Large unmanned aerial vehicle full-state flight safety control method and system

Abstract

The application discloses a large unmanned aerial vehicle full-state flight safety control method and system, relates to the technical field of unmanned aerial vehicle flight safety control, and collects body vibration frequency spectrum, structural strain and attitude angular velocity data; a basic safety level is generated according to the attitude angular velocity, the vibration frequency spectrum is input into a pre-trained prediction model to output an instability risk index, the instability risk index is used for correcting and compensating the basic safety level, and a comprehensive safety level is obtained; real-time aeroelastic modal parameters are identified on the basis of the structural strain and the vibration frequency spectrum; the comprehensive safety level and the real-time aeroelastic modal parameters are used as dynamic constraint boundaries, real-time safety energy budget is combined, reversible active actuator matrix is subjected to rolling time domain optimization to generate an optimal control sequence; and the optimal control sequence is executed to adjust the effective resistance area until the attitude is stabilized, so that the flight safety of the large unmanned aerial vehicle in all states is ensured.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight safety control technology, and in particular to a method and system for all-state flight safety control of large UAVs. Background Technology

[0002] Large drones are increasingly used in logistics, surveying, agricultural protection, and security monitoring. However, with the increase in size and weight, flight safety issues are becoming increasingly prominent. The crash rate of large drones is significantly higher than that of small consumer drones, mainly due to the following two reasons: Category 1: Normal Vibration and Structural Fatigue Issues. Large UAVs have high structural flexibility and aspect ratio, making them susceptible to multiple excitation sources during flight, including periodic excitation forces generated by engine / motor rotational imbalance, aeroelastic flutter of the wing / fuselage in high-speed airflow, and dynamic responses of external loads. Long-term accumulation of these excitations can lead to structural fatigue crack propagation, decreased precision load imaging quality, and increased measurement noise from flight control sensors. Existing normal vibration suppression solutions, such as passive vibration isolation, are ineffective against low-frequency (<30Hz) resonance; increasing structural stiffness significantly increases weight; and active vibration control has a short stroke and high power consumption, all of which struggle to balance vibration reduction effectiveness with engineering feasibility.

[0003] The second category: Extreme loss of control and emergency landing issues. When encountering severe accidents such as control surface jamming, complete sensor failure, extreme wind shear, or mid-air collision, the drone may enter fatal modes such as rapid roll, dive, or spiral. Existing whole-aircraft parachutes are bulky and heavy and ineffective at low altitudes, self-destruct systems are unacceptable for civilian use, and power redistribution is ineffective in cases of complete loss of control. There is a lack of reliable low-altitude emergency recovery methods. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a method and system for full-state flight safety control of large unmanned aerial vehicles, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of this invention is implemented as follows: In a first aspect, the present invention provides a method for all-state flight safety control of a large unmanned aerial vehicle (UAV), the method comprising the following steps: S1: Collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage; S2: Generate a basic safety level based on the attitude angular velocity data, and input the vibration spectrum data into a pre-trained prediction model to output an instability risk index. Use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level. S3: Based on the structural strain data and the vibration spectrum data, the recursive subspace identification algorithm is used to identify the real-time aeroelastic modal parameters online; S4: Using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and combining them with a preset real-time safety energy budget, the reversible active actuator matrix is ​​subjected to rolling time-domain optimization to generate the optimal control sequence; S5: Execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure, and when the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​judged to be saturated, trigger the deployment of the variable area aerodynamic drag actuator, and adjust its effective drag area with the current attitude angular velocity as the closed-loop feedback quantity until attitude stability is restored.

[0006] As a preferred embodiment of the present invention, step S1 specifically includes: Vibration spectrum data of multiple key nodes of the fuselage are collected through a distributed piezoelectric vibration sensor array; Strain data of the fuselage structure were acquired using a fiber Bragg grating strain sensor array. Attitude angular velocity data is collected using an attitude sensor.

[0007] As a preferred embodiment of the present invention, the specific process of generating the basic safety level based on the attitude angular velocity data in step S2 includes: Based on multiple preset safety states, the current attitude angular velocity data is matched with the angular velocity threshold ranges corresponding to the multiple safety states to generate the basic safety level.

[0008] As a preferred embodiment of the present invention, the specific process of inputting the vibration spectrum data into the pre-trained prediction model to output the instability risk index in step S2 includes: The prediction model is a spatiotemporal graph attention network model. The graph nodes of the spatiotemporal graph attention network model correspond to sensors distributed on multiple key nodes of the fuselage. The edge weights between nodes are dynamically updated according to the fuselage structure transfer function. Using the vibration spectrum data within the first time window in the past as the input to the spatiotemporal graph attention network model, the model outputs the predicted peak angular velocity and flutter divergence probability values ​​for multiple future time windows. The instability risk index is obtained by weighting and fusing the predicted peak angular velocity value and the flutter divergence probability value.

