Aircraft wing-body docking method

Abstract

The present invention provides a method for aircraft wing-body docking, which uses an improved particle swarm optimization algorithm to optimize and analyze the attitude adjustment process of aircraft wing-body docking. By defining the position and tolerance of each assembly point in detail, the relative position of the key points of the wing and body in three-dimensional space is evaluated and optimized to achieve precise docking. This process needs to comprehensively consider the various deformations and displacements that the components may encounter in the actual environment, and calculate the deviation between the current assembly attitude monitoring point and the ideal state to evaluate the quality of the assembly. Based on the information provided by the supplier and deformation analysis, the displacement of important monitoring points is measured and evaluated in detail. The RRT* algorithm is used to plan the path of the progressive stage in the aircraft wing-body docking process, that is, the path planning from the initial point to the target point. The core of the algorithm is to efficiently generate and expand a search tree, which starts from the initial state and gradually expands to the target state. This incremental sampling algorithm can quickly find the initial path.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing, and in particular to a method for connecting an aircraft wing to a body. Background Art

[0002] The aircraft wing-to-body docking process directly impacts the structural integrity and production efficiency of the aircraft. Traditional wing-to-body docking methods rely on a staged assembly process, consisting of an approach phase and an alignment phase. During the approach phase, the wing is gradually moved closer to the fuselage. The main task of this phase is to ensure the wing's rough spatial positioning, but the movement process from the initial position to the final point still needs to be precisely controlled, as any initial position error will affect the assembly efficiency of the subsequent alignment phase. After the approach phase is completed, the wing will be in close proximity to the fuselage, but not yet finalized.

[0003] Next comes the alignment phase, where the wing is precisely inserted or docked into the designated interface on the fuselage through fine-tuning. This involves fine-tuning the wing's angle, position, and height until all pre-set docking points meet the designed tolerances. The alignment phase is the most critical part of the entire assembly process, as it directly determines the final quality of the docking. Traditionally, the alignment phase relies on high-precision measuring equipment and experienced operators to manually adjust the docking points.

[0004] The above process analysis shows that traditional wing-body docking methods are not only time-consuming but can also result in substandard assembly due to insufficient precision. With the rapid development of aviation technology, the requirements for assembly speed and precision are also increasing. This has created an urgent need for a more efficient and precise docking technology to meet the demands of modern aviation manufacturing.

[0005] To address this issue, an embodiment of the present invention proposes a reverse path planning method based on the fusion of an improved particle swarm optimization (PSO) and a fast randomized tree star algorithm (RRT*). By utilizing historical assembly data, this algorithm can reversely deduce the optimal path from the initial state to the final state of successful attitude adjustment. This reverse planning method can not only significantly improve assembly efficiency, but also enhance the repeatability and accuracy of the process. By implementing this technology, it can provide a new solution for the aviation manufacturing industry to meet the increasingly complex assembly challenges. Summary of the Invention

[0006] To address the above-mentioned problems in the prior art, this application proposes an aircraft wing-body docking method. By defining the position and tolerance of each assembly point in detail, evaluating the relative positions of key wing-body points in three-dimensional space, and comprehensively considering the various deformations and displacements of aircraft wing-body components in actual environments, this method achieves intelligent reverse planning of the aircraft wing-body docking path. The specific steps include:

[0007] First, the wing-body docking is optimized based on the improved particle swarm algorithm. The positions and velocities of all particles in the aircraft wing-body are randomly initialized, the velocity of each particle is updated, and the position of each particle is updated based on the updated velocity. It is determined whether the new position exceeds the maximum allowable deviation. If the maximum allowable deviation is exceeded, the position based on the deviation is updated and the fitness value of each particle is calculated. Then, it is determined whether the termination condition is met. If the maximum allowable deviation is not exceeded, it is directly determined whether the termination condition is met. If the termination condition is met, the optimization result is output. If the termination condition is not met, the algorithm returns to updating the velocity of each particle.

