A metal 3D printing model three-dimensional modeling system

Abstract

This invention relates to the field of metal 3D printing and computer graphics 3D modeling technology, specifically a 3D modeling system for metal 3D printing models. The system includes units for ideal benchmark construction, parametric injection and simulation, dual-track differential extraction, coupling decision, and inverse compensation. The system generates a thermal stress simulation model by constructing a zero-stress ideal model and injecting physical constraints. Its core is to extract the actual and theoretical residual fields of the measured point cloud and the simulation model relative to the ideal model, and calculate their topological 3D similarity. When the similarity meets the overlap condition, the system locks the true thermal stress region and performs inverse compensation; otherwise, it is considered noise rejection. This invention achieves physical-level decoupling between real thermal deformation and environmental noise, eliminating geometric errors in the initial stage and significantly reducing the false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of metal 3D printing and computer graphics 3D modeling technology, specifically a 3D modeling system for metal 3D printing models. Background Technology

[0002] In the current metal 3D printing and computer graphics modeling environment, models often undergo displacement and deformation during the forming process due to complex physical manufacturing constraints and thermodynamic factors. To repair these deformations, existing solutions generally adopt the method of directly comparing the differences between measured 3D point cloud data and the initial design model, extracting deformation features and performing compensation.

[0003] Although this scheme has basic deformation analysis capabilities, traditional 3D modeling systems are prone to introducing geometric approximation errors when discretizing smooth surfaces, and the measured data are often mixed with environmental noise such as scanner optical distortion and placement posture deviation. This original comparison scheme, which relies heavily on static geometric inspection, cannot effectively separate the real thermodynamic deformation from environmental measurement noise and non-thermal stress damage, resulting in a high false alarm rate for defects and a lack of thermophysical law verification for feature extraction, which severely limits the accuracy of 3D model topology reconstruction and reverse compensation.

[0004] Therefore, how to accurately remove environmental measurement noise under complex geometric surfaces, precisely identify true thermal stress defects, and achieve high-fidelity reverse compensation 3D modeling has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a three-dimensional modeling system for metal 3D printing models. Specifically, the technical solution of this invention includes:

[0006] Ideal reference construction unit: Obtain initial ideal geometric boundary data, construct a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generate a zero-stress ideal three-dimensional model;

[0007] Parametric injection and simulation unit: acquires physical manufacturing constraint parameters, actively superimposes the physical manufacturing constraint parameters onto the zero-stress ideal three-dimensional model, drives the zero-stress ideal three-dimensional model to undergo displacement, and generates a thermal stress simulation state model;

[0008] Dual-track differential extraction unit: acquires measured 3D point cloud data, calculates the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generates a real residual field; simultaneously calculates the difference between the thermal stress simulation model and the zero-stress ideal 3D model, and generates a theoretical residual field;

[0009] Coupled Decision and Reverse Compensation Unit: Calculates the topological 3D similarity between the actual residual field and the theoretical residual field; determines whether the topological 3D similarity satisfies a preset coincidence condition; when the topological 3D similarity satisfies the preset coincidence condition, determines the corresponding region as a true thermal stress pain point region, inverts the theoretical residual field and superimposes it onto the zero-stress ideal 3D model, generating a pre-reverse compensated 3D printing model; when the topological 3D similarity does not satisfy the preset coincidence condition, determines the corresponding region as a noise region, and removes the residual data of the corresponding region to maintain the original state of the zero-stress ideal 3D model.

[0010] Furthermore, the method of obtaining initial ideal geometric boundary data, constructing a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generating a zero-stress ideal three-dimensional model includes: parsing a preset three-dimensional design file to extract the initial ideal geometric boundary data; reconstructing the initial ideal geometric boundary data based on first principles using a preset non-uniform rational B-spline surface or high-precision voxels; constructing the pure mathematical ideal geometric field under preset ideal physical boundary conditions of no gravity, absolute isotherm, and no thermal stress; and using the pure mathematical ideal geometric field as the reference origin for difference calculation to generate the zero-stress ideal three-dimensional model.

[0011] Furthermore, the method of obtaining physical manufacturing constraint parameters, actively superimposing these parameters onto the zero-stress ideal 3D model, and driving the zero-stress ideal 3D model to undergo displacement to generate a thermal stress simulation model includes: obtaining preset thermal property parameters of the metal material and a preset printing process vector as the physical manufacturing constraint parameters; wherein, the thermal property parameters of the metal material include the coefficient of thermal expansion and latent heat of phase change; the printing process vector includes laser power, scanning path, and layer thickness; a built-in thermodynamic active simulation engine is used to convert the thermal property parameters of the metal material and the printing process vector into a physical parameter vector; the physical parameter vector is used as an interference factor and actively superimposed onto the zero-stress ideal 3D model; and the thermodynamic active simulation engine drives the voxels in the zero-stress ideal 3D model to undergo displacement, thereby generating the thermal stress simulation model.

[0012] Furthermore, the method of acquiring measured 3D point cloud data, calculating the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generating a real residual field includes: collecting historical in-situ 3D scan point cloud data of the same topology as the measured 3D point cloud data; aligning the measured 3D point cloud data with the zero-stress ideal 3D model in spatial coordinates; calculating the displacement vector difference between the aligned measured 3D point cloud data and the zero-stress ideal 3D model at corresponding spatial nodes to generate a real residual vector field as the real residual field; wherein, the real residual field includes real thermal deformation characteristics and environmental noise characteristics.

