Large-stroke aircraft airfoil static force loading system for aircraft strength test

Abstract

The invention discloses a large-stroke aircraft airfoil static force loading system for aircraft strength testing, and particularly relates to the technical field of aviation structure testing, and the large-stroke aircraft airfoil static force loading system comprises a multi-region loading execution module for realizing distributed autonomous loading and state sensing; the boundary coordination control module realizes force field continuity between loading areas through flexible fusion; the asynchronous self-synchronization adjustment module adjusts a loading step pitch based on local feedback and a loading offset prediction function; the redundant closed-loop feedback control module adopts a force value and displacement dual-channel path; the loading path modeling and reproduction module realizes historical path playback and loading track reproduction by constructing a track data set; according to the invention, continuous coordination of multi-region loading force fields is realized, the adaptive adjustment capability of loading step pitch and frequency is improved, and the loading behavior is dynamically optimized through a path difference feedback mechanism; a redundancy feedback and path reproduction mechanism is configured, and the fault tolerance and loading consistency of the system are enhanced; and intelligent scheduling is realized based on the shared table, so that the loading precision, stability and efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft structure testing. More specifically, the present invention relates to a large-stroke aircraft wing surface static loading system for aircraft strength testing. Background Art

[0002] During the service life cycle of an aircraft structure, it needs to withstand external loads under various complex working conditions. In particular, its wing surface, as a core load-bearing component, its static strength and stiffness characteristics directly affect the safety and flight performance of the whole aircraft. To verify the force response of the wing surface under extreme loads or specific load conditions, systematic static strength tests need to be carried out on the ground. This test usually simulates forces such as lift, inertia force, and aerodynamic force during flight through a loading device to evaluate the force distribution, deformation behavior, and damage tolerance ability of the structure.

[0003] In the prior art, for the static loading of large-size aircraft wing surface structures, a centralized control method is mainly adopted, that is, a central control system uniformly schedules multiple loading actuators for collaborative loading. Although this method is applicable to conventional rigid wing surface structures, when facing a new generation of composite material wing surface structures with characteristics such as large aspect ratio, high flexibility, and non-uniform stiffness distribution, the following problems exist:

[0004] Discontinuous loading force field distribution: Due to insufficient distribution density of loading points or response lag, transitional faults occur in the force on the wing surface area, affecting the authenticity of the structural response; Difficulty in coordinating boundary regions: At the junctions of multiple loading areas, due to large stiffness differences and asynchronous feedback, loading mismatch, deformation concentration, or crack induction often occur; Unstable response of loading behavior: The existing system has a lag in adjusting loading errors, is prone to loading jumps, and even excites non-linear responses; Insufficient path reproduction ability: Lack of accurate path modeling and real-time feedback mechanism, resulting in uncontrollable and irreproducible loading behavior and poor stability of test data.

[0005] Especially in the trend of increasing wingspan of the next-generation wide-body aircraft and aircraft, how to achieve precise loading, intelligent adjustment, and path controllability of large wing surface flexible structures in static tests has become a key problem that urgently needs to be broken through in aircraft structure test technology. Therefore, the present invention proposes a large-stroke aircraft wing surface static loading system for aircraft strength testing in order to solve the above problems. Summary of the Invention

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A large-stroke aircraft wing surface static loading system for aircraft strength testing, comprising:

[0008] Multi-region loading execution module, which divides the aircraft wing surface structure into multiple loading regions. Each region is configured with an execution unit having local loading control capabilities. The execution unit includes a local controller and a sensing element, and performs loading actions without relying on central continuous instructions, and senses the stress state and displacement changes at its location.

[0009] Boundary coordination control module. There is a boundary transition zone between the loading regions. By setting flexible fusion factors and distributed cooperative adjustment rules, a continuous and adjustable force field edge connection relationship is formed between different loading regions to meet the distribution response requirements of high-aspect-ratio wing surfaces under high-load torsion conditions.

[0010] Asynchronous self-synchronization adjustment module. Based on the local feedback results, the loading unit adjusts the loading step through a loading offset prediction function, and can achieve the approximation and fusion of local loading behaviors to the global loading trend without a global synchronization signal.

[0011] Redundant closed-loop feedback control module. The loading unit is configured with two different types of feedback paths, and automatically calls the backup path or generates a fusion signal to form a stable feedback judgment when the main path fails.

[0012] Loading path modeling and reproduction module. Records the force value changes, spatial displacement paths and loading sequences of each loading unit during the loading process, forms a data trajectory set that can be replayed, and realizes the reproduction control of the loading path by comparing the fitting error between the current path and the historical path during subsequent loading processes.

[0013] In the multi-region loading execution module, the local controller of each loading execution unit includes an embedded microprocessor structure, which has the capabilities of loading state perception, action planning and local fitting of the loading curve. When performing the loading task, it dynamically generates a local loading curve based on historical stress data and current feedback information. This curve is updated at least once in each loading cycle, and the update algorithm performs bivariate optimization based on the local error expansion function within the region and the structural stiffness adjustment parameter, so as to enhance the adaptability of the loading path to the target structure, and control the fitting error of the loading curve within a set threshold range. This threshold is set in advance according to the wing surface material properties and test conditions to ensure that the loading behavior maintains microscopic consistency within each region, and at the same time enhances the matching ability of the loading structure to the local nonlinear response of the structure.

[0014] The boundary coordination control module includes an adjustable flexible force fusion factor. The setting of this factor is based on the relative deformation gradient difference between the boundary units between the loading regions, and combines the force expansion trajectories collected in historical experiments to achieve smooth loading transition by optimizing the calculation of the boundary transition function during continuous boundary loading.

[0015] The selection of the boundary transition function includes three types: linear transition function, logarithmic smoothing function, or cubic spline transition function. Each function type is selected according to the average stiffness change rate of the target wing surface in the boundary region;

[0016] During the loading process, it is updated once in each iteration step through the distributed collaborative adjustment rule, and a boundary coordination matrix is established by the feedback data of adjacent boundary loading units to balance the loading direction and intensity differences and achieve the continuous force field connection of the multi-region loading state.

[0017] In a preferred embodiment, in the asynchronous self-synchronization adjustment module, the loading offset prediction function is constructed based on the historical loading trend and the current real-time error fitting curve during the loading process. The prediction model adopted is any one of the three models: weighted moving average function, time series residual regression model, or local sample non-linear mapping function. The output result of the prediction function is used to guide the dynamic correction of the loading step size.

[0018] In a preferred embodiment, each loading execution unit autonomously adjusts the loading step frequency and the loading micro-displacement amplitude according to the difference degree of the loading behavior between its affiliated loading region and the adjacent region. Through the strategy of minimizing the loading offset between regions, the distributed loading behavior gradually approaches the global synchronization target, while avoiding systematic oscillation and loading mutation problems. This asynchronous self-synchronization adjustment process consists of three stages: local prediction - correction - re-prediction, and the prediction window length can be dynamically adjusted according to the loading complexity.

[0019] In a preferred embodiment, the two different types of feedback paths configured in the redundant closed-loop feedback control module are respectively a force value sensing path based on a resistance strain gauge and a displacement sensing path based on the laser interferometric ranging principle. The two paths collect data at different sampling frequencies during the loading process. Among them, the force value sampling interval is set at the millisecond level, and the displacement data sampling is set at the micron level. At the end of each loading cycle, data fitting processing is carried out through a preset fusion calculation model, and it is judged whether the deviation value between the two paths exceeds the preset tolerance interval. If it exceeds, the fault-tolerant switching mechanism is triggered, and the multi-path fusion correction control strategy is enabled to adaptively adjust the loading action to avoid unstable oscillation during the loading process and ensure the reliability and safety tolerance of the loading response.

[0020] In a preferred embodiment, the loading path modeling and reproduction module records the dynamic parameters of the loading process by point-by-point sampling, and synchronously maps the force value, loading order, loading time, and spatial position through constructing a high-dimensional path matrix. The method for constructing the path matrix adopts a three-dimensional Bezier curve path interpolation method to construct an initial model. During the testing process, the real-time collected data is fitted with the historical model, the loading trajectory is adjusted in real time through the fitting residual, and the path record is updated to form a path data set that can be called multiple times. During the path reproduction stage, the system preferentially calls a set of historical path data with the highest fitting degree as the current loading reference trajectory, and sets a deviation threshold to judge the path reproduction accuracy. If the path fitting deviation exceeds the critical value, the path is automatically de-escalated and the current loading strategy is re-planned.

