A multi-propeller power cooperative control system

Abstract

The application provides a multi-screw power cooperative control system and relates to the technical field of data processing.The system is used for fitting the coupling effect difference between screws through the thrust gradient change of adjacent screws, accumulating in accordance with the physical adjacent relationship, identifying the internal structure constraint of the whole multi-screw power chain, superimposing the interaction power between different screws, calculating the non-uniform energy coupling degree between the multi-screws, obtaining the power distribution index, reversely mapping the local strength area of the screw coupling matrix, establishing the power transmission path between the screws, calculating the synchronous stability degree between the main propelling screw and the auxiliary propelling screw, obtaining the dynamic balance factor, mutually normalizing the change trend of the power distribution index and the dynamic balance factor to form a cooperative vector, defining the instantaneous power adjustment sequence of each screw, and obtaining a control instruction matrix.The application ensures the stable navigation of a fleet by controlling the screw power cooperation.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a multi-propeller power cooperative control system. Background Technology

[0002] In existing multi-ship combined navigation systems, a single main propulsion vessel typically drives multiple barges in tandem via mechanical towing or hydraulic connection. Some vessels are structurally equipped with multiple propellers to achieve thrust distribution and directional control. However, in traditional control systems, each propeller is usually controlled by an independent motor, and the power coordination between the main vessel and auxiliary vessels relies on simple speed signals or power commands, which may not be able to achieve dynamic balance control of the overall propulsive torque. When the fleet is in complex sea conditions such as cross waves, oblique winds, or shallow shoals and backflows, the force state of propellers at different positions varies greatly, which may cause problems such as local thrust overload, course deviation, or uneven energy distribution.

[0003] For example, when a combined fleet is sailing in coordination at sea, the main propulsion vessel and multiple auxiliary vessels may need to adjust their respective propeller thrusts simultaneously to maintain formation stability. However, they may lack the perception and feedback correction of real-time navigation status. When external disturbances cause a sudden change in the force on a vessel, the system may not be able to dynamically compensate among multiple propellers, resulting in situations such as steering lag, uneven structural stress, or increased rolling of the fleet. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-propeller power cooperative control system, which aims to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A multi-propeller power cooperative control system, the system comprising:

[0007] The feature module is used to acquire the operating data of each propeller, and to obtain the difference feature data by fitting the coupling effect difference between propellers through the thrust gradient change of adjacent propellers.

[0008] The coupling module is used to partition and accumulate the gradient components in the difference feature set according to the physical adjacency relationship, identify the internal structural constraints of the overall power chain of the multi-propeller, and obtain the propeller coupling matrix.

[0009] The index module is used to superimpose the interaction power between different propellers based on the propeller coupling matrix, calculate the degree of non-uniform energy coupling between multiple propellers, and obtain the power distribution index.

[0010] The correlation module is used to reverse map the local intensity region of the propeller coupling matrix according to the power distribution index, establish the power transfer path between propellers, and obtain the propulsion coordination matrix.

[0011] The factor module is used to calculate the degree of synchronization stability between the main propeller and the auxiliary propeller based on the propulsion coordination matrix, and obtain the dynamic balance factor.

[0012] The fusion module is used to complement and fuse the power distribution index and the dynamic balance factor, and normalize the changing trends of the two to form a cooperative vector, thus obtaining a multi-propeller cooperative vector group.

[0013] The instruction module is used to recursively expand the multi-propeller cooperative vector group, define the instantaneous power adjustment sequence of each propeller, and obtain the control instruction matrix.

[0014] Furthermore, the feature module includes:

[0015] The analysis unit is used to analyze the operating data of each propeller, and to simultaneously separate the speed, torque and fluid resistance to obtain the operating parameter matrix.

[0016] The gradient calculation unit is used to extract the thrust gradient of adjacent propellers based on the operating parameter matrix, calculate the thrust increment and angle change ratio in each time interval, and obtain the gradient change matrix.

[0017] The coupling analysis unit is used to establish a thrust response correlation model between propellers based on the gradient change matrix, extract the coupling difference values ​​of each adjacent propeller, and obtain the coupling difference matrix.

[0018] The feature generation unit is used to perform weighted fusion and normalization processing on the coupling difference values ​​of each propeller according to the coupling difference matrix, identify the differences in coupling effects between propellers, and obtain difference feature data.

[0019] Furthermore, the coupling module includes:

[0020] The partitioning construction unit is used to determine the adjacency relationship of each propeller in the physical layout based on the difference feature data, establish a position index, and obtain the propeller adjacency set;

[0021] The cumulative calculation unit is used to partition and accumulate the gradient components in the difference feature data according to the adjacency relationship of adjacent propellers based on the adjacent set of propellers, so as to obtain the gradient accumulation matrix.

[0022] The constraint identification unit is used to extract the transmission chain nodes between multiple propellers based on the gradient accumulation matrix, perform constraint calibration on the torque correlation between each transmission chain node, and obtain the power chain constraint table.

[0023] The matrix generation unit is used to map the constraint relationships of each partition to a unified coordinate system based on the power chain constraint table, ensuring the internal structure of the overall power chain and obtaining the propeller coupling matrix.

[0024] Furthermore, the constraint recognition unit includes:

[0025] The node extraction unit is used to identify points with significant gradient changes in multiple propellers based on the gradient accumulation matrix, and define the points as nodes of the transmission chain to obtain a node index set.

[0026] The torque calculation unit is used to calculate the torque components and direction differences between any adjacent transmission chain nodes based on the node index set, and obtain the torque correlation matrix.

[0027] The constraint modeling unit is used to construct a set of constraint equations between nodes based on the moment correlation matrix, and to normalize the set of constraint equations to obtain a constraint coefficient table.

[0028] The calibration generation unit is used to determine the force boundary conditions of each node in the overall power chain of the multi-propeller based on the constraint coefficient table, identify the torque transmission relationship between the multi-propellers, and obtain the power chain constraint table.

[0029] Furthermore, the index module includes:

[0030] The power distribution index unit is used to calculate the coupling weight of the power chain as the distance decreases along the path length based on the propeller coupling matrix, thus obtaining the topological coupling term; and to calculate the geometric equilibrium value of the power ratio of any two propellers based on the power ratio of each propeller, thus obtaining the scale ratio term.

[0031] Based on whether the included angle and rotation direction of any two propeller propulsion axes are consistent, the directional modulation of the alignment degree is calculated to obtain the directional consistency term; the magnitude amplification of the difference in the rate of change of any two propellers is calculated to obtain the dynamic difference term; the topological coupling term, scale ratio term, directional consistency term and dynamic difference term are fused together to calculate the local coupling aggregation term;

[0032] Based on the local coupling aggregation term, the scale factor of the compressed or stretched numerical range is calculated to obtain the scale shaping term; based on the dispersion of the overall propeller power ratio, the global non-uniformity modification term is calculated; the local coupling aggregation term, the scale shaping term, and the global non-uniformity modification term are fused to obtain the power distribution index.

[0033] Furthermore, the associated module includes:

[0034] The region identification unit is used to determine the local intensity regions in the propeller coupling matrix that are higher than the preset power threshold based on the power distribution index, thereby obtaining the intensity region set;

[0035] The reverse mapping unit is used to map the energy distribution nodes in each intensity region to propeller number indices based on the intensity region set, establish the energy backtracking relationship between propeller pairs, and obtain the backtracking mapping table.

[0036] The path construction unit is used to connect propeller nodes with energy backtracking relationships sequentially according to the backtracking mapping table to form a continuous power transfer link and obtain a transfer path set;

[0037] The coordination matrix unit is used to convert the energy ratio of each path segment into matrix entry weights based on the transmission path set, determine the power transmission path between propellers, and obtain the propulsion coordination matrix.

[0038] Furthermore, the cooperative matrix unit includes:

[0039] The energy allocation unit is used to quantify the energy difference between the starting and ending nodes of each path segment according to the transmission path set, establish an energy ratio table, and obtain the path energy set.

[0040] The weight correction unit is used to perform symmetric balancing based on the path energy set, and to perform bidirectional correction and normalization on the weight values ​​of the bidirectional transmission paths in the energy ratio table to generate an energy weight matrix.

[0041] The topology mapping unit is used to map the energy weight distribution to the physical connection topology between propellers based on the energy weight matrix, forming a topology association table;

[0042] The propulsion coordination matrix unit is used to incorporate the energy weights of each path segment into the matrix according to the physical adjacency relationship based on the topology association table, and to determine the power transfer path relationship between propellers, thus obtaining the propulsion coordination matrix.

[0043] Furthermore, the factor module includes:

[0044] The dynamic balance factor unit is used to calculate the instantaneous response difference between each pair of main propellers and auxiliary propellers based on the propulsion coordination matrix, and obtain the time offset term; based on the output power amplitude of each propeller, it compares the power fluctuation direction and amplitude ratio of the main propeller and auxiliary propeller in the same time period, calculates the degree of coordination amplitude between the two, and obtains the amplitude coupling term.

[0045] Based on the inertia parameters and relative spatial positions of each propeller, the difference in inertial response between the main propeller and the auxiliary propeller is measured, the degree of inertia matching is calculated, and the inertia compensation term is obtained; based on the instantaneous frequency changes of each propeller, the average operating frequency and fluctuation range of the propellers are statistically analyzed, the group frequency consistency among each propeller is calculated, and the frequency synchronization term is obtained.

[0046] By fusing the timing offset term, amplitude coupling term, inertia compensation term, and frequency synchronization term, the overall synchronization stability among multiple propellers is calculated, and the dynamic balance factor is obtained.