[0013] Furthermore, the method of simultaneously calculating the difference between the thermal stress simulation model and the zero-stress ideal three-dimensional model to generate the theoretical residual field includes: extracting the first three-dimensional coordinates of each voxel in the thermal stress simulation model; extracting the second three-dimensional coordinates of the corresponding voxels in the zero-stress ideal three-dimensional model; deriving the purely theoretical stress deformation characteristics between the thermal stress simulation model and the zero-stress ideal three-dimensional model based on thermophysical equations; calculating the vector difference between the first three-dimensional coordinates and the second three-dimensional coordinates to generate a theoretical residual vector field, which serves as the theoretical residual field.

[0014] Furthermore, the method for calculating the topological three-dimensional similarity between the actual residual field and the theoretical residual field includes: extracting a first deformation tensor of the actual residual field; extracting a second deformation tensor of the theoretical residual field; calculating the Hausdorff distance between the first deformation tensor and the second deformation tensor; calculating the cosine similarity of the deformation tensor angle between the first deformation tensor and the second deformation tensor; and fusing the Hausdorff distance and the cosine similarity of the deformation tensor angle to generate the topological three-dimensional similarity.

[0015] Further, the method for determining whether the topological three-dimensional similarity satisfies a preset overlap condition; when the topological three-dimensional similarity satisfies the preset overlap condition, determining the corresponding region as a true thermal stress pain point region includes: setting a preset similarity threshold as the preset overlap condition; comparing the topological three-dimensional similarity with the preset similarity threshold; when the topological three-dimensional similarity is greater than or equal to the preset similarity threshold, determining that the deformation vectors of the actual residual field and the theoretical residual field in the corresponding region are highly overlapped, thus satisfying the preset overlap condition; in response to satisfying the preset overlap condition, removing scanner noise and geometric visual errors in the actual residual field, and determining that the corresponding region is the true thermal stress pain point region.

[0016] Furthermore, the method of inverting the theoretical residual field and superimposing it onto the zero-stress ideal 3D model to generate a pre-reverse-compensated 3D printing model includes: extracting the theoretical residual field corresponding to the true thermal stress pain point region; performing an inversion operation on the theoretical residual field to generate a reverse deformation compensation vector field; directly superimposing the reverse deformation compensation vector field onto the spatial nodes corresponding to the zero-stress ideal 3D model; reconstructing the 3D topology based on the superimposed spatial node coordinates to generate the pre-reverse-compensated 3D printing model, thus completing closed-loop modeling.

[0017] Furthermore, the method of determining the corresponding region as a noise region and removing the residual data of the corresponding region to maintain the original state of the zero-stress ideal three-dimensional model when the topological three-dimensional similarity does not meet the preset overlap condition includes: when the topological three-dimensional similarity is less than the preset similarity threshold, determining that the actual residual field and the theoretical residual field do not overlap in the corresponding region and do not meet the preset overlap condition; in response to not meeting the preset overlap condition, determining that the deformation of the corresponding region is caused by non-thermal stress physical factors and marking it as the noise region; shielding the theoretical residual field and the actual residual field corresponding to the noise region, and retaining the original geometric topology of the zero-stress ideal three-dimensional model in the corresponding region.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. This system effectively overcomes the shortcomings of traditional 3D modeling systems, which easily introduce geometric approximation errors when discretizing smooth surfaces, and the original comparison scheme, which is difficult to separate real thermal stress deformation from environmental measurement noise;

[0020] The system utilizes a pre-defined non-uniform rational B-spline surface or high-precision voxels to reconstruct the initial ideal geometric boundary data, eliminating fundamental errors in static geometric inspection at the source. The system extracts the first deformation tensor of the real residual field and the second deformation tensor of the theoretical residual field, and generates a topological three-dimensional similarity by fusing the Hausdorff distance and the cosine similarity of the angle between the deformation tensors. By comparing this similarity with a pre-defined similarity threshold, the system can accurately identify features.

[0021] When the topological 3D similarity meets the preset overlap condition, the system actively removes scanner noise and geometric visual errors mixed in the real residual field and locks the real thermal stress pain point area; when the preset overlap condition is not met, the system attributes the deformation to non-thermal stress physical factors, marks it as a noise area and masks the corresponding residual data; this successfully achieves physical-level decoupling between real thermal deformation and environmental noise and non-thermal stress damage, while preserving the original geometric topology of the zero-stress ideal 3D model, and significantly reducing the false alarm rate caused by traditional static geometric comparison;

[0022] 2. This system solves the technical problem that traditional deformation extraction schemes lack verification of thermophysical laws, which severely limits the accuracy of 3D model topology reconstruction and inverse compensation;

[0023] The system uses a built-in thermodynamic active simulation engine to convert the thermal properties of metallic materials, including the coefficient of thermal expansion and latent heat of phase change, as well as the printing process vector, including laser power, scanning path and layer thickness, into a physical parameter vector. This vector is actively superimposed as an interference factor to drive the zero-stress ideal 3D model to generate a thermal stress simulation model, and derives the pure theoretical residual field based on the thermophysical equations. After identifying the true thermal stress pain point area, the system directly extracts the noise-free theoretical residual field of that area, performs an inversion operation, generates an inverse deformation compensation vector field, and superimposes it onto the spatial nodes of the zero-stress ideal 3D model to reconstruct the 3D topology.

