Aircraft large component butt joint tolerance distribution and self-adaptive attitude adjustment method

Abstract

The invention discloses a butt joint tolerance distribution and self-adaptive attitude adjustment method for large components of an aircraft, and belongs to the technical field of digital assembly and automatic manufacturing of aircrafts. The objective of the invention is to solve the technical problems of unreasonable tolerance distribution and the like caused by dependence on a single measurement means in the docking process of large components of an existing airplane. In the posture adjusting process, a laser tracker and a visual sensor are synchronously utilized to obtain three-dimensional coordinates and key docking feature images of a target point, and a fusion posture model is generated through space-time registration and fusion; accessing a preset tolerance database, and calculating an optimal attitude adjustment target pose which meets all interface tolerance constraints and enables an attitude adjustment displacement weighting norm to be minimum; and automatically generating a posture adjusting path, performing real-time closed-loop correction on the target posture and path in the movement process, and dynamically adjusting and optimizing the target weight according to the load of the posture adjusting platform. The method is mainly used for achieving high-precision, high-efficiency and high-reliability automatic butt joint assembly of large components such as aircraft wings and fuselages.

Description

Technical Field

[0001] This invention belongs to the field of aircraft digital assembly and automated manufacturing technology, specifically involving a method for the allocation of docking tolerances and adaptive attitude adjustment for large aircraft components. Background Technology

[0002] In the field of automated docking and assembly of large aircraft components, existing technologies mainly rely on discrete measurement and attitude adjustment processes. Typically, laser trackers are used to measure the component's attitude, followed by planning and executing an attitude adjustment path based on the measurement results. This method faces several inherent challenges in practical applications.

[0003] First, relying on a single laser tracker for measurement has limitations. While laser trackers can provide high-precision absolute pose data in a global coordinate system, their measurement results reflect the position of target points fixed on the components. This makes it difficult to directly and accurately capture the subtle relative positional relationships between key mating features directly related to the docking function. For example, the relative deviations of key features such as hole groups, surfaces, or docking contours on two mating components cannot be fully characterized by the absolute coordinates of the target points. This disconnect between measurement data and functional requirements means that attitude adjustment calculations based solely on absolute pose may not ensure that all critical interfaces simultaneously meet their respective tolerance requirements.

[0004] Secondly, the reliance on a single measurement data source also introduces reliability risks. In complex workshop environments, the laser tracker's measurement beam may be momentarily interrupted or generate noise due to personnel, equipment obstructions, or air disturbances, leading to data loss or decreased accuracy. In such cases, the system lacks available redundant data for verification or supplementation, potentially resulting in calculations based on inaccurate pose information, leading to layout errors or pose adjustment failures. Although attempts have been made to introduce other sensors such as vision sensors, fusing data from different sensors presents challenges. Laser data and visual images originate from different coordinate systems, possessing different data structures and temporal sequences. Without precise spatiotemporal synchronization and coordinate system unification, simple data overlay not only fails to provide effective information but may also cause confusion in system state judgment due to data conflicts. Establishing a stable and reliable fusion mechanism is a practical engineering challenge.

[0005] Furthermore, existing attitude adjustment processes are mostly open-loop or have only limited feedback. That is, the target pose is calculated and the path is planned before the attitude adjustment path is executed; once execution begins, it moves according to the preset path until the endpoint, at which point measurement and verification are performed. During this process, if the actual movement trajectory of the component deviates from the expected trajectory due to component deformation, platform servo errors, or unexpected interference, the system cannot detect and correct it midway through the movement. This deviation accumulates, potentially causing the component to collide with surrounding structures during movement, or revealing that the docking tolerance is still out of tolerance after reaching the target position, requiring time-consuming manual intervention and readjustment, thus reducing the efficiency and reliability of the docking process.

[0006] Furthermore, even with closed-loop control, ensuring process smoothness during path correction remains a challenge. When the system replans the path based on new measurement data, simply generating a new trajectory starting from the current stationary point and ending at the newly calculated target point, while ignoring the actual speed of the components at the current moment, may result in sudden speed changes at the junction of the old and new trajectories. This could lead to mechanical shocks, component vibrations, or exceed the response capability of the drive unit, affecting attitude adjustment accuracy and equipment safety.

[0007] In summary, existing technologies face challenges such as incomplete measurement information leading to inaccurate tolerance compliance judgments, difficulties in multi-source data fusion affecting system robustness, lack of real-time adaptive capabilities during attitude adjustment, and the potential for motion shocks caused by path replanning. These issues hinder further improvements in the precision, efficiency, and automation level of large aircraft component docking and assembly. Summary of the Invention

[0008] One object of the embodiments of the present invention is to solve at least the above-mentioned problems and / or defects, and to provide at least the advantages described below.

[0009] Another objective of this invention is to provide a method for allocating docking tolerances and adaptive attitude adjustment for large aircraft components.

[0010] This addresses the problems of existing aircraft component docking methods, which rely on a single measurement method, resulting in unreasonable tolerance allocation, lack of adaptive closed-loop correction during attitude adjustment, and difficulty in simultaneously optimizing interface tolerance compliance and attitude adjustment platform motion performance.

[0011] This addresses the problem of inaccurate fusion pose models caused by the difficulty in accurately fusing multi-source, heterogeneous measurement data from laser trackers and vision sensors due to differences in coordinate systems and time sequences.

[0012] This addresses the issue that during closed-loop correction, a simple switch in the attitude adjustment path can cause sudden changes in motion speed, leading to equipment shock and vibration, which in turn affects the stability and safety of attitude adjustment.

