A method and system for controlling laser hybrid welding

Abstract

This invention provides a control method and system for laser hybrid welding. The invention relates to the field of laser hybrid welding control technology. The method includes acquiring the original welding process parameters for laser hybrid welding and CT imaging of the titanium alloy dental implant abutment blank after preliminary welding; determining multiple preliminary porosity suppression welding schemes based on the original welding process parameters, the three-dimensional morphology data of the weld after preliminary welding of the abutment blank, and micropore characteristic information; acquiring CT imaging of each preliminary porosity suppression welding scheme after welding; determining a target porosity suppression welding scheme based on the CT imaging of each preliminary porosity suppression welding scheme; and performing laser hybrid welding on the remaining initial titanium alloy dental implant abutment blanks based on the target porosity suppression welding scheme. This method can efficiently and accurately determine the optimal porosity suppression welding scheme for titanium alloy dental implant abutment blanks.

Description

Technical Field

[0001] This invention relates to the field of laser hybrid welding control technology, and specifically to a control method and system for laser hybrid welding. Background Technology

[0002] Laser-assisted composite welding technology, as an advanced precision joining method, is crucial for ensuring the structural integrity and dimensional accuracy of titanium alloy dental implant abutments. In the practical application of laser-assisted composite welding of titanium alloy dental implant abutment blanks, micropores easily form in the welding area. These micropores are tiny cavities within the weld seam, which significantly degrade the mechanical properties of the weld joint, such as reducing its fatigue strength, tensile strength, and toughness. For dental implant abutments, which are small and precision components that need to serve in the complex stress environment of the oral cavity over a long period, the presence of internal micropores is not only a potential stress concentration point and could become the origin of fatigue cracks, but it can also affect their long-term reliability and safety. Currently, traditional methods of porosity suppression welding mainly rely on optimizing single welding parameters, such as adjusting laser power, welding speed, or shielding gas flow rate. However, these methods have significant limitations; the porosity suppression optimization process heavily depends on the operator's subjective experience and lacks objective, quantitative guidance. This is not only time-consuming and labor-intensive, but also involves high trial-and-error costs and low efficiency. Traditional porosity suppression welding methods have unstable porosity suppression effects, making it difficult to effectively guarantee the yield rate. Especially for active metals like titanium alloys, which are very sensitive to the welding environment and have a complex porosity formation mechanism, it is more difficult to suppress porosity through conventional experience.

[0003] Therefore, how to efficiently and accurately determine the optimal porosity suppression welding scheme for titanium alloy dental implant abutment blanks is a problem that urgently needs to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to efficiently and accurately determine the optimal porosity suppression welding scheme for titanium alloy dental implant abutment blanks.

[0005] According to a first aspect, the present invention provides a control method for laser hybrid welding, comprising: acquiring original welding process parameters for laser hybrid welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; determining three-dimensional morphological data and micropore feature information of the weld seam after preliminary welding of the abutment blank based on the CT imaging of the titanium alloy dental implant abutment blank after preliminary welding; and determining multiple preliminary porosity suppression welding schemes based on the original welding process parameters for laser hybrid welding, the three-dimensional morphological data of the weld seam after preliminary welding of the abutment blank, and the micropore feature information, thereby performing porosity suppression welding. The welding scheme includes the mixing ratio of high-purity argon and nitrogen in laser hybrid welding and the oscillation amplitude of the laser beam; based on the multiple preliminary porosity suppression welding schemes, laser hybrid welding is performed on multiple initial titanium alloy dental implant abutment blanks, and CT images of each preliminary porosity suppression welding scheme are obtained after welding; based on the CT images of each preliminary porosity suppression welding scheme, a target porosity suppression welding scheme is determined; based on the target porosity suppression welding scheme, the remaining initial titanium alloy dental implant abutment blanks are laser hybrid welded to produce qualified titanium alloy dental implant abutment finished products.

[0006] In one possible implementation, based on the original welding process parameters of the laser-hybrid welding, the three-dimensional morphological data of the weld after the initial welding of the base blank, and the micropore feature information, multiple preliminary porosity suppression welding schemes are determined. The porosity suppression welding schemes include the mixing ratio of high-purity argon and nitrogen in the laser-hybrid welding and the laser beam oscillation amplitude. This includes: determining multiple significant porosity point information based on the three-dimensional morphological data of the weld after the initial welding of the base blank and the micropore feature information; clustering the multiple significant porosity point information to obtain K clusters; determining multiple significant porosity regions based on the K clusters; determining the high-purity argon and nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region based on the multiple significant porosity regions and the original welding process parameters of the laser-hybrid welding; and determining multiple preliminary porosity suppression welding schemes based on the high-purity argon and nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region.

[0007] In one possible implementation, determining the target porosity suppression welding scheme based on the CT image after welding each initially selected porosity suppression welding scheme includes: constructing a porosity suppression welding atlas, which includes multiple initially selected porosity suppression nodes and multiple edges between the nodes. The node features of each initially selected porosity suppression node include an initially selected porosity suppression welding scheme and a CT image after welding the initially selected porosity suppression welding scheme. The edges between the nodes represent the similarity between the initially selected porosity suppression welding schemes. The target porosity suppression welding scheme is determined by processing the porosity suppression welding atlas based on a graph neural network.

[0008] In one possible implementation, the stomatal feature information includes stomatal number and morphological information, stomatal distribution coordinates, and stomatal volume percentage.

