Real-time fault control reconfiguration method based on redundant propulsion system

Abstract

The present application relates to low altitude heavy load multi-rotor control technical field, especially in real-time fault control reconstruction method based on redundancy propulsion system, the present application is according to the abnormal situation of the propulsion system of each aircraft according to the operation data of propulsion system and high-frequency mutation tension component, the abnormal situation of the propulsion system of each aircraft is analyzed according to load state data and low-frequency oscillation tension component;According to the abnormal situation of the rope connection of each aircraft in combination with formation topology data, distributed dynamic reconstruction is executed, and the formation configuration of the reconstructed aircraft is obtained;In combination with low-frequency oscillation tension component, the load stability of the rope connection of each aircraft after reconstruction is predicted;Distributed fault-tolerant control instructions of each aircraft are generated and issued.

Description

Technical Field

[0001] This invention relates to the field of low-altitude heavy-load multirotor control technology, and in particular to a real-time fault control reconfiguration method based on a redundant propulsion system. Background Technology

[0002] During coordinated lifting operations, multiple aircraft form a strongly coupled rigid-flexible hybrid system with the load via cables. The aircraft must not only resist their own disturbances but also coordinate the distribution of load gravity and inertial forces through the cables. As the aircraft propulsion systems operate under long-term high loads, complex airflow, and load oscillation impacts, the thrusters are prone to redundant propulsion system failures such as thrust reduction, speed fluctuations, and abnormal current. Simultaneously, the lifting cable connections may loosen, experience reduced stiffness, or suffer partial failure under repeated tension fluctuations, load oscillations, and external impacts. If these failures are not identified and controlled in a timely manner, insufficient thrust from a single aircraft or abnormal lifting cable connections can be transmitted to the entire formation via cable coupling, leading to load attitude instability, formation divergence, and even major accidents such as load falling or aircraft collisions.

[0003] However, existing technologies for fault-tolerant control of multi-aircraft collaborative hoisting systems do not combine cable tension vectors with real-time aircraft attitude data for spectral decomposition of low-frequency oscillations and mid-to-high-frequency abrupt changes. For example, when the propulsion system experiences early thrust fluctuations, without combining the aircraft attitude quaternion coordinate transformation with cable direction projection, it is difficult to separate the mid-to-high-frequency abrupt change components reflecting propulsion anomalies from the tension signal coupled with load swing, thus making it impossible to accurately analyze the degree of propulsion system anomalies and hoisting cable connection anomalies. At the same time, existing technologies also lack continuous prediction of the dynamic thrust margin and hoisting cable connection load stability after reconstruction, making it difficult for the fault-tolerant control reconstruction results to accurately match the actual health status and load stability requirements of each aircraft, and failing to provide a comprehensive basis for the generation of distributed fault-tolerant commands. Summary of the Invention

[0004] To overcome the defects and shortcomings of existing technologies, this invention provides a real-time fault control reconfiguration method based on redundant propulsion systems. By decomposing the cable tension vector into low-frequency oscillation components and mid-to-high-frequency abrupt change components, and combining propulsion operation data and load status data, the method calculates propulsion system anomaly indicators and cable connection anomaly indicators. It then performs distributed dynamic reconfiguration based on the overall health and formation topology, and predicts dynamic thrust margin and cable connection load stability based on the reconfigured configuration to generate continuous fault-tolerant control commands.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, embodiments of the present invention provide a real-time fault control reconfiguration method based on a redundant propulsion system, comprising the following steps: Step S1: Through the distributed sensing units deployed on each collaborative hoisting aircraft, synchronously collect the propulsion system operation data, cable tension vector data, and load status data of each aircraft; at the same time, acquire the aircraft formation topology data and the real-time attitude data of the aircraft. Step S2: Combining cable tension vector data and aircraft real-time attitude data, decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium-to-high frequency abrupt tension components caused by propulsion system anomalies. Step S3: Analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and combine the formation topology data to perform distributed dynamic reconstruction to obtain the reconstructed aircraft formation configuration. Step S4: Based on the reconstructed aircraft formation configuration and combined with the low-frequency oscillation tension component, predict the dynamic thrust margin required for each aircraft to suppress load swing during the reconstruction transition, and predict the load stability of the suspender cable connection of each aircraft after reconstruction. Step S5: Based on the predicted stability results of the rigging connection load of each aircraft after reconstruction, generate and issue distributed fault-tolerant control commands to each aircraft.

[0006] In one implementation of the present invention, the cable tension vector is decomposed into a low-frequency oscillating tension component caused by load oscillation and a mid-to-high frequency abrupt tension component caused by propulsion system anomalies; specifically, the following steps are included: S21. Obtain the attitude quaternion of each aircraft in the real-time attitude data of the aircraft, and at the same time obtain the cable tension vector directly measured by the force sensor on the corresponding aircraft in the sensor coordinate system; generate the rotation matrix of each aircraft according to the attitude quaternion; multiply the rotation matrix by the cable tension vector in the sensor coordinate system to obtain the inertial frame cable tension vector of each aircraft. S22. Obtain the aircraft position coordinates from the real-time attitude data and the load position coordinates from the load status data, and calculate the unit vector of the cable direction from each aircraft to the load. S23. Decompose the inertial tether tension vector of each aircraft into the component along the tether direction and the component perpendicular to the tether direction of each aircraft. S24. Obtain the load swing angle from the load status data, perform a short-time Fourier transform on the load swing angle, extract the peak frequency at each moment in the time spectrum as the instantaneous frequency of the load swing, and use the instantaneous frequency of the load swing as the main frequency of the load swing. S25. Using the load swing main frequency as the center frequency, set the bandwidth, and perform zero-phase adaptive bandpass filtering on the vertical cable direction component of each aircraft to obtain the low-frequency oscillation tension component of each aircraft in the inertial frame that is in the same direction as the vertical cable direction component. S26. Subtract the static tension component and the low-frequency oscillation tension component from the inertial tether tension vector of each aircraft to obtain the residual tension component of each aircraft; wherein, the static tension component is the low-pass average value of the component along the tether direction of the corresponding aircraft within a preset sliding window. S27. Perform high-pass filtering on the residual tension components of each aircraft to obtain the mid-to-high frequency abrupt tension components of each aircraft; the cutoff frequency of the high-pass filter is higher than the load swing frequency band.

[0007] In one implementation of the present invention, the abnormal conditions of the propulsion system of each aircraft are analyzed based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; specifically including: S31. Extract the real-time current sequence and real-time speed sequence of each aircraft thruster within a preset sliding window from the propulsion system operation data; calculate the corresponding thruster current fluctuation rate and speed fluctuation rate based on the real-time current sequence and real-time speed sequence of each aircraft thruster within the preset sliding window. S32. Based on the mid-to-high frequency abrupt tension components of each aircraft, calculate the proportion of mid-to-high frequency abrupt tension energy and the peak factor of mid-to-high frequency abrupt tension for each aircraft. S33. Combine the corresponding thruster current fluctuation rate, corresponding thruster speed fluctuation rate, mid-to-high frequency sudden tension energy ratio, and mid-to-high frequency sudden tension peak factor of the corresponding aircraft into the propulsion anomaly feature vector of the corresponding aircraft, input it into the pre-trained propulsion anomaly analysis model, and obtain the propulsion system anomaly of the corresponding aircraft.

[0008] In one implementation of the present invention, the abnormal connection status of the suspension cables of each aircraft is analyzed based on load status data and low-frequency oscillation tension components; including the following specific steps: S34. Extract the load swing angle signal sequence and load swing angular velocity sequence within the preset sliding window from the load state data; based on the load swing angle signal sequence, load swing angular velocity sequence, and low-frequency oscillation tension components of each aircraft, analyze the correlation coefficient between the load swing angle and the low-frequency oscillation tension component modulus, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy of the corresponding aircraft. S35. Combine the correlation coefficient between the load swing angle and the low-frequency oscillation tension component mode of the corresponding aircraft, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy into the abnormal feature vector of the suspension rope connection of the corresponding aircraft. Input the vector into the pre-trained suspension rope connection anomaly analysis model to obtain the abnormal situation of the suspension rope connection of the corresponding aircraft.