[0047] Furthermore, the fusion module includes:

[0048] The energy mapping unit is used to perform time-series expansion based on the dynamic distribution index, establish a mapping relationship between the rate of change of the dynamic distribution index and the energy offset amplitude, and obtain the energy change matrix.

[0049] The stabilization modulation unit is used to perform frequency domain transformation based on the dynamic balance factor, and couple the amplitude distribution of the dynamic balance factor with the phase stability in different frequency bands to obtain the frequency domain modulation matrix.

[0050] The dual-domain fusion unit is used to establish a nonlinear cross-mapping between the time and frequency domains based on the energy change matrix and the frequency modulation matrix, extract the resonance region and form a fusion tensor to obtain the energy-stable correlation matrix.

[0051] The cooperative vector generation unit is used to extract and normalize the principal component directions of the fusion tensor based on the energy-stable correlation matrix, and transform them into a multi-dimensional cooperative vector group to obtain a multi-propeller cooperative vector group.

[0052] Furthermore, the instruction module includes:

[0053] The recursive setting unit is used to arrange the cooperative vectors in layers according to the propeller number and direction of action based on the multi-propeller cooperative vector group, establish a multi-level expansion index table, and obtain the recursive parameter set.

[0054] The power deconstruction unit is used to calculate the vector change rate and energy offset between adjacent layers based on the recursive parameter set, describe the power change trend, and obtain the power gradient matrix.

[0055] The sequence generation unit is used to expand the power change rate of each propeller into an adjustment sequence on the time axis according to the power gradient matrix, so as to obtain the instantaneous power adjustment sequence set;

[0056] The instruction construction unit is used to arrange the adjustment sequences of each propeller according to the number and time order based on the instantaneous power adjustment sequence set, so as to obtain the control instruction matrix.

[0057] The above-described solution of the present invention has at least the following beneficial effects:

[0058] This invention accumulates gradient components in the difference feature set according to the physical adjacency of the propellers, giving the original local feature data a logical structure of spatial topology. This enables information mapping from points to surfaces and explicitly encodes the local interactions of the propellers into a two-dimensional data structure, where the rows and columns of the matrix correspond to physical nodes and the matrix values ​​correspond to the numerical intensity of gradient accumulation. Through this partitioning accumulation and structural constraint mapping, a calculable, inverseable, and block-based overall kinetic chain data expression is formed, enabling subsequent operations to be performed in a unified coordinate system and realizing the transformation of the original data into modelable structural data.

[0059] This invention calculates the degree of non-uniform energy coupling based on the interaction power between propellers, and superimposes data from different dimensions in the coupling matrix to form a set of dynamic distribution indices that can be used to characterize the energy distribution state. This enables the energy distribution to have monotonic variation characteristics in the numerical space, distinguishes the relative relationships of different coupling states, and realizes a dimensionality reduction description of complex power chain information. This allows the system to describe the coupling distribution state of the overall propulsion structure with limited data dimensions, providing directly quantifiable input data for subsequent applications.

[0060] This invention achieves a structural mapping from local intensity values ​​to energy transfer paths by reverse mapping the local intensity regions of the propeller coupling matrix. This transforms the energy distribution of the system from a static description into dynamic path data. Local regions above a threshold are mapped to propeller number indices, establishing a power transfer sequence between propellers. The energy ratio relationship of different path segments is represented by the matrix entry weights, allowing manipulation of energy flow in the numerical domain. This provides a data link for dynamic control, linearizes the multidimensional spatial coupling relationship, and makes the power transfer relationship computable in the data space.

[0061] This invention calculates the synchronization stability between the main propeller and auxiliary propeller based on the propulsion coordination matrix, integrates parameter data from multiple dimensions, and enables numerical comparison and weighting of data from different sources within the same scale system. The output dynamic balance factor serves as a comprehensive result of the overall dynamic coordination state of the system. It has direct data operability, and the combined quantity obtained from the relevant calculations of multiple data sources can maintain consistent numerical continuity and stability in the data space. It completes the mapping from multiple input parameters to a single numerical feature, providing the system with a data-based balance constraint input.

[0062] This invention achieves complementary fusion of the dynamic distribution index and the dynamic equilibrium factor, and normalizes their changing trends to form a multi-dimensional collaborative vector group. Nonlinear cross-fusion between different data domains is realized through time series expansion and frequency domain mapping. The dynamic distribution index represents the spatial energy distribution characteristics, and the dynamic equilibrium factor represents the stability of time-varying changes. The normalized mapping of the two forms a cross-domain tensor data structure. The system transforms high-dimensional composite data into a small number of collaborative vector groups. Each vector contains consistency information of the propeller group in terms of energy and time. A large number of matrix variables that need to be processed simultaneously are mapped into small-scale vector groups, laying the input foundation for the generation of control commands. Attached Figure Description

[0063] Figure 1 This is a flowchart of a multi-propeller power cooperative control system provided in an embodiment of the present invention. Detailed Implementation

[0064] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0065] like Figure 1 As shown, an embodiment of the present invention proposes a multi-propeller power cooperative control system, the system comprising:

[0066] The feature module is used to acquire the operating data of each propeller, and to obtain the difference feature data by fitting the coupling effect difference between propellers through the thrust gradient change of adjacent propellers.

[0067] The coupling module is used to partition and accumulate the gradient components in the difference feature set according to the physical adjacency relationship, identify the internal structural constraints of the overall power chain of the multi-propeller, and obtain the propeller coupling matrix.

[0068] The index module is used to superimpose the interaction power between different propellers based on the propeller coupling matrix, calculate the degree of non-uniform energy coupling between multiple propellers, and obtain the power distribution index.

[0069] The correlation module is used to reverse map the local intensity region of the propeller coupling matrix according to the power distribution index, establish the power transfer path between propellers, and obtain the propulsion coordination matrix.

[0070] The factor module is used to calculate the degree of synchronization stability between the main propeller and the auxiliary propeller based on the propulsion coordination matrix, and obtain the dynamic balance factor.

[0071] The fusion module is used to complement and fuse the power distribution index and the dynamic balance factor, and normalize the changing trends of the two to form a cooperative vector, thus obtaining a multi-propeller cooperative vector group.

[0072] The instruction module is used to recursively expand the multi-propeller cooperative vector group, define the instantaneous power adjustment sequence of each propeller, and obtain the control instruction matrix.

[0073] In this embodiment of the invention, the feature module is used to acquire the operating data of each propeller, fit the difference in coupling effect between propellers by the thrust gradient change of adjacent propellers to obtain difference feature data, and transform the operating data into comparable difference features and bind them to a clear adjacency index, realizing the transformation from independent channel signals to interactive expressions; the coupling module is used to partition and accumulate the gradient components in the difference feature set according to physical adjacency, identify the internal structural constraints of the overall power chain of the multi-propeller, obtain the propeller coupling matrix, and embed the adjacent differences into a globally consistent two-dimensional matrix structure. The coupling relationship is fixed under a unified coordinate and index system to avoid duplicate counting or expansion. The solution is not closed due to the fracture of the propeller; the index module is used to superimpose the interaction power between different propellers based on the propeller coupling matrix, calculate the degree of non-uniform energy coupling between multiple propellers, and obtain the dynamic distribution index. It can provide a quantitative ranking of the coupling state of different spatial regions at the same time without changing the topology, avoiding the judgment split caused by relying on multi-channel thresholds; the association module is used to reverse map the local intensity region of the propeller coupling matrix based on the dynamic distribution index, establish the power transfer path between propellers, and obtain the propulsion coordination matrix. The power transfer is expressed from static intensity to a set of paths with clear edges and weights, providing a data channel for subsequent operations;

[0074] The factor module calculates the synchronization stability between the main propeller and the auxiliary propeller based on the propulsion coordination matrix, obtaining the dynamic balance factor. It projects observables onto a common time and index system, maintaining alignment with the coordination matrix in data structure to ensure consistent numerical references across modules and avoid computational ambiguity caused by the disordered superposition of multiple sources. The fusion module complementarily fuses the power distribution index and the dynamic balance factor, normalizing their changing trends to form a coordination vector, resulting in a multi-propeller coordination vector set. This explicitly stores the co-occurrence relationship between energy state and synchronization state in the form of matrix entries, carrying key linkage information with a small data size for easy subsequent processing. The command module recursively expands the multi-propeller coordination vector set, defining the instantaneous power adjustment sequence for each propeller to obtain a control command matrix. It transcribes the coordination relationship into directly executable time-series control data, facilitating deterministic writing by the execution end according to the matrix entries at each time step, ensuring the end-to-end data link from feature extraction to control output is closed and traceable.

[0075] This includes acquiring the operating data of each propeller, specifically:

[0076] The operational data specifically includes three types of parameters: rotational speed, torque, and fluid resistance. These three parameters constitute the sole input set for subsequent thrust gradient variation fitting and coupling effect calculation. First, propeller rotational speed data characterizes the propeller's rotational motion state and directly reflects its power output. Each propeller shaft is equipped with a high-precision angular velocity sensor or photoelectric encoder, which detects rotational pulse signals to obtain angular velocity changes and converts them into rotational speed data per unit time. During main controller startup, the sampling frequency and initial zero position of all encoders are uniformly calibrated to ensure consistency of propeller rotational speed data across the time axis. Second, torque data reflects the propeller's inherent energy output in overcoming fluid resistance to generate thrust in water, serving as the fundamental reference for thrust change rate. Torque is measured in real-time by a strain gauge torque sensor mounted on the drive shaft. The sensor converts axial stress changes into voltage signals, which are then converted from analog to digital to generate a discrete sampling sequence. The system collects torque data from each propeller at fixed time steps and uses a unified timestamp synchronization mechanism to align torque signals from different channels to the same sampling time. Finally, fluid resistance data reflects the interaction between the propeller and the surrounding fluid, and is a crucial variable affecting thrust output and energy distribution. Fluid resistance is measured by hydrodynamic pressure sensing units located at the bottom of the hull or the propeller hub. These sensors detect dynamic pressure changes on the propeller blade surface and perform real-time calculations via an integrated controller to obtain the fluid reaction resistance experienced by the propeller. All of the above data is acquired in real-time by a local data acquisition unit at a fixed sampling period. Subsequently, buffering and time calibration are performed at local nodes. By employing a unified time synchronization mechanism, the sampling times of each propeller node are aligned to generate operational data.