[0009] According to a second aspect, the present invention provides a control system for laser hybrid welding, comprising: an acquisition module for acquiring original welding process parameters for laser hybrid welding and CT images of a titanium alloy dental implant abutment blank after preliminary welding; a data analysis module for determining, based on the CT images of the titanium alloy dental implant abutment blank after preliminary welding, the three-dimensional morphological data of the weld seam and micropore feature information of the abutment blank after preliminary welding; and a scheme generation module for determining, based on the original welding process parameters for laser hybrid welding, the three-dimensional morphological data of the weld seam of the abutment blank after preliminary welding, and the micropore feature information, a plurality of preliminary porosity suppression welding schemes, wherein the porosity suppression welding... The scheme includes: a high-purity argon and nitrogen mixing ratio for laser hybrid welding and a laser beam oscillation amplitude; a welding testing module for performing laser hybrid welding on multiple initial titanium alloy dental implant abutment blanks based on the multiple preliminary porosity suppression welding schemes, and acquiring CT images of each preliminary porosity suppression welding scheme after welding; a scheme optimization module for determining a target porosity suppression welding scheme based on the CT images of each preliminary porosity suppression welding scheme after welding; and a finished product production module for performing laser hybrid welding on the remaining initial titanium alloy dental implant abutment blanks based on the target porosity suppression welding scheme to produce qualified finished titanium alloy dental implant abutments.

[0010] In one possible implementation, the scheme generation module is further configured to: determine multiple significant porosity point information based on the three-dimensional morphology data of the weld after preliminary welding of the base blank and the micropore feature information; cluster the multiple significant porosity point information to obtain K clusters; determine multiple significant porosity regions based on the K clusters; determine the high-purity argon and nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region based on the multiple significant porosity regions and the original welding process parameters of the laser hybrid welding; and determine multiple preliminary porosity suppression welding schemes based on the high-purity argon and nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region.

[0011] In one possible implementation, the scheme optimization module is further configured to: construct a porosity suppression welding atlas, the porosity suppression welding atlas including multiple initially selected porosity suppression nodes and multiple edges between the nodes, each initially selected porosity suppression node having node features including an initially selected porosity suppression welding scheme and a CT image of the initially selected porosity suppression welding scheme after welding, the edges between the nodes representing the similarity between the initially selected porosity suppression welding schemes; and process the porosity suppression welding atlas based on a graph neural network to determine a target porosity suppression welding scheme.

[0012] In one possible implementation, the stomatal feature information includes stomatal number and morphological information, stomatal distribution coordinates, and stomatal volume percentage.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring original welding process parameters for laser hybrid welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; determining three-dimensional morphological data and micropore feature information of the weld seam after preliminary welding of the abutment blank based on the CT imaging of the titanium alloy dental implant abutment blank after preliminary welding; and determining three-dimensional morphological data and micropore feature information of the weld seam after preliminary welding of the abutment blank based on the original welding process parameters for laser hybrid welding and the three-dimensional morphological data of the weld seam after preliminary welding of the abutment blank. Based on the micropore characteristic information, multiple preliminary porosity suppression welding schemes are determined. These schemes include the mixing ratio of high-purity argon and nitrogen in laser-assisted composite welding and the laser beam oscillation amplitude. Based on these preliminary porosity suppression welding schemes, multiple initial titanium alloy dental implant abutment blanks are subjected to laser-assisted composite welding, and CT images of each preliminary porosity suppression welding scheme after welding are obtained. Based on the CT images of each preliminary porosity suppression welding scheme after welding, a target porosity suppression welding scheme is determined. Based on the target porosity suppression welding scheme, the remaining initial titanium alloy dental implant abutment blanks are subjected to laser-assisted composite welding to produce qualified finished titanium alloy dental implant abutments.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned control method for laser hybrid welding. The method includes: acquiring the original welding process parameters for laser hybrid welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; determining the three-dimensional morphological data of the weld seam and micropore feature information of the abutment blank after preliminary welding based on the CT imaging of the titanium alloy dental implant abutment blank after preliminary welding; and determining the three-dimensional morphological data of the weld seam and micropore feature information of the abutment blank after preliminary welding based on the original welding process parameters for laser hybrid welding, the three-dimensional morphological data of the weld seam after preliminary welding of the abutment blank, and the micropore feature information. Information was collected to determine multiple preliminary porosity suppression welding schemes, including the mixing ratio of high-purity argon and nitrogen in laser hybrid welding and the laser beam oscillation amplitude. Based on the multiple preliminary porosity suppression welding schemes, multiple initial titanium alloy dental implant abutment blanks were laser hybrid welded, and CT images of each preliminary porosity suppression welding scheme after welding were obtained. Based on the CT images of each preliminary porosity suppression welding scheme after welding, a target porosity suppression welding scheme was determined. Based on the target porosity suppression welding scheme, the remaining initial titanium alloy dental implant abutment blanks were laser hybrid welded to produce qualified finished titanium alloy dental implant abutments.

[0015] This invention provides a control method and system for laser hybrid welding. The method includes acquiring the original welding process parameters for laser hybrid welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; determining the three-dimensional morphological data and micropore feature information of the weld after preliminary welding of the abutment blank based on the CT imaging of the abutment blank; and determining multiple preliminary porosity suppression welding schemes based on the original welding process parameters, the three-dimensional morphological data of the weld after preliminary welding of the abutment blank, and the micropore feature information. The porosity suppression welding schemes include a mixture of high-purity argon and nitrogen for laser hybrid welding. The method involves several steps: First, laser composite welding is performed on multiple initial titanium alloy dental implant abutment blanks based on the selected porosity suppression welding schemes, and CT images of each initial porosity suppression welding scheme are obtained after welding. Then, a target porosity suppression welding scheme is determined based on the CT images of each initial porosity suppression welding scheme. Finally, the remaining initial titanium alloy dental implant abutment blanks are laser composite welded based on the target porosity suppression welding scheme to produce qualified finished titanium alloy dental implant abutment products. This method can efficiently and accurately determine the optimal porosity suppression welding scheme for titanium alloy dental implant abutment blanks. Attached Figure Description

[0016] Figure 1 A schematic flowchart of a laser composite welding control method provided in an embodiment of the present invention;

[0017] Figure 2A schematic diagram of a titanium alloy dental implant abutment blank provided in an embodiment of the present invention;

[0018] Figure 3 A schematic flowchart illustrating a method for determining multiple preliminary porosity suppression welding schemes, provided in an embodiment of the present invention;

[0019] Figure 4 This is a flowchart illustrating a welding scheme for determining target porosity suppression, provided by an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of a laser-based composite welding control system provided in an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0022] In this embodiment of the invention, the following are provided: Figure 1 The method for controlling laser hybrid welding, as shown, includes steps S1 to S6:

[0023] Step S1: Obtain the original welding process parameters of laser hybrid welding and CT imaging of the titanium alloy dental implant abutment blank after preliminary welding.

