Space engine 3D printing interlayer metallurgical bonding enhancing method

Abstract

The invention discloses an airspace engine 3D printing interlayer metallurgical bonding enhancing method which comprises the following steps: S1, slicing an airspace engine 3D model, and constructing a printing path diagram structure; s2, constructing a lattice consistency map of the current printing layer; s3, performing graph structure alignment on the lattice consistency graphs of the current printing layer and the previous printing layer to generate a cross-layer lattice graph sequence; s4, inputting the cross-layer lattice atlas sequence into the graph neural control body model, obtaining a disturbance sensitivity vector and a disturbance score value of each node, and outputting a correction control strategy vector; s5, executing a cross-layer disturbance projection operation based on the corrected control strategy vector, and generating a control intervention instruction; and S6, repeatedly executing the steps S2 to S5 to form a complete lattice atlas evolution record sequence. According to the method, the lattice atlas modeling and the graph neural control body model are utilized to realize the metallurgical enhancement between the printing layers of the airspace engine, and the method has the advantages of closed loop regulation and control, structural continuity and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of metallurgical bonding control, and particularly to a method for enhancing the interlayer metallurgical bonding of 3D printing of aerospace engines. Background Art

[0002] In the field of aerospace, with the continuous improvement of the structural complexity of engines and the extreme development of service conditions, higher requirements are put forward for the manufacturing precision, microstructure properties and overall service life of metal components. Metal 3D printing technology, especially additive manufacturing processes such as selective laser melting (SLM) and electron beam melting (EBM), has become an important means for manufacturing key components of aerospace engines because it can achieve the integrated forming of complex structures, has high material utilization rate and a large adjustable range of parameters. However, in the printing process of layer-by-layer stacking, the quality problem of interlayer metallurgical bonding has always been an important bottleneck restricting its reliability. Due to the non-uniformity of heat input conditions, the change of cooling rate and the non-coordination of grain growth directions between different layers, it is very easy to form metallurgical defects such as microcracks, pores and grain boundary mismatches in the interlayer region, thereby affecting the tissue continuity and mechanical property consistency of the material.

[0003] In the prior art, researchers mostly adopt methods of static parameter optimization or image monitoring during the printing process to improve the interlayer bonding quality. The static optimization method usually combines and adjusts parameters such as printing power, scanning speed and scanning strategy through experimental design in order to obtain a forming result with higher macroscopic density; while the image monitoring method relies on infrared thermal imagers, optical cameras or laser reflection imaging equipment to monitor the printing process online, and is used to identify possible molten pool instability, abnormal spatter or surface texture defects. Although these methods improve the printing stability to a certain extent, they generally have limitations such as response lag, uncontrollability of microstructure, and lack of understanding of cross-layer structures and intelligent feedback mechanisms.

[0004] In terms of tissue modeling, some studies have tried to introduce machine learning models to perform pattern recognition on image data, but most of them stay at the static analysis of the single-layer printing state and lack a systematic modeling of the evolution process of grains with the printing level. Especially in terms of lattice-level structure control, the existing methods often cannot jointly model the cross-layer lattice direction consistency, thermal field conduction continuity and stress evolution trend, resulting in the control means being difficult to provide an implementable prediction and intervention plan from the perspective of microscopic metallurgical behavior.

[0005] In addition, traditional control methods basically rely on the feedforward control logic of preset printing paths and parameters, lacking the closed-loop adjustment ability based on the actual forming state, and unable to achieve dynamic identification and local adjustment of key control nodes in the path map structure. Even if some advanced systems introduce path replanning mechanisms, they are mostly based on the compensation of macroscopic printing errors and cannot be accurately mapped to tissue offsets, thermal-stress coupling perturbations, or microscopic defect predictions at the lattice level. Therefore, it is difficult to essentially enhance metallurgical continuity.

[0006] Therefore, how to provide a method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] An object of the present invention is to propose a method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines. The present invention uses the construction of lattice consistency maps and the method of graph neural control body modeling to accurately perceive and dynamically model the grain evolution state of each layer during the 3D printing process of aerospace engines. Combining the perturbation scoring and cross-layer projection mechanisms, it outputs a path-level control strategy and realizes real-time closed-loop intervention, effectively improving the quality of interlayer metallurgical bonding, and having the advantages of high control accuracy, strong adjustment feedforward, good structural continuity, and adaptability to complex-shaped paths.

