Flight stability control method and system based on flight transition stage

Abstract

The invention provides a flight stability control method and system based on a flight transition stage, and the method comprises the steps: firstly obtaining a real-time flight parameter set comprising a power system state parameter, a real-time attitude parameter and an environment disturbance parameter in an aircraft transition stage; determining a flight control parameter set comprising a power distribution weight, a rotor wing tilting angle adjustment threshold value and a redundancy safety compensation coefficient according to the real-time flight parameter set, generating a dynamic adjustment instruction for synchronously adjusting the power output proportion and the tilting angle offset of a plurality of tilting rotor wings according to the flight control parameter set, and then regulating and controlling a power distribution system in real time. A rotor wing tilting control signal and a power redistribution result are generated, and finally the flight state is adjusted according to the result, so that the stability index of the aircraft meets the preset threshold value, multi-parameter accurate regulation and control can be integrated, and the flight stability of the aircraft in the transition stage is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and more particularly, to a flight stability control method and system based on the flight transition stage. Background Art

[0002] In the field of aircraft flight technology, flight stability control has always been a crucial research direction. Especially during the flight transition stage, the aircraft faces more complex and variable operating conditions, posing extremely high challenges to stability control.

[0003] Existing aircraft flight stability control methods do not establish an effective analysis model to comprehensively process various flight parameters. They often adopt simple empirical formulas or fixed parameter adjustment strategies and cannot dynamically and accurately determine appropriate control parameters according to the complex situations faced by the aircraft in real time, resulting in poor control effects.

[0004] In terms of the control of tilt-rotor aircraft, most of the existing technologies independently adjust the power output and tilt angle of each rotor, lacking a synchronous coordination adjustment mechanism. This independent adjustment method is prone to causing disharmony among the rotors and may lead to serious problems such as the loss of control of the aircraft attitude during the flight transition stage.

[0005] In addition, existing methods lack an effective safety compensation mechanism to cope with sudden situations or parameter fluctuations. When the aircraft encounters unexpected disturbances or system parameters deviate, it is difficult to ensure its stability and safety. Summary of the Invention

[0006] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide a flight stability control method based on the flight transition stage, the method comprising:

[0007] Obtain a set of real-time flight parameters of the aircraft during the transition stage, the set of real-time flight parameters including power system state parameters, real-time attitude parameters, and environmental disturbance parameters;

[0008] Based on the set of real-time flight parameters, determine a set of flight control parameters through a flight stability analysis model, the set of flight control parameters including power distribution weights, tilt angle adjustment thresholds of the rotors, and redundant safety compensation coefficients;

[0009] Generate a dynamic adjustment instruction for the power system according to the set of flight control parameters, the dynamic adjustment instruction being used to synchronously adjust the power output ratio and tilt angle offset of multiple tilt rotors;

[0010] Based on the dynamic adjustment instruction, perform real-time regulation on the power distribution system of the aircraft to generate a tilt control signal for the rotors and a power re-distribution result;

[0011] Adjust the flight state of the aircraft during the transition phase according to the rotor tilting control signal and the result of power reallocation, so that the stability index of the aircraft meets a preset threshold.

[0012] On the other hand, an embodiment of the present invention further provides a flight stability control system based on the flight transition phase, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0013] Based on the above aspects, the embodiments of the present invention obtain a multi-dimensional real-time flight parameter set such as the power system state, real-time attitude, and environmental disturbance, and use the flight stability analysis model to determine the flight control parameter set, fully considering the complex interaction and coupling relationship between various parameters, and can adaptively generate key control parameters such as power distribution weights, rotor tilting angle adjustment thresholds, and redundant safety compensation coefficients, effectively improving the pertinence and accuracy of control. Then, generate dynamic adjustment instructions according to the flight control parameter set, and perform real-time regulation on the power distribution system, realizing the synchronous and coordinated adjustment of the power output ratio and tilting angle offset of multiple tilt rotors, avoiding the conflicts and disharmonies that may occur in the independent adjustment of each part in traditional control, enabling the aircraft to respond quickly and smoothly to external changes during the transition phase. Finally, adjust the flight state based on the regulation result to ensure that the stability index meets the preset threshold, achieving the goal of maintaining the stability of the aircraft during the complex and changeable flight transition phase, greatly improving the ability of the aircraft to cope with complex working conditions during the transition phase, and significantly enhancing flight safety and stability. Description of the Drawings

[0014] Figure 1 It is a schematic flowchart of the execution process of the flight stability control method based on the flight transition phase provided by the embodiment of the present invention.

[0015] Figure 2 It is a schematic diagram of an exemplary hardware and software component of the flight stability control system based on the flight transition phase provided by the embodiment of the present invention. Detailed Embodiments

[0016] The present invention will be specifically described below in conjunction with the drawings of the specification. Figure 1 It is a schematic flowchart of the flight stability control method based on the flight transition phase provided by an embodiment of the present invention. The flight stability control method based on the flight transition phase will be introduced in detail below.

[0017] Step S110: Obtain the set of real-time flight parameters of the aircraft during the transition phase, where the set of real-time flight parameters includes power system state parameters, real-time attitude parameters, and environmental disturbance parameters.

[0018] In this embodiment, taking the example of a new aircraft with multiple tilt rotors during the transition from vertical takeoff to horizontal cruise, for obtaining the power system state parameters, sensors are equipped on each tilt rotor of the aircraft. For example, torque sensors are installed near the rotation axis of each rotor to measure its torque output value. Suppose at a certain moment, the torque output value of rotor 1 detected by the sensor is 150 Newton-meters, and the torque output value of rotor 2 is 140 Newton-meters, etc. At the same time, rotor speed sensors can measure the real-time speed of each rotor. If the speed of rotor 1 is 1200 revolutions per minute at this time, and the speed of rotor 2 is 1150 revolutions per minute, etc. Also, the tilt angle of each tilt rotor can be accurately measured. For example, the tilt angle of rotor 1 is 30 degrees, and the tilt angle of rotor 2 is 25 degrees, thus jointly constituting the power system state parameters.

[0019] Regarding obtaining the real-time attitude parameters, the inertial measurement unit of the aircraft can accurately measure the pitch angle, roll angle, and yaw angle of the aircraft. Suppose at a certain instant during the transition phase, the measured pitch angle is 10 degrees, which means the nose of the aircraft is lifted 10 degrees relative to the horizontal direction; the roll angle is 5 degrees, indicating that the aircraft is tilted to one side to a certain extent; the yaw angle is 15 degrees, reflecting the deviation angle of the aircraft's heading relative to a certain reference direction.

[0020] For environmental disturbance parameters, the wind speed sensor detects that the wind speed at this time is 10 meters per second, and by measuring the wind speed at different positions and analyzing the data, the direction of the wind speed can be obtained. The air pressure change gradient sensor finds that the air pressure drops by 2 hectopascals per kilometer in the horizontal direction, indicating that there are certain changes in the air pressure environment around the aircraft. The turbulence intensity sensor detects that the turbulence intensity is at a medium level. For example, the energy fluctuation range of the turbulence is between a certain specific value. The above environmental disturbance parameters reflect the influence of the external environment of the aircraft on the flight state. Finally, the dynamic system state parameters, real-time attitude parameters, and environmental disturbance parameters are fused together to obtain the real-time flight parameter set of the aircraft during the transition stage. And, in order to improve the accuracy of the data, noise filtering is performed on this real-time flight parameter set. The Kalman filter algorithm is used for smoothing the real-time attitude parameters. Assuming that there are some fluctuations in the original measured pitch angle due to noise, a more accurate corrected pitch angle value is obtained after being processed by the Kalman filter algorithm; the sliding window mean algorithm is used for the dynamic system state parameters. For example, the average rotational speed of each rotor and the stable torque output value are calculated within a time window of the past 10 seconds; time series analysis is performed on the environmental disturbance parameters to identify sudden wind speed change events. For example, the situation where the wind speed suddenly increases from 8 meters per second to 12 meters per second within a certain short period of time is accurately identified, and at the same time, invalid turbulence intensity data is removed. Finally, the three processed parameters are aligned according to the time stamp to generate a standardized flight parameter matrix, and then this standardized flight parameter matrix is input into the anomaly detection model to mark and replace the parameter values that exceed the physical limits. For example, if the measured rotational speed value of a rotor exceeds the maximum rotational speed that the aircraft may have during normal operation due to a sensor failure, then this abnormal value will be marked and replaced with a reasonable value.

[0021] Step S120, based on the real-time flight parameter set, determine a flight control parameter set through a flight stability analysis model. The flight control parameter set includes power distribution weights, rotor tilt angle adjustment thresholds, and redundant safety compensation coefficients.

[0022] Still taking the above scenario as an example, relevant data is extracted from the obtained and processed real-time flight parameter set. First is the real-time load distribution data in the dynamic system state parameters. Assuming that based on information such as the torque output value and rotational speed of each rotor, it can be analyzed that at the current moment, the load of the aircraft is more concentrated on the tilt rotors at the front of the fuselage. For example, the front rotors bear 60% of the total load, while the rear rotors bear 40%. At the same time, the dynamic change gradient of the real-time attitude parameters is extracted. For example, within a short period of time in the past, the change rate of the pitch angle is 2 degrees per second, the change rate of the roll angle is 1 degree per second, and the change rate of the yaw angle is 0.5 degree per second.

[0023] Then, input the real-time load distribution data into the first neural network model. The first neural network model has been pre-trained with a large amount of flight data and can output the initial power distribution weights of each tilt-rotor and the range of redundant safety compensation coefficients according to the input load distribution data. Suppose that according to the current load distribution situation, the results output by the first neural network model are: the initial power distribution weight of rotor 1 is 0.3, the initial power distribution weight of rotor 2 is 0.25, etc., and the range of redundant safety compensation coefficients is from 1.1 to 1.3.

[0024] Next, input the dynamic change gradient into the second neural network model, which has also been trained. After inputting the dynamic change gradient, the model outputs the tilt-rotor angle adjustment threshold and the angle correction priority sequence. For example, the tilt-rotor angle adjustment threshold output by the second neural network model is 5 degrees, and the angle correction priority sequence is to first correct the tilt-rotor angle corresponding to the pitch angle, followed by the one corresponding to the roll angle, and finally the one corresponding to the yaw angle.