[0009] As a preferred embodiment of the present invention, the specific process of correcting and compensating the basic safety level using the instability risk index to obtain the comprehensive safety level in step S2 includes: Obtain the first risk indicator threshold corresponding to the warning level and the second risk indicator threshold corresponding to the danger level among the multiple safety states; Obtain instability risk indicators at multiple sampling moments within a historical time window, and calculate risk trend parameters based on the changing trends of the instability risk indicators at the current moment and the instability risk indicators within the historical time window; The current instability risk index is compared with the first risk index threshold and the second risk index threshold to determine the level correction range, thus obtaining the comprehensive safety level, including: When the instability risk index does not exceed the first risk index threshold but the risk rise rate exceeds the preset rate threshold, the basic safety level is adjusted up by one level, and the adjusted level is used as the comprehensive safety level. When the instability risk index exceeds the first risk index threshold but does not exceed the second risk index threshold, the basic security level is revised upward to the pre-warning level, and the pre-warning level is used as the comprehensive security level. When the instability risk index exceeds the second risk index threshold, the basic safety level is revised upward to the danger level, and the danger level is used as the comprehensive safety level.

[0010] As a preferred embodiment of the present invention, the specific process of identifying real-time aeroelastic modal parameters online using the recursive subspace identification algorithm in step S3 includes: Using the vibration spectrum data as input data and the structural strain data as output data, past input matrix, past output matrix, future input matrix and future output matrix are constructed respectively, and then merged row by row to obtain the expanded dimension data matrix at each time point; Based on the extended dimension data matrix at each time point, calculate the orthogonal projection from the future output row space to the past input-output joint row space to obtain the projection matrix, and perform singular value decomposition on the projection matrix to obtain the generalized observable matrix. Eigenvalue decomposition is performed on the generalized observable matrix to obtain eigenvalues. Based on the eigenvalues, the natural frequency values ​​and modal damping ratio values ​​corresponding to the bending and torsional modes in the UAV fuselage structure are calculated. The natural frequency values ​​and modal damping ratio values ​​are used as the real-time aeroelastic modal parameters.

[0011] As a preferred embodiment of the present invention, the specific process of performing rolling time-domain optimization on the reversible active actuator matrix in step S4 includes: Obtain the current state parameters of each actuator in the reversible active actuator matrix, and construct an aeroelastic dynamics model using the current state parameters and the real-time aeroelastic modal parameters; The aeroelastic dynamics model uses the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and the real-time safety energy budget as energy constraint. Using the current state parameters of each actuator as the initial state, and the safety constraint boundary and the energy constraint boundary as constraints, a cost function with vibration suppression and energy consumption as optimization objectives is established. Within each control cycle, the control instruction sequence of each actuator that satisfies the constraints and minimizes the cost function is determined as the optimal control sequence.

[0012] As a preferred embodiment of the present invention, the specific implementation process of executing the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure in step S5, and triggering the deployment of the variable area aerodynamic drag actuator when the comprehensive safety level reaches the emergency level or when the reversible active actuator matrix is ​​determined to be saturated, is as follows: Obtain the control command values ​​of each actuator in the current control cycle of the optimal control sequence, and drive each actuator in the reversible active actuator matrix to act according to the control command values ​​of each actuator, so as to adjust the stiffness and damping characteristics of the fuselage structure. During the execution of the optimal control sequence, the comprehensive safety level is acquired in real time, and the working status of each actuator in the reversible active actuator matrix is ​​monitored. When the overall safety level reaches the emergency level, or when the overall safety level does not reach the emergency level but the adjustment capability of the reversible active actuator matrix is ​​determined to be saturated based on the working status of each actuator, a trigger signal is generated and sent to the variable area aerodynamic drag actuator, causing the variable area aerodynamic drag actuator to unfold from the retracted state into the flow field.

[0013] As a preferred embodiment of the present invention, the specific process of adjusting the effective drag area using the current attitude angular velocity as the closed-loop feedback quantity in step S5 until attitude stability is restored includes: After the variable area pneumatic drag actuator is deployed, the current attitude angular velocity is collected in real time, the preset target attitude angular velocity is obtained, and the angular velocity deviation between the current attitude angular velocity and the target attitude angular velocity is calculated. Based on the angular velocity deviation, an area adjustment command is generated through a closed-loop control algorithm. The variable area pneumatic resistance actuator is driven to operate according to the area adjustment command to adjust its effective resistance area. During the adjustment process, when the absolute value of the angular velocity deviation converges to a preset dead zone range and remains within the dead zone range for a preset period of time, the attitude is determined to have returned to stability.