[0008] A method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines according to an embodiment of the present invention includes the following steps: S1. Perform slicing processing on the 3D model of the aerospace engine component to generate the printing path data of each printing layer, and construct a printing path map structure; S2. Collect the infrared thermal image, optical texture image, laser reflection image, and printing parameters of the current printing layer, and construct a corresponding lattice consistency map in combination with the printing path map structure; S3. Align the lattice consistency maps of the current printing layer and the previous printing layer in terms of graph structure, establish a cross-layer connection edge set, and generate a cross-layer lattice map sequence; S4. Input the cross-layer lattice map sequence into the graph neural control body model, obtain the perturbation sensitivity vector and perturbation scoring value of each node, and output the corrected control strategy vector of the current printing layer; S5. Based on the corrected control strategy vector, perform a cross-layer perturbation projection operation, project the area to be corrected in the current printing layer onto the cross-layer printing path map structure through spatial mapping, and generate a control intervention instruction; S6. Based on the control intervention instruction, repeat steps S2 to S5, iteratively complete the lattice consistency maps of all printing layers layer by layer, and finally form a complete lattice map evolution record sequence.

[0009] Optionally, the S1 specifically includes the spatial coordinate information, scanning order, scanning direction, and scanning speed of each layer of printing path points, and constructs a printing path map structure, where nodes represent path points, edges represent the printing order relationship between adjacent path points, and edge attributes include the scanning direction vector of the path segment, the geometric distance between the start and end points, and the scanning speed gradient.

[0010] Optionally, the S2 specifically includes: S21. Collect the infrared thermal image corresponding to the current printing layer to obtain the two-dimensional image information of the thermal field distribution during the printing process; S22. Collect the optical texture map and laser reflection map of the current printing layer to obtain the surface microstructure characteristics and energy coupling region distribution; S23. Extract the printing parameters of the current printing layer, where the printing parameters include printing power and cooling time interval; S24. Perform spatial registration on the infrared thermal image, optical texture map, laser reflection map, and printing parameters, perform pixel-level correction, and uniformly map them to the coordinate system of the printing path map structure; S25. Construct a lattice consistency map based on the printing path map structure. In the lattice consistency map, the nodes are the lattice candidate centers corresponding to the path points, and the node attributes include crystal orientation estimation tensors, temperature gradient values, and surface stress state vectors.

[0011] Optionally, the S3 specifically includes: S31. Obtain the lattice consistency maps of the current printing layer and the previous printing layer, perform preliminary registration on the node sets of the lattice consistency maps of the current printing layer and the previous printing layer, establish an initial matching node pair set based on the spatial coordinate position relationship, and preliminarily screen out the node pairs whose distance exceeds the preset threshold ; S32. Calculate the crystal grain direction difference, temperature thermal gradient direction difference, and surface stress state similarity between the initially screened initial matching node pairs, and perform weighted fusion to obtain a comprehensive matching score for evaluating the lattice consistency between cross-layer nodes; S33. Screen the node pairs with the comprehensive matching score higher than the set threshold again, construct a cross-layer connection edge set, and the cross-layer connection edge attributes include crystal orientation difference, thermal continuity index, and historical evolution identifier; S34. Add the cross-layer connection edge set to the lattice consistency map of the current printing layer to construct an extended lattice map including a cross-layer connection structure; S35. Perform structural integration on the extended lattice map and the lattice consistency map of the previous printing layer to form a cross-layer lattice map structure; S36. Use the cross-layer lattice atlas structure as the lattice state at the current time step during the printing process to generate a cross-layer lattice atlas sequence.

[0012] Optionally, the specific steps of S4 include: S41. Input the cross-layer lattice atlas sequence into the graph neural control body model, which includes a deformation fitting perception unit, a graph-graph coupling perturbation unit, and a residual regulation network; S42. The deformation fitting perception unit extracts the lattice principal axis direction tensors of each node at different time steps to construct a set of crystal orientation evolution trajectories , where represents the lattice principal axis direction tensor of the -th node at time step ; S43. The deformation fitting perception unit uses Bessel surface fitting to solve and predict the crystal orientation tensor , and calculates the crystal orientation offset residual: ; where, represents the crystal orientation offset residual of the -th node, represents the Frobenius norm; S44. The graph-graph coupling perturbation unit embeds the crystal orientation offset residual, temperature gradient change, and stress tensor change of the node into the perturbation sensitivity vector: ; where, represents the perturbation sensitivity vector of the -th node, represents the activation function, , and represent weight matrices, represents the offset, represents the temperature gradient change of the -th node, represents the stress tensor change of the -th node; S45. The residual regulation network calculates the metallurgical perturbation score based on the perturbation feature vector : ; where, represents the metallurgical perturbation score of the -th node, , and represent control factor weights, represents the Euclidean norm; S46. Sort all nodes in descending order according to the scores and select the first nodes with metallurgical disturbance scores as the areas to be corrected. For each area to be corrected, generate a local control strategy vector using the corresponding disturbance sensitivity vector: ; wherein, represents the local control strategy vector of the th control target area, represents the control weight matrix, represents the control bias term; Aggregate the local control strategy vectors of all areas to be corrected and perform boundary smoothing processing to form the correction control strategy vector of the current printing layer.

