Supportless 3D Printing Method, Component, Device and Equipment with Optimized Overhang Structure

Abstract

The present invention discloses a supportless 3D printing method, component, device and equipment with optimized overhang structures. The method includes: identifying overhang structure features of a geometric topology model of a component to be formed to generate target geometric parameters; obtaining process parameters of the overhang structure and inputting them together with the target geometric parameters into a machine learning model to predict matching actual geometric parameters; calculating the deformation amount of the overhang structure based on the actual geometric parameters and the target geometric parameters; performing deformation compensation on the overhang structure according to the deformation amount to generate a compensated and corrected geometric topology model; and applying the corrected model as the target printing model of the component to be formed to supportless 3D printing. The present invention provides a more accurate target printing model for supportless 3D printing, effectively improving the forming quality and printing success rate of overhang structures and realizing high-quality supportless additive manufacturing.

Description

Technical Field

[0001] The present invention relates to the field of additive manufacturing technology, and in particular to a support-free 3D printing method, component, device and equipment for optimizing overhang structures. Background Art

[0002] In the 3D printing process, the design of the support structure plays a vital role in building components with complex shapes. For example, in typical complex components of aerospace, there are a large number of overhanging structures represented by cavities or overhangs. However, when using traditional 3D printing technology, the overhanging structure often faces problems such as warping, powder sticking and slag hanging on the lower surface during the forming process, which greatly reduces the forming accuracy and surface finish, and may even lead to forming failure.

[0003] In traditional 3D printing processes, such as the LPBF (Laser Powder Bed Fusion) process, due to the particularity of the overhanging structure, warping deformation and powder sticking and slag on the lower surface are prone to occur during the forming process. These problems directly affect the forming quality and surface finish of the overhanging structure, especially when the overhang angle is higher than a certain value. In order to overcome these problems, it is usually necessary to add support structures, whose main functions are to conduct heat from the molten pool, reduce deformation caused by residual stress, and resist the force of the scraper. The addition of support ensures the forming of the overhanging structure to a certain extent, but it also brings a series of problems, such as increasing the difficulty and time of pre- and post-processing, poor surface quality after removing the support, and reduced raw material utilization.

[0004] Therefore, there is an urgent need in the prior art for a support-free 3D printing process that can ensure the forming accuracy and surface quality of the overhang structure, so as to achieve high-quality support-free additive manufacturing of the overhang structure. Summary of the invention

[0005] Embodiments of the present invention provide a support-free 3D printing method, component, device and equipment for optimizing overhang structures, which are used to achieve high-quality support-free additive manufacturing of overhang structures.

[0006] To achieve the above object, the embodiments of the present invention adopt the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a supportless 3D printing method with optimized overhang structure. The method includes: performing feature recognition on a geometric topology model of a component to be formed, which includes at least an overhang structure, to generate target geometric parameters of the overhang structure; obtaining process parameters of the overhang structure and using them together with the target geometric parameters as inputs, and predicting and outputting matching actual geometric parameters through a machine learning model, where the machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters, and actual geometric parameters of the overhang structure; calculating the deformation amount of the overhang structure based on the actual geometric parameters and the target geometric parameters; performing deformation compensation on the overhang structure according to the deformation amount to generate a compensated and corrected geometric topology model; and using the compensated and corrected geometric topology model as the target printing model of the component to be formed for supportless 3D printing.

[0008] In an alternative embodiment of the first aspect, the performing feature recognition on a geometric topology model of a component to be formed, which includes at least an overhang structure, includes: separately extracting a contour and a skeleton from the obtained geometric topology model of the component to be formed; performing reverse fitting on the contour and the skeleton; and obtaining features including at least the overhang structure according to the reverse fitting result.

[0009] In an alternative embodiment of the first aspect, the obtaining features including at least the overhang structure according to the reverse fitting result includes: obtaining a structure with an overhang angle greater than a preset threshold from the reverse fitting result to confirm it as the overhang structure.