[0008] Then, the aircraft wing-body docking process is optimized. A set of random offset positions of the aircraft wing is defined as the initial state using random configuration and target configuration to simulate the uncertain starting conditions in actual situations. Random points are generated in space, and the node closest to the newly generated point in the initial state is found as the nearest node. The new node is created by turning from the nearest node to the random point. The cost of reaching the new node from the starting point to the nearest node is calculated. The new node is added to the tree, and the parent node of the new node is set to the nearest node. If the new node is within the threshold distance from the target, a path from the start to the target is constructed; if the new node is not within the threshold distance from the target, the path is returned to turning from the nearest node to the random point to create the new node.

[0009] Preferably, the maximum allowable deviation range takes into account the displacement of key points. Assume that for any key point i, its ideal position coordinate is (X i ,Y i ,Z i ), define D i is the design position of point i, V i is the standard deviation data provided by the supplier, Δ i is the actual measured deformation of point i, T i is the total tolerance of point i after considering the supplier data and deformation data; tolerance T i It can be expressed by the following formula:

[0010] T i =V i +α△ i (1)

[0011] Among them, α is a scaling factor used to adjust the weight of the deformed data in the total tolerance.

[0012] Preferably, for each key point, there is a single axial distance difference between the actual position and the ideal position, a symmetry tolerance, and a distance difference between specific paired points, and the tolerance T is set to i The calculation is broken down into five types of deviations:

[0013]

[0014] Where ΔX i,j is the cross-axis deviation, ΔY i,j is the uniaxial deviation, and is the desired distance between points i and j along the X and Y axes, ΔZ i is the absolute position deviation, z expected,i is its expected position, is the symmetry deviation, X L , Z L and X R , Z R are the coordinates of the symmetrical points on the left and right components, Δcomposite is the composite deviation, and δ is used to consider additional tolerance or correction.

[0015] Preferably, the posture evaluation index is designed as a complex fitness function "evaluate_assembly", which serves as the loss function for the subsequent improved particle swarm algorithm to optimize the posture adjustment process; the particle swarm algorithm is used to initialize a series of candidate solutions, where each particle represents a possible set of spatial coordinates of the wing, and the position and velocity of the particle are gradually iteratively updated through simulation to seek to minimize the fitness function; in this process, the motion trajectory of each particle is based on its current position, combined with the information of the individual historical best position and the group historical best position, so as to effectively guide the particle to move to an area with better fitness.

[0016] Preferably, the steps in optimizing the aircraft wing-body docking process include:

[0017] An initial state is defined, consisting of a set of randomly offset positions of the aircraft wing to simulate the uncertain starting conditions in real-world situations. In each iteration, the algorithm first randomly generates a point. To speed up the algorithm's convergence, a point is generated around the target every ten iterations, and a wider area is generated the rest of the time.

[0018] Find the node closest to the newly generated point in the current tree as the nearest node; use the "steer" function to generate a new node from the nearest node toward the new point. This function ensures that the new node moves along a certain step length in the direction from the nearest node to the new point, thereby gradually approaching the target;

[0019] For each new node, the total cost from the starting point to that point is calculated and the new node is connected to the tree; if the new node is within a preset threshold, the goal is considered to have been reached and the search ends.

[0020] The above technical features can be combined in various suitable ways or replaced by equivalent technical features, as long as the purpose of the present invention can be achieved.

[0021] The intelligent reverse planning method for aircraft wing-body docking path provided by the present invention has at least the following beneficial effects compared with the prior art:

[0022] This algorithm not only significantly improves the efficiency and accuracy of the docking process, but also ensures the feasibility and repeatability of the solution. The resulting 3D path visualization clearly demonstrates the algorithm's performance in dynamic space, validating its potential for practical industrial applications. Furthermore, this research not only provides a new algorithmic framework for tackling similar engineering problems but also offers valuable insights for future research in automation and intelligent manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be described in more detail below based on embodiments and with reference to the accompanying drawings, wherein:

[0024] Figure 1 This is a schematic diagram of key monitoring points in the wing-fuselage docking area;

[0025] Figure 2 The flowchart of the improved PSO algorithm applied to the posture adjustment stage is shown in Figure 2.

