Airplane body self-adaptive load distribution method and system based on digital twin inversion

Abstract

The application discloses an aviation projectile self-adaptive load distribution method and system based on digital twin inversion, and relates to the technical field of load control.The application inverses the real stress of a structure accurately by fusing an agent model and real-time sensing based on digital twin, and solves the model distortion problem; the actual load is taken as a multivariable PID feedback at the decision end, the control target is improved to the structure load space, the mean coupling and fuzzy self-adaptation are combined, and quick response and multi-station intelligent coordination are realized; and the execution end cooperatively fuses instructions through a decoupling matrix, so that local adjustment does not damage global balance.

Description

Technical Field

[0001] This invention relates to the field of load control technology, and in particular to an adaptive load allocation method and system for airborne missiles based on digital twin inversion. Background Technology

[0002] In the fields of airborne missile structure testing and flight load testing, accurate load distribution is a crucial foundation for ensuring the effectiveness of structural test verification. Traditional load calculation methods are based on the lever principle and idealized assumptions, completing mass distribution and support reaction force calculations through preset sites, providing a static theoretical benchmark for the design phase. However, this method is essentially an open-loop, one-time calculation process, unable to respond to load distribution changes caused by manufacturing errors, assembly gaps, and dynamic deformation during testing, creating a gap that is difficult to bridge between theoretical design and physical reality.

[0003] To compensate for the shortcomings of theoretical calculations, digital twin monitoring technology has emerged. This technology, by fusing sensor data with finite element models, enables real-time reconstruction and visualization of the actual strain field of a structure, accurately identifying load deviations caused by factors such as eccentric loading and contact nonlinearity. However, existing digital twin systems are still primarily limited to state perception and diagnosis, forming a "monitoring-only" loop that lacks direct control over the loading system and cannot translate identified deviations into corrective actions. At the load loading level, traditional PID control technology is widely used to drive individual actuators. However, its control loop is usually based on the force or displacement feedback of the actuator itself, rather than the overall stress state of the structure. This actuator-centric control strategy, while ensuring the tracking accuracy of local commands, cannot perceive and correct global load distribution imbalances caused by structural coupling and system deviations, creating a contradictory situation of "precise control but deviation from the target."

[0004] Currently, these three technologies remain independent: static calculation provides the target, digital twins explain the deviation, and PID control executes the instructions. However, they lack an integrated intelligent closed loop that connects them into a "perception-decision-execution" system. Existing technologies have not yet been able to realize an adaptive system that dynamically adjusts the load distribution across multiple sites based on the actual structural response, making it difficult to achieve truly accurate, reliable, and safe load verification for airborne missiles when facing actual engineering deviations. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive load allocation method and system for airborne missiles based on digital twin inversion, in order to solve the technical problem that existing technologies cannot form an integrated intelligent closed loop of "perception-decision-execution", which leads to the inability of load allocation to be adaptively optimized in real time according to the actual response of the structure.

[0006] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0007] An adaptive load allocation method for airborne missile bodies based on digital twin inversion includes:

[0008] Sensor data from various stations on the airborne projectile are collected and the target load vector is calculated. A deviation parameter-full-field strain proxy model is constructed using radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state. The sensor data includes strain gauge data, displacement data, and force data.

[0009] Based on the deviation strain field and sensor data of the current state of each site, the corresponding digital twin deviation strain field is constructed through the bridge function. Combined with the offline calibration load inversion matrix and the interface stiffness weight matrix, the corresponding actual load vector of the current state is generated.

[0010] Calculate the coupling control deviation between the actual load vector and the target load vector for each current state; based on the coupling control deviation, calculate the comprehensive accuracy index; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error;

[0011] Based on the comprehensive accuracy index and the actual load vector of the current state, load allocation is performed through the mean-coupled fuzzy PID collaborative control method to generate load allocation control commands for the current state.

[0012] In the above process, the present invention solves three major technical problems in the allocation of airborne missile loads by constructing a three-in-one closed-loop control architecture of "digital twin inversion - mean coupled fuzzy PID - comprehensive accuracy evaluation", namely, theoretical model distortion, perception and control separation, and multi-site cooperative instability. At the perception level, a digital twin strain field is constructed by fusing a radial basis function surrogate model with a bridge function. Combined with an offline calibration inversion matrix and interface stiffness weights, high-precision inversion of the actual stress state of the structure under limited measurement points is achieved, breaking through the limitations of traditional theoretical calculations that rely on idealized assumptions. At the decision level, the actual load obtained from the inversion is used as the feedback input of the mean-coupled fuzzy PID controller. By weighted fusion of independent deviations and global average deviations, a coupled deviation is formed, which elevates the control target from the actuator space to the structural load space. Combined with a fuzzy rule base, dynamic adaptive tuning of control parameters is achieved, solving the problem of single-site PID control being "precise but local" and unable to respond to global imbalances. At the execution level, control commands from each site are collaboratively fused based on the decoupling matrix identified by the system, actively compensating for coupling interference between sites and avoiding the "pressing down the gourd and the ladle floats up" adjustment oscillations in traditional control. Ultimately, by introducing a comprehensive accuracy index that includes both load magnitude error and global allocation ratio error as a convergence criterion, the system is ensured to maintain dynamic stability while pursuing amplitude accuracy, thus achieving a fundamental leap from "open-loop static calculation" to "closed-loop dynamic tracking" and from "local precise control" to "global collaborative optimization".

[0013] Further, generating the corresponding deviation strain field for the current state includes:

[0014] Sensors are installed in the airborne missile body at a certain site to collect corresponding sensor data;

[0015] The airborne projectile body at a certain site is simplified into a beam model with n mass elements, and the target load vector at the site is calculated.

[0016] Based on sensor data, the rigid body rotation matrix of the loading plane of the airborne projectile is solved by the least squares method;

[0017] Based on the rigid body rotation matrix, the real-time loading deviation parameters of the current state are calculated, and a deviation parameter-full-field strain proxy model is constructed by radial basis function and finite element simulation method to generate the deviation strain field of the current state.

[0018] In the above process, on the one hand, the present invention uses beam model and rigid body transformation to quantify physical deviations into six-dimensional parameters, enabling the finite element proxy model to be accurately invoked for actual working conditions; on the other hand, the present invention transforms online finite element calculation into millisecond-level function query through the "deviation parameter → full field strain" mapping established by radial basis functions, solving the technical problems of inaccurate strain field prediction caused by manufacturing and assembly deviations in traditional methods, and the inability of online simulation time to meet the real-time requirements of closed-loop control, providing a high-fidelity and high-timeliness basic strain field input for subsequent digital twin fusion and load inversion.