[0010] In an alternative embodiment of the first aspect, the target geometric parameters and the actual geometric parameters include at least one of geometric configuration dimensions, overhang angle, aspect ratio, span, and constraint conditions.

[0011] In an alternative embodiment of the first aspect, the process parameters include at least one of laser power, scanning strategy, spot diameter, and powder layer thickness.

[0012] In an alternative embodiment of the first aspect, the performing deformation compensation on the overhang structure according to the deformation amount includes: dividing the geometric topology model into triangular patches as units; and separately performing independent overhang structure deformation compensation on each triangular patch according to the deformation amount.

[0013] In an alternative embodiment of the first aspect, the machine learning model indicates the mapping relationship between the target geometric parameters, process parameters, actual geometric parameters, process parameters, and forming quality of the overhang structure.

[0014] In an alternative embodiment of the first aspect, the method further includes: using at least the target geometric parameters of the overhanging structure as input, predicting and outputting optimized process parameters matching the overhanging structure through the machine learning model; and using the optimized process parameters as the target process parameters of the overhanging structure for support-free 3D printing.

[0015] In an alternative embodiment of the first aspect, the using at least the target geometric parameters of the overhanging structure as input includes: using the target geometric parameters of the overhanging structure and the forming quality as input.

[0016] In an alternative embodiment of the first aspect, the predicting and outputting optimized process parameters matching the overhanging structure through the machine learning model includes: dividing the angle region to which it belongs according to the feature recognition result of the overhanging structure, and using the divided angle region as at least one of the target geometric parameters of the overhanging structure as input; predicting and outputting optimized process parameters matching the angle region to which the overhanging structure belongs through the machine learning model.

[0017] In an alternative embodiment of the first aspect, there are at least three types of angle regions, namely Q1, Q2, and Q3.

[0018] In an alternative embodiment of the first aspect, Q1 < 20°, 20° ≤ Q2 < 35°, Q3 ≥ 35°.

[0019] In an alternative embodiment of the first aspect, the process parameters include at least one of the molten pool size and morphology, molten pool temperature distribution, and stress distribution.

[0020] In an alternative embodiment of the first aspect, the forming quality includes at least one of defect characteristics, roughness, and dimensional accuracy.

[0021] In a second aspect, an embodiment of the present invention provides a component printed according to the method described in any one of the first aspect.

[0022] In a third aspect, an embodiment of the present invention provides a supportless 3D printing device, including a model correction unit and a 3D printing unit; wherein, the model correction unit includes: a feature recognition unit configured to perform feature recognition on the geometric topology model of the to-be-formed component, including at least overhang structures, to generate target geometric parameters of the overhang structures; a data prediction unit configured to obtain process parameters of the overhang structures and use them together with the target geometric parameters as inputs, and predict and output matching actual geometric parameters through a machine learning model, where the machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters, and actual geometric parameters of the overhang structures; a deformation amount calculation unit configured to calculate the deformation amount of the overhang structures based on the actual geometric parameters and the target geometric parameters; a deformation compensation unit configured to perform deformation compensation on the overhang structures according to the deformation amount to generate a compensated and corrected geometric topology model; wherein, the 3D printing unit is configured to use the compensated and corrected geometric topology model as the target printing model of the to-be-formed component for supportless 3D printing.

[0023] In a fourth aspect, an embodiment of the present invention provides an electronic device, including: at least one processor; at least one memory, the at least one memory being coupled to the at least one processor and configured to store instructions executed by the at least one processor, and when the instructions are executed by the at least one processor, the electronic device is caused to execute the method according to any one of the first aspect.

[0024] In a fifth aspect, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the method according to any one of the first aspect is implemented.

[0025] In a sixth aspect, an embodiment of the present invention provides a computer program product including computer-executable instructions, and when the computer-executable instructions are executed by a processor, the computer is caused to implement the method according to any one of the first aspect.

