A method for deformation prevention machining of thin-walled aerospace components

Abstract

The application discloses a deformation prevention machining method for a thin-walled aerospace component and belongs to the technical field of thin-walled aerospace component machining. The method comprises the following steps: obtaining an initial three-dimensional digital model and initial machining stress field data of the thin-walled aerospace component; constructing a surface layer topological structure according to the initial three-dimensional digital model, and extracting curvature distribution characteristics and thickness distribution characteristics; coupling the initial machining stress field data with the curvature distribution characteristics and the thickness distribution characteristics to generate a machining deformation sensitivity atlas; defining a zero-deformation propagation path on the machining deformation sensitivity atlas and performing iterative optimization to obtain a target machining track; and generating a numerical control machining program according to the target machining track to control a machining device to machine the component. The method realizes active inhibition of deformation of the thin-walled aerospace component by constructing a sensitivity atlas and optimizing a low-deformation propagation path.

Description

Technical Field

[0001] This invention relates to the field of thin-walled aerospace component processing technology, specifically to a deformation-resistant processing method for thin-walled aerospace components. Background Technology

[0002] Thin-walled aerospace components are highly susceptible to unexpected bending, warping, and other deformations during machining due to the release of residual stress within the material, severely impacting their geometric accuracy and performance. Existing machining methods typically rely on operational experience or conservative cutting parameters to passively address deformation, lacking quantitative analysis of the coupling effect between component geometry and stress field. After obtaining a 3D model of the workpiece, conventional methods directly plan equally spaced or fixed toolpaths, but these paths fail to consider the influence of curvature variations and wall thickness differences in different regions of the component on deformation propagation.

[0003] Existing technical solutions suffer from the following drawbacks. Firstly, traditional path planning methods treat component geometry information and initial stress state separately, failing to construct a quantitative map that comprehensively reflects the sensitivity of geometry to stress release. This leads to unfavorable overlap between stress release paths and material removal sequences in areas with drastic curvature changes or abrupt thickness shifts, causing localized deformation accumulation. Secondly, even when considering stress factors, existing methods only perform simple global stress equalization, failing to seek a specific route at the processing trajectory level that can block or minimize deformation accumulation along the path. This means the processing sequence still follows empirical settings and cannot adaptively avoid highly sensitive areas that exacerbate deformation propagation. The core issues that urgently need to be addressed are how to quantify and explicitly characterize the processing deformation sensitivity of each region under the coupling effect of geometric features and the initial stress field, and how to determine and optimize a processing trajectory that can actively suppress deformation propagation on this quantitative map. Summary of the Invention

[0004] The present invention aims to provide a deformation-resistant processing method for thin-walled aerospace components, in order to solve the problems in the prior art of lacking quantitative characterization of the sensitivity of component geometry and stress coupling, and being unable to actively plan processing trajectories to suppress deformation propagation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a method for anti-deformation processing of thin-walled aerospace components. The method includes: acquiring an initial three-dimensional digital model and initial processing stress field data of the thin-walled aerospace component; constructing a surface topology based on the initial three-dimensional digital model, and extracting the curvature distribution characteristics and thickness distribution characteristics of the surface topology; coupling the initial processing stress field data with the curvature distribution characteristics and the thickness distribution characteristics to generate a processing deformation sensitivity map; defining a zero-deformation propagation path on the processing deformation sensitivity map, and iteratively optimizing the zero-deformation propagation path to obtain a target processing trajectory; generating a CNC machining program based on the target processing trajectory, and controlling the machining equipment to process the thin-walled aerospace component. This method, by constructing a sensitivity map that integrates residual stress and geometric features, and planning a low-sensitivity processing trajectory, suppresses processing deformation of thin-walled components from the source of the process.

[0006] When acquiring the initial three-dimensional digital model and initial processing stress field data, the point cloud data of the component can be collected by a three-dimensional scanning device and reconstructed into a mesh to obtain the initial three-dimensional digital model. At the same time, the residual stress distribution of the blank state can be simulated by finite element analysis software and discretized onto the mesh nodes of the model to form initial processing stress field data containing the stress tensors of each node, thereby providing accurate raw data for deformation tendency assessment.

[0007] In the process of constructing the surface topology and extracting geometric features, the initial 3D digital model undergoes surface mesh extraction to obtain a surface topology composed of triangular facets. Inferior triangular facets with areas smaller than a preset minimum area threshold or interior angles smaller than a preset minimum angle threshold are removed to improve the stability of feature extraction. The average curvature and Gaussian curvature of each triangular facet are calculated, normalized, and combined to obtain curvature distribution features. Simultaneously, the perpendicular distance from each mesh node to the central axis of the model is calculated and normalized to obtain thickness distribution features. This joint curvature and thickness characterization can sensitively identify geometrically weak areas on thin-walled components that are prone to unstable deformation.

[0008] When generating the machining deformation sensitivity map, the stress tensor of each mesh node is decomposed into normal stress components and tangential stress components. The normal stress component is multiplied by the thickness value of the corresponding node to obtain the thickness-stress coupling coefficient; the tangential stress component is multiplied by the average curvature of the triangular facet to obtain the curvature-stress coupling coefficient. The two coupling coefficients are superimposed on the corresponding nodes and facets to generate the machining deformation sensitivity map. This map deeply integrates stress magnitude, direction, and local geometric features, accurately locating the parts with the highest deformation risk during machining.