[0026] Figure 3 Schematic diagram of the RRT* algorithm flow applied to the progressive stage;

[0027] Figure 4 is the normalized loss curve of the pose adjustment stage;

[0028] Figure 5 The expansion steps and node connection relationship of the RRT* algorithm;

[0029] Figure 6 The final path obtained by the RRT* algorithm from backtracking through the search tree. DETAILED DESCRIPTION

[0030] The present invention will be further described below with reference to the accompanying drawings.

[0031] The technical problem to be solved by the present invention is to address the deficiencies and needs of the above-mentioned existing technologies, and to achieve the following by establishing an optimization and planning method in stages:

[0032] (1) Wing-body docking attitude optimization based on improved particle swarm optimization (PSO)

[0033] (2) Progressive path planning for wing-body docking based on RRT*

[0034] These two parts consider the wing-body docking process in reverse, that is, they first consider the second stage and use the optimization results of the second stage as the moving target of the first stage, thus achieving reverse path planning.

[0035] First, an improved particle swarm optimization (PSO) algorithm was used to optimize and analyze the alignment process for the aircraft's wing-body docking. By defining the position and tolerance of each assembly point in detail, the relative positions of key wing-body points in three-dimensional space were evaluated and optimized to achieve precise docking. This process comprehensively considers the various deformations and displacements that components may encounter in real-world environments, calculating the deviation between the current assembly posture monitoring points and the ideal state to evaluate assembly quality. Based on information provided by the supplier and deformation analysis, the displacement of key monitoring points was thoroughly measured and evaluated.

[0036] The RRT* algorithm is then used to plan the path for the incremental phase of the wing-body docking process, specifically the path from the initial point to the target. The core of the algorithm lies in efficiently generating and expanding a search tree, starting from the initial state and gradually expanding toward the target state. This incremental sampling algorithm can quickly find the initial path.

[0037] In one embodiment, the displacement of key points is considered based on the supplier data as part of the tolerance range and reflected in the fitness function. Figure 1 As shown in Figure 1, it is a schematic diagram of the key monitoring points and deformation of the wing-body docking area. Assume that for any key point i, its ideal position coordinate is (X i ,Y i ,

[0038] Z i ), define D i is the design position of point i, V i is the standard deviation data provided by the supplier, Δ i is the actual measured deformation of point i, T i It is the total tolerance of point i after taking into account the supplier data and deformation data. Tolerance T i It can be expressed by the following formula:

[0039] T i =V i +α△ i (1)

[0040] Among them, α is a scaling factor used to adjust the weight of the deformed data in the total tolerance.

[0041] For each key point, there is a single axial distance difference between the actual position and the ideal position, a symmetry tolerance, and a distance difference between specific paired points, so the tolerance T can be i The calculation is broken down into five types of deviations:

[0042]

[0043] Where ΔXi,j is the cross-axis deviation, ΔY i,j is the uniaxial deviation, and is the desired distance between points i and j along the X and Y axes, ΔZ i is the absolute position deviation, z expected,i is its expected position, is the symmetry deviation, X L , Z L and X R , Z R are the coordinates of the symmetrical points on the left and right components, Δcomposite is the composite deviation, and δ is used to consider additional tolerance or correction.

[0044] By integrating the above formulas, the pose evaluation metric is designed as the fitness function "evaluate_assembly," which serves as the loss function for the subsequent improved particle swarm optimization process. The particle swarm algorithm initializes a series of candidate solutions, where each particle represents a possible set of spatial coordinates for the wing. Through simulation, the particle positions and velocities are iteratively updated to minimize the fitness function. During this process, each particle's trajectory is adjusted based on its current position, combined with information about its individual and group historical best positions, effectively guiding the particle toward areas of improved fitness.