[0026] Based on the above solutions, the supportless 3D printing method provided by the embodiments of the present invention realizes the accurate prediction of the actual geometric parameters of overhang structures through the feature recognition of overhang structures and the application of a machine learning model. By calculating and correcting the deformation amount, the deformation problem of overhang structures during the printing process is effectively solved. By intelligently predicting and outputting optimized process parameters that match the overhang structures through the machine learning model, the forming quality and surface finish of supportless 3D printing are effectively improved. This method combines advanced 3D printing technology and intelligent machine learning means, providing high controllability for the manufacturing of components under supportless conditions, and finally achieving high-quality and high-efficiency 3D printing in a supportless state. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0028] Figure 1 Shows an exemplary form of the overhang structure, where (a) is the cavity form and (b) is the suspended form;

[0029] Figure 2 Shows an exemplary image of the forming defect of the overhang structure under the traditional 3D printing process, where (a) is the phenomenon of powder sticking and slag hanging, and (b) is the phenomenon of warping deformation;

[0030] Figure 3 Shows an exemplary form of the overhang structure with supports added, where (a) is the cavity form with supports added and (b) is the suspended form with supports added;

[0031] Figure 4 Shows a schematic flow diagram of the supportless 3D printing method provided by the embodiments of the present invention;

[0032] Figure 5 Shows a schematic process diagram of inverse fitting of the contour and the skeleton provided by the embodiments of the present invention, where (a) is before inverse fitting and (b) is after inverse fitting;

[0033] Figure 6 Shows a schematic structural diagram of the neural network model provided by the embodiments of the present invention;

[0034] Figure 7 Shows a schematic environmental diagram of the neural network model provided by the embodiments of the present invention;

[0035] Figure 8 Shows a schematic diagram of the correction of the overhang angle provided by the embodiments of the present invention;

[0036] Figure 9 Shows another schematic flow diagram of the supportless 3D printing method provided by the embodiments of the present invention;

[0037] Figure 10 Shows a schematic application structure diagram of the neural network model provided by the embodiments of the present invention;

[0038] Figure 11 Shows a schematic prediction flow diagram of a process parameter provided by the embodiments of the present invention;

[0039] Figure 12 Shows a schematic diagram of the matching between the overhanging structure and process parameters at a specific angle provided by an embodiment of the present invention;

[0040] Figure 13 Shows a schematic block diagram of a supportless 3D printing device provided by an embodiment of the present invention;

[0041] Figure 14 Shows a schematic block diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0042] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals indicate the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present invention, and should not be construed as limiting the present invention.

[0043] Figure 1 Shows an exemplary form of the overhanging structure, Figure 2 Shows the forming defect phenomenon of the overhanging structure under the traditional 3D printing process. As Figure 1 shown, the overhanging structure refers to an area with cavities or overhangs in a component, which has a special geometric form and usually shows fewer or no support points across the space. The morphological characteristics of the overhanging structure are of great significance for constructing lightweight, complex and functional components. As Figure 2 shown, in the traditional 3D printing process, such structures are prone to forming problems, including warping deformation, powder sticking and slag hanging on the lower surface, etc., seriously reducing the forming accuracy and surface finish of the overhanging structure, and even leading to forming failure in severe cases. Therefore, special treatment methods are required to optimize its manufacturing process.

[0044] Figure 3 Shows an exemplary form of the overhanging structure with added support. As Figure 3 shown, in 3D printing, in order to solve problems such as warping deformation and powder sticking and slag hanging on the lower surface that may occur during the forming process of the overhanging structure, it is often necessary to add support structures. These support structures usually provide stable support by adding temporary support materials under the overhanging structure to prevent the component from deforming or having other quality problems during the forming process. The process of adding support is usually configured in 3D printing software to achieve support customization and optimization for the overhanging structure. However, although the support structure can improve the forming success rate to a certain extent, it also introduces a series of problems, such as increasing the difficulty and time of pre- and post-processing, poor surface quality after removing the support, decreased raw material utilization rate, and difficult removal of support for complex overhanging structures such as internal cavities, resulting in inability to form.