[0009] When planning the target processing trajectory, the grid node with the lowest sensitivity value on the processing deformation sensitivity map is selected as the starting point of the path, and the node with the second lowest sensitivity value is selected as the ending point of the path. Connecting the two points generates an initial zero-deformation propagation path, and the sum of the sensitivities of all nodes on this path is calculated as the total sensitivity of the path. Within the neighborhood of this path, the node positions are continuously and randomly perturbed, and the total sensitivity of the new path is recalculated. If the new total sensitivity is less than the current total sensitivity, the new path is updated as the zero-deformation propagation path. This perturbation and comparison are repeated until the iterative convergence condition is met, and the converged path is taken as the target processing trajectory. Preferably, the convergence condition is that the relative rate of change of the total sensitivity is lower than a preset threshold or the number of iterations reaches a preset upper limit during a preset number of iterations. Through this iterative search, the target processing trajectory extends along the low-sensitivity region as much as possible, effectively avoiding the deformation-sensitive region.

[0010] When generating a CNC machining program based on the target machining trajectory, the target trajectory is mapped onto the surface of the initial 3D digital model to obtain a sequence of tool contact points. For two adjacent contact points in the sequence, the difference between the coordinates of the latter point and the former point is calculated as the feed direction vector, and the Euclidean distance between the two points is calculated as the feed step size. Based on the feed direction vector, feed step size, and preset tool diameter and depth of cut, a linear interpolation path is determined, the offset vector of the tool center point relative to the contact point is calculated, the contact point is translated to obtain the tool center point path, and then it is converted into G-code format conforming to the ISO standard. The machining equipment executes layered cutting from the edge of the thin-walled component towards the center according to this code, and the depth of cut of each layer decreases as the machining deformation sensitivity value of the corresponding area of ​​that layer increases. This layered variable depth of cut strategy further avoids deformation caused by cutting force concentration and sudden stress release.

[0011] As a preferred embodiment of the present invention, vibration signals and cutting force signals of the machining area are acquired in real time during the machining process and converted into time-domain feature data. Fourier transform is performed on the time-domain feature data to extract the dominant frequency amplitude and the secondary frequency amplitude. The ratio of the dominant frequency amplitude to the secondary frequency amplitude is calculated as the real-time frequency feature ratio. Simultaneously, the sensitivity value corresponding to the current machining position is read from the machining deformation sensitivity map, and its reciprocal is taken to obtain the reference frequency feature ratio. The absolute value of the difference between the real-time frequency feature ratio and the reference frequency feature ratio is calculated and then divided by the reference frequency feature ratio to obtain the real-time deviation. When the real-time deviation exceeds a preset deviation threshold, machining is paused, and the feed speed of the unmachined portion in the target machining trajectory is adjusted, thereby achieving online closed-loop suppression of machining deformation and ensuring the final dimensional accuracy and surface quality of the component.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This method generates a processing deformation sensitivity map by coupling initial processing stress field data, curvature distribution characteristics, and thickness distribution characteristics. The technique involves decomposing the stress tensor into normal and tangential components. The normal stress component is multiplied by the thickness value of the corresponding node to obtain a thickness-stress coupling coefficient. Simultaneously, the tangential stress component is multiplied by the average curvature of the corresponding triangular facet to obtain a curvature-stress coupling coefficient. These two coefficients are then superimposed to generate the processing deformation sensitivity map. The effect of this method is that it integrates the previously isolated residual stress data, the degree of curvature of the component surface, and wall thickness variation—three influencing factors—into a single, spatially continuous sensitivity scalar field. This directly quantifies and explicitly characterizes the tendency of any location on the component surface to deform during material removal, avoiding the blindness in deformation risk assessment caused by the separation of geometric and physical field information in traditional methods.

[0013] The technique involves iteratively optimizing the zero-deformation propagation path on a machining deformation sensitivity map. This involves selecting low-value points on the sensitivity map as the path's start and end points and connecting them to generate an initial path. After calculating the total path sensitivity, the path node positions are randomly perturbed within the neighborhood. Paths with lower total sensitivity are repeatedly compared and retained until convergence, yielding the target machining trajectory. The effect of this method is that, with minimizing the total path sensitivity as the optimization objective, the machining trajectory automatically searches for and approximates a spatial curve with the weakest cumulative deformation effect on the sensitivity map. This allows the cutting sequence to proactively avoid high-sensitivity areas and block the continuous propagation of deformation along the trajectory direction, transforming passive deformation response into proactive stress release path guidance. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a flowchart of the anti-deformation processing method; Figure 2 This is a flowchart of the process for obtaining the initial three-dimensional digital model and processing stress field data of thin-walled aerospace components; Figure 3 This is a flowchart of the process for extracting the surface topology and calculating the curvature and thickness distribution characteristics of thin-walled aerospace components; Figure 4 This is a flowchart for generating the processing deformation sensitivity map. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] See Figure 1 This invention provides a deformation-resistant machining method for thin-walled aerospace components. The method includes: acquiring an initial three-dimensional digital model and initial machining stress field data of the thin-walled aerospace component; constructing the surface topology of the thin-walled aerospace component based on the initial three-dimensional digital model, and extracting the curvature distribution characteristics and thickness distribution characteristics of the surface topology; coupling the initial machining stress field data with the curvature distribution characteristics and thickness distribution characteristics to generate a machining deformation sensitivity map; defining a zero-deformation propagation path on the machining deformation sensitivity map, and iteratively optimizing the zero-deformation propagation path to obtain a target machining trajectory; generating a CNC machining program based on the target machining trajectory to control the machining equipment to machine the thin-walled aerospace component. By coupling the surface geometric characteristics and internal stress state of the component, a data distribution reflecting the machining deformation sensitivity of each region is obtained. A continuous path that minimizes the deformation propagation effect is planned and optimized on this path, and a tool movement trajectory is generated based on this path, thereby suppressing machining deformation during material removal.

[0018] In practice, the process of obtaining the initial three-dimensional digital model and initial processing stress field data of thin-walled aerospace components is as follows.