[0024] This mechanism shifts the compensation benchmark from the measured point cloud with mixed measurement noise to the theoretical residuals derived from first-principles calculations, ensuring that every compensation action is based on a rigorous thermophysical causal chain. This not only avoids introducing external physical environment interference into closed-loop modeling, but also ensures the high accuracy of the conversion of physical manufacturing constraints into computer graphics 3D modeling, thereby generating a pre-reverse-compensated 3D printed model with extremely high dimensional accuracy and structural fidelity. Attached Figure Description

[0025] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0026] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0028] like Figure 1 As shown, a three-dimensional modeling system for metal 3D printing models includes:

[0029] Ideal reference building unit: Obtain initial ideal geometric boundary data, construct a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generate a zero-stress ideal three-dimensional model;

[0030] Parametric injection and simulation unit: acquires physical manufacturing constraint parameters, actively superimposes the physical manufacturing constraint parameters onto the zero-stress ideal 3D model, drives the zero-stress ideal 3D model to undergo displacement, and generates a thermal stress simulation state model;

[0031] Dual-track differential extraction unit: acquires measured 3D point cloud data, calculates the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generates a realistic residual field; simultaneously calculates the difference between the thermal stress simulation model and the zero-stress ideal 3D model, and generates a theoretical residual field;

[0032] Coupled Decision and Reverse Compensation Unit: Calculates the topological 3D similarity between the actual residual field and the theoretical residual field; determines whether the topological 3D similarity meets the preset coincidence condition; when the topological 3D similarity meets the preset coincidence condition, the corresponding region is determined to be the true thermal stress pain point region, the theoretical residual field is inverted and superimposed onto the zero-stress ideal 3D model to generate a pre-reverse compensated 3D printing model; when the topological 3D similarity does not meet the preset coincidence condition, the corresponding region is determined to be the noise region, and the residual data of the corresponding region is removed to maintain the original state of the zero-stress ideal 3D model.

[0033] This embodiment provides a 3D modeling system for metal 3D printing models, aiming to establish a reverse conversion mechanism from physical manufacturing constraints to topological reconstruction of 3D modeling using computer graphics. Taking the defect investigation application scenario of aero-engine turbine blades as an example, when a blade forming task is received, the ideal reference building unit obtains the initial ideal geometric boundary data. This data comes from the original boundary representation data output by the computer-aided design software, representing the original design form without physical environmental interference. Based on the initial ideal geometric boundary data, the ideal reference building unit constructs a pure mathematical ideal geometric field after excluding real physical factors such as gravity and temperature gradients, and uses this as an absolute zero-degree reference to generate a zero-stress ideal 3D model to establish an absolutely smooth geometric reference system.

[0034] The parametric injection and simulation unit acquires physical manufacturing constraint parameters, which are data sets reflecting the properties of specific metal materials and the working state of specific printers. The physical manufacturing constraint parameters are actively superimposed onto the zero-stress ideal three-dimensional model. By simulating the thermodynamic boundary conditions during actual printing, the parametric injection and simulation unit drives the zero-stress ideal three-dimensional model to undergo displacement to generate a thermal stress simulation state model.

[0035] During this period, the dual-track differential extraction unit acquires measured 3D point cloud data, calculates the difference between the measured 3D point cloud data and the zero-stress ideal 3D model to generate the real residual field, and simultaneously calculates the difference between the thermal stress simulation model and the zero-stress ideal 3D model to generate the theoretical residual field. After the dual-track differential extraction is completed, the coupling decision and reverse compensation unit calculates the topological 3D similarity between the real residual field and the theoretical residual field, and determines whether the topological 3D similarity meets the preset coincidence condition. When the topological 3D similarity meets the preset coincidence condition, the coupling decision and reverse compensation unit determines that the corresponding region is the true thermal stress pain point region, and then inverts the theoretical residual field and superimposes it onto the zero-stress ideal 3D model to generate a pre-reverse compensated 3D printing model.

[0036] Conversely, the corresponding area is determined to be a noise area, and the residual data of the corresponding area is removed to maintain the original state of the zero-stress ideal three-dimensional model. This technical solution actively synthesizes a model with theoretical thermal stress defects in digital space, stripping away the environmental measurement noise hidden under the complex geometric appearance, and verifying the robustness of the system in a variable manufacturing environment.

[0037] In this embodiment, the method for obtaining initial ideal geometric boundary data, constructing a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generating a zero-stress ideal three-dimensional model includes: parsing a preset three-dimensional design file to extract the initial ideal geometric boundary data; reconstructing the initial ideal geometric boundary data based on first principles using a preset non-uniform rational B-spline surface or high-precision voxels; constructing a pure mathematical ideal geometric field under preset ideal physical boundary conditions of no gravity, absolute isotherm, and no thermal stress; and using the pure mathematical ideal geometric field as the reference origin for difference calculation to generate a zero-stress ideal three-dimensional model.

[0038] This embodiment further specifies the internal logic of the ideal benchmark construction unit. In the application scenario of high-fidelity parsing and model building of 3D design source files, the ideal benchmark construction unit parses the preset 3D design files and extracts the initial ideal geometric boundary data. Since the forced discretization of smooth surfaces when exporting files by traditional 3D modeling software introduces geometric approximation errors, the ideal benchmark construction unit, based on first principles, uses preset non-uniform rational B-spline surfaces or high-precision voxels to reconstruct the initial ideal geometric boundary data to eliminate initial errors. This reconstruction operation aims to establish an absolutely smooth reference system in a mathematical sense.

[0039] Under the pre-defined ideal physical boundary conditions of zero gravity, absolute isotherm, and no thermal stress, the ideal reference building unit constructs a pure mathematical ideal geometric field. The pure mathematical ideal geometric field is used as the reference origin for difference calculation to generate a zero-stress ideal three-dimensional model. The design of using non-uniform rational B-spline surface for reconstruction avoids misjudging the meshing discretization error as thermal stress deformation error during subsequent difference calculation, reflecting the technical consideration of high fidelity in the basic data cleaning stage.