[0014] Secondly, the present invention provides a large-scale unmanned aerial vehicle (UAV) full-state flight safety control system, applied to any of the methods described above, the system comprising: The data acquisition module is used to collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage. The safety prediction module is used to generate a basic safety level based on the attitude angular velocity data, input the vibration spectrum data into a pre-trained prediction model to output an instability risk index, and use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level. The parameter identification module is used to identify real-time aeroelastic modal parameters online based on the structural strain data and the vibration spectrum data using a recursive subspace identification algorithm. The optimal control module is used to perform rolling time-domain optimization on the reversible active actuator matrix, using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and combining them with a preset real-time safety energy budget, to generate the optimal control sequence; The optimal adjustment module is used to execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure. When the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​determined to be saturated, the variable area aerodynamic drag actuator is triggered to deploy and its effective drag area is adjusted with the current attitude angular velocity as the closed-loop feedback until attitude stability is restored.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires vibration spectrum, structural strain, and attitude angular velocity data through unified sensing by multiple sensors; utilizes a pre-trained spatiotemporal graph attention network to predict future instability risks and correct and compensate for safety levels; employs a recursive subspace identification algorithm to identify real-time aeroelastic modal parameters online, and combines real-time safety energy budgets to generate the optimal control sequence through rolling time-domain optimization; in emergency situations, a variable area aerodynamic drag actuator is triggered, and the effective drag area is adjusted in a closed loop using attitude angular velocity as feedback until stability is restored, thereby achieving full-spectrum safety coverage from normal vibration to extreme runaway, ensuring the flight safety of large UAVs in all states. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a method for all-state flight safety control of a large unmanned aerial vehicle (UAV) provided in an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of a large unmanned aerial vehicle (UAV) full-state flight safety control system provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" refers to a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0019] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0020] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0021] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0022] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0023] Firstly, this invention provides a method for all-state flight safety control of a large unmanned aerial vehicle (UAV). Please refer to the appendix. Figure 1 The method includes the following steps: S1: Collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage.

[0024] In this embodiment, a distributed piezoelectric vibration sensor array (such as a PZT piezoelectric ceramic sensor) is arranged at key structural locations such as the wing leading edge, wing root, fuselage main spars, and tail. The sampling frequency is set to 1~10kHz to cover the structural modal frequency range of 0.5~500Hz. The collected vibration signals are then subjected to Fourier transform to generate vibration spectrum data.

[0025] Fiber Bragg grating (FBG) strain sensor arrays are arranged along the wing span and fuselage longitudinal direction, with a sampling frequency of 1kHz, to measure the dynamic strain data of the structure.

[0026] A three-axis MEMS gyroscope with a sampling frequency of 200Hz is placed near the center of gravity of the fuselage to measure attitude angular velocity data.

[0027] S2: Generate a basic safety level based on the attitude angular velocity data, and input the vibration spectrum data into a pre-trained prediction model to output an instability risk index. Use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level.

[0028] In this embodiment, based on multiple preset safety states, the current attitude angular velocity data is matched with the angular velocity threshold range corresponding to the multiple safety states to generate the basic safety level.

[0029] Specifically, the pre-defined multi-level safety state model includes at least six progressively increasing safety states, from lowest to highest: Normal State (L1), with a three-axis angular velocity threshold range of <30° for all axes; Caution State (L2), with a three-axis angular velocity threshold range of 30°~50° for any axis; Pre-warning State (L3), with a three-axis angular velocity threshold range of 50°~80° for any axis; Warning State (L4), with a three-axis angular velocity threshold range of 80°~120° for any axis; and Danger State (L5), with a three-axis angular velocity threshold range of >200° for any axis. The angular velocity threshold ranges for each level are pre-calibrated through flight testing.

[0030] The maximum value of the current attitude angular velocity is matched with the three-axis angular velocity threshold range corresponding to each level. If it falls within a certain level range, the corresponding basic safety level is generated.

[0031] Furthermore, the vibration spectrum data is input into a pre-trained prediction model to output an instability risk index. The specific process includes: The prediction model is a spatiotemporal graph attention network model. The graph nodes of this model correspond to piezoelectric vibration sensors distributed across 12 key nodes of the fuselage. The edge weights between nodes are dynamically updated based on the fuselage structure transfer function. Specifically, the model uses vibration spectrum data from the past 5-second time window (each node contains energy values ​​from 32 frequency bands within the 0.5~500Hz frequency range) as input, and outputs predicted peak angular velocity values ​​for the next 10, 20, and 30-second time windows. The predicted peak angular velocity values ​​and flutter divergence probability values ​​are then weighted and fused using the following formula to generate an instability risk index:

[0032] in, Representing the current time Instability risk indicators The weighting coefficients represent the predicted angular velocity values. This is used to balance the contribution of angular velocity prediction and flutter probability to the risk index; in this embodiment, it is set to 0.5. Represents the future time window of the predictive model output. Predicted peak angular velocity within the range, The lower limit of the angular velocity threshold representing the warning level (L4) is set to 80° / s in this embodiment. This represents the flutter divergence probability value at the current moment, as output by the prediction model.