[0019] See Figure 2 Point cloud data of thin-walled aerospace components was acquired using a 3D scanning device. The 3D scanning device, a laser scanner, projects a laser beam onto the surface of the thin-walled aerospace component, and a built-in camera captures the reflected laser stripe images. Based on the principle of triangulation, the 3D spatial coordinates of each sampling point on the component surface are calculated to form the raw point cloud data. During the acquisition process, the thin-walled aerospace component is scanned from multiple perspectives to ensure that the point cloud data covers the entire surface of the component. Preprocessing operations are performed on the raw point cloud data, including outlier removal and multi-view point cloud registration. Outlier removal uses a statistical filtering method, calculating the average distance from each point to its nearest neighbors. Points whose average distance exceeds the global average distance plus a preset standard deviation are discarded as outliers. Multi-view point cloud registration uses an iterative nearest-point algorithm. By finding corresponding point pairs between adjacent viewpoints, solving for the rotation matrix and translation vector, the point clouds from each viewpoint are transformed to a unified coordinate system, obtaining the registered complete point cloud data.

[0020] Point cloud data is reconstructed using a mesh model to obtain an initial 3D digital model. The mesh reconstruction employs the Poisson surface reconstruction method, constructing an indicator function whose gradient approximates the normal vector field of the point cloud. Isosurfaces are extracted by solving the Poisson equation, generating a closed mesh model composed of triangular facets. This closed mesh model is then simplified to control the number of triangular facets, resulting in the initial 3D digital model. The initial 3D digital model includes the coordinates of the mesh nodes and the topological connections of the triangular facets; the mesh nodes are the vertices of the triangular facets.

[0021] The residual stress distribution of thin-walled aerospace components in their blank state was simulated using finite element analysis (FEM) software. An initial 3D digital model was imported into the FEM software, and a solid model of the blank was created based on the blank dimensions of the thin-walled aerospace component. The shape of the solid model included the final component shape, while retaining uniform machining allowances. Appropriate material properties were assigned to the solid model, including elastic modulus, Poisson's ratio, yield strength, and thermophysical parameters. An initial residual stress field was set: if the blank manufacturing process was known, residual stress was obtained through process simulation; otherwise, the measured residual stress distribution was used, applying a depth-dependent stress distribution function within the blank. Boundary constraints were applied to prevent rigid body displacement of the solid model. The static analysis solver was then used to calculate the stress equilibrium state of the solid model under the initial residual stress field, obtaining the residual stress distribution. The residual stress distribution includes the stress tensor at each element integration point.

[0022] The residual stress distribution is discretized onto each mesh node of the initial 3D digital model to obtain the initial processing stress field data. The discretization process is as follows: extract the spatial coordinates of each mesh node in the initial 3D digital model, locate the blank solid model element containing that mesh node in the residual stress distribution, and interpolate the stress tensor at the integration point to the mesh node position using the shape function of that element to obtain the stress tensor at the mesh node. After performing the interpolation operation on all mesh nodes, the initial processing stress field data is formed. In the initial processing stress field data, each mesh node stores a stress tensor, which contains three normal stress components and three shear stress components. The stress tensor reflects the spatial differences in the initial stress state inside the blank.

[0023] In specific implementation, please refer to Figure 3 The process of constructing the surface topology of thin-walled aerospace components based on the initial three-dimensional digital model and extracting curvature distribution features and thickness distribution features is as follows.

[0024] A surface mesh is extracted from the initial 3D digital model to obtain a surface topology composed of triangular facets. The extraction process involves traversing all triangular facets in the initial 3D digital model and identifying edges with at least one edge belonging to only one triangular facet as boundary edges. These boundary edges form closed loops that define the outer surface region of the initial 3D digital model. All triangular facets belonging to the outer surface region are extracted to form a set of surface triangular facets. This set of surface triangular facets is then reconstructed through topological connectivity to generate the surface topology composed of triangular facets. Each triangular facet in the surface topology retains its vertex coordinates and normal vector direction information from the initial 3D digital model.

[0025] When extracting the surface mesh, inferior triangular facets with an area smaller than a preset minimum area threshold or an interior angle smaller than a preset minimum angle threshold are removed. The preset minimum area threshold is set to 5% of the average area of ​​all triangular facets in the surface topology, and the preset minimum angle threshold is set to 10 degrees. Triangular facets with an area smaller than the preset minimum area threshold are classified as narrow facets, which may produce numerical singularities in subsequent curvature calculations. Triangular facets with an interior angle smaller than the preset minimum angle threshold are classified as elongated facets, as excessive distortion in the shape of elongated facets will reduce the accuracy of curvature estimation. The surface triangular facet set is traversed, and the area and three interior angles of each triangular facet are calculated. When the area of ​​a triangular facet is smaller than the preset minimum area threshold or any interior angle of a triangular facet is smaller than the preset minimum angle threshold, the triangular facet is marked as an inferior triangular facet. The marked inferior triangular facets are removed from the surface triangular facet set, and the triangular facets around the removed area are re-stitched to maintain the continuity of the surface topology. The stitching operation is achieved through local edge flipping and vertex merging, ensuring that no holes appear in the surface topology.