[0021] In a preferred embodiment, a data sharing channel is configured between the multi-region loading execution module and the boundary coordination control module. A loading status sharing table is established through a multi-thread information synchronization mechanism. This sharing table is regularly pushed with status data by each loading unit and is regularly read and integrated. During the loading task, the loading order and priority scheduling strategy are dynamically updated according to the loading force change trend in the sharing table and the feedback stability of each unit. This strategy adopts a weight promotion - feedback delay penalty mechanism, enabling the loading units with fast response and good fitting to obtain higher instruction frequencies and control priorities, thereby improving the overall loading response efficiency and reducing the probability of loading path oscillation.

[0022] In a preferred embodiment, a dynamic path feedback linkage mechanism is established between the asynchronous self-synchronization adjustment module and the loading path modeling and reproduction module. After each loading action is completed, a multi-parameter fitting analysis is performed on the current actual loading path and the historical path, and the path difference index is calculated. This index consists of three parts, namely the total loading force difference ratio, the path shape residual factor, and the loading point time offset rate. The calculation result is fed back to the next round of loading step adjustment link as a path correction factor to be embedded, controlling the loading path to gradually converge to the historical best path to achieve continuous path evolution and dynamic fine-tuning capabilities during the loading process.

[0023] The technical effects and advantages of the present invention:

[0024] The present invention constructs multi-region loading execution units on the aircraft wing surface structure, and introduces a boundary coordination control module between regions, sets a flexible fusion factor and a loading transition function, realizing a smooth connection of mechanical transitions between different loading regions, and significantly improving the continuity of the force field during the loading process. The cooperative algorithm that dynamically adjusts according to the deformation gradient and historical trajectory between the boundary loading units can effectively solve problems such as sudden changes in loading force and regional faults in traditional loading systems, and is particularly suitable for wing surface structures with large aspect ratios and non-uniform stiffness distributions, making the loading response closer to the real structural behavior under flight conditions.

[0025] By introducing an asynchronous self - synchronization adjustment module, the present invention enables the loading unit to autonomously complete the dynamic correction of the loading step, loading frequency, and micro - displacement amplitude based on local feedback data and a loading offset prediction function without the need for a global synchronization signal. The loading control process constructs a "prediction - correction - re - prediction" closed - loop mechanism, and combines the fitting residual and path difference index fed back by the path modeling module to achieve real - time evolution and fine - tuning of the loading behavior, effectively reducing sudden jumps, non - linear response excitation, and local error accumulation during the loading process, improving the loading accuracy and path stability, and ensuring the consistency and controllability of the large - structure flexible wing surface loading.

[0026] The present invention configures a redundant closed - loop feedback control module in each loading execution unit, combines a dual - channel high - frequency sampling mechanism for force value and displacement, and enhances the system's ability to identify and adaptively correct abnormal feedback or error surges. At the same time, a high - dimensional path matrix is constructed through the path modeling and reproduction module to form a callable historical loading trajectory library, and the path with the highest fitting degree is dynamically selected as the reference trajectory during the path reproduction stage to achieve path - consistent loading control. In addition, through the loading status sharing table and multi - thread scheduling mechanism, the system can adjust the instruction frequency and scheduling priority of the loading unit in real time, enabling the unit with faster response and better fitting to obtain the loading dominance, further improving the overall loading efficiency and safety tolerance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] For the convenience of those skilled in the art to understand, the present invention will be further described below with reference to the accompanying drawings;

[0028] Figure 1 is the schematic diagram of a large - stroke aircraft wing surface static loading system for aircraft strength testing in the present invention;

[0029] Figure 2 is the schematic diagram of the redundant closed - loop feedback control module in the present invention;

[0030] Figure 3 is the schematic diagram of the loading offset prediction function in the asynchronous self - synchronization adjustment module of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0032] Refer to Figures 1-3 to obtain the following embodiments:

[0033] Example 1:

[0034] A large-stroke aircraft wing surface static loading system for aircraft strength testing, comprising:

[0035] A multi-region loading execution module that divides the aircraft wing surface structure into multiple loading regions. Each region is configured with an execution unit having local loading control capabilities. The execution unit includes a local controller and a sensing element, and performs loading actions without relying on central continuous instructions, and senses the force state and displacement changes at its location. The role of the multi-region loading execution module is to "split" the entire large wing surface structure into multiple controlled loading regions, solving the problem that traditional centralized control cannot adapt to high aspect ratio wing surfaces. Each region has a local intelligent unit (controller + sensor) to achieve autonomous execution and in-situ data acquisition. Its essential role is to provide the basis for system execution and data acquisition.

[0036] A boundary coordination control module. There is a boundary transition zone between the loading regions. By setting a flexible fusion factor and a distributed collaborative adjustment rule, a continuous and adjustable force field edge connection relationship is formed between different loading regions to adapt to the distribution response requirements of high aspect ratio wing surfaces under high load torsion states. By setting the regional boundary transition zone, it prevents the occurrence of "discontinuous force values" or "loading tearing" at the edges between loading units. Using the flexible fusion factor and adjustment rules to achieve "soft connection" rather than "hard splicing". Its essential role is to eliminate the physical and control faults caused by "regional segmentation" and achieve continuous loading response.

[0037] An asynchronous self-synchronization adjustment module. The loading unit is based on the local feedback result and adjusts the loading step through a loading offset prediction function, and can achieve the approximation and fusion of local loading behaviors to the global loading trend without a global synchronization signal. Each loading point does not require a central unified instruction, but adjusts the loading step according to local feedback through the "loading offset prediction function". Based on the "loading error propagation model + regional difference weight factor" to judge synchronization and form a "loosely coupled control cluster". The asynchronous self-synchronization adjustment module is the decision-making core, making the system both autonomous and collaborative.

[0038] A redundant closed-loop feedback control module. The loading unit is configured with two different types of feedback paths, and automatically calls the backup path or generates a fusion signal to form a stable feedback judgment when the main path fails. Each loading point has two different types of feedback paths (such as force + displacement), and the backup path is enabled when the main path fails. It can generate a fusion signal to ensure the closed-loop stability of the loading behavior. Its role is to provide real-time safety protection to prevent out-of-control, oscillation or cumulative loading offset.

[0039] The loading path modeling and reproduction module records the force value changes, spatial displacement paths, and loading sequences of each loading unit during the loading process, forming a data trajectory set that can be replayed. During subsequent loading processes, by comparing the fitting error between the current path and the historical path, the loading path reproduction control is realized. It records information such as force values, displacements, and sequences in real time; forms a historical path trajectory set, which is used for path reproduction and error correction during subsequent loading; improves the comparability and consistency of fatigue tests / repeated loading experiments. Its essential function is to endow the loading system with the "memory" ability, forming a function of historical path traceability and precise reproduction.

[0040] The entire large-stroke aircraft wing surface static loading system consists of five core modules. These modules cooperate with each other and are linked layer by layer during the system operation, forming a closed-loop, high-precision, and strong-robustness loading control system. Its cooperation logic can be described according to the following process:

[0041] The start of the loading task begins with the multi-region loading execution module. According to the geometric shape of the wing surface structure and the test requirements, this module divides the entire wing surface into several loading regions. Each region deploys loading execution units with local control and sensing capabilities, which can independently execute loading actions and collect local force values and displacement states.

[0042] To avoid force field fractures or local transition mutations between loading regions during the loading process, the boundary coordination control module intervenes at the junctions between loading regions. By setting flexible fusion factors and adjustment rules, the loading force presents a continuous and adjustable transition state between different regions. This mechanism ensures the physical response consistency of the structure during the overall force-bearing process.

[0043] After receiving the loading instruction, the loading execution units in each region do not rely on a unified central scheduling signal, but rely on the asynchronous self-synchronization adjustment module for control. Based on the local real-time feedback of each execution unit, this module adjusts the loading step through the loading offset prediction function, so that although each loading point acts asynchronously, it generally tends to the synchronous trend of the target force field. This "loose coupling" control method improves the system's adaptability to large-scale loading distributions.