[0009] In one implementation of the present invention, distributed dynamic reconstruction is performed in conjunction with formation topology data to obtain the reconstructed aircraft formation configuration; including the following specific contents: S36. Calculate the cooperative transport health of the corresponding aircraft based on the abnormal conditions of the aircraft's propulsion system and the abnormal conditions of the sling connection. The weighted sum of the abnormal conditions of the corresponding aircraft's propulsion system and the abnormal conditions of the corresponding aircraft's sling connection is obtained by subtracting the weighted sum from the value. S37. Obtain aircraft formation topology data. Under the distributed communication topology, each aircraft exchanges its cooperative transport health with its neighboring aircraft. Minimize the local cost function of each aircraft through a distributed optimization algorithm to obtain the reconstructed aircraft formation configuration. In one implementation of the present invention, the dynamic thrust margin required to suppress load swing during the reconfiguration transition of each aircraft is predicted; this includes the following specific steps: S41. Based on the reconstructed aircraft formation configuration, calculate the desired cable direction unit vector from each aircraft to the load. S42. Based on the low-frequency oscillation tension components of each aircraft, estimate the maximum additional thrust amplitude required for each aircraft to suppress load swing during the reconfiguration transition. S43. Based on the reconstructed aircraft formation configuration, calculate the reference acceleration of each aircraft from its current position to its reconstructed position, and multiply the mass of the corresponding aircraft by the reference acceleration to obtain the maneuvering acceleration thrust required by the corresponding aircraft; take the sum of the static hovering thrust, the maximum additional thrust amplitude, and the maneuvering acceleration thrust required by the corresponding aircraft as the dynamic thrust requirement of the corresponding aircraft. S44. Calculate the remaining thrust of each aircraft based on the abnormal conditions of its propulsion system. S45. Divide the difference between the remaining thrust of each aircraft and the dynamic thrust requirement of the corresponding aircraft by the remaining thrust of the corresponding aircraft to obtain the dynamic thrust margin of each aircraft.

[0010] In one implementation of the present invention, the stability of the sling connection load of each aircraft after reconfiguration is predicted; the specific steps include the following: S46. Extract the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and reconstruct the geometric uniformity characteristics of the configuration for each aircraft. S47. Combine the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and geometric uniformity features of the reconstructed configuration of each aircraft into the stability feature vector of the corresponding aircraft, and input it into the pre-trained suspension cable connection load stability prediction model to obtain the suspension cable connection load stability of the corresponding aircraft.

[0011] In one implementation of the present invention, based on the predicted stability of the rigging connection load of each aircraft after reconstruction, distributed fault-tolerant control commands for each aircraft are generated and issued; including the following specific steps: S51. Obtain the load stability of the sling connection of each aircraft; S52. Based on the stability of the rigging connection load and the dynamic thrust margin of each aircraft, calculate the control gain scaling factor and thrust distribution weight of each aircraft. S53. Based on the reconstructed aircraft formation configuration, using the stability of the rigging load of each aircraft as the transition speed adjustment factor, generate a smooth reference trajectory for each aircraft from its current position to the reconstructed target position. S54. Combine the smoothed reference trajectory, control gain scaling factor, and thrust distribution weight of each aircraft into a distributed fault-tolerant control command, and send it to the corresponding aircraft through the inter-aircraft communication network; each aircraft adjusts the control gain and thrust distribution of its own position controller and attitude controller according to the received distributed fault-tolerant control command to complete the fault-tolerant control.

[0012] Secondly, embodiments of the present invention also provide a real-time fault control reconfiguration system based on a redundant propulsion system, including: The data acquisition module is used to synchronously collect propulsion system operation data, cable tension vector data, and load status data of each aircraft through distributed sensing units deployed on each collaborative hoisting aircraft; at the same time, it acquires aircraft formation topology data and real-time attitude data of the aircraft. The tension decomposition module is used to combine cable tension vector data and aircraft real-time attitude data to decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium- and high-frequency abrupt tension components caused by propulsion system anomalies. The formation reconfiguration module is used to analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; to analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and to perform distributed dynamic reconfiguration based on the formation topology data to obtain the reconfigured aircraft formation configuration. The load prediction module is used to predict the dynamic thrust margin required by each aircraft to suppress load swing during the reconfiguration transition based on the reconfigured aircraft formation configuration and the low-frequency oscillation tension component, and to predict the load stability of the suspender cable connection of each aircraft after reconfiguration. The control module generates and issues distributed fault-tolerant control commands to each aircraft based on the predicted stability of the rigging connection load of each aircraft after reconstruction.

[0013] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. This invention improves the accuracy of fault identification by performing spectral decomposition on cable tension to distinguish between propulsion abnormalities and suspension rope connection abnormalities; 2. This invention avoids threshold jumps and improves control smoothness by outputting continuous dimensionless anomaly and stability indicators; 3. This invention reduces the risk of secondary instability and ensures reconfiguration safety by predicting the dynamic thrust margin and load stability after reconfiguration; 4. This invention enhances the fault tolerance capability of a multi-aircraft collaborative hoisting system through distributed dynamic reconfiguration and fault-tolerant control. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the overall process of the real-time fault control and reconfiguration method based on redundant propulsion systems of the present invention. Figure 2 This is an analysis flowchart of step S3 in the real-time fault control reconfiguration method based on redundant propulsion system of the present invention; Figure 3 This is an analysis flowchart of step S4 in the real-time fault control reconfiguration method based on redundant propulsion system of the present invention; Figure 4 This is a schematic diagram of the real-time fault control and reconfiguration system based on a redundant propulsion system according to the present invention. Detailed Implementation

[0015] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0016] Example 1

[0017] In the field of industrial control, multi-aircraft collaborative lifting systems are key equipment for aerial transport and emergency rescue of large-size, heavy-mass loads. Their operational stability directly affects the success or failure of lifting missions, equipment safety, and the safety of personnel and property on site. Redundant propulsion systems, as the core power components of each aircraft, are prone to abnormalities such as thrust reduction and speed fluctuations under long-term high loads and load oscillation impacts. Simultaneously, the lifting cable connections may experience a decrease in stiffness or partial failure under repeated tension fluctuations. If these faults are not identified and controlled in a timely manner, they will be transmitted to the entire formation through cable coupling, leading to loss of load attitude control or even a crash.

[0018] Traditional multi-aircraft cooperative hoisting fault-tolerant control strategies are mainly based on centralized threshold detection or preset rules. By monitoring parameters such as thruster current, speed, and total cable tension, formation reconfiguration or thrust redistribution is triggered when an anomaly is detected, thereby avoiding operational risks.

[0019] However, these existing technologies have significant technical limitations, as follows: First, traditional control strategies rely solely on a single information source or total threshold for fault diagnosis, neglecting the spectral differences between low-frequency tension changes caused by normal load oscillations and mid-to-high-frequency abrupt tension changes caused by propulsion system anomalies. In practice, this often results in normal load oscillations being misjudged as propulsion faults, or early propulsion anomalies being masked by load oscillations and thus missed.

[0020] Secondly, existing methods do not distinguish between propulsion system anomalies and suspension rope connection anomalies, resulting in vague identification of fault sources. This leads to fault-tolerant reconfiguration strategies failing to accurately target the fault type, unreasonable reconfiguration configurations, and difficulty in effectively isolating faulty units.

[0021] Third, existing methods lack continuous prediction of dynamic thrust margin and stability of suspender cable connection load during formation reconfiguration. Furthermore, the quantification of fault severity often uses graded thresholds or subjective evaluation, leading to jumps in control commands. After reconfiguration, the system may still experience secondary faults due to insufficient thrust or load instability.