[0025] Generate dynamic safety constraint conditions based on the range of redundant safety compensation coefficients and the angle correction priority sequence. For example, based on the range of redundant safety compensation coefficients from 1.1 to 1.3, it is stipulated that when making power adjustments and tilt-rotor angle adjustments, it is necessary to ensure that the adjusted system can operate safely within this coefficient range; according to the angle correction priority sequence, determine the order of multiple angles to be corrected simultaneously, so as to ensure flight safety.

[0026] Finally, fuse the initial power distribution weights, the tilt-rotor angle adjustment threshold, and the dynamic safety constraint conditions to generate a set of flight control parameters. For example, in the finally obtained set of flight control parameters, the power distribution weights are rotor 1: 0.3, rotor 2: 0.25, etc., the tilt-rotor angle adjustment threshold is 5 degrees, and the redundant safety compensation coefficient is 1.2.

[0027] Step S130, generate a dynamic adjustment instruction for the power system according to the set of flight control parameters, and the dynamic adjustment instruction is used to synchronously adjust the power output ratio and the tilt angle offset of multiple tilt-rotors.

[0028] In the scenario of the above-mentioned aircraft, extract the power distribution weights, the tilt-rotor angle adjustment threshold, and the redundant safety compensation coefficient from the set of flight control parameters. For example, the power distribution weights are rotor 1: 0.3, rotor 2: 0.25, the tilt-rotor angle adjustment threshold is 5 degrees, and the redundant safety compensation coefficient is 1.2.

[0029] Calculate the reference value of the power output ratio of each tilt-rotor according to the power distribution weight. Assume that the total power output of the aircraft is 1000 kW. Then the reference value of the power output ratio of rotor 1 is 1000×0.3 = 300 kW, and the reference value of the power output ratio of rotor 2 is 1000×0.25 = 250 kW. Then, based on the redundant safety compensation factor, correct the safety margin of the reference value of the power output ratio. Since the redundant safety compensation factor is 1.2, the adjustment amount of the power output ratio of rotor 1 is 300×1.2 = 360 kW, and the adjustment amount of the power output ratio of rotor 2 is 250×1.2 = 300 kW.

[0030] Determine the initial offset of the tilt angle of each rotor according to the tilt angle adjustment threshold of the rotor. Here, the tilt angle adjustment threshold is 5 degrees. Assume that the initial offset of the tilt angle of rotor 1 is set to 3 degrees, and the initial offset of the tilt angle of rotor 2 is 2 degrees. Then, combined with the redundant safety compensation factor, perform dynamic range constraint on the initial offset of the tilt angle. After calculation, the final offset of the tilt angle of each rotor is obtained. For example, the offset of the tilt angle of rotor 1 is 3×1.2 = 3.6 degrees, and the offset of the tilt angle of rotor 2 is 2×1.2 = 2.4 degrees.

[0031] Pair the power output ratio adjustment amount with the tilt angle offset according to the rotor number to generate an initial adjustment instruction set containing rotor identifiers, power ratio adjustment items, and angle offset items. For example, for rotor 1, the generated initial adjustment instruction is (rotor 1, 360 kW, 3.6 degrees), and for rotor 2 it is (rotor 2, 300 kW, 2.4 degrees).

[0032] Sort the execution order of the instructions in the initial adjustment instruction set according to the preset rotor control priority rule to generate a power distribution priority queue and a tilt angle adjustment sequence. Assume that the control priority is determined in the order of power distribution weight from high to low. Then, rotor 1 is first in the power distribution priority queue, and then rotor 2; the tilt angle adjustment sequence also adjusts the tilt angle of rotor 1 first, and then the tilt angle of rotor 2.

[0033] Based on the power distribution priority queue, convert the power ratio adjustment item into a torque adjustment instruction for each rotor motor. For example, according to the power output ratio adjustment amount and the performance parameters of the motor, calculate that the torque adjustment instruction for the rotor 1 motor is to increase by 100 N·m (assumed specific value, calculated according to the actual motor performance), and the torque adjustment instruction for the rotor 2 motor is to increase by 80 N·m. Convert the angle offset item into a pulse width modulation instruction for the rotor drive mechanism according to the tilt angle adjustment sequence. For example, for the 3.6-degree tilt angle offset of rotor 1, convert it into a specific pulse width modulation instruction to control the rotor drive mechanism to achieve the adjustment of this angle, and the same applies to rotor 2.

[0034] Align the torque adjustment command and the pulse width modulation command in time sequence according to the time synchronization signal to generate a dynamic adjustment command containing a synchronization timestamp. For example, at a certain time point t1, the torque adjustment command for rotor 1 and the pulse width modulation command for the tilt angle adjustment are issued simultaneously, and the corresponding commands for rotor 2 are issued at time point t2.

[0035] Finally, the instruction verification module verifies the logical consistency between the power output ratio adjustment amount and the tilt angle offset amount in the dynamic adjustment command. If a conflict is detected, for example, it is found that the torque adjustment calculated according to the power output ratio adjustment amount will cause the tilt angle to exceed the limit of the tilt angle offset amount, then the conflicting command is proportionally reduced based on the redundant safety compensation coefficient. Suppose a conflict is found between the torque adjustment command and the tilt angle adjustment command for rotor 1, and the torque adjustment command is proportionally reduced according to the redundant safety compensation coefficient of 1.2, and a new torque adjustment command is recalculated to generate a verified dynamic adjustment command.

[0036] Step S140, based on the dynamic adjustment command, perform real-time regulation on the power distribution system of the aircraft to generate a rotor tilt control signal and a power reallocation result.

[0037] In this scenario, parse the power output ratio adjustment amount and the tilt angle offset amount in the dynamic adjustment command to generate a power distribution priority queue for each rotor. For example, according to the previously determined power distribution priority queue, rotor 1 is in the priority position, and rotor 2 is next.

[0038] Based on the power distribution priority queue, calculate the real-time torque output value and the tilt angle correction amount for each rotor through the power distribution algorithm. Suppose the power distribution algorithm calculates, according to the dynamic model of the aircraft and the current flight state, that for rotor 1, the real-time torque output value is 400 Newton-meters (this value is calculated based on multiple factors such as the power output ratio adjustment amount, the weight of the aircraft, and the aerodynamic force), and the tilt angle correction amount is 3.5 degrees; for rotor 2, the real-time torque output value is 320 Newton-meters, and the tilt angle correction amount is 2.2 degrees.

[0039] Generate a power reallocation signal according to the real-time torque output value. For example, convert the real-time torque output value into a corresponding voltage or current signal and send it to the motors of each rotor to adjust the output power of the motors. Convert the tilt angle correction amount into a pulse control signal for the rotor drive motor. For example, convert the tilt angle correction amount of 3.5 degrees into a pulse signal with a specific frequency and duty cycle and send it to the drive motor of rotor 1, and send the pulse signal corresponding to 2.2 degrees to the drive motor of rotor 2.

[0040] Verify the compatibility of the power redistribution signal and the pulse control signal through the redundant safety module. First, obtain the operating state data of the current power distribution system, including the historical torque output values of each rotor and the tilt angle error range. Assume that the average value of the historical torque output of rotor 1 over a period of time is 350 N·m, and the tilt angle error range is ±0.5°. Conduct a deviation analysis on the real-time torque output value of 400 N·m in the power redistribution signal and the historical torque output value of 350 N·m, and calculate the torque deviation coefficient as (400 - 350) / 350 = 0.14 (this is just a simple example of the calculation method). Match the tilt angle correction amount of 3.5° in the pulse control signal with the tilt angle error range of ±0.5° to generate the angle correction safety level. Assume that 3.5° is within the allowable error range, and the angle correction safety level is safe.

[0041] Calculate the comprehensive safety assessment score of the power distribution system based on the torque deviation coefficient and the angle correction safety level. For example, according to the preset calculation method, the comprehensive safety assessment score = torque deviation coefficient × 0.6 + score corresponding to the angle correction safety level × 0.4 (here, the score corresponding to the angle correction safety level being safe is assumed to be 0.8), and calculate the comprehensive safety assessment score. If the comprehensive safety assessment score is lower than the preset safety threshold, trigger the activation instruction of the redundant power channel, and regenerate the power redistribution signal and the pulse control signal. If the score is higher than the threshold, generate the verified power redistribution result and the rotor tilt control signal.

[0042] Step S150, adjust the flight state of the aircraft in the transition stage according to the rotor tilt control signal and the power redistribution result, so that the stability index of the aircraft meets the preset threshold.

[0043] In this embodiment, the rotor tilt control signal can be sent to the rotor drive controller, and the rotor drive controller adjusts the tilt angle of each rotor to the target position. For example, send the pulse control signal corresponding to the tilt angle correction amount of rotor 1 obtained previously to the drive controller of rotor 1, and the drive controller controls the drive motor of rotor 1 to adjust the tilt angle of rotor 1 from the current 30° to the target 33.5°; the same applies to rotor 2, and its tilt angle is adjusted from 25° to 27.2°.

[0044] According to the power redistribution result, adjust the motor output power of each rotor through the power distribution controller. According to the previously calculated power redistribution result, the power distribution controller sends corresponding control signals to the motors of each rotor, so that the motor output power of rotor 1 is adjusted to the power value corresponding to 400 N·m of torque, and the motor output power of rotor 2 is adjusted to the power value corresponding to 320 N·m of torque.

[0045] Real-time collect the adjusted flight state data, including the attitude angular velocity of the aircraft, the altitude change rate, and the rotor vibration amplitude. For example, the adjusted attitude angular velocity of the aircraft measured by the attitude sensor is 1 degree per second (a reasonable value assumed here), the altitude change rate detected by the altitude sensor is 0.5 meters per second, and the rotor vibration amplitude measured by the rotor vibration sensor is 0.1 meters.