[0044] During the loading execution process, each loading unit is connected to the redundant closed-loop feedback control module. By configuring two different types of feedback paths, the stress state and displacement information during the loading process are collected in real time. Once a feedback path fails, the system will automatically call the backup path. The main path can be set as one of the two paths, and the backup path is the other, or judgment and correction are carried out through multi-path signal fusion, so as to ensure the continuity and stability of the entire loading behavior.

[0045] Meanwhile, the loading path modeling and reproduction module samples each data node in the entire loading process to establish a high-dimensional path trajectory database. After each loading cycle is completed, this module compares the actual execution path with the historical path, calculates the fitting error, and feeds back the path difference information to the asynchronous self-synchronization adjustment module, which serves as an important basis for adjusting the next loading step size, enabling the loading system to possess the capabilities of path self-learning and dynamic optimization.

[0046] In the multi-region loading execution module, the local controller of each loading execution unit includes an embedded microprocessor structure, which has the capabilities of loading state perception, motion planning, and local fitting of the loading curve. When executing the loading task, it dynamically generates a local loading curve based on the historical stress data and the current feedback information. This curve is updated at least once in each loading cycle, and the update algorithm performs bivariate optimization based on the local error expansion function within the region and the structural stiffness adjustment parameter, enhancing the adaptability of the loading path to the target structure and controlling the fitting error of the loading curve within a set threshold range. This threshold is set in advance according to the material properties of the wing surface and the test conditions to ensure that the loading behavior maintains microscopic consistency in each region and simultaneously enhances the matching ability of the loading structure to the local nonlinear response of the structure.

[0047] In each loading execution unit, an embedded microprocessor with low power consumption and strong real-time processing capabilities (such as the STM32H7 series) is configured, which is mainly responsible for the following four tasks:

[0048] Collect the force value and displacement feedback output by the local sensor;

[0049] Read the historical loading data (stress, deformation response) of the corresponding region from the local storage unit;

[0050] Perform local loading curve fitting according to the "loading error expansion function" and "structural stiffness adjustment parameter" in the present invention;

[0051] Execute fitting optimization update at least once within each loading cycle, and output new loading step values, speeds, and directions according to the results.

[0052] The local error expansion function within the region is used to describe: the magnitude of the error between the currently fitted loading curve and the actual loading feedback, and the degree of influence of this error on adjacent loading units.

[0053] The calculation of this error quantity is based on the following principles:

[0054] Take the difference between the predicted loading value and the actual feedback value at the current loading point, and perform a square process to reflect the absolute magnitude of the error;

[0055] Consider the distance between the loading point and other surrounding loading units. The closer the distance, the greater the impact; the farther the distance, the more the impact gradually attenuates.

[0056] The error contribution of each loading point is weighted according to its distance from the target point. A commonly used weighting method is the exponential decay model, that is, the farther the distance, the smaller its error contribution.

[0057] The output value of the entire error expansion function is: the comprehensive error estimate of this loading point, which reflects the matching degree and coordination of its current fitted curve to the structure of this region.

[0058] The structural stiffness adjustment parameter is used to describe: the deviation between the structural stiffness response of the current loading unit and the historical reference stiffness response.

[0059] The specific calculation method is as follows:

[0060] First, measure the change in the force value generated by the current loading unit within a small loading displacement to obtain the change in the force value corresponding to the unit displacement, that is, the current stiffness.

[0061] Then compare this stiffness value with the reference stiffness value at the same loading position in the historical database to obtain the difference between the two.

[0062] If this difference is large, it indicates that there is a significant change in the current structural response, and the fitted loading curve needs to be adjusted.

[0063] In addition, different weight coefficients can be set for the importance degree of each loading unit. For example, the weight of the loading unit near the wing root can be higher than that of the wing tip.

[0064] The finally obtained value of the stiffness adjustment parameter reflects whether the current loading behavior is suitable for the structural elasticity or stiffness state of this region.

[0065] The local controller takes these two values, the "local error expansion function within the region" and the "structural stiffness adjustment parameter", as joint inputs and performs overall fitting optimization through a weighted combination method. The goal is to make both the error term and the stiffness deviation term as small as possible. The optimization methods that the system can adopt include but are not limited to: gradient descent optimization method; least squares fitting method; local spline interpolation combined with the method of minimizing fitting error.

[0066] The optimization goal is to generate a loading curve that can:

[0067] Maximally fit the current structural response trend; control the prediction error within the deviation range allowed by the material and structure; maintain the microscopic consistency of the loading points in physical space.

[0068] In the present invention, the fitting error of each loading cycle is controlled within a threshold range. The sources of this threshold value include the following three dimensions: structural material parameters, such as the maximum safe strain value of composite materials; experimental condition settings, such as the maximum allowable loading error percentage, which is usually set within three percent; and the geometric distribution and historical error trend of the structure in this area, especially lower error tolerances need to be set for stress concentration or weak areas. The system takes the minimum value among the above three error sources as the allowable range of fitting error for this loading point. If the fitting error of the currently generated loading curve exceeds this range, the fitting correction mechanism will be triggered to re-plan the loading path.

[0069] Taking an actual loading unit as an example, it is located in the third area of the wing surface and is numbered as the seventh loading execution point. Its local controller extracts the average stress response curve of this point during the past five loading processes before the loading starts, and collects the real-time force value changes in the current cycle. Subsequently, it calculates the error between the predicted curve and the feedback value, and evaluates the influence of this error on the surrounding loading points. The system also measures the displacement increment and the corresponding stress value increment of this loading point in the current cycle, calculates the current stiffness, and judges whether the stiffness deviation is acceptable after comparing it with the historical reference stiffness. Finally, the two values are input into the optimization module to perform loading curve correction output and control the next loading action of this point. If the error of the finally generated loading curve is controlled within the preset threshold, the controller will execute the loading according to this curve; otherwise, it will enter the re-fitting stage.

[0070] The boundary coordination control module includes an adjustable flexible force fusion factor. The setting of this factor is based on the relative deformation gradient difference between the boundary units in the loading areas, and combines the force expansion trajectories collected in historical experiments. During the continuous loading at the boundary, the loading smooth transition is achieved by optimizing and calculating the boundary transition function.

[0071] The selection of the boundary transition function includes three types: linear transition function, logarithmic smoothing function, or cubic spline transition function. Each function type is selected according to the average stiffness change rate of the target wing surface in the boundary area;

[0072] During the loading process, it is updated once in each iteration step through the distributed cooperative adjustment rule, and a boundary coordination matrix is established through the feedback data of adjacent boundary loading units to balance the loading direction and intensity differences and achieve the continuous force field connection of the multi-region loading state.

[0073] The adjustable flexible force fusion factor is a key parameter in the boundary coordination control module to realize the force transition control between different loading areas. This factor controls the distribution weight of the loading force at the area junction, that is, it determines the proportion of the loading forces from the left and right areas that act on the boundary loading unit together.

[0074] To make this fusion factor clear and implementable, its value is not set based on subjective experience, but is jointly determined by the following two physical and data-based dimensions and undergoes standardized fusion.

[0075] The first dimension: the relative deformation gradient difference between boundary elements, namely the "structural response difference coefficient": between adjacent loading elements at the boundary of the loading area, measure the displacement change rate per unit length (i.e., the deformation gradient), and this index reflects the local differences in structural rigidity and deformation ability.

[0076] Measurement method: For two adjacent loading elements at the boundary, collect their displacements after loading respectively; use the geometric distance between the two points as the denominator to calculate the deformation value per unit length; calculate for the boundary elements of the two regions respectively, and take their difference as the "deformation gradient difference".

[0077] Numerical processing: Normalize this difference to the interval [0,1]. For example, when the difference is close to zero, it indicates that the regional rigidities are approximately the same, and the boundary transition can be rigidly processed, and the fusion factor tends to zero (rigid connection); when the difference is large, it indicates that the flexible responses of the two regions are different, and the fusion factor should be increased to enhance flexible buffering.

[0078] The second dimension: the difference value of the force expansion trajectories collected in historical experiments, namely the "loading path difference coefficient": refers to the expansion trend of the force paths of the boundary loading elements over time in multiple loading cycles obtained through historical loading experiments, which is used to predict whether the loading behaviors in this area during the current loading process converge or deviate.

[0079] Extraction method: Retrieve the force-time curves of the boundary loading elements under multiple similar working conditions in the database; fit them with the real-time loading curve of the boundary points in the current loading cycle; calculate the curve fitting difference through a sliding time window or the Bezier residual fitting method; take the maximum difference, average residual or deviation rate as the trajectory difference index.