[0013] Optionally, the S5 specifically includes: S51. Based on the correction control strategy vector, extract the spatial coordinates of all nodes in the area to be corrected, and establish a cross-layer disturbance projection relationship according to the spatial position correspondence between the printing path map structure of the current printing layer and the printing path map structure of the next printing layer; S52. Perform a cross-layer disturbance projection operation according to the cross-layer disturbance projection relationship, project the local control strategy vector corresponding to the area to be corrected in the current printing layer to the area to be corrected corresponding to the next printing layer through a spatial mapping method, and calculate the projection weight according to the spatial distance between nodes to realize the disturbance projection of the local control strategy vector of the area to be corrected corresponding to the next printing layer; S53. Perform boundary smoothing processing on the local control strategy vector after disturbance projection in the next printing layer to eliminate the control discontinuity at the disturbance boundary and form a printing path map structure of the next printing layer after disturbance optimization; S54. Generate a control intervention instruction according to the printing path map structure of the next printing layer after disturbance optimization.

[0014] Optionally, the control intervention instruction includes the spatial position of the path point, the scanning direction, the laser power adjustment amount, and the scanning timing adjustment information.

[0015] Optionally, the control rules of the control intervention instruction are as follows: If the following conditions are met: metallurgical disturbance score , change amount of the scanning path direction , laser power adjustment amount , change amount of the scanning timing , and there is no cross-layer abnormal connection of the path point in the printing path map structure, then output a standard printing control instruction to maintain the original parameters; wherein Represents the set metallurgical disturbance score threshold, Represents the set threshold for the change in the scanning path direction, Represents the set threshold for the adjustment amount of the laser power, Represents the set threshold for the change in the scanning time sequence; If the metallurgical disturbance score is satisfied , when there is , , Any one of the above situations, and the path structure where it is located does not form a cross-layer continuous disturbance area, then output the corresponding single-parameter fine-tuning type control intervention instruction, and only adjust one of the path direction, laser power or scanning time sequence; If the metallurgical disturbance score is satisfied , when there is , , Two or more of the above situations, and the path structure where it is located belongs to the node group of the disturbance connected area in the mapping relationship, then output the multi-parameter joint control intervention instruction, and synchronously modify the path direction, laser power and scanning time sequence.

[0016] The beneficial effects of the present invention are: First of all, different from the traditional method that relies on static parameter tuning and single-layer monitoring, the present invention accurately captures multi-physical field parameters such as the grain direction estimation, temperature gradient distribution, and stress state in the printing path area by constructing the lattice consistency map of each printing layer, and maps them to the path map structure to accurately express the local metallurgical behavior in the form of a graph structure. This map is not only used for single-layer analysis, but also establishes a structure alignment relationship between layers to generate a cross-layer lattice map sequence, realizing the modeling and storage of the lattice continuity evolution process for the first time, and making up for the defect that the prior art cannot express the microscopic evolution trend.

[0017] Secondly, the graph neural control body model proposed by the present invention integrates a deformation fitting perception unit, a graph coupling perturbation mechanism and a residual scoring network, and can process the complex structure dependence relationship in the lattice map sequence in an end-to-end manner. By extracting the perturbation sensitivity vector and calculating the metallurgical disturbance score, the present invention can automatically identify the control key path units that may occur crystal orientation mismatch or thermal-stress mutation, and output targeted local control strategies, including multi-dimensional control means such as scanning direction adjustment, laser power compensation and path order optimization. This strategy generation method starting from the graph structure breaks through the limitations of the traditional image classification or path heuristic correction, and truly realizes the precise feedback and intervention for the tissue evolution process.

[0018] In addition, to ensure that the control strategy has feedforward regulation capabilities, the present invention further designs a disturbance cross-layer projection mechanism, which projects the abnormal areas in the current printing layer onto the path structure of the next printing layer through spatial mapping, pre-setting intervention operations before printing, and enhancing the foresight and stability of metallurgical continuity control. Through the fusion output of the control strategy vector, control intervention instructions can be dynamically generated to guide the real-time update of the subsequent path graph structure and printing parameters, constructing a closed-loop control system for the printing hierarchy driven by data.