[0046] Compare the adjusted flight state data with the preset stability indicators to generate a flight state deviation value. Extract the allowable range of attitude angular velocity, the upper limit of altitude change rate, and the threshold of rotor vibration amplitude from the preset stability indicators. Assume that the allowable range of attitude angular velocity is 0.8 degrees to 1.2 degrees per second, the upper limit of altitude change rate is 1 meter per second, and the threshold of rotor vibration amplitude is 0.15 meters. Calculate the deviation percentage of the attitude angular velocity of the aircraft from the allowable range of attitude angular velocity to generate the first deviation factor. Here, the attitude angular velocity is 1 degree per second, the middle value of the allowable range is (0.8 + 1.2) / 2 = 1 degree, and the deviation percentage is (1 - 1) / 1 = 0, so the first deviation factor is 0. Calculate the ratio of the altitude change rate to the upper limit of altitude change rate to generate the second deviation factor, that is, 0.5 / 1 = 0.5. Calculate the absolute value of the difference between the rotor vibration amplitude and the threshold of rotor vibration amplitude to generate the third deviation factor, that is, |0.1 - 0.15| = 0.05. Perform a weighted sum of the first deviation factor, the second deviation factor, and the third deviation factor. Assume that the weights are 0.3, 0.4, and 0.3 respectively. Then the flight state deviation value = 0×0.3 + 0.5×0.4 + 0.05×0.3 = 0.215.

[0047] If the flight state deviation value exceeds the tolerance threshold, update the flight control parameter set based on the adaptive control algorithm and regenerate the dynamic adjustment instruction. Assume that the tolerance threshold is 0.3. Since 0.215 is less than 0.3, the current flight state is within the acceptable range, and there is no need to update the flight control parameter set and regenerate the dynamic adjustment instruction. If the flight state deviation value exceeds 0.3, for example, it is 0.35, then the adaptive control algorithm will be activated. According to the current flight state data and the preset rules, readjust the flight control parameters such as the power distribution weight, the rotor tilting angle adjustment threshold, and the redundant safety compensation coefficient, and then regenerate the dynamic adjustment instruction according to the previous steps to regulate the power distribution system of the aircraft again until the stability indicators of the aircraft meet the preset threshold.

[0048] Based on the above steps, the embodiment of the present invention obtains a multi-dimensional real-time flight parameter set including the power system state, real-time attitude, and environmental disturbances, and uses a flight stability analysis model to determine the flight control parameter set, fully considering the complex interactions and coupling relationships among the parameters. It can adaptively generate key control parameters such as power distribution weights, tilt angle adjustment thresholds of the rotors, and redundant safety compensation coefficients, effectively improving the pertinence and accuracy of control. Then, a dynamic adjustment instruction is generated based on the flight control parameter set, and the power distribution system is regulated in real time, realizing the synchronous and coordinated adjustment of the power output ratio and tilt angle offset of multiple tilt rotors, avoiding the conflicts and incoordination that may occur in the independent adjustment of each part in traditional control, enabling the aircraft to respond quickly and smoothly to external changes during the transition phase. Finally, the flight state is adjusted based on the regulation result to ensure that the stability index meets the preset threshold, achieving the goal of maintaining the stability of the aircraft during the complex and changeable flight transition phase, greatly improving the ability of the aircraft to handle complex working conditions during the transition phase, and significantly enhancing flight safety and stability.

[0049] In a possible implementation manner, step S120 includes:

[0050] Step S121, extracting the real-time load distribution data of the power system state parameters and the dynamic change gradient of the real-time attitude parameters from the real-time flight parameter set.

[0051] Specifically, the power system state parameters contain a lot of information about each tilt-rotor. Among them, the extraction of real-time load distribution data requires comprehensive consideration of multiple factors. For example, the load distribution situation is determined by data such as the torque output value and rotational speed of each tilt-rotor. Suppose the aircraft has four tilt-rotors, namely Rotor 1, Rotor 2, Rotor 3, and Rotor 4. After measurement, the torque output value of Rotor 1 is 150 Newton-meters and the rotational speed is 1200 revolutions per minute; the torque output value of Rotor 2 is 140 Newton-meters and the rotational speed is 1150 revolutions per minute; the torque output value of Rotor 3 is 130 Newton-meters and the rotational speed is 1100 revolutions per minute; the torque output value of Rotor 4 is 120 Newton-meters and the rotational speed is 1050 revolutions per minute. According to these data and the pre-set calculation method, the proportion of the load borne by each rotor is calculated. For example, the proportion of the total load borne by Rotor 1 is (150×1200)÷(150×1200 + 140×1150 + 130×1100 + 120×1050). After a detailed written calculation process, first calculate the numerator 150×1200 which is equal to 180000, then calculate the denominator part, 140×1150 is equal to 161000, 130×1100 is equal to 143000, 120×1050 is equal to 126000. Add these three results and then add 180000 to get 610000. So the proportion of the total load borne by Rotor 1 is 180000÷610000 which is approximately equal to 0.295. By analogy, the load proportions borne by Rotor 2, Rotor 3, and Rotor 4 can be calculated, and this is the real-time load distribution data.

[0052] Regarding the dynamic change gradient of the real-time attitude parameters, within a certain time period, for example, within the past 5 seconds, the pitch angle of the aircraft changes from 8 degrees to 10 degrees, and its dynamic change gradient is (10 - 8)÷5 which is equal to 0.4 degrees per second; the roll angle changes from 3 degrees to 5 degrees, and the dynamic change gradient is (5 - 3)÷5 which is equal to 0.4 degrees per second; the yaw angle changes from 12 degrees to 15 degrees, and the dynamic change gradient is (15 - 12)÷5 which is equal to 0.6 degrees per second.

[0053] Step S122: Input the real-time load distribution data into the first neural network model, and output the initial power distribution weights of each tilt-rotor and the range of redundant safety compensation coefficients.

[0054] The first neural network model performs calculations and analyses based on the input load distribution data, and outputs the initial power distribution weights of each tilt-rotor and the range of redundancy safety compensation coefficients. For example, the results output by the first neural network model are that the initial power distribution weight of rotor 1 is 0.3, the initial power distribution weight of rotor 2 is 0.25, the initial power distribution weight of rotor 3 is 0.2, and the initial power distribution weight of rotor 4 is 0.25. At the same time, the range of redundancy safety compensation coefficients is from 1.1 to 1.3, and this output result is obtained by comparing and analyzing the input load distribution data with historical data based on the algorithm logic inside the model.

[0055] Step S123: Input the dynamic change gradient into the second neural network model, and output the tilt-rotor angle adjustment threshold and the angle correction priority sequence.

[0056] Specifically, after inputting the calculated dynamic change gradient above, the second neural network model outputs the tilt-rotor angle adjustment threshold and the angle correction priority sequence. For example, the tilt-rotor angle adjustment threshold output by the second neural network model is 5 degrees. For the angle correction priority sequence, the second neural network model determines to first correct the tilt-rotor angle corresponding to the yaw angle, followed by the pitch angle, and finally the roll angle according to the flight characteristics of the aircraft and the influence degree of attitude changes.

[0057] Step S124: Generate dynamic safety constraint conditions according to the range of redundancy safety compensation coefficients and the angle correction priority sequence.

[0058] For example, based on the range of redundancy safety compensation coefficients from 1.1 to 1.3, it means that during the subsequent power adjustment and tilt-rotor angle adjustment processes, it is necessary to ensure that the adjusted result is within this range of redundancy safety compensation coefficients to ensure the safe operation of the system. For example, when calculating the power output, the adjusted value must satisfy being between 1.1 times and 1.3 times the original calculated value. According to the angle correction priority sequence, when it is necessary to correct the tilt-rotor angles corresponding to multiple angles simultaneously, the operations must be carried out in the order of yaw angle first, then pitch angle, and finally roll angle to ensure flight safety.

[0059] Step S125: Integrate the initial power distribution weights, the tilt-rotor angle adjustment threshold, and the dynamic safety constraint conditions to generate the set of flight control parameters.

[0060] For example, in the finally obtained set of flight control parameters, the power distribution weights are rotor 1: 0.3, rotor 2: 0.25, rotor 3: 0.2, rotor 4: 0.25; the tilt-rotor angle adjustment threshold is 5 degrees; the redundancy safety compensation coefficient is taken as 1.2. This set of flight control parameters will be used for generating subsequent dynamic adjustment instructions for the power system.

[0061] For example, in a possible implementation, step S130 includes:

[0062] Step S131, extract the power distribution weight, the tilt angle adjustment threshold of the tilt-rotor, and the redundant safety compensation coefficient from the set of flight control parameters.

[0063] For example, in the previous example, the power distribution weight is rotor 1: 0.3, rotor 2: 0.25, rotor 3: 0.2, rotor 4: 0.25; the tilt angle adjustment threshold of the tilt-rotor is 5 degrees; the redundant safety compensation coefficient is 1.2.

[0064] Step S132, calculate the power output ratio reference value of each tilt-rotor according to the power distribution weight, and perform a safety margin correction on the power output ratio reference value based on the redundant safety compensation coefficient to generate the power output ratio adjustment amount of each rotor.

[0065] Assume that the total power output of the aircraft is 1000 kW. Then, for rotor 1, its power output ratio reference value is 1000 × 0.3 = 300 kW; for rotor 2, its power output ratio reference value is 1000 × 0.25 = 250 kW; for rotor 3, its power output ratio reference value is 1000 × 0.2 = 200 kW; for rotor 4, its power output ratio reference value is 1000 × 0.25 = 250 kW. Then, perform a safety margin correction on the power output ratio reference value based on the redundant safety compensation coefficient. For rotor 1, its power output ratio adjustment amount is 300 × 1.2 = 360 kW; for rotor 2, its power output ratio adjustment amount is 250 × 1.2 = 300 kW; for rotor 3, its power output ratio adjustment amount is 200 × 1.2 = 240 kW; for rotor 4, its power output ratio adjustment amount is 250 × 1.2 = 300 kW. This process is to determine the power output adjustment range of each rotor while ensuring safety redundancy.

[0066] Step S133, determine the initial tilt angle offset of each rotor according to the tilt angle adjustment threshold of the tilt-rotor, and perform a dynamic range constraint on the initial tilt angle offset in combination with the redundant safety compensation coefficient to generate the tilt angle offset of each rotor.