[0080] Numerical normalization: The smaller the difference, the more stable and predictable the boundary loading behavior is, and the smaller the fusion factor; the larger the difference, the more significant the difference between the historical and current loading behaviors, and the fusion factor should be increased to buffer unexpected disturbances.

[0081] The final calculation method of the fusion factor: Denote the above two difference quantities as the "structural response difference coefficient" and the "loading path difference coefficient" respectively, and adopt the weighted average fusion algorithm as follows:

[0082] The value of the fusion factor = the first difference coefficient multiplied by the first weight coefficient, plus the second difference coefficient multiplied by the second weight coefficient.

[0083] Among them: the first weight coefficient represents the degree dominated by the geometric flexibility of the structure; the second weight coefficient represents the degree of influence of the historical path consistency on the current loading safety; the values of the two weight coefficients can be preset according to the structural position (such as the wing root, wing tip) and the material properties of the wing surface. For example: for high-flexibility materials, the second weight can be set higher than the first; for rigid regions or node connection regions, the first weight can be set higher.

[0084] The result of the final fusion factor is limited between zero and one: approaching zero indicates a transition to a more rigid state; approaching one indicates a more flexible coordination; it usually fluctuates between 0.2 and 0.8.

[0085] Suppose in a certain loading, the normalized difference in deformation gradient at the boundary between the third region and the fourth region is 0.6, and the historical path fitting residual is 0.4. The first weight coefficient is set to 0.7, and the second weight coefficient is set to 0.3. Then: the fusion factor = 0.6 multiplied by 0.7, plus 0.4 multiplied by 0.3; finally, the fusion factor is approximately 0.54, that is, the current boundary adopts a flexible force fusion distribution strategy with a ratio of 54%.

[0086] The boundary transition function refers to a mathematical function that smoothly transitions the loading force from one region to another in the boundary region to avoid loading mutations.

[0087] In the present invention, three function forms are supported: linear transition function: used for boundary regions with small stiffness changes, with a linear transition behavior, suitable for wing surfaces with medium curvature;

[0088] Logarithmic smoothing function: used for highly flexible wing surfaces, which can quickly achieve exponential growth or decay from low loading to high loading, suitable for flexible structure interfaces;

[0089] Cubic spline transition function: used to finely control the change of force values in the boundary region, which can adaptively control the slope and second derivative of the function to ensure continuity and smoothness.

[0090] Selection strategy: The selection of the specific function type is based on the following parameters: First, evaluate the average stiffness change rate of the boundary region, that is, the stiffness difference between the loading units on both sides of the boundary divided by the boundary width; if the change rate is lower than the preset threshold, a linear function is selected; if the change rate exceeds a certain degree, it indicates that the nonlinear response of the structure is significant, and logarithmic or cubic spline can be preferentially selected; the function can be adaptively selected at the initialization stage of each loading and dynamically adjusted during the loading process.

[0091] The boundary coordination matrix is a control matrix used to allocate the direction and intensity during the boundary loading process. Each row of the matrix corresponds to the current state of a boundary loading unit, and each column corresponds to the loading coupling degree between it and adjacent units.

[0092] The construction method is as follows: collect the loading direction (such as normal direction, tangential direction) and force value feedback of each loading unit at the boundary in the current iteration step;

[0093] Construct a coupling factor array with the angle difference of the loading direction and the force value difference as the weight factors;

[0094] Fill the coupling factors into the matrix to form a two-dimensional correlation structure;

[0095] During the loading adjustment process, this matrix is used to dynamically adjust the force application direction and amplitude of the loading unit to achieve continuous force field between regions.

[0096] The fuzzy weighting strategy can be adopted for matrix filling; use the normalization factor to unify the direction difference and force difference into the same dimension; the matrix update frequency is synchronized with the loading adjustment iteration step to ensure fast response of the boundary region.

[0097] Example scenario: Assume that the aircraft wing surface is divided into six loading regions, and there is a flexible structure transition region between the third and fourth regions. During the test, it is found that the boundary stiffness change rate is as high as a difference of one hundred Newtons per meter, which is a typical high-flexibility discontinuous transition scenario.

[0098] Implementation process: The system automatically selects a cubic spline transition function according to the stiffness change rate; uses this function to construct a transition force path from the boundary loading unit in the third region to the loading unit in the fourth region; at the same time, set the flexible force fusion factor to a relatively high initial value (such as seventy percent flexible weight) to adapt to the fast loading response of this region; in each loading cycle, update the boundary coordination matrix through the feedback displacement changes of each boundary loading unit; the coordination matrix adjusts the loading direction and magnitude, gradually approaching the ideal transition force field state to avoid loading jumps or fault phenomena.

[0099] In the asynchronous self-synchronizing adjustment module, the loading offset prediction function is constructed based on the historical loading trend and the current real-time error fitting curve during the loading process. Any one of the three models, namely the weighted moving average function, the time series residual regression model, or the local sample non-linear mapping function, is selected as the prediction model. The output result of the prediction function is used to guide the dynamic correction of the loading step size.

[0100] The loading offset prediction function is the control core of the asynchronous self-synchronizing adjustment module of the present invention, which is used to predict the loading trend offset value of the loading unit in the future micro-step, and accordingly dynamically adjust the loading step size and the rhythm of the loading action to achieve the adaptive convergence control of the system without a centralized instruction.

[0101] The inputs of this function include: historical loading trends (i.e., sequences of loading response data over several past cycles); the error change curve measured in real time during the current loading cycle (i.e., the difference between the expected loading value and the current feedback value); and its output is a loading offset estimate, which is used to determine whether the current loading behavior deviates from the overall system trend and to correct the current loading micro-step behavior accordingly.

[0102] Weighted moving average function: By averaging the loading error values over several past cycles with different weights, it predicts the loading offset trend at the next moment. Implementation method: Set a sliding window with a fixed length, for example, five loading cycles; set weights for the error data of each cycle within the window, with higher weights for cycles closer to the current cycle; sum up these weighted errors to obtain the current predicted offset value; the weight settings can be set according to linear decay, exponential decay, or engineering experience. Parameter source: The sliding window length is determined by the loading frequency and the wing surface response delay; the weight factor can be adjusted according to the position of the loading point in the structure (such as the wing root, wing tip). Characteristics: Simple to implement, fast response, suitable for areas with small loading fluctuations.

[0103] Time series residual regression model: Regarding the historical loading error data as a time series, it constructs a prediction function by fitting the trend of its residuals. Implementation method: First, fit the error sequence over a past period with a standard loading model; calculate the difference between the real error and the fitted error, that is, the residual sequence; establish a linear or nonlinear regression model (such as autoregressive moving average model) for the residual sequence; use this model to predict the offset change amount in the next loading cycle. Parameter source: The residual sequence is directly collected from the error fitting curve; the model order and lag period can be selected through criteria such as AIC and BIC; the regression coefficients are obtained by least squares or maximum likelihood estimation. Characteristics: Can identify the error change trend, suitable for medium-speed change areas or transition section areas.

[0104] Local sample nonlinear mapping function: It establishes a sample subset around the current loading state, extracts features from this subset to construct a nonlinear response mapping, and is suitable for scenarios with complex or violently fluctuating error change patterns. Implementation method: Extract multiple historical samples from the database that are most similar to the current loading unit state; use nonlinear modeling tools such as locally weighted regression, radial basis neural network, or support vector regression; establish a nonlinear functional relationship between the loading offset value and the input state variables (such as loading position, loading path residual, loading speed) within this local space; output the loading offset estimate for the current micro-step adjustment. Parameter source: Similar samples are screened by Euclidean distance or Mahalanobis distance; the nonlinear modeling parameters are determined by the training process; the mapping model can be dynamically updated to adapt to changes in the loading environment. Characteristics: Strong learning ability, suitable for complex structure areas, suitable for local response nonlinear areas.

[0105] Regardless of which prediction model is used, its prediction results will be input into the loading step adjustment mechanism as a "loading offset estimate": if the prediction result shows that the deviation direction is consistent with the system's expected direction, the step length will be extended or the loading frequency will be increased; if the deviation direction is opposite, the system will reduce the step length or even suspend loading to wait for the error to converge; the correction amplitude can be determined based on the predicted value amplitude and the current error absolute value to avoid excessive adjustment.