[0019] Finally, in combination with the process of reconstructing the lattice map layer by layer, the present invention establishes a complete sequence of lattice map evolution records, providing a long-term and effective data basis for printing quality traceability, model retraining, and process self-adaptation, and significantly enhancing the micro-metallurgical bonding strength and overall structural reliability during the 3D printing of key components of aerospace engines. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 is the overall flowchart of a method for enhancing interlayer metallurgical bonding in 3D printing of aerospace engine components proposed by the present invention; Figure 2 is the flowchart of graph structure alignment and cross-layer connection construction for a method for enhancing interlayer metallurgical bonding in 3D printing of aerospace engine components proposed by the present invention; Figure 3 is the structural schematic diagram of the graph neural control body model for a method for enhancing interlayer metallurgical bonding in 3D printing of aerospace engine components proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0022] Refer to Figures 1 - 3 , a method for enhancing interlayer metallurgical bonding in 3D printing of aerospace engine components, comprising the following steps: S1. Perform slicing processing on the 3D model of the aerospace engine component to generate the printing path data of each printing layer and construct a printing path graph structure; S2. Collect the infrared thermal image, optical texture image, laser reflection image, and printing parameters of the current printing layer, and construct a corresponding lattice consistency map in combination with the printing path graph structure; S3. Align the lattice consistency maps of the current printing layer and the previous printing layer in terms of graph structure, establish a cross-layer connection edge set, and generate a cross-layer lattice map sequence; S4, inputting the cross-layer lattice graph sequence into the graph neural control body model, obtaining the disturbance sensitivity vector and disturbance score value of each node, and outputting the correction control strategy vector of the current printing layer; S5, based on the correction control strategy vector, performing a cross-layer disturbance projection operation, projecting the area to be corrected in the current printing layer into the cross-layer printing path graph structure by spatial mapping, and generating a control intervention instruction; S6. Based on the control intervention instruction, steps S2 to S5 are repeatedly executed to iterate the lattice consistency maps of all printed layers layer by layer, and finally form a complete lattice map evolution record sequence.

[0023] This method proposes an additive manufacturing process control mechanism that combines graph structure modeling and disturbance perception control. The printing path graph structure is generated by slicing layer by layer, and after each layer is completed, the lattice consistency map is constructed by combining multimodal printing data. Then, cross-layer connections are established between the maps to generate a complete lattice evolution sequence. The evolution state of the graph nodes and edges is modeled by the graph neural control body model to achieve the identification of key disturbance areas and the output of control strategies, and disturbance projection and parameter update are performed in subsequent printing layers, finally achieving closed-loop control of the entire printing process. This scheme not only improves the metallurgical continuity of aerospace engine components during the layer-by-layer construction process, but also significantly enhances the adaptability of the printing process to the evolution of the microstructure. While ensuring metallurgical quality, it also takes into account the real-time adjustment capabilities of the forming path optimization and control instructions.

[0024] In this embodiment, the S1 specifically includes the spatial coordinate information, scanning order, scanning direction and scanning speed of each layer of printing path points, and constructs a printing path graph structure, in which nodes represent path points, edges represent the printing order relationship between adjacent path points, and edge attributes include the scanning direction vector of the path segment, the geometric distance between the starting and ending points, and the scanning speed gradient.

[0025] In the path data construction stage, the printing points of each layer are abstracted as graph nodes, and the topological edge structure is constructed through sequence and spatial relationship, which not only retains the structural information of the printing trajectory, but also provides a basis for the subsequent spatial alignment with thermal field images and lattice estimation results. By introducing edge attributes such as scanning direction vector, path segment distance and velocity gradient, the path graph has the ability to express parameter sensitivity, making subsequent mapping and control more physically interpretable. The expression of this graph structure opens up the connection channel between spatial position and printing parameters, laying the foundation for the modeling of micro-grain behavior and the targeted output of control strategies, and improving the engineering expressiveness and controllability of the graph data structure in the entire additive manufacturing process.

[0026] In this implementation manner, S2 specifically includes: S21. Collect the infrared thermal image corresponding to the current printing layer to obtain the two-dimensional image information of the thermal field distribution during the printing process; S22. Collect the optical texture map and laser reflection map of the current printing layer to obtain the surface microstructure characteristics and the distribution of the energy coupling region; S23. Extract the printing parameters of the current printing layer, where the printing parameters include printing power and cooling time interval; S24. Perform spatial registration on the infrared thermal image, optical texture map, laser reflection map and printing parameters, perform pixel-level correction, and uniformly map them to the coordinate system of the printing path map structure; S25. Construct a lattice consistency map based on the printing path map structure. The nodes in the lattice consistency map are the lattice candidate centers of the regions corresponding to the path points, and the node attributes include crystal orientation estimation tensors, temperature gradient values and surface stress state vectors.