[0022] In view of the above-mentioned technical limitations, such as Figure 1 As shown, this embodiment provides a real-time fault control reconfiguration method based on a redundant propulsion system; specifically, it includes the following steps: Step S1: Through the distributed sensing units deployed on each collaborative hoisting aircraft, synchronously collect the propulsion system operation data, cable tension vector data, and load status data of each aircraft; at the same time, acquire the aircraft formation topology data and the real-time attitude data of the aircraft. Step S2: Combining cable tension vector data and aircraft real-time attitude data, decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium-to-high frequency abrupt tension components caused by propulsion system anomalies. Step S3: Analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and combine the formation topology data to perform distributed dynamic reconstruction to obtain the reconstructed aircraft formation configuration. Step S4: Based on the reconstructed aircraft formation configuration and combined with the low-frequency oscillation tension component, predict the dynamic thrust margin required for each aircraft to suppress load swing during the reconstruction transition, and predict the load stability of the suspender cable connection of each aircraft after reconstruction. Step S5: Based on the predicted stability results of the rigging connection load of each aircraft after reconstruction, generate and issue distributed fault-tolerant control commands to each aircraft.

[0023] In one specific embodiment, the cable tension vector is decomposed into a low-frequency oscillating tension component caused by load oscillation and a mid-to-high frequency abrupt tension component caused by propulsion system anomalies; specifically including the following steps: S21. Obtain the attitude quaternion of each aircraft in the real-time attitude data of the aircraft, and at the same time obtain the cable tension vector directly measured by the force sensor on the corresponding aircraft in the sensor coordinate system; generate the rotation matrix of each aircraft according to the attitude quaternion; wherein, the rotation matrix is ​​obtained by the attitude quaternion of the corresponding aircraft according to the quaternion to rotation matrix formula; multiply the rotation matrix by the cable tension vector in the sensor coordinate system to the left to obtain the inertial frame cable tension vector of each aircraft. S22. Obtain the aircraft position coordinates from the real-time aircraft pose data and the load position coordinates from the load status data, and calculate the cable direction unit vector from each aircraft to the load; the cable direction unit vector is the result of dividing the difference between the load position coordinates and the aircraft position coordinates by the modulus of the difference between the load position coordinates and the aircraft position coordinates. S23. Decompose the inertial tether tension vector of each aircraft into the component along the tether direction and the component perpendicular to the tether direction of each aircraft; wherein, the dot product of the inertial tether tension vector of each aircraft with the corresponding unit vector of the tether direction of the aircraft is multiplied by the corresponding unit vector of the tether direction of the aircraft to obtain the component along the tether direction of each aircraft; the vector obtained by subtracting the component along the tether direction of the aircraft from the inertial tether tension vector of each aircraft is taken as the component perpendicular to the tether direction of each aircraft. To facilitate understanding of the physical meaning of signal processing in subsequent steps S24 to S27, the two components obtained from the decomposition in S23 are further explained below: The component along the cable direction mainly consists of the static projection of the load gravity in the cable direction and the elastic deformation force of the cable itself. Its amplitude variation reflects the weight distribution of the load borne by the aircraft and belongs to the low-frequency quasi-static component. The component perpendicular to the cable direction is mainly caused by the load swing inertial force, Coriolis force, and horizontal disturbance of the aircraft. The cable tension change caused by the load swing is concentrated in this vertical component. Therefore, the subsequent step S25 uses the vertical cable direction component as input to extract the low-frequency oscillating tension component coupled with the load swing. This decomposition method provides a clear mapping relationship for the subsequent analysis of the physical meaning of the signal.

[0024] S24. Obtain the load swing angle from the load status data, perform a short-time Fourier transform on the load swing angle, extract the peak frequency at each moment in the time spectrum as the instantaneous frequency of the load swing, and use the instantaneous frequency of the load swing as the main frequency of the load swing. S25. Using the load swing frequency as the center frequency and setting the bandwidth, perform zero-phase adaptive bandpass filtering on the vertical cable direction component of each aircraft to obtain the low-frequency oscillation tension component of each aircraft in the inertial frame, which is in the same direction as the vertical cable direction component. The "zero-phase adaptive bandpass filtering" refers to bandpass filtering using a forward-backward bilateral filter (i.e., the FLTFFIT function). Its characteristic is that the phase response of the filter is zero, meaning there is no phase delay between the filtered signal and the unfiltered signal. This ensures that the extracted low-frequency oscillation tension component and the load swing angle signal maintain an accurate correspondence on the time axis, avoiding systematic deviations in subsequent calculations of the correlation coefficient and phase difference between the load swing angle and the low-frequency oscillation tension component due to phase shift. The "adaptive" aspect is reflected in the fact that the center frequency of the bandpass filter is the instantaneous load swing frequency extracted in real time in step S24. This frequency is dynamically updated according to the actual state of the load swing, rather than being fixed to a certain preset frequency value.

[0025] S26. Subtract the static tension component and the low-frequency oscillation tension component from the inertial tether tension vector of each aircraft to obtain the residual tension component of each aircraft; wherein, the static tension component is the low-pass average value of the component along the tether direction of the corresponding aircraft within a preset sliding window; wherein, the low-frequency oscillation component is a vector and its direction is vertical; it should be noted that the above-mentioned inertial tether tension vector, static tension component, low-frequency oscillation tension component and residual tension component are all in three-dimensional vector form. The static tension component is a vector along the tether direction, and the low-frequency oscillation tension component is a vector perpendicular to the tether direction, and the two satisfy a linear superposition relationship with the original inertial tether tension vector; It should be noted that the duration of the "preset sliding window" involved in the preceding and following steps in this embodiment is uniformly set to 2 seconds. This value is based on the following: the inherent period of load swing in a multi-aircraft cooperative hoisting system is typically between 0.5 and 3 seconds (depending on the length of the hoisting rope and the load mass). A window duration of 2 seconds is approximately 1 to 4 times the typical value of the inherent period of load swing. This allows for the inclusion of tension change information within at least one complete swing cycle to extract stable statistical features, while avoiding excessive smoothing of mid-to-high frequency transient signals caused by propulsion system anomalies due to an excessively long window. Simultaneously, this window duration matches the typical sampling frequency of each aircraft's onboard sensors (100Hz is exemplarily used in this embodiment), enabling feature updates to be completed within 200 sampling points, thus balancing real-time feature calculation and statistical stability. In practical applications, the window duration can be adaptively adjusted according to the specific hoisting rope length, load mass, and sensor sampling frequency under the guidance of the above principles. The 2-second value used in this embodiment is an exemplary preferred value and not the only limitation.

[0026] S27. Perform high-pass filtering on the residual tension components of each aircraft to obtain the mid-to-high frequency abrupt tension components of each aircraft; the cutoff frequency of the high-pass filter is higher than the load swing frequency band.

[0027] It should be noted that although the residual tension component was obtained by subtracting the static tension component and the low-frequency oscillation tension component from the inertial tether tension vector in step S26, in actual engineering applications, the zero-phase adaptive bandpass filter used in step S25 is not an ideal filter. Its transition band has a certain width of frequency leakage, which may cause some of the energy in the low-frequency oscillation tension component to leak into the residual tension component. In addition, high-frequency interference sources caused by non-propulsion system anomalies such as aircraft body vibration, aerodynamic disturbance, and force sensor quantization noise may also couple into the residual signal. Therefore, in this embodiment, step S27 further performs high-pass filtering on the residual tension component with a cutoff frequency higher than the load swing frequency band. The purpose is to: (1) filter out the residual swing component introduced by the low-frequency leakage of the bandpass filter, and ensure that the mid-to-high frequency sudden tension component and the load swing component are strictly separated in the frequency domain; (2) suppress the high-frequency measurement noise caused by non-propulsion system anomalies, and avoid misjudging aerodynamic or vibration interference as propulsion system anomalies, thereby improving the signal purity and diagnostic accuracy of propulsion system anomaly analysis.

[0028] It should be further explained that in step S2 of this embodiment, the cable tension vector is decomposed into low-frequency oscillating tension components and mid-to-high-frequency abrupt tension components based on frequency domain characteristics. Essentially, this utilizes the inherent differences in frequency band distribution between load oscillation and propulsion system anomalies to achieve preliminary extraction and separation of the two types of disturbance components at the signal level. The anomaly analysis in step S3 does not simply reuse the frequency domain decomposition result. Instead, it further integrates multi-dimensional information such as propulsion system operating data (current, speed sequence) and load state data (swing angle, angular velocity sequence), constructing multi-dimensional feature vectors and inputting them into a neural network for quantitative diagnosis. Therefore, steps S2 and S3 of this embodiment constitute a progressive logical relationship of feature extraction and comprehensive judgment, which is well-known in the art.