[0013] To achieve the above-mentioned objectives, the present invention employs the following technical solution: A method for tolerance allocation and adaptive attitude adjustment of large aircraft components includes the following steps: S1) During the attitude adjustment process, the three-dimensional coordinate pose data of the target point on the part to be docked is obtained simultaneously using a laser tracker, and the image data of the key docking features between the part to be docked and the reference part is obtained using a vision sensor; the three-dimensional coordinate pose data and the image data are spatiotemporally registered and fused to generate a fused pose model containing the absolute pose of the part and the details of the relative docking features. S2) Access the preset tolerance database, which defines the allowable deviation range of all key docking interfaces in six degrees of freedom; using the fused pose model as input and the tolerance database as constraints, calculate an optimal pose adjustment target pose through a multi-objective optimization algorithm. This optimization process includes two objectives: the first objective is to maximize the safety margin that satisfies the tolerance constraints of all docking interfaces, and the second objective is to minimize the weighted norm of the pose adjustment displacement vectors of all support points, where the weight coefficients are set based on the motion performance of each pose adjustment drive unit; the target pose must ensure that the expected deviation of all docking interfaces falls within their corresponding allowable deviation range, and that the weighted norm of the pose adjustment displacement vectors required by all support points is minimized; S3) Based on the difference between the optimal pose adjustment target pose and the current fused pose model, automatically generate a collision-free and smooth pose adjustment path; control the multi-degree-of-freedom pose adjustment platform to drive the docking component to move according to the pose adjustment path, and repeat steps S1) and S2) in real time during the movement to perform closed-loop correction of the pose adjustment target pose and the pose adjustment path until the docking component reaches the optimal pose adjustment target pose. In the closed-loop correction process of step S3, the weight coefficient of the attitude adjustment displacement vector in the second target is dynamically adjusted according to the real-time measured load of the attitude adjustment platform. When the load is close to the threshold, the weight coefficient is increased to prioritize ensuring motion stability and equipment safety.

[0014] Preferably, the spatiotemporal registration and fusion in step S1) specifically involves: Using a fixed calibration plate or specific feature points on a component as a common reference, a transformation relationship between the laser tracker coordinate system and the vision sensor coordinate system is established. Through timestamp synchronization, the three-dimensional coordinate pose data and image data acquired at the same time are aligned, and the three-dimensional coordinate pose data are smoothed using Kalman filtering or extended Kalman filtering algorithms. At the same time, the feature point coordinates extracted from the image data are used to supplement and correct the smoothed pose data to generate the fused pose model.

[0015] Preferably, the tolerance database in step S2) is stored in matrix form, with each row corresponding to a key docking interface. Its contents include, but are not limited to, the interface number, the theoretical position in the component coordinate system, and the allowable positive deviation value and allowable negative deviation value in each degree of freedom direction.

[0016] Preferably, in step S3), generating a collision-free and smooth attitude adjustment path specifically involves: using B-spline curves or fifth-order polynomial interpolation to plan a continuous motion trajectory with continuous first and second derivatives between the initial attitude adjustment pose and the optimal attitude adjustment target pose; and during the planning process, performing dynamic virtual collision interference checks based on the three-dimensional model of the component to ensure that the path is collision-free.

[0017] Preferably, the closed-loop correction in step S3) specifically involves: during the attitude adjustment path execution process, re-executing steps S1) and S2) at a fixed frequency; if the deviation between the newly calculated optimal attitude adjustment target pose and the endpoint pose of the current execution path exceeds a preset threshold, then based on the latest fused pose model and the optimal attitude adjustment target pose, the unexecuted portion of the attitude adjustment path is replanned online. The preset threshold is a position deviation of 0.05 mm to 0.2 mm and an attitude angle deviation of 0.01° to 0.05°.

[0018] Preferably, in step S1), when there is a non-negligible conflict between the laser tracker data and the visual sensor data, the following data reliability arbitration strategy is executed: A first confidence weight is set for the laser tracker data, and a second confidence weight is set for the visual sensor data; the first confidence weight is positively correlated with the geometric configuration accuracy factor of the target point relative to the laser tracker, and the second confidence weight is positively correlated with the image sharpness and illumination uniformity of the key docking features; The conflicting data are weighted and fused based on the first and second confidence weights to generate the fused pose model.

[0019] Preferably, after establishing the coordinate system transformation relationship using specific feature points on a fixed calibration plate or component as a common reference, an online self-calibration step is also included: During the orientation adjustment process, the multi-degree-of-freedom orientation adjustment platform is periodically controlled to cause the component to be docked to produce a small, known, preset motion; By comparing the actual motion calculated from laser tracker data with the actual motion calculated from vision sensor data, the current transformation relationship residual between the laser tracker coordinate system and the vision sensor coordinate system is calculated. When the residual exceeds the allowable range, the original transformation relationship is updated using the latest calculated transformation relationship to achieve online self-calibration of the coordinate system. The allowable range is a translation residual of 0.05 mm to 0.2 mm and a rotation residual of 0.02° to 0.1°. Small, known preset movements are translations of 1 mm to 5 mm and rotations of 0.5° to 2°.

[0020] Preferably, the multi-objective optimization algorithm adopts a non-dominated sorting genetic algorithm with elite retention, and during algorithm initialization, an initial population based on historical successful orientation adjustment cases is introduced: from the historical successful orientation adjustment cases, the orientation adjustment target pose corresponding to the case that is similar to the current part model and initial pose to be docked is extracted, and it is injected into the initial population as a high-quality gene to accelerate optimization convergence and avoid getting trapped in local optima.

[0021] Preferably, when performing dynamic virtual collision interference checks based on the 3D model of the component, a dynamic safety boundary is also extended for the component to be docked and the reference component. The size of the dynamic safety boundary is positively correlated with the velocity value of the component to be docked at the corresponding position calculated based on the attitude adjustment path. When replanning the posture adjustment path online, the following smooth transition strategy is adopted to avoid motion shock: Using the current actual pose and velocity of the components to be docked as the new starting state, and the newly calculated optimal pose target as the endpoint, a B-spline curve or fifth-order polynomial trajectory is regenerated, ensuring that the newly generated trajectory is continuous with the currently executing trajectory in pose and velocity at the starting point. The basic safety boundary is set to 10 mm to 20 mm. When the velocity of the components to be docked is in the range of 0 mm / s to 50 mm / s, the safety boundary size increases linearly with velocity, up to a maximum of 150% to 200% of the basic value (i.e., 15 mm to 40 mm).

