodynamic moment for trimming the control surfaces under the current flight conditions, Given the nominal local control performance matrix, the unconstrained control surface commands are as shown in formula (20): (20) In the formula, For the first Unconstrained control surface candidate commands at each sampling time. Represents the Moore–Penrose generalized inverse; remember For the first The control surface command is finally sent to the control surface after applying position and rate constraints at each sampling time. and These are the lower limit vector and upper limit vector of each control surface position, respectively. This is a vector composed of the maximum permissible positive velocities of each control surface. The control surface command issued at the previous sampling time is used; control surface position and rate constraints are applied to the unconstrained control surface candidate command obtained by equation (20) as shown in equation (21): (21) The constrained control surface command is sent to the control surface, and steps 1 to 8 are repeated at the next sampling time, thus forming a closed-loop fault-tolerant control.

Images