[0106] Assume that the fifth loading unit in the second area is undergoing cyclic loading: the error has gradually increased over the past six cycles, and the error between the current cycle and the expected value has expanded to five percent; the system selects a weighted sliding average model, sets a five-cycle window, and decreases the weights sequentially from 0.4 for the latest cycle; after weighted processing, the offset estimate is negative, indicating that the current loading trend deviates from the overall synchronization trend; the system reduces the next microstep loading displacement by ten percent and adjusts the loading direction to match the historical trend; at the same time, the offset value is recorded in the path modeling module for closed-loop correction of the path prediction error.

[0107] Each loading execution unit autonomously adjusts the loading step frequency and loading micro-displacement amplitude according to the degree of difference in loading behavior between its loading area and the adjacent area. Through the inter-area loading offset minimization strategy, the distributed loading behavior gradually approaches the global synchronization target, while avoiding systematic oscillation and loading mutation problems. The asynchronous self-synchronization adjustment process consists of three stages: local prediction-correction-re-prediction, and the prediction window length can be dynamically adjusted according to the loading complexity.

[0108] The degree of loading behavior difference is used to quantify the behavior gap between the current loading behavior of a loading execution unit and the loading units in its adjacent area, reflecting "whether they are synchronized" or "whether there is an offset".

[0109] Calculation method: Compare the loading force change rate (i.e. loading acceleration) of the current loading unit with that of the adjacent unit per unit time;

[0110] Compare whether the loading displacement directions are consistent. The larger the angle between the directions, the greater the difference.

[0111] The above two types of differences are weighted and combined to construct a comprehensive loading behavior difference index; the higher the difference value, the more it indicates that the loading behavior of the unit is not synchronized with the neighboring area, and it should be actively self-adjusted.

[0112] Parameter source: All parameters are obtained through real-time feedback without central instructions; a synchronous reference table can be set to set the maximum difference threshold allowed between regions; it can also be dynamically adjusted in combination with the historical consistency standard provided by the path modeling module.

[0113] Dynamic adjustment rules for loading step frequency and loading micro-displacement amplitude: The loading step frequency refers to the number of loading operations executed per unit time. If the deviation of the current loading point is small, the frequency can be maintained or slightly increased. If the deviation is large, the frequency should be decreased or the execution interval should be increased. If there is severe oscillation in the system, the loading should be paused until the predicted trend stabilizes. The loading micro-displacement amplitude refers to the movement amount of the loading head in a single loading operation.

[0114] The system can adjust the displacement proportionally according to the predicted offset value output by the "loading offset prediction function". Generally, upper and lower limits are set to prevent the loading mechanism from exceeding the limit. The combination rule: The frequency and amplitude need to be considered synchronously to prevent loading mutations caused by "high frequency + large displacement". A joint adjustment matrix can be set to jointly control according to the predicted offset and the degree of difference.

[0115] Strategy for minimizing loading offset between regions: By actively controlling the behavior of each loading unit, the loading offset in multiple loading regions is gradually reduced and approaches the overall loading target of the system.

[0116] Implementation method: Establish a prediction error comparison matrix between each loading point and its adjacent points; within each micro-loading cycle, select the loading behavior "closest to the average state" as the convergence benchmark; all loading points are slightly adjusted towards this benchmark behavior to form cluster-style self-organizing synchronization; this strategy can be implemented using fuzzy control, local linear adjustment, or adaptive learning methods.

[0117] Mechanism design for avoiding systematic oscillation and loading mutation problems: Oscillation identification method: If the direction of the loading error continuously alternates between positive and negative in two or more consecutive loading cycles; or the loading prediction value jumps by more than a certain amplitude within a short period of time, it is judged as oscillation; Coping mechanism: Pause the loading behavior; Reduce the micro-displacement amplitude; Narrow the prediction window length to enhance the short-term trend control ability.

[0118] Loading mutation determination: When the jump amplitude of the loading displacement or force value exceeds the system preset threshold for two consecutive cycles; the system triggers a "soft reset", returns to the previous fitting state and performs a secondary prediction.

[0119] Three-stage mechanism of local prediction - correction - re-prediction (prediction closed-loop): The first stage: Local prediction, use the loading offset prediction function to output the current trend and predict the loading response in the next one or two cycles. The second stage: Correction, according to the real-time feedback value, correct the error of the prediction result in the previous stage and adjust the behavior curve. The third stage: Re-prediction, use the new behavior state and the corrected error, re-enter the prediction function to obtain the next loading behavior suggestion and achieve dynamic iteration. This mechanism ensures that each loading unit can still maintain an efficient and stable "self-regulation - self-correction" ability under decentralized control.

[0120] The dynamic adjustment of the prediction window length refers to the number of historical data periods used by the prediction function, and its length affects the sensitivity and stability of the system prediction. Adjustment method: The higher the loading complexity (i.e., the greater the change in the force field gradient within the loading area), the shorter the prediction window is set; in areas with a relatively stable loading trend or a strong periodicity, the window can be set longer to enhance robustness; the system can set upper and lower limits (such as 3 periods to 10 periods) to automatically select the optimal length.

[0121] The loading offset prediction function provides a directional judgment of the current loading trend and an offset amplitude estimation; this part, as the execution layer, uses this estimation to adjust the loading rhythm and action details; the two together form a local closed-loop system of "prediction - decision - execution", and each loading unit operates independently; multiple loading units achieve cluster collaboration and synchronous trend convergence of the entire system through difference comparison and minimum offset strategies.

[0122] Suppose the loading unit 8 in the third area has a predicted offset direction of its current loading behavior consistent with the main trend of the system, and the offset amplitude is a small positive differential value. The system increases its stepping frequency by ten percent and keeps the loading amplitude unchanged; at the same time, it detects that the loading difference between it and the adjacent units (7 and 9) is below the system average; it is determined that no significant adjustment is required and only slight following is needed; if the loading trend continues to be consistent, the system will automatically extend the prediction window to 6 periods in the next cycle to improve prediction stability.

[0123] The two different types of feedback paths configured in the redundant closed-loop feedback control module are respectively the force value sensing path constructed based on resistance strain gauges and the displacement sensing path constructed based on the principle of laser interferometric ranging. The two paths collect data at different sampling frequencies during the loading process. Among them, the force value sampling interval is set at the millisecond level, and the displacement data sampling is set at the micron level. At the end of each loading cycle, data fitting processing is performed through a preset fusion calculation model, and it is judged whether the deviation value between the two paths exceeds the preset tolerance interval. If it exceeds, the fault-tolerant switching mechanism is triggered, and the multi-path fusion correction control strategy is enabled to adaptively adjust the loading action to avoid unstable oscillations during the loading process and ensure the reliability and safety tolerance of the loading response.

[0124] The force value sensing path uses resistance strain gauges pasted on the stressed structural parts of the loading mechanism; the resistance of the strain gauges changes linearly with the external stress; the strain data is converted into a voltage signal through a bridge circuit; after analog-to-digital conversion, it is sent to the local controller as the immediate feedback source of the loading force value; the force value data is collected at a high frequency (usually once per millisecond) during the loading process.

[0125] The displacement sensing path uses laser interferometry ranging technology to emit a laser beam to the loading head or the reference target surface; calculates the spatial displacement change of the loading head through the change amount of the interference fringes; this path has extremely high resolution and is usually used to capture tiny displacement changes at the micron level; the sampling frequency is lower than that of the force value path, but the sampling accuracy is extremely high; it can be set to trigger a record every change of several microns.

[0126] The system has a dual-clock synchronization mechanism: the force value data uses a uniform time step in milliseconds; the displacement data uses trigger sampling in units of distance (such as every five microns of movement); within each loading cycle, the system aligns data with different frequencies on the time axis through time tags or position tags; constructs a "loading behavior response table" uniformly for subsequent fusion calculations.

[0127] Fusion model objective: Construct a bivariate fitting model that can simultaneously consider the "loading force change trend" and the "displacement response speed" to capture whether the system loading behavior is consistent and stable. Implementation method: At the end of each loading cycle, correspond the force value data sequence and the displacement data sequence within this cycle; establish a fitting function: for example, use linear fitting, polynomial regression, locally weighted regression, etc.; compare the definition of the pre-deviation amount:

[0128] Deviation value = the maximum difference (or mean square error) between the fitting result of the force value path and the fitting result of the displacement path within the current cycle; the maximum allowable difference (referred to as the tolerance interval) can be set. For example, the system stipulates that the maximum allowable error does not exceed five percent. Tolerance setting basis: Set by the engineer according to the structure type, loading position, and test safety level; different material structures can set different thresholds (such as setting a low tolerance for carbon fiber structures and a relatively high tolerance for metal structures) to measure the deviation between the model output and the true feedback data, and calculate the difference in fitting residuals between the two paths.