[0067] For example, the tilt angle adjustment threshold here is 5 degrees. Assume that the initial tilt angle offset of rotor 1 is 3 degrees, the initial tilt angle offset of rotor 2 is 2 degrees, the initial tilt angle offset of rotor 3 is 2.5 degrees, and the initial tilt angle offset of rotor 4 is 1.5 degrees. Combining with the redundant safety compensation coefficient to dynamically constrain the range of the initial tilt angle offset, for rotor 1, its tilt angle offset is 3×1.2 = 3.6 degrees; for rotor 2, its tilt angle offset is 2×1.2 = 2.4 degrees; for rotor 3, its tilt angle offset is 2.5×1.2 = 3 degrees; for rotor 4, its tilt angle offset is 1.5×1.2 = 1.8 degrees.

[0068] Step S134, pair the power output ratio adjustment amount with the tilt angle offset according to the rotor number to generate an initial adjustment instruction set including rotor identifiers, power ratio adjustment items, and angle offset items.

[0069] For example, for rotor 1, the generated initial adjustment instruction is (rotor 1, 360 kW, 3.6 degrees); for rotor 2, it is (rotor 2, 300 kW, 2.4 degrees); for rotor 3, it is (rotor 3, 240 kW, 3 degrees); for rotor 4, it is (rotor 4, 300 kW, 1.8 degrees).

[0070] Step S135, sort the execution order of the instructions in the initial adjustment instruction set according to the preset rotor control priority rule to generate a power distribution priority queue and a tilt angle adjustment sequence.

[0071] Assume that the control priority is determined in the order from high to low of the power distribution weight. Then, rotor 1 is the first in the power distribution priority queue, followed by rotor 2 and rotor 4 (because their power distribution weights are the same, and the order can be further refined according to other rules. Here, assume rotor 2 first and then rotor 4), and finally rotor 3. The tilt angle adjustment sequence is also to adjust the tilt angle of rotor 1 first, then rotor 2, then rotor 4, and finally rotor 3.

[0072] Step S136, based on the power distribution priority queue, convert the power ratio adjustment item into a torque adjustment instruction for each rotor motor, and convert the angle offset item into a pulse width modulation instruction for the rotor drive mechanism according to the tilt angle adjustment sequence.

[0073] For example, for rotor 1, it is known that the torque-power relationship of its motor is 2 Newton-meters of torque corresponding to each kilowatt (set according to the performance parameters of the motor here). Then, according to the power output ratio adjustment amount of 360 kilowatts, the torque adjustment command of its motor can be calculated as 360×2 = 720 Newton-meters; for rotor 2, following the same calculation method, according to its power output ratio adjustment amount of 250 kilowatts, the torque adjustment command of its motor is 250×2 = 500 Newton-meters; for rotor 3, its torque adjustment command is 240×2 = 480 Newton-meters; for rotor 4, its torque adjustment command is 300×2 = 600 Newton-meters. Convert the angle offset term into the pulse width modulation command of the rotor drive mechanism according to the tilt angle adjustment sequence, and this conversion needs to be carried out according to the characteristics of the rotor drive mechanism. For example, for the 3.6-degree tilt angle offset of rotor 1, by querying the pre-set correspondence table between the tilt angle and the pulse width modulation command (this correspondence table is obtained based on the performance test of the rotor drive mechanism), the corresponding pulse width modulation command is obtained, and the same applies to rotors 2, 3, and 4.

[0074] Step S137, align the torque adjustment command and the pulse width modulation command in time sequence according to the time synchronization signal to generate a dynamic adjustment command including a synchronous timestamp.

[0075] For example, set the time interval of the time synchronization signal to 0.1 seconds. At the time point t = 0 seconds, set the synchronous timestamp for the torque adjustment command of rotor 1 and the pulse width modulation command for tilt angle adjustment. At t = 0.1 seconds, set the synchronous timestamp for the corresponding commands of rotor 2, and so on, to generate a dynamic adjustment command including a synchronous timestamp.

[0076] Step S138, verify the logical consistency of the power output ratio adjustment amount and the tilt angle offset amount in the dynamic adjustment command through the command verification module. If a conflict is detected, proportionally reduce the conflicting command based on the redundant safety compensation coefficient to generate a verified dynamic adjustment command.

[0077] For example, the verification module checks whether the torque adjustment of each rotor calculated according to the power output ratio adjustment amount will cause the tilt angle to exceed the limit of the tilt angle offset amount. Suppose it is found that for rotor 3, the result calculated according to its torque adjustment command may cause the tilt angle to exceed the limit of its 3-degree tilt angle offset amount. In this case, proportionally reduce the conflicting command based on the redundant safety compensation coefficient. For rotor 3, proportionally reduce its torque adjustment command according to the redundant safety compensation coefficient of 1.2. The original torque adjustment command is 480 Newton-meters, and the reduced torque adjustment command is 480÷1.2 = 400 Newton-meters, thereby generating a verified dynamic adjustment command.

[0078] In a possible implementation manner, step S140 includes:

[0079] Step S141, parsing the power output ratio adjustment amount and tilt angle offset amount in the dynamic adjustment instruction, and generating a power distribution priority queue for each rotor.

[0080] In this embodiment, dynamic adjustment instructions for each rotor have been obtained previously. For example, the power output ratio adjustment amount of rotor 1 is 360 kW, and the tilt angle offset amount is 3.6 degrees; the power output ratio adjustment amount of rotor 2 is 300 kW, and the tilt angle offset amount is 2.4 degrees; the power output ratio adjustment amount of rotor 3 is 240 kW, and the tilt angle offset amount is 3 degrees; the power output ratio adjustment amount of rotor 4 is 300 kW, and the tilt angle offset amount is 1.8 degrees. According to the previously set power distribution priority rule, such as in the order of decreasing power distribution weight, the generated power distribution priority queue is first rotor 1, followed by rotor 2 and rotor 4 (since the two had the same power distribution weight and a predetermined order before), and finally rotor 3.

[0081] Step S142, based on the power distribution priority queue, calculating the real-time torque output value and tilt angle correction amount for each rotor through a power distribution algorithm.

[0082] For example, the power distribution algorithm can comprehensively consider various factors of the aircraft, such as the weight of the aircraft, aerodynamic characteristics, and the current flight attitude. For rotor 1, assuming that the aircraft is in the current flight state, according to the power distribution algorithm, combined with its power output ratio adjustment amount of 360 kW, considering that the total weight of the aircraft is 5000 kg and relevant factors such as the force coefficient of the air on the rotor (these coefficients are obtained through prior aerodynamic tests and calculations), calculate its real-time torque output value. Assuming that the torque coefficient generated by each kilowatt of power is 2 N·m (this is a relationship determined according to the design of the rotor motor and the overall aircraft), then the real-time torque output value of rotor 1 is 360×2 = 720 N·m. For the tilt angle correction amount of 3.6 degrees, according to the requirements for aircraft attitude adjustment in the power distribution algorithm, combined with the current attitude data (such as the pitch angle, roll angle, yaw angle and their change gradients mentioned above), after a complex calculation process, it is determined that this tilt angle correction amount meets the adjustment requirements of the current flight state. For rotor 2, similarly according to the power distribution algorithm, combined with its power output ratio adjustment amount of 300 kW, calculated according to the torque coefficient of 2 N·m per kilowatt, its real-time torque output value is 300×2 = 600 N·m, and the tilt angle correction amount of 2.4 degrees is also determined according to the algorithm combined with the aircraft attitude data. For rotor 3, its real-time torque output value is 240×2 = 480 N·m, and the tilt angle correction amount of 3 degrees is also the result of the algorithm combined with the attitude data. For rotor 4, the real-time torque output value is 300×2 = 600 N·m, and the tilt angle correction amount of 1.8 degrees is also determined based on the relevant algorithm and attitude data.

[0083] Step S143, generate a power redistribution signal according to the real-time torque output value, and convert the tilt angle correction amount into a pulse control signal for the rotor drive motor.

[0084] For example, convert the real-time torque output value of 720 N·m of rotor 1 into a corresponding electrical signal or other control signals (the specific signal type depends on the control mechanism of the aircraft power system), and this signal will be sent to the motor control system of rotor 1 to adjust the output power of the motor, thereby realizing power redistribution. At the same time, convert the tilt angle correction amount into a pulse control signal for the rotor drive motor. For example, for the tilt angle correction amount of 3.6 degrees of rotor 1, according to the pre-set correspondence between the tilt angle and the pulse control signal (this relationship is obtained through tests and characteristic analysis of the rotor drive motor), convert 3.6 degrees into a specific pulse control signal, and this pulse control signal will be sent to the drive motor of rotor 1 to control the tilt angle of the rotor. For other rotors, such as rotor 2, rotor 3, and rotor 4, similarly generate a power redistribution signal according to their respective real-time torque output values and convert it into a pulse control signal according to the tilt angle correction amount.

[0085] Step S144: Verify the compatibility between the power redistribution signal and the pulse control signal through the redundant safety module, and generate a verified power redistribution result and a rotor tilting control signal.

[0086] In a possible implementation, step S144 includes:

[0087] Step S1441: Obtain the operation status data of the current power distribution system, where the operation status data includes the historical torque output values of each rotor and the tilting angle error range.

[0088] Assume that during a previous flight, the average historical torque output value of rotor 1 was 650 Newton - meters, and the tilting angle error range was plus or minus 0.5 degrees; the average historical torque output value of rotor 2 was 550 Newton - meters, and the tilting angle error range was plus or minus 0.4 degrees; the average historical torque output value of rotor 3 was 420 Newton - meters, and the tilting angle error range was plus or minus 0.3 degrees; the average historical torque output value of rotor 4 was 580 Newton - meters, and the tilting angle error range was plus or minus 0.4 degrees.

[0089] Step S1442: Conduct a deviation analysis on the real - time torque output value in the power redistribution signal and the historical torque output value, and generate a torque deviation coefficient.