[0019] Further, generating the actual load vector corresponding to the current state includes:

[0020] Based on strain gauge data from strain gauge measurement points, the corresponding bridge function factors are calculated using Latin hypercube sampling and radial basis function network.

[0021] A digital twin bias strain field is generated based on the bridge function factor and the bias strain field of the current state.

[0022] Based on sensor data, the mechanical influence area of ​​each site is constructed;

[0023] Based on the material constitutive relation and digital twin bias strain field of the projectile, the element stress distribution is generated, and combined with the finite element simulation method, a six-dimensional load vector for each station is generated.

[0024] Simulate the strain response of each mechanically affected region under different working conditions to generate an offline calibration load inversion matrix; set the interface stiffness weight matrix through the connection unit or contact interface of each mechanically affected region.

[0025] Based on the six-dimensional load vector, offline calibration load inversion matrix, and interface stiffness weight matrix of each site, the actual load vector of the current state of each site is generated.

[0026] In the above process, this invention, based on Latin hypercube sampling and radial basis function network calculation of bridge function factors, dynamically fuses the measured strain with the deviation strain field predicted by the surrogate model to generate a high-fidelity digital twin deviation strain field, solving the technical problem that relying solely on simulation models cannot reflect real manufacturing and assembly deviations. Secondly, by constructing the mechanical influence region of each site, the global strain field is reduced to a local region according to the force transmission path, which reduces the scale of inversion calculation and avoids interference from unrelated strain noise on load identification. Thirdly, by adopting an offline calibration method, the load inversion matrix is ​​constructed by simulating the strain response of each influence region under different unit load conditions, transforming the complex mechanical inverse problem into online matrix multiplication, realizing millisecond-level mapping from strain to load, and meeting the real-time requirements of closed-loop control. Finally, an interface stiffness weight matrix is ​​introduced to weight and correct the strain components according to the equivalent stiffness of the connecting unit or contact interface, effectively compensating for load transmission path deviations caused by factors such as bolt loosening, contact state changes, and sudden changes in local stiffness, extending the linear mapping relationship from an idealized uniform assumption to a non-uniform weighted mapping that conforms to the actual physical force transmission characteristics.

[0027] Furthermore, the overall accuracy index for calculation includes:

[0028] Based on the actual load vector and target load vector of each current state, calculate the independent load deviation vector of each station in the current state;

[0029] Calculate the global deviation vector based on the deviation vectors of each independent load;

[0030] Based on the global deviation vector and the independent load deviation vector of each station, the coupling control deviation of each station is calculated.

[0031] Based on the coupling control deviation, actual load vector, and target load vector of each station, and combining the sum of the actual main loads of each station with the ratio of the actual loads of each station, the comprehensive accuracy index is calculated.

[0032] Furthermore, the calculation of the comprehensive accuracy index by combining the sum of the actual main loads of each site with the actual load ratio of each site includes:

[0033] Based on the site load magnitude error, coupling control deviation, actual load vector, and target load vector of each site, the corresponding site load magnitude error is calculated;

[0034] Based on the main load components of each site, the sum of the actual main loads of the sites is calculated, and combined with the theoretical load ratio, the actual load ratio of each site is generated.

[0035] The global allocation ratio error is calculated based on the theoretical load ratio, actual load ratio, and coupling control deviation of each station.

[0036] The overall accuracy index is calculated based on the site load magnitude error and the global allocation ratio error of each site.

[0037] Furthermore, the formula corresponding to the independent load deviation vector is:

[0038] ;

[0039] in, Indicates the first Target load vector of each site Indicates the first Current status of each site The actual load vector, Indicates the first Independent load deviation vectors for each site;

[0040] The formula corresponding to the global deviation vector is:

[0041] ;

[0042] in, Represents the global deviation vector. Indicates the total number of sites. This represents the summation function;

[0043] The formula corresponding to the coupling control deviation is:

[0044] ;

[0045] in, Indicates the first Current status of each site Coupling control deviation, Indicates the first Site weight coefficient for each site.

[0046] In the aforementioned process, this invention constructs a three-tiered progressive error calculation system of "independent deviation - global deviation - coupled deviation," and integrates the site load magnitude error and the global allocation ratio error to form a comprehensive accuracy index, achieving a refined and multi-dimensional evaluation of the load allocation status of multiple sites. On the one hand, the weighted fusion of the independent deviation of each site and the global average deviation yields the coupled control deviation, enabling the evaluation index to reflect both the degree of deviation at local sites and the overall load imbalance state. On the other hand, load amplitude error and allocation ratio error are simultaneously introduced, and the dynamic statistical characteristics of the coupled deviation are used to correct both, allowing the comprehensive accuracy index to automatically amplify during control oscillations and provide timely warnings when the ratio is out of balance. This scheme solves the technical shortcomings of traditional load evaluation methods that only focus on single-site amplitude error and cannot identify proportional imbalances and control dynamic stability issues among multiple sites. It provides an accurate convergence criterion for mean-coupled fuzzy PID control, effectively guiding the generation and termination decisions of control commands.

[0047] Furthermore, the load allocation control command for generating the current state includes:

[0048] Determine whether the overall accuracy index is less than the preset convergence threshold; if so, use the load allocation control command of the previous state as the load allocation control command of the current state.

[0049] Conversely, based on the actual load vector of the current state, the standardized deviation and rate of change of each station under different degrees of freedom in the current state are generated;

[0050] The standardized deviation and rate of change of each station under the current state are used as the input data of the fuzzy PID controller, and the control parameters of the fuzzy PID controller are adjusted based on the fuzzy rule base to obtain the updated control parameters.

[0051] Based on the updated control parameters, calculate the control output for each station under different degrees of freedom;

[0052] Integrate the control outputs of all stations and perform coordinated decoupling through a decoupling matrix to generate the final coordinated control increment;

[0053] By combining the final coordinated control increment and the load allocation control command from the previous state, the load allocation control command for the current state is generated.