[0045] In view of this, embodiments of the present invention aim to provide a supportless 3D printing method for the particularity of the overhanging structure to achieve high-quality forming of the overhanging structure.

[0046] Figure 4 The flowchart of the supportless 3D printing method provided by the embodiments of the present invention is shown. As Figure 4 shown, the supportless 3D printing method 100 provided in some embodiments of the present invention includes steps 101-105, etc.

[0047] 101. Perform feature recognition on the geometric topology model of the component to be formed, including at least the overhanging structure, and generate the target geometric parameters of the overhanging structure.

[0048] The geometric topology model is a mathematical model used to represent the shape and structure of the component to be formed, serving as the basis for constructing the component during the 3D printing process. This model describes the geometric shape, size, and internal structure of the component in a mathematical way, and it can take various forms, such as three-dimensional point clouds or CAD models, etc. In the geometric topology model, each point, edge, and face has specific geometric attributes, describing the spatial form of the component. Points represent the vertices or specific positions of the component, edges represent the line segments connecting two points, and faces represent the planar regions formed by the boundaries, that is, this model can comprehensively depict the external shape and internal structure of the component.

[0049] Performing feature recognition on the geometric topology model, including at least the overhanging structure, is to analyze the model to find parts including but not limited to the overhanging structure (other non-overhanging structure features can also be recognized in addition to the overhanging structure), for example, by detecting specific shape features, spatial distributions, or other geometric attributes. In the recognition of the overhanging structure, a series of geometric attributes can be considered, such as cavities, suspended areas, areas with fewer or no support points, etc. By analyzing the geometric shape, topological structure, and possible geometric features of the overhanging structure of the model, to determine the position and form of the overhanging structure and generate the corresponding target geometric parameters.

[0050] The target geometric parameters refer to the geometric features and parameters that are expected for the component to finally take shape during the supportless 3D printing process, such as requirements in aspects including the shape, size, angle, etc. of the component, and are also the characteristics of the initial geometric topology model.

[0051] In some embodiments, the target geometric parameters include at least one of geometric configuration dimensions, overhanging angle, aspect ratio, span, and constraint conditions, preferably including all of these parameters.

[0052] Geometric configuration dimensions refer to the dimensions of the component in all directions in three-dimensional space, including length, width, and height, etc. For example, it can cover the overall size, shape of the component, as well as the relative positions and proportional relationships between its various parts.

[0053] The overhang angle refers to the angle between the overhanging structure and the horizontal plane or the vertical plane. To unify the standard, the present invention preferably refers to the former, that is, the angle between the overhanging structure and the horizontal plane. For the overhanging structure, its overhang angle determines the inclination degree of the structure.

[0054] The aspect ratio is the ratio between the length and the width of the component, which reflects the relative dimensional relationship of the component in two main directions.

[0055] The span refers to the spatial distance spanned by the overhanging structure, that is, the maximum dimension of the structure in the horizontal direction, which describes the lateral extension degree of the overhanging structure.

[0056] The constraint condition refers to the restriction or support (non-printing support) received by the overhanging structure in a specific direction. For example, it includes fixed supports at the bottom or side of the structure, or other forms of external constraints, which have an important impact on the shape and stability of the overhanging structure.

[0057] In one implementation manner of performing feature recognition including at least the overhanging structure on the geometric topology model of the to-be-formed component, first, the contour and the skeleton are respectively extracted from the obtained geometric topology model of the to-be-formed component, then the extracted contour and skeleton are reversely fitted, and finally, features including at least the overhanging structure are obtained according to the reverse fitting result.