[0033] Furthermore, relying solely on the current instability risk indicator for correction may lead to false triggers due to instantaneous fluctuations. Therefore, this invention introduces the changing trend of the risk indicator within a historical time window to distinguish between a continuously deteriorating risk trend and occasional instantaneous disturbances.

[0034] To obtain historical time windows, this embodiment takes a sequence of instability risk indicators from multiple sampling intervals within a 30-second period. A linear fit is then performed on this sequence, and the fit slope is used as a risk trend parameter to characterize the rate of change of the risk indicators over time. The rate of change of the risk trend parameter is defined as:

[0035] in, The rate of change of the risk trend parameter at the current moment. This is the instability risk indicator (dimensionless) at the previous sampling time. The sampling interval. Preset rate of change threshold. The risk trend parameter change rate is taken as 1.5 to 2.0 times the average rate of change under historical normal flight conditions; in this example, it is taken as 2.0 times. ,in, This represents the average rate of change under historical normal flight conditions.

[0036] Furthermore, the first risk indicator threshold corresponding to the warning level and the second risk indicator threshold corresponding to the danger level are obtained among the multiple safety states; Specifically, the calibration method for the first risk indicator threshold (corresponding to the pre-warning level) and the second risk indicator threshold (corresponding to the danger level) is as follows: collect samples of instability risk indicators of the UAV under known normal flight conditions, and calculate their mean values. and standard deviation ,set up , .in, This is the arithmetic mean of the instability risk index under normal conditions. This represents the sample standard deviation of the instability risk index under normal conditions.

[0037] The current instability risk index is compared with the first risk index threshold and the second risk index threshold to determine the level correction range, thus obtaining the comprehensive safety level. The logic for determining the level correction range is as follows: (1) When and When the level correction range is zero, the overall security level is the same as the basic security level. (2) When but When necessary, the basic security level will be revised upward by one level, and the revised level will be used as the overall security level. (3) When When this happens, the basic security level will be revised upward to the pre-warning level (L3), and the pre-warning level will be used as the overall security level. (4) When When this happens, the basic safety level will be revised upwards to the danger level (L5), and the danger level will be used as the overall safety level.

[0038] S3: Based on the structural strain data and the vibration spectrum data, the real-time aeroelastic modal parameters are identified online using a recursive subspace identification algorithm.

[0039] In this embodiment, the mass distribution and structural dynamic characteristics of the UAV continuously change during flight due to fuel consumption, external payload dropping, or icing. This invention employs the Recursive Subspace Identification (N4SID) algorithm to identify real-time aeroelastic modal parameters online, enabling continuous tracking of the structural dynamic characteristics.

[0040] Using specific frequency band energy from vibration spectrum data as input data Structural strain data is used as output data. ,in The sampling time number. For the number of input variables, The number of output variables (in this embodiment) The value is 12, corresponding to the number of pressure strain sensors. Given the data length, construct a matrix of past input data blocks. Past output data block matrix Future input data block matrix and future output data block matrix Based on the past input data block matrix For example:

[0041] in, This represents the number of rows in the past; in this example, it is taken as 10. Number of columns. Past output data block matrix. Future input data block matrix and future output data block matrix Constructed in the same manner, the future data block matrix is ​​shifted backward to correspond to the time period. There are 10 sampling points. The input data block matrix and the output data block matrix corresponding to the same time moment are merged row by row to obtain the extended dimension data matrix for each time moment.

[0042] Furthermore, the orthogonal projection of the future output row space onto the past input-output joint row space is calculated to obtain the projection matrix. Orthogonal projection refers to the process of removing components related to past input and output data from future output data and then applying this to the projection matrix. After weighting, singular value decomposition is performed to obtain the singular value distribution:

[0043] in, It is a left singular vector matrix. It is a right singular vector matrix. Let be a diagonal matrix, where, These are the singular values ​​arranged in descending order.

[0044] Determining the system order based on singular value distribution When a clear inflection point appears in the singular value sequence (i.e., the subsequent singular values ​​decrease sharply and tend to level off), the number of singular values ​​before the inflection point is taken as the system order. After determining the order, extract the preceding... The left singular vectors corresponding to the singular values ​​constitute the generalized observable matrix. ,in, Left singular vector matrix The former List, diagonal matrix The former Principal submatrix of order 1.

[0045] Eigenvalue decomposition is performed on the generalized observable matrix to obtain eigenvalues. Based on the eigenvalues, the natural frequency values ​​and modal damping ratio values ​​corresponding to the bending and torsional modes in the UAV fuselage structure are calculated. The natural frequency values ​​and modal damping ratio values ​​are used as the real-time aeroelastic modal parameters.

[0046] Among them, by eigenvalues The specific formula for calculating the natural frequency value is as follows:

[0047] in, For the first The undamped natural frequencies of the first mode, The sampling period is Represents the eigenvalues The modulus after taking the natural logarithm.