[0022] Preferably, the weighting coefficients are dynamically adjusted based on the real-time measured load of the attitude adjustment platform, specifically through a fuzzy controller: The inputs to the fuzzy controller are the load rate and the load rate change rate, and the output is the adjustment amount of the weight coefficient; When the load rate is high and the load rate change rate is positive, a large positive adjustment amount is output to quickly increase the weighting coefficient and reduce the optimization priority of the attitude adjustment displacement. The load safety threshold is set to 85% ~ 95% of the rated load of the attitude adjustment platform drive unit. Load rate input range: 0% ~ 120% (percentage of rated load). Load rate change rate input range: -10% / s ~ +10% / s. Weighting coefficient adjustment amount output range: 0.5 ~ 3.0 (as a multiplier, applied to the initial weighting coefficient).

[0023] Compared with the prior art, the advantages and beneficial technical effects of the present invention are: This invention utilizes a multi-sensor fusion pose model and tolerance database for multi-objective optimization and closed-loop attitude adjustment, which can effectively improve the compliance rate and allocation rationality of docking tolerance. At the same time, it adaptively balances docking accuracy and equipment motion efficiency during attitude adjustment, thereby enhancing the intelligence level and robustness of the entire docking system.

[0024] This invention establishes a precise and reliable coordinate system transformation and time synchronization mechanism through data fusion, and uses a filtering algorithm to fuse heterogeneous data, thereby generating a more accurate and stable component fusion pose model, providing a high-quality data foundation for subsequent optimization calculations.

[0025] This invention manages complex multi-interface tolerance information in a standardized matrix form through a tolerance database structure, enabling optimization algorithms to efficiently and unambiguously read and utilize the constraints of each degree of freedom, thereby improving the efficiency and accuracy of tolerance judgment and optimization calculation.

[0026] This invention utilizes a path planning method, employing high-order continuous mathematical curves for trajectory planning and combining it with virtual collision checks. This ensures a smooth, stable, and absolutely safe posture adjustment process, effectively avoiding the risks of jamming, vibration, or collisions during movement.

[0027] This invention achieves real-time monitoring and feedback control of the entire attitude adjustment process through a closed-loop correction mechanism. It can promptly correct path deviations caused by various factors, significantly improve the accuracy and success rate of the final docking, and reduce the need for manual intervention.

[0028] This invention employs a data arbitration strategy to intelligently weight data based on the quality of measurement conditions when data conflicts occur. This allows the system to output the most reliable fused pose results even under complex conditions, greatly improving the reliability of the system's decision-making in non-ideal measurement environments.

[0029] This invention, through an online self-calibration process, can automatically detect and correct errors in the transformation relationship between sensor coordinate systems, avoiding long-term measurement accuracy degradation caused by calibration parameter drift, ensuring the stability and accuracy of the system during long-term operation, and reducing maintenance downtime.

[0030] This invention optimizes the algorithm's initialization strategy and leverages historical success experience to guide the search direction, which significantly improves the convergence speed and solution quality of the optimization algorithm. This enables the system to plan high-performance attitude adjustment schemes more quickly and improves docking efficiency.

[0031] This invention enables the system to adaptively adjust its safety strategy based on real-time motion status through dynamic safety boundaries and smooth transition strategies, and achieves seamless connection during path replanning. This ensures both safety and attitude adjustment efficiency, and completely avoids the mechanical shock that may be caused by replanning.

[0032] This invention uses fuzzy control to dynamically adjust the weight coefficients, enabling the system to respond more smoothly and logically to load changes, avoiding abrupt changes in control commands, and further ensuring the stability of the posture adjustment platform and the safety of the equipment under heavy or variable load conditions.

[0033] Other advantages, objectives, and features of the embodiments of the present invention will be apparent in part from the following description, and in part will be understood by those skilled in the art through study and practice of the embodiments of the present invention. Detailed Implementation

[0034] To further illustrate the technical means and effects of this invention, the following embodiments are provided for further explanation. The specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0035] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0036] According to one embodiment of the present invention, a method for tolerance allocation and adaptive attitude adjustment of large aircraft components includes the following steps: S1) During the attitude adjustment process, the three-dimensional coordinate pose data of the target point on the part to be docked is obtained simultaneously using a laser tracker, and the image data of the key docking features between the part to be docked and the reference part is obtained using a vision sensor; the three-dimensional coordinate pose data and the image data are spatiotemporally registered and fused to generate a fused pose model containing the absolute pose of the part and the details of the relative docking features. S2) Access the preset tolerance database, which defines the allowable deviation range of all key docking interfaces in six degrees of freedom; using the fused pose model as input and the tolerance database as constraints, calculate an optimal pose adjustment target pose through a multi-objective optimization algorithm. The target pose must ensure that the expected deviation of all docking interfaces falls within their corresponding allowable deviation range, and that the weighted norm of the pose adjustment displacement vectors required by all support points is minimized. The optimization objectives include: the first objective is to maximize the safety margin that satisfies the tolerance constraints of all docking interfaces, and the second objective is to minimize the weighted norm of the pose adjustment displacement vectors of all support points, where the weight coefficients are set based on the motion performance of each pose adjustment drive unit. S3) Based on the difference between the optimal pose adjustment target pose and the current fused pose model, automatically generate a collision-free and smooth pose adjustment path; control the multi-degree-of-freedom pose adjustment platform to drive the docking component to move according to the pose adjustment path, and repeat steps S1) and S2) in real time during the movement to perform closed-loop correction of the pose adjustment target pose and the pose adjustment path until the docking component reaches the optimal pose adjustment target pose. In the closed-loop correction process of step S3, the weight coefficient of the attitude adjustment displacement vector in the second target is dynamically adjusted according to the real-time measured load of the attitude adjustment platform. When the load is close to the threshold, the weight coefficient is increased to prioritize motion stability and equipment safety.

[0037] An embodiment of a method for tolerance allocation and adaptive attitude adjustment for large aircraft component docking can be implemented as follows: During the docking and assembly process of the aircraft wing and fuselage, a laser tracker and several vision sensors are first deployed at the attitude adjustment station. At the start of attitude adjustment, the laser tracker and vision sensors are simultaneously activated. The laser tracker continuously measures the three-dimensional coordinates of several target points mounted on the wing. Simultaneously, the vision sensors simultaneously capture image sequences of key docking features between the docking holes on the wing and the corresponding hole systems on the fuselage.