[0024] The initial welding process parameters for laser hybrid welding are the technical parameters set at the beginning of the laser hybrid welding process. These parameters include laser power, welding speed, defocusing amount, and shielding gas flow rate.

[0025] Titanium alloy dental implant abutment blanks are the initial workpieces used to process titanium alloy dental implant abutments. The titanium alloy dental implant abutment blanks are made of titanium alloy material and have a basic structural shape that is adapted to dental implants. Figure 2 This is a schematic diagram of a titanium alloy dental implant abutment blank provided in an embodiment of the present invention.

[0026] CT imaging of titanium alloy dental implant abutment blanks after preliminary welding refers to the preliminary welding operation performed on titanium alloy dental implant abutment blanks without specific measures to suppress microporosity. After the welding operation is completed, the blank is fully scanned by a CT scanning device, and the final image generated can show the structural state of the blank inside and the weld area after preliminary welding.

[0027] Step S2: Based on the CT imaging of the titanium alloy dental implant abutment blank after preliminary welding, determine the three-dimensional morphological data and micropore feature information of the weld after preliminary welding of the abutment blank.

[0028] In some embodiments, an imaging analysis model can be used to determine the three-dimensional morphological data and micropore feature information of the weld seam after preliminary welding of the abutment blank. The imaging analysis model is a convolutional neural network model. The input to the imaging analysis model is a CT image of the titanium alloy dental implant abutment blank after preliminary welding. The output of the imaging analysis model is the three-dimensional morphological data and micropore feature information of the weld seam after preliminary welding of the abutment blank.

[0029] Convolutional Neural Network (CNN) models are a type of feedforward neural network that incorporates convolutional computations and has a deep structure. A CNN consists of an input layer, convolutional layers, pooling layers, fully connected layers, and an output layer. Convolutional layers are responsible for extracting local features from an image and can capture information such as edges and textures by performing convolution operations between the kernel and the input image. Pooling layers reduce the dimensionality of the feature maps to reduce computational cost while preserving key features. Fully connected layers integrate the extracted features and map them to the sample label space.

[0030] The three-dimensional morphological data of the weld after the initial welding of the base blank is the geometric contour data of the weld joint area in three-dimensional space output by the imaging analysis model. The three-dimensional morphological data of the weld after the initial welding of the base blank includes the weld width, penetration depth, reinforcement height, and surface waviness.

[0031] Micropore feature information is a detailed description of the micropores present in and around the weld seam area of ​​the titanium alloy dental implant abutment blank after preliminary welding, output by an imaging analysis model. This micropore feature information includes the number and morphology of pores, pore distribution coordinates, and pore volume percentage.

[0032] Morphological information describes the shape characteristics of micropores in the weld area after the initial welding of the titanium alloy dental implant abutment blank, specifically including the shape of the micropores, surface smoothness, and pore diameter uniformity.

[0033] The porosity ratio is the ratio between the total volume of all micropores in the weld area and the total volume of the weld area after the initial welding of the titanium alloy dental implant abutment blank.

[0034] The CT images of titanium alloy dental implant abutment blanks after initial welding contain the distribution of X-ray absorption coefficients of the internal materials of the abutment blanks. High-density metal materials and low-density porosity defects can show significant differences in grayscale values. The CT imaging slice sequence preserves the complete spatial topology of the weld area, which can provide a complete and accurate image data source for the model to extract three-dimensional morphological data and micropore feature information of the weld.

[0035] Convolutional neural networks (CNNs) can extract features layer by layer from CT images of titanium alloy dental implant abutment blanks after initial welding using multiple convolutional kernels. In shallow convolutional layers, CNNs can identify brightness abrupt changes in CT images, thus constructing the interface contour between the weld and the base material, as well as the edge boundaries of micropores. As the network depth increases, CNNs can correlate two-dimensional slice features along the channel dimension in deeper feature maps to construct the three-dimensional spatial structure of the weld. The model can determine the weld depth and remaining height by analyzing the grayscale continuity of the weld area and calculate the distribution density based on the clustering of micropore features. The model can also locate micropores by identifying the closed geometric features of low grayscale regions and calculate the physical diameter of the micropores based on pixel size. Then, in the fully connected layer, the convolutional neural network can map the extracted weld space features into specific geometric parameters to generate three-dimensional morphological data of the weld after the initial welding of the base blank. At the same time, it transforms the distribution features of micropores into micropore feature information, and finally determines the accurate three-dimensional morphological data of the weld and micropore feature information after the initial welding of the base blank.

[0036] Step S3: Based on the original welding process parameters of the laser hybrid welding, the three-dimensional morphology data of the weld after the initial welding of the base blank, and the micropore feature information, determine a number of preliminary porosity suppression welding schemes. The porosity suppression welding schemes include the mixing ratio of high-purity argon and nitrogen in laser hybrid welding and the oscillation amplitude of the laser beam.

[0037] In some embodiments, Figure 3 This is a flowchart illustrating a method for determining multiple preliminary porosity suppression welding schemes according to an embodiment of the present invention. The determination of multiple preliminary porosity suppression welding schemes includes steps S31 to S35:

[0038] Step S31: Based on the three-dimensional morphological data of the weld after the initial welding of the base blank and the micropore feature information, determine the information of multiple significant pore points.

[0039] In some embodiments, a salient feature analysis model can be used to determine multiple salient porosity points. The salient feature analysis model is a deep neural network model. The inputs to the salient feature analysis model are the three-dimensional morphological data of the weld after preliminary welding of the base blank and the micropore feature information; the output of the salient feature analysis model is the information on multiple salient porosity points.