[0077] In a preferred embodiment of the present invention, the feature module includes:

[0078] The analysis unit is used to analyze the operating data of each propeller, and to simultaneously separate the speed, torque and fluid resistance to obtain the operating parameter matrix.

[0079] The gradient calculation unit is used to extract the thrust gradient of adjacent propellers based on the operating parameter matrix, calculate the thrust increment and angle change ratio in each time interval, and obtain the gradient change matrix.

[0080] The coupling analysis unit is used to establish a thrust response correlation model between propellers based on the gradient change matrix, extract the coupling difference values ​​of each adjacent propeller, and obtain the coupling difference matrix.

[0081] The feature generation unit is used to perform weighted fusion and normalization processing on the coupling difference values ​​of each propeller according to the coupling difference matrix, identify the differences in coupling effects between propellers, and obtain difference feature data.

[0082] In this embodiment of the invention, the analysis unit is used to analyze the operating data of each propeller, synchronously separate the rotational speed, torque, and fluid resistance to obtain the operating parameter matrix, avoiding numerical aliasing caused by cross-channel phase misalignment and the bias of dimensions on weighting and differentiation; the gradient calculation unit is used to extract the thrust gradient of adjacent propellers based on the operating parameter matrix, calculate the thrust increment and angle change ratio in each time interval, and obtain the gradient change matrix, providing deterministic input for identifying coupling differences and constructing coupling relationships; the coupling analysis unit is used to establish a thrust response correlation model between propellers based on the gradient change matrix, extract the coupling difference values ​​of each adjacent propeller, and obtain the coupling difference matrix, avoiding structural jumps caused by instantaneous fluctuations and providing stable input for subsequent feature generation; the feature generation unit is used to perform weighted fusion and normalization processing on the coupling difference values ​​of each propeller based on the coupling difference matrix, identify the coupling effect differences between propellers, obtain difference feature data, ensure that the fusion process is traceable and reproducible, guarantee the comparability across time periods and configurations, and form standardized feature loads.

[0083] The gradient calculation unit is used to extract the thrust gradient of adjacent propellers based on the operating parameter matrix, calculate the thrust increment and angle change ratio within each time interval, and obtain the gradient change matrix, specifically including:

[0084] First, an adjacency table is established based on the actual arrangement of the propellers on the hull, ensuring that each propeller is clearly associated with its neighboring propellers. Then, at fixed time intervals, differential calculations are performed on the thrust data of each propeller at consecutive time points, subtracting the thrust value from the previous moment to obtain the change in thrust over time. This reflects the increase or decrease trend of propeller thrust under the current sea state, transforming static measurement data into a dynamic signal. The system further performs spatial differential calculations on the thrust data between adjacent propellers, comparing the difference in thrust magnitude between two adjacent propellers at the same time to characterize the difference in their power distribution in space. Simultaneously, the system records the propulsion direction based on the propeller's arrangement angle and calculates the rate of change of angle between adjacent propellers over consecutive time periods, reflecting the rate of change of propulsion direction for different propellers within the same time range. Once the time variation, spatial difference component, and angle variation ratio are calculated, the system integrates these data to form a set of feature data that characterizes the power difference between adjacent propellers. This data is stored in the form of a matrix, where each row corresponds to the dynamic change information of a pair of adjacent propellers, and each column corresponds to different types of change indicators, thus obtaining a gradient change matrix.

[0085] The coupling analysis unit is used to establish a thrust response correlation model between propellers based on the gradient change matrix, extract the coupling difference values ​​of each adjacent propeller, and obtain the coupling difference matrix, specifically including:

[0086] First, the system reads the feature data of each pair of adjacent propellers from the gradient change matrix to establish their correspondence. Then, within a certain time window, the thrust change, time change, and angle change rate of adjacent propellers are used as input data, and the thrust difference between adjacent propellers is used as the output. A response model between the propellers is constructed through regression fitting, calculating the coefficients of mutual influence between each pair of adjacent propellers during thrust change, distinguishing the direct response relationship between propellers from random disturbance factors. After the response model of each pair of adjacent propellers is established, the system performs statistical analysis on the model parameters, such as comparing the amplitude relationship and directional consistency between different parameters, determining the degree of difference in response intensity between propellers, and converting these differences into a weighted numerical value to quantitatively describe the coupling difference between adjacent propellers. The higher the value, the stronger the inconsistency in the dynamic response; the lower the value, the tighter the coupling between them. After completing the calculations for all adjacent propellers, the system fills these coupling difference values ​​into a two-dimensional matrix according to the propeller numbers. The rows and columns of the matrix correspond to the propeller numbers, and each value in the matrix represents the degree of coupling difference between the two propellers, thus obtaining the coupling difference matrix.

[0087] In a preferred embodiment of the present invention, the coupling module includes:

[0088] The partitioning construction unit is used to determine the adjacency relationship of each propeller in the physical layout based on the difference feature data, establish a position index, and obtain the propeller adjacency set;

[0089] The cumulative calculation unit is used to partition and accumulate the gradient components in the difference feature data according to the adjacency relationship of adjacent propellers based on the adjacent set of propellers, so as to obtain the gradient accumulation matrix.

[0090] The constraint identification unit is used to extract the transmission chain nodes between multiple propellers based on the gradient accumulation matrix, perform constraint calibration on the torque correlation between each transmission chain node, and obtain the power chain constraint table.

[0091] The matrix generation unit is used to map the constraint relationships of each partition to a unified coordinate system based on the power chain constraint table, ensuring the internal structure of the overall power chain and obtaining the propeller coupling matrix.

[0092] In this embodiment of the invention, a partitioning construction unit is used to determine the adjacency relationship of each propeller in physical layout based on the difference feature data, establish a position index, and obtain a propeller adjacency set, which can distinguish between directly influential nodes and unrelated nodes, avoiding the introduction of false associations later; an accumulation calculation unit is used to accumulate the gradient components in the difference feature data according to the adjacency relationship of adjacent propellers based on the propeller adjacency set, and obtain a gradient accumulation matrix, which reflects the spatial clustering and difference of thrust response between propellers, enabling the subsequent differentiation of energy transfer intensity in different regions; a constraint identification unit is used to extract the transmission chain nodes between multiple propellers based on the gradient accumulation matrix, perform constraint calibration on the torque correlation between each transmission chain node, and obtain a power chain constraint table, realizing the transition from statistical gradient data to physical mechanical constraints and avoiding contradictory coupling paths; a matrix generation unit is used to map the constraint relationship of each partition to a unified coordinate system based on the power chain constraint table, ensuring the internal structure of the overall power chain, obtaining a propeller coupling matrix, realizing a global unified expression of local constraint data, and enabling the constraint intensity of different propeller regions to be compared under the same numerical system.

[0093] The cumulative calculation unit is used to partition and accumulate the gradient components in the difference feature data according to the adjacency relationship of adjacent propellers based on the adjacent propeller set, to obtain the gradient accumulation matrix, specifically including:

[0094] First, the gradient components of each propeller node are extracted from the characteristic difference data. Based on the propeller's neighbor set, the spatial neighbor numbers of each propeller node are sequentially retrieved. For any propeller node, the gradient component differences between it and its neighboring nodes are read, and these differences are arranged according to the time step to form a local gradient difference sequence. Subsequently, the system performs an accumulation operation on this sequence within each time window, using a weighted summation method. The weights are determined by the spatial distance between neighboring nodes, the propeller axis angle, or the fluid interference coefficient. The result of this operation reflects the comprehensive thrust gradient change experienced by the node at a specific moment. The system writes the weighted accumulated value of each node into a temporary storage matrix and continues to perform the same operation on all nodes. After completing the calculation for all nodes, the system normalizes the accumulated results within each time window to ensure that the numerical scales of different time intervals are consistent. After normalization, the system reorders the accumulated values ​​of each propeller node in the time dimension according to the node number and combines them in matrix form. The rows of this matrix represent propeller numbers, the columns represent time steps or calculation intervals, and the numerical entries in the matrix are the cumulative gradient intensities of the corresponding nodes within that time window, forming a gradient accumulation matrix.