[0026] The average curvature and Gaussian curvature of each triangular facet in the surface topology are calculated and normalized to form a curvature distribution characteristic. The calculation method for the average curvature and Gaussian curvature of each triangular facet is as follows: using discrete differential geometry, a local quadratic surface is constructed within the local region consisting of the triangular facet and its one-neighboring triangular facets. The average curvature and Gaussian curvature are calculated using the principal curvatures of the fitted quadratic surface. The average curvature is the arithmetic mean of the two principal curvatures, and the Gaussian curvature is the product of the two principal curvatures. This process is repeated for all triangular facets in the surface topology to obtain the average curvature and Gaussian curvature values ​​for each facet. Next, the average curvature values ​​of all triangular facets are normalized using a linear mapping method, mapping all average curvature values ​​linearly to the interval [0,1]. The mapping formula is: for the average curvature value of the i-th triangular facet, the normalized average curvature value is equal to the original average curvature value minus the difference between the minimum average curvature value of all triangular facets, divided by the difference between the maximum and minimum average curvature values ​​of all triangular facets. Similarly, the Gaussian curvature values ​​of all triangular facets are normalized and mapped to the interval [0,1]. The normalized average curvature value and the normalized Gaussian curvature value of each triangular facet are combined into a two-dimensional vector, which serves as the curvature distribution feature of that triangular facet. The curvature distribution features of all triangular facets collectively constitute the curvature distribution feature of the thin-walled aerospace component.

[0027] The thickness distribution characteristics are obtained by calculating the vertical distance from each grid node in the surface topology to the central axis of the initial 3D digital model and normalizing the vertical distances of all grid nodes. The central axis of the initial 3D digital model is obtained as follows: the initial 3D digital model is extracted as a skeleton using a Laplacian contraction-based skeleton extraction algorithm. The surface topology is iteratively contracted inward until it converges into a skeleton composed of a set of one-dimensional nodes and connecting edges. The central axis is then fitted to the skeleton. The central axis is the curved surface at the midpoint of the component's wall thickness direction. For each grid node in the surface topology, the vertical distance from the grid node to the central axis is calculated. The vertical distance is calculated by searching for the nearest projection point on the central axis; the Euclidean distance between the grid node and the projection point is the vertical distance. This process is repeated for all grid nodes in the surface topology to obtain the vertical distance for each grid node. Next, the vertical distances of all grid nodes are normalized using a linear mapping method, which maps all vertical distance values ​​linearly to the interval [0,1]. The mapping method is as follows: for the j-th grid node, the normalized vertical distance value is equal to the original vertical distance value minus the difference between the minimum vertical distance values ​​of all grid nodes, divided by the difference between the maximum and minimum vertical distance values ​​of all grid nodes. The normalized vertical distance values ​​constitute the thickness distribution feature, and the thickness distribution feature corresponds one-to-one with the grid nodes.

[0028] In specific implementation, please refer to Figure 4 The process of coupling the initial processing stress field data with the curvature distribution characteristics and thickness distribution characteristics to generate a processing deformation sensitivity map is as follows.

[0029] The stress tensor of each grid node in the initial processing stress field data is decomposed into normal stress components and tangential stress components. For each grid node in the initial processing stress field data, the stress tensor stored in that grid node is read. The stress tensor is a 3x3 symmetric matrix containing three normal stress components and three shear stress components. At that grid node, the local surface normal vector of the surface topology is calculated. The local surface normal vector is obtained by taking the area-weighted average of the normal vectors of all triangular facets adjacent to that grid node. Using the local surface normal vector as the normal direction and the plane perpendicular to the local surface normal vector as the tangential plane, the stress tensor is projected onto the normal direction and the tangential plane direction to obtain the normal stress components and tangential stress components. The normal stress component is the normal stress value of the stress tensor in the direction of the local surface normal vector. The normal stress component is calculated by multiplying the stress tensor on the left by the transpose of the local surface normal vector, and then multiplying it on the right by the local surface normal vector to obtain the scalar form of the normal stress component. The tangential stress component is the set of stress components of the stress tensor in the tangential plane. The tangential stress component is calculated by subtracting the portion corresponding to the normal stress component from the stress tensor to obtain the tangential stress tensor, and then taking the maximum principal stress value of the tangential stress tensor as the tangential stress component. Stress tensor decomposition is performed on all mesh nodes in the initial processing stress field data, resulting in one normal stress component and one tangential stress component for each mesh node.

[0030] The thickness-stress coupling coefficient is obtained by multiplying the normal stress component by the thickness value of the corresponding mesh node in the thickness distribution characteristics. For each mesh node, the normal stress component is extracted from the initial processing stress field data, and the corresponding thickness value is extracted from the thickness distribution characteristics. The thickness value is the normalized value of the vertical distance from the mesh node to the axial surface in the initial 3D digital model. The product of the normal stress component and the thickness value is the thickness-stress coupling coefficient. The thickness-stress coupling coefficient characterizes the combined effect of the normal stress and local wall thickness on processing deformation at that mesh node.

[0031] The curvature-stress coupling coefficient is obtained by multiplying the tangential stress component by the average curvature of the corresponding triangular facet in the curvature distribution feature. The curvature distribution feature corresponds one-to-one with the triangular facet, while the tangential stress component corresponds one-to-one with the mesh node. The method for associating the tangential stress component of the mesh node with the triangular facet is as follows: for each triangular facet, obtain the three mesh nodes constituting the facet, extract the tangential stress components of each of the three mesh nodes, and take the arithmetic mean of the three tangential stress components as the representative value of the tangential stress of that triangular facet. The average curvature corresponding to that triangular facet is extracted from the curvature distribution feature; the average curvature is a normalized value. The representative value of the tangential stress of the triangular facet is directly multiplied by the average curvature; the product is the curvature-stress coupling coefficient. The curvature-stress coupling coefficient characterizes the combined effect of the tangential stress and the surface geometric curvature on the processing deformation of the component in the triangular facet region.