[0027] During the process of constructing the lattice consistency map, the hierarchical thermal field information is obtained through the infrared thermal image, the microscopic surface topography is obtained through the optical texture map, the energy coupling region is captured through the laser reflection map, and then combined with the printing parameters to jointly act on the construction of the map node features, realizing the multi-source perception of the printing state. After the image registration is completed in the unified coordinate system, the map nodes can accurately correspond to the lattice candidate structure centers within the printing path point regions, and assign them attributes such as crystal orientation estimation, temperature gradient and stress state. This construction method significantly improves the fitting degree of the map to the true state of the lattice, and provides the basic data structure for subsequent cross-layer structure alignment and perturbation analysis, with clear physical interpretability and executability.

[0028] In this embodiment, the specific steps of S3 are as follows: S31. Obtain the lattice consistency maps of the current printing layer and the previous printing layer, perform preliminary registration on the node sets of the lattice consistency maps of the current printing layer and the previous printing layer, establish an initial matching node pair set according to the spatial coordinate position relationship, and preliminarily screen out the node pairs whose distance exceeds the preset threshold ; S32. Calculate the crystal grain direction difference, temperature thermal gradient direction difference and surface stress state similarity between the initially matched node pairs after preliminary screening, and perform weighted fusion to obtain a comprehensive matching score for evaluating the lattice consistency between cross-layer nodes; S33. Screen the node pairs with the comprehensive matching score higher than the set threshold again, and construct a cross-layer connection edge set. The attributes of the cross-layer connection edges include crystal orientation difference, thermal continuity index and historical evolution identifier; S34. Add the cross-layer connection edge set to the lattice consistency map of the current printing layer to construct an extended lattice map including cross-layer connection structures; S35. Integrate the extended lattice map and the lattice consistency map of the previous printing layer structurally to form a cross-layer lattice map structure; S36. Use the cross-layer lattice map structure as the lattice state at the current time step during the printing process to generate a cross-layer lattice map sequence.

[0029] For the lattice consistency maps between different printing layers, potential cross-layer node connection relationships are screened out through preliminary coordinate registration and physical property similarity scoring, and then edge connections are constructed based on the matching score and the thermal-mechanical continuity index, thereby establishing a sequence of graph structures with directionality and evolution properties. While maintaining the spatial topological continuity, this sequence truly reflects the transfer paths of lattice direction offsets and thermal coupling perturbations between layers. By cross-layer expanding and integrating each layer's map into a time series sequence, it can provide structured time series input for the subsequent graph neural control volume model, effectively enhancing the prediction ability for the development trend of metallurgical deviations, and featuring strong cross-layer structural continuity and high dynamic evolution expression ability.

[0030] In this embodiment, the specific steps of S4 are as follows: S41. Input the cross-layer lattice map sequence into the graph neural control volume model, which includes a deformation fitting perception unit, a graph-graph coupling perturbation unit, and a residual regulation network; S42. The deformation fitting perception unit extracts the lattice principal axis direction tensors of each node at different time steps to construct a set of crystal orientation evolution trajectories , where represents the lattice principal axis direction tensor of the th node at the time step ; S43. The deformation fitting perception unit uses B-spline surface fitting to solve and predict the crystal orientation tensor , and calculates the crystal orientation offset residual: ; where represents the crystal orientation offset residual of the th node, represents the Frobenius norm; S44. The graph-graph coupling perturbation unit embeds the crystal orientation offset residual, temperature gradient change, and stress tensor change of the node into the perturbation sensitivity vector: ; where represents the perturbation sensitivity vector of the th node, represents the activation function, , and represent weight matrices, Indicates the offset, represents the temperature gradient change of the th node, represents the stress tensor change of the th node; S45. The residual regulation network is based on the perturbation eigenvector , and calculates the metallurgical perturbation score: ; wherein, represents the metallurgical perturbation score of the th node, , and represent the control factor weights, represents the Euclidean norm; S46. Sort all nodes in descending order according to the score , and select the first nodes with the metallurgical perturbation score as the area to be corrected, and generate a local control strategy vector for each area to be corrected using the corresponding perturbation sensitivity vector: ; wherein, represents the local control strategy vector of the th control target area, represents the control weight matrix, represents the control bias term; Aggregate the local control strategy vectors of all areas to be corrected, and through boundary smoothing processing, form the correction control strategy vector of the current printing layer.

[0031] With the help of the graph neural control body model, the input lattice map sequence is transformed into the perturbation score and strategy generation results of each path control unit. Calculate the crystal orientation offset residual through the deformation fitting perception unit, combine the thermal gradient and stress tensor changes, and construct a multi-dimensional perturbation eigenvector. After screening by the perturbation scoring function, the key control areas are identified, and the scanning angle perturbation, power adjustment and sequence offset values are output using the control transformation function, and aggregated to form a set of control strategy vectors for the printing layer. This strategy has the advantages of strong physical feature perception ability, good spatial distribution adaptability, and fine control granularity, realizes the closed-loop mapping from physical data-driven to path parameter optimization, and significantly enhances the organization control ability of the printing process.