[0095] The matrix generation unit is used to map the constraint relationships of each partition to a unified coordinate system based on the powertrain constraint table, ensuring the internal structure of the overall powertrain and obtaining the propeller coupling matrix. Specifically, it includes:

[0096] First, the system calls the kinetic chain constraint table, which records the node numbers, spatial coordinates, direction vectors, and constraint strength coefficients of the transmission chain nodes within each partition. This data is preprocessed to ensure the consistency of the coordinate systems of all nodes. If some partitions use local coordinate systems, such as relative coordinates established with a local hull section as the origin, the system will convert them to the global hull coordinate system using a coordinate transformation matrix. Coordinate transformation can be performed using homogeneous coordinate representation, achieved through a combination of rotation and translation matrices. After coordinate unification, the system constructs a corresponding constraint connection matrix based on the constraint strength values ​​of each pair of adjacent nodes in the constraint table. The rows and columns of the constraint connection matrix correspond to node numbers, and the matrix entries represent the constraint strength or directional parameters between the node pairs. When multiple constraints exist, the system merges the constraint values ​​of identical node pairs through superposition operations to form equivalent constraint weights. At this stage, the system normalizes all constraint strengths, limiting their values ​​to the [0,1] interval. Normalization can be based on maximum / minimum values ​​or standard deviation. Next, the system optimizes the structure of the normalized constraint matrix to eliminate isolated nodes and duplicate connections. Isolated nodes refer to propeller nodes that have not formed effective constraint relationships. When the system detects such nodes, it assigns nearest-neighbor constraint relationships to them through interpolation compensation or adjacency expansion to ensure the connectivity of the matrix. Redundant connections refer to multiple redundant constraints pointing to the same node pair in the constraint table; these are merged through averaging or weighted fusion. After structural optimization, the system maps this matrix to a global propeller index system, ensuring that each propeller node has a unique corresponding row and column in the matrix. Then, the system calculates the constraint projection angles between all nodes to determine whether the constraint directions are on the same force chain. If the constraint directions of a node pair are continuous, it is marked as part of the main power transmission chain in the matrix; if the directions differ significantly, it is marked as a secondary link or compensation channel. Finally, the overall constraint matrix is ​​symmetricized to ensure that the torque conservation relationship is maintained in numerical calculations, generating the propeller coupling matrix.

[0097] In a preferred embodiment of the present invention, the constraint recognition unit includes:

[0098] The node extraction unit is used to identify points with significant gradient changes in multiple propellers based on the gradient accumulation matrix, and define the points as nodes of the transmission chain to obtain a node index set.

[0099] The torque calculation unit is used to calculate the torque components and direction differences between any adjacent transmission chain nodes based on the node index set, and obtain the torque correlation matrix.

[0100] The constraint modeling unit is used to construct a set of constraint equations between nodes based on the moment correlation matrix, and to normalize the set of constraint equations to obtain a constraint coefficient table.

[0101] The calibration generation unit is used to determine the force boundary conditions of each node in the overall power chain of the multi-propeller based on the constraint coefficient table, identify the torque transmission relationship between the multi-propellers, and obtain the power chain constraint table.

[0102] In this embodiment of the invention, the node extraction unit is used to identify points with significant gradient changes in the multi-propeller system based on the gradient accumulation matrix, and define these points as nodes in the transmission chain, thereby obtaining a node index set and automatically identifying key structural nodes in the multi-propeller power chain, providing structured data for subsequent processing; the torque calculation unit is used to calculate the torque components and direction differences between any adjacent transmission chain nodes based on the node index set, obtaining a torque correlation matrix, realizing an explicit description of the mechanical interactions in the multi-propeller system, and ensuring that the data structure is reversible and decomposable in numerical operations; the constraint modeling unit is used for... Based on the torque correlation matrix, a set of constraint equations between nodes is constructed, and the constraint equations are normalized to obtain a constraint coefficient table. This realizes the mathematical transformation from torque matrix data to constraint parameter model, making data comparable under different ship types and power distribution conditions. The calibration generation unit is used to determine the force boundary conditions of each node in the overall multi-propeller power chain based on the constraint coefficient table, identify the torque transmission relationship between multiple propellers, and obtain the power chain constraint table. Through numerical boundary definition, a quantitative description of the force range of each node is realized, making the power chain structure visualized and traceable.

[0103] The torque calculation unit is used to calculate the torque components and direction differences between any adjacent nodes in the transmission chain based on the node index set, thereby obtaining the torque correlation matrix. Specifically, it includes:

[0104] First, the system retrieves the position information, propulsion direction, and rotation parameters of the propeller corresponding to each node within the ship's coordinate system. It then performs comprehensive calculations on the spatial distance, directional consistency, and thrust difference between adjacent nodes. For each pair of adjacent nodes, the system first determines their relative positional relationship in space and calculates the change in relative distance as a quantitative indicator of the lever arm. Subsequently, based on the thrust gradient change and propulsion direction of each node, the system calculates the local torque components formed by the two nodes, numerically evaluating the rotational contribution of the two nodes to the overall power chain of the system. Furthermore, it calculates the angle between the propulsion directions of adjacent nodes to reflect the consistency between the propeller rotation direction and the propulsion direction. The calculation results for all node pairs are then compiled into a matrix, where each entry represents the torque interaction strength and directional difference between adjacent nodes, resulting in a torque correlation matrix.

[0105] The constraint modeling unit is used to construct a set of constraint equations between nodes based on the moment correlation matrix, and to normalize the constraint equations to obtain a constraint coefficient table, which specifically includes:

[0106] First, the system scans all node relationships within the torque correlation matrix to identify data entries with stable torque correlations. Then, based on these valid entries, the system constructs a set of equations representing the node coordination relationships. For each pair of nodes with torque correlations, the system assumes a balance constraint relationship between the torque changes of the two nodes; if one increases, the other must compensate proportionally. After forming the complete set of equations, the system solves them using computational methods, obtaining the constraint coefficient for each node through numerical decomposition. This coefficient represents the proportion of the node's response to torque transmission in the entire power chain. To ensure the comparability of values ​​between different nodes, the system normalizes all coefficients, making them uniformly applicable within the same numerical range. The system also performs a consistency check on the solution results. If the constraint coefficients of some nodes vary too much, the system re-solves by resampling or extending the time window until the constraint coefficients stabilize. Finally, a data table containing node numbers and corresponding constraint coefficients is formed. Each record includes the constraint value of the node and the confidence index in the solution process, resulting in a constraint coefficient table, which is used to reflect the relative force coordination relationship between the nodes in the entire multi-propeller power chain.

[0107] The calibration generation unit is used to determine the force boundary conditions of each node in the overall power chain of the multi-propeller system based on the constraint coefficient table, identify the torque transmission relationship between the multi-propellers, and obtain the power chain constraint table, specifically including:

[0108] First, based on the values ​​recorded in the constraint coefficient table and combined with the historical thrust data saved by the nodes in the preceding calculations, the average force and fluctuation range of each node within a certain time range are calculated, forming the upper and lower limits of the force for each node, thus giving the force range of each node a quantifiable range. Next, the constraint coefficients of each node are compared. When there is a difference in the constraint coefficients of two adjacent nodes, the direction corresponding to this difference is considered the main direction of torque transmission; that is, the node with the larger constraint value is the active node, and the node with the smaller constraint value is the passive node. The system further combines the directional parameters between nodes with the historical force data to perform a proportional calculation of the torque transmission intensity, obtaining the torque transmission weight between each node. The system performs a topology consistency check on the transmission paths between all nodes. If a closed loop or duplicate path is detected, the system automatically removes the path with the smaller weight to eliminate non-physical feedback loops. The system outputs the node force boundaries, transmission directions, and transmission weights to obtain the kinetic chain constraint table.

[0109] In a preferred embodiment of the present invention, the index module includes:

[0110] The power distribution index unit is used to calculate the coupling weight of the power chain as the distance decreases along the path length based on the propeller coupling matrix, thus obtaining the topological coupling term; and to calculate the geometric equilibrium value of the power ratio of any two propellers based on the power ratio of each propeller, thus obtaining the scale ratio term.

[0111] Based on whether the included angle and rotation direction of any two propeller propulsion axes are consistent, the directional modulation of the alignment degree is calculated to obtain the directional consistency term; the magnitude amplification of the difference in the rate of change of any two propellers is calculated to obtain the dynamic difference term; the topological coupling term, scale ratio term, directional consistency term and dynamic difference term are fused together to calculate the local coupling aggregation term;

[0112] Based on the local coupling aggregation term, the scale factor of the compressed or stretched numerical range is calculated to obtain the scale shaping term; based on the dispersion of the overall propeller power ratio, the global non-uniformity modification term is calculated; the local coupling aggregation term, the scale shaping term, and the global non-uniformity modification term are fused to obtain the power distribution index.

[0113] In this embodiment of the invention, the power distribution index unit is used to calculate the coupling weight of the power chain, which decreases with distance along the path length, based on the propeller coupling matrix, to obtain a topological coupling term. This ensures that subsequent measurements are dominated by spatially achievable transmission chains, avoiding unreasonable amplification of non-physically distant items in the index. Based on the power proportion of each propeller, the geometric equilibrium value of the power proportion of any two propellers is calculated to obtain a scale ratio term, stably characterizing the contribution of two propellers of similar scale to energy coupling, avoiding the dominance of large values ​​on one side. Based on whether the included angle and rotation direction of the propulsion axes of any two propellers are consistent, the direction modulation of the alignment degree is calculated to obtain a direction consistency term, converting discrete attitude and rotation direction information into a continuous modulation factor, objectively reflecting the consistency of the propellers in the thrust direction. The magnitude amplification of the difference in the rate of change of any two propellers is calculated to obtain a dynamic difference term, which reflects the change in... The magnitude of the rate difference represents the relative temporal evolution difference, avoiding the judgment of similar states based solely on instantaneous amplitude. The topological coupling term, scale ratio term, direction consistency term, and dynamic difference term are fused to calculate the local coupling aggregation term, quantifying the comprehensive measure of local coupling strength and avoiding subjective weighting settings for cross-dimensional summation. Based on the local coupling aggregation term, the scale factor for compressing or stretching the numerical range is calculated to obtain the scale shaping term, avoiding information loss due to oversaturation or sparsity of entries during subsequent merging. Based on the dispersion of the overall propeller power proportion, a global non-uniformity modification term is calculated to objectively reflect the impact of overall non-uniformity on local entries. The local coupling aggregation term, scale shaping term, and global non-uniformity modification term are fused to obtain the dynamic distribution index, ensuring that objective and calculable indicators exist at different decision granularities, and ensuring that the index can be directly compared and reused across batches and operating conditions.