[0040] In this embodiment, the method of obtaining physical manufacturing constraint parameters, actively superimposing these parameters onto a zero-stress ideal 3D model, and driving the zero-stress ideal 3D model to undergo displacement to generate a thermal stress simulation model includes: obtaining preset thermal property parameters of the metal material and a preset printing process vector as physical manufacturing constraint parameters; wherein, the thermal property parameters of the metal material include the coefficient of thermal expansion and latent heat of phase change; the printing process vector includes laser power, scanning path, and layer thickness; a built-in thermodynamic active simulation engine is used to convert the thermal property parameters of the metal material and the printing process vector into a physical parameter vector; the physical parameter vector is used as an interference factor and actively superimposed onto the zero-stress ideal 3D model; and the thermodynamic active simulation engine drives the voxels in the zero-stress ideal 3D model to undergo displacement, thereby generating a thermal stress simulation model.

[0041] This embodiment is a further specification of the parametric injection and simulation implementation mechanism. For the application scenario of thermophysical risk assessment of aero-engine turbine blades, the parametric injection and simulation unit obtains the preset thermal property parameters of the metal material and the preset printing process vector as physical manufacturing constraint parameters. The thermal property parameters of the metal material include the coefficient of thermal expansion and the latent heat of phase change, and the printing process vector includes the laser power, scanning path and layer thickness.

[0042] The built-in thermodynamic active simulation engine of the parameterized injection and simulation unit transforms the thermal property parameters of the metal material and the printing process vector into physical parameter vectors, and actively superimposes the physical parameter vectors as interference factors into the zero-stress ideal three-dimensional model. In the process of transforming into physical parameter vectors, the thermodynamic active simulation engine specifically obtains the unit linear energy by calculating the quotient of the laser power and the scanning speed derived from the scanning path, maps the unit linear energy with the layer thickness into a three-dimensional volume heat source, and calculates the local thermal expansion strain vector of the corresponding mesh area as the physical parameter vector based on the thermal expansion coefficient and the latent heat of phase change.

[0043] Specifically, when mapping unit linear energy to a three-dimensional volumetric heat source based on layer thickness, a Gaussian heat source distribution model is used to calculate the local heat flux density. The calculation formula is:

[0044]

[0045] in, For laser power, For the effective spot radius, This represents the radial distance from the voxel node to the laser center. Represented by natural constant An exponential function with base 0. Pi; furthermore, when calculating the local thermal expansion strain vector based on the coefficient of thermal expansion and latent heat of phase change, the engine calculates the node temperature rise based on the local heat flux density, material density, material specific heat capacity, and the effective action time of the laser scan over the voxel. The specific calculation formula is as follows:

[0046]

[0047] in, This represents the actual heated surface area of ​​the voxel. The effective action time of laser scanning. Density of metallic materials Voxel volume The specific heat capacity of the material; when the node temperature rises Before the phase transition temperature is reached, the strain vector magnitude is ,in, The coefficient of thermal expansion of the aforementioned metallic material; when the node temperature rises... When crossing the phase transition temperature, it is necessary to introduce equivalent temperature compensation for the latent heat of phase transition. At this time, the strain vector magnitude is corrected as follows:

[0048]

[0049] in, As the latent heat of phase transition, the strain vector expands along the normal direction of the local temperature gradient. Finally, the voxels in the zero-stress ideal 3D model are driven to displace using a thermodynamic active simulation engine to generate a thermal stress simulation model. When driving the voxels to displace, the simulation engine applies this physical parameter vector as the initial strain field to the 3D topological nodes of the zero-stress ideal 3D model, and uses the finite element equilibrium iterative algorithm to solve the total stiffness matrix of the structure, obtaining the 3D displacement change of each voxel node. This step transforms the invisible physical process in the metal printing process into a quantifiable digital displacement vector, providing the system with a comparison benchmark dominated by thermophysical laws, so that subsequent defect identification is based on a rigorous thermodynamic causal chain.

[0050] In this embodiment, the method for acquiring measured 3D point cloud data, calculating the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generating a real residual field includes: collecting historical in-situ 3D scan point cloud data of the same topological structure as measured 3D point cloud data; aligning the measured 3D point cloud data with the zero-stress ideal 3D model in spatial coordinates; calculating the displacement vector difference between the aligned measured 3D point cloud data and the zero-stress ideal 3D model at corresponding spatial nodes to generate a real residual vector field as the real residual field; wherein, the real residual field includes real thermal deformation characteristics and environmental noise characteristics.

[0051] This embodiment further specifies the real-world residual field extraction method. In the application scenario of processing measured data backtracking and defect analysis of aero-engine turbine blades, the dual-track differential extraction unit collects historical in-situ 3D scanning point cloud data of similar topological structures as measured 3D point cloud data, and aligns the measured 3D point cloud data with the zero-stress ideal 3D model in spatial coordinates. During spatial coordinate alignment, the system extracts the surface feature points of the zero-stress ideal 3D model and uses the iterative nearest point algorithm for rigid registration. By continuously iterating and calculating the minimum Euclidean distance from each point in the measured 3D point cloud data to the set of feature points on the model surface, and solving the optimal rotation and translation transformation matrix according to the least squares method, the system continues until the difference of the objective function between two iterations is less than the preset convergence threshold, thereby completing the macroscopic pose unification. After the alignment operation is completed, since the measured 3D point cloud data is essentially a scattered and disordered set of points, while the zero-stress ideal 3D model has been reconstructed into a continuous non-uniform rational B-spline surface or high-precision voxels, there is objectively no natural one-to-one correspondence between the two nodes.

[0052] Therefore, the dual-track differential extraction unit uses a K-tree nearest neighbor search algorithm to map the aligned, disordered measured 3D point cloud data to the corresponding voxel nodes of the discretized zero-stress ideal 3D model. Specifically, it uses a K-tree search to find the nearest neighbor node to the center of the target voxel. The measured point cloud coordinates, among which The value is a preset positive integer, and this is calculated using an inverse distance weighted interpolation algorithm. The weighted average displacement value of each point cloud coordinate is assigned to the corresponding voxel node to complete the structured resampling of the data, thereby establishing a one-to-one spatial node relationship.