[0029] In a specific embodiment, such as Figure 2 As shown, based on the propulsion system operating data and mid-to-high frequency abrupt tension components, the abnormal conditions of the propulsion systems of each aircraft are analyzed; specifically including: S31. Extract the real-time current sequence and real-time speed sequence of each aircraft thruster within a preset sliding window from the propulsion system operation data; calculate the corresponding thruster current fluctuation rate and speed fluctuation rate based on the real-time current sequence and real-time speed sequence of each aircraft thruster within the preset sliding window. The standard deviation of the corresponding real-time current sequence of the thruster is divided by the sum of the average value of the corresponding thruster current sequence and a preset positive number to obtain the current fluctuation rate of the corresponding thruster. The standard deviation of the corresponding real-time speed sequence of the thruster is divided by the sum of the average value of the corresponding real-time speed sequence of the thruster and a preset positive number to obtain the speed fluctuation rate of the corresponding thruster. It should be noted that the parameters of preset positive number and preset sliding window are involved multiple times in the following steps. For ease of understanding and implementation by those skilled in the art, this embodiment provides the following unified explanation: The preset positive number is used to prevent the denominator from reaching zero; in this embodiment, it is exemplarily set to 0.001, which is much smaller than the effective dimensions of each signal and will not have a substantial impact on the calculation results. The duration of the preset sliding window is used to capture the real-time data stream to calculate local statistical characteristics; in this embodiment, the sliding window duration in each step is exemplarily set to 2 seconds. It should be noted that in practical applications, the above parameters can be adaptively adjusted according to the sensor sampling frequency, communication delay, and inherent period of load swing of the specific aircraft. The values ​​used in this embodiment are merely exemplary preferred values ​​and are not the only limitations.

[0030] S32. Based on the mid-to-high frequency abrupt tension components of each aircraft, calculate the proportion of mid-to-high frequency abrupt tension energy and the peak factor of mid-to-high frequency abrupt tension for each aircraft. The square integral of the mid-to-high frequency abrupt tension component mode of the corresponding aircraft within the preset sliding window is divided by the sum of the square integrals of the low-frequency oscillation tension component mode and the mid-to-high frequency abrupt tension component mode of the corresponding aircraft within the preset sliding window to obtain the proportion of mid-to-high frequency abrupt tension energy. The mid-to-high frequency abrupt tension component modulus of the corresponding aircraft within the preset sliding window is divided by the root mean square value of the mid-to-high frequency abrupt tension component modulus of the corresponding aircraft within the preset sliding window to obtain the mid-to-high frequency abrupt tension peak factor. S33. Combine the corresponding thruster current fluctuation rate, corresponding thruster speed fluctuation rate, mid-to-high frequency sudden tension energy ratio, and mid-to-high frequency sudden tension peak factor of the corresponding aircraft into the propulsion anomaly feature vector of the corresponding aircraft, and input it into the pre-trained propulsion anomaly analysis model to obtain the propulsion system anomaly of the corresponding aircraft; the propulsion anomaly analysis model adopts the sigmoid output function, and the value range of the propulsion system anomaly is 0 to 1.

[0031] It should be noted that all training data in this embodiment comes from a multi-aircraft collaborative hoisting semi-physical simulation experimental platform. This platform includes 4 to 8 real multi-rotor aircraft, actual hoisting ropes and loads, a real-time control system, and a fault injection module. The fault injection module can apply different degrees of thrust loss (ranging from 0% to 100%, with a step size of 5%) to the propulsion system and different degrees of stiffness reduction (ranging from 0% to 100%, with a step size of 5%) to the hoisting rope connection, covering the full fault spectrum from normal to complete failure, and collecting a total of 2000 sets of valid experimental data.

[0032] For example, the process of building the anomaly analysis model in this embodiment is as follows: Acquire hardware-in-the-loop simulation data under various historical propulsion system failure conditions. Each set of experimental data includes propulsion system operation data and corresponding mid-to-high frequency abrupt tension component data. In each set of experimental data, the thruster current fluctuation rate, thruster speed fluctuation rate, proportion of mid-to-high frequency sudden tension energy, and peak factor of mid-to-high frequency sudden tension are calculated, and these four indicators are combined into a historical thrust anomaly feature vector; the propulsion system anomaly training label for each set of experimental data is obtained, and the propulsion system anomaly training label is the actual thrust loss percentage, wherein the actual thrust loss percentage is equal to the difference between the maximum thrust under normal conditions and the actual remaining thrust under the current conditions divided by the maximum thrust under normal conditions; The historical propulsion anomaly feature vectors of all experimental data and the corresponding propulsion system anomaly training labels are used to form a propulsion anomaly dataset. The propulsion anomaly dataset is then divided into a propulsion anomaly training set and a propulsion anomaly verification set according to a preset ratio of 8:2. A feedforward neural network model is constructed, comprising an input layer, a first hidden layer, a second hidden layer, and an output layer. The input layer has 4 units, corresponding to the thruster current fluctuation rate, thruster speed fluctuation rate, mid-to-high frequency sudden tension energy ratio, and mid-to-high frequency sudden tension peak factor. The first hidden layer has 16 units with ReLU activation function, the second hidden layer has 8 units with ReLU activation function, and the output layer has 1 unit with sigmoid activation function, outputting abnormal conditions of the propulsion system. The feedforward neural network model was trained on the propulsion anomaly training set. The training process used the Adam optimizer with a learning rate of 0.001, a batch size of 32, a maximum training epoch of 200, and a mean squared error loss function. During training, L2 regularization was applied to the first and second hidden layers with a regularization coefficient of 0.0001. Dropout layers were added after the first and second hidden layers with a dropout ratio of 0.2. An early stopping strategy was adopted, and training was stopped when the root mean square error on the propulsion anomaly validation set no longer decreased for 20 consecutive epochs. The model parameters with the smallest root mean square error on the validation set were saved. The trained model is validated by advancing an anomaly validation set, and the root mean square error and coefficient of determination on the validation set are calculated. When the root mean square error of the validation set is less than or equal to the preset error threshold of 0.05, the current model is output as the advancing anomaly analysis model.

[0033] In a specific embodiment, such as Figure 2 As shown, based on the load status data and low-frequency oscillation tension components, the abnormalities in the sling connections of each aircraft are analyzed; including the following specific steps: S34. Extract the load swing angle signal sequence and load swing angular velocity sequence within the preset sliding window from the load state data; based on the load swing angle signal sequence, load swing angular velocity sequence, and low-frequency oscillation tension components of each aircraft, analyze the correlation coefficient between the load swing angle and the low-frequency oscillation tension component modulus, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy of the corresponding aircraft. Pearson correlation calculations are performed on the load swing angle signal sequence within the preset sliding window and the corresponding low-frequency oscillation tension component mode sequence of the aircraft within the preset sliding window to obtain the correlation coefficient between the load swing angle and the low-frequency oscillation tension component mode. The natural logarithmic value of the ratio of consecutive peak values ​​of the low-frequency oscillation tension component corresponding to the aircraft is used as the logarithmic decay rate of the low-frequency oscillation tension component. The "logarithmic decay rate" is a characteristic quantity used in vibration analysis to characterize the rate of decay of an oscillation signal; it is defined as the natural logarithmic value of the ratio of two adjacent peak values ​​(consecutive peak values) in the same direction of the oscillation signal. A larger logarithmic decay rate indicates faster oscillation decay, greater system damping, and healthier damping characteristics in the suspension cable connection state. Conversely, a smaller logarithmic decay rate indicates slower oscillation decay, less system damping, and potential abnormalities such as decreased stiffness or damping degradation in the suspension cable connection state. In practical engineering, to reduce the interference of single peak measurement noise on the calculation results, the average of the logarithms of multiple adjacent peak values ​​can be used as the final logarithmic decay rate output. In this embodiment, the average of the logarithmic values ​​of the ratio of three consecutive peak values ​​is used as an example.