[0038] The acquired 3D coordinate pose data and image data are spatiotemporally registered and fused. Specifically, a pre-calibrated transformation matrix is ​​used to unify the coordinates of feature points extracted from the visual sensor images to the measurement coordinate system of the laser tracker. The Kalman filter algorithm is used to smooth the laser tracker data, and the hole center coordinates extracted with high precision from the image are used to supplement and verify the smoothed pose details. Finally, a fused pose model is generated that simultaneously includes the absolute spatial pose of the wing and its relative docking feature relationship with the fuselage.

[0039] Subsequently, the system accesses a pre-set tolerance database. This database explicitly defines the upper and lower limits of permissible deviations in six degrees of freedom for all critical interfaces connecting the wing and fuselage. Using the fused pose model at the current moment as input and all clauses of the tolerance database as hard constraints, a multi-objective optimization algorithm is initiated for calculation. The primary objective of this algorithm is to find a target pose that maximizes the overall safety margin while meeting the tolerance requirements for the expected deviations of all interfaces. Secondly, based on meeting the primary objective, the algorithm also needs to optimize the calculation to minimize the weighted norm of the attitude adjustment displacement vectors required for the multiple support points driving the wing motion. The weights are set based on the historical motion performance and current state of each electric cylinder drive unit.

[0040] Based on the difference between the calculated optimal target pose and the current fused pose model, the system automatically plans a collision-free and smooth attitude adjustment path. This path is generated using fifth-order polynomial interpolation to ensure continuous velocity and acceleration during motion. The multi-degree-of-freedom attitude adjustment platform begins driving the wing along this path. During motion, the system repeats the entire process of measurement, fusion, and optimization calculation at a fixed frequency in real time. Once the optimal target pose calculated based on the latest data changes significantly, the system immediately replans the remaining unexecuted attitude adjustment paths online based on the latest target pose and the current actual wing pose. Furthermore, during closed-loop correction, the system monitors the load of each attitude adjustment drive unit in real time. When the load of a drive unit approaches its safety threshold, the system automatically increases its weight coefficient in the optimization objective function, ensuring that the replanned path prioritizes the motion stability and safety of that unit, even if this may mean a small increase in attitude adjustment displacement. Through this continuous cycle of perception, decision-making, execution, and feedback, the wing finally reaches a docking pose that meets all tolerance requirements and has optimal motion performance.

[0041] Existing technologies also use laser trackers to measure target points on components and perform attitude adjustment based on this absolute pose. However, relying solely on laser trackers as a measurement method lacks direct visual perception of key docking features. The calculation of the target pose relies primarily on theoretical models and preset tolerances, resulting in a relatively singular optimization objective. It typically focuses only on final pose accuracy, neglecting the motion displacement of each support point on the attitude adjustment platform as a significant optimization goal. The entire attitude adjustment process is essentially open-loop; once the adjustment path begins, it is no longer corrected based on real-time measurement data. The path replanning function is either missing or very weak, unable to handle unexpected deviations during movement. Furthermore, the system lacks the ability to dynamically adjust optimization strategies based on the real-time load of the attitude adjustment platform.

[0042] Compared with existing technologies, this invention offers a series of positive and beneficial effects. By fusing multi-source sensor data, this method constructs a more comprehensive and accurate component pose model, providing a more reliable information foundation for subsequent decision-making. Utilizing a multi-objective optimization algorithm to comprehensively consider docking tolerance compliance and attitude adjustment efficiency, the final attitude adjustment scheme is more economical and reliable while ensuring quality. The introduction of a real-time closed-loop correction mechanism significantly enhances the system's ability to resist interference and uncertainties, ensuring the final accuracy of the attitude adjustment process. The strategy of dynamically adjusting the optimization target weights based on the load further enhances the system's adaptability and the safety of equipment operation. Overall, this method makes the docking process of large aircraft components more intelligent, robust, and efficient.

[0043] According to one embodiment of the present invention, preferably, the spatiotemporal registration and fusion in step S1) specifically comprises: Using a fixed calibration plate or specific feature points on a component as a common reference, a transformation relationship between the laser tracker coordinate system and the vision sensor coordinate system is established. Through timestamp synchronization, the three-dimensional coordinate pose data and image data acquired at the same time are aligned, and the three-dimensional coordinate pose data are smoothed using Kalman filtering or extended Kalman filtering algorithms. At the same time, the feature point coordinates extracted from the image data are used to supplement and correct the smoothed pose data to generate the fused pose model.

[0044] An embodiment for achieving spatiotemporal registration and fusion is as follows: During the initialization phase of the measurement system, a high-precision calibration plate is fixedly installed in the measurement field. This calibration plate simultaneously displays a reflective target sphere that can be measured by a laser tracker and a special pattern that can be clearly identified by a visual sensor. By simultaneously using the laser tracker to measure the three-dimensional coordinates of the target sphere and the visual sensor to capture images of the pattern, the precise transformation relationship between the laser tracker coordinate system and the visual sensor coordinate system is calculated. In subsequent attitude adjustment, within each data acquisition cycle, the system assigns a unified timestamp to the target point coordinate data acquired by the laser tracker and the image data acquired by the visual sensor. Based on this timestamp, the three-dimensional data and two-dimensional images at the same moment are paired. The sequential coordinate data acquired by the laser tracker is smoothed using a Kalman filter algorithm to suppress random measurement noise. Simultaneously, the pixel coordinates of key docking features are extracted from the synchronized visual images and mapped to the three-dimensional measurement space through coordinate system transformation relationships. These visual feature point data are used to supplement the smoothed laser tracker pose data, especially to provide a correction reference when the laser data may be transiently missing or drifted due to occlusion, ultimately generating a stable and reliable fused pose model.