[0032] In this embodiment, the S5 specifically includes: S51. Extract the spatial coordinates of all nodes within the area to be corrected based on the corrected control strategy vector, and establish a cross-layer perturbation projection relationship according to the spatial position correspondence between the printing path map structure of the current printing layer and that of the next printing layer; S52. Perform a cross-layer perturbation projection operation based on the cross-layer perturbation projection relationship, project the local control strategy vector corresponding to the area to be corrected in the current printing layer onto the corresponding area to be corrected in the next printing layer through a spatial mapping method, and calculate the projection weight based on the spatial distance between nodes to achieve the perturbation projection of the local control strategy vector for the area to be corrected in the next printing layer; S53. Perform boundary smoothing processing on the perturbed projected local control strategy vector in the next printing layer to eliminate the control discontinuity at the perturbation boundary, and form a printing path map structure for the next printing layer after perturbation optimization; S54. Generate a control intervention instruction according to the printing path map structure of the next printing layer after perturbation optimization.

[0033] For the generated corrected control strategy vector, the cross-layer intervention control is transmitted through the perturbation area recognition and path point spatial projection mechanism. This mechanism establishes a spatial mapping relationship between the spatial position and control vector of the key area in the current printing layer and the path structure of the next printing layer, and projects the control instruction to the area to be corrected in the target layer to achieve the feedforward transmission and influence continuation of the control strategy. Subsequently, combined with the boundary smoothing processing and local control fusion mechanism, the continuity and stability of the control adjustment are ensured. Finally, these perturbation results are encapsulated into control intervention instructions for direct parsing and execution by the printing control system, significantly improving the predictability and control accuracy of printing parameter adjustment.

[0034] In this embodiment, the control intervention instruction includes the spatial position of the path point, the scanning direction, the laser power adjustment amount, and the scanning timing adjustment information.

[0035] In this embodiment, the control rules of the control intervention instruction are as follows: If the following conditions are met: metallurgical perturbation score , change amount of scanning path direction , laser power adjustment amount , change amount of scanning timing , and there is no cross-layer abnormal connection of the path point in the printing path map structure, then output a standard printing control instruction to maintain the original parameters; Where represents the set metallurgical perturbation score threshold, represents the set change amount threshold of the scanning path direction, represents the set laser power adjustment amount threshold, represents the set change amount threshold of the scanning timing; If the metallurgical disturbance score is satisfied , when any one of the following situations occurs , , and the path structure does not form a cross-layer continuous disturbance region, a single-parameter fine-tuning type control intervention instruction is output, and only one of the path direction, laser power, or scanning timing is adjusted; If the metallurgical disturbance score is satisfied , when two or more of the following situations occur , , and the path structure belongs to the node group of the disturbance connected region in the mapping relationship, a multi-parameter joint control intervention instruction is output to synchronously modify the path direction, laser power, and scanning timing.

[0036] Example 1

[0037] To verify the feasibility of the present invention in implementation, the present invention is applied to the 3D printing process of a certain type of high thrust-to-weight ratio aerospace engine nozzle assembly. The nozzle assembly is made of Inconel 718 superalloy material and is constructed layer by layer through the selective laser melting process. The overall structure has extremely high interlayer thermal stress concentration and microstructural grain boundary mismatch risks. The interlayer metallurgical defect rate under the traditional process always remains between 2% and 3%, seriously restricting the achievement of the nozzle life index. In this example, the middle contraction section with a large length-to-diameter ratio and severe geometric mutation in the nozzle structure is selected as the key area, and the method of the present invention is deployed in this area to perform the whole process of lattice consistency map construction, cross-layer graph structure alignment, disturbance perception modeling, and printing path closed-loop adjustment.

[0038] During the printing process, the system first collects the infrared thermal image, optical texture image, and laser reflection image after each layer is printed, and constructs a lattice consistency map by combining the position information, power, and speed parameters of each scanning path point. Subsequently, this map is aligned with the map of the previous printed layer for structure alignment and cross-layer connection, and a map sequence is generated and input into the graph neural control body model for disturbance sensitivity evaluation. After the first layer is printed, the system identifies an abnormal area with an average crystal orientation deviation residual reaching 3.1 degrees and a temperature gradient mutation reaching 18.4 K / mm, and the disturbance score is as high as 0.85. Based on this, the system adjusts the scanning angle, reduces the power, and locally delays the scanning timing, generates a control intervention instruction, and then executes the printing. During the subsequent iterative process from the second layer to the fifth layer, the system continuously constructs the lattice map and outputs the strategy. The response of the disturbance area gradually decreases, the control adjustment amplitude decreases, and the metallurgical state tends to be stable.