[0032] The thickness stress coupling coefficient and curvature stress coupling coefficient are superimposed onto the corresponding mesh nodes and triangular facets to generate a machining deformation sensitivity map. The thickness stress coupling coefficient is defined on the mesh nodes, and the curvature stress coupling coefficient is defined on the triangular facets. The curvature stress coupling coefficient is mapped from the triangular facets to the mesh nodes as follows: For each mesh node, all triangular facets containing that mesh node are obtained, and the arithmetic mean of the curvature stress coupling coefficients of these triangular facets is calculated as the representative value of the curvature stress coupling for that mesh node. For each mesh node, the thickness stress coupling coefficient is added to the representative value of the curvature stress coupling to obtain the machining deformation sensitivity value of the mesh node. The calculation of the machining deformation sensitivity value is expressed as follows:

[0033]

[0034] in, Indicates the first The processing deformation sensitivity value of each mesh node. The range of values ​​for is the interval of real numbers; Indicates the first Thickness stress coupling coefficient of each mesh node It is obtained by multiplying the normal stress component by the thickness value; Indicates the first The representative value of curvature stress coupling of each grid node. By including the first The curvature stress coupling coefficients of all triangular faces of a grid node are obtained by taking the arithmetic mean. This represents the thickness stress weighting factor. The value is 0.6. The value is based on the fact that in the processing deformation of thin-walled components, the contribution of the coupling effect between normal stress and wall thickness to the deformation accounts for about 60%. Indicates the curvature stress weighting factor. The value is 0.4. The values ​​are based on the following: the contribution of the coupling effect of tangential stress and curvature to deformation accounts for approximately 40%. The sum of the thickness stress weighting factor and the curvature stress weighting factor is one.

[0035] The processing deformation sensitivity value of each mesh node is mapped to the corresponding mesh node, and based on the mesh nodes, it is displayed on the 3D mesh model through color rendering or numerical annotation to form a processing deformation sensitivity map. Each mesh node in the processing deformation sensitivity map is assigned a processing deformation sensitivity value, and the magnitude of the processing deformation sensitivity value reflects the degree to which the region where the mesh node is located is prone to processing deformation during material removal.

[0036] In practice, a zero-deformation propagation path is defined on the processing deformation sensitivity map, and the zero-deformation propagation path is iteratively optimized to obtain the target processing trajectory as follows.

[0037] On the machining deformation sensitivity map, the mesh node with the lowest sensitivity value is selected as the path start point, and the mesh node with the second lowest sensitivity value is selected as the path end point. The machining deformation sensitivity values ​​of all mesh nodes in the machining deformation sensitivity map are traversed, and a sorting algorithm is used to find the mesh node with the lowest and second lowest machining deformation sensitivity values. When multiple mesh nodes have the same lowest machining deformation sensitivity value, the mesh node whose spatial location is closest to the component boundary is selected as the path start point. When multiple mesh nodes have the same second lowest machining deformation sensitivity value, the mesh node whose spatial location is farthest from the path start point in Euclidean distance is selected as the path end point. Both the path start point and the path end point are mesh nodes on the surface topology.

[0038] Connecting the starting and ending points of the path generates an initial zero-deformation propagation path. The connection method employs a geodesic path generation approach based on grid topology. On the triangular mesh of the surface topology, with the starting point as the source and the ending point as the target, the Dijkstra algorithm searches for the shortest path along the mesh edges. The Dijkstra algorithm operates on a graph composed of mesh nodes, where the nodes are the mesh nodes in the surface topology and the edges are the edges of the triangular meshes. The weight of each edge is set to the average of the processing deformation sensitivity values ​​of the mesh nodes at its two endpoints. The resulting path consists of a series of consecutive mesh nodes arranged sequentially; this series of mesh nodes constitutes the initial zero-deformation propagation path. The total number of mesh nodes on the zero-deformation propagation path is denoted as N, where N is an integer greater than or equal to two.

[0039] The sum of the sensitivity values ​​of all mesh nodes along the zero-deformation propagation path is calculated as the total path sensitivity. The total path sensitivity is calculated as follows:

[0040]

[0041] in, Indicates the overall path sensitivity. The value is a real number; This represents the total number of mesh nodes contained in the zero-deformation propagation path; This indicates the sequence number of the mesh node along the zero-deformation propagation path. The range of values ​​is from arrive Integers; Indicates the sequence number is The processing deformation sensitivity value corresponding to the mesh node. Read directly from the processing deformation sensitivity map.

[0042] Within the neighborhood of the zero-deformation propagation path, randomly perturb the positions of the path nodes and recalculate the total sensitivity of the new path after the perturbation. The neighborhood of the zero-deformation propagation path is defined as follows: for each internal grid node on the zero-deformation propagation path, the internal grid node refers to the node with the sequence number... arrive For each internal grid node, all adjacent grid nodes directly connected to it via a grid edge in the surface topology are obtained. These adjacent grid nodes constitute the neighborhood candidate set of the internal grid node. A random adjacent grid node is selected from the neighborhood candidate set to replace the current internal grid node, completing the position perturbation of the internal grid node. After performing random perturbation on each internal grid node along the zero-deformation propagation path, the perturbed grid node sequence is reconnected to form a new zero-deformation propagation path. The starting point and ending point of the new zero-deformation propagation path remain the same. The total path sensitivity is recalculated for the new zero-deformation propagation path to obtain the new total path sensitivity.

[0043] When the total sensitivity of the new path is less than the total sensitivity of the path, the perturbed path is updated to the current zero-deformation propagation path, and the updated zero-deformation propagation path continues to participate in subsequent perturbations and comparisons. When the total sensitivity of the new path is greater than or equal to the total sensitivity of the path, the perturbed path is discarded, and the original zero-deformation propagation path remains unchanged.