[0040] Deep neural network models include deep neural networks (DNNs), which are artificial neural networks with multiple hidden layers. By stacking multiple layers of neurons between the input and output layers, deep neural networks can learn complex nonlinear mappings between data. Each neuron in a layer receives the output of the neuron in the previous layer, which is then processed through a linearly weighted sum and a nonlinear activation function before being passed to the next layer. Deep neural networks are suitable for extracting valuable target information from complex data.

[0041] Significant pore points are micropores that are representative and have a key impact, identified through a significant feature analysis model.

[0042] Each significant pore point includes its precise three-dimensional location coordinates, core pore size parameters, weld functional area affiliation, stress-related distance parameters, and local aggregation density parameters.

[0043] The classification of weld functional zones refers to the specific category of the weld functional zone where porosity is located, which includes the fusion zone, heat-affected zone, and weld center zone.

[0044] The fusion zone is the transition area between the weld and the titanium alloy base material.

[0045] The heat-affected zone is the area in which the titanium alloy base material does not melt under the heat of welding, but its microstructure and properties change.

[0046] The central area of ​​the weld is the core melting and crystallization zone of the weld.

[0047] The stress-related distance parameter is the spatial distance data between the pores and the key stress-bearing parts of the weld. The stress-related distance parameter includes the shortest straight-line distance between the pores and the key section of the weld tensile strength, and the shortest straight-line distance between the pores and the stress concentration point of the weld.

[0048] The local aggregation density parameter is the distribution density of other micropores within a preset range around a single pore. The local aggregation density parameter includes the number of micropores within a specific radius range centered on the pore, such as the number of micropores within a radius of 200 μm, and the average spacing between micropores within the specific radius range.

[0049] The three-dimensional morphological data of the weld after the initial welding of the base blank can provide the overall geometric constraint boundary of the weld and clarify the flow and solidification trajectory of the molten pool. The micropore feature information describes the physical properties and spatial distribution of the defects in detail, thus providing a complete geometric and defect attribute basis for the model to determine which pores are located in the stress concentration area or the typical defect location caused by improper process parameters.

[0050] Deep neural networks, with their multi-level feature learning and complex correlation modeling capabilities, can deeply mine the potential correlation between weld 3D morphology data and microporosity features. Deep neural networks can extract the spatial distribution patterns of weld geometric parameters such as weld penetration and weld reinforcement height, and combine this with the number, shape, and distribution coordinates of microporosity to accurately capture key feature dimensions affecting weld quality. Deep neural networks can also use fully connected layers to construct feature mapping relationships, quantifying the spatial correlation between porosity size, distribution density, and weld functional areas and stress-critical parts as weights, thereby filtering out porosity that significantly affects weld strength and sealing performance. By autonomously learning the correlation between welding defects and quality, deep neural networks avoid the subjectivity of manual selection, accurately identifying representative and significant porosity points.

[0051] Step S32: Cluster the multiple significant pore point information to obtain K clusters.

[0052] The clustering method used is K-means clustering. K-means clustering is an unsupervised learning algorithm used to divide a dataset into k distinct clusters, each containing data points with similar characteristics, while the characteristics of different clusters differ significantly. The value of K can be pre-input manually.

[0053] The K clusters are K sets into which information on multiple salient stomata are divided according to spatial proximity or feature similarity. Each of the K clusters includes information on all salient stomata belonging to that cluster and the coordinates of the cluster's center point.

[0054] The process of clustering multiple salient stomata using the K-means clustering algorithm is as follows: First, K salient stomata are randomly selected as initial cluster centers. Then, the Euclidean distance between each remaining salient stomata and these K initial cluster centers is calculated, and each salient stomata is assigned to the category represented by its nearest cluster center. After one round of assignment, the average coordinates of all salient stomata within each cluster are recalculated, and this average is used to update the new center of the cluster. This process of assignment and center updating is repeated until the position of the cluster centers no longer changes significantly or the preset number of iterations is reached, ultimately resulting in K clusters with clear boundaries and consistent internal characteristics.

[0055] Step S33: Determine multiple significant stomatal regions based on the K clusters.

[0056] In some embodiments, a salient region segmentation model can be used to determine multiple salient stomatal regions. The salient region segmentation model is a deep neural network model. The input to the salient region segmentation model is the K clusters, and the output of the salient region segmentation model is the multiple salient stomatal regions.

[0057] Multiple salient porosity regions are a set of concentrated porosity distribution regions that have a key impact on the safety of welded structures, formed by integrating and defining salient porosity points based on the common characteristics and spatial distribution patterns of K clusters in the salient region partitioning model.

[0058] Each significant stomatal region can include significant stomatal points with similar characteristics or spatial proximity into the same continuous spatial range based on the spatial correlation, size consistency, and common functional attributes of the stomata.

[0059] K clusters can group numerous, scattered individual salient stomata into categories based on characteristics such as size consistency and common functional attributes. This effectively simplifies the complexity of individual salient stomata data and avoids the inefficiency and difficulty in pattern extraction caused by directly analyzing individual salient stomata. K clusters retain the key features of individual salient stomata, present the characteristic homology of salient stomata through classification and aggregation, and implicitly contain the spatial distribution correlation patterns of salient stomata. This allows the model to quickly integrate salient stomata regions with clear boundaries based on the common characteristics and spatial distribution patterns of clusters, without having to analyze a massive number of individual salient stomata one by one. This accurately reflects the concentrated distribution state of salient stomata.

[0060] Deep neural networks possess powerful features for feature association mining and spatial modeling, enabling precise identification of multiple salient porosity regions based on K clusters. Deep neural networks can deeply analyze potential relationships within and between clusters, extracting common features of each cluster through a multi-level network structure, such as the size consistency and functional homology of salient pore points. Simultaneously, deep neural networks can capture spatial distribution patterns between clusters, including their proximity and density differences. By autonomously learning the physical mechanisms of porosity formation during welding, deep neural networks can transform feature associations and spatial patterns into quantified weights. The model can determine the logic for cluster integration or subdivision, merging clusters with homologous features and spatial proximity into the same region. Furthermore, deep neural networks can accurately split heterogeneous clusters, ultimately forming well-defined salient porosity regions.