[0090] For rotor 1, the real - time torque output value in the power redistribution signal is 720 Newton - meters, and the average historical torque output value is 650 Newton - meters. Calculate the deviation value as 720 - 650 = 70 Newton - meters. The torque deviation coefficient is the deviation value divided by the historical value, that is, 70÷650 (here is the detailed calculation process: first calculate 70÷650, reduce 70 and 650 by 10 times to become 7÷65, 7÷65 is approximately equal to 0.1077) is approximately equal to 0.1077. For rotor 2, the real - time torque output value is 600 Newton - meters, and the historical value is 550 Newton - meters on average. The deviation value is 600 - 550 = 50 Newton - meters. The torque deviation coefficient is 50÷550 (calculation process: 50÷550, reduce by 10 times to become 5÷55, 5÷55 is approximately equal to 0.0909) is approximately equal to 0.0909. For rotor 3, the real - time torque output value is 480 Newton - meters, and the historical value is 420 Newton - meters on average. The deviation value is 480 - 420 = 60 Newton - meters. The torque deviation coefficient is 60÷420 (calculation process: 60÷420, reduce by 60 times to become 1÷7, 1÷7 is approximately equal to 0.1429) is approximately equal to 0.1429. For rotor 4, the real - time torque output value is 600 Newton - meters, and the historical value is 580 Newton - meters on average. The deviation value is 600 - 580 = 20 Newton - meters. The torque deviation coefficient is 20÷580 (calculation process: 20÷580, reduce by 20 times to become 1÷29, 1÷29 is approximately equal to 0.0345) is approximately equal to 0.0345.

[0091] Step S1443: Match the tilt angle correction amount in the pulse control signal with the tilt angle error range to generate an angle correction safety level.

[0092] For rotor 1, the tilt angle correction amount is 3.6 degrees, and the tilt angle error range is plus or minus 0.5 degrees. Since 3.6 degrees is within the error range, the angle correction safety level is safe (a specific level value can be set, for example, 1 represents the highest safety level). For rotor 2, the tilt angle correction amount is 2.4 degrees, and the tilt angle error range is plus or minus 0.4 degrees. Since 2.4 degrees is within the error range, the angle correction safety level is safe. For rotor 3, the tilt angle correction amount is 3 degrees, and the tilt angle error range is plus or minus 0.3 degrees. Since 3 degrees exceeds the error range, the angle correction safety level is dangerous (set to 0 to represent the dangerous level). For rotor 4, the tilt angle correction amount is 1.8 degrees, and the tilt angle error range is plus or minus 0.4 degrees. Since 1.8 degrees is within the error range, the angle correction safety level is safe.

[0093] Step S1444: Calculate the comprehensive safety assessment score of the power distribution system based on the torque deviation coefficient and the angle correction safety level.

[0094] Assume that the weight of the set torque deviation coefficient is 0.6 and the weight of the angle correction safety level is 0.4. For rotor 1, the comprehensive safety assessment score is equal to the torque deviation coefficient 0.1077 multiplied by 0.6 plus the angle correction safety level 1 multiplied by 0.4 (calculation process: 0.1077×0.6 = 0.06462, 1×0.4 = 0.4, 0.06462 + 0.4 = 0.46462), approximately equal to 0.46462. For rotor 2, the comprehensive safety assessment score is equal to 0.0909×0.6 plus 1×0.4 (calculation process: 0.0909×0.6 = 0.05454, 0.05454 + 0.4 = 0.45454), approximately equal to 0.45454. For rotor 3, the comprehensive safety assessment score is equal to 0.1429×0.6 plus 0×0.4 (calculation process: 0.1429×0.6 = 0.08574, 0.08574 + 0 = 0.08574), approximately equal to 0.08574. For rotor 4, the comprehensive safety assessment score is equal to 0.0345×0.6 plus 1×0.4 (calculation process: 0.0345×0.6 = 0.0207, 0.0207 + 0.4 = 0.4207), approximately equal to 0.4207.

[0095] Step S1445: If the comprehensive safety assessment score is lower than the preset safety threshold, trigger the activation instruction of the redundant power channel and regenerate the power reallocation signal and the pulse control signal.

[0096] Assume that the preset safety threshold is 0.4. For rotor 3, since its comprehensive safety assessment score of 0.08574 is lower than 0.4, an activation instruction for the redundant power channel will be triggered, and a power redistribution signal and a pulse control signal will be regenerated. The regeneration process will again consider various state parameters of the aircraft, following the previous calculation method, but will be adjusted based on meeting the redundant safety constraint conditions to ensure the safety of the power distribution system and the stable flight of the aircraft. For other rotors, such as rotor 1, rotor 2, and rotor 4, since their comprehensive safety assessment scores are higher than the preset safety threshold, their power redistribution signals and pulse control signals will be used as the results passed the verification for subsequent adjustment of the flight state of the aircraft.

[0097] Step S145, when a power distribution conflict is detected, trigger an activation instruction for the redundant power channel based on the dynamic safety constraint conditions.

[0098] In a possible implementation manner, step S150 includes:

[0099] Step S151, send the rotor tilting control signal to the rotor drive controller, and adjust the tilting angles of each rotor to the target position through the rotor drive controller.

[0100] The tilting control signals of each rotor have been determined previously. For example, for rotor 1, the tilting angle corresponding to its tilting control signal is 3.6 degrees. This tilting control signal is sent to the drive controller of rotor 1, and the drive controller controls the operation of the drive motor of rotor 1 according to this signal, so that the tilting angle of rotor 1 is gradually adjusted to the target position of 3.6 degrees. For rotor 2, the tilting angle corresponding to its tilting control signal is 2.4 degrees, and its drive controller also controls the motor to adjust the tilting angle to 2.4 degrees. For rotor 3, the tilting angle is 3 degrees, and it reaches this target position after being adjusted by the drive controller. For rotor 4, the tilting angle is 1.8 degrees, and it is also adjusted in place by the drive controller.

[0101] Step S152, according to the power redistribution result, adjust the motor output power of each rotor through the power distribution controller.

[0102] For example, the previously calculated power redistribution result shows that the real-time torque output value of rotor 1 is 720 Newton-meters. The power distribution controller adjusts the output power of the motor of rotor 1 according to this result, so that the torque output by the motor can reach 720 Newton-meters. This adjustment is achieved based on the performance characteristics and control mechanism of the motor. For rotor 2, its real-time torque output value is 600 Newton-meters, and the power distribution controller makes corresponding adjustments to the output power of its motor; for rotor 3, the real-time torque output value is 480 Newton-meters, and the power distribution controller adjusts its motor power accordingly; for rotor 4, the real-time torque output value is 600 Newton-meters, and the power distribution controller also adjusts the output power of the motor.

[0103] Step S153, collect the adjusted flight state data in real time. The adjusted flight state data includes the attitude angular velocity of the aircraft, the altitude change rate, and the rotor vibration amplitude.

[0104] These data are collected by the attitude sensor, altitude sensor, and rotor vibration sensor on the aircraft. Assume that the attitude sensor collects the attitude angular velocity of the aircraft as 1.1 degrees per second, the altitude sensor detects the altitude change rate as 0.6 meters per second, and the rotor vibration sensor measures the rotor vibration amplitude as 0.12 meters.

[0105] Step S154, compare the adjusted flight state data with the preset stability index to generate a flight state deviation value.

[0106] In a possible implementation, step S154 includes:

[0107] Step S1541, extract the allowable range of attitude angular velocity, the upper limit of altitude change rate, and the rotor vibration amplitude threshold from the preset stability index.

[0108] Assume that the preset allowable range of attitude angular velocity is from 0.9 degrees to 1.3 degrees per second, the upper limit of altitude change rate is 1 meter per second, and the rotor vibration amplitude threshold is 0.15 meters.

[0109] Step S1542, calculate the deviation percentage of the attitude angular velocity of the aircraft from the allowable range of attitude angular velocity to generate a first deviation factor.

[0110] For example, the middle value of the allowable range of attitude angular velocity is (0.9 + 1.3) ÷ 2 = 1.1 degrees. The attitude angular velocity is 1.1 degrees per second, the deviation value is 1.1 - 1.1 = 0 degrees, and the deviation percentage is 0 ÷ 1.1 (detailed calculation process: 0 divided by 1.1 equals 0) = 0. So the first deviation factor is 0.

[0111] Step S1543, calculate the ratio of the altitude change rate to the upper limit of altitude change rate to generate a second deviation factor.

[0112] For example, the rate of change of altitude is 0.6 meters per second, the upper limit of the rate of change of altitude is 1 meter per second, and the ratio is 0.6÷1 = 0.6. Therefore, the second deviation factor is 0.6.

[0113] Step S1544: Calculate the absolute value of the difference between the amplitude of the rotor vibration and the threshold of the amplitude of the rotor vibration to generate a third deviation factor.

[0114] For example, the amplitude of the rotor vibration is 0.12 meters, the threshold of the amplitude of the rotor vibration is 0.15 meters, and the absolute value of the difference is |0.12 - 0.15| = 0.03 meters. Therefore, the third deviation factor is 0.03.

[0115] Step S1545: Perform a weighted sum of the first deviation factor, the second deviation factor, and the third deviation factor to generate the flight state deviation value.

[0116] For example, assume that the weight of the first deviation factor is 0.3, the weight of the second deviation factor is 0.4, and the weight of the third deviation factor is 0.3. The flight state deviation value is equal to the first deviation factor 0 multiplied by 0.3 plus the second deviation factor 0.6 multiplied by 0.4 plus the third deviation factor 0.03 multiplied by 0.3 (detailed calculation process: 0×0.3 = 0, 0.6×0.4 = 0.24, 0.03×0.3 = 0.009, 0 + 0.24 + 0.009 = 0.249), which is equal to 0.249.

[0117] Step S155: If the flight state deviation value exceeds the tolerance threshold, update the flight control parameter set based on the adaptive control algorithm and regenerate the dynamic adjustment instruction.