[0048] From eigenvalues The specific formula for calculating the modal damping ratio is as follows:

[0049] in, For the first The damping ratio of the first mode has a value range of 0 to 1. Represents the eigenvalues The real part after taking the natural logarithm.

[0050] Based on the mode shape directions corresponding to each mode, which are obtained by extracting the mode shape vectors corresponding to each column of the generalized observable matrix, a mode is determined to be a bending mode when the mode shape mainly moves along the direction perpendicular to the chord, and a torsional mode when the mode shape mainly exhibits rotational motion of the cross section about the spanwise axis. The calculated modes are then classified into bending and torsional modes. The natural frequencies are sorted from smallest to largest, and the natural frequency values ​​corresponding to the first three bending modes and the first two torsional modes are selected. And modal damping ratio , as the real-time aeroelastic modal parameter. Where the subscript... These represent the first three bending modes, respectively, with subscripts indicating the order of bending. These represent the first two torsional modes, respectively.

[0051] Using the aforementioned natural frequency and modal damping ratio as parameters, an aeroelastic dynamics model for the current moment is constructed. Specifically, the state vector is formed by the coordinates of the first three bending modes and the first two torsional modes (a total of 5 modal coordinates) and their first derivatives (a total of 10 state variables). The torques of the three reversible active actuators are used as input vectors to establish a state space model in the modal space.

[0052] S4: Using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and combined with the preset real-time safety energy budget, the reversible active actuator matrix is ​​optimized in the rolling time domain to generate the optimal control sequence.

[0053] In this embodiment, the reversible active actuator matrix consists of three types of active adjustment actuators distributed inside the fuselage structure of the UAV: ​​(1) prestressed carbon fiber tendons, embedded in the pre-embedded channel of the fuselage, with the pre-tightening force adjusted by electromagnetic fine-tuning anchors, used for low-frequency (<30Hz) natural frequency adjustment, with a response time of 100~500ms; (2) magnetorheological fluid dampers, embedded in the gradient damping layer between the fuselage skin and the frame, with the damping coefficient changed by adjusting the magnetic field strength of the electromagnetic coil array, used for mid-to-high frequency (30~200Hz) vibration energy dissipation, with a response time of <10ms; (3) shape memory alloy rods, embedded in the longitudinal central beam of the fuselage, with the contraction force generated by the heating drive circuit to actively offset vibration, used for low-frequency large-amplitude vibration active force offset (<5Hz, amplitude>3°), with a response time of 100~300ms.

[0054] Furthermore, within each control cycle, the current state parameters of each actuator in the reversible active actuator matrix are obtained. The current state parameters include the current input command value, the current output response value, and the available adjustment margin. An aeroelastic dynamics model is constructed using the current state parameters and the real-time aeroelastic modal parameters.

[0055] The aeroelastic dynamics model uses the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and the real-time safety energy budget as energy constraint.

[0056] Specifically, the safety constraint boundaries are determined as follows: based on the comprehensive safety level, a preset correspondence is queried to obtain the upper limit values ​​of the fuselage structure response constraints and the upper limit values ​​of the attitude angular velocity constraints. For example, when the comprehensive safety level is pre-warning state L3, the upper limit value of the attitude angular velocity constraint is set to 80° / s, indicating that the upper limit value of the modal coordinate amplitude constraint is set to 80% of the allowable value; when the comprehensive safety level is warning state L4, the upper limit value of the attitude angular velocity constraint is set to 120° / s, indicating that the upper limit value of the modal coordinate amplitude constraint is set to 100% of the allowable value.

[0057] Furthermore, the real-time secure energy budget The initial value of the safety energy budget (in Wh) reserved for the current moment is dynamically set according to the current flight phase: 35% of the total energy during takeoff or climb (altitude <100m), 15% during cruise, 30% during landing or approach, and 40% during severe weather or known risky airspace, as determined by the ground station. Each time the reversible active actuator is invoked during flight, the actual energy consumption is calculated based on the current integral and deducted from the real-time safety energy budget. Deduct from the middle; when the steady state lasts for more than 30 seconds, slowly recharge at a rate of 0.5W.

[0058] Then, take the current state parameters of each actuator as the initial state. With safety and energy constraints as conditions, the following cost function is established. :

[0059] in, To determine the number of time-domain steps for prediction, this embodiment uses 20. Step index in the time domain For the first The predicted state vector value for the step; The target state vector; For the first Step-by-step control increments for each actuator; , , For the first The control input vectors of the three actuators are as follows: This is the control force for prestressed carbon fiber reinforcement; The excitation current of the magnetorheological fluid damper; The heating voltage for the shape memory alloy rod; This is the state weight matrix, used to measure the importance of vibration suppression; The incremental weight matrix is ​​used to control the control quantity and suppress drastic changes. The energy penalty weight matrix, and the real-time safe energy budget It is negatively correlated, and its specific form is as follows: ,in, As the baseline weight matrix, Total battery energy (unit: Wh) It is a natural exponential function.