[0061] In some embodiments, determining multiple significant stomatal regions based on the K clusters includes steps S331 to S333:

[0062] Step S331: Based on the K clusters, determine the spatial distribution pattern of the clusters, the porosity hazard of each cluster, the weld location sensitivity of each cluster, and the defect superposition intensity between each cluster and its adjacent clusters.

[0063] In some embodiments, deep neural networks can be used to determine the spatial distribution pattern of clusters, the porosity hazard of each cluster, the weld location sensitivity of each cluster, and the defect superposition intensity between each cluster and its neighboring clusters.

[0064] The determination of the spatial distribution pattern of the cluster is a specific category label used to clarify the overall spatial organization pattern of the cluster in the weld area after the overall spatial arrangement of the cluster is standardized and classified by a deep neural network.

[0065] Specific categories include isolated and dispersed type, chain-like extension type, sheet-like aggregation type, and core-radiation type.

[0066] The hazard level of porosity aggregation for each cluster is a numerical quantification index of the potential hazard to welding quality of that cluster, assessed by a deep neural network.

[0067] The weld position sensitivity of each cluster is a numerical parameter determined by a deep neural network based on the three-dimensional spatial position of each cluster in the weld, representing the degree of influence of each cluster on the weld performance.

[0068] The defect superposition intensity of each cluster and its neighboring clusters is determined by a deep neural network, which measures the combined effect of a single cluster and its surrounding neighboring clusters on weld quality. Defect superposition intensity is a quantitative indicator reflecting the synergistic destructive effect of clusters.

[0069] Deep neural networks possess the ability to efficiently process high-dimensional data and uncover deep correlations. K clusters contain information such as 3D coordinates, relative distances, density, and the number, morphology, and compactness of pores within each cluster, exhibiting complex nonlinear relationships. Deep neural networks, through hierarchical operations of multiple neurons, can automatically extract hidden spatial patterns and intrinsic connections from this data without requiring manually pre-defined feature rules. Deep neural networks can integrate global cluster information to identify inherent characteristics of the overall arrangement, thereby classifying cluster spatial distribution patterns. Simultaneously, the model can focus on the correlation between the local attributes of individual clusters and their 3D spatial positioning, enabling quantitative assessment of the hazard of pore aggregation and the sensitivity of weld location for each cluster. Deep neural networks can also accurately capture inter-cluster interactions, thus determining the defect superposition intensity between each cluster and its neighboring clusters. After training with sufficient samples, the model can establish stable data mapping relationships and output consistent and reliable evaluation results under different cluster data scenarios.

[0070] Step S332: Based on the K clusters, the porosity hazard of each cluster, the weld location sensitivity of each cluster, and the defect superposition intensity between each cluster and adjacent clusters, determine multiple core concern clusters and multiple regular concern clusters.

[0071] In some embodiments, a deep neural network may be used to determine multiple core clusters of interest and multiple regular clusters of interest.

[0072] The core concern clusters are the set of clusters that pose the greatest threat to welding quality and require priority attention, as determined by deep neural networks.

[0073] The regular clusters of interest are other clusters identified by deep neural networks that have a certain potential impact on welding quality but have a lower priority than the core clusters of interest.

[0074] Deep neural networks can efficiently integrate complex information across multiple dimensions. After receiving input information such as K clusters, the porosity hazard of each cluster, the weld location sensitivity of each cluster, and the defect superposition intensity between each cluster and its neighboring clusters, the model can calculate the comprehensive risk score for each cluster. The score calculation integrates the severity of the hazard, the influence weight of location sensitivity, and the defect superposition effect with neighboring clusters, while also performing global calibration by referencing the overall distribution of the K clusters. Subsequently, the model can classify clusters with high comprehensive risk scores as core concern clusters and clusters with scores in the medium range as regular concern clusters, based on a preset risk classification standard.

[0075] Step S333: Based on the spatial distribution pattern of the clusters, the multiple core clusters of interest, and the multiple regular clusters of interest, determine multiple significant stomatal regions.

[0076] In some embodiments, a deep neural network may be used to determine multiple salient stomatal regions.

[0077] Deep neural networks can encode cluster spatial distribution patterns as categorical features using one-hot encoding. The one-hot encoded features are then dimensionally aligned with the spatial parameters and attributes of core and regular clusters of interest. These spatial parameters and attributes of multiple core and regular clusters of interest are then input into the input layer. In the hidden layers, the deep neural network can activate specific regions based on the cluster spatial distribution pattern to generate weight paths. If the cluster spatial distribution pattern is chain-like and extending, the deep neural network increases weight connections along the extension direction, tending to merge spatially adjacent core and regular clusters of interest into a continuous, elongated region. If the cluster spatial distribution pattern is isolated and scattered, the deep neural network tends to define separate region boundaries for each core cluster of interest. Next, the deep neural network can calculate the spatial neighborhood relationships between multiple core and regular clusters of interest. The model constructs an initial field with multiple core clusters of interest as region centers and evaluates whether nearby regular clusters of interest are within the influence radius of this initial field. Deep neural networks can make judgments through nonlinear operations. If multiple regular clusters of interest are close enough to multiple core clusters of interest and conform to the topological characteristics of the cluster spatial distribution pattern, they are included in the same region for envelope. If multiple regular clusters of interest are far apart, the model will decide whether to divide them into secondary regions independently or ignore them based on their own attributes.

[0078] Step S34: Based on the multiple significant porosity regions and the original welding process parameters of the laser composite welding, determine the high-purity argon and nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region.