[0118] For example, assume that the tolerance threshold is 0.3. Since 0.249 is less than 0.3, the current flight state deviation is within the acceptable range, and there is no need to update the flight control parameter set and regenerate the dynamic adjustment instruction. If the flight state deviation value exceeds 0.3, for example, it is 0.35, then the adaptive control algorithm will be activated. The adaptive control algorithm will re-adjust the flight control parameters according to the current flight state data, such as the attitude angular velocity of 1.1 degrees per second, the rate of change of altitude of 0.6 meters per second, the amplitude of the rotor vibration of 0.12 meters, and the preset rules. For example, re-evaluate the power distribution weights. It may be found that the power distribution weight of rotor 1 needs to be adjusted to 0.28, rotor 2 to 0.26, rotor 3 to 0.22, and rotor 4 to 0.24 according to the current flight state. For the rotor tilting angle adjustment threshold, it may be adjusted to 4.5 degrees, and the redundant safety compensation coefficient may be adjusted to 1.15, etc. Then, regenerate the dynamic adjustment instruction according to the previous steps and regulate the power distribution system of the aircraft again until the stability index of the aircraft meets the preset threshold.

[0119] In a possible implementation, step S110 includes:

[0120] Step S111, collecting real-time attitude parameters of the aircraft through the inertial measurement unit of the aircraft, where the real-time attitude parameters include pitch angle, roll angle, and yaw angle.

[0121] For example, at a certain moment, the inertial measurement unit detects that the pitch angle of the aircraft is 12 degrees, which means the nose of the aircraft is lifted 12 degrees upward relative to the horizontal direction; the roll angle is 4 degrees, indicating that the aircraft is tilted to one side to a certain extent; the yaw angle is 10 degrees, reflecting the deviation angle of the aircraft's heading relative to a certain reference direction.

[0122] Step S112, collecting real-time rotational speeds, torque output values, and tilting angles of each tilt-rotor through a rotor sensor to generate power system state parameters.

[0123] Suppose the aircraft has four tilt-rotors, namely rotor 1, rotor 2, rotor 3, and rotor 4. The rotor sensor detects that the real-time rotational speed of rotor 1 is 1250 revolutions per minute, the torque output value is 160 Newton-meters, and the tilting angle is 32 degrees; the real-time rotational speed of rotor 2 is 1200 revolutions per minute, the torque output value is 150 Newton-meters, and the tilting angle is 28 degrees; the real-time rotational speed of rotor 3 is 1150 revolutions per minute, the torque output value is 140 Newton-meters, and the tilting angle is 25 degrees; the real-time rotational speed of rotor 4 is 1100 revolutions per minute, the torque output value is 130 Newton-meters, and the tilting angle is 22 degrees. These data constitute the power system state parameters and can reflect the working states of each tilt-rotor.

[0124] Step S113, collecting environmental disturbance parameters through an environmental sensor, where the environmental disturbance parameters include wind speed, air pressure change gradient, and turbulence intensity.

[0125] For example, the environmental sensor detects that the wind speed is 8 meters per second, and based on the air pressure measurement data at different positions, it is calculated that the air pressure drops by 1.5 hectopascals per kilometer in the horizontal direction, which is the air pressure change gradient. At the same time, through a dedicated turbulence sensor, the turbulence intensity is detected to be at a medium level, and the specific value is assumed to be a certain quantization value (according to the measurement and calibration of the sensor).

[0126] Step S114, fusing the power system state parameters, real-time attitude parameters, and environmental disturbance parameters to generate the set of real-time flight parameters.

[0127] Step S115, performing noise filtering processing on the set of real-time flight parameters.

[0128] In a possible implementation, step S115 includes:

[0129] Step S1151, perform smoothing processing on the real-time attitude parameters using the Kalman filtering algorithm to generate the corrected pitch angle, roll angle, and yaw angle.

[0130] Taking the pitch angle as an example, the originally measured pitch angle may fluctuate due to factors such as sensor noise. During a certain period, a series of pitch angle values are measured, such as 11.8 degrees, 12.2 degrees, 11.9 degrees, etc. The Kalman filtering algorithm will perform weighted calculations based on the previous state estimate value and the current measurement value to obtain a more accurate corrected pitch angle value. Suppose that after being processed by the Kalman filtering algorithm, the corrected pitch angle is 12.0 degrees. The same processing is also carried out for the roll angle and yaw angle. For example, the original roll angle measurement values are 3.8 degrees, 4.2 degrees, 3.9 degrees, etc., and the corrected roll angle after filtering is 4.0 degrees; the original yaw angle measurement values are 9.8 degrees, 10.2 degrees, 10.1 degrees, etc., and the corrected yaw angle after filtering is 10.0 degrees.

[0131] Step S1152, use the moving window average algorithm for the dynamic system state parameters to calculate the average rotation speed of each rotor and the stable torque output value.

[0132] Taking rotor 1 as an example, set the moving window size to the data of the past 10 seconds. Within these 10 seconds, the rotation speeds of rotor 1 are different values such as 1240 revolutions per minute, 1250 revolutions per minute, 1245 revolutions per minute, etc. Add these 10 rotation speed values and then divide by 10 to obtain the average rotation speed. The detailed calculation process is as follows: Suppose the sum of these 10 rotation speed values is 12450 revolutions (1240 + 1250 + 1245 + ……), then the average rotation speed is 12450 ÷ 10 which is 1245 revolutions per minute. The same calculation is also carried out for the torque output value. Suppose the torque output values within 10 seconds are 158 Newton - meters, 160 Newton - meters, 159 Newton - meters, etc., and the sum of these values is assumed to be 1590 Newton - meters, then the stable torque output value is 1590 ÷ 10 which is 159 Newton - meters. The average rotation speed and the stable torque output value of rotors 2, 3, and 4 are also calculated in the same way.

[0133] Step S1153, perform time - series analysis on the environmental disturbance parameters to identify wind speed mutation events and eliminate invalid turbulence intensity data.

[0134] For example, over a period of time, the measured wind speed data forms a time series. For example, the measured wind speed values at consecutive time points are 7 m / s, 8 m / s, 9 m / s, 15 m / s, 10 m / s, etc. When the wind speed suddenly increases from 9 m / s to 15 m / s, this is identified as a wind speed mutation event. For the turbulence intensity data, due to possible mismeasurements or interference of the sensor, some turbulence intensity data may not conform to the actual physical laws or have too large a difference from the surrounding data, and these invalid turbulence intensity data will be excluded.

[0135] Step S1154: Align the processed real-time attitude parameters, power system state parameters, and environmental disturbance parameters according to the time stamps to generate a standardized flight parameter matrix.

[0136] The time stamp is an identifier that marks each data acquisition moment, ensuring that the data at the same moment of each parameter can be accurately corresponding. For example, at a specific time point t1, the corresponding corrected pitch angle, the average rotational speed of each rotor, the wind speed, etc. are arranged together in a certain order to form a row of the standardized flight parameter matrix. The data at different time points are arranged in sequence to form a complete standardized flight parameter matrix.

[0137] Step S1155: Input the standardized flight parameter matrix into the anomaly detection model to mark and replace the parameter values that exceed the physical limits.

[0138] The anomaly detection model is trained based on a large amount of normal flight data, and it can identify the parameter values that do not conform to the physical laws or exceed the normal operating range of the aircraft. For example, if due to a sensor failure, the average rotational speed of a certain rotor is measured as 5000 revolutions per minute, which obviously exceeds the possible range of the rotor rotational speed during the normal operation of this aircraft, the anomaly detection model will mark this abnormal value and replace it with a reasonable value according to the design parameters and normal operating range of the aircraft. For example, according to the normal range of the rotor rotational speed of this type of aircraft, it is replaced with a reasonable value of about 1200 revolutions per minute.

[0139] In a possible implementation manner, the training method of the flight stability analysis model includes:

[0140] Step S210: Obtain a historical flight data set, where the historical flight data set includes multiple flight parameter samples in the transition stage and corresponding stability label data.

[0141] In this embodiment, taking a certain set of numerous historical flight data as an example, the flight parameter samples in the transition stage contain rich information. In terms of the state parameters of the power system, for multiple tilt-rotors, for example, the rotational speed of rotor 1 may be 1,300 revolutions per minute, the torque output value is 165 Newton-meters, and the tilt angle is 33 degrees; the rotational speed of rotor 2 is 1,250 revolutions per minute, the torque output value is 155 Newton-meters, and the tilt angle is 30 degrees; the rotational speed of rotor 3 is 1,200 revolutions per minute, the torque output value is 145 Newton-meters, and the tilt angle is 27 degrees, etc.

[0142] For the real-time attitude parameters, for example, the pitch angle is 11 degrees, which represents the angle at which the head of the aircraft is lifted upward relative to the horizontal position; the roll angle is 6 degrees, which reflects the degree of left-right tilt of the aircraft; the yaw angle is 13 degrees, which reflects the deviation of the aircraft's heading relative to a certain reference direction.

[0143] For the environmental disturbance parameters, the wind speed may be 9 meters per second, and this value has a direct impact on the flight state of the aircraft; the air pressure change gradient is a decrease of 1.8 hectopascals per kilometer, which is related to the change in the air pressure environment around the aircraft; the turbulence intensity is a certain quantified value, assumed to be at a medium-high level, which will affect the stability of the aircraft.

[0144] The stability label data is data related to the set of flight control parameters. For example, in the stability label data of a certain flight, in terms of the power distribution weight, the power distribution weight of rotor 1 is 0.32, which means that in the current flight state, the total power is distributed to rotor 1 according to this ratio; the power distribution weight of rotor 2 is 0.28; the power distribution weight of rotor 3 is 0.25, etc. The tilt angle adjustment threshold of the rotor is 4.5 degrees, which is a critical value for the tilt angle adjustment of the rotor to ensure stable flight in this flight state. The redundant safety compensation coefficient is 1.15, and this redundant safety compensation coefficient is used to ensure a certain safety margin during power distribution and rotor adjustment. These stability label data are obtained through analysis and summary from actual flights, and reflect the control parameter conditions corresponding to the stable flight of the aircraft under specific flight parameter samples.

[0145] Step S220, construct a convolutional neural network model, and the convolutional neural network model includes a parameter encoding layer, a multi-modal feature fusion layer, and a control parameter prediction layer.