[0114] In a preferred embodiment of the present invention, the association module includes:

[0115] The region identification unit is used to determine the local intensity regions in the propeller coupling matrix that are higher than the preset power threshold based on the power distribution index, thereby obtaining the intensity region set;

[0116] The reverse mapping unit is used to map the energy distribution nodes in each intensity region to propeller number indices based on the intensity region set, establish the energy backtracking relationship between propeller pairs, and obtain the backtracking mapping table.

[0117] The path construction unit is used to connect propeller nodes with energy backtracking relationships sequentially according to the backtracking mapping table to form a continuous power transfer link and obtain a transfer path set;

[0118] The coordination matrix unit is used to convert the energy ratio of each path segment into matrix entry weights based on the transmission path set, determine the power transmission path between propellers, and obtain the propulsion coordination matrix.

[0119] In this embodiment of the invention, the region identification unit is used to determine the local intensity regions in the propeller coupling matrix that are higher than a preset power threshold based on the power distribution index, thereby obtaining a set of intensity regions. Each region is given a clear boundary, scale, and intensity metric to avoid unconstrained combination across the entire domain. The reverse mapping unit is used to map the energy distribution nodes in each intensity region to propeller number indices based on the intensity region set, establishing an energy backtracking relationship between propeller pairs, and obtaining a backtracking mapping table. This achieves the mapping from matrix entries to physical number fields, facilitating horizontal comparisons between different regions and different time slices. The path construction unit is used to determine the path based on the backtracking mapping table. By sequentially connecting propeller nodes with energy backtracking relationships, a continuous power transfer link is formed, resulting in a transfer path set. The edge-level backtracking relationships are organized into power transfer links with definite start, end, and order, so that energy interaction is no longer represented only by isolated edge weights, but is recorded in a serialized structure, providing replicable rules for the future. The coordination matrix unit is used to convert the energy ratio of each path segment into matrix entry weights based on the transfer path set, determine the power transfer path between propellers, and obtain the propulsion coordination matrix. This allows power transfer to be rewritten from a sequential structure to a unified matrix structure at the data layer, converting the dimensional differences of different regions and different paths to the same scale.

[0120] The reverse mapping unit is used to map the energy distribution nodes in each intensity region to propeller number indices based on the intensity region set, establish the energy backtracking relationship between propeller pairs, and obtain the backtracking mapping table, specifically including:

[0121] First, for each intensity region, the energy distribution entries within it are read. This process is then iterated through by region label, and the matrix coordinates of each entry are reconstructed into the corresponding propeller number pair. Simultaneously, the system verifies whether the number pair satisfies the adjacency or transitivity relationship based on the ship's drawings or a pre-defined physical adjacency list. Entries that do not satisfy physical adjacency are directly eliminated at this stage. Within the same region, the energy interaction intensity between each propeller pair is calculated, and multiple records of the same pair within the same time slice are aggregated to form a candidate relationship record with a unique index. Subsequently, the system determines the direction attribute for each candidate relationship. The direction is determined based on the intensity comparison of the same propeller pair in both directions: if the interaction intensity from one propeller to another is greater than the reverse intensity, the energy backtracking direction is confirmed as from the stronger side to the weaker side; if the bidirectional intensity is equal and no direction can be determined, the entry is marked as having an uncertain direction and is not temporarily added to the backtracking table. After the direction is determined, a dimensionless relative contribution ratio is calculated for each directed relation. This ratio is obtained by comparing the intensity of the entry with the sum of the intensities of all outward entries from the same source propeller within the same time slice, thus grouping the intensities of different regions and times into a comparable unified scale. After obtaining the direction and ratio, the system performs a cross-regional and cross-time merging. For entries from the same source propeller and the same destination propeller that appear consecutively in adjacent time slices, the system adds a continuity marker based on temporal continuity. For the same pair of relations that appears repeatedly in different intensity regions, the system merges them according to a preset convergence strategy, typically using regional intensity, regional area, or regional confidence as weighted comparison values ​​for weighted summation, and retaining the indexes of all source regions for traceability. After merging, the system applies threshold and quality control rules, discarding entries with excessively small comparison values, uncertain directions, or lacking physical adjacency support, retaining only relations that meet the minimum intensity, minimum ratio, and structural consistency requirements. Finally, the system obtains a backtracking mapping table in the form of a standardized data table for all the filtered directed relations.

[0122] The path construction unit is used to sequentially connect propeller nodes with energy backtracking relationships according to the backtracking mapping table, forming a continuous power transfer link and obtaining a transfer path set, specifically including:

[0123] First, a node set is established based on the numbers of all propellers appearing in the graph, and a weighted set of directed edges is established for each directed backtracking relationship, forming a directed weighted graph structure for path growth. To determine the starting point set of the path, the system counts the incoming and outgoing edges of each node, prioritizing nodes with zero incoming edges as starting points; if no such nodes exist, nodes with a departure strength not less than an incoming strength are selected as candidate starting points. After the starting points are determined, the system begins to grow paths one by one: starting from the current node, all available outgoing relationships are retrieved, and without violating physical adjacency and temporal order constraints, the edge with the largest backtracking ratio is selected as the next expansion direction, and the next node is added to the current path sequence, while recording the edge weight and source region label of that segment. During path growth, if a node that is about to return to a previously encountered node is detected, the system triggers a loop handling mechanism. Loop processing employs two strategies: First, it uses the local minimum edge weight as the disconnection criterion, marking the relationship with the smallest weight within the loop as redundant and removing it from the current expansion set. Second, it temporarily shrinks several nodes within the loop into an equivalent supernode, using the minimum edge weight of this supernode as the effective upper limit for subsequent expansion, maintaining the unidirectional expandability of the path through equivalent line segments. The path stopping conditions include three aspects: first, reaching a terminal node without outward relationships; second, the path length reaching a preset upper limit; and third, the currently available maximum outward ratio falling below a minimum threshold. When any condition is met, path growth stops at that branch, and the system saves the formed node sequence and corresponding edge weight sequence as a transmission path, assigning it a unique path identifier. After all starting points have completed expansion, the system performs overlap processing and consistency merging on the resulting multiple paths. For paths with identical node sequences but different edge weight statistics, the system merges them according to source weight and time coverage to generate a single representative path. For partially overlapping paths, the system segments and splices them using shared subsequences as boundaries, ensuring that each retained path has a clear starting point, ending point, and intermediate order. After the path set is finally determined, a mapping relationship from path to matrix entry is established: the segment contribution is calculated for each path's adjacent node pairs. The segment contribution is usually represented by the bottleneck value of the edge weights along the path or the normalized average value; then, the path convergence coefficient is set according to the overall strength or length of the path, and the segment contribution is written back to the unified advancement coordination matrix entry. The path identifier and source region are recorded on the entry to ensure that multiple paths with the same pair of numbers can be accurately accumulated or retrieved individually in subsequent statistics, thus obtaining a backtracking mapping table.

[0124] In a preferred embodiment of the present invention, the cooperative matrix unit includes:

[0125] The energy allocation unit is used to quantify the energy difference between the starting and ending nodes of each path segment according to the transmission path set, establish an energy ratio table, and obtain the path energy set.

[0126] The weight correction unit is used to perform symmetric balancing based on the path energy set, and to perform bidirectional correction and normalization on the weight values ​​of the bidirectional transmission paths in the energy ratio table to generate an energy weight matrix.

[0127] The topology mapping unit is used to map the energy weight distribution to the physical connection topology between propellers based on the energy weight matrix, forming a topology association table;

[0128] The propulsion coordination matrix unit is used to incorporate the energy weights of each path segment into the matrix according to the physical adjacency relationship based on the topology association table, and to determine the power transfer path relationship between propellers, thus obtaining the propulsion coordination matrix.

[0129] In this embodiment of the invention, the energy allocation unit is used to quantify the energy difference between the starting and ending nodes of each path segment according to the transmission path set, establish an energy ratio table, obtain the path energy set, and describe the directional difference of any path segment, so as to maintain the consistency of calculation caliber among fleets of different sizes and power levels; the weight correction unit is used to perform symmetrical balancing according to the path energy set, and to perform bidirectional correction and normalization on the weight values ​​of bidirectional transmission paths in the energy ratio table, generating an energy weight matrix. The bidirectional entries are kept in pairs in the data, and the directional information is explicitly encoded with symbols, which facilitates subsequent processing. Item-by-item mapping; Topology mapping unit, used to map the energy weight distribution to the physical connection topology between propellers according to the energy weight matrix, forming a topology association table. Logical edges that do not exist physically are cleared in the data, and the energy weight distribution and actual connection relationship are aligned at the entry level; Propulsion coordination matrix unit, used to merge the energy weights of each path segment into the matrix according to the physical adjacency relationship according to the topology association table, and determine the power transfer path relationship between propellers to obtain the propulsion coordination matrix. Off-diagonal elements reflect the directed or undirected coupling relationship of power transfer, while diagonal elements play the role of row closure and local conservation, serving as subsequent input.