[0045] However, traditional particle swarm algorithms (PSOs) have several limitations when applied to aircraft wing-body docking scenarios. First, traditional PSOs converge quickly in the early stages of iteration. While this can quickly reduce errors, it can prematurely lock onto a local optimal solution rather than a global optimal solution, resulting in undesirable wing docking positions. Second, there's a lack of clear guidance on how to set parameters like particle velocity and weighting factors for wing-body docking tasks. These parameters typically require continuous adjustment through experimentation based on the specific task, a process characterized by uncertainty that can lead to docking failure. Finally, wing-body docking involves precise control of multiple spatial coordinates, representing a high-dimensional optimization problem. Traditional PSOs are inefficient when handling high-dimensional problems and are easily limited by dimensionality, making it difficult to effectively explore all regions of the solution space. To address these limitations, embodiments of the present invention offer improvements specifically for wing-body docking scenarios, including mechanisms for position update and velocity adjustment, as well as limiting the maximum allowable deviation of particles from their initial positions. Particle behavior is implemented within the Particle class to ensure effective and practical search results.

[0046] Next, the mechanism of algorithm improvement is described in detail, such as Figure 2 As shown in Figure 1, it is a flow chart of the improved PSO algorithm applied in the posture adjustment stage.

[0047] Step 201, dynamically adjust particle speed and position:

[0048] In particle swarm optimization, each particle's position update depends on the combined influence of its historical best position, the global best position, and random factors. In an embodiment of the invention, the speed and position update formulas of the standard particle swarm algorithm are improved by introducing an adaptive inertia weight w, as well as individual and social learning factors c1 and c2. This setting not only enhances the algorithm's search capability but also improves the convergence speed. The particle swarm algorithm speed and position update formulas are as follows:

[0049]

[0050] in, is the velocity of particle i at time t; w is the inertia weight, which is used to control the continuity of particle velocity; c1 and c2 are learning factors, representing individual cognitive components and social cognitive components respectively; r1 and r2 are random numbers in the interval [0,1]; is the optimal position found by particle i so far, g best is the global optimal position; is the position of particle i at time t.

[0051] Step 202: Improve the update formula of the dynamic inertia weight and the adaptive learning factor:

[0052]

[0053] Among them, w max and w min is the maximum and minimum value of the inertia weight, I i is the current iteration number, I max is the maximum number of iterations. and are the starting and ending values ​​of the learning factor c1, and and are the starting and ending values ​​of the learning factor c2. Based on this approach, the inertia weight w can be reduced as the number of iterations increases, c1 decreases linearly from a higher value, and c2 increases linearly from a lower value, thereby increasing the global search ability in the early stages of the search process and increasing the local search ability in the later stages of the iteration.

[0054] Step 203: Limit the maximum offset:

[0055] To prevent invalid solutions during the particle search process, a mechanism is added to the particle class to limit the maximum distance that a particle can deviate from its base position. When the particle's offset exceeds the preset maximum value, it will automatically adjust back to the maximum offset limit. First, calculate the particle offset according to the following formula:

[0056]

[0057] Step 204: compare whether the offset meets the maximum allowable offset threshold. If the offset is greater than the maximum allowable value, adjust the particle position so that the offset does not exceed the threshold range:

[0058]

[0059] Among them, μ represents the offset, p base Indicates the reference point, ∥μ∥ indicates the modulus of the offset, μ max Represents the offset threshold. This measure ensures the effectiveness of the search and prevents the algorithm from drifting blindly in high-dimensional space.

[0060] Step 205: Customized design of fitness function:

[0061] A custom fitness function was designed based on the specific requirements and tolerances of aircraft wing-body assembly. This function not only considers the absolute positional differences between components but also incorporates structural symmetry and specific geometric constraints. This enables the algorithm to more precisely optimize for the actual assembly quality issues.

[0062] Through these improvements, the particle swarm optimization algorithm proposed in the embodiment of the present invention can more effectively solve the attitude adjustment problem in aircraft wing-body docking, achieving a leap from theory to practical application.

[0063] Next, the RRT* algorithm is used to optimize the progressive phase of the aircraft wing-body docking process. The algorithm flow chart is shown in the following figure. Figure 3 shown.