[0058] The contour reflects the edge contour of the component's outer shape and is the projection boundary of the component on the two-dimensional plane. During the extraction process of the contour, for example, methods of graphics and computational geometry are used. By analyzing the surface information of the geometric topology model, the outer contour boundary of the component is detected and extracted to construct the contour line, thereby forming the overall contour of the component. The skeleton, as the main internal support structure of the component, is an abstract representation used to describe the main shape of the component. During the extraction process of the skeleton, for example, methods such as mathematical morphology and skeletonization algorithms are used. By analyzing the voxel or point cloud data of the geometric topology model, the skeleton structure of the component is extracted. During this process, by analyzing the internal points of the geometric topology model, unnecessary detail information is removed, and the main support structure is retained to form the skeleton.

[0059] Figure 5 Shows a schematic diagram of the process of reversely fitting the contour and the skeleton provided by the embodiment of the present invention. As Figure 5As shown, the reverse fitting of the extracted contour and skeleton is to restore a more abstract and simplified representation from the geometric features of the component, so as to better capture the specific shape of the overhang structure. In the reverse fitting of the contour and skeleton, the extracted contour and skeleton information are abstracted into a more simplified geometric representation to better describe the overall shape and internal structure features. Corresponding fitting algorithms (such as Bezier curve / surface fitting, least squares curve / surface fitting, B-spline curve / surface fitting, etc.) can be used to fit the contour edge data and skeleton data, so as to restore the overall shape and internal structure features of the component.

[0060] After the fitting is completed, it is necessary to further extract the key features including special shapes such as overhang structures, that is, analyze the results of the reverse fitting to identify the features of the overhang structure. For example, by analyzing the continuity of the suspended area, the inclination degree of the structure (overhang angle), the distribution of support points, etc., to determine whether there is an overhang structure, and generate corresponding target geometric parameters after identifying the overhang structure.

[0061] In the implementation method of overhang structure feature recognition, the results of reverse fitting provide an estimate of the internal structure of the component. Among them, the key feature is the overhang angle, which reflects the inclination degree of the overhang structure relative to the horizontal plane. By setting a preset threshold, the overhang angle (obtained by calculating the slope) in the reverse fitting results can be screened to confirm whether there is an overhang structure. For example, the structure with an overhang angle greater than the preset threshold obtained from the reverse fitting results is confirmed as an overhang structure. Through this implementation method, the overhang structure in the component (geometric topology model) can be quickly and effectively identified to provide accurate target geometric parameters for subsequent unsupported 3D printing.

[0062] 102. Obtain the process parameters of the overhang structure and use them together with the target geometric parameters as inputs, and predict and output the matching actual geometric parameters through a machine learning model. The machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters and actual geometric parameters of the overhang structure.

[0063] In some embodiments, the process parameters include at least one of laser power, scanning strategy, spot diameter, powder layer thickness, and preferably include all of these parameters.

[0064] The laser power refers to the energy intensity of the laser beam when melting the molten metal powder, which directly affects the temperature and depth of the molten pool. Appropriately adjusting the laser power can control the temperature distribution during the forming process, thereby affecting the melting condition and structural strength of the component.

[0065] The scanning strategy involves the movement path and scanning speed of the laser beam. Different scanning strategies affect the surface quality, forming speed, and energy distribution of the component. Reasonably selecting the scanning strategy can improve the forming accuracy and surface finish of the component.

[0066] The spot diameter represents the diameter of the laser beam at the focus, which constrains the energy density at the focus and thus affects the melting condition of the component.

[0067] The powder spreading layer thickness refers to the thickness of each layer of powder spreading. The selection of the powder spreading layer thickness will affect the forming speed, surface finish, and internal structure of the component. For overhanging structures, reasonable adjustment of the powder spreading layer thickness can reduce warping deformation and the need for support structures.