[0130] The weight correction unit is used to perform symmetric balancing based on the path energy set, and to perform bidirectional correction and normalization on the weight values ​​of the bidirectional transmission paths in the energy ratio table to generate an energy weight matrix. Specifically, it includes:

[0131] First, the path energy set is read within the same time window. For each record pointing from one propeller to another, a search is conducted to determine if a corresponding record exists in the opposite direction. If not, it is filled with zero difference or the estimated value from the previous time window, ensuring that any pair of propellers forms paired entries in both directions within the time window. Next, the direction weights are calculated using the energy difference information of the paired entries. The calculation method ensures that the weights for two opposite directions are one positive and one negative, with equal numerical values, thus explicitly preserving the direction attribute in the data. Then, organized by propeller row, the direction weights pointing from the same propeller to all adjacent propellers are processed on the same scale: first, the total weight of all external connections for that propeller within the current time window is calculated; then, this total is used to proportionalize the weights of each external connection, resulting in a comparable normalized result. When the total is too small or zero, a very small stabilizing factor is introduced to avoid incalculable situations. Before normalization, entries that significantly deviate from the norm can undergo amplitude limiting or sliding statistical processing to suppress the influence of occasional noise on the results. All normalized weights are assembled into a sparse two-dimensional data structure according to the propeller numbering order. A closed term is added to the diagonal position of each row so that the sum of the off-diagonal elements of each row cancels out the diagonal element, forming a locally conserved row structure, thus obtaining the energy weight matrix.

[0132] The topology mapping unit is used to map the energy weight distribution to the physical connection topology between propellers based on the energy weight matrix, forming a topology association table, which specifically includes:

[0133] First, based on the actual arrangement of the ship formation, a physical adjacency relationship consisting only of zeros and ones is constructed to indicate whether any two propellers are structurally directly adjacent or have a valid physical connection. This relationship can be derived from design drawings, installation lists, or online identification results, and is frozen as a read-only reference after its establishment. Then, the energy weight matrix obtained in the previous step is aligned item by item with this physical adjacency relationship: all entries without a physical direct connection are set to zero, retaining only entries that correspond one-to-one with physical connections, thereby eliminating spurious connections that may arise from statistical associations at the data level. To facilitate subsequent assembly and retrieval, the retained non-zero entries are uniformly sorted according to propeller number and time window, and each record is converted into a key-value entry containing a start-point number, end-point number, time window index, and weight value; if the formation topology has directional constraints, the record's directional attribute is displayed in the key-value entry. Subsequently, a connectivity check is performed on the key-value entries within the current time window, assigning consistent block identifiers to entries belonging to the same connected component for parallel computation or block assembly, ultimately forming a topology association table.

[0134] The propulsion coordination matrix unit is used to incorporate the energy weights of each path segment into the matrix according to their physical adjacency based on the topology association table, and to determine the power transfer path relationship between propellers, thus obtaining the propulsion coordination matrix, which specifically includes:

[0135] First, a two-dimensional matrix of all zeros is initialized within the defined propeller numbering space for the current time window. Then, each record in the topological association table is traversed, and the weight value of the record is written into the corresponding off-diagonal position in the matrix. Each time an off-diagonal entry is written, the diagonal position of that row is simultaneously updated, ensuring that the sum of all off-diagonal elements cancels out the diagonal elements, thus maintaining row closure during assembly. If the directional relationship of power transfer needs to be expressed, the asymmetric structure of the matrix is ​​preserved; if modal analysis or eigenvalue decomposition is required, a symmetric equivalent matrix is ​​generated as needed after assembly, while the original matrix is ​​retained for backtracking directional information. After all entries are written, scale constraints or spectral radius limits can be applied to the matrix row-by-row to ensure that the numerical range remains within a preset interval, facilitating subsequent stable calculations. To facilitate the clarification of power transfer path relationships at the matrix level, off-diagonal positions can be labeled based on preset thresholds: entries with significantly positive values ​​are considered as effective transfer segments from the corresponding propeller in the row to the corresponding propeller in the column, entries with significantly negative values ​​are considered as effective transfer segments in the opposite direction, and entries with values ​​close to zero are considered as having no significant transfer within the current time window; this labeling serves only as an index and does not change the values ​​of the matrix itself, thus yielding the propulsion coordination matrix.

[0136] In a preferred embodiment of the present invention, the factor module includes:

[0137] The dynamic balance factor unit is used to calculate the instantaneous response difference between each pair of main propellers and auxiliary propellers based on the propulsion coordination matrix, and obtain the time offset term; based on the output power amplitude of each propeller, it compares the power fluctuation direction and amplitude ratio of the main propeller and auxiliary propeller in the same time period, calculates the degree of coordination amplitude between the two, and obtains the amplitude coupling term.

[0138] Based on the inertia parameters and relative spatial positions of each propeller, the difference in inertial response between the main propeller and the auxiliary propeller is measured, the degree of inertia matching is calculated, and the inertia compensation term is obtained; based on the instantaneous frequency changes of each propeller, the average operating frequency and fluctuation range of the propellers are statistically analyzed, the group frequency consistency among each propeller is calculated, and the frequency synchronization term is obtained.

[0139] By fusing the timing offset term, amplitude coupling term, inertia compensation term, and frequency synchronization term, the overall synchronization stability among multiple propellers is calculated, and the dynamic balance factor is obtained.

[0140] In this embodiment of the invention, the dynamic balance factor unit is used to calculate the instantaneous response difference between each pair of main propellers and auxiliary propellers based on the propulsion coordination matrix, obtaining a timing offset term to provide a locatable peak hysteresis and avoid the sensitivity of pure threshold triggering to short-term noise; based on the output power amplitude of each propeller, the power fluctuation direction and amplitude ratio of the main propeller and auxiliary propeller in the same time period are compared to calculate the degree of coordination amplitude between the two, obtaining an amplitude coupling term to suppress the influence of individual anomalies, and the median is used to further resist heavy-tail error. The combination of the two forms an objective characterization of coordinated fluctuation; based on the inertia parameters and relative spatial positions of each propeller, the timing difference between the main propeller and auxiliary propeller is determined. The difference in inertial response between propellers is used to calculate the degree of inertia matching and obtain the inertia compensation term. This avoids the geometric distance or inertia difference from dominating alone, ensuring numerical stability and comparability across pairs. Based on the instantaneous frequency changes of each propeller, the average operating frequency and fluctuation range of the propellers are statistically analyzed to calculate the group frequency consistency between propellers and obtain the frequency synchronization term. This enables horizontal comparability and ensures that there is still computable output under operating conditions such as start-up and shutdown. The timing offset term, amplitude coupling term, inertia compensation term, and frequency synchronization term are fused to calculate the comprehensive synchronization stability between multiple propellers and obtain the dynamic balance factor. This provides a fine-grained characterization of the state of any master-slave pair and achieves a single-valued description of the overall synchronization stability.

[0141] In a preferred embodiment of the present invention, the fusion module includes:

[0142] The energy mapping unit is used to perform time-series expansion based on the dynamic distribution index, establish a mapping relationship between the rate of change of the dynamic distribution index and the energy offset amplitude, and obtain the energy change matrix.

[0143] The stabilization modulation unit is used to perform frequency domain transformation based on the dynamic balance factor, and couple the amplitude distribution of the dynamic balance factor with the phase stability in different frequency bands to obtain the frequency domain modulation matrix.

[0144] The dual-domain fusion unit is used to establish a nonlinear cross-mapping between the time and frequency domains based on the energy change matrix and the frequency modulation matrix, extract the resonance region and form a fusion tensor to obtain the energy-stable correlation matrix.

[0145] The cooperative vector generation unit is used to extract and normalize the principal component directions of the fusion tensor based on the energy-stable correlation matrix, and transform them into a multi-dimensional cooperative vector group to obtain a multi-propeller cooperative vector group.

[0146] In this embodiment of the invention, the energy mapping unit is used to perform time-series expansion based on the dynamic distribution index, establish a mapping relationship between the rate of change of the dynamic distribution index and the energy offset amplitude, obtain an energy change matrix, retain two complementary types of information: local rapid changes and global bias, and realize deterministic mapping and traceable recording from vector indices to time-series matrices; the stabilization modulation unit is used to perform frequency domain transformation based on the dynamic balance factor, couple the amplitude distribution and phase stability of the dynamic balance factor in different frequency bands to obtain a frequency domain modulation matrix, and convert the time-domain dynamic balance factor into quantifiable frequency-domain weights, which is convenient for interaction with time-domain objects. The system performs the following operations: a dual-domain fusion unit, which establishes a nonlinear cross-mapping between the time and frequency domains based on the energy change matrix and the frequency modulation matrix, extracts the resonance region and forms a fusion tensor to obtain an energy-stable correlation matrix, realizing a closed-loop construction from a three-dimensional tensor index to a two-dimensional correlation graph, ensuring a clear data foundation for subsequent operations; and a cooperative vector generation unit, which extracts and normalizes the principal component directions of the fusion tensor based on the energy-stable correlation matrix, transforming them into a multi-dimensional cooperative vector group, obtaining a multi-propeller cooperative vector group, completing the deterministic dimensionality reduction and reconstruction from a cross-domain matrix to a low-dimensional executable representation, and providing an input interface for subsequent operations.

[0147] The energy mapping unit is used to perform time-series expansion based on the dynamic distribution index, establish a mapping relationship between the rate of change of the dynamic distribution index and the energy shift amplitude, and obtain the energy change matrix, specifically including:

[0148] First, the power distribution index sequence of each propeller is collected within a continuous time window. The timestamps are aligned according to a uniform sampling period. For missing samples, spline interpolation or linear interpolation is used to complete the data, and high-frequency noise is suppressed by a polynomial smoothing filter or moving average. Then, the sequence of each propeller within the window is decentered and variance standardized to allow data from different sources and with different dimensions to be compared and superimposed in the same numerical domain. The difference between adjacent time points is calculated as the instantaneous rate of change on the smoothed sequence, and mirror extension or boundary preservation is used at the beginning and end of the window to avoid length shortening and boundary distortion. At each time point, the group reference level of all propellers, such as the mean or median, is calculated, and the relative energy shift is characterized by the offset of each propeller relative to the reference level. The instantaneous rate of change and the relative energy shift are linearly superimposed using calibrated weighting coefficients, and amplitude shaping is performed through a reversible compressed nonlinear mapping to characterize the spikes caused by the rate of change and the gradual changes caused by the shift on the same scale. The vectors obtained for all propellers at each time point are stacked row by row in chronological order to form an energy change matrix.