[0044] Repeat the perturbation and comparison process until a preset iterative convergence condition is met. The zero-deformation propagation path at convergence is taken as the target processing trajectory. The preset iterative convergence condition is that the relative rate of change of the total path sensitivity in a preset number of consecutive iterations is lower than a preset threshold, or the number of iterations reaches a preset upper limit. The preset number of iterations is set to ten, the preset threshold is set to one percent, and the preset upper limit is set to five hundred. After each iteration, whether or not an update is performed, the total path sensitivity after that iteration is recorded, and the relative rate of change of the total path sensitivity in the most recent ten consecutive iterations is calculated. The relative rate of change is calculated by subtracting the minimum value from the maximum value of the total path sensitivity in ten iterations, and then dividing by the absolute value of the average value of the total path sensitivity in ten iterations. When the relative rate of change is lower than one percent, the zero-deformation propagation path is considered to have converged. When the number of iterations reaches five hundred, even if the relative rate of change is not lower than one percent, the iteration is forcibly terminated, and the current zero-deformation propagation path is taken as the target processing trajectory. The iteration count starts from the first execution of perturbation and comparison.

[0045] In practice, the process of generating a CNC machining program based on the target machining trajectory and controlling the machining equipment to process thin-walled aerospace components is as follows.

[0046] The target machining trajectory is mapped onto the surface of an initial 3D digital model of a thin-walled aerospace component to obtain a sequence of tool contact points in 3D space. The target machining trajectory consists of a series of sequentially arranged mesh nodes, each with 3D spatial coordinates. These mesh nodes are directly used as tool contact points, and a sequence of tool contact points is generated according to the arrangement order of the mesh nodes in the target machining trajectory. The tool contact point sequence is represented as follows: ,in, This indicates the total number of tool contact points. It is an integer greater than or equal to two. Each tool contact point is contained in the initial three-dimensional digital model coordinate system. Coordinate values coordinate values ​​and Coordinate values.

[0047] Based on the position coordinates of two adjacent tool contact points in the tool contact point sequence, calculate the tool feed direction vector and feed step size. Extract the first and second tool contact points from the sequence; the second contact point is the next contact point in the sequence after the first. Calculate the difference between the position coordinates of the second and first tool contact points to obtain the feed direction vector. The feed direction vector is a three-dimensional vector representing the spatial direction of the tool movement from the first to the second tool contact point. Calculate the Euclidean distance between the second and first tool contact points to obtain the feed step size. The Euclidean distance is calculated as follows: [The text abruptly ends here, so the translation stops as well.] Coordinate value minus the first tool contact point The square of the difference in coordinate values, the second tool contact point Coordinate value minus the first tool contact point The square of the difference in coordinate values, and the second tool contact point Coordinate value minus the first tool contact point Sum the squares of the differences in the coordinate values, and then take the square root of the sum. Traverse all adjacent tool contact point pairs in the tool contact point sequence to obtain the feed direction vector and feed step size corresponding to each pair of adjacent tool contact points.

[0048] Based on the feed direction vector and feed step, combined with the preset tool diameter and depth of cut, the toolpath code for the CNC machining program is generated. The preset tool diameter is the cutting edge size of the tool selected for machining. The tool diameter is pre-selected based on the material properties and machining accuracy requirements of the thin-walled aerospace component, and its value typically ranges from 2 mm to 20 mm. The preset depth of cut is the depth to which the tool cuts into the material along the tool axis in a single pass. The depth of cut is pre-set based on the overall wall thickness range of the thin-walled aerospace component, and its value ranges from 0.1 mm to 2 mm. The process of generating the toolpath code is as follows: Based on the feed direction vector and feed step, the linear interpolation path between two adjacent tool contact points is determined. The linear interpolation path is a straight line segment extending one feed step from the first tool contact point along the feed direction vector. Based on the preset tool diameter and depth of cut, the offset vector of the tool center point relative to the tool contact point is calculated. The offset vector is calculated as follows: When the tool is a ball end mill, the offset vector equals the difference between the tool radius multiplied by the surface unit normal vector and the tool radius multiplied by the unit vector in the tool axis direction, plus the sum of the depth of cut multiplied by the unit vector in the tool axis direction. The tool radius is half the tool diameter. The surface unit normal vector is the unit outward normal vector of the initial 3D digital model surface corresponding to the tool contact point. The unit vector in the tool axis direction is the unit outward normal vector of the machining coordinate system. The positive axis unit vector. Each tool contact point on the linear interpolation path is translated according to the offset vector to obtain the tool center point path. The coordinates of each tool center point on the tool center point path are the coordinates of the tool contact point plus the offset vector. The tool center point path is converted into G-code format conforming to ISO standards to generate the tool path code for the CNC machining program. The conversion process includes: inserting a rapid positioning command G00 at the start of the tool center point path, inserting a linear interpolation command G01 between adjacent tool center points, and attaching the corresponding feed rate command F value and spindle speed command S value. The feed rate command F value and spindle speed command S value are selected from a standard cutting parameter database based on the tool diameter, depth of cut, and material type of thin-walled aerospace components.