[0129] Fault tolerance switching mechanism trigger conditions: If the fitting deviation within the current cycle exceeds the tolerance interval, or there is an increasing trend of deviation in two consecutive cycles; Execution action: The system enters the "fault tolerance mode"; automatically pauses the feedback control from the main path, temporarily switches to the standby path to dominate the loading behavior, and calls the historical fusion model for behavior smoothing prediction to temporarily maintain the loading action.

[0130] In the fault tolerance mode, read the data of both paths simultaneously, establish a real-time difference comparison table; use fusion strategies (such as least squares weighting, Bayesian fusion, or fuzzy weighted average) to synthesize the results of the two paths; if a certain path still has a relatively high weight of stability, then assign a higher reference value; generate a corrected loading instruction and transmit it to the execution unit. The output results are used for: dynamically fine-tuning the loading step; adjusting the loading speed; rewriting the prediction parameters of the next round of loading behavior for the asynchronous adjustment module to call.

[0131] Assume that the second region loading unit 9 is performing cyclic loading. The frequency of the force value data transmitted back by its resistance strain gauge is once per millisecond, and the displacement change feedback by the laser rangefinder is recorded once every five micrometers. During the current loading cycle, the system records a total of 200 groups of force value data and 70 groups of displacement data. Using polynomial regression to fit the two data sequences respectively, it is found that the difference in fitting residuals reaches 6% of the preset tolerance upper limit. The system triggers the fault tolerance mechanism and simultaneously runs the fusion control module to smoothly correct the next loading behavior.

[0132] The loading path modeling and reproduction module records the dynamic parameters of the loading process by point-by-point sampling, and synchronously maps the force value, loading order, loading time, and spatial position through constructing a high-dimensional path matrix. The method for constructing the path matrix uses the three-dimensional Bezier curve path interpolation method to construct the initial model. During the test process, the real-time collected data is fitted with the historical model, and the loading trajectory is adjusted in real time through the fitting residuals, and the path record is updated to form a path data set that can be called multiple times. During the path reproduction stage, the system preferentially calls a set of historical path data with the highest fitting degree as the current loading reference trajectory, and sets a deviation threshold to judge the path reproduction accuracy. If the path fitting deviation exceeds the critical value, the path is automatically de-escalated and the current loading strategy is re-planned.

[0133] During the execution of each loading behavior, the system records the following data for each loading point in real time: the current loading force value; the sequential number of the current loading execution (e.g., which step of the loading); the timestamp when the current loading occurs; the spatial position coordinates of the current loading head or loading execution mechanism (such as three-dimensional X-Y-Z coordinates). Each set of complete parameters collected constitutes a "loading sampling point". Sampling frequency: It can be set to sample once per loading micro-step, or it can be set to a fixed interval time (such as once every 10 milliseconds); the accuracy requirement can be adjusted in combination with the material response frequency.

[0134] Synchronous mapping construction method of the high-dimensional path matrix: Data structure description: The path matrix is a tabular structure indexed based on the loading step sequence. Each row represents a loading behavior; each row contains: force value, loading order, loading time, loading position; synchronous mapping means that these information have a one-to-one correspondence, that is: what is the force value of the 10th step of loading; at what time does this step occur; where is the loading point in space; what is its order in the entire loading path. Structural characteristics: It can be implemented through list nesting, four-dimensional tensor structure, or hash mapping; this path matrix serves as the digital coding basis of the historical trajectory for subsequent path calls and comparisons.

[0135] The three-dimensional Bezier curve is an interpolation method used to generate smooth spatial paths, capable of generating continuously differentiable paths with a set of control points; in this system, the position parameters of multiple sampling points are used as control point inputs to construct a curve that fits the entire loading path.

[0136] Implementation steps: Extract several key nodes from the initial round of loading experiments (for example, take one point every 5 points); use the three-dimensional coordinates of these nodes to establish a Bezier control point sequence; call the Bezier interpolation algorithm to generate a continuous spatial loading path; attach the loading force and time synchronized with each point position on the path to the generated curve to form a "spatial loading path model". Parameter description: The number of control points determines the smoothness and fitting ability of the path; the Bezier weight function is distributed at equal intervals or according to the actual loading rhythm; the initial model can be stored as an independent template for multiple loading batches to call.

[0137] Implementation process: During the current loading process, the real-time data (force value, displacement, time) collected by the system is continuously compared with the historical Bezier path model; the comparison method uses cumulative point-by-point difference or moving window mean square error comparison; if the current loading path deviates from the historical model trend, the system performs path correction through the "residual feedback mechanism": the residual is defined as the weighted sum of the coordinate deviation, time deviation, and force value deviation between the current real-time point and the historical curve point; the correction methods include: updating the control point position, adjusting the fitting interpolation density, and local path regeneration.

[0138] Construction and call mechanism of the path dataset: Content of the path dataset: The path matrix of all historical loading processes, its corresponding Bezier model; fitting residual values (which can be used as path priority sorting indicators); label information of each model (loading object, structure number, area number, etc.);

[0139] Ways that can be called multiple times: When the loading system calls the "path reproduction" function, first screen out the matching historical path models through conditional retrieval; the system matches the current loading data with the candidate paths one by one through the fitting algorithm; the fitting degree is jointly scored by multiple factors such as the mean residual and path trend similarity; the system selects the path with the highest score as the "benchmark trajectory" for the current loading to reproduce the path.

[0140] Deviation threshold setting and path degradation processing mechanism: The deviation threshold can be set in advance by engineering personnel, such as the path fitting error shall not exceed three percent; or automatically recommended by the system according to the structure type, for example, the tolerance can be relaxed for flexible structures and the accuracy needs to be improved for rigid structures; the error measurement methods can use dynamic Euclidean distance, mean Bezier fitting residual, maximum trajectory offset, etc.

[0141] System behavior after exceeding the critical value: The system determines that the current path model is no longer applicable to the current working condition; starts the "path degradation" strategy, that is, no longer uses the historical path and switches to the real-time dynamic prediction mode; at the same time, records the current loading behavior as a new path model for future use; the loading control logic switches to the "prediction - feedback - replanning" adaptive path generation mechanism.

[0142] Suppose a wing loading test of an ultra-high aspect ratio aircraft is in progress, and the target area is the third area. The loading unit retrieves three historical paths with relatively high fitting degrees from the historical path database: The system compares the 15th sampling point of the current loading with the 15th point of the historical path and finds that the deviation of the current point from path A is 5% and from path B is 2%; the system preferentially selects path B as the reference trajectory for this loading; during the loading process, the control points of path B are locally automatically adjusted to adapt to the current microenvironment disturbance; if the deviation in the subsequent 30th step exceeds 3%, the system performs a degradation process and enters a new path regeneration process.

[0143] A data sharing channel is configured between the multi-region loading execution module and the boundary coordination control module. A loading status sharing table is established through a multi-threaded information synchronization mechanism. This sharing table is regularly pushed with status data by each loading unit and is regularly read and integrated. During the loading task, the loading order and priority scheduling strategy are dynamically updated based on the loading force change trend in the sharing table and the feedback stability of each unit. This strategy adopts a weight increase - feedback delay penalty mechanism, enabling the loading unit with fast response and good fitting to obtain a higher instruction frequency and control priority, thereby improving the overall loading response efficiency and reducing the probability of loading path oscillation.

[0144] A communication channel is established between the loading control systems of each region for information sharing outside local control. Implementation forms: Bus communication (such as CAN, RS485), Ethernet data link, or local wireless network can be used; the channel is responsible for synchronously receiving and sending status data from each loading unit at a fixed period and does not participate in the issuance of the loading control instructions themselves.

[0145] The communication content includes: the current loading force value; the loading speed and direction; the prediction error and feedback stability status within this cycle; the feedback deviation of the previous loading behavior.