[0045] Existing technologies may use laser trackers and vision sensors independently. While they may establish initial coordinate system transformations, they lack a rigorous timestamp synchronization mechanism, resulting in a time lag between 3D coordinates and 2D image data. In terms of data processing, they may only perform simple moving average filtering on the laser data, or directly use the raw data without incorporating visually extracted feature points as effective correction information into the pose calculation process. When the laser data experiences a brief anomaly, the system may only be able to wait for it to recover or report an error directly, unable to utilize parallel vision information for compensation.

[0046] This invention establishes a rigorous spatiotemporal unified framework for multi-sensor data. Through filtering and complementary data processing, it effectively improves the accuracy and reliability of the fused pose model, builds a solid data foundation for subsequent pose adjustment decisions, and enhances the robustness of the system in complex measurement environments.

[0047] According to one embodiment of the present invention, preferably, the tolerance database in step S2) is stored in matrix form, with each row corresponding to a key docking interface. The content includes, but is not limited to, the interface number, the theoretical position in the component coordinate system, and the allowable positive deviation value and allowable negative deviation value in each degree of freedom direction.

[0048] An embodiment of the construction and use of a tolerance database can be described as follows: During the system deployment phase, a tolerance database stored in matrix form is created based on the design drawings and process specifications of major aircraft components. Each row of this matrix uniquely corresponds to a key docking interface, such as a hole group or mating surface with a specific number. Each row sequentially records the identifier of the interface, its theoretical three-dimensional position coordinates in the component coordinate system, and the allowable positive and negative deviation values ​​along the three translational and three rotational degrees of freedom. These deviation values ​​are explicitly stored in numerical form. During attitude adjustment optimization calculations, the optimization algorithm directly reads this tolerance database matrix. For each docking interface, the algorithm compares its predicted deviation with the corresponding allowable positive and negative values ​​in the database and incorporates them as inequality constraints into the optimization model. This structured storage method enables the program to automatically traverse and check the tolerance compliance of all interfaces.

[0049] Existing technologies may manage tolerance information in the form of unstructured documents or scattered parameters. For example, tolerance requirements for different interfaces may be recorded in paper charts or various electronic configuration files. During orientation calculations, operators or programs need to consult and interpret these documents one by one and then manually convert them into constraints usable by the algorithm. This approach is not only inefficient, but also prone to omissions or misinterpretations when there are many interfaces, leading to incomplete or inaccurate tolerance constraint settings, which in turn affects the correctness of the orientation results.

[0050] This invention uses a standardized matrix structure to centrally manage all interface tolerance data, making information retrieval efficient and error-free, ensuring the integrity and accuracy of optimization calculation constraints, and providing indispensable data support for achieving precise tolerance allocation.

[0051] According to one embodiment of the present invention, as a preferred embodiment, the generation of a collision-free and smooth attitude adjustment path in step S3) specifically involves: using B-spline curves or fifth-order polynomial interpolation to plan a continuous motion trajectory with continuous first and second derivatives between the initial attitude adjustment pose and the optimal attitude adjustment target pose; and during the planning process, dynamic virtual collision interference checks are performed based on the three-dimensional model of the component to ensure that the path is collision-free.

[0052] An embodiment for generating a collision-free smooth attitude adjustment path can be implemented as follows: After calculating the optimal attitude adjustment target pose, the path planning module is activated. This module uses the current initial pose of the part to be docked and the optimal attitude adjustment target pose as the starting and ending points of the path. A spatial motion trajectory is constructed between these two points using a fifth-order polynomial interpolation method. This fifth-order polynomial ensures that the position, velocity, and acceleration of the generated trajectory are continuous. During the trajectory construction process, the system simultaneously initiates a dynamic virtual collision interference check process. This process, based on a three-dimensional digital model of the part to be docked and the surrounding environment, performs dense sampling along the planned trajectory, calculating the minimum distance between the part and the static environment and other dynamic objects at each sampling point. Once any potential collision risk point with a distance less than the safety threshold is detected, the path planning module adjusts the parameters of the fifth-order polynomial and regenerates the trajectory until the entire path is confirmed to meet the collision-free requirements at all locations.

[0053] Existing technologies may employ simple linear interpolation or low-order polynomials to plan paths. Linear interpolation can lead to abrupt velocity changes at the start and end points, while low-order polynomials may fail to guarantee the continuity of acceleration. Both of these situations can cause vibration and shock to the attitude adjustment platform. Furthermore, collision checks may be performed as a one-time static check after path planning is complete, or based solely on a simplified bounding box for a rough assessment, making it difficult to detect potential interference between fine structures, or failing to effectively integrate obstacle avoidance functionality during the planning phase.

[0054] This invention utilizes high-order continuous curves to plan trajectories and ensures smoothness from a kinematic perspective. At the same time, it integrates detailed dynamic collision detection into the planning process, thereby ensuring the high stability and absolute safety of the posture adjustment movement and effectively avoiding equipment damage and unexpected interruptions during the posture adjustment process.

[0055] According to one embodiment of the present invention, as a preferred embodiment, the closed-loop correction in step S3) specifically involves: during the execution of the pose adjustment path, re-executing steps S1) and S2) at a fixed frequency; if the deviation between the newly calculated optimal pose adjustment target pose and the endpoint pose of the current execution path exceeds a preset threshold, then based on the latest fused pose model and the optimal pose adjustment target pose, the pose adjustment path of the unexecuted part is replanned online.

[0056] An embodiment of closed-loop correction can be described as follows: During the movement of the wing component along a predetermined path driven by the attitude adjustment platform, the system re-executes the data acquisition, fusion, and optimization calculation process at a fixed frequency of ten times per second. This means that every 0.1 seconds, the system generates a new fused pose model based on the latest measurement data and recalculates the optimal attitude adjustment target pose based on this model. The system continuously compares the newly calculated target pose with the endpoint pose of the currently executed attitude adjustment path. If the deviation between the two in position or attitude angle exceeds a preset tolerance threshold, such as a position deviation greater than 0.1 mm or an attitude angle deviation greater than 0.01 degrees, it is determined that the endpoint of the current path is no longer optimal. At this time, the system does not immediately stop the movement, but instead uses the current actual pose of the component as a new starting point and the newly calculated optimal attitude adjustment target pose as a new endpoint to replan the remaining unexecuted path segments online, generating an updated and more optimized motion trajectory, and instructing the attitude adjustment platform to smoothly transition to execute this new path.