[0039] In this embodiment, with the dynamic execution of the control strategy, the crystal orientation residual gradually decreases from 3.1° to 2.1°, the temperature gradient perturbation decreases from 18.4 K / mm to 14.2 K / mm, and the range of change in the stress tensor also becomes smooth accordingly. Eventually, the metallurgical defect rate drops significantly from the initial 2.5% to 0.9%. The average adjustment amount of the control instruction is reduced from the initial 1.00 to 0.68, indicating that the system can already adaptively regulate complex lattice changes, avoid unnecessary intervention, and ensure metallurgical consistency at the same time. After printing is completed, X-ray tomography is used to verify the continuity of the grain boundaries. No obvious microcracks or interlayer pores are found inside the nozzle contraction section, and the surface quality and grain structure show good consistency, meeting the conditions for direct assembly and application.

[0040] This embodiment effectively proves the adaptability and control accuracy of the method of the present invention in the printing of high-complexity aerospace engine parts. The map-driven intervention control mechanism implemented based on lattice-level perception breaks through the bottlenecks of traditional static parameter setting and image monitoring, not only improving the microscopic metallurgical quality, but also bringing new capabilities of controllable structure, adjustable path, and traceable evolution to the aerospace-level printing process, and has high engineering application and promotion value.

[0041] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered by the protection scope of the present invention.

Claims

1. A method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines, characterized in that, The steps are as follows: S1. Slice the 3D model of the aerospace engine component to generate the printing path data for each printing layer, and construct a printing path graph structure; S2. Collect the infrared thermal image, optical texture image, laser reflection image and printing parameters of the current printing layer, and construct a corresponding lattice consistency map in combination with the printing path graph structure; S3. Align the lattice consistency maps of the current printing layer and the previous printing layer in terms of graph structure, establish a cross-layer connection edge set, and generate a cross-layer lattice map sequence; S4. Input the cross-layer lattice map sequence into the graph neural control body model, obtain the perturbation sensitivity vector and perturbation score value of each node, and output the correction control strategy vector of the current printing layer; S5. Based on the correction control strategy vector, perform a cross-layer perturbation projection operation, project the area to be corrected in the current printing layer to the cross-layer printing path graph structure through spatial mapping, and generate a control intervention instruction; S6. Based on the control intervention instruction, repeat steps S2 to S5 layer by layer to complete the lattice consistency maps of all printing layers through iterative processing, and finally form a complete lattice map evolution record sequence.

2. The method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines according to claim 1, characterized in that, The S1 specifically includes the spatial coordinate information, scanning order, scanning direction and scanning speed of each printing path point, and constructs a printing path graph structure, where the nodes represent path points, the edges represent the printing order relationship between adjacent path points, and the edge attributes include the scanning direction vector of the path segment, the geometric distance between the start and end points, and the scanning speed gradient.

3. The method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines according to claim 1, characterized in that, The S2 specifically includes: S21. Collect the infrared thermal image corresponding to the current printing layer to obtain the two-dimensional image information of the thermal field distribution during the printing process; S22. Collect the optical texture image and laser reflection image of the current printing layer to obtain the surface microstructure characteristics and the distribution of the energy coupling region; S23. Extract the printing parameters of the current printing layer, and the printing parameters include the printing power and the cooling time interval; S24. Perform spatial registration on the infrared thermal image, optical texture image, laser reflection image and printing parameters, perform pixel-level correction, and uniformly map them to the coordinate system of the printing path graph structure; S25. Construct a lattice consistency map based on the printing path graph structure. In the lattice consistency map, the nodes are the lattice candidate centers of the regions corresponding to the path points, and the node attributes include the crystal orientation estimation tensor, the temperature gradient value and the surface stress state vector.

4. The method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines according to claim 1, characterized in that, The S3 specifically includes: S31. Obtain the lattice consistency map of the current printing layer and the previous printing layer, perform preliminary registration on the node sets of the lattice consistency maps of the current printing layer and the previous printing layer, establish an initial set of matching node pairs based on the spatial coordinate position relationship, and preliminarily screen out the node pairs whose distance exceeds the preset threshold ; S32. Calculate the grain direction difference, temperature thermal gradient direction difference and surface stress state similarity between the initially screened initial matching node pairs, and perform weighted fusion to obtain a comprehensive matching score for evaluating the lattice consistency between cross-layer nodes; S33. Screen the node pairs with the comprehensive matching score higher than the set threshold again to construct a cross-layer connection edge set, and the cross-layer connection edge attributes include the crystal orientation difference, the thermal continuity index, and the historical evolution identifier; S34. Add the cross-layer connection edge set to the lattice consistency map of the current printing layer to construct an extended lattice map including a cross-layer connection structure; S35. Integrate the structure of the extended lattice map and the lattice consistency map of the previous printing layer to form a cross-layer lattice map structure; S36. Use the cross-layer lattice map structure as the lattice state at the current time step during the printing process to generate a cross-layer lattice map sequence.