[0054] In the above process, this invention uses the comprehensive accuracy index as the control entry point and achieves adaptive switching of "achieving the target means steady state, and failing to achieve the target means adjustment" by setting a convergence threshold, thus avoiding ineffective adjustment and system oscillation. In the decision-making stage, the standardized deviation and rate of change of different degrees of freedom of each station are calculated based on the actual load vector of the current state, and used as the input of the fuzzy PID controller. The control parameters are tuned online based on the fuzzy rule base, enabling the controller to have the adaptive ability to optimize its own performance in real time according to dynamic working conditions, solving the technical problems of fixed parameters and difficulty in dealing with uncertainties such as manufacturing deviations and nonlinear coupling in traditional PID controllers. In the execution stage, the control outputs of all stations are collaboratively fused through the decoupling matrix obtained by system identification to generate the final collaborative control increment, which is superimposed with the load allocation control command of the previous state to form the current command. This incremental collaborative architecture not only ensures the smooth continuity of control actions, but also actively compensates for coupling interference between multiple stations, fundamentally solving the contradiction of "local adjustment causing global instability" and "precise control but deviation from the target" in traditional control, and ultimately achieving rapid convergence, smooth transition and accurate matching of load allocation.

[0055] An adaptive load distribution system for airborne missile bodies based on digital twin inversion includes:

[0056] The strain field generation module is used to collect sensor data from various stations on the airborne projectile and calculate the target load vector. It constructs a deviation parameter-full-field strain proxy model through radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state.

[0057] The load calculation module is used to construct the corresponding digital twin deviation strain field based on the deviation strain field and sensor data of the current state of each site through the bridge function, and generate the corresponding actual load vector of the current state by combining the offline calibration load inversion matrix and the interface stiffness weight matrix.

[0058] The index calculation module is used to calculate the coupling control deviation between the actual load vector and the target load vector at each current state; based on the coupling control deviation, it calculates the comprehensive accuracy index; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error;

[0059] The load allocation module is used to allocate loads based on a comprehensive accuracy index and the actual load vector of the current state, using a mean-coupled fuzzy PID collaborative control method, and to generate load allocation control commands for the current state. (See attached diagram.)

[0060] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0062] Figure 2 This is a flowchart of the deviation strain field generation process in Embodiment 1 of the present invention;

[0063] Figure 3 This is a system structure diagram in Embodiment 1 of the present invention;

[0064] Figure 4 This is a structural diagram of the aircraft missile body in Embodiment 2 of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0066] Please see Figure 1 This embodiment provides an adaptive load allocation method for airborne missile bodies based on digital twin inversion. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0067] An adaptive load allocation method for airborne missile bodies based on digital twin inversion includes:

[0068] S1. Collect sensor data from each station on the airborne missile body and calculate the target load vector. Construct a deviation parameter-full-field strain proxy model using radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state (current control cycle). The sensor data includes strain gauge data, displacement data, and force data. The types of stations include suspended, contact, and distributed.

[0069] Taking a site as an example, S1 includes:

[0070] S1-1. Sensors are installed in the airborne projectile at a certain site to collect corresponding sensor data; the sensors include strain gauge array sensors, displacement sensors and force sensors.

[0071] In this embodiment, strain gauge array sensors cover the main load-bearing areas of the aircraft body. Displacement sensors are arranged at the loading interface of the aircraft body. Force sensors are installed at the output ends of each actuator of the aircraft body. In addition, by selecting three non-collinear reference points on the aircraft body, the transformation relationship from the local coordinate system to the global coordinate system O-XYZ of each sensor is established. This refers to the local-to-global coordinate transformation relationship, ensuring coordinate consistency in subsequent sensor data fusion. Represents the rotation matrix. O represents the translation vector, and X, Y, and Z represent the horizontal, vertical, and y axes of the global coordinate system, respectively.

[0072] S1-2. Simplify the airborne missile body at a certain site into a beam model with n mass elements, and calculate the target load vector at the site.

[0073] Specifically, based on the "Calculation Method for Flight Load of Aerial Bombs," the theoretical load is allocated, simplifying the bomb body into a beam model containing n mass elements. Taking one mass element as an example, the element mass is... The centroid coordinates are Among them, the "Calculation Method for Flight Load of Aerial Bombs" is a standard document of the People's Republic of China's weaponry industry.

[0074] According to the formula:

[0075] ;

[0076] ;

[0077] Calculate the first Target payload vector of each site .in, This represents the summation function. Represents gravitational acceleration. Represents the coordinates of the pivot point. Indicates the first The equivalent lever arm length of each station, This represents the safety factor (usually taken as 1.5). Represents the transpose matrix. , , These represent the force vectors in three directions (X, Y, Z). , , These represent the torque vectors in the three directions, respectively.

[0078] S1-3. Based on sensor data, the rigid body rotation matrix of the loading plane of the airborne projectile is solved by the least squares method.

[0079] Specifically, let the coordinates of the measuring point before the airborne projectile is loaded be... The coordinates of the measured points after loading are The rigid body transformation of the loading plane is solved using the least squares method, i.e., according to the formula:

[0080] ;

[0081] Calculate the rigid body rotation matrix .in, This indicates the total number of measuring points (the number of displacement sensors). , , These represent the first measuring point, the second measuring point, and the third measuring point, respectively. The coordinates of each measuring point before loading. , , These represent the first measuring point, the second measuring point, and the third measuring point, respectively. The coordinates of each measuring point after loading. Represents the translation vector of a rigid body. Describes the minimum value function. It represents the sum of squares.

[0082] S1-4. Based on the rigid body rotation matrix, calculate the real-time loading deviation parameters of the current state, and construct a deviation parameter-full-field strain proxy model through radial basis function and finite element simulation method to generate the deviation strain field of the current state.

[0083] Specifically, first, the rigid body rotation matrix... Decomposed into Euler angles, the real-time loading deviation parameters are obtained. The corresponding expression is:

[0084] ;

[0085] in, , , These represent the rotation angles of the projectile around the X-axis, Y-axis, and Z-axis, respectively. , , These represent the translation amounts in the three directions, respectively.

[0086] like Figure 2 As shown, a high-precision finite element parametric model (i.e., deviation parameter-full-field strain proxy model) of the airborne missile body is established by finite element simulation, which includes geometric structure, material properties, boundary conditions and connection relationships.

[0087] In the finite element simulation method, Latin hypercube sampling is used to generate 90 representative sample points in the deviation parameter space. , equivalent to Each set of sample points includes A deviation parameter is used to ensure uniform sample coverage of the parameter range. Each set of sample points is input into the deviation parameter-full-field strain surrogate model for static simulation calculations. The strain value at each finite element integration point is extracted to obtain the full-field strain distribution corresponding to that sample point. The strain values ​​of each sample point in all finite elements are recorded and used as input-output data pairs for that finite element. (The last sentence appears to be incomplete and requires further context.) Taking a finite element as an example, with each set of sample points as input and the corresponding strain value as output, a deviation parameter-full-field strain surrogate model is constructed using radial basis functions (RBF). The corresponding formula is:

[0088] ;

[0089] in, Represents the spatial coordinates of a finite element. Indicates the first The deviation strain field corresponding to each finite element element. Indicates the number of sample points. Indicates the first The weighting coefficients of each finite element. Represents the Gaussian radial basis functions. Indicates the first A loading deviation parameter with sample points This represents the Euclidean norm of a vector.