[0034] Divide the absolute value of the phase difference between the load swing angle signal and the low-frequency oscillation tension component of the corresponding aircraft by π to obtain the phase difference normalized value. The ratio of low-frequency oscillation energy to load swing kinetic energy is obtained by dividing the square integral of the low-frequency oscillation tension component mode of the corresponding aircraft within the preset sliding window by the sum of the square integral of the load swing angular velocity mode within the preset sliding window and a preset positive number. S35. The correlation coefficient between the load swing angle and the low-frequency oscillation tension component mode, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy of the corresponding aircraft are combined into the abnormal feature vector of the suspension rope connection of the corresponding aircraft. The vector is then input into the pre-trained suspension rope connection anomaly analysis model to obtain the abnormal suspension rope connection status of the corresponding aircraft. The suspension rope connection anomaly analysis model adopts the sigmoid output function, and the value range of the abnormal suspension rope connection status is from 0 to 1.

[0035] For example, the construction process of the suspension rope connection anomaly analysis model in this embodiment is as follows: Acquire hardware-in-the-loop simulation data under multiple historical sets of different suspension rope connection conditions. Each set of experimental data includes load state data and corresponding low-frequency oscillation tension component data. In each set of experimental data, the correlation coefficient between the load swing angle and the low-frequency oscillation tension component modulus, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy were calculated. These four indicators were combined into a historical suspension rope connection anomaly feature vector. The suspension rope connection anomaly training label for each set of experimental data was obtained. The suspension rope connection anomaly training label is the percentage decrease in actual suspension rope connection stiffness. The percentage decrease in actual suspension rope connection stiffness is equal to the difference between the stiffness under normal connection state and the actual stiffness under the current state divided by the stiffness under normal connection state. The historical rope connection anomaly feature vectors of all experimental data and the corresponding rope connection anomaly training labels are used to form a rope connection anomaly dataset. The rope connection anomaly dataset is then divided into a rope connection anomaly training set and a rope connection anomaly verification set according to a preset ratio of 8:2. A feedforward neural network model is constructed, which includes an input layer, a first hidden layer, a second hidden layer, and an output layer. The input layer has 4 units, corresponding to the four abnormal rope connection features. The first hidden layer has 16 units and the activation function is ReLU. The second hidden layer has 8 units and the activation function is ReLU. The output layer has 1 unit and the activation function is sigmoid, outputting the abnormal rope connection situation. The feedforward neural network model was trained on the training set of suspension rope connection anomalies. The training process used the Adam optimizer with a learning rate of 0.001, a batch size of 32, a maximum number of training epochs of 200, and the mean squared error loss function. During training, L2 regularization was applied to the first and second hidden layers with a regularization coefficient of 0.0001. Dropout layers were added after the first and second hidden layers respectively, with a dropout ratio of 0.2. An early stopping strategy was adopted, and training was stopped when the root mean square error on the suspension rope connection anomaly validation set no longer decreased for 20 consecutive epochs. The model parameters with the smallest root mean square error on the validation set were saved. The trained model is validated using a suspension rope connection anomaly validation set, and the root mean square error and coefficient of determination on the validation set are calculated. When the root mean square error of the validation set is less than or equal to the preset error threshold of 0.05, the current model is output as the suspension rope connection anomaly calculation model.

[0036] In a specific embodiment, such as Figure 2 As shown, and combined with formation topology data, distributed dynamic reconstruction is performed to obtain the reconstructed aircraft formation configuration; including the following specific contents: S36. Calculate the cooperative transport health of the corresponding aircraft based on the abnormal conditions of the aircraft's propulsion system and the abnormal conditions of the sling connection. The weighted sum of the abnormal conditions of the corresponding aircraft's propulsion system and the abnormal conditions of the corresponding aircraft's sling connection is obtained by subtracting the weighted sum from the value. S37. Obtain aircraft formation topology data. Under the distributed communication topology, each aircraft exchanges its cooperative transport health with its neighboring aircraft. Minimize the local cost function of each aircraft through a distributed optimization algorithm to obtain the reconstructed aircraft formation configuration. The local cost function includes the sum of squares of the expected relative position errors of adjacent aircraft and the sum of squares of the expected position deviation adjusted according to the cooperative transport health of each aircraft.

[0037] For example, in this embodiment, the reconstructed aircraft formation configuration is obtained by minimizing the local cost function of each aircraft using a distributed optimization algorithm. The specific steps are as follows: A1. Obtain the formation topology data of the aircraft and determine the set of adjacent aircraft and the adjacency weight of each aircraft; initialize the target position of each aircraft to the current real-time position; determine the expected relative position vector between each adjacent aircraft according to the formation task; obtain the cooperative transport health of each aircraft and send the cooperative transport health of each aircraft to the adjacent aircraft through the inter-aircraft communication network. A2. Calculate the expected health adjustment position for each aircraft based on its cooperative transport health status. The formula for calculating the expected health adjustment position is as follows: ; In the formula, The desired position is adjusted according to the health of the corresponding aircraft, where poi is the original desired position of the corresponding aircraft, hi is the cooperative transport health of the corresponding aircraft, and ri is the direction vector from the formation center to the outside of the corresponding aircraft. A3. Define a local cost function for each aircraft. The calculation formula for the local cost function is as follows: ; In the formula, Ji is the local cost function corresponding to the i-th aircraft, aij is the adjacency weight between the corresponding aircraft and the j-th aircraft, bi is the position adjustment weight, which controls the traction strength of health on the target position, pi and pj are the current target positions of the i-th aircraft and the j-th aircraft, respectively, dij is the expected relative position vector between the i-th aircraft and the j-th aircraft, and Ni is the set of neighboring aircraft of the i-th aircraft; For norm squaring operations; A4. Each aircraft uses its current target position obtained from exchanging information with neighboring aircraft to calculate the gradient of its local cost function with respect to its own target position; the formula for calculating the gradient is as follows: ; In the formula, Let be the gradient of the local cost function of the i-th aircraft with respect to its own target position; A5. Each aircraft updates its target position using gradient descent based on the gradient. The updated target position is equal to the target position before the update minus the difference between the step size and the gradient. A6. Each aircraft sends the updated target position to all neighboring aircraft and receives the updated target position from neighboring aircraft. A7. Repeat steps A4-A6 until the convergence condition is met. The convergence condition is that the maximum value of the change in the target position of all aircraft in two adjacent iterations is less than the preset convergence threshold, or the preset maximum number of iterations is reached. After the convergence condition is met, stop the iteration and output the final target position of each aircraft as the reconstructed aircraft formation configuration.

[0038] In a specific embodiment, such as Figure 3 As shown, the dynamic thrust margin required to suppress load swing during the reconfiguration transition of each aircraft is predicted; including the following specific steps: S41. Based on the reconstructed aircraft formation configuration, calculate the desired cable direction unit vector from each aircraft to the load; divide the difference between the load position coordinates and the corresponding reconstructed aircraft position coordinates by the modulus of the difference between the load position coordinates and the corresponding reconstructed aircraft position coordinates to obtain the desired cable direction unit vector. S42. Based on the low-frequency oscillation tension components of each aircraft, estimate the maximum additional thrust amplitude required for each aircraft to suppress load sway during the reconfiguration transition. The maximum additional thrust amplitude is the maximum value of the low-frequency oscillation tension component magnitude of the corresponding aircraft within a preset sliding window. The preset sliding window refers to a time interval of length T prior to the current moment, and the maximum additional thrust amplitude is calculated in real-time based on the low-frequency oscillation tension components within the current real-time sliding window. The implicit engineering assumption is that the dominant frequency and amplitude envelope of the load sway will not undergo drastic changes during the short reconfiguration transition period; therefore, the historical maximum value within the current sliding window can be used as an effective conservative estimate of the additional thrust required to suppress sway in the initial stage of reconfiguration. If the load state changes significantly during actual flight, the sliding window in this embodiment can be updated in real-time, possessing adaptive adjustment capabilities.

[0039] It should be noted that the method described above for estimating the maximum additional thrust amplitude based on the maximum value of the low-frequency oscillation tension component modulus within a preset sliding window is effective only if the natural frequency and amplitude envelope of the load swing do not undergo drastic changes during the short period of the reconfiguration transition, i.e., the load swing exhibits slow time-varying characteristics. This premise can be met under most conventional and stable hoisting conditions. In this case, using the historical maximum value within the current sliding window as a conservative estimate of the additional thrust required to suppress the swing in the early stage of reconfiguration is reasonable and effective.