[0049] The toolpath code is transmitted to the CNC system of the machining equipment, which then controls the equipment to perform layered cutting on the thin-walled aerospace component according to the toolpath code. The transmission method involves sending the toolpath code of the CNC machining program to the CNC system memory of the machining equipment via a serial communication interface or Ethernet communication interface. The CNC system parses the toolpath code line by line and drives the servo motors and spindle motors to execute the corresponding movements. Layered cutting is performed progressively from the edge of the thin-walled aerospace component towards the center, with the cutting depth of each layer decreasing as the machining deformation sensitivity value of the corresponding area increases. The implementation method of layered cutting is as follows: the entire machining area is divided into multiple machining layers along the tool axis direction, with each layer covering the entire projected area of ​​the target machining trajectory. For each machining layer, the machining deformation sensitivity values ​​of all grid nodes within the covered area are read, and the maximum value of these values ​​is calculated as the maximum sensitivity of that machining layer. The cutting depth of a machining layer is determined based on its maximum sensitivity. The cutting depth is calculated as follows: when the maximum sensitivity is less than or equal to a preset lower sensitivity threshold, the cutting depth is the preset maximum cutting depth; when the maximum sensitivity is greater than or equal to a preset upper sensitivity threshold, the cutting depth is the preset minimum cutting depth; when the maximum sensitivity is between the preset lower and upper sensitivity thresholds, the cutting depth is determined by linear interpolation. The linear interpolation formula is: cutting depth equals the preset maximum cutting depth minus (maximum sensitivity minus preset lower sensitivity threshold) divided by (preset upper sensitivity threshold minus preset lower sensitivity threshold) multiplied by (preset maximum cutting depth minus preset minimum cutting depth). The preset lower sensitivity threshold is 0.2, the preset upper sensitivity threshold is 0.8, the preset maximum cutting depth is 1.5 mm, and the preset minimum cutting depth is 0.3 mm. Machining begins at the edge of the component and proceeds sequentially towards the center until all machining layers are completed.

[0050] During the machining process of thin-walled aerospace components, real-time monitoring and adaptive adjustment are performed. Vibration and cutting force signals within the machining area are acquired in real time. Vibration signals are acquired by installing an accelerometer on the spindle box or workpiece fixture of the machining equipment. The accelerometer converts the mechanical vibration generated during machining into a voltage signal. After amplification and filtering by a signal conditioner, the voltage signal is converted from analog to digital by a data acquisition card at a preset sampling frequency to obtain the time-series data of the vibration signal. Cutting force signals are acquired by installing a force sensor under the worktable or on the tool holder of the machining equipment. The force sensor converts the cutting force into a charge signal. After amplification and filtering by a charge amplifier, the charge signal is converted from analog to digital by a data acquisition card at a preset sampling frequency to obtain the time-series data of the cutting force signal. The preset sampling frequency is set to 10 kHz.

[0051] The vibration signal and cutting force signal are converted into time-domain feature data. The conversion method is as follows: the root mean square (RMS) value of the vibration signal time series data is calculated to obtain the RMS value of the vibration signal; the peak value of the cutting force signal time series data is calculated to obtain the peak value of the cutting force signal. The combination of the RMS value of the vibration signal and the peak value of the cutting force signal constitutes the time-domain feature data.

[0052] The real-time deviation of the current processing is calculated by comparing the time-domain feature data with the sensitivity values ​​at the corresponding positions in the processing deformation sensitivity map. The calculation process involves performing a Fourier transform on the time-domain feature data using the Fast Fourier Transform algorithm to convert the time-domain feature data to the frequency domain, obtaining the spectrum data. The frequency component with the largest amplitude in the spectrum data is extracted as the dominant frequency, and the amplitude corresponding to the dominant frequency is recorded as the dominant frequency amplitude. Extract the frequency component with the second largest amplitude from the spectral data as the second frequency, and record the amplitude corresponding to the second frequency as the second frequency amplitude. Main frequency amplitude The vertical axis value corresponding to the highest amplitude point in the spectrum is the secondary frequency amplitude. This represents the vertical coordinate value corresponding to the second highest amplitude point in the spectrum. Calculate the main frequency amplitude. With secondary frequency amplitude The ratio of the two values ​​is used to obtain the real-time frequency characteristic ratio. , Read the machining deformation sensitivity value corresponding to the current machining position from the machining deformation sensitivity map. The current machining position is the mesh node corresponding to the currently executing tool contact point in the tool contact point sequence. Calculate the machining deformation sensitivity value. The reciprocal of the reference frequency characteristic ratio is used to obtain the reference frequency characteristic ratio. , When the processing deformation sensitivity value When it is zero, the reference frequency characteristic ratio is... Set a preset maximum value, which is one thousand. Calculate the real-time deviation. Real-time deviation The calculation formula is expressed as:

[0053]

[0054] in, Indicates the real-time deviation. The range of values ​​for is non-negative real numbers; Indicates the real-time frequency characteristic ratio. Through the main frequency amplitude Divide by the second frequency amplitude get; Represents the reference frequency characteristic ratio. By processing deformation sensitivity values The reciprocal of the product is obtained; This represents the amplitude corresponding to the main frequency in the spectrum data; This represents the amplitude corresponding to the second frequency in the spectrum data; This represents the machining deformation sensitivity value corresponding to the current machining position, read from the machining deformation sensitivity map.

[0055] When the real-time deviation exceeds the preset deviation threshold, machining is paused and the feed rate of the unmachined portion of the target machining trajectory is adjusted. The preset deviation threshold is set to 0.3. The control method for pausing machining is as follows: the CNC system executes the pause command M00, stopping the spindle rotation and feed motion. The method for adjusting the feed rate of the unmachined portion of the target machining trajectory is as follows: for the unexecuted tool contact point sequence segment in the target machining trajectory, the original feed rate value is multiplied by an adjustment factor, which is equal to one minus the real-time deviation. When the adjustment factor is less than 0.2, it is set to 0.2 to avoid excessively low feed rates. The adjusted feed rate is applied to the toolpath code of subsequent CNC machining programs, replacing the original F instruction value, and then the machining process is resumed. The resumption of the machining process is confirmed by the operator before the cycle start command is executed.

[0056] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for preventing deformation in the processing of thin-walled aerospace components, characterized in that, The method includes: To obtain the initial three-dimensional digital model and initial processing stress field data of thin-walled aerospace components; The surface topology of the thin-walled aerospace component is constructed based on the initial three-dimensional digital model, and the curvature distribution characteristics and thickness distribution characteristics of the surface topology are extracted. The initial processing stress field data is coupled with the curvature distribution characteristics and the thickness distribution characteristics to generate a processing deformation sensitivity map; A zero-deformation propagation path is defined on the processing deformation sensitivity map, and the zero-deformation propagation path is iteratively optimized to obtain the target processing trajectory; A CNC machining program is generated based on the target machining trajectory, and the machining equipment is controlled to process the thin-walled aerospace component.