[0149] The stabilization modulation unit is used to perform frequency domain transformation based on the dynamic balance factor, coupling the amplitude distribution of the dynamic balance factor in different frequency bands with the phase stability to obtain the frequency domain modulation matrix, specifically including:

[0150] First, a dynamic balance factor sequence consistent with the aforementioned time window and sampling rhythm is obtained, and interval normalization is performed to eliminate the influence of dimensional differences and drift. A short-time discrete Fourier transform is performed with a fixed length and fixed step size to obtain segmented amplitude spectra and phase sequences. Within the entire window, the dispersion of phase difference and the proportion of phases falling within the tolerance range are statistically analyzed for each frequency band to jointly measure phase stability. To prevent occasional spikes from biasing the frequency band weights, the amplitude spectrum is robustly summarized using the mean or quantiles within the window, and then combined with phase stability using multiplicative or other deterministic coupling rules to form candidate frequency band weights. Subsequently, normalization is performed on all frequency bands of each propeller to make the sum of the weights of each frequency band constant, and a lower limit is applied to truncate the smallest weights to avoid subsequent numerical underflow. The weights of each propeller in each frequency band are concatenated in a row-by-row manner to obtain the frequency domain modulation matrix.

[0151] The dual-domain fusion unit is used to establish a nonlinear cross-mapping between the time and frequency domains based on the energy change matrix and the frequency modulation matrix, extract the resonance region, form a fusion tensor, and obtain the energy-stable correlation matrix, specifically including:

[0152] First, the energy change matrix is ​​normalized row-wise and a noise threshold is set. Entries below the threshold are gated and zeroed, resulting in a sparse time-domain matrix that retains only the intensity at effective moments. The frequency-domain modulation matrix is ​​normalized column-wise and sparsified, retaining only the frequency bands with the highest contribution from each propeller, and clearing the remaining entries to highlight the main frequency components and reduce computational burden. Based on this, a nonlinear multiplicative cross-mapping between time intensity and frequency band weights is constructed for each propeller. The time axis and frequency axis are combined pairwise to generate a three-dimensional fusion tensor, which is then aggregated along the propeller dimension to obtain a two-dimensional energy-stable response map in both time and frequency dimensions. Connectivity analysis and peak persistence discrimination are performed on the graph, setting the minimum duration and minimum bandwidth. Then, morphological closing operations are performed to remove isolated holes and edge spikes, forming a set of resonant regions that satisfy persistence and continuity. These resonant regions are then written back to the propeller-propeller relationship plane according to the predetermined projection rules: the time domain vector is normalized at each time slice and the common participation of each propeller pair is calculated; the frequency domain weights are normalized at each frequency band and the common weights of each propeller pair are calculated; and the joint contributions of these time and frequency are accumulated in all resonant regions to construct a symmetric and diagonally consistent energy-stable correlation matrix.

[0153] In a preferred embodiment of the present invention, the instruction module includes:

[0154] The recursive setting unit is used to arrange the cooperative vectors in layers according to the propeller number and direction of action based on the multi-propeller cooperative vector group, establish a multi-level expansion index table, and obtain the recursive parameter set.

[0155] The power deconstruction unit is used to calculate the vector change rate and energy offset between adjacent layers based on the recursive parameter set, describe the power change trend, and obtain the power gradient matrix.

[0156] The sequence generation unit is used to expand the power change rate of each propeller into an adjustment sequence on the time axis according to the power gradient matrix, so as to obtain the instantaneous power adjustment sequence set;

[0157] The instruction construction unit is used to arrange the adjustment sequences of each propeller according to the number and time order based on the instantaneous power adjustment sequence set, so as to obtain the control instruction matrix.

[0158] In this embodiment of the invention, the recursive setting unit is used to arrange the cooperative vectors hierarchically according to the propeller number and direction of action based on the multi-propeller cooperative vector group, establish a multi-level expansion index table, and obtain a recursive parameter set to avoid ambiguity caused by many-to-one mapping and provide a definite reference value for subsequent processing. The power deconstruction unit is used to calculate the vector change rate and energy offset between adjacent levels based on the recursive parameter set, describe the power change trend, and obtain a power gradient matrix to ensure that the information of the parent layer and the child layer is numerically compatible, to ensure the cumulativeity under time progression, and to ensure consistent comparison with existing execution values. The sequence generation unit is used to expand the power change rate of each propeller into an adjustment sequence on the time axis based on the power gradient matrix to obtain an instantaneous power adjustment sequence set, ensuring that the sequence progresses gradually within the allowable domain, and the quantization constraint maps continuous values ​​to discrete code values ​​that the actuator can recognize. The instruction construction unit is used to arrange the adjustment sequences of each propeller according to the number and time order based on the instantaneous power adjustment sequence set to obtain a control instruction matrix, complete the organization from sequence to instruction matrix, and ensure that the execution end interprets and executes the sequence unambiguously according to the number and time order.

[0159] The power deconstruction unit is used to calculate the vector change rate and energy offset between adjacent layers based on the recursive parameter set, describe the power change trend, and obtain the power gradient matrix, specifically including:

[0160] First, at the start of each control cycle, the multi-level expanded index and recursive parameter set established in the previous step are read. Cooperative vectors with parent-child relationships are time-aligned, and all data is resampled to the same equally spaced time scale. When the original records contain missing data, duplicates, or overlapping times, they are repaired using window interpolation and window boundary truncation to ensure that each time scale, each number, and each direction corresponds to only one unique data point. Then, the difference between the parent and child layers is calculated based on the inter-layer allocation ratio, and orthogonal decomposition is performed on the directional components to ensure that the difference accurately falls under the two-dimensional coordinates of the number and direction. To avoid noise amplification caused by direct differencing, a fixed-width sliding fit and smoothing process is applied to the difference sequence before calculating the rate of change, and then the rate of change is calculated according to the equally spaced time scale. The energy offset is based on the actual execution value at the end of the previous cycle. The product of the rate of change and the time step is gradually accumulated along the time axis, and the accumulated result is compared with the baseline to obtain the offset. All entries then undergo out-of-bounds checks: when the rate of change or offset touches a mechanically, electrically, or software-defined boundary, the system replaces the original value with the boundary value and records the reason for the out-of-bounds violation and the location of the violation in a separate table, ultimately obtaining the power gradient matrix.

[0161] The sequence generation unit is used to expand the power change rate of each propeller into an adjustment sequence on the time axis based on the power gradient matrix, thereby obtaining a set of instantaneous power adjustment sequences, specifically including:

[0162] First, for each propeller number and its directional components, a time series starting from the actual execution value of the previous cycle is established. At each time scale, the corresponding rate of change and energy offset are read. Based on the time scale coefficient and energy amortization coefficient given in the recursive parameter set, the predicted increment of that time scale is obtained and added to the actual value of the previous time scale to form the unconstrained target value. Subsequently, three types of constraints are applied in sequence: first, the rate of change constraint, which limits the change between adjacent time scales to the maximum allowable range; second, the amplitude constraint, which limits the target value to the minimum and maximum allowable range of the execution channel; and third, the resolution constraint, which rounds the continuous target value according to the actuator resolution so that it falls into the set of discrete code values ​​that can be issued. To avoid frequent switching between increasing and decreasing directions in a short period of time, the system performs directional consistency detection on a fixed-width time window, merges segments in the same direction, and inserts a minimum dwell time at the direction switching position to ensure that the segment has a sustainable dwell time at the execution end. For the moment marked in the out-of-bounds record, the system retains both the value before and after the constraint for easy traceability. Finally, a time-stamped instantaneous power adjustment sequence is formed for each number. The sequence covers all directional components under that number and includes identification information from the constraint and out-of-bounds processing. After merging the sequences of all numbers, an instantaneous power adjustment sequence set is formed.

[0163] The instruction construction unit is used to arrange the adjustment sequences of each propeller according to their number and time order based on the instantaneous power adjustment sequence set, thereby obtaining a control instruction matrix, which specifically includes:

[0164] First, a primary time index and a secondary index for numbers are established, and a fixed direction column order is pre-defined for each number to ensure that the column position of the same direction remains consistent across any time row. Then, a two-dimensional control instruction data table is assembled, with time as the row and numbers and their direction components as columns. Before writing to a time row, the system checks if there is any overwriting update for that time scale; if so, the latest entry replaces the historical entry, ensuring that only a unique valid value is retained at the same time position. To support reliable deployment and on-site reproduction, the system converts each row into a framed representation with a frame header and a checksum field: the frame header includes a timestamp, unit description, direction column order version, and row-level summary checksum; this is followed by the discrete code values ​​of all numbers and directions for that time row; the row-level checksum and its source summary are appended to the end of the row. The source summary points to the version number of the index and parameters, as well as the number of out-of-bounds flags contained in the row. Simultaneously, overall metadata is generated at the full table level, including the start and end times of coverage, time step, number list, direction column order, channel resolution, and boundary descriptions. In terms of storage organization, the data table is maintained with a circular buffer. Newly arrived time windows will overwrite old windows according to the time index, so that the reading end always obtains the instructions valid for the current cycle, and finally obtains the control instruction data table, which can be directly converted into a bus transmission frame sequence or mapped by the execution end into a periodic task table. The execution end reads in a fixed order from the time row to the number column group and then to the direction column, so that the discrete code value of the corresponding number and direction can be written into the actuator channel at each time scale, completing the deterministic arrangement and delivery from the adjustment sequence to the executable control instructions.