[0053] The dual-track differential extraction unit calculates the displacement vector difference between the resampled and aligned measured 3D point cloud data and the zero-stress ideal 3D model at the corresponding spatial nodes, generating a real residual vector field as the real residual field. The real residual field not only records the real shrinkage and collapse of the blade due to heat, i.e., the real thermal deformation characteristics, but also inevitably mixes in the optical distortion noise of the scanner lens and the small posture deviation of the operator when placing the parts, i.e., the environmental noise characteristics.

[0054] By establishing a rigorous mapping and calculating the displacement vector difference, the dual-track differential extraction unit objectively and completely captures the deformation information fed back from the physical world, eliminates the logical breaks caused by data structure mismatch, preserves the full picture of physical distortion in the real manufacturing environment, and provides underlying data support for subsequent multi-dimensional data comparison.

[0055] In this embodiment, the method for simultaneously calculating the difference between the thermal stress simulation model and the zero-stress ideal three-dimensional model to generate the theoretical residual field includes: extracting the first three-dimensional coordinates of each voxel in the thermal stress simulation model; extracting the second three-dimensional coordinates of the corresponding voxels in the zero-stress ideal three-dimensional model; deriving the purely theoretical stress deformation characteristics between the thermal stress simulation model and the zero-stress ideal three-dimensional model based on thermophysical equations; calculating the vector difference between the first three-dimensional coordinates and the second three-dimensional coordinates to generate a theoretical residual vector field, which serves as the theoretical residual field.

[0056] This embodiment is a further specification of the theoretical residual field extraction logic; in the pure physical deduction application scenario of establishing the standard deformation tolerance baseline, the dual-track differential extraction unit extracts the first three-dimensional coordinates of each voxel in the thermal stress simulation state model, and at the same time extracts the second three-dimensional coordinates of the corresponding voxel in the zero-stress ideal three-dimensional model.

[0057] The dual-track differential extraction unit derives the pure theoretical stress deformation characteristics between the thermal stress simulation model and the zero-stress ideal 3D model based on thermophysical equations. In the step of deriving the pure theoretical stress deformation characteristics based on thermophysical equations, the specific derivation mechanism is as follows: using voxels as the basic calculation unit, the dynamic temperature field gradient in the 3D printing process is input, and the theoretical thermal strain tensor of each voxel is calculated according to the isotropic linear elastic thermal stress constitutive equation. By solving the entire deformation field spatially, the overall macroscopic deformation deviation state of the model is obtained, thereby removing the complex external force interference in actual processing.

[0058] In the specific discretization calculation, since the zero-stress ideal three-dimensional model has been divided into discrete high-precision voxels, the above spatial integration solution is equivalently transformed into summing the finite element matrix over all voxel nodes; for any nth voxel node... Individual elements, among which It is a positive integer, and , The total number of voxels in a zero-stress ideal three-dimensional model, and its theoretical thermal strain tensor. Based on the theory of isotropic thermal expansion, the calculation is as follows:

[0059]

[0060] in, This represents the dynamic temperature rise of the voxel. The coefficient of thermal expansion is... It is the identity matrix, and the specific dimensions of the identity matrix are... The system constructs a global stiffness matrix. With global thermal load vector :

[0061]

[0062] in, The integral symbol is used. For the first Spatial integral region of individual elements; For the first Strain-displacement matrix of individual elements, superscript Represents matrix transpose. This is the transpose of the strain-displacement matrix, used to map nodal displacements to strains within the element. The elastic modulus of the material obtained above Compared with Poisson The specific expression for the constructed isotropic linear elasticity matrix is ​​as follows:

[0063]

[0064] in, The volume is a voxel; the solution is obtained by solving a system of linear equations. ,in, The global stiffness matrix. This is a global displacement vector that includes the displacements of all nodes. The global thermal load vector is used to obtain the overall macroscopic deformation deviation state of the model; the dual-track differential extraction unit calculates the vector difference between the first three-dimensional coordinate and the second three-dimensional coordinate to generate the theoretical residual vector field as the theoretical residual field;

[0065] To accurately quantify this theoretical spatial deviation, the dual-track differential extraction unit calculates the theoretical deformation deviation vector for any voxel within the model. The calculation logic involves directly subtracting the first three-dimensional coordinate from the second three-dimensional coordinate, i.e., subtracting the second three-dimensional coordinate from the first three-dimensional coordinate to obtain the theoretical deformation deviation vector. The first three-dimensional coordinate marks the spatial position of the voxel after being driven by the thermodynamic active simulation engine, while the second three-dimensional coordinate marks the original spatial position of the voxel under an absolutely isothermal state without thermal stress. The theoretical residual field generated by this mechanism is immune to external physical environment interference and serves as an ideal digital field, providing a reliable theoretical reference for subsequent identification of genuine and fake defects.

[0066] In this embodiment, the method for calculating the topological three-dimensional similarity between the actual residual field and the theoretical residual field includes: extracting the first deformation tensor of the actual residual field; extracting the second deformation tensor of the theoretical residual field; and calculating the Hausdorff distance between the sets of displacement vectors corresponding to the first deformation tensor and the second deformation tensor.

[0067] Calculate the cosine similarity of the deformation tensor angle between the first deformation tensor and the second deformation tensor; fuse Hausdorff distance and the cosine similarity of the deformation tensor angle to generate a topological 3D similarity.

[0068] This embodiment is a further concretization of the topological three-dimensional similarity calculation model; in the application scenario of authenticity verification and anomaly identification of multi-dimensional spatial deformation features, the coupled decision and reverse compensation unit extracts the first deformation tensor of the real residual field, and at the same time extracts the second deformation tensor of the theoretical residual field.