[0079] In some embodiments, an adaptation range determination model can be used to determine the adaptation range for the high-purity argon to nitrogen mixing ratio and the laser beam oscillation amplitude for each significant porosity region. The adaptation range determination model is a Transformer model. The inputs to the adaptation range determination model are the multiple significant porosity regions and the original welding process parameters of the laser hybrid welding. The outputs of the adaptation range determination model are the adaptation ranges for the high-purity argon to nitrogen mixing ratio and the laser beam oscillation amplitude for each significant porosity region.

[0080] The Transformer model is a deep learning model based on a self-attention mechanism. It primarily consists of an encoder and a decoder, abandoning traditional recurrent and convolutional structures and relying entirely on attention mechanisms to capture the global dependencies between elements in the input sequence. Self-attention allows the model to focus on information from all other positions in the input sequence while processing information at a particular position, assigning different weights. Multi-head attention enables the model to learn features in parallel from different representation subspaces. The Transformer model excels at processing sequential data and establishing long-range associations, and can effectively handle multidimensional nonlinear mappings between complex process parameters and defect features.

[0081] The high-purity argon to nitrogen mixing ratio adaptation range for each significant porosity region is determined by the adaptation range determination model. This range represents the effective range of high-purity argon to nitrogen mixing ratios that can suppress micropore formation within the region.

[0082] The laser beam oscillation amplitude adaptation range for each significant pore region is determined by the adaptation range determination model, which is the range of laser beam oscillation amplitude values ​​that can effectively suppress the micropores in that region.

[0083] Multiple prominent porosity regions clearly indicate the location and severity of defect concentration, reflecting the instability of the local molten pool state. The original welding process parameters of laser-hybrid welding provide the initial process baseline for the generation of these defects. The Transformer model can infer the range of process parameter adjustments that can correct the current defect state based on these two data.

[0084] The Transformer model uses an encoder to vectorize the features of multiple salient porosity regions and the original welding process parameters and corresponding defect results of laser-assisted hybrid welding. A self-attention mechanism calculates the correlation weights between the features of salient porosity regions and the original process parameters, capturing the relationship between porosity formation in specific regions and insufficient gas ratios or beam oscillation. For example, the model can identify that porosity in a certain region is caused by poor molten pool fluidity and is related to an excessively low high-purity argon-nitrogen mixing ratio. Based on the features extracted by the encoder and the learned process mechanisms—such as how adding nitrogen can improve porosity in titanium alloy welds and increasing oscillation can refine grains and expel gas—the decoder can predict the target parameter range for eliminating porosity in that region. The model can generate a parameter range with confidence levels by analyzing repair cases of similar defects in a historical database. By comprehensively considering the needs of different regions, the Transformer model can ultimately output the appropriate high-purity argon-nitrogen mixing ratio range and the appropriate laser beam oscillation amplitude range for each salient porosity region.

[0085] Step S35: Based on the high-purity argon and nitrogen mixing ratio adaptation range of each significant porosity region and the laser beam oscillation amplitude adaptation range, determine multiple preliminary porosity suppression welding schemes.

[0086] In some embodiments, a scheme generation model can be used to determine multiple preliminary porosity suppression welding schemes. The scheme generation model is a deep neural network model. The inputs to the scheme generation model are the high-purity argon and nitrogen mixing ratio adaptation range for each significant porosity region and the laser beam oscillation amplitude adaptation range. The output of the scheme generation model is multiple preliminary porosity suppression welding schemes.

[0087] The initial porosity suppression welding scheme is a set of exclusive combinations of high-purity argon and nitrogen mixing ratios and laser beam oscillation amplitudes generated by the scheme generation model for each significant porosity region. Each initial porosity suppression welding scheme corresponds to a specific set of high-purity argon and nitrogen mixing ratios and laser beam oscillation amplitudes for each significant porosity region.

[0088] The deep neural network first normalizes the boundary values ​​of the high-purity argon-nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region and inputs them into the input layer. In the hidden layer, the deep neural network performs cross-analysis on the high-purity argon-nitrogen mixing ratio adaptation range and the laser beam oscillation amplitude adaptation range for each significant porosity region. If multiple significant porosity regions exist, and the corresponding high-purity argon-nitrogen mixing ratio adaptation range or laser beam oscillation amplitude adaptation range for each significant porosity region is not completely consistent, the model calculates the common overlapping part of these ranges. If there are no completely overlapping ranges, the model assigns weights based on the defect severity of each significant porosity region, and determines the comprehensive feasible region that can take into account the porosity suppression requirements of all regions through weighted fusion. Then, the network weights can be used to perform intelligent sampling and nonlinear mapping within the defined high-purity argon-nitrogen mixing ratio adaptation range and laser beam oscillation amplitude adaptation range for each significant porosity region. The model tends to select key parameter points that show a high probability of porosity suppression in historical data features. Deep neural networks simulate the coupling effects between process parameters through the weighted connections of neurons, and eliminate outliers that, although located within the optimal range of high-purity argon-nitrogen mixing ratios or laser beam oscillation amplitudes for each significant porosity region, may produce negative effects when combined. In the output layer, the deep neural network can inversely normalize the optimal feature solutions calculated by the hidden layer into specific high-purity argon-nitrogen mixing ratio values ​​and laser beam oscillation amplitude values, thereby generating multiple sets of differentiated numerical combinations, i.e., outputting multiple initial porosity suppression welding schemes.

[0089] Step S4: Based on the multiple preliminary porosity suppression welding schemes, laser composite welding is performed on multiple initial titanium alloy dental implant abutment blanks, and CT images of each preliminary porosity suppression welding scheme after welding are obtained.

[0090] Multiple initial titanium alloy dental implant abutment blanks refer to titanium alloy dental implant abutment blanks that have not undergone any welding operations and have uniform initial structure and material characteristics.

[0091] The CT images following each initial porosity suppression welding scheme are obtained by performing a comprehensive CT scan on each initial titanium alloy dental implant abutment blank after laser composite welding using the corresponding initial porosity suppression welding scheme.