[0146] Specifically, the parameter encoding layer is mainly responsible for extracting features from the input flight parameter samples in the transition stage, converting the flight parameter samples into high-dimensional feature vectors, which involves a series of convolution operations, pooling operations, and the application of non-linear activation functions. For example, through convolution operations, the convolution kernel slides on the flight parameter sample data to extract local features. For data such as the rotor speed and torque in the power system state parameters, the convolution kernel can capture the local correlations in the data; the pooling operation downsamples the features, reducing the amount of data while retaining important features; the non-linear activation function (such as the ReLU function) introduces non-linearity into the model, enabling the model to learn more complex relationships.

[0147] The function of the multi-modal feature fusion layer is to fuse features of different modalities. In this scenario, the multi-modal feature fusion layer processes the correlation analysis between the high-dimensional feature vectors obtained from the parameter encoding layer and the external environment features. The external environment features may include some macroscopic meteorological data or geographical information, etc. Although not detailed in the previous examples, these information will also affect flight in actual situations. The role of this multi-modal feature fusion layer is to fuse these multi-faceted information together to provide a comprehensive feature representation for the subsequent prediction of control parameters.

[0148] The control parameter prediction layer includes the first neural network model and the second neural network model of the foregoing embodiments. The control parameter prediction layer is the part that predicts the set of flight control parameters based on the fused features. For example, the information processed by the previous two layers can be used to generate a prediction result for the set of flight control parameters through an internal mapping relationship, such as predicting the power distribution weight, the adjustment threshold of the rotor tilting angle, and the redundant safety compensation coefficient, etc.

[0149] Step S230: Input the flight parameter samples in the transition stage into the parameter encoding layer to extract high-dimensional feature vectors.

[0150] Taking a certain flight parameter sample in the transition stage mentioned before as an example, when it is input into the parameter encoding layer, the whole process is complex and progressive.

[0151] For the power system state parameter part, such as the data of the rotor speed, torque, and tilting angle of rotor 1, etc., they will first be decomposed into small data blocks, and the convolution kernel slides one by one on these data blocks. Assuming the size of the convolution kernel is 3×3 (just for illustration here), when sliding on the rotor speed data of rotor 1, the weighted sum of 3 consecutive rotor speed data is calculated each time (the weights are determined by the convolution kernel), and this process is carried out on the entire rotor speed data to obtain a series of local features. The same operation is performed for the torque and tilting angle data.

[0152] Then, a pooling operation is performed. For example, max pooling is adopted to select the maximum value within a 2×2 area as the representative, which can reduce the amount of data. After alternating operations of multiple convolutional layers and pooling layers, the features of these dynamic system state parameters are gradually abstracted and refined.

[0153] A similar processing process is also applied to the real-time attitude parameters and environmental disturbance parameters. Finally, all these processed features are combined with the results processed by a non-linear activation function (such as the ReLU function, which sets values less than 0 to 0 and keeps values greater than 0 unchanged) to form a high-dimensional feature vector. This high-dimensional feature vector contains an abstract feature representation of various information in the flight parameter samples. It is no longer the original flight parameters but a feature representation form that can be better processed by the subsequent layers of the model.

[0154] Step S240: Through the multi-modal feature fusion layer, perform correlation analysis on the high-dimensional feature vector and the external environmental features to generate a fusion feature matrix.

[0155] Step S250: Through the control parameter prediction layer, map the fusion feature matrix to the stability label data to generate the prediction result of the flight control parameter set.

[0156] The control parameter prediction layer uses the previously obtained fusion feature matrix to predict the flight control parameter set. This process is achieved through an internal mapping function, which is learned during the model training process.

[0157] Taking the previously generated fusion feature matrix as an example, it contains rich flight-related information. The control parameter prediction layer predicts the values of flight control parameters such as the power distribution weight, the rotor tilting angle adjustment threshold, and the redundant safety compensation coefficient based on this information. For example, it predicts that the power distribution weight is rotor 1: 0.33, rotor 2: 0.27, etc.; the rotor tilting angle adjustment threshold is 4.6 degrees; and the redundant safety compensation coefficient is 1.16. The above prediction results are obtained based on the information in the fusion feature matrix and the patterns learned by the model during the training process.

[0158] Step S260: Use the mean squared error loss function to train the prediction result and the stability label data until the model converges.

[0159] For example, the mean squared error loss function is an index used to measure the error between the prediction result and the actual stability label data.

[0160] When calculating the mean squared error, it is necessary to compare the prediction results with each corresponding element in the stability label data. For example, for the power distribution weight, the power distribution weight of rotor 1 in the prediction result is 0.33, while in the stability label data it is 0.32, so their difference is 0.33 - 0.32 = 0.01. For the rotor tilt angle adjustment threshold, the predicted value is 4.6 degrees, and the actual label data is 4.5 degrees, with a difference of 4.6 - 4.5 = 0.1 degrees; for the redundant safety compensation coefficient, the predicted value is 1.16 and the actual value is 1.15, with a difference of 1.16 - 1.15 = 0.01.

[0161] Then calculate the square of the difference for each element. The square of the difference for the power distribution weight is 0.01^2 = 0.0001; the square of the difference for the rotor tilt angle adjustment threshold is 0.1^2 = 0.01; the square of the difference for the redundant safety compensation coefficient is 0.01^2 = 0.0001.

[0162] Finally, find the average of these squared values. Assuming only these three elements are involved in the calculation, the mean squared error is (0.0001 + 0.01 + 0.0001) ÷ 3 = 0.0034.

[0163] During the training process, the parameters of the convolutional neural network model, such as the weights of the convolutional kernels and the connection weights of the fully connected layers, are continuously adjusted according to this mean squared error. Through the backpropagation algorithm, the mean squared error is propagated backward from the output layer to the previous layers, and the parameters of each layer are adjusted according to the error. For example, if the weight of a certain convolutional kernel has a greater impact on the mean squared error, then a larger adjustment will be made to this weight; if the connection weight of a certain fully connected layer has a smaller impact on the mean squared error, then the adjustment amplitude will be relatively smaller.

[0164] As the number of training times increases, the mean squared error will gradually decrease. When the mean squared error no longer decreases significantly, that is, the model converges. At this time, the model has learned enough knowledge to accurately predict the flight control parameter set according to the flight parameter samples in the transition stage, thus providing reliable support for the flight control of the aircraft in the transition stage.

[0165] Among them, step S240 includes:

[0166] Step S241, dividing the high-dimensional feature vector into a power system feature subset, a rotor state feature subset, and an environmental feature subset.

[0167] The power system feature subset contains features related to the power system extracted from high-dimensional feature vectors, such as the feature parts related to the power output and power distribution of each rotor; the rotor state feature subset contains features related to the state of the rotor, such as the tilting angle and vibration state of the rotor; the environmental feature subset contains features related to environmental factors, such as the manifestation of factors such as wind speed and air pressure change gradient in the high-dimensional feature vector.

[0168] Step S242, perform cross-attention calculation on the power system feature subset and the rotor state feature subset to generate a first fused feature vector.

[0169] In this calculation process, the complex mutual relationship between the power system and the rotor state needs to be considered. For example, there may be a specific relationship between the power output of a certain rotor and its tilting angle. When the tilting angle of the rotor changes, its power output will also change accordingly. The cross-attention calculation will dynamically allocate attention weights according to the importance of this relationship. Specifically, during the calculation, by comparing the correlation between each element of the power system feature subset and the rotor state feature subset, a higher weight is given to the element combinations with high correlation. For example, if a strong correlation is found between the torque output of rotor 1 and the tilting angle, then a larger weight will be given to the element combination corresponding to this relationship when calculating the first fused feature vector. Specifically, assume that the correlation coefficient between an element of the power system feature subset and an element of the rotor state feature subset is 0.8 (this coefficient is obtained through a specific calculation method, such as calculating their covariance, etc.), and the correlation coefficients between other elements are relatively low. Then, when generating the first fused feature vector by weighted summation, these two element combinations with high correlation will account for a relatively large share in the summation result.

[0170] Step S243, perform a graph convolution operation on the first fused feature vector and the environmental feature subset to construct a global feature relationship graph.

[0171] Specifically, the first fusion feature vector and the environmental feature subset can be regarded as nodes in a graph. For example, an element in the first fusion feature vector and the wind speed element in the environmental feature subset can be regarded as two nodes in the graph. Edges are constructed based on their relationships. If there is an influence relationship between them, such as the wind speed affecting the power output-related elements in the first fusion feature vector, then an edge will be constructed between these two nodes. Through graph convolution operations, information is propagated on this graph. The graph convolution operation will perform a weighted sum of the information of neighboring nodes according to the weights of the edges (the weights reflect the strength of the relationship between nodes), thereby updating the information of each node. After multiple graph convolution operations, the constructed global feature relationship graph can reflect the mutual relationships among various features of the aircraft in the global scope, including the complex interaction relationships among the power system, rotor state, and environmental features.

[0172] Step S244: Perform temporal dependence modeling on the global feature relationship graph through a gated recurrent unit to generate a dynamic fusion feature sequence.

[0173] Since the flight state of the aircraft changes over time, it is necessary to consider this temporal dependence. There are two gating mechanisms inside the gated recurrent unit, namely the update gate and the forget gate. Taking the state of the global feature relationship graph at several consecutive time steps as an example, when the update gate determines which information needs to be updated, it can calculate an update coefficient based on the current input global feature relationship graph information and the hidden state at the previous moment (the hidden state is a state information maintained inside the gated recurrent unit for recording previous information). The forget gate determines which information needs to be forgotten and also calculates a forget coefficient based on the current input and the hidden state at the previous moment. Through these two gating mechanisms, the gated recurrent unit selectively updates and transmits information, thereby generating a dynamic fusion feature sequence. For example, if at a certain time step, the attitude angle of the aircraft changes significantly, then the gated recurrent unit will adjust the element values in the dynamic fusion feature sequence according to the previous state and the current input to reflect the temporal characteristics of this attitude change.

[0174] Step S245: Input the dynamic fusion feature sequence into a fully connected layer for dimensionality reduction processing to output the fusion feature matrix.