[0064] Step 301 defines an initial state, consisting of a set of randomly offset positions of the aircraft's wings to simulate the uncertain starting conditions encountered in real-world scenarios. In each iteration, the algorithm first randomly generates a point. To speed up convergence, a point is generated around the target every ten iterations, with the remaining points generated over a wider area.

[0065] Step 302: Find the node in the current tree closest to the newly generated point as the nearest node. Use the "steer" function to generate a new node from the nearest node toward the new point. This function ensures that the new node moves along a certain step size in the direction from the nearest node to the new point, thereby gradually approaching the target.

[0066] In step 303, the total cost from the starting point to each new node is calculated and the new node is connected to the tree. If the new node is close enough to the target state (within a preset threshold), the target is considered to have been reached and the search ends. Once the target node is found, this method demonstrates that the RRT* algorithm not only effectively handles path planning problems in high-dimensional spaces but also adapts to the uncertainty of the initial state, making it an ideal choice for solving path planning in the progressive phase of aircraft wing-body docking.

[0067] Next, the specific implementation of the algorithm.

[0068] The pseudo code of the aircraft wing-body docking path inverse planning algorithm based on the improved PSO-RRT* is shown in Appendix 1.

[0069] Table 1

[0070]

[0071]

[0072]

[0073] Step 401 sets 30 particles to maintain computational efficiency and performs 150 iterations. For the maximum offset setting, historical measurements of qualified docking states are used as predefined target configurations. Particle positions are randomly initialized within the range [-1, 1], and velocities are initialized within the smaller range [-0.1, 0.1] to promote initial exploratory behavior. The inertia weight is initialized to 0.6, the individual learning factor c1 is reduced from an initial value of 2.5 to 0.5, and the social learning factor c2 is increased from an initial value of 0.5 to 2.5.

[0074] Step 402: Output loss. The normalized loss curve of the posture adjustment stage is shown in the attached figure. Figure 4 shown.

[0075] Step 403: Obtain key point information representing the spatial structure of the left wing using the improved PSO algorithm, as shown in Table 2.

[0076] Table 2

[0077]

[0078] In step 404, the optimization result is used as the target endpoint of the path planning in the progressive stage, and the key parameters of the RRT* algorithm are set. In order to ensure the adequacy and efficiency of the search process, 2000 iterations are set. When generating new nodes, the step length is set to 200 units in length, which not only ensures the accuracy of the search, but also avoids the generation of overly dense nodes due to too small a step length, thereby accelerating the convergence of the algorithm. In order to determine whether the target has been reached, a threshold of 100 units in length is set. When the average distance between the new node and the target point is less than this value, the target is considered to have been successfully reached. Every 10 iterations of the algorithm, a random point with a Gaussian distribution and a perturbation of 2000 units in length is generated around the target point to increase the search coverage of the target area and improve the convergence speed. At other times, the random points are evenly distributed in the range of [-10000, 10000] to ensure that the entire search space is fully explored. The optimization of the RRT* algorithm designed in this study makes the search process more efficient, and the generated path shows good smoothness and feasibility while meeting spatial and mechanical constraints. The expansion steps of the algorithm and the node connection relationship are shown in the attached figure. Figure 5 As shown in the figure, the final path obtained by backtracking from the search tree is shown in the attached figure. Figure 6 shown.

[0079] To date, a reverse planning algorithm for aircraft wing-body docking paths has been successfully developed and demonstrated, based on a fusion of an improved PSO and RRT* algorithm. Using a particle swarm optimization algorithm, the algorithm optimizes the attitude adjustment process, enabling rapid identification of the optimal solution within complex spatial constraints. Furthermore, using the RRT* algorithm, precise path planning from the initial state to the final successful attitude adjustment state is achieved, demonstrating its effectiveness in addressing highly challenging spatial planning problems.

[0080] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be employed in conjunction with other described embodiments.