5. The method for enhancing the interlayer metallurgical bonding in 3D printing of aerospace engines according to claim 1, characterized in that, The S4 specifically includes: S41. Input the cross-layer lattice map sequence into the graph neural control body model, where the graph neural control body model includes a deformation fitting perception unit, a graph-graph coupling perturbation unit, and a residual regulation network; S42. The deformation fitting perception unit extracts the lattice principal axis direction tensors of each node at different time steps, and constructs a set of crystal orientation evolution trajectories , where represents the lattice principal axis direction tensor of the -th node at the time step ; S43. The deformation fitting perception unit uses Bessel surface fitting to solve and predict the crystal orientation tensor , and calculates the crystal orientation offset residual: ； Among them, represents the crystal orientation offset residual of the th node, represents the Frobenius norm; S44. The graph-graph coupling perturbation unit embeds the crystal orientation offset residuals, temperature gradient changes, and stress tensor changes of nodes into a perturbation sensitivity vector: ； Among them, represents the perturbation sensitivity vector of the th node, represents the activation function, , and represent the weight matrices, represents the offset, represents the change in temperature gradient of the th node, represents the change in stress tensor of the th node; S45. The residual regulation network calculates a metallurgical disturbance score based on the disturbance feature vector , and ； Among them, represents the metallurgical disturbance score of the th node, , and represent the control factor weights, represents the Euclidean norm; S46. Sort all nodes in descending order according to the score and select the first metallurgical disturbance scoring nodes as the area to be corrected. For each area to be corrected, generate a local control strategy vector using the corresponding disturbance sensitivity vector: ； Among them, represents the local control strategy vector of the th control target area, represents the control weight matrix, represents the control bias term; Aggregate the local control strategy vectors of all regions to be corrected, and through boundary smoothing, form the corrected control strategy vector of the current printing layer.

6. A method for enhancing the interlayer metallurgical bonding in 3D printing of a space engine according to claim 1, wherein, The specific steps of S5 are as follows: S51. Based on the corrected control strategy vector, extract the spatial coordinates of all nodes within the region to be corrected, and establish a cross-layer perturbation projection relationship according to the spatial position correspondence between the printing path graph structure of the current printing layer and the printing path graph structure of the next printing layer; S52. Perform a cross-layer perturbation projection operation according to the cross-layer perturbation projection relationship, project the local control strategy vector corresponding to the region to be corrected in the current printing layer onto the corresponding region to be corrected in the next printing layer through a spatial mapping method, and calculate the projection weight according to the spatial distance between nodes to achieve the perturbation projection of the local control strategy vector of the corresponding region to be corrected in the next printing layer; S53. Perform boundary smoothing on the perturbed projected local control strategy vector in the next printing layer to eliminate the control discontinuity at the perturbation boundary, and form the printing path graph structure of the next printing layer after perturbation optimization; S54. Generate a control intervention instruction according to the printing path graph structure of the next printing layer after perturbation optimization.

7. A method for enhancing the interlayer metallurgical bonding in 3D printing of a space engine according to claim 1, wherein, The control intervention instruction includes the spatial position of the path point, the scanning direction, the laser power adjustment amount, and the scanning timing adjustment information.

8. A method for enhancing the interlayer metallurgical bonding in 3D printing of a space engine according to claim 1, wherein, The control rules of the control intervention instruction are as follows: If the following conditions are met: metallurgical disturbance score , the change amount of the scanning path direction , the adjustment amount of the laser power , the change amount of the scanning time sequence , and there is no cross-layer abnormal connection of the path points in the printing path diagram structure, then a standard printing control instruction to maintain the original parameters is output; wherein represents the set metallurgical disturbance score threshold, represents the set threshold of the change amount of the scanning path direction, represents the set threshold of the laser power adjustment amount, represents the set threshold of the change amount of the scanning time sequence; If the metallurgical disturbance score is satisfied , when any one of the following occurs , , , and the path structure does not form a cross-layer continuous disturbance area, then output the corresponding single-parameter fine-tuning type control intervention instruction, and only adjust one of the path direction, laser power, or scanning timing; If the metallurgical disturbance score is satisfied , when there are , , and two or more of these situations occur, and the path structure is a node group in the disturbance connected area of the mapping relationship, then output a multi-parameter joint control intervention instruction to synchronously modify the path direction, laser power, and scanning timing.

Images