[0146] The sharing table is a centralized dynamic data table structure that records the operating status of all loading units. Multi-threaded mechanism description: Each loading unit runs an independent thread or asynchronous task and periodically pushes data; data is collected from all threads in each cycle and written into the sharing table; all loading regions share this table to achieve synchronous read and write.

[0147] Table Structure Description: Each row of the shared table corresponds to a loading execution unit, including fields such as: unit number; last loading time; loading amplitude; current feedback error; path fitting degree of the previous cycle; current status flag (such as "stable", "abnormal", "needs adjustment").

[0148] Dynamically update the loading order and priority scheduling strategy based on the loading force change trend and feedback stability in the shared table: The system judges which loading units respond faster and more reliably according to the information summarized in the shared table, so as to decide which units execute the next loading first. Method for judging the loading force change trend: Compare the loading force values of each loading unit in the past 3-5 cycles; calculate the slope, volatility, stable interval, etc. of the loading force; The loading unit with a stable trend and a high linearity of the fitting curve is regarded as a "stable source".

[0149] Judgment of feedback stability: Statistically analyze the change of the feedback error value of each unit; if the error remains within the set threshold for multiple consecutive cycles, it is marked as "feedback stable"; if the error mutates or oscillates, it is marked as "feedback unstable".

[0150] Priority calculation logic of the "weight increase - feedback delay penalty mechanism": Weight increase part: Assign a higher priority weight to the unit with small loading error, high path fitting degree, and fast response speed; The weight score is constructed by the following formula: First, set three indicators: inverse ratio score of loading error, fitting degree score, and inverse ratio of response delay; Each item is normalized on a percentage basis; The final priority score is obtained through linear weighted average.

[0151] Feedback delay penalty mechanism: If the feedback of a certain loading unit lags (such as exceeding 30% of the average feedback cycle), or there is a lack of feedback; then its weight is set with a penalty factor (such as multiplied by 0.7), and its priority is temporarily downgraded; Prevent "units with poor feedback" from dominating the loading rhythm and reduce path stability.

[0152] Instruction frequency: The number of control signals that the loading unit can receive per unit time; Control priority: The central scheduler preferentially selects it to execute the next loading action; It can be achieved by sorting the task queue of the scheduling controller, for example, arranging high-priority units in the front list; The system can set the minimum priority frequency and the maximum priority cycle to avoid over-biasing towards a certain unit.

[0153] High-priority loading points respond faster to system adjustment instructions; The system actively avoids units with abnormal feedback and prone to path errors; The scheduling is more robust, the loading rhythm is more uniform, avoiding path discontinuity and sudden jumps; It operates in coordination with the aforementioned path modeling and residual correction module to jointly form a feedback closed-loop.

[0154] Example description: In actual tests, there are a total of 6 loading execution units in the first area. Unit No. 4 in the past 3 loading cycles: The change in the loading force shows a linear increase, and the error is controlled within the set tolerance; the feedback signal delay is less than 30% of the average value; the fitting residual score is the highest in the whole group. After the system scheduler reads the shared table data: It assigns a weight score of 92 to this loading unit; compared with the average weight of 73 of other units, it is significantly higher; it places it at the top of the priority queue for the next loading task; if its response is stable, the system will maintain its high-frequency scheduling state in multiple loading cycles; if there is a subsequent feedback delay or a sharp increase in error, the system will automatically reduce its weight and switch to other candidate units.

[0155] A dynamic path feedback linkage mechanism is established between the asynchronous self-synchronizing adjustment module and the loading path modeling and reproduction module. After each loading action is completed, the current actual loading path is subjected to multi-parameter fitting analysis with the historical path, and the path difference index is calculated. This index consists of three parts, namely the total loading force difference ratio, the path shape residual factor, and the loading point time offset rate. The calculation result is fed back to the next round of loading step adjustment link as a path correction factor to be embedded, controlling the loading path to gradually converge to the historical best path to achieve continuous path evolution and dynamic fine-tuning capabilities during the loading process.

[0156] The present invention establishes a dynamic closed-loop channel between the loading path execution and the loading control prediction, enabling the actual execution result of the path to affect the loading control algorithm in real time. After each loading action is completed, the loading path modeling module immediately fits the current execution path with the historical path; the difference index generated by the fitting result is transmitted to the asynchronous self-synchronizing adjustment module; in the next round of loading control, this difference index is embedded as a parameter affecting control quantities such as the loading step, frequency, and direction; realizing a closed-loop chain of control - path execution - difference judgment - control adjustment.

[0157] Implementation method of multi-parameter fitting analysis (path difference analysis) operation process: Align the path data (point-by-point force value, displacement, time) of the current loading cycle with the corresponding points of the historical path data; use methods such as sliding window and node matching to establish a two-axis comparison structure of time and space; conduct segmented analysis on the overall path profile, the loading peak position, and the time characteristics of the loading sequence; the fitting result is not limited to the spatial trajectory, but also includes the behavioral characteristics of the loading sequence.

[0158] Total loading force difference ratio: Defined as the ratio difference between the total force value of all loading points in the current loading path and the total force value of the reference path; by integrating all loading force values point by point, the total loading energy is obtained; compare the difference ratio between the current path sum and the historical path sum; this value is used to reflect the overall force matching degree of the structure.

[0159] Path shape residual factor: It reflects the deviation of the current path in geometric form; compares the displacement differences between the current loading path and the historical path in three-dimensional space; uses the point-by-point trajectory difference and the overall path curvature trend difference to evaluate the fitting degree, corresponding to the judgment of "whether the path is offset and whether the bending form is consistent", and can be scored and evaluated by a machine learning model such as a convolutional neural network model or an expert pre-trained.

[0160] Loading point time offset rate: Compares the difference between the actual loading time of each loading point in the current loading behavior and the loading time of this point in the historical path; counts the proportion of the average offset or the maximum offset in the historical loading cycle; used to judge "whether the loading rhythm is advanced or lagged".

[0161] The system sets the weights of the three factors according to the loading task requirements; for example, for a precision-first task, the weight of the path shape residual can be increased; for a real-time-first task, the weight of the time offset rate can be increased. Each index is normalized to a percentage value range; the three are weighted and summed after multiplying by the weights to obtain the total value of the "path difference index"; the system sets whether path adjustment is required and the adjustment amplitude according to this value.

[0162] Control embedding mechanism: Before the asynchronous self-synchronization adjustment module executes the loading step adjustment, it reads the path difference index of the previous round; uses this index as the "path feedback correction factor" to adjust the next loading step amplitude, loading speed or path prediction strategy; if the difference value is larger, the loading behavior is more conservative (i.e., smaller steps, slower loading); if the difference value is extremely low, it is allowed to improve the loading efficiency or execute the predicted path extension. Joint use with other factors: Together with the output of the loading offset prediction function, it constitutes the "behavior adjustment parameter set"; participates in the micro-step frequency control, loading micro-amplitude adjustment, and scheduling priority strategy update.

[0163] The system continuously records the change of the difference index in multiple loading cycles. If the path difference index continuously decreases, it indicates that the loading path is gradually converging, and the system can automatically shorten the path prediction window in subsequent cycles to improve efficiency. If the index rebounds or oscillates, the system will trigger a fine-tuning mechanism, including: adjusting the interpolation density of control points, reconstructing the non-linear path fitting model, local path substitution or automatically switching the reference path.

[0164] Example illustration: In a wing static load test, the system calls the historical path model X as the reference trajectory. After the end of the third loading cycle, the system fits the current path with the X path and finds that: the total load force difference ratio is 2.5%; the path shape residual factor is 3%; the time offset rate is 5%; the system assigns weights of 30%, 40%, and 30% to the three items respectively, and calculates that the comprehensive path difference index is approximately 3.75%; in the next loading cycle, the asynchronous self-synchronization module reduces the loading step amplitude by 5% accordingly and adjusts the prediction window length to be shortened to four cycles; if the subsequent path difference continues to decrease, the system will gradually relax the loading rhythm and finally converge to the spatial and temporal characteristics of the historical path X.