Claims

1. A method for docking an aircraft wing and body, characterized in that: By defining the position and tolerance of each assembly point in detail, evaluating the relative positions of key wing-body points in three-dimensional space, and comprehensively considering the various deformations and displacements of aircraft wing-body components in the actual environment, intelligent reverse planning of the aircraft wing-body docking path is achieved. The specific steps include: First, the wing-body docking is optimized based on the improved particle swarm algorithm. The positions and velocities of all particles in the aircraft wing-body are randomly initialized, the velocity of each particle is updated, and the position of each particle is updated based on the updated velocity. It is determined whether the new position exceeds the maximum allowable deviation. If the maximum allowable deviation is exceeded, the position based on the deviation is updated and the fitness value of each particle is calculated. Then, it is determined whether the termination condition is met. If the maximum allowable deviation is not exceeded, it is directly determined whether the termination condition is met. If the termination condition is met, the optimization result is output. If the termination condition is not met, the algorithm returns to updating the velocity of each particle. Then, the aircraft wing-body docking process is optimized. A set of random offset positions of the aircraft wing is defined as the initial state using random configuration and target configuration to simulate the uncertain starting conditions in actual situations. Random points are generated in space, and the node closest to the newly generated point in the initial state is found as the nearest node. The new node is created by turning from the nearest node to the random point. The cost of reaching the new node from the starting point to the nearest node is calculated. The new node is added to the tree, and the parent node of the new node is set to the nearest node. If the new node is within the threshold distance from the target, a path from the start to the target is constructed; if the new node is not within the threshold distance from the target, the path is returned to turning from the nearest node to the random point to create the new node.

2. The aircraft wing-body docking method according to claim 1, characterized in that: The maximum allowable deviation range takes into account the displacement of key points. Assume that for any key point i, its ideal position coordinates are (X i ,Y i ,Z i ), define D i is the design position of point i, V i is the standard deviation data provided by the supplier, Δ i is the actual measured deformation of point i, T i is the total tolerance of point i after considering the supplier data and deformation data; tolerance T i Expressed as follows: T i =V i +α△ i (1) Among them, α is a scaling factor used to adjust the weight of the deformed data in the total tolerance.

3. The aircraft wing-body docking method according to claim 2, characterized in that: For each key point, there is a single axial distance difference between the actual position and the ideal position, a symmetry tolerance, and a distance difference between specific paired points. The tolerance T i The calculation is broken down into five types of deviations: Where ΔX i,j is the cross-axis deviation, ΔY i,j is the uniaxial deviation, and is the desired distance between points i and j along the X and Y axes, ΔZ i is the absolute position deviation, z expected,i is its expected position, is the symmetry deviation, X L , Z L and X R , Z R are the coordinates of the symmetrical points on the left and right components, Δcomposite is the composite deviation, and δ is used to consider additional tolerance or correction.

4. The aircraft wing-body docking method according to claim 3, characterized in that: The pose evaluation metric is designed as a complex fitness function "evaluate_assembly", which serves as the loss function for the subsequent improved particle swarm optimization process. The particle swarm algorithm is used to initialize a series of candidate solutions, where each particle represents a possible set of spatial coordinates of the wing. The particle positions and velocities are then iteratively updated through simulation to minimize the fitness function. During this process, each particle's trajectory is adjusted based on its current position, combined with information about both its individual and group historical best positions, effectively guiding the particle toward areas with improved fitness.

5. The aircraft wing-body docking method according to claim 4, characterized in that: The steps in optimizing the aircraft wing-body alignment process include: An initial state is defined, consisting of a set of randomly offset positions of the aircraft wing to simulate the uncertain starting conditions in real-world situations. In each iteration, the algorithm first randomly generates a point. To speed up the algorithm's convergence, a point is generated around the target every ten iterations, and a wider area is generated the rest of the time. Find the node closest to the newly generated point in the current tree as the nearest node; use the "steer" function to generate a new node from the nearest node toward the new point. This function ensures that the new node moves along a certain step length in the direction from the nearest node to the new point, thereby gradually approaching the target; For each new node, the total cost from the starting point to the new node is calculated and the new node is connected to the tree; if the new node is within the preset threshold, the goal is considered to have been reached and the search ends.

Images