[0090] The weighting coefficients for each sample point are determined by solving the deviation parameter-full-field strain surrogate model, thus obtaining the mapping relationship of the finite element. ).

[0091] The above process is then repeated for all finite element elements to form a complete mapping function library, which includes the deviation parameter-full-field strain surrogate model corresponding to each sample point of each finite element. For subsequent use, the corresponding deviation parameter-full-field strain surrogate model is directly called to obtain the corresponding deviation strain field.

[0092] Based on real-time loading deviation parameters By calling the corresponding deviation parameter-full-field strain surrogate model in the mapping function library, the deviation strain field under the current state can be obtained. .

[0093] exist Figure 2 middle, , , These represent the first sample point, the second sample point, and the last sample point, respectively. Indicates the first The first sample point Each deviation parameter, and so on, can be used to determine... , , The meaning of parameters such as... Indicates the first The first sample point The deviation strain field of each finite element, and so on, can be deduced. , , The meaning of parameters such as...

[0094] S2. Based on the deviation strain field and sensor data of the current state of each station, construct the corresponding digital twin deviation strain field through the bridge function, and generate the corresponding actual load vector of the current state by combining the offline calibration load inversion matrix and the interface stiffness weight matrix.

[0095] Taking a single site as an example, S2 includes:

[0096] S2-1. Based on strain gauge data from strain gauge measurement points, the corresponding bridge function factor is calculated using Latin hypercube sampling and radial basis function network. The corresponding formula is as follows:

[0097] ;

[0098] in, Indicates the first Each strain gauge measurement point Indicates the first Bridge function factor corresponding to each strain gauge measurement point Indicates the first Strain gauge data at each strain gauge measurement point Indicates the first The deviation strain field corresponding to each strain gauge measuring point.

[0099] It should be explained that the bridge function method is characterized by its simplicity, high accuracy, and high construction efficiency, making it a widely used data fusion method.

[0100] To achieve rapid mapping from deviation parameters to the full-field strain response, a surrogate model is established using Latin hypercube sampling and radial basis function networks. At the measurement point of the strain gauge (i.e., the location of the strain gauge sensor), the residual between the deviation strain field at that location and the strain gauge data is calculated to form a discrete bridge function factor. A continuous bridge function correction field is then constructed using spatial interpolation. .

[0101] S2-2. Based on the bridge function factor and the deviation strain field of the current state, generate a digital twin deviation strain field;

[0102] Specifically, the deviation strain field of the current state is linearly superimposed with the bridge function factor to generate a digital twin deviation strain field covering the entire structure, thereby enabling real-time and reliable prediction of strain values ​​at any coordinate position.

[0103] The formula corresponding to S2-2 is:

[0104] ;

[0105] in, This represents the bias strain field of the digital twin.

[0106] S2-3. Based on sensor data, construct the mechanical influence area for each site;

[0107] Specifically, a corresponding mechanical influence region is established for each site in the high-precision finite element parametric model. For the first... For each site, the set of finite element elements that have a direct force transmission path to that site is determined in the high-precision finite element parametric model, and this set is used as the mechanical influence region of that site. Mechanically Influenced Area This typically includes shell elements near the loading nodes, connecting structural elements, and fasteners or contact elements involved in load transfer. This approach allows for the identification of the primary load transfer paths within the structure at each site in the digital twin model, thus establishing a stable mechanical foundation for subsequent load inversion calculations.

[0108] S2-4. Based on the material constitutive relation and digital twin bias strain field of the projectile, the element stress distribution is generated, and combined with the finite element simulation method, a six-dimensional load vector for each station is generated; the material constitutive relation includes the anisotropy, viscoelasticity, and strain rate effect of the projectile structure;

[0109] It needs to be explained that the mechanical influence area at a certain site... Within this region, operations such as nodal force synthesis and stress integration are performed on the finite element elements.

[0110] According to the formula:

[0111] ;

[0112] Transform the digital twin's biased strain field into a single-element stress distribution. ;in, , Let represent the isotropic constitutive tensor and the anisotropic correction tensor of the finite element in the high-precision finite element parameterized model, respectively. Represents elastic strain components, This indicates the integration operation. The fundamental constitutive tensor of a material. This indicates tensor dual composite operations. Represents elastic strain rate. Represents the constitutive tensor of the material. This indicates the initial stress or initial strain term.

[0113] The element stress distribution is converted into the nodal equivalent load of each node under each finite element. Considering integrating along the force point region to the internal contour of the polygonal micro-element shape, this embodiment uses an interpolated strain-compatible model. The modified strain energy metric and Pettov-Galerkin optimization framework are positive. The classical nodal integral is extended to a multi-scale yield surface enhancement form, that is, the formula corresponding to the nodal equivalent load of each node is converted into the element stress distribution:

[0114] ;

[0115] in, This represents the total number of Gaussian integration points. Indicates the first The equivalent load of each node. Indicates summation and multiplication. This represents the Gaussian product weights. Represents the coordinates of the Gaussian integration point. express The shape function value, Represents the unit tensor or correction factor. This represents the tensor product operation. Represents the generalized strain-displacement operator. Indicates the first Each node at the integration point Displacement value at that location, Represents the element stress distribution (i.e., element stress distribution) ), This represents the weighted residual correction factor. This represents the Jacobian determinant.

[0116] To accurately obtain the true stress state of the loading point in the structure, this invention employs the nodal force synthesis method, selecting a mechanical reference point. The equivalent loads at the nodes of all finite element units in each mechanically affected region are vector-synthesized to calculate the total moment in the mechanically affected region. In this process, the spatial non-uniformity of shear stress distribution, material anisotropy, and interactions between multilayer plate structures are fully considered, and a yield quantum correction term is introduced. This extends the classical composite topology formula into a unified expression that incorporates the effects of volume forces, boundary stress contributions, and corrections for higher-order terms, thereby enabling accurate load synthesis for complex structures under non-uniform stress fields. Therefore, the formula for the total moment in the mechanically affected region is:

[0117] ;

[0118] in, Indicates the first The total torque in the mechanically affected area Indicates the first The position vectors of each node. Represents the yield quantum correction term an approximate quantity, Represents a generalized external product operator with a range of magnitude. Indicates the first The nodal force vector of each node. To represent a triple integral, This represents the position vector of the integration point. Represents the stress divergence term. It represents a combination of terms in a higher-order polynomial.