[0057] Existing technologies may not recalculate the target pose midway after the orientation path begins execution, or they may only monitor at a low frequency and lack online replanning capabilities. They typically wait until the component reaches the end of the preset path before performing final pose measurement and verification. If the tolerance requirements are not met at this point, the entire process must start from scratch, requiring re-measurement, calculation, and planning, resulting in long orientation cycles and low efficiency. This static method cannot correct accumulated errors generated during motion.

[0058] This invention endows the attitude adjustment system with a high degree of dynamic adaptability. Through high-frequency real-time feedback and online path optimization, it can actively correct various deviations during the motion process, ensuring that the attitude adjustment process always evolves towards the most ideal target, thereby significantly improving the success rate and accuracy of the final docking.

[0059] According to one embodiment of the present invention, preferably, in step S1), when there is a non-negligible conflict between the laser tracker data and the visual sensor data, the following data reliability arbitration strategy is executed: A first confidence weight is set for the laser tracker data, and a second confidence weight is set for the visual sensor data; the first confidence weight is positively correlated with the geometric configuration accuracy factor of the target point relative to the laser tracker, and the second confidence weight is positively correlated with the image sharpness and illumination uniformity of the key docking features; The conflicting data are weighted and fused based on the first and second confidence weights to generate the fused pose model.

[0060] One embodiment of data confidence arbitration operates as follows: During data fusion, when the system detects a significant inconsistency between the location of a feature point calculated from laser tracker data and the location of the same feature point extracted from a visual sensor image, an arbitration strategy is initiated. The system continuously evaluates the confidence weight of the laser tracker data, which is related to the spatial geometry of all target points relative to the laser tracker at the current moment. If the target points are well-distributed and the configuration accuracy factor is high, the laser data is assigned a higher confidence score. Simultaneously, the system evaluates the confidence weight of the visual data, which depends on the sharpness of key docking features and the uniformity of illumination in the currently acquired image. If the image has low noise, sharp feature edges, and uniform illumination, the visual data is assigned a higher confidence score. Subsequently, based on these two dynamically calculated weights, the system performs a weighted average fusion of the conflicting laser and visual data, rather than simply discarding either, thereby generating the most reliable fused pose model under the current measurement conditions.

[0061] When faced with conflicting multi-source data, existing technologies may employ simple rule-based processing, such as always prioritizing laser tracker data while ignoring visual data, or vice versa. They may also use fixed-weight fusion methods, such as averaging. These methods fail to consider the impact of instantaneous changes in the measurement environment on the quality of data from specific sensors. When the quality of data from a particular sensor deteriorates due to temporary interference, it may still be treated the same as high-quality data or incorrectly discarded, leading to fusion results that deviate from the true value.

[0062] This invention introduces an intelligent data quality assessment and arbitration mechanism, enabling the data fusion process to adapt to complex and changing working conditions, prioritizing the adoption of more reliable data sources, thereby maintaining the accuracy and reliability of the fused pose model under various adverse conditions, and enhancing the system's decision-making intelligence and overall robustness.

[0063] According to one embodiment of the present invention, preferably, after establishing the coordinate system transformation relationship using specific feature points on a fixed calibration plate or component as a common reference, an online self-calibration step is further included: During the orientation adjustment process, the multi-degree-of-freedom orientation adjustment platform is periodically controlled to cause the component to be docked to produce a small, known, preset motion; By comparing the actual motion calculated from laser tracker data with the actual motion calculated from vision sensor data, the current transformation relationship residual between the laser tracker coordinate system and the vision sensor coordinate system is calculated. When the residual exceeds the allowable range, the original transformation relationship is updated using the latest calculated transformation relationship to achieve online self-calibration of the coordinate system.

[0064] One embodiment of online self-calibration can be performed as follows: After the attitude adjustment system completes initial coordinate system calibration and begins operation, at regular intervals, for example, after completing 5% of the attitude adjustment stroke, the system automatically inserts a small calibration motion. This motion command drives the multi-DOF attitude adjustment platform to perform a known, minute translation and rotation, for example, translating one millimeter along the X-axis and rotating 0.5 degrees around the Z-axis. The system then accurately records the actual motion of the component calculated from the laser tracker data, and also records the actual motion of the component calculated from the vision sensor data through image processing. Theoretically, these two motion quantities should be consistent through the initial coordinate system transformation relationship. By comparing these two sets of actual motion quantities, the system can calculate the residual error of the transformation relationship between the current laser tracker coordinate system and the vision sensor coordinate system. If this residual exceeds the system's allowable accuracy range, the system automatically updates and corrects the original calibration matrix using the latest calculated transformation parameters, thereby achieving online self-calibration of the sensor coordinate system without stopping the system or introducing external calibration tools.

[0065] Existing technologies typically perform offline calibration only once during system installation and commissioning. During long-term use, factors such as vibration, temperature changes, or mechanical stress relaxation can cause the relative position or internal parameters of the sensor to drift slowly, gradually rendering the initial calibration parameters ineffective. However, the system itself cannot detect this drift and continues to use outdated calibration parameters for data fusion, introducing imperceptible systematic errors. These errors accumulate over time, affecting long-term measurement and attitude adjustment accuracy.

[0066] This invention enables the system to perform self-monitoring and calibration. It can actively detect and correct minute changes between sensor coordinate systems, effectively suppressing long-term systematic errors caused by calibration parameter drift, ensuring the measurement system's sustained accuracy and stability throughout its entire lifecycle, and reducing reliance on external manual maintenance.

[0067] According to one embodiment of the present invention, preferably, the multi-objective optimization algorithm employs a non-dominated sorting genetic algorithm with elite retention, and an initial population based on historical successful pose adjustment cases is introduced during algorithm initialization: From the historical successful attitude adjustment cases, extract the attitude adjustment target pose corresponding to cases that are similar to the current part model and initial pose to be docked, and inject them as high-quality genes into the initial population to accelerate optimization convergence and avoid getting trapped in local optima.