[0092] CT imaging after welding each initial porosity suppression welding scheme can show the weld quality and microporosity suppression effect inside the blank after welding based on each initial porosity suppression welding scheme.

[0093] Step S5: Determine the target porosity suppression welding scheme based on the CT imaging after welding of each initially selected porosity suppression welding scheme.

[0094] In some embodiments, Figure 4 This is a flowchart illustrating a target porosity suppression welding scheme provided by an embodiment of the present invention. The process of determining the target porosity suppression welding scheme includes steps S51 to S52:

[0095] Step S51: Construct a porosity suppression welding atlas. The porosity suppression welding atlas includes multiple initially selected porosity suppression nodes and multiple edges between the nodes. The node features of each initially selected porosity suppression node include an initially selected porosity suppression welding scheme and a CT image of the initially selected porosity suppression welding scheme after welding. The edges between the nodes represent the similarity between the initially selected porosity suppression welding schemes.

[0096] A graph is a data structure composed of vertices and edges, used to represent the relationships between nodes. A porosity suppression welding graph is a graph structure that can describe the correlation between different welding schemes and their effects. A porosity suppression welding graph includes multiple initially selected porosity suppression nodes and edges connecting these nodes.

[0097] Multiple preliminary porosity suppression nodes are the basic units in the porosity suppression welding atlas, and each node corresponds to a preliminary porosity suppression welding scheme.

[0098] The node features of the initially selected porosity suppression nodes can comprehensively reflect the execution basis and welding effect of the initially selected porosity suppression welding scheme. The node features of each initially selected porosity suppression node include an initially selected porosity suppression welding scheme and a CT image of the initially selected porosity suppression welding scheme after welding.

[0099] The edges between nodes are lines connecting two nodes, and the edges represent the similarity between the initially selected porosity suppression welding schemes. The edges between nodes can be used to quantify the similarity in parameter settings between the two initially selected porosity suppression welding schemes.

[0100] In some embodiments, deep neural networks can be used to determine the similarity between preliminary porosity suppression welding schemes.

[0101] Step S52: Process the porosity suppression welding map based on graph neural network to determine the target porosity suppression welding scheme.

[0102] Graph Neural Networks (GNNs) are deep learning models capable of processing graph data. GNNs allow nodes in a graph to exchange information with their neighbors through a message-passing mechanism. In each graph convolutional operation, nodes can aggregate feature information from their neighbors and update their state by combining it with their own features. GNNs can capture the topological features and node attribute features of a graph, thereby effectively analyzing the relationships between nodes and global distribution patterns. GNNs are suitable for tasks such as node classification, link prediction, and graph attribute regression. The input of the GNN is the porosity suppression welding graph, and the output of the GNN is the target porosity suppression welding scheme.

[0103] The target porosity suppression welding scheme was determined by analyzing the porosity suppression welding spectrum using a graph neural network. It is the welding scheme with the best microporosity suppression effect and the best weld quality.

[0104] The target porosity suppression welding scheme can meet the welding quality requirements of titanium alloy dental implant abutment blanks to the greatest extent.

[0105] Graph Neural Networks (GNNs) can determine target porosity suppression welding schemes based on porosity suppression welding maps. The core of this capability lies in the GNN's ability to perform deep analysis of the map data. GNNs can use node embedding techniques to transform the features of each initially selected porosity suppression node into a low-dimensional vector, and utilize edge weights to efficiently capture local correlations between nodes and the global porosity suppression welding map structure. Through message passing mechanisms, GNNs can aggregate the feature information of each initially selected porosity suppression node with its neighbors, focusing on integrating the effect data corresponding to schemes with similar parameter settings, thereby achieving deep fusion and complementarity of different node features. This approach allows GNNs to overcome the limitations of a single initially selected scheme, deeply learning and refining a more universal and superior complex nonlinear relationship between scheme parameters and welding effects. Simultaneously, GNNs can fully absorb the advantageous features of all initially selected schemes. Based on the learned global optimal rules, and through deep integration and optimization of the advantages of all initially selected schemes, the GNN ultimately determines the target porosity suppression welding scheme.

[0106] Step S6: Based on the target porosity suppression welding scheme, perform laser composite welding on the remaining initial titanium alloy dental implant abutment blanks to produce qualified finished titanium alloy dental implant abutments.

[0107] Once the target porosity suppression welding scheme is determined, laser composite welding is performed on the remaining initial titanium alloy dental implant abutment blanks of the same batch based on the target porosity suppression welding scheme to ensure that the generation of porosity defects during the laser composite welding process is suppressed most effectively.

[0108] Based on the same inventive concept Figure 5 This is a schematic diagram of a laser hybrid welding control system provided in an embodiment of the present invention. The laser hybrid welding control system includes:

[0109] The acquisition module 71 is used to acquire the original welding process parameters of laser hybrid welding and CT images of the titanium alloy dental implant abutment blank after preliminary welding.

[0110] Data analysis module 72 is used to determine the three-dimensional morphological data and micropore feature information of the weld after the initial welding of the titanium alloy dental implant abutment blank based on CT imaging after the initial welding of the abutment blank.

[0111] The scheme generation module 73 is used to determine multiple preliminary porosity suppression welding schemes based on the original welding process parameters of the laser hybrid welding, the three-dimensional morphological data of the weld after the initial welding of the base blank, and the micropore feature information. The porosity suppression welding schemes include the mixing ratio of high-purity argon and nitrogen in laser hybrid welding and the oscillation amplitude of the laser beam.

[0112] The welding test module 74 is used to perform laser composite welding on multiple initial titanium alloy dental implant abutment blanks based on the multiple preliminary porosity suppression welding schemes, and to obtain CT images of each preliminary porosity suppression welding scheme after welding.

[0113] The scheme optimization module 75 is used to determine the target porosity suppression welding scheme based on the CT imaging after welding of each initially selected porosity suppression welding scheme.