[0119] Then, based on the vector sum of the nodal forces in each mechanical influence region, determine the corresponding resultant force, and the corresponding formula is:

[0120] ;

[0121] in, Indicates the first The combined efforts of each site Indicates the first Force in each direction.

[0122] Then, based on the resultant force and the total moment, a six-dimensional load vector is constructed for each station, namely:

[0123] ;

[0124] in, Indicates the first The six-dimensional load vector of each site, , , They represent the first The force of a station in three directions, , , They represent the first The torque of a station in three directions.

[0125] S2-5. Simulate the strain response of each mechanically affected region under different working conditions and generate an offline calibration load inversion matrix; set the interface stiffness weight matrix through the connection unit or contact interface of each mechanically affected region.

[0126] It should be explained that in the high-precision finite element parametric model, a unit six-dimensional load (three forces and three moments) is applied sequentially to the target site, and the mechanical influence area is calculated under each working condition. The strain response within the load is obtained by solving the linear relationship between the load and the strain response using the least squares method, yielding the inversion matrix. .

[0127] Considering that the load transfer path in complex connection structures may be affected by differences in connection stiffness or changes in contact state, this embodiment introduces an interface stiffness weight matrix. The interface stiffness weight matrix can be determined based on the mechanically influential region. The equivalent contact stiffness of the connecting or contacting elements within the interface (obtainable through finite element contact analysis, stiffness testing, or empirical formulas) is used to obtain the weighting coefficients of each finite element in the load transfer process. Finite element elements with higher equivalent contact stiffness indicate stronger load transfer capacity and should be assigned higher weights in load inversion. All weighting coefficients are then integrated to form the interface stiffness weighting matrix. .

[0128] S2-6. Based on the six-dimensional load vector of each station, the offline calibration load inversion matrix, and the interface stiffness weight matrix, generate the actual load vector of the current state of each station.

[0129] Specifically, to achieve rapid solution for load inversion calculations, this embodiment expresses the relationship between the digital twin strain field and the site load as a matrix operator. The mechanical influence area... The strain information within is combined in a certain order to form a regional strain vector. Then, according to the formula:

[0130] ;

[0131] Construct an inversion-linear mapping relationship between the digital twin strain field and the site load.

[0132] However, the inversion-linear mapping relationship is based on the idealized linear elastic continuum assumption and cannot reflect the non-uniform load transfer caused by complex connection factors such as bolted connections, contact interfaces, and multi-layer plate structures in actual airborne missile bodies—the contribution weight of strain to the load at different sites varies due to differences in local stiffness. Therefore, an interface stiffness weight matrix is ​​introduced to improve the inversion-linear mapping relationship, and the corresponding formula is:

[0133] ;

[0134] In this embodiment, by constructing an interface stiffness weight matrix, each strain component is weighted and corrected according to the equivalent stiffness of the connecting unit or contact interface. This extends the uniform linear mapping into a non-uniform mapping that "distributes weights according to the force transmission path," thus transforming the inversion relationship into... This improvement significantly enhances the accuracy and stability of load inversion in complex connection structures, effectively compensating for transmission path deviations caused by factors such as bolt loosening, changes in contact state, and sudden changes in local stiffness, ensuring that the inversion results can truly reflect the actual stress state of the structure.

[0135] The improved inversion-linear mapping relationship is used to generate the actual load vectors of different stations under the current state.

[0136] S3. Calculate the coupling control deviation between the actual load vector and the target load vector for each current state; calculate the comprehensive accuracy index based on the coupling control deviation; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error;

[0137] S3 includes:

[0138] S3-1. Based on the actual load vector and target load vector of each current state, calculate the independent load deviation vector of each station in the current state. The corresponding formula is:

[0139] ;

[0140] in, Indicates the first Target load vector of each site Indicates the first Current status of each site The actual load vector, Indicates the first Independent load deviation vectors for each site.

[0141] S3-2. Based on the independent load deviation vectors, calculate the global deviation vector. The corresponding formula is:

[0142] ;

[0143] in, Represents the global deviation vector. This indicates the total number of sites.

[0144] S3-3. Based on the global deviation vector and the independent load deviation vector of each station, calculate the coupled control deviation of each station. The corresponding formula is:

[0145] ;

[0146] in, Indicates the first Current status of each site Coupling control deviation, Indicates the first Site weight coefficient for each site. The range of values ​​is .

[0147] S3-4. Based on the coupling control deviation, actual load vector, and target load vector of each station, and combining the sum of the actual main loads of each station with the actual load ratio of each station, calculate the comprehensive accuracy index.

[0148] S3-4 includes:

[0149] S3-4-1. Based on the site load magnitude error, coupling control deviation, actual load vector, and target load vector of each site, calculate the corresponding site load magnitude error;

[0150] It needs to be explained that traditional site load magnitude error only focuses on the amplitude deviation between the actual load vector and the theoretical target load vector. This single-dimensional evaluation method cannot reflect the dynamic stability of the control process. Even if the load amplitude deviation is small at a certain moment, if the control system is in a state of violent oscillation (when the coupling control error is large), the actual loading may still become unstable, causing the structure to be subjected to unexpected dynamic load impacts, threatening the safety of the test and the reliability of the results. Therefore, this embodiment introduces the statistical characteristics of coupling control error into the evaluation index, incorporating dynamic stability into the assessment. When the coupling error is small, the exponential term approaches 1, and the index mainly reflects the amplitude deviation; when the coupling error is large, the exponential term increases significantly (note that the exponent sign is negative), thus automatically amplifying the comprehensive error, forcing the control system to suppress oscillations while pursuing load amplitude accuracy, ensuring that the load distribution process is both "accurate" and "stable".

[0151] Therefore, taking a single site as an example, the formula corresponding to the site load magnitude error is:

[0152] ;

[0153] in, Indicates the first The site load magnitude error (used to measure the degree of deviation in amplitude between the actual load vector and the theoretical target load vector at each loading site). Let represent the logarithmic function with base e. Indicates the current state Next Independent load deviation vectors for each site The The load deviation component of each degree of freedom.

[0154] S3-4-2. Based on the main load components of each station, calculate the total actual main load of each station, and combine it with the theoretical load ratio to generate the actual load ratio of each station.