[0060] Within each control cycle, a sequence of control commands is determined that satisfies both the safety and energy constraints and minimizes the cost function. ,in, In this embodiment, the control period is set to 20ms. This sequence is used as the optimal control sequence, and only the control command value of the first control period is output to the reversible active actuator matrix for execution. When the next control period arrives, the optimization solution is re-solved based on the updated state to achieve rolling time-domain control.

[0061] Furthermore, when the real-time safety energy budget is lower than a preset threshold, such as 5% of the total energy, it is determined to enter the energy-saving warning mode: the high-power actuators in the reversible active actuator matrix, such as the heating drive of the shape memory alloy rod, are suspended, and the judgment threshold of the comprehensive safety level is dynamically tightened by 20%, that is, the upgrade is triggered earlier to trigger passive or low-power devices in advance.

[0062] S5: Execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure, and when the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​judged to be saturated, trigger the deployment of the variable area aerodynamic drag actuator, and adjust its effective drag area with the current attitude angular velocity as the closed-loop feedback quantity until attitude stability is restored.

[0063] In this implementation, the control command values ​​of each actuator in the current control cycle of the optimal control sequence are obtained, and the prestressed carbon fiber tendons, magnetorheological fluid dampers and shape memory alloy rods in the reversible active actuator matrix are driven to adjust the stiffness and damping characteristics of the fuselage structure.

[0064] During the execution of the optimal control sequence, the overall safety level is acquired in real time and the operating status of each actuator is monitored. When the overall safety level reaches the emergency level, or when the overall safety level does not reach the emergency level but the reversible active actuator matrix adjustment capability is determined to be saturated based on the operating status of each actuator, a trigger signal is generated and sent to the variable area aerodynamic drag actuator, causing it to unfold from its retracted state into the flow field.

[0065] The method for determining the saturation of the reversible active actuator matrix adjustment capability is as follows: Monitor whether the current output response value of each actuator has reached its upper limit of usable adjustment margin, i.e., the prestressing force of the prestressed carbon fiber reinforcement reaches more than 95% of the maximum allowable prestressing force, or the shape memory alloy rod reaches more than 90% of the maximum design shrinkage, or the magnetic field strength of the magnetorheological fluid damper reaches the saturation magnetic field strength. When any actuator reaches its saturation state and the attitude angular velocity continues to increase, it is determined that the adjustment capability is saturated.

[0066] Specifically, the variable area pneumatic resistance actuator includes: a miniature conical umbrella and a matching servo winch that pops out from the stored state, the effective windward area being changed by adjusting the length of the umbrella ropes via the servo winch; an inflatable airbag and a matching dual-way solenoid valve that inflates and unfolds by compressed gas; the dual-way solenoid valve controlling the inflation and deflation to adjust the internal pressure from 0.5 to 2.0 bar; and an electrically driven linkage mechanism that drives an openable skin speed reducer to achieve continuous adjustment of the opening angle from 0 to 60°.

[0067] After the variable area pneumatic drag actuator deploys, its effective drag area is adjusted using the current attitude angular velocity as the closed-loop feedback until attitude stability is restored. The specific process is as follows: Real-time acquisition of current attitude angular velocity Obtain the preset target attitude angular velocity In this embodiment, the target value is 0° / s, which means returning to a level flight attitude without roll / pitch / yaw. The angular velocity deviation is calculated. :

[0068] The area adjustment command is generated using a PID closed-loop control algorithm.

[0069] in, This represents the adjustment amount of the effective resistance area at the current moment; a positive value increases the area, and a negative value decreases the area. To control the gain proportionally, this embodiment uses 0.8; For integral control gain, this embodiment takes a value of 0.1; The differential control gain is set to 0.05 in this embodiment; The angular velocity deviation from the initial time 0 to the current time Time integral; This is the first derivative of the angular velocity deviation with respect to time.

[0070] Drive the servo winch or dual-way solenoid valve according to the area adjustment command: when At this time, the servo winch releases the parachute ropes to increase the windward area, or the dual-channel solenoid valve inflates the airbag to increase its volume; when At this time, the servo winch winds up the paracord to reduce the windward area, or the dual-channel solenoid valve releases air to reduce the airbag volume.

[0071] Furthermore, a dual condition is adopted for determining attitude recovery stability: when the absolute value of the angular velocity deviation... And this condition continues to be met for a preset duration. At that time, it is determined that the attitude has returned to stability. Among them, To define the preset dead zone range, this embodiment uses 2° / s. To preset the stabilization time, this embodiment uses 2 seconds. When both conditions are met simultaneously, the controller determines that the attitude has stabilized and controls the variable area pneumatic drag actuator to partially retract or maintain the current state.