[0068] It should be understood that in addition to the above-mentioned process parameters such as laser power, scanning strategy, spot diameter, and powder spreading layer thickness, other conventional process parameters involved in the 3D printing process are also covered in the present invention.

[0069] It should be understood that the machine learning model involved in step 102 refers to the model that has been constructed and trained. After receiving the input data (target geometric parameters and process parameters), it predicts the actual geometric parameters by learning the patterns of historical data. The actual geometric parameters also include at least one of the features of geometric configuration dimensions, overhang angle, aspect ratio, span, and constraint conditions, and are the geometric parameters that the overhanging structure can actually achieve under the conditions of process parameters through 3D printing.

[0070] Figure 6 The structural schematic diagram of the neural network model provided by the embodiment of the present invention is shown. Figure 7 The environmental schematic diagram of the neural network model provided by the embodiment of the present invention is shown. Here, taking this machine learning model as an example of a neural network (NeuralNetworks, NN) model, the process of its construction and training is described. This neural network model can specifically be a deep neural network (DNN) model. It should be understood that in addition to the neural network model, other machine learning algorithms such as decision trees, naive Bayes, and support vector machines can also be used, and a suitable model can be selected according to specific requirements.

[0071] Such as Figure 6 - 7As shown in the figure, the environment 200 includes a model construction device 201, an initial neural network model 202, a database 203, a model training device 204, and an applied neural network model 205. In the environment 200, the model construction device 201 is used to define the structure of the neural network, including the number of nodes in the input layer, hidden layer, and output layer, the connection method between layers, and the selection of activation functions, so as to create the initial neural network model 202. The initial neural network model 202 is an untrained model, and its weights and biases are randomly initialized. Therefore, it is necessary to use the model training device 204 to train the initial neural network model 202 using the data stored in the database 203.

[0072] The database 203 stores sample data for training the neural network, including target geometric parameters, process parameters, and corresponding actual geometric parameters. The combination of these sample data forms the training set of the initial neural network model 202. By learning the patterns and relationships of these sample data, the initial neural network model 202 can form an applied neural network model 205 after training to predict the input target geometric parameters and process parameters and output the corresponding actual geometric parameters.

[0073] The training process of the model training device 204 for the initial neural network model 202 goes through multiple rounds of iteration. Its main training process is as follows:

[0074] (1) At the beginning of training, it is necessary to initialize the neural network so that parameters such as weights and biases are set to random initial values.

[0075] (2) Use the sample data in the training set to calculate the output of the neural network through forward propagation. For each sample, the input is the target geometric parameter and the process parameter, and the output is the predicted value of the actual geometric parameter corresponding to the model. During the forward propagation process, the weights and biases of the model are used to calculate the output of each neuron.

[0076] (3) Calculate the gap between the output of the neural network and the actual geometric parameter, that is, the loss. Commonly used loss functions such as mean square error (MSE), etc. Its goal is to make the error between the predicted value and the actual value as small as possible.

[0077] (4) Use the loss value to adjust the weights and biases in the neural network through the backpropagation algorithm. This is a process of calculating the gradient of each parameter with respect to the loss through the chain rule and then using an optimization algorithm (such as gradient descent) to update the parameters.

[0078] (5) Repeat steps (2), (3), and (4) to continuously adjust the model parameters to make the loss gradually decrease. This process will go through multiple epochs. Each epoch represents that the model conducts a complete learning on the entire training set, that is, uses the entire training set for a complete training process.

[0079] (6) Whether the termination condition is met is determined by setting the training termination condition (for example, a certain number of training rounds is reached or the loss drops to a certain threshold). If the termination condition is met, the model training ends and the applied neural network model 205 is obtained.

[0080] In addition, during the training process, a validation set can be set to verify the performance of the model on unseen data, so as to adjust the hyperparameters of the model in time to prevent overfitting or underfitting.

[0081] After the above steps, the model training device 204 continuously optimizes the parameters of the neural network model so that it can accurately predict the actual geometric parameters