[0155] It needs to be explained that the first Current status of each site The main load component is set as ( According to the formula:

[0156] ;

[0157] Calculate the total actual main load of the site .

[0158] Then, according to the formula:

[0159] ;

[0160] Calculate the theoretical load ratio .in, Indicates the first The target primary payload components of each site.

[0161] Based on the actual sum of the main loads and the main load components of the site, calculate the first... Actual load ratio of each site The corresponding formula is:

[0162] .

[0163] S3-4-3. Calculate the global allocation ratio error based on the theoretical load ratio, actual load ratio, and coupling control deviation of each station.

[0164] It should be explained that, in order to simultaneously consider the impact of control error on proportional stability, a coupling error term is introduced based on the original proportional error index. Therefore, the formula corresponding to S3-4-3 is:

[0165] ;

[0166] in, This indicates the global allocation ratio error. Indicates the number of degrees of freedom.

[0167] The global allocation ratio error not only reflects the degree of deviation between the load ratio of multiple sites and the theoretical allocation, but also reflects the stability of the multi-site collaborative control process through the control error term.

[0168] S3-4-4. Based on the site load magnitude error and global allocation ratio error of each site, calculate the comprehensive accuracy index. The corresponding formula is:

[0169] ;

[0170] in, This indicates the overall accuracy index. , These represent the site error weights and the global weights, respectively. This represents the maximum value function.

[0171] S4. Based on the comprehensive accuracy index and the actual load vector of the current state, load allocation is performed through the mean-coupled fuzzy PID collaborative control method to generate the load allocation control command for the current state.

[0172] S4 includes:

[0173] S4-1. Determine whether the overall accuracy index is less than the preset convergence threshold; if so, use the load allocation control command of the previous state as the load allocation control command of the current state; otherwise, proceed to S4-2.

[0174] S4-2. Based on the actual load vector in the current state, generate the standardized deviation and rate of change of each station under different degrees of freedom in the current state. The corresponding formula is:

[0175] ;

[0176] ;

[0177] in, Indicates the first The first site Standardization bias of each degree of freedom Indicates the first The first site The rate of change of each degree of freedom , These represent the preset minimum and maximum values ​​of the coupling control deviation, respectively. Indicates the sampling time interval. Indicates the previous state Next Independent load deviation vectors for each site The The load deviation component of each degree of freedom.

[0178] S4-3. The standardized deviation and rate of change of each station under the current state are used as input data for the fuzzy PID controller, and the control parameters of the fuzzy PID controller are adjusted based on the fuzzy rule base to obtain the updated control parameters; the updated control parameters include the proportional coefficient of the fuzzy PID controller. Integral coefficient Differential coefficients .

[0179] Specifically, S4-3 is a mean-coupled fuzzy PID collaborative control method. First, the standardized deviations and rates of change of different degrees of freedom of each station in the current state are used as input variables and fed into the fuzzy PID controllers of each degree of freedom of the corresponding station. Then, fuzzification processing is performed according to a preset fuzzy rule base: the precise input values ​​are mapped to preset fuzzy sets (e.g., {negative large (NB), negative medium (NM), negative small (NS), zero (ZO), positive small (PS), positive medium (PM), positive large (PB)}) through membership functions (such as triangular or Gaussian functions), and the probability of each input value belonging to each fuzzy set is calculated, completing the conversion from precise to fuzzy quantities. On this basis, the proportional coefficient, integral coefficient, and derivative coefficient are adjusted online based on the fuzzy inference mechanism. The output fuzzy quantity obtained by fuzzy inference is converted into precise values ​​using the centroid method (i.e., taking the centroid of the area enclosed by the membership function curve and the horizontal axis), obtaining the incremental values ​​of the proportional coefficient, integral coefficient, and derivative coefficient, which are then superimposed with the initial values ​​to generate the actual control parameters for the current control cycle.

[0180] Therefore, this method weights and fuses the independent load deviations of each station with the global average deviation to form a "coupled deviation" that simultaneously reflects local demand and overall balance. This "coupled deviation" serves as the control input, and its rate of change and the coupled deviation itself are used as input variables for the fuzzy PID controller. This allows the control system to not only perceive the load deviation of individual stations but also understand the global imbalance state of the entire loading system. Based on this, a pre-set fuzzy rule base (e.g., "if the deviation is large and the rate of change is negative, then significantly increase the proportional gain") is used to adjust the proportional, integral, and derivative parameters of each PID controller in real time and intelligently, enabling the controller to adaptively optimize its control performance online according to dynamic operating conditions. Simultaneously, the decoupling matrix obtained through system identification is used to collaboratively fuse the output commands of the controllers at each station, actively compensating for coupling interference between multiple stations. The advantages of this method are as follows: On the one hand, the fuzzy adaptive mechanism enables the system to cope with load deviation changes caused by uncertainties such as manufacturing tolerances and assembly gaps, without the need for repeated manual parameter tuning, which significantly improves the robustness and adaptability of the control; on the other hand, the combined effect of collaborative decoupling and mean coupling ensures the global stability during the multi-site load adjustment process, avoiding the coupling oscillation problem of "pressing down the gourd and the ladle floats up" in traditional control, and finally achieving rapid convergence, smooth transition and precise matching of load distribution.

[0181] S4-4. Based on the updated control parameters, calculate the control output for each station under different degrees of freedom. That is, substitute the updated control parameters into the classic PID incremental formula to calculate the control output under that degree of freedom.

[0182] S4-5. Integrate the control outputs of all stations and perform coordinated decoupling through a decoupling matrix to generate the final coordinated control increment;

[0183] It needs to be explained that, according to the formula:

[0184] ;

[0185] Integrate the control outputs of all stations to generate the overall control vector. .in, , These represent the control outputs of the first and sixth degrees of freedom at the first station, respectively. , They represent the first The control outputs for the first and sixth degrees of freedom of each station.

[0186] During the initial calibration phase, standard test signals are sequentially input to each degree of freedom of each actuator of the projectile, and response data of all load sensors are collected synchronously. A dynamic coupling model between the stations is constructed through a multivariable system identification algorithm, and the generalized inverse of its steady-state gain matrix is ​​extracted as the decoupling matrix.

[0187] Then, according to the formula:

[0188] ;

[0189] Calculate the final cooperative control increment Ultimately, collaborative control increment The vector, composed of control increments calculated independently for all stations and all degrees of freedom, reflects the regulation requirements of each station's own control loop. This represents the decoupling matrix.