[0069] Specifically, given that the underlying data structure of both the real and theoretical residual fields is a displacement vector field, the system transforms the displacement vectors into a first and second deformation tensor reflecting the local deformation state by taking partial derivatives and calculating symmetric gradients in a three-dimensional spatial coordinate system, thus establishing a unified tensor computation basis. The coupled decision and inverse compensation units calculate the Hausdorff distance between the first and second deformation tensors and the cosine similarity of the deformation tensor angle between them. The Hausdorff distance and the cosine similarity of the deformation tensor angle are then fused to generate a topological three-dimensional similarity. This similarity aims to accurately evaluate both spatial scale and deformation direction, and its calculation logic is as follows:

[0070]

[0071] in, For topological 3D similarity, Distance weights For direction weights, This is the distance attenuation coefficient, whose dimension is the reciprocal of length, to ensure that the input parameters of the exponential function are dimensionless. Let be the Hausdorff distance between the sets of neighborhood displacement vectors corresponding to the first deformation tensor and the second deformation tensor. For the first feature point set, For the second feature point set, For the first deformation tensor, For the second deformation tensor, and These are the tensor norms of the first deformation tensor and the second deformation tensor, respectively;

[0072] In this calculation process, to clarify the distance and similarity measurement rules between tensors, since the deformation tensor physically represents a structured set of deformation states of multiple spatial nodes within a corresponding region, an appropriate data flow logic needs to be adopted when performing measurements in different dimensions; and in calculating Hausdorff distance... To map the deformation tensor features back to the physical distance dimension, the system extracts all displacement vectors within the voxel nodes corresponding to the first and second deformation tensors and their preset three-dimensional neighborhood spaces, forming a set of displacement vectors, which are then used as the first feature point set. With the second feature point set Calculate the point set any vector to a set of points The maximum value of the shortest distance, and the point set Any vector to a set of points The maximum value of the shortest distance is taken as the larger of the two values ​​as the Hausdorff distance, which measures the most extreme mismatch between the two sets of tensors on the local spatial distribution contour.

[0073] In calculating tensor dot product At that time, the system strictly maintains the ordered matrix structure of the first and second deformed tensors as high-dimensional arrays, and calculates them using the Frobenius inner product rule, that is, multiplying the corresponding elements in the first and second deformed tensors one by one and then summing them; tensor norm and Similarly, the calculation is based on the ordered matrix structure and uses the Frobenius norm rule, which involves taking the square root of the sum of squares of all elements in the tensor; where the distance weights... with direction weight Satisfy normalization constraints ;

[0074] To ensure the adaptability of the similarity metric, the system does not use a fixed constant, but dynamically calculates the weights based on the signal-to-noise ratio of the real residual field: the variance of the Frobenius norm of the first deformation tensor in its neighborhood is calculated as the local volatility, and when the local volatility... Greater than the preset smoothing threshold When the system determines that the high-frequency noise in the region is significant, it dynamically calculates and reduces the directional weights based on the attenuation function. Its function expression is:

[0075]

[0076] in, The initial direction weights are set. This is the attenuation adjustment coefficient. For local volatility, A preset smoothing threshold is set; this threshold is obtained based on the baseline white noise variance calibration at the scanner's factory settings; to prevent excessive weight amplification that could lead to negative distance weights, an upper limit for direction weights is set. The actual directional weights are:

[0077]

[0078] in, This represents the function that takes the minimum value; the system is based on normalization constraints. Increase distance weight accordingly This is to reduce the interference of local directional mutations on the overall similarity evaluation;

[0079] Conversely, if the local volatility is less than or equal to the preset smoothing threshold, the directional weight is increased. To enhance sensitivity to deformation trends, this computational model overcomes the limitations of the single distance threshold judgment method by accurately splitting and independently operating the underlying data structure, while eliminating the problem of self-contradictory parameter types. Even if there is slight translational noise in the overall real point cloud data, as long as its local deformation and distortion trend is consistent with the thermodynamic theory, the coupled decision and reverse compensation unit can still identify regional similarity, demonstrating the superiority of the multi-dimensional feature cascade mechanism in noise resistance.

[0080] In this embodiment, determining whether the topological three-dimensional similarity meets a preset overlap condition; when the topological three-dimensional similarity meets the preset overlap condition, determining the corresponding region as a true thermal stress pain point region includes: setting a preset similarity threshold as a preset overlap condition; comparing the topological three-dimensional similarity with the preset similarity threshold; when the topological three-dimensional similarity is greater than or equal to the preset similarity threshold, determining that the deformation vectors of the actual residual field and the theoretical residual field in the corresponding region are highly overlapped, thus meeting the preset overlap condition; in response to meeting the preset overlap condition, removing scanner noise and geometric visual errors in the actual residual field, and determining the corresponding region as a true thermal stress pain point region.

[0081] This embodiment is a further specification of the real thermal stress pain point area judgment mechanism; in the real thermal stress defect confirmation application scenario, the coupling judgment and reverse compensation unit sets a preset similarity threshold as a preset overlap condition. The basis for setting this threshold is to perform statistical analysis on the topological similarity distribution of a large number of historical samples and extract the lower limit of the confidence interval.