[0175] Each neuron in the fully connected layer is connected to all neurons in the previous layer. During this process, the neurons in the fully connected layer perform linear combination and non-linear transformation on the elements in the dynamically fused feature sequence. For example, for a certain element in the dynamically fused feature sequence, the neurons in the fully connected layer multiply the element according to its connection weights, then add a bias term, and then pass through a non-linear activation function (such as the tanh function) to obtain a new value. By performing such operations on all elements in the dynamically fused feature sequence, the high-dimensional dynamically fused feature sequence is converted into a low-dimensional fused feature matrix. The fused feature matrix synthesizes information from multiple aspects such as the dynamic system, rotorcraft state, and environmental features, and retains features that are important for predicting flight control parameters during the dimensionality reduction process.

[0176] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of a flight stability control system 100 based on a flight transition phase that can implement the inventive concept provided by some embodiments of the present invention. For example, the processor 120 can be used on the flight stability control system 100 based on the flight transition phase and is used to execute the functions in the present invention.

[0177] The flight stability control system 100 based on the flight transition phase can be a general-purpose server or a special-purpose server, both of which can be used to implement the flight stability control method based on the flight transition phase of the present invention. Although only one server is shown in the present invention, for convenience, the functions described in the present invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0178] For example, the flight stability control system 100 based on the flight transition phase can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the flight stability control system 100 based on the flight transition phase can also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present invention can be implemented according to these program instructions. The flight stability control system 100 based on the flight transition phase also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0179] For ease of explanation, only one processor is described in the flight stability control system 100 based on the flight transition phase. However, it should be noted that the flight stability control system 100 based on the flight transition phase in the present invention may also include multiple processors. Therefore, the steps performed by one processor described in the present invention may also be jointly performed or separately performed by multiple processors. For example, if the processor of the flight stability control system 100 based on the flight transition phase performs step A and step B, it should be understood that step A and step B may also be jointly performed by two different processors or separately performed in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0180] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the flight stability control method based on the flight transition phase as described above is implemented.

[0181] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. A flight stability control method based on the flight transition phase, characterized in that: The method comprises: Acquire a real-time flight parameter set of the aircraft in the transition phase, wherein the real-time flight parameter set includes power system state parameters, real-time attitude parameters, and environmental disturbance parameters; Based on the real-time flight parameter set, determining a flight control parameter set through a flight stability analysis model, wherein the flight control parameter set includes a power distribution weight, a rotor tilt angle adjustment threshold, and a redundant safety compensation coefficient; generating a dynamic adjustment instruction for a power system according to the flight control parameter set, wherein the dynamic adjustment instruction is used to synchronously adjust a power output ratio and a tilt angle offset of a plurality of tilt rotors; Based on the dynamic adjustment instruction, the power distribution system of the aircraft is controlled in real time to generate a rotor tilt control signal and a power redistribution result; The transitional flight state of the aircraft is adjusted according to the rotor tilt control signal and the power redistribution result, so that the stability index of the aircraft meets a preset threshold.

2. The flight stability control method based on the flight transition phase according to claim 1, characterized in that: The step of determining a flight control parameter set based on the real-time flight parameter set by using a flight stability analysis model comprises: Extracting real-time load distribution data of power system state parameters and dynamic change gradient of real-time attitude parameters from the real-time flight parameter set; Inputting the real-time load distribution data into a first neural network model, and outputting the initial power distribution weight and redundant safety compensation coefficient range of each tilt-rotor; Inputting the dynamic change gradient into a second neural network model, and outputting a rotor tilt angle adjustment threshold and an angle correction priority sequence; Generate dynamic safety constraint conditions according to the redundant safety compensation coefficient range and the angle correction priority sequence; The flight control parameter set is generated by integrating the initial power distribution weight, the rotor tilt angle adjustment threshold and the dynamic safety constraint condition.

3. The flight stability control method based on the flight transition phase according to claim 1, characterized in that: The real-time regulation of the power distribution system of the aircraft based on the dynamic adjustment instruction to generate a rotor tilt control signal and a power redistribution result includes: Analyzing the power output ratio adjustment amount and the tilt angle offset in the dynamic adjustment instruction to generate a power allocation priority queue for each rotor; Based on the power allocation priority queue, the real-time torque output value and tilt angle correction value of each rotor are calculated by a power allocation algorithm; generating a power redistribution signal according to the real-time torque output value, and converting the tilt angle correction amount into a pulse control signal of a rotor drive motor; Verifying the compatibility of the power redistribution signal and the pulse control signal through a redundant safety module, and generating a verified power redistribution result and a rotor tilt control signal; When a power distribution conflict is detected, an activation command of a redundant power channel is triggered based on the dynamic safety constraint condition.

4. The flight stability control method based on the flight transition phase according to claim 3, characterized in that: The verifying the compatibility of the power redistribution signal and the pulse control signal by the redundant safety module includes: Acquiring operating status data of the current power distribution system, wherein the operating status data includes a torque output history value and a tilt angle error range of each rotor; Performing deviation analysis on the real-time torque output value in the power redistribution signal and the torque output historical value to generate a torque deviation coefficient; Matching the tilt angle correction amount in the pulse control signal with the tilt angle error range to generate an angle correction safety level; Calculating a comprehensive safety assessment score of the power distribution system according to the torque deviation coefficient and the angle correction safety level; If the comprehensive safety assessment score is lower than a preset safety threshold, the activation instruction of the redundant power channel is triggered, and the power redistribution signal and the pulse control signal are regenerated.

5. The flight stability control method based on the flight transition phase according to claim 1, characterized in that: The step of adjusting the transitional flight state of the aircraft according to the rotor tilt control signal and the power redistribution result includes: The rotor tilt control signal is sent to a rotor drive controller, and the rotor drive controller is used to adjust the tilt angle of each rotor to a target position; According to the power redistribution result, adjusting the motor output power of each rotor by a power distribution controller; Collecting adjusted flight status data in real time, wherein the adjusted flight status data includes the aircraft attitude angular velocity, altitude change rate and rotor vibration amplitude; Comparing the adjusted flight status data with a preset stability index to generate a flight status deviation value; If the flight state deviation value exceeds a tolerance threshold, the flight control parameter set is updated based on an adaptive control algorithm and a dynamic adjustment instruction is regenerated.

6. The flight stability control method based on the flight transition phase according to claim 5, characterized in that: The step of comparing the flight status data with a preset stability index to generate a flight status deviation value includes: Extracting the attitude angular velocity allowable range, the altitude change rate upper limit and the rotor vibration amplitude threshold from the preset stability index; Calculating the deviation percentage between the attitude angular velocity of the aircraft and the allowable range of the attitude angular velocity to generate a first deviation factor; Calculating a ratio of the altitude change rate to the altitude change rate upper limit to generate a second deviation factor; calculating an absolute value of a difference between the rotor vibration amplitude and the rotor vibration amplitude threshold to generate a third deviation factor; The first deviation factor, the second deviation factor and the third deviation factor are weightedly summed to generate the flight status deviation value.

7. The flight stability control method based on the flight transition phase according to claim 1, characterized in that: The step of obtaining a real-time flight parameter set of the aircraft in the transition phase includes: The real-time attitude parameters of the aircraft are collected by an inertial measurement unit of the aircraft, wherein the real-time attitude parameters include a pitch angle, a roll angle, and a yaw angle; The real-time rotation speed, torque output value and tilt angle of each tilt rotor are collected through the rotor sensor to generate the power system state parameters; Collecting environmental disturbance parameters through environmental sensors, wherein the environmental disturbance parameters include wind speed, air pressure gradient and turbulence intensity; Fusion of the power system state parameters, real-time attitude parameters and environmental disturbance parameters to generate the real-time flight parameter set; A noise filtering process is performed on the real-time flight parameter set.

8. The flight stability control method based on the flight transition phase according to claim 7, characterized in that: The performing noise filtering on the real-time flight parameter set includes: The real-time attitude parameters are smoothed using a Kalman filter algorithm to generate corrected pitch angles, roll angles, and yaw angles; The sliding window mean algorithm is used for the power system state parameters to calculate the average speed of each rotor and the torque output stability value; Performing time series analysis on the environmental disturbance parameters to identify wind speed mutation events and eliminate invalid turbulence intensity data; Align the processed real-time attitude parameters, power system state parameters and environmental disturbance parameters according to timestamps to generate a standardized flight parameter matrix; The normalized flight parameter matrix is ​​input into an anomaly detection model to flag and replace parameter values ​​that exceed physical limits.

9. The flight stability control method based on the flight transition phase according to claim 1, characterized in that: The training method of the flight stability analysis model comprises: Acquire a historical flight data set, wherein the historical flight data set includes a plurality of transition phase flight parameter samples and corresponding stability label data; Constructing a convolutional neural network model, wherein the convolutional neural network model includes a parameter encoding layer, a multimodal feature fusion layer, and a control parameter prediction layer; Inputting the transition phase flight parameter samples into the parameter encoding layer to extract high-dimensional feature vectors; The high-dimensional feature vector is correlated with the external environment feature through the multimodal feature fusion layer to generate a fusion feature matrix; Mapping the fused feature matrix to the stability label data through the control parameter prediction layer to generate a prediction result of the flight control parameter set; The prediction result and the stability label data are trained using a mean square error loss function until the model converges; The step of performing correlation analysis on the high-dimensional feature vector and the external environment feature through the multimodal feature fusion layer to generate a fusion feature matrix includes: Dividing the high-dimensional feature vector into a power system feature subset, a rotor state feature subset, and an environment feature subset; Performing cross-attention calculation on the power system feature subset and the rotor state feature subset to generate a first fusion feature vector; Performing a graph convolution operation on the first fused feature vector and a subset of environmental features to construct a global feature relationship graph; Modeling the temporal dependency of the global feature relationship graph through a gated recurrent unit to generate a dynamic fusion feature sequence; The dynamic fusion feature sequence is input into the fully connected layer for dimensionality reduction processing, and the fusion feature matrix is ​​output.

10. A flight stability control system based on the flight transition phase, characterized in that: The flight stability control system based on the flight transition phase includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the flight stability control method based on the flight transition phase as described in any one of claims 1 to 9 above.

Images