[0190] S4-6. Combine the final collaborative control increment and the load allocation control command of the previous state to generate the load allocation control command for the current state.

[0191] In an incremental control architecture, the final collaborative control increment Essentially, it's an adjustment calculated based on the current deviation, not an absolute command value. If... Using the load allocation control command directly as the current state is equivalent to discarding the load foundation established in the previous state. This not only leads to drastic changes in control actions, disrupting system stability, but also fails to inherit historical control effects, making it difficult to achieve smooth and continuous load tracking across multiple sites. Therefore, superimposing the load allocation control command from the previous state with the current collaborative control increment to generate the load allocation control command for the current state ensures both the cumulative and continuous nature of control, and ensures that the adjustment process strictly follows the engineering logic of "gradually correcting based on the existing load." This guarantees that the load at multiple sites can smoothly and monotonically converge to the theoretical target.

[0192] Therefore, the formula corresponding to S4-6 is:

[0193] ;

[0194] in, Load distribution control commands indicating the current state. This indicates the load distribution control command for the previous state.

[0195] In summary, this invention constructs a complete integrated intelligent closed-loop system encompassing perception, decision-making, and execution, achieving a fundamental leap from traditional open-loop static calculation to closed-loop dynamic tracking in airborne missile load allocation. At the perception level, by integrating finite element proxy models with real-time sensor data using digital twin technology, the true stress state of each structural station is accurately inverted, solving the model distortion problem caused by manufacturing deviations and assembly errors in traditional theoretical calculations. At the decision-making level, the actual load obtained through inversion is innovatively used as the feedback input of a multivariable PID controller, elevating the control objective from the actuator space to the structural load space. Combined with a mean-value coupling mechanism and a fuzzy adaptive algorithm, the control system can quickly respond to dynamic deviations and intelligently coordinate the coupling effects of multiple stations. At the execution level, the decoupling matrix obtained through system identification enables the coordinated fusion of multi-channel commands, ensuring that local adjustments do not disrupt global balance.

[0196] like Figure 3 As shown, an adaptive load distribution system for airborne missile bodies based on digital twin inversion includes:

[0197] The strain field generation module is used to collect sensor data from various stations on the airborne projectile and calculate the target load vector. It constructs a deviation parameter-full-field strain proxy model through radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state.

[0198] The load calculation module is used to construct the corresponding digital twin deviation strain field based on the deviation strain field and sensor data of the current state of each site through the bridge function, and generate the corresponding actual load vector of the current state by combining the offline calibration load inversion matrix and the interface stiffness weight matrix.

[0199] The index calculation module is used to calculate the coupling control deviation between the actual load vector and the target load vector at each current state; based on the coupling control deviation, it calculates the comprehensive accuracy index; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error;

[0200] The allocation module is used to allocate load based on the comprehensive accuracy index and the actual load vector of the current state, and to generate load allocation control commands for the current state through the mean-coupled fuzzy PID collaborative control method.

[0201] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0202] Example 2:

[0203] This embodiment is based on Embodiment 1.

[0204] Using a certain type of aircraft missile body as the test object, its structure is as follows: Figure 4 As shown, the projectile has a total length of 4000mm and is equipped with four stations. The target load vector value of station 1 is 3940N, that of station 2 is 3483N, that of station 3 is 3110N, and that of station 4 is 2748N.

[0205] In this embodiment, three sets of comparative experiments were set up: control group A adopted traditional open-loop control, which was statically loaded based on theoretical load without real-time adjustment; control group B simply connected digital twin monitoring and PID control, and manually intervened to adjust the command after monitoring and displaying deviations; the experimental group adopted the method of Example 1. The experiment used an actual manufactured aircraft body, whose objectively existing engineering deviations include the centroid position offset caused by the manufacturing tolerance of each section (maximum offset ±8mm), the nonlinearity of connection stiffness caused by assembly gaps, the batch difference of material elastic modulus (the measured value is 0.93~1.06 times the theoretical value), and the loading point offset caused by the fixture positioning error (about ±2mm). The above deviations are all inherent objective existences of the actual structure, reflecting the inherent gap between the theoretical model and physical reality.

[0206] Loading was performed according to the target load vector values ​​(3940N, 3483N, 3110N, 2748N) at each station. When the projectile was loaded to 80% of the target value, the test system initiated closed-loop adjustment: strain gauge array (strain gauge data), displacement sensor, and force sensor collected data in real time; the displacement data was used to calculate the current loading deviation parameters; the basic strain field was obtained by querying the finite element database containing the deviation; and the digital twin strain field was generated by fusing it with the measured strain. The actual load vectors at each station were then obtained through inversion. The actual load vector values ​​corresponding to stations 1 to 4 were 3895N, 3715N, 3058N, and 2812N, respectively, with deviations from the theoretical values ​​of -1.1%, +6.7%, -1.7%, and +2.3%, respectively. Subsequently, the mean-coupled fuzzy PID controller calculated the independent deviations and global average deviations at each station, forming a coupled deviation as the control input. The fuzzy rule base was used to tune the PID parameters online, and the decoupling matrix was used to fuse and generate collaborative control commands to drive the actuators to adjust collaboratively. After four iterations of approximately 2.5 seconds, the load error at each station decreased to within 3%. The convergence condition was met at the 18th control cycle of 3.6 seconds, and the system entered a steady state.

[0207] Evaluation indicators Control group A Control group B This invention Site load size error 16.5% (Site 2) 8.7% (Site 2) 2.1% (Site 3) Global allocation ratio error 11.3% 6.2% 1.5% Adjusting the convergence time none 152s 3.2s Comprehensive accuracy indicators 14.2% 7.3% 1.9%

[0208] As shown in Table 1, the maximum station load error of control group A was 16.5% (station 2), the global allocation ratio error was 11.3%, and the comprehensive accuracy index was 14.2%; the maximum station load error of control group B was 8.7%, the global allocation ratio error was 6.2%, the adjustment convergence time was 152s, and the comprehensive accuracy index was 7.3%; the maximum station load error of the experimental group was 2.1%, the global allocation ratio error was 1.5%, the adjustment convergence time was 3.2s, and the comprehensive accuracy index was 1.9%. Although control group A was loaded strictly according to the theoretical load, due to the failure to consider objective engineering deviations such as manufacturing tolerances and assembly gaps, the overall stress state of the projectile deviated significantly from the design target, verifying the "difficult-to-bridge gap between theoretical design and physical reality" mentioned in the background technology. Although control group B could detect deviations through digital twins, manual adjustment was lagging and difficult to coordinate the coupling effects of multiple stations, the adjustment process was time-consuming, and the accuracy was still not ideal. The experimental group used the actual load inverted in real time by the digital twin as PID feedback, and automatically completed the multi-site coordinated adjustment within 3.2s through mean-coupled coordinated control. The overall accuracy index of the present invention is the smallest, which fully proves that the present invention effectively compensates for the inherent deviation between theoretical calculation and physical reality. It has significant advantages in load allocation accuracy, response speed and coordinated capability.