[0165] For better understanding, the principles of the present invention are sorted out as follows:

[0166] Distributed execution architecture and basic control of loading behavior: The present invention is based on a large-stroke static loading system for strength testing of aircraft wing surfaces. The overall architecture adopts a multi-region loading execution mode, divides the entire wing surface structure into multiple loading regions, and deploys loading execution units with local control and feedback capabilities in each region. Each loading unit realizes loading state perception, local path fitting, and autonomous adjustment of loading steps through an embedded microprocessor, and dynamically generates a higher-matching loading curve in combination with the structural stiffness and historical stress response of the region where it is located. The units do not rely on central continuous control instructions, but form the basis of asynchronous but converging loading behavior through autonomous adjustment of the loading step and the amplitude of the loading micro-displacement. Boundary coordination control modules are set between regions to set boundary transition zones, and flexible fusion factors and transition functions are used to control the continuous change of the loading force at the boundary, ensuring the overall smooth transition of the force field and avoiding regional fractures or loading mutations.

[0167] Asynchronous self-synchronization loading control logic and feedback closed-loop mechanism: In the loading control process, the asynchronous self-synchronization adjustment module plays a core algorithm decision-making role. Through the loading offset prediction function, combined with the historical loading trend and the current error change, it dynamically calculates the offset direction and amplitude of the loading behavior, and adjusts the loading frequency and step accordingly to achieve a three-stage closed-loop of local prediction - feedback correction - re-prediction within the system. To ensure the stability and robustness of the system operation, two feedback paths are configured in the loading execution unit to collect the loading force value and the spatial displacement respectively, and the residual difference between the two paths is compared through a fusion calculation model. When the deviation exceeds the preset tolerance, the fault tolerance mechanism and the multi-path correction control strategy are automatically triggered to keep the loading action stable. At the same time, the loading path modeling and reproduction module records the loading force, loading displacement, loading time, and loading point sequence in real time, constructs a high-dimensional path matrix, and constructs a continuous spatial path model through three-dimensional Bezier interpolation to support the traceability, reproducibility, and historical comparison of the loading process trajectory.

[0168] Dispatch coordination and path linkage are realized to achieve self-evolving control of the loading behavior: Each loading area is interconnected through a data sharing channel. A multi-threaded synchronous data sharing mechanism is established between the loading execution module and the boundary coordination module, and with the loading status sharing table as the core carrier, the periodic push and integration of loading data are realized. The system dynamically adjusts the priorities of the loading units and the scheduling order of the loading tasks according to multi-dimensional indicators such as loading error, path fitting degree, and feedback response speed in the sharing table. Through the "weight increase - feedback delay penalty" strategy, the loading units with high stability, fast response, and excellent fitting obtain higher execution frequencies. After each loading action ends, the system fits the current loading path with the historical path and calculates the path difference index, which consists of the total loading force difference, trajectory shape residual, and time offset. This difference value is passed to the adjustment module as a path correction factor and directly participates in the adjustment of the next loading step size, enabling the loading path to be dynamically corrected and gradually converge to the optimal trajectory, thereby constructing a continuous evolution and intelligent adaptive control system for the loading behavior.

[0169] The above formulas are all dimensionless and take their numerical calculations. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0170] It should be understood that in various embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0171] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0172] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and modules can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here.

[0173] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present application and should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A large-stroke static loading system for aircraft wing surfaces in aircraft strength testing, characterized in that Comprising: A multi-region loading execution module that divides the aircraft wing structure into multiple loading regions. Each region is configured with an execution unit having local loading control capabilities. The execution unit includes a local controller and a sensing element, and performs loading actions without relying on central continuous instructions, and senses the force state and displacement changes at its location; A boundary coordination control module that has a boundary transition zone between loading regions. By setting a flexible fusion factor and a distributed collaborative adjustment rule, a continuous and adjustable force field edge connection relationship is formed between different loading regions to meet the distribution response requirements of high aspect ratio wings under high load torsion states; An asynchronous self-synchronization adjustment module. Based on the local feedback results, the loading unit adjusts the loading step through a preset loading offset prediction function, and can achieve the approximation and fusion of local loading behaviors to the global loading trend without a global synchronization signal; A redundant closed-loop feedback control module. The loading unit is configured with two different types of feedback paths, and automatically calls the backup path or generates a fusion signal to form a stable feedback judgment when the main path fails; A loading path modeling and reproduction module that records the force value changes, spatial displacement paths, and loading sequences of each loading unit during the loading process, forms a replayable data trajectory set, and realizes loading path reproduction control by comparing the fitting error between the current path and the historical path during subsequent loading processes.

2. The large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 1, characterized in that, In the multi-region loading execution module, the local controller of each loading execution unit includes an embedded microprocessor structure, which has the capabilities of loading state sensing, action planning, and local fitting of the loading curve. When performing the loading task, it dynamically generates a local loading curve based on historical stress data and current feedback information. This curve is updated at least once in each loading cycle, and the update algorithm performs bivariate optimization based on the local error expansion function within the region and the structural stiffness adjustment parameter. The fitting error of the loading curve is controlled within a set threshold range, and this threshold is set in advance according to the wing material properties and test conditions.

3. The large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 2, wherein, The boundary coordination control module includes an adjustable flexible force fusion factor, and the setting of this factor is based on the relative deformation gradient difference between the boundary units between loading regions, and combines the force extension trajectories collected in historical experiments to achieve smooth loading transition through optimizing the calculation of the boundary transition function during continuous boundary loading.

4. A large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 3, characterized in that, The selection of the boundary transition function includes three types: linear transition function, logarithmic smoothing function, or cubic spline transition function. Each function type is selected according to the average stiffness change rate of the target wing in the boundary area; During the loading process, it is updated once in each iteration step through the distributed collaborative adjustment rule, and a boundary coordination matrix is established through the feedback data of adjacent boundary loading units.

5. The large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 4, characterized in that, In the asynchronous self-synchronization adjustment module, the loading offset prediction function is constructed based on the historical loading trend and the current real-time error fitting curve during the loading process. The prediction model adopted selects any one of the three models: weighted moving average function, time series residual regression model, or local sample non-linear mapping function. The output result of the prediction function is used to guide the dynamic correction of the loading step.

6. The large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 5, characterized in that, Each loading execution unit autonomously adjusts the loading step frequency and the amplitude of the loading micro-displacement according to the difference in loading behavior between its affiliated loading area and the adjacent area. Through the strategy of minimizing the loading offset between regions, the distributed loading behavior gradually approaches the global synchronization goal.

7. A large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 6, characterized in that, The two different types of feedback paths configured in the redundant closed-loop feedback control module are respectively the force sensing path based on resistance strain gauges and the displacement sensing path based on the principle of laser interferometric ranging. The two paths collect data at different sampling frequencies during the loading process. Among them, the force value sampling interval is set at the millisecond level, and the displacement data sampling is set at the micron level. At the end of each loading cycle, data fitting processing is performed through a preset fusion calculation model, and it is judged whether the deviation value between the two paths exceeds the preset tolerance interval. If it exceeds, the fault-tolerant switching mechanism is triggered, and the multi-path fusion correction control strategy is enabled to adaptively adjust the loading action.

8. A large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 7, characterized in that The loading path modeling and reproduction module uses the point-by-point sampling method to record the dynamic parameters of the loading process, and synchronously maps the force value, loading order, loading time and spatial position by constructing a high-dimensional path matrix. The path matrix construction method uses the three-dimensional Bessel curve path interpolation method to construct the initial model. During the test, the real-time collected data is fitted with the historical model, the loading trajectory is adjusted in real time through the fitting residual, and the path record is updated to form a path data set that can be called multiple times.

9. The large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 8, characterized in that, A data sharing channel is configured between the multi-region loading execution module and the boundary coordination control module. A loading status sharing table is established through a multi-thread information synchronization mechanism. This sharing table is regularly pushed with status data by each loading unit and regularly read and integrated. During the loading task, according to the change trend of the loading force in the sharing table and the feedback stability of each unit, the loading order and the priority scheduling strategy are dynamically updated. This strategy adopts the weight promotion-feedback delay penalty mechanism, so that the loading unit with fast response and good fitting obtains a higher instruction frequency and control priority.

10. A large-stroke aircraft wing surface static loading system for aircraft strength testing according to claim 9, characterized in that, A dynamic path feedback linkage mechanism is established between the asynchronous self-synchronization adjustment module and the loading path modeling and reproduction module. After each loading action is completed, the current actual loading path and the historical path are subjected to multi-parameter fitting analysis, and the path difference index is calculated. This index consists of three parts, namely the total loading force difference ratio, the path shape residual factor and the loading point time offset rate. The calculation result is fed back to the next round of loading step adjustment link as a path correction factor to be embedded.

Images