2. The anti-deformation processing method for thin-walled aerospace components according to claim 1, characterized in that, The acquisition of the initial three-dimensional digital model and initial processing stress field data of the thin-walled aerospace component includes: The point cloud data of the thin-walled aerospace component is acquired by a 3D scanning device, and the point cloud data is reconstructed into a grid to obtain the initial 3D digital model. The residual stress distribution of the thin-walled aerospace component in the blank state is simulated by finite element analysis software, and the residual stress distribution is discretized to each mesh node of the initial three-dimensional digital model to obtain the initial processing stress field data.

3. The anti-deformation processing method for thin-walled aerospace components according to claim 1, characterized in that, Based on the initial three-dimensional digital model, the surface topology of the thin-walled aerospace component is constructed, and the curvature distribution characteristics and thickness distribution characteristics of the surface topology are extracted, including: The surface mesh of the initial three-dimensional digital model is extracted to obtain the surface topology structure composed of triangular facets; The average curvature and Gaussian curvature of each triangular facet in the surface topology are calculated, and the average curvature and Gaussian curvature of all triangular facets are normalized and then combined to form the curvature distribution feature. The thickness distribution characteristics are obtained by calculating the vertical distance from each grid node in the surface topology to the axial plane in the initial three-dimensional digital model and normalizing the vertical distances of all grid nodes.

4. The anti-deformation processing method for thin-walled aerospace components according to claim 3, characterized in that, When extracting the surface mesh of the initial three-dimensional digital model, inferior triangular facets with an area smaller than a preset minimum area threshold or an interior angle smaller than a preset minimum angle threshold are removed.

5. The anti-deformation processing method for thin-walled aerospace components according to claim 1, characterized in that, The initial processing stress field data is coupled with the curvature distribution characteristics and the thickness distribution characteristics to generate a processing deformation sensitivity map, including: The stress tensor of each grid node in the initial processing stress field data is decomposed into normal stress components and tangential stress components. The thickness stress coupling coefficient is obtained by multiplying the normal stress component with the thickness value of the corresponding grid node in the thickness distribution feature. The curvature-stress coupling coefficient is obtained by multiplying the tangential stress component by the average curvature of the corresponding triangular facet in the curvature distribution feature. The thickness stress coupling coefficient and the curvature stress coupling coefficient are superimposed on the corresponding mesh nodes and triangular facets to generate the processing deformation sensitivity map.

6. The anti-deformation processing method for thin-walled aerospace components according to claim 1, characterized in that, A zero-deformation propagation path is defined on the processing deformation sensitivity map, and the zero-deformation propagation path is iteratively optimized to obtain the target processing trajectory, including: On the processing deformation sensitivity map, the grid node with the lowest sensitivity value is selected as the starting point of the path, and the grid node with the second lowest sensitivity value is selected as the ending point of the path. Connect the starting point and the ending point of the path to generate the initial zero-deformation propagation path; The sum of the sensitivity values ​​of all grid nodes on the zero-deformation propagation path is calculated as the total path sensitivity; The positions of randomly perturbed path nodes within the neighborhood of the zero-deformation propagation path are then randomly adjusted, and the total sensitivity of the new path after the perturbation is recalculated. When the total sensitivity of the new path is less than the total sensitivity of the path, the perturbed path is updated to the current zero-deformation propagation path, and the perturbation and comparison are repeated until the preset iterative convergence condition is met. The zero-deformation propagation path at the time of convergence is taken as the target processing trajectory.

7. A method for preventing deformation processing of thin-walled aerospace components according to claim 6, characterized in that, The preset iterative convergence condition is that the relative rate of change of the total sensitivity of the path is lower than a preset threshold or the number of iterations reaches a preset upper limit during a preset number of consecutive iterations.

8. A method for preventing deformation processing of thin-walled aerospace components according to claim 1, characterized in that, Generating a CNC machining program based on the target machining trajectory and controlling the machining equipment to process the thin-walled aerospace component includes: The target machining trajectory is mapped onto the surface of the initial three-dimensional digital model of the thin-walled aerospace component to obtain the tool contact point sequence in three-dimensional space; Based on the position coordinates of two adjacent tool contact points in the tool contact point sequence, calculate the tool feed direction vector and feed step size; Based on the feed direction vector and the feed step size, combined with the preset tool diameter and depth of cut, the tool path code of the CNC machining program is generated; The toolpath code is transmitted to the CNC system of the machining equipment, which then controls the machining equipment to perform layered cutting machining on the thin-walled aerospace component according to the toolpath code.

9. A method for preventing deformation processing of thin-walled aerospace components according to claim 8, characterized in that, The layered cutting process is carried out in a progressive manner from the edge of the thin-walled aerospace component towards the center, and the cutting depth of each layer decreases as the processing deformation sensitivity value of the corresponding region of that layer increases.

10. A method for preventing deformation processing of thin-walled aerospace components according to claim 8, characterized in that, The step of calculating the tool feed direction vector and feed step size based on the position coordinates of two adjacent tool contact points in the tool contact point sequence includes: Extract the first tool contact point and the second tool contact point from the tool contact point sequence, wherein the second tool contact point is the next contact point of the first tool contact point in the sequence; The feed direction vector is obtained by calculating the difference between the position coordinates of the second tool contact point and the position coordinates of the first tool contact point. Calculate the Euclidean distance between the second tool contact point and the first tool contact point to obtain the feed step size.

Images