[0040] To further enhance the robustness of this method under complex operating conditions (such as sudden wind disturbances or abrupt changes in load mass), this embodiment also provides the following adaptive adjustment mechanism: Real-time monitoring of the instantaneous value of the low-frequency oscillation tension component magnitude; if this instantaneous value exceeds the product of the maximum value within the current sliding window and a preset proportional coefficient (exemplarily 1.2 in this embodiment), it is determined that a significant change in the load oscillation state has occurred. At this point, the maximum value of the sliding window is immediately updated to the current instantaneous value, and the maximum additional thrust amplitude is recalculated without waiting for the window to naturally slide and update. Through this adaptive mechanism, a rapid response can be achieved when the load state changes abruptly, and the thrust margin estimate can be adjusted in a timely manner, ensuring the conservatism and safety of thrust allocation during the reconfiguration transition.

[0041] S43. Based on the reconstructed aircraft formation configuration, calculate the reference acceleration of each aircraft from its current position to its reconstructed position, and multiply the mass of the corresponding aircraft by the reference acceleration to obtain the maneuvering acceleration thrust required by the corresponding aircraft; take the sum of the static hovering thrust, the maximum additional thrust amplitude, and the maneuvering acceleration thrust required by the corresponding aircraft as the dynamic thrust requirement of the corresponding aircraft. S44. Calculate the remaining thrust of each aircraft based on the abnormal propulsion system conditions of each aircraft; multiply the difference between value one and the abnormal propulsion system condition of the corresponding aircraft by the maximum thrust of the corresponding aircraft to obtain the remaining thrust of the corresponding aircraft. S45. Divide the difference between the remaining thrust of each aircraft and the dynamic thrust requirement of the corresponding aircraft by the remaining thrust of the corresponding aircraft to obtain the dynamic thrust margin of each aircraft. If the division result is less than 0, the dynamic thrust margin is 0; if the division result is greater than 1, the dynamic thrust margin is 1; otherwise, the original value is retained.

[0042] In a specific embodiment, such as Figure 3 As shown, the stability of the sling connection load of each aircraft after reconstruction is predicted; including the following specific steps: S46. Extract the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and reconstructed configuration geometric uniformity features of each aircraft; wherein, the neighborhood average dynamic thrust margin is the arithmetic mean of the dynamic thrust margins of each aircraft adjacent to the corresponding aircraft; the reconstructed configuration geometric uniformity features are the variance of the angle between the unit vector of the desired cable direction of the corresponding aircraft and the unit vector of the desired cable direction of each adjacent aircraft; wherein, if the number of adjacent aircraft of the corresponding aircraft is less than 2, the angle variance cannot be effectively calculated, and in this case, the reconstructed configuration geometric uniformity features of the aircraft are taken as a preset neutral value ( In this embodiment, 0.5 is used as an example. This value corresponds to a neutral state where geometric uniformity has no significant bias effect on the stability of the suspension rope connection load. Under the sigmoid normalization output framework, 0.5 represents a neutral level where the input feature neither positively promotes nor negatively inhibits the output, avoiding abnormal jumps in stability prediction due to insufficient neighbor numbers. At the same time, during implementation, it is also possible to temporarily remove this feature from the stability feature vector (i.e., not participate in the stability prediction at this moment), and only use the other three features to input the suspension rope connection load stability prediction model. Both processing methods are within the protection scope of this invention.

[0043] S47. Combine the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and geometric uniformity features of the reconstructed configuration of each aircraft into a stability feature vector of the corresponding aircraft, and input it into a pre-trained suspension cable connection load stability prediction model to obtain the suspension cable connection load stability of the corresponding aircraft; the suspension cable connection load stability prediction model adopts the sigmoid output function, and the value range of the suspension cable connection load stability is 0 to 1.

[0044] For example, the construction process of the load stability prediction model for the suspension rope connection in this embodiment is as follows: Acquire hardware-in-the-loop simulation data under multiple historical sets of different reconfiguration configurations and different dynamic thrust margins. Each set of experimental data includes the reconfigured aircraft formation configuration, low-frequency oscillation tension components, and corresponding load motion state data. In each set of experimental data, the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and geometric uniformity characteristics of the reconstructed configuration are calculated, and these four indicators are combined into a historical stability feature vector. The suspension cable connection load stability training label for each set of experimental data is obtained. The suspension cable connection load stability training label is the normalized value of the load swing decay time constant. The load swing decay time constant is the time required for the amplitude to decay to 1 / e of the initial amplitude during the free decay process of the load swing angle. The reciprocal of the load swing decay time constant is normalized. Specifically, the minimum and maximum values ​​of the load swing decay time constant under the action of the corresponding aircraft in each set of experiments are used as the lower and upper bounds of normalization, respectively. The current load swing decay time constant is mapped to the range of 0 to 1 as the training label. The historical stability feature vectors of all experimental data and the corresponding suspension rope connection load stability training labels are used to form a stability dataset. The stability dataset is then divided into a stability training set and a stability validation set according to a preset ratio of 8:2. A feedforward neural network model is constructed, which includes an input layer, a first hidden layer, a second hidden layer, and an output layer. The input layer has 4 units, corresponding to the four stability features. The first hidden layer has 16 units and the activation function is ReLU. The second hidden layer has 8 units and the activation function is ReLU. The output layer has 1 unit and the activation function is sigmoid, which outputs the stability of the load connected by the suspension rope.

[0045] The feedforward neural network model was trained on a stability training set using the Adam optimizer with a learning rate of 0.001, a batch size of 32, a maximum training epoch of 200, and a mean squared error loss function. During training, L2 regularization was applied to the first and second hidden layers with a regularization coefficient of 0.0001. Dropout layers were added after the first and second hidden layers with a dropout ratio of 0.2. An early stopping strategy was adopted, stopping training when the root mean square error on the stability validation set no longer decreased for 20 consecutive epochs, and saving the model parameters with the smallest root mean square error on the validation set. The trained model is validated using a stability validation set, and the root mean square error and coefficient of determination are calculated on the validation set. When the root mean square error of the validation set is less than or equal to the preset error threshold of 0.05, the current model is output as the stability prediction model for the suspension rope connection load.

[0046] The technical logic relationship between steps S4 and S5 in this embodiment is further clarified here: Although both the "dynamic thrust margin" calculated in step S4 and the "rope connection load stability" predicted in step S4 are used as input parameters for generating fault-tolerant control commands in step S5, their functional roles in control command generation are different. The "dynamic thrust margin" reflects the sufficiency of each aircraft's current remaining thrust capacity relative to the thrust required for reconfiguration, directly determining whether each aircraft can safely perform reconfiguration maneuvers. Therefore, it is used to calculate thrust allocation weights to ensure that the thrust allocation does not exceed the physical limits of each aircraft. The "rope connection load stability," on the other hand, reflects the ability of each aircraft to maintain load attitude stability via ropes after reconfiguration, determining the speed of the reconfiguration transition process. Therefore, it is used to generate a transition speed adjustment factor for a smooth reference trajectory. Both, from the two orthogonal dimensions of "sufficient thrust" and "smooth reconfiguration," jointly ensure the safety of the reconfiguration process, and neither is dispensable. If thrust allocation is based solely on dynamic thrust margin without considering the stability of the tethered connection load, the reconfiguration process may result in excessive additional dynamic loads on the aircraft due to excessive load swing, leading to a deterioration of the tethered connection and potentially causing secondary instability. Conversely, if trajectory velocity is adjusted solely based on load stability without considering thrust margin, a dangerous situation may arise where thrust demand exceeds the aircraft's remaining thrust capacity. Therefore, the two outputs of step S4 constitute a dual safety guarantee for the generation of distributed fault-tolerant control commands in step S5.