[0082] The specific logic for extracting the lower limit of the confidence interval is as follows: collect the topological three-dimensional similarity data of real thermal stress defect samples and pure scanning equipment noise samples that have been manually verified through multiple rounds of verification; fit the similarity distribution curve of the real defect samples using a probability density function; and select the value corresponding to the cumulative probability reaching 95% confidence level on one side as the preset similarity threshold; compare the topological three-dimensional similarity with the preset similarity threshold using the coupled decision and reverse compensation units; when the topological three-dimensional similarity is greater than or equal to the preset similarity threshold, it is determined that the deformation vectors of the actual residual field and the theoretical residual field in the corresponding region are highly coincident, thus satisfying the preset coincidence condition;

[0083] In response to meeting the preset overlap conditions, the coupling decision and reverse compensation unit actively removes scanner noise and geometric visual errors from the real residual field, determining the corresponding area as the true thermal stress pain point area. When actively removing scanner noise and geometric visual errors from the real residual field, the unit specifically clears the data of the environmental noise displacement vector corresponding to the area below the threshold in physical memory or assigns it a zero-weight label, so that subsequent compensation is driven only by verified real defect data. By setting a strict similarity threshold for double verification, the coupling decision and reverse compensation unit requires that the identified defects must find supporting evidence in thermodynamic theory deduction, solving the technical problem that traditional static geometric inspection software is prone to falsely reporting defects and improving the accuracy of defect feature extraction.

[0084] In this embodiment, the method of inverting the theoretical residual field and superimposing it onto the zero-stress ideal 3D model to generate a pre-inversely compensated 3D printing model includes: extracting the theoretical residual field corresponding to the true thermal stress pain point region; performing an inversion operation on the theoretical residual field to generate an inverse deformation compensation vector field; directly superimposing the inverse deformation compensation vector field onto the spatial nodes corresponding to the zero-stress ideal 3D model; reconstructing the 3D topology based on the superimposed spatial node coordinates to generate a pre-inversely compensated 3D printing model, thus completing closed-loop modeling.

[0085] This embodiment further specifies the reverse compensation execution process. For the application scenario of performing directional digital compensation for verified deformation gaps, the coupling decision and reverse compensation unit extracts the theoretical residual field corresponding to the true thermal stress pain point area, performs an inversion operation on the theoretical residual field to generate a reverse deformation compensation vector field, and directly superimposes the reverse deformation compensation vector field onto the spatial nodes corresponding to the zero-stress ideal three-dimensional model. Based on the coordinates of the superimposed spatial nodes, the three-dimensional topology is reconstructed to generate a pre-reverse-compensated three-dimensional printing model.

[0086] The coupling decision and inverse compensation unit uses a pure theoretical residual field for inverse compensation, aiming to achieve inverse compensation of digital model deviations to physical real deformations. The use of a theoretical residual field instead of a real residual field with mixed measurement noise for compensation ensures that the compensation amount conforms to the first principles of physics, avoids introducing environmental noise into the model, and improves the surface finish and dimensional accuracy of the compensated model.

[0087] In this embodiment, when the topological three-dimensional similarity does not meet the preset overlap condition, the corresponding region is determined to be a noise region, and the residual data of the corresponding region is removed to maintain the original state of the zero-stress ideal three-dimensional model. This includes: when the topological three-dimensional similarity is less than the preset similarity threshold, it is determined that the actual residual field and the theoretical residual field do not overlap in the corresponding region and do not meet the preset overlap condition; in response to not meeting the preset overlap condition, it is determined that the deformation of the corresponding region is caused by non-thermal stress physical factors and is marked as a noise region; the theoretical residual field and the actual residual field corresponding to the noise region are shielded, and the original geometric topology of the zero-stress ideal three-dimensional model is retained in the corresponding region.

[0088] This embodiment further specifies the noise region processing mechanism. In the application scenario of intercepting false alarms caused by mechanical collisions due to non-thermal physical factors, when the topological three-dimensional similarity is less than the preset similarity threshold, the coupling decision and reverse compensation unit determines that the actual residual field and the theoretical residual field do not coincide in the corresponding region, and determines that the preset coincidence condition is not met. In response to the failure to meet the preset coincidence condition, the coupling decision and reverse compensation unit determines that the deformation of the corresponding region is caused by non-thermal stress physical factors, such as mechanical collision deformation caused by the operator accidentally dropping the part before scanning or systematic error caused by scanner calibration drift, and marks it as a noise region.

[0089] The coupled decision and reverse compensation unit forcibly shields the theoretical and actual residual fields corresponding to the noise region, preserving the original geometric topology of the zero-stress ideal three-dimensional model in the corresponding region. This mechanism provides a safety strategy to prevent overcompensation, resolutely identifying deformations that do not conform to thermodynamic deduction as noise and shielding them, protecting the original design intent of non-thermal stress sensitive areas, and verifying the stability of the system when dealing with massive complex structures.

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A metal 3D printing model three-dimensional modeling system, characterized in that, include: Ideal reference construction unit: Obtain initial ideal geometric boundary data, construct a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generate a zero-stress ideal three-dimensional model; Parametric injection and simulation unit: acquires physical manufacturing constraint parameters, actively superimposes the physical manufacturing constraint parameters onto the zero-stress ideal three-dimensional model, drives the zero-stress ideal three-dimensional model to undergo displacement, and generates a thermal stress simulation state model; Dual-track differential extraction unit: acquires measured 3D point cloud data, calculates the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generates a real residual field; simultaneously calculates the difference between the thermal stress simulation model and the zero-stress ideal 3D model, and generates a theoretical residual field; Coupled decision and reverse compensation unit: Calculates the topological three-dimensional similarity between the actual residual field and the theoretical residual field; Determine whether the topological 3D similarity meets the preset overlap condition; when the topological 3D similarity meets the preset overlap condition, determine that the corresponding region is a true thermal stress pain point region, invert the theoretical residual field and superimpose it onto the zero-stress ideal 3D model to generate a pre-compensated 3D printing model; when the topological 3D similarity does not meet the preset overlap condition, determine that the corresponding region is a noise region, and remove the residual data of the corresponding region to maintain the original state of the zero-stress ideal 3D model.