[0165] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multi-propeller power cooperative control system, characterized in that, The system includes: The feature module is used to acquire the operating data of each propeller, and to obtain the difference feature data by fitting the coupling effect difference between propellers through the thrust gradient change of adjacent propellers. The coupling module is used to partition and accumulate the gradient components in the difference feature set according to the physical adjacency relationship, identify the internal structural constraints of the overall power chain of the multi-propeller, and obtain the propeller coupling matrix. The index module is used to superimpose the interaction power between different propellers based on the propeller coupling matrix, calculate the degree of non-uniform energy coupling between multiple propellers, and obtain the power distribution index. The association module is used to reverse map the local intensity region of the propeller coupling matrix according to the power distribution index, establish the power transfer path between propellers, and obtain the propulsion coordination matrix. The factor module is used to calculate the degree of synchronization stability between the main propeller and the auxiliary propeller based on the propulsion coordination matrix, and obtain the dynamic balance factor. The fusion module is used to complement and fuse the power distribution index and the dynamic balance factor, and normalize the changing trends of the two to form a cooperative vector, thus obtaining a multi-propeller cooperative vector group. The instruction module is used to recursively expand the multi-propeller cooperative vector group, define the instantaneous power adjustment sequence of each propeller, and obtain the control instruction matrix.

2. The multi-propeller power cooperative control system according to claim 1, characterized in that, The feature module includes: The analysis unit is used to analyze the operating data of each propeller, and to simultaneously separate the speed, torque and fluid resistance to obtain the operating parameter matrix. The gradient calculation unit is used to extract the thrust gradient of adjacent propellers based on the operating parameter matrix, calculate the thrust increment and angle change ratio in each time interval, and obtain the gradient change matrix. The coupling analysis unit is used to establish a thrust response correlation model between propellers based on the gradient change matrix, extract the coupling difference values ​​of each adjacent propeller, and obtain the coupling difference matrix. The feature generation unit is used to perform weighted fusion and normalization processing on the coupling difference values ​​of each propeller according to the coupling difference matrix, identify the differences in coupling effects between propellers, and obtain difference feature data.

3. The multi-propeller power cooperative control system according to claim 2, characterized in that, The coupling module includes: The partitioning construction unit is used to determine the adjacency relationship of each propeller in the physical layout based on the difference feature data, establish a position index, and obtain the propeller adjacency set; The cumulative calculation unit is used to partition and accumulate the gradient components in the difference feature data according to the adjacency relationship of adjacent propellers based on the adjacent propeller set, so as to obtain the gradient accumulation matrix. The constraint identification unit is used to extract the transmission chain nodes between multiple propellers based on the gradient accumulation matrix, perform constraint calibration on the torque correlation between each transmission chain node, and obtain the power chain constraint table. The matrix generation unit is used to map the constraint relationships of each partition to a unified coordinate system based on the power chain constraint table, ensuring the internal structure of the overall power chain and obtaining the propeller coupling matrix.

4. A multi-propeller power cooperative control system according to claim 3, characterized in that, The constraint recognition unit includes: The node extraction unit is used to identify points with significant gradient changes in multiple propellers based on the gradient accumulation matrix, and define the points as nodes of the transmission chain to obtain a node index set. The torque calculation unit is used to calculate the torque components and direction differences between any adjacent transmission chain nodes based on the node index set, and obtain the torque correlation matrix. The constraint modeling unit is used to construct a set of constraint equations between nodes based on the moment correlation matrix, and to normalize the set of constraint equations to obtain a constraint coefficient table. The calibration generation unit is used to determine the force boundary conditions of each node in the overall power chain of the multi-propeller based on the constraint coefficient table, identify the torque transmission relationship between the multi-propellers, and obtain the power chain constraint table.

5. A multi-propeller power cooperative control system according to claim 4, characterized in that, The index module includes: The power distribution index unit is used to calculate the coupling weight of the power chain as the distance decreases along the path length based on the propeller coupling matrix, thus obtaining the topological coupling term; and to calculate the geometric equilibrium value of the power ratio of any two propellers based on the power ratio of each propeller, thus obtaining the scale ratio term. Based on whether the included angle and rotation direction of any two propeller propulsion axes are consistent, the directional modulation of the alignment degree is calculated to obtain the directional consistency term; the magnitude amplification of the difference in the rate of change of any two propellers is calculated to obtain the dynamic difference term; the topological coupling term, scale ratio term, directional consistency term and dynamic difference term are fused together to calculate the local coupling aggregation term; Based on the local coupling aggregation term, the scale factor of the compressed or stretched numerical range is calculated to obtain the scale shaping term; based on the dispersion of the overall propeller power ratio, the global non-uniformity modification term is calculated; the local coupling aggregation term, the scale shaping term, and the global non-uniformity modification term are fused to obtain the power distribution index.

6. A multi-propeller power cooperative control system according to claim 5, characterized in that, The associated module includes: The region identification unit is used to determine the local intensity regions in the propeller coupling matrix that are higher than the preset power threshold based on the power distribution index, thereby obtaining the intensity region set; The reverse mapping unit is used to map the energy distribution nodes in each intensity region to propeller number indices based on the intensity region set, establish the energy backtracking relationship between propeller pairs, and obtain the backtracking mapping table. The path construction unit is used to connect propeller nodes with energy backtracking relationships sequentially according to the backtracking mapping table to form a continuous power transfer link and obtain a transfer path set; The coordination matrix unit is used to convert the energy ratio of each path segment into matrix entry weights based on the transmission path set, determine the power transmission path between propellers, and obtain the propulsion coordination matrix.

7. A multi-propeller power cooperative control system according to claim 6, characterized in that, The coordination matrix includes: The energy allocation unit is used to quantify the energy difference between the starting and ending nodes of each path segment according to the transmission path set, establish an energy ratio table, and obtain the path energy set. The weight correction unit is used to perform symmetric balancing based on the path energy set, and to perform bidirectional correction and normalization on the weight values ​​of the transmitted paths in the energy ratio table to generate an energy weight matrix. The topology mapping unit is used to map the energy weight distribution to the physical connection topology between propellers based on the energy weight matrix, forming a topology association table; The propulsion coordination matrix unit is used to incorporate the energy weights of each path segment into the matrix according to the physical adjacency relationship based on the topology association table, and to determine the power transfer path relationship between propellers, thus obtaining the propulsion coordination matrix.

8. A multi-propeller power cooperative control system according to claim 7, characterized in that, The factor module includes: The dynamic balance factor unit is used to calculate the instantaneous response difference between each pair of main propellers and auxiliary propellers based on the propulsion coordination matrix, and obtain the time offset term; based on the output power amplitude of each propeller, it compares the power fluctuation direction and amplitude ratio of the main propeller and auxiliary propeller in the same time period, calculates the degree of coordination amplitude between the two, and obtains the amplitude coupling term. Based on the inertia parameters and relative spatial positions of each propeller, the difference in inertial response between the main propeller and the auxiliary propeller is measured, the degree of inertia matching is calculated, and the inertia compensation term is obtained; based on the instantaneous frequency changes of each propeller, the average operating frequency and fluctuation range of the propellers are statistically analyzed, the group frequency consistency among each propeller is calculated, and the frequency synchronization term is obtained. By fusing the timing offset term, amplitude coupling term, inertia compensation term, and frequency synchronization term, the overall synchronization stability among multiple propellers is calculated, and the dynamic balance factor is obtained.

9. A multi-propeller power cooperative control system according to claim 8, characterized in that, The fusion module includes: The energy mapping unit is used to perform time-series expansion based on the dynamic distribution index, establish a mapping relationship between the rate of change of the dynamic distribution index and the energy offset amplitude, and obtain the energy change matrix. The stabilization modulation unit is used to perform frequency domain transformation based on the dynamic balance factor, and couple the amplitude distribution of the dynamic balance factor with the phase stability in different frequency bands to obtain the frequency domain modulation matrix. The dual-domain fusion unit is used to establish a nonlinear cross-mapping between the time and frequency domains based on the energy change matrix and the frequency modulation matrix, extract the resonance region and form a fusion tensor to obtain the energy-stable correlation matrix. The cooperative vector generation unit is used to extract and normalize the principal component directions of the fusion tensor based on the energy-stable correlation matrix, and transform them into a multi-dimensional cooperative vector group to obtain a multi-propeller cooperative vector group.

10. A multi-propeller power cooperative control system according to claim 9, characterized in that, The instruction module includes: The recursive setting unit is used to arrange the cooperative vectors in layers according to the propeller number and direction of action based on the multi-propeller cooperative vector group, establish a multi-level expansion index table, and obtain the recursive parameter set. The power deconstruction unit is used to calculate the vector change rate and energy offset between adjacent layers based on the recursive parameter set, describe the power change trend, and obtain the power gradient matrix. The sequence generation unit is used to expand the power change rate of each propeller into an adjustment sequence on the time axis according to the power gradient matrix, so as to obtain the instantaneous power adjustment sequence set; The instruction construction unit is used to arrange the adjustment sequences of each propeller according to the number and time order based on the instantaneous power adjustment sequence set, so as to obtain the control instruction matrix.

Images