[0047] In one specific embodiment, based on the predicted stability of the rigging connection load of each aircraft after reconstruction, distributed fault-tolerant control commands are generated and issued to each aircraft; including the following specific steps: S51. Obtain the load stability of the sling connection of each aircraft; S52. Based on the rigging connection load stability and dynamic thrust margin of each aircraft, calculate the control gain scaling factor and thrust allocation weight of each aircraft; wherein, the control gain scaling factor is equal to the rigging connection load stability of the corresponding aircraft multiplied by the preset benchmark control gain scaling factor; obtain the product of the rigging connection load stability and the dynamic thrust margin of the corresponding aircraft; based on the maximum and minimum values ​​of the corresponding product results of all aircraft, perform extreme value normalization processing on the product results of each aircraft to obtain the weight coefficient of each aircraft in the total thrust allocation, which is used as the thrust allocation weight of each aircraft; S53. Based on the reconstructed aircraft formation configuration, using the stability of the rigging connection load of each aircraft as the transition speed adjustment factor, generate a smooth reference trajectory for each aircraft from its current position to the reconstructed target position; the transition speed of the smooth reference trajectory is positively correlated with the stability of the rigging connection load of the corresponding aircraft, that is, the higher the stability, the faster the transition speed, and the lower the stability, the slower the transition speed. It should be noted that generating a smooth reference trajectory for each aircraft from its current position to the reconstructed target position specifically includes the following steps: Obtain the current position coordinates and reconstructed target position coordinates of each aircraft, and calculate the position difference vector of each aircraft; the position difference vector is equal to the reconstructed target position coordinates minus the current position coordinates of the corresponding aircraft. Based on the load stability of the sling connection of each aircraft, the transition time length of each aircraft is calculated; the transition time length is equal to the preset basic transition time divided by the sum of the load stability of the sling connection of the corresponding aircraft and a preset positive number; wherein, when the transition time length is less than the preset minimum transition time, the transition time length is taken as the preset minimum transition time; when the transition time length is greater than the preset maximum transition time, the transition time length is taken as the preset maximum transition time, thus obtaining the transition time length of each aircraft; The greater the stability of the load connected by the sling, the shorter the transition time and the faster the transition speed; conversely, the smaller the stability of the load connected by the sling, the longer the transition time and the slower the transition speed. The transition time and the stability of the load connected by the sling are in a continuous monotonic mapping relationship, and no threshold switching is set. This embodiment calculates the transition time through a continuous monotonic mapping, avoiding sudden changes in the reference trajectory velocity caused by threshold switching, and improving the stability and smoothness of the multi-aircraft cooperative hoisting system during the fault-tolerant reconfiguration process. Starting from the current position of each aircraft, ending at the reconstructed target position of the corresponding aircraft, and taking the transition time of the corresponding aircraft as the total duration, a fifth-order polynomial interpolation is used to generate the reference position trajectory, reference velocity trajectory, and reference acceleration trajectory of each aircraft; the fifth-order polynomial interpolation satisfies the position boundary conditions, velocity boundary conditions, and acceleration boundary conditions of the starting and ending points, and the velocity and acceleration of the starting and ending points are both zero; The reference position trajectory, reference velocity trajectory, and reference acceleration trajectory of each aircraft are output as smooth reference trajectories.

[0048] S54. Combine the smoothed reference trajectory, control gain scaling factor, and thrust distribution weight of each aircraft into a distributed fault-tolerant control command, and send it to the corresponding aircraft through the inter-aircraft communication network; each aircraft adjusts the control gain and thrust distribution of its own position controller and attitude controller according to the received distributed fault-tolerant control command to complete the fault-tolerant control.

[0049] It should be noted that the values ​​of the setting parameters (such as weights and thresholds) not specifically described in this embodiment are determined as follows: Propulsion system operation data, cable tension vector data, load status data, real-time aircraft attitude data, and formation topology data are acquired synchronously during the actual operation of the multi-aircraft collaborative hoisting system (2000 sets). The abnormal conditions of the propulsion system, the abnormal conditions of the hoisting cable connection, the dynamic thrust margin, and the load stability of the hoisting cable connection are calculated for each aircraft according to the aforementioned steps, and the collaborative transport health is calculated. Simultaneously, the judgment results of whether any abnormal events such as thruster failure, hoisting cable connection failure, or load instability actually occurred under the corresponding operating conditions are obtained for 2000 sets. The calculation results of various indicators and the corresponding abnormal event judgment results during the 2000 sets of operation are imported into the MATLAB Curve Fitting Toolbox fitting software. Binary logistic regression is used to perform maximum likelihood estimation fitting for each individual indicator, with the highest determination coefficient on the validation set as the objective to determine the values ​​of each setting parameter (such as weights and thresholds).

[0050] Example 2

[0051] like Figure 4 As shown, this embodiment provides a real-time fault control reconfiguration system based on a redundant propulsion system, including: The data acquisition module is used to synchronously collect propulsion system operation data, cable tension vector data, and load status data of each aircraft through distributed sensing units deployed on each collaborative hoisting aircraft; at the same time, it acquires aircraft formation topology data and real-time attitude data of the aircraft. The tension decomposition module is used to combine cable tension vector data and aircraft real-time attitude data to decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium- and high-frequency abrupt tension components caused by propulsion system anomalies. The formation reconfiguration module is used to analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; to analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and to perform distributed dynamic reconfiguration based on the formation topology data to obtain the reconfigured aircraft formation configuration. The load prediction module is used to predict the dynamic thrust margin required by each aircraft to suppress load swing during the reconfiguration transition based on the reconfigured aircraft formation configuration and the low-frequency oscillation tension component, and to predict the load stability of the suspender cable connection of each aircraft after reconfiguration. The control module generates and issues distributed fault-tolerant control commands to each aircraft based on the predicted stability of the rigging connection load of each aircraft after reconstruction.

[0052] The steps for implementing the corresponding functions of each parameter and each unit module in the real-time fault control reconfiguration system based on the redundant propulsion system of the present invention can be referred to the parameters and steps in the embodiments of the real-time fault control reconfiguration method based on the redundant propulsion system above, and will not be repeated here.

[0053] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments for IoT devices and media are relatively simple in description because they are fundamentally similar to the method embodiments; relevant parts can be referred to the descriptions in the method embodiments.

[0054] The systems, media, and methods provided in the embodiments of the present invention are in one-to-one correspondence. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0058] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0059] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0060] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0061] It should also be noted that the terms include, encompass, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, the presence of additional identical elements in the process, method, article, or apparatus that includes elements is not excluded.

[0062] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A real-time fault control reconfiguration method based on a redundant propulsion system, characterized in that, Includes the following steps: Step S1: Through the distributed sensing units deployed on each collaborative hoisting aircraft, synchronously collect propulsion system operation data, cable tension vector data, and load status data of each aircraft; at the same time, acquire aircraft formation topology data and real-time attitude data of the aircraft. Step S2: Combining cable tension vector data and aircraft real-time attitude data, decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium-to-high frequency abrupt tension components caused by propulsion system anomalies. Step S3: Analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and combine the formation topology data to perform distributed dynamic reconstruction to obtain the reconstructed aircraft formation configuration. Step S4: Based on the reconstructed aircraft formation configuration and combined with the low-frequency oscillation tension component, predict the dynamic thrust margin required for each aircraft to suppress load swing during the reconstruction transition, and predict the load stability of the suspender cable connection of each aircraft after reconstruction. Step S5: Based on the predicted stability results of the rigging connection load of each aircraft after reconstruction, generate and issue distributed fault-tolerant control commands to each aircraft.

2. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 1, characterized in that, The process of decomposing the cable tension vector into a low-frequency oscillating tension component caused by load oscillation and a mid-to-high frequency abrupt tension component caused by propulsion system anomalies includes the following steps: S21. Obtain the attitude quaternion of each aircraft in the real-time attitude data of the aircraft, and at the same time obtain the cable tension vector directly measured by the force sensor on the corresponding aircraft in the sensor coordinate system; generate the rotation matrix of each aircraft according to the attitude quaternion; multiply the rotation matrix by the cable tension vector in the sensor coordinate system to obtain the inertial frame cable tension vector of each aircraft. S22. Obtain the aircraft position coordinates from the real-time attitude data and the load position coordinates from the load status data, and calculate the unit vector of the cable direction from each aircraft to the load. S23. Decompose the inertial tether tension vector of each aircraft into the component along the tether direction and the component perpendicular to the tether direction of each aircraft. S24. Obtain the load swing angle from the load status data, perform a short-time Fourier transform on the load swing angle, extract the peak frequency at each moment in the time spectrum as the instantaneous frequency of the load swing, and use the instantaneous frequency of the load swing as the main frequency of the load swing. S25. Using the load swing main frequency as the center frequency, set the bandwidth, and perform zero-phase adaptive bandpass filtering on the vertical cable direction component of each aircraft to obtain the low-frequency oscillation tension component of each aircraft in the inertial frame that is in the same direction as the vertical cable direction component. S26. Subtract the static tension component and the low-frequency oscillation tension component from the inertial tether tension vector of each aircraft to obtain the residual tension component of each aircraft; wherein, the static tension component is the low-pass average value of the component along the tether direction of the corresponding aircraft within a preset sliding window. S27. Perform high-pass filtering on the residual tension components of each aircraft to obtain the mid-to-high frequency abrupt tension components of each aircraft; the cutoff frequency of the high-pass filter is higher than the load swing frequency band.

3. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 2, characterized in that, The analysis of propulsion system anomalies for each aircraft is based on propulsion system operating data and mid-to-high frequency abrupt tension components; specifically including: S31. Extract the real-time current sequence and real-time speed sequence of each aircraft thruster within a preset sliding window from the propulsion system operation data; calculate the corresponding thruster current fluctuation rate and speed fluctuation rate based on the real-time current sequence and real-time speed sequence of each aircraft thruster within the preset sliding window. S32. Based on the mid-to-high frequency abrupt tension components of each aircraft, calculate the proportion of mid-to-high frequency abrupt tension energy and the peak factor of mid-to-high frequency abrupt tension for each aircraft. S33. Combine the corresponding thruster current fluctuation rate, corresponding thruster speed fluctuation rate, mid-to-high frequency sudden tension energy ratio, and mid-to-high frequency sudden tension peak factor of the corresponding aircraft into the propulsion anomaly feature vector of the corresponding aircraft, and input it into the pre-trained propulsion anomaly analysis model to obtain the propulsion system anomaly of the corresponding aircraft.

4. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 3, characterized in that, The abnormal connection of the suspension ropes of each aircraft is analyzed based on the load status data and the low-frequency oscillation tension component. The specific steps include the following: S34. Extract the load swing angle signal sequence and load swing angular velocity sequence within the preset sliding window from the load state data; based on the load swing angle signal sequence, load swing angular velocity sequence, and low-frequency oscillation tension components of each aircraft, analyze the correlation coefficient between the load swing angle and the low-frequency oscillation tension component modulus, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy of the corresponding aircraft. S35. Combine the correlation coefficient between the load swing angle and the low-frequency oscillation tension component mode of the corresponding aircraft, the logarithmic decay rate of the low-frequency oscillation tension component, the normalized value of the phase difference between the load swing angle and the low-frequency oscillation tension component, and the ratio of low-frequency oscillation energy to load swing kinetic energy into the abnormal feature vector of the suspension rope connection of the corresponding aircraft. Input the vector into the pre-trained suspension rope connection anomaly analysis model to obtain the abnormal situation of the suspension rope connection of the corresponding aircraft.

5. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 4, characterized in that, The above, combined with the formation topology data, is used to perform distributed dynamic reconstruction to obtain the reconstructed aircraft formation configuration; Includes the following specific content: S36. Calculate the cooperative transport health of the corresponding aircraft based on the abnormal conditions of the aircraft's propulsion system and the abnormal conditions of the sling connection. The weighted sum of the abnormal conditions of the corresponding aircraft's propulsion system and the abnormal conditions of the corresponding aircraft's sling connection is obtained by subtracting the weighted sum from the value. S37. Obtain aircraft formation topology data. Under the distributed communication topology, each aircraft exchanges its cooperative transport health with its neighboring aircraft. Minimize the local cost function of each aircraft through a distributed optimization algorithm to obtain the reconstructed aircraft formation configuration.

6. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 5, characterized in that, The predicted dynamic thrust margin required by each aircraft to suppress load swing during the reconfiguration transition; The specific steps include the following: S41. Based on the reconstructed aircraft formation configuration, calculate the desired cable direction unit vector from each aircraft to the load. S42. Based on the low-frequency oscillation tension components of each aircraft, estimate the maximum additional thrust amplitude required for each aircraft to suppress load swing during the reconfiguration transition. S43. Based on the reconstructed aircraft formation configuration, calculate the reference acceleration of each aircraft from its current position to its reconstructed position, and multiply the mass of the corresponding aircraft by the reference acceleration to obtain the maneuvering acceleration thrust required by the corresponding aircraft; take the sum of the static hovering thrust, the maximum additional thrust amplitude, and the maneuvering acceleration thrust required by the corresponding aircraft as the dynamic thrust requirement of the corresponding aircraft. S44. Calculate the remaining thrust of each aircraft based on the abnormal conditions of its propulsion system. S45. Divide the difference between the remaining thrust of each aircraft and the dynamic thrust requirement of the corresponding aircraft by the remaining thrust of the corresponding aircraft to obtain the dynamic thrust margin of each aircraft.

7. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 6, characterized in that, The text describes and predicts the load stability of the sling connections of each aircraft after reconstruction. The specific steps include the following: S46. Extract the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and reconstruct the geometric uniformity characteristics of the configuration for each aircraft. S47. Combine the dynamic thrust margin, neighborhood average dynamic thrust margin, logarithmic decay rate of low-frequency oscillation tension component, and geometric uniformity features of the reconstructed configuration of each aircraft into the stability feature vector of the corresponding aircraft, and input it into the pre-trained suspension cable connection load stability prediction model to obtain the suspension cable connection load stability of the corresponding aircraft.

8. The real-time fault control reconfiguration method based on a redundant propulsion system according to claim 7, characterized in that, Based on the predicted stability of the suspender cable connection load of each aircraft after reconstruction, distributed fault-tolerant control commands for each aircraft are generated and issued. The specific steps include the following: S51. Obtain the load stability of the rigging connections of each aircraft; S52. Based on the stability of the rigging connection load and the dynamic thrust margin of each aircraft, calculate the control gain scaling factor and thrust distribution weight of each aircraft. S53. Based on the reconstructed aircraft formation configuration, using the stability of the rigging load of each aircraft as the transition speed adjustment factor, generate a smooth reference trajectory for each aircraft from its current position to the reconstructed target position. S54. Combine the smoothed reference trajectory, control gain scaling factor, and thrust distribution weight of each aircraft into a distributed fault-tolerant control command, and send it to the corresponding aircraft through the inter-aircraft communication network; each aircraft adjusts the control gain and thrust distribution of its own position controller and attitude controller according to the received distributed fault-tolerant control command to complete the fault-tolerant control.

9. A real-time fault control and reconfiguration system based on a redundant propulsion system, implemented based on any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to synchronously collect propulsion system operation data, cable tension vector data, and load status data of each aircraft through distributed sensing units deployed on each collaborative hoisting aircraft; at the same time, it acquires aircraft formation topology data and real-time attitude data of the aircraft. The tension decomposition module is used to combine cable tension vector data and aircraft real-time attitude data to decompose the cable tension vector into low-frequency oscillating tension components caused by load swing and medium- and high-frequency abrupt tension components caused by propulsion system anomalies. The formation reconfiguration module is used to analyze the abnormal conditions of the propulsion system of each aircraft based on the propulsion system operation data and the mid-to-high frequency abrupt tension components; to analyze the abnormal conditions of the suspension cable connection of each aircraft based on the load status data and the low-frequency oscillation tension components; and to perform distributed dynamic reconfiguration based on the formation topology data to obtain the reconfigured aircraft formation configuration. The load prediction module is used to predict the dynamic thrust margin required by each aircraft to suppress load swing during the reconfiguration transition based on the reconfigured aircraft formation configuration and the low-frequency oscillation tension component, and to predict the load stability of the suspender cable connection of each aircraft after reconfiguration. The control module generates and issues distributed fault-tolerant control commands to each aircraft based on the predicted stability of the rigging connection load of each aircraft after reconstruction.

Images