[0114] The finished product production module 76 is used to perform laser composite welding on the remaining initial titanium alloy dental implant abutment blanks based on the target porosity suppression welding scheme, so as to produce qualified finished titanium alloy dental implant abutments.

[0115] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0116] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A control method of laser composite welding characterized by, The method comprises the following steps: acquiring original welding process parameters of laser composite welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; determining weld three-dimensional morphology data and micro-pore feature information of the titanium alloy dental implant abutment blank after preliminary welding based on the CT imaging of the titanium alloy dental implant abutment blank after preliminary welding; determining a plurality of initial pore inhibition welding schemes based on the original welding process parameters of the laser composite welding, the weld three-dimensional morphology data of the titanium alloy dental implant abutment blank after preliminary welding and the micro-pore feature information, wherein the pore inhibition welding scheme comprises a high-purity argon and nitrogen mixed ratio of the laser composite welding and a laser beam swing amplitude, and the determination of the plurality of initial pore inhibition welding schemes based on the original welding process parameters of the laser composite welding, the weld three-dimensional morphology data of the titanium alloy dental implant abutment blank after preliminary welding and the micro-pore feature information comprises: determining a plurality of significant pore point information based on the weld three-dimensional morphology data of the titanium alloy dental implant abutment blank after preliminary welding and the micro-pore feature information; performing clustering based on the plurality of significant pore point information to obtain K clusters; determining a plurality of significant pore regions based on the K clusters; determining a high-purity argon and nitrogen mixed ratio adaptive interval and a laser beam swing amplitude adaptive interval of each significant pore region based on the plurality of significant pore regions and the original welding process parameters of the laser composite welding; determining a plurality of initial pore inhibition welding schemes based on the high-purity argon and nitrogen mixed ratio adaptive interval of each significant pore region and the laser beam swing amplitude adaptive interval; performing laser composite welding on a plurality of initial titanium alloy dental implant abutment blanks based on the plurality of initial pore inhibition welding schemes respectively, and acquiring CT imaging after welding of each initial pore inhibition welding scheme; determining a target pore inhibition welding scheme based on the CT imaging after welding of each initial pore inhibition welding scheme; performing laser composite welding on the remaining initial titanium alloy dental implant abutment blanks based on the target pore inhibition welding scheme to produce qualified titanium alloy dental implant abutment finished products.

2. The control method of laser compound welding according to claim 1, wherein The determination of the target pore inhibition welding scheme based on the CT imaging after welding of each initial pore inhibition welding scheme comprises: constructing a pore inhibition welding graph, wherein the pore inhibition welding graph comprises a plurality of initial pore inhibition nodes and a plurality of edges between the nodes, the node features of each initial pore inhibition node comprise an initial pore inhibition welding scheme and CT imaging after welding of the initial pore inhibition welding scheme, and the edges between the nodes represent the similarity between the initial pore inhibition welding schemes; determining the target pore inhibition welding scheme by processing the pore inhibition welding graph based on a graph neural network.

3. The control method of laser compound welding according to claim 1, wherein The pore feature information comprises pore quantity and morphology information, pore distribution coordinates and pore volume proportion.

4. A control system for laser hybrid welding, characterized in that The method comprises the following steps: an acquisition module, configured to acquire original welding process parameters of laser composite welding and CT imaging of a titanium alloy dental implant abutment blank after preliminary welding; a data analysis module configured to determine weld seam three-dimensional morphology data and micro-pore feature information of the titanium alloy dental implant abutment blank after the initial welding based on CT imaging of the titanium alloy dental implant abutment blank after the initial welding; a scheme generation module configured to determine a plurality of primary porosity suppression welding schemes based on the original laser hybrid welding process parameters, the weld seam three-dimensional morphology data of the abutment blank after the initial welding, and the micro-pore feature information, wherein the porosity suppression welding scheme includes a high-purity argon and nitrogen mixed ratio of the laser hybrid welding and a laser beam swing amplitude, and the scheme generation module is further configured to: determine a plurality of significant porosity point information based on the weld seam three-dimensional morphology data of the abutment blank after the initial welding and the micro-pore feature information; cluster the plurality of significant porosity point information to obtain K clusters; determine a plurality of significant porosity regions based on the K clusters; determine a high-purity argon and nitrogen mixed ratio adaptive range and a laser beam swing amplitude adaptive range of each significant porosity region based on the plurality of significant porosity regions and the original laser hybrid welding process parameters; determine a plurality of primary porosity suppression welding schemes based on the high-purity argon and nitrogen mixed ratio adaptive range of each significant porosity region and the laser beam swing amplitude adaptive range; a welding test module configured to perform laser hybrid welding on a plurality of initial titanium alloy dental implant abutment blanks based on the plurality of primary porosity suppression welding schemes and obtain CT imaging of each primary porosity suppression welding scheme after welding; a scheme optimization module configured to determine a target porosity suppression welding scheme based on the CT imaging of each primary porosity suppression welding scheme after welding; a finished product production module configured to perform laser hybrid welding on the remaining initial titanium alloy dental implant abutment blanks based on the target porosity suppression welding scheme to produce qualified titanium alloy dental implant abutment finished products.

5. The control system for laser composite welding of claim 4, wherein, The scheme optimization module is further configured to: construct a porosity suppression welding graph, wherein the porosity suppression welding graph includes a plurality of primary porosity suppression nodes and a plurality of edges between the nodes, each primary porosity suppression node has a node feature including a primary porosity suppression welding scheme and CT imaging of the primary porosity suppression welding scheme after welding, and the edges between the nodes represent the similarity between the primary porosity suppression welding schemes; determine the target porosity suppression welding scheme based on processing of the porosity suppression welding graph by a graph neural network.

6. The control system for laser composite welding of claim 4, wherein, The porosity feature information includes the number and morphology information of the pores, pore distribution coordinates, and pore volume proportion.

7. An electronic device, comprising: comprise: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the laser hybrid welding control method of any one of claims 1 to 3.

8. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the laser hybrid welding control method of any one of claims 1 to 3.

Images