[0068] An embodiment of the optimization algorithm is described below: When starting to calculate the optimal orientation target pose, the system initiates a non-dominated sorting genetic algorithm with elite retention. During the population initialization phase, the system does not generate initial solutions completely randomly, but first accesses a database storing historical successful orientation cases. The system searches and matches in the database based on the model code of the component to be docked and the deviation characteristics of its initial pose. It identifies several historical successful cases most similar to the current working condition and extracts the final orientation target pose parameters used at that time from these cases. These pose parameters, proven successful in practice, are directly injected into the initial population of the genetic algorithm as high-quality individuals. Subsequently, the algorithm begins its regular iterative process, including selection, crossover, mutation, and selection based on non-dominated sorting. These high-quality genes provide a good starting point for the population, effectively guiding the search direction.

[0069] Existing techniques typically use completely random initial population generation when employing similar optimization algorithms. While this method can theoretically cover the entire solution space, the quality of individuals in the population may generally be poor, falling far short of the optimal solution. This results in the algorithm requiring a considerable number of iterations to converge to a promising region, leading to long computational times. Furthermore, in complex solution spaces, this random initialization method makes it easier for the search process to prematurely get trapped in a local optimum, making it difficult to escape and find a globally better solution.

[0070] This invention utilizes historical successful experience to intelligently guide the optimization algorithm, significantly improving the quality of the starting point of the search. This not only accelerates the convergence speed of the algorithm and shortens the planning time, but also increases the possibility of finding a high-quality, globally optimal pose adjustment solution, making the pose adjustment process more efficient and reliable.

[0071] According to one aspect of the present invention, based on the aforementioned aspect, as a preferred embodiment, when performing dynamic virtual collision interference checks based on the three-dimensional model of the component, a dynamic safety boundary is further extended for the component to be docked and the reference component. The size of the dynamic safety boundary is positively correlated with the velocity value of the component to be docked at the corresponding position calculated based on the attitude adjustment path. When replanning the posture adjustment path online, the following smooth transition strategy is adopted to avoid motion shock: Using the actual pose and velocity of the component to be docked at the current moment as the new starting state, and the newly calculated optimal pose adjustment target pose as the endpoint, a B-spline curve or fifth-order polynomial trajectory is regenerated, and it is ensured that the newly generated trajectory is continuous with the currently executed trajectory in pose and velocity at the starting point.

[0072] One embodiment for enhancing path safety and smoothness operates as follows: During dynamic virtual collision interference checks, the system does not directly use the original 3D models of the parts to be docked and the reference parts. Instead, it wraps their outer surfaces with a dynamic safety boundary. The size of this safety boundary is not fixed but dynamically adjusted based on the velocity value of the part to be docked at its current position on the planned path. When the part moves to a high-speed segment of the path, the safety boundary automatically expands to a larger size to allow for a longer braking distance and reaction time. When the part moves at a low speed or is close to rest, the safety boundary maintains a smaller basic size. During online path replanning, the system first obtains the actual pose and actual velocity vector of the part to be docked at the current moment. It uses this actual state as the starting point of the new path and the newly calculated optimal pose of the target pose as the ending point to regenerate a fifth-order polynomial trajectory. During trajectory generation, it strictly ensures that the position and velocity of the new trajectory at the starting point are completely continuous with the current actual state of the part, thereby achieving a smooth and seamless transition from the old trajectory to the new trajectory.

[0073] Existing technologies may use fixed safety boundaries. These boundaries can be overly conservative at low speeds, impacting efficiency, while at high speeds they may be insufficient to mitigate risks. Path replanning may only consider positional continuity, neglecting velocity continuity. When a new path is switched and executed, a velocity jump occurs at the transition point, causing significant jerking or impact on the attitude adjustment platform, adversely affecting the mechanical structure and drive system.

[0074] This invention introduces a dynamic safety strategy that adapts to the motion state, achieving a better balance between safety and efficiency. Simultaneously, through a smooth path transition technology ensuring continuous state, it completely eliminates motion shocks during replanning, guaranteeing the long-term stability and lifespan of the equipment.

[0075] According to one preferred embodiment of the present invention, the aircraft large component docking tolerance allocation and adaptive attitude adjustment method dynamically adjusts the weight coefficients based on the real-time measured load of the attitude adjustment platform, specifically through a fuzzy controller: The inputs to the fuzzy controller are the load rate and the load rate change rate, and the output is the adjustment amount of the weight coefficient; When the load rate is high and the load rate change rate is positive, a large positive adjustment amount is output to quickly increase the weight coefficient and reduce the optimization priority of the attitude adjustment displacement.

[0076] The system incorporates a fuzzy controller specifically for managing the adjustment of weighting coefficients. This controller has two input variables: the real-time load rate of the attitude adjustment platform (the ratio of the current load to the rated load) and the rate of change of the load rate, representing the trend and speed of load change. The output variable is the adjustment amount of the weighting coefficients. The fuzzy controller internally uses fuzzy rules based on expert experience. For example, one rule states that if the load rate is high and the rate of change of the load rate is positive, then a large positive adjustment amount is output. During attitude adjustment, the system monitors the load data of each drive unit in real time, calculates the load rate and its rate of change, and transmits these as inputs to the fuzzy controller. The controller derives a specific adjustment amount based on fuzzy inference. This adjustment amount is used to increase the current value of the weighting coefficients, thereby giving greater weight to the attitude adjustment displacement vector in the optimization objective. This prompts the system to plan paths that are more inclined to reduce the motion amplitude of high-load drive units, prioritizing their stable operation and safety.

[0077] Existing technologies may employ simple threshold methods for control. For example, the weighting coefficient might be suddenly set to a large fixed value only when the load exceeds a certain fixed threshold. This control method is very abrupt and easily causes frequent jumps in the weighting coefficient near the threshold, leading to drastic changes in the optimization objective and unstable oscillations in the attitude adjustment path, thus failing to achieve a smooth and compliant control effect.