2. The metal 3D printing model three-dimensional modeling system according to claim 1, wherein, The method for obtaining initial ideal geometric boundary data, constructing a pure mathematical ideal geometric field based on the initial ideal geometric boundary data, and generating a zero-stress ideal three-dimensional model includes: parsing a preset three-dimensional design file to extract the initial ideal geometric boundary data; reconstructing the initial ideal geometric boundary data based on first principles using a preset non-uniform rational B-spline surface or high-precision voxels; constructing the pure mathematical ideal geometric field under preset ideal physical boundary conditions of no gravity, absolute isotherm, and no thermal stress; and using the pure mathematical ideal geometric field as the reference origin for difference calculation to generate the zero-stress ideal three-dimensional model.

3. The metal 3D printing model three-dimensional modeling system of claim 2, wherein, The method of acquiring physical manufacturing constraint parameters, actively superimposing these parameters onto the zero-stress ideal 3D model, and driving the zero-stress ideal 3D model to undergo displacement to generate a thermal stress simulation model includes: acquiring preset thermal property parameters of the metal material and a preset printing process vector as the physical manufacturing constraint parameters; wherein, the thermal property parameters of the metal material include the coefficient of thermal expansion and latent heat of phase change; the printing process vector includes laser power, scanning path, and layer thickness; a built-in thermodynamic active simulation engine converts the thermal property parameters of the metal material and the printing process vector into a physical parameter vector; the physical parameter vector is used as an interference factor and actively superimposed onto the zero-stress ideal 3D model; and the thermodynamic active simulation engine drives the voxels in the zero-stress ideal 3D model to undergo displacement, thereby generating the thermal stress simulation model.

4. The metal 3D printing model three-dimensional modeling system according to claim 3, characterized in that, The method for acquiring measured 3D point cloud data, calculating the difference between the measured 3D point cloud data and the zero-stress ideal 3D model, and generating a real residual field includes: collecting historical in-situ 3D scan point cloud data of the same topology as the measured 3D point cloud data; aligning the measured 3D point cloud data with the zero-stress ideal 3D model in spatial coordinates; calculating the displacement vector difference between the aligned measured 3D point cloud data and the zero-stress ideal 3D model at corresponding spatial nodes to generate a real residual vector field as the real residual field; wherein, the real residual field includes real thermal deformation characteristics and environmental noise characteristics.

5. The metal 3D printing model three-dimensional modeling system according to claim 4, characterized in that, The method for synchronously calculating the difference between the thermal stress simulation model and the zero-stress ideal three-dimensional model to generate a theoretical residual field includes: extracting the first three-dimensional coordinates of each voxel in the thermal stress simulation model; extracting the second three-dimensional coordinates of the corresponding voxels in the zero-stress ideal three-dimensional model; deriving the purely theoretical stress deformation characteristics between the thermal stress simulation model and the zero-stress ideal three-dimensional model based on thermophysical equations; calculating the vector difference between the first three-dimensional coordinates and the second three-dimensional coordinates to generate a theoretical residual vector field, which serves as the theoretical residual field.

6. The three-dimensional modeling system for metal 3D printing models according to claim 5, characterized in that, The method for calculating the topological three-dimensional similarity between the actual residual field and the theoretical residual field includes: extracting a first deformation tensor of the actual residual field; extracting a second deformation tensor of the theoretical residual field; calculating the Hausdorff distance between the first deformation tensor and the second deformation tensor; calculating the cosine similarity of the deformation tensor angle between the first deformation tensor and the second deformation tensor; and fusing the Hausdorff distance and the cosine similarity of the deformation tensor angle to generate the topological three-dimensional similarity.

7. A three-dimensional modeling system for metal 3D printing models according to claim 6, characterized in that, The method for determining whether the topological 3D similarity satisfies a preset overlap condition, and determining the corresponding region as a true thermal stress pain point region when the topological 3D similarity satisfies the preset overlap condition, includes: setting a preset similarity threshold as the preset overlap condition; comparing the topological 3D similarity with the preset similarity threshold; when the topological 3D similarity is greater than or equal to the preset similarity threshold, determining that the deformation vectors of the actual residual field and the theoretical residual field highly overlap in the corresponding region, thus satisfying the preset overlap condition; in response to satisfying the preset overlap condition, removing scanner noise and geometric visual errors in the actual residual field, and determining that the corresponding region is the true thermal stress pain point region.

8. A three-dimensional modeling system for metal 3D printing models according to claim 7, characterized in that, The method of inverting the theoretical residual field and superimposing it onto the zero-stress ideal 3D model to generate a pre-inversely compensated 3D printing model includes: extracting the theoretical residual field corresponding to the true thermal stress pain point region; performing an inversion operation on the theoretical residual field to generate an inverse deformation compensation vector field; directly superimposing the inverse deformation compensation vector field onto the spatial nodes corresponding to the zero-stress ideal 3D model; reconstructing the 3D topology based on the superimposed spatial node coordinates to generate the pre-inversely compensated 3D printing model, thus completing closed-loop modeling.

9. A three-dimensional modeling system for metal 3D printing models according to claim 8, characterized in that, The method of determining the corresponding region as a noise region and removing the residual data of the corresponding region to maintain the original state of the zero-stress ideal three-dimensional model when the topological three-dimensional similarity does not meet the preset overlap condition includes: when the topological three-dimensional similarity is less than the preset similarity threshold, determining that the actual residual field and the theoretical residual field do not overlap in the corresponding region and do not meet the preset overlap condition; in response to not meeting the preset overlap condition, determining that the deformation of the corresponding region is caused by non-thermal stress physical factors and marking it as the noise region; shielding the theoretical residual field and the actual residual field corresponding to the noise region, and retaining the original geometric topology of the zero-stress ideal three-dimensional model in the corresponding region.

Images