[0072] At low safety levels, such as normal state (L1) or alert state (L2), the deviation of the current attitude angular velocity As a control variable, the servo winch or dual-channel solenoid valve is controlled by the above-mentioned PID closed-loop control algorithm to make the miniature conical umbrella or inflatable airbag deploy to no more than 5% of the maximum deployment area, generating an asymmetric small drag difference on both sides of the wing, which serves as an active aerodynamic damper to assist in stabilization.

[0073] Secondly, the present invention also provides a full-state flight safety control system for large unmanned aerial vehicles (UAVs), in conjunction with the appendix. Figure 2 The system includes: The data acquisition module is used to collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage. The safety prediction module is used to generate a basic safety level based on the attitude angular velocity data, input the vibration spectrum data into a pre-trained prediction model to output an instability risk index, and use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level. The parameter identification module is used to identify real-time aeroelastic modal parameters online based on the structural strain data and the vibration spectrum data using a recursive subspace identification algorithm. The optimal control module is used to perform rolling time-domain optimization on the reversible active actuator matrix, using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and in conjunction with a preset real-time safety energy budget, to generate the optimal control sequence. The optimal adjustment module is used to execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure. When the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​determined to be saturated, the variable area aerodynamic drag actuator is triggered to deploy and its effective drag area is adjusted with the current attitude angular velocity as the closed-loop feedback until attitude stability is restored.

[0074] In this embodiment, the specific functions and implementation methods of each module are completely consistent with the corresponding steps in the method described in the first aspect, and will not be repeated here.

[0075] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for all-state flight safety control of a large unmanned aerial vehicle (UAV), characterized in that, The method includes the following steps: S1: Collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage; S2: Generate a basic safety level based on the attitude angular velocity data, and input the vibration spectrum data into a pre-trained prediction model to output an instability risk index. Use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level. S3: Based on the structural strain data and the vibration spectrum data, the recursive subspace identification algorithm is used to identify the real-time aeroelastic modal parameters online; S4: Using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and combining them with a preset real-time safety energy budget, the reversible active actuator matrix is ​​subjected to rolling time-domain optimization to generate the optimal control sequence; S5: Execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure, and when the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​judged to be saturated, trigger the deployment of the variable area aerodynamic drag actuator, and adjust its effective drag area with the current attitude angular velocity as the closed-loop feedback quantity until attitude stability is restored.

2. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 1, characterized in that, Step S1 specifically includes: Vibration spectrum data of multiple key nodes of the fuselage are collected through a distributed piezoelectric vibration sensor array; Strain data of the fuselage structure were acquired using a fiber Bragg grating strain sensor array. Three-axis attitude angular velocity data are collected using an attitude sensor.

3. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 1, characterized in that, In step S2, the specific process of generating the basic safety level based on the attitude angular velocity data includes: Based on multiple preset safety states, the current attitude angular velocity data is matched with the angular velocity threshold ranges corresponding to the multiple safety states to generate the basic safety level.

4. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 3, characterized in that, In step S2, the specific process of inputting the vibration spectrum data into the pre-trained prediction model to output an instability risk index includes: The prediction model is a spatiotemporal graph attention network model. The graph nodes of the spatiotemporal graph attention network model correspond to sensors distributed on multiple key nodes of the fuselage. The edge weights between nodes are dynamically updated according to the fuselage structure transfer function. Using the vibration spectrum data within the first time window in the past as the input to the spatiotemporal graph attention network model, the model outputs the predicted peak angular velocity and flutter divergence probability values ​​for multiple future time windows. The instability risk index is obtained by weighting and fusing the predicted peak angular velocity value and the flutter divergence probability value.

5. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 4, characterized in that, In step S2, the specific process of correcting and compensating the basic safety level using the instability risk index to obtain the comprehensive safety level includes: Obtain the first risk indicator threshold corresponding to the warning level and the second risk indicator threshold corresponding to the danger level among the multiple safety states; Obtain instability risk indicators at multiple sampling moments within a historical time window, and calculate risk trend parameters based on the changing trends of the instability risk indicators at the current moment and the instability risk indicators within the historical time window; The current instability risk index is compared with the first risk index threshold and the second risk index threshold to determine the level correction range, thus obtaining the comprehensive safety level, including: When the instability risk index does not exceed the first risk index threshold but the risk rise rate exceeds the preset rate threshold, the basic safety level is adjusted up by one level, and the adjusted level is used as the comprehensive safety level. When the instability risk index exceeds the first risk index threshold but does not exceed the second risk index threshold, the basic security level is revised upward to the pre-warning level, and the pre-warning level is used as the comprehensive security level. When the instability risk index exceeds the second risk index threshold, the basic safety level is revised upward to the danger level, and the danger level is used as the comprehensive safety level.

6. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 5, characterized in that, In step S3, the specific process of using the recursive subspace identification algorithm to identify real-time aeroelastic modal parameters online includes: Using the vibration spectrum data as input data and the structural strain data as output data, past input matrix, past output matrix, future input matrix and future output matrix are constructed respectively, and then merged row by row to obtain the expanded dimension data matrix at each time point; Based on the extended dimension data matrix at each time point, calculate the orthogonal projection from the future output row space to the past input-output joint row space to obtain the projection matrix, and perform singular value decomposition on the projection matrix to obtain the generalized observable matrix. Eigenvalue decomposition is performed on the generalized observable matrix to obtain eigenvalues. Based on the eigenvalues, the natural frequency values ​​and modal damping ratio values ​​corresponding to the bending and torsional modes in the UAV fuselage structure are calculated. The natural frequency values ​​and modal damping ratio values ​​are used as the real-time aeroelastic modal parameters.

7. A method for full-state flight safety control of a large unmanned aerial vehicle (UAV) according to claim 6, characterized in that, In step S4, the specific process of performing rolling time-domain optimization on the reversible active actuator matrix includes: Obtain the current state parameters of each actuator in the reversible active actuator matrix, and construct an aeroelastic dynamics model using the current state parameters and the real-time aeroelastic modal parameters; The aeroelastic dynamics model uses the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and the real-time safety energy budget as energy constraint. Using the current state parameters of each actuator as the initial state, and the safety constraint boundary and the energy constraint boundary as constraints, a cost function with vibration suppression and energy consumption as optimization objectives is established. Within each control cycle, the control instruction sequence of each actuator that satisfies the constraints and minimizes the cost function is determined as the optimal control sequence.

8. The method for full-state flight safety control of a large unmanned aerial vehicle according to claim 7, characterized in that, In step S5, the specific implementation process of executing the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure, and triggering the deployment of the variable area aerodynamic drag actuator when the comprehensive safety level reaches the emergency level or when the reversible active actuator matrix is ​​determined to be saturated, is as follows: Obtain the control command values ​​of each actuator in the current control cycle of the optimal control sequence, and drive each actuator in the reversible active actuator matrix to act according to the control command values ​​of each actuator, so as to adjust the stiffness and damping characteristics of the fuselage structure. During the execution of the optimal control sequence, the comprehensive safety level is acquired in real time, and the working status of each actuator in the reversible active actuator matrix is ​​monitored. When the overall safety level reaches the emergency level, or when the overall safety level does not reach the emergency level but the adjustment capability of the reversible active actuator matrix is ​​determined to be saturated based on the working status of each actuator, a trigger signal is generated and sent to the variable area aerodynamic drag actuator, causing the variable area aerodynamic drag actuator to unfold from the retracted state into the flow field.

9. A method for full-state flight safety control of a large unmanned aerial vehicle (UAV) according to claim 8, characterized in that, In step S5, the specific process of adjusting the effective drag area using the current attitude angular velocity as the closed-loop feedback quantity until attitude stability is restored includes: After the variable area pneumatic drag actuator is deployed, the current attitude angular velocity is collected in real time, the preset target attitude angular velocity is obtained, and the angular velocity deviation between the current attitude angular velocity and the target attitude angular velocity is calculated. Based on the angular velocity deviation, an area adjustment command is generated through a closed-loop control algorithm. The variable area pneumatic resistance actuator is driven to operate according to the area adjustment command to adjust its effective resistance area. During the adjustment process, when the absolute value of the angular velocity deviation converges to a preset dead zone range and remains within the dead zone range for a preset period of time, the attitude is determined to have returned to stability.

10. A large unmanned aerial vehicle (UAV) full-state flight safety control system, applied to the method described in any one of claims 1 to 9, characterized in that, The system includes: The data acquisition module is used to collect vibration spectrum data, structural strain data, and attitude angular velocity data of the large UAV fuselage. The safety prediction module is used to generate a basic safety level based on the attitude angular velocity data, input the vibration spectrum data into a pre-trained prediction model to output an instability risk index, and use the instability risk index to correct and compensate the basic safety level to obtain a comprehensive safety level. The parameter identification module is used to identify real-time aeroelastic modal parameters online based on the structural strain data and the vibration spectrum data using a recursive subspace identification algorithm. The optimal control module is used to perform rolling time-domain optimization on the reversible active actuator matrix, using the comprehensive safety level and the real-time aeroelastic modal parameters as dynamic constraint boundaries, and combining them with a preset real-time safety energy budget, to generate the optimal control sequence; The optimal adjustment module is used to execute the optimal control sequence to adjust the stiffness and damping characteristics of the fuselage structure. When the comprehensive safety level reaches the emergency level or the reversible active actuator matrix is ​​determined to be saturated, the variable area aerodynamic drag actuator is triggered to deploy and its effective drag area is adjusted with the current attitude angular velocity as the closed-loop feedback until attitude stability is restored.

Images