[0209] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0210] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for adaptive load allocation of airborne missile bodies based on digital twin inversion, characterized in that, include: Sensor data from various stations on the airborne projectile are collected and the target load vector is calculated. A deviation parameter-full-field strain proxy model is constructed using radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state. The sensor data includes strain gauge data, displacement data, and force data. Based on the deviation strain field and sensor data of the current state of each site, the corresponding digital twin deviation strain field is constructed through the bridge function. Combined with the offline calibration load inversion matrix and the interface stiffness weight matrix, the corresponding actual load vector of the current state is generated. Calculate the coupling control deviation between the actual load vector and the target load vector for each current state; Based on the coupling control deviation, a comprehensive accuracy index is calculated; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error. Based on the comprehensive accuracy index and the actual load vector of the current state, load allocation is performed through the mean-coupled fuzzy PID collaborative control method to generate load allocation control commands for the current state.

2. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 1, characterized in that, The generation of the corresponding deviation strain field for the current state includes: Sensors are installed in the airborne missile body at a certain site to collect corresponding sensor data; The airborne projectile body at a certain site is simplified into a beam model with n mass elements, and the target load vector at the site is calculated. Based on sensor data, the rigid body rotation matrix of the loading plane of the airborne projectile is solved by the least squares method; Based on the rigid body rotation matrix, the real-time loading deviation parameters of the current state are calculated, and a deviation parameter-full-field strain proxy model is constructed by radial basis function and finite element simulation method to generate the deviation strain field of the current state.

3. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 1, characterized in that, The generation of the actual load vector corresponding to the current state includes: Based on strain gauge data from strain gauge measurement points, the corresponding bridge function factors are calculated using Latin hypercube sampling and radial basis function network. A digital twin bias strain field is generated based on the bridge function factor and the bias strain field of the current state. Based on sensor data, the mechanical influence area of ​​each site is constructed; Based on the material constitutive relation and digital twin bias strain field of the projectile, the element stress distribution is generated, and combined with the finite element simulation method, a six-dimensional load vector for each station is generated. Simulate the strain response of each mechanically affected region under different working conditions to generate an offline calibration load inversion matrix; set the interface stiffness weight matrix through the connection unit or contact interface of each mechanically affected region. Based on the six-dimensional load vector, offline calibration load inversion matrix, and interface stiffness weight matrix of each site, the actual load vector of the current state of each site is generated.

4. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 1, characterized in that, The overall accuracy index for calculation includes: Based on the actual load vector and target load vector of each current state, calculate the independent load deviation vector of each station in the current state; Calculate the global deviation vector based on the deviation vectors of each independent load; Based on the global deviation vector and the independent load deviation vector of each station, the coupling control deviation of each station is calculated. Based on the coupling control deviation, actual load vector, and target load vector of each station, and combining the sum of the actual main loads of each station with the ratio of the actual loads of each station, the comprehensive accuracy index is calculated.

5. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 4, characterized in that, The calculation of the comprehensive accuracy index by combining the sum of the actual main loads of the stations with the proportion of the actual loads of each station includes: Based on the site load magnitude error, coupling control deviation, actual load vector, and target load vector of each site, the corresponding site load magnitude error is calculated; Based on the main load components of each site, the sum of the actual main loads of the sites is calculated, and combined with the theoretical load ratio, the actual load ratio of each site is generated. The global allocation ratio error is calculated based on the theoretical load ratio, actual load ratio, and coupling control deviation of each station. The overall accuracy index is calculated based on the site load magnitude error and the global allocation ratio error of each site.

6. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 5, characterized in that, The formula corresponding to the independent load deviation vector is: ； in, Indicates the first Target load vector of each site Indicates the first Current status of each site The actual load vector, Indicates the first Independent load deviation vectors for each site; The formula corresponding to the global deviation vector is: ； in, Represents the global deviation vector. Indicates the total number of sites. This represents the summation function; The formula corresponding to the coupling control deviation is: ； in, Indicates the first Current status of each site Coupling control deviation, Indicates the first Site weight coefficient for each site.

7. The adaptive load allocation method for airborne missile bodies based on digital twin inversion according to claim 1, characterized in that, The load allocation control command for generating the current state includes: Determine whether the overall accuracy index is less than the preset convergence threshold; if so, use the load allocation control command of the previous state as the load allocation control command of the current state. Conversely, based on the actual load vector of the current state, the standardized deviation and rate of change of each station under different degrees of freedom in the current state are generated; The standardized deviation and rate of change of each station under the current state are used as the input data of the fuzzy PID controller, and the control parameters of the fuzzy PID controller are adjusted based on the fuzzy rule base to obtain the updated control parameters. Based on the updated control parameters, calculate the control output for each station under different degrees of freedom; Integrate the control outputs of all stations and perform coordinated decoupling through a decoupling matrix to generate the final coordinated control increment; By combining the final coordinated control increment and the load allocation control command from the previous state, the load allocation control command for the current state is generated.

8. An adaptive load allocation system for airborne missile bodies based on digital twin inversion, used to implement the adaptive load allocation method for airborne missile bodies based on digital twin inversion as described in any one of claims 1 to 7, characterized in that, include: The strain field generation module is used to collect sensor data from various stations on the airborne projectile and calculate the target load vector. It constructs a deviation parameter-full-field strain proxy model through radial basis functions and finite element simulation to generate the corresponding deviation strain field for the current state. The load calculation module is used to construct the corresponding digital twin deviation strain field based on the deviation strain field and sensor data of the current state of each site through the bridge function, and generate the corresponding actual load vector of the current state by combining the offline calibration load inversion matrix and the interface stiffness weight matrix. The index calculation module is used to calculate the coupling control deviation between the actual load vector and the target load vector for each current state. Based on the coupling control deviation, a comprehensive accuracy index is calculated; the comprehensive accuracy index includes the site load magnitude error and the global allocation ratio error. The allocation module is used to allocate load based on the comprehensive accuracy index and the actual load vector of the current state, and to generate load allocation control commands for the current state through the mean-coupled fuzzy PID collaborative control method.

Images