[0078] This invention utilizes the ability of fuzzy control to handle uncertainty, enabling a more refined and intelligent response to load conditions. It can smoothly and rationally adjust optimization strategies based on the severity and changing trends of the load, effectively ensuring equipment safety while maintaining the continuity and stability of the attitude adjustment process.

[0079] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the embodiments of the present invention. Other modifications can be readily implemented by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the embodiments of the present invention are not limited to the specific details.

Claims

1. An aircraft major component docking tolerance allocation and adaptive alignment method, characterized in that, The method comprises the following steps: S1) acquiring three-dimensional coordinate pose data of target points on the component to be docked, and acquiring image data of key docking features between the component to be docked and the reference component, performing space-time registration and fusion of the three-dimensional coordinate pose data and the image data, and generating a fusion pose model containing the absolute pose of the component and the details of the relative docking features; S2) accessing a pre-set tolerance database, the tolerance database defining the allowable deviation range of all key docking interfaces in six degrees of freedom; taking the fusion pose model as input and the tolerance database as constraint condition, an optimal pose adjustment target pose is calculated through a multi-objective optimization algorithm, the optimization process including two objectives: the first objective is to maximize the safety margin that meets all docking interface tolerance constraints, and the second objective is to minimize the weighted norm of all support point pose adjustment displacement vectors, wherein the weight coefficients are set based on the motion efficiency of each pose adjustment driving unit; S3) generating a pose adjustment path according to the difference between the optimal pose adjustment target pose and the current fusion pose model; driving the component to be docked to move according to the pose adjustment path, and repeatedly performing steps S1) and S2) in real time during the movement to close-loop correct the pose adjustment target pose and the pose adjustment path until the component to be docked reaches the optimal pose adjustment target pose; In the closed-loop correction process of step S3), the weight coefficients of the pose adjustment displacement vectors in the second objective are dynamically adjusted according to the real-time measured pose adjustment platform load, and the weight coefficients are increased when the load approaches the threshold.

2. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, The space-time registration and fusion in step S1) is specifically: A fixed calibration plate or specific feature points on a component are used as a common reference to establish the conversion relationship between the laser tracker coordinate system and the vision sensor coordinate system; through timestamp synchronization, the three-dimensional coordinate pose data and the image data collected at the same time are aligned, and Kalman filtering or extended Kalman filtering algorithm is used to smooth the three-dimensional coordinate pose data, and the feature point coordinates extracted from the image data are used to supplement and correct the smoothed pose data to generate the fusion pose model.

3. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, The tolerance database in step S2) is stored in matrix form, each row corresponding to a key docking interface, and its content includes but is not limited to interface number, theoretical position in component coordinate system, and allowable positive deviation value and allowable negative deviation value in each degree of freedom direction.

4. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, The method for generating a collision-free and smooth pose adjustment path in step S3) is: a B-spline curve or a quintic polynomial interpolation method is used to plan a continuous and first and second derivative continuous motion trajectory between the pose adjustment starting pose and the optimal pose adjustment target pose.

5. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 4, wherein, The closed-loop correction in step S3) is specifically: steps S1) and S2) are re-executed at a fixed frequency during the execution of the pose adjustment path; if the deviation between the newly calculated optimal pose adjustment target pose and the end pose of the currently executed path exceeds a preset threshold, the unexecuted part of the pose adjustment path is re-planned online based on the latest fusion pose model and the optimal pose adjustment target pose.

6. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, In step S1), when there is a non-negligible conflict between the laser tracker data and the vision sensor data, the following data credibility arbitration strategy is performed: a first confidence weight is set for the laser tracker data, and a second confidence weight is set for the vision sensor data; the first confidence weight is positively correlated with a geometric configuration accuracy factor of a target point relative to the laser tracker, and the second confidence weight is positively correlated with image clarity and uniformity of illumination of the key docking feature; the conflicting data is weighted and fused based on the first and second confidence weights to generate the fused pose model.

7. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 2, wherein, After establishing the coordinate system conversion relationship using a fixed calibration plate or specific feature points on a component as a common reference, an online self-calibration step is further included: During the pose adjustment process, a multi-degree-of-freedom pose adjustment platform is periodically controlled to cause a preset motion of the component to be docked; By comparing the actual motion calculated from the laser tracker data with the actual motion calculated from the vision sensor data, a current conversion relationship residual error between the laser tracker coordinate system and the vision sensor coordinate system is solved; When the current conversion relationship residual error exceeds an allowable range, the original conversion relationship is updated using the latest solved conversion relationship to realize online self-calibration of the coordinate system.

8. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, In step S2), the optimization calculation uses a multi-objective optimization algorithm, which uses a non-dominated sorting genetic algorithm with elitism, and introduces an initial population based on historical successful pose adjustment cases during algorithm initialization: From the historical successful pose adjustment cases, the pose adjustment target pose corresponding to the case similar to the current component model and initial pose is extracted and injected into the initial population as a high-quality gene to accelerate optimization convergence and avoid falling into local optimum.

9. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 5, wherein, When performing dynamic virtual collision interference checking based on the three-dimensional model of the component, a dynamic safety boundary is further expanded for the component to be docked and the reference component; The size of the dynamic safety boundary is positively correlated with the speed value of the component to be docked at the corresponding position calculated according to the pose adjustment path; When the pose adjustment path is re-planned online, the following smooth transition strategy is used to avoid motion impact: A B-spline curve or quintic polynomial trajectory is regenerated using the actual pose and speed of the component to be docked at the current time as the new starting state and the newly solved optimal pose adjustment target pose as the end point, and it is ensured that the newly generated trajectory is continuous in pose and speed with the trajectory being currently executed at the starting point.

10. The aircraft major component docking tolerance assignment and adaptive alignment method of Claim 1, wherein, The weight coefficient is dynamically adjusted according to the real-time measured pose adjustment platform load, which is specifically realized by a fuzzy controller: The input of the fuzzy controller is the load rate and the load rate change rate, and the output is the adjustment amount of the weight coefficient; When the load rate is high and the load rate change rate is positive, a positive adjustment amount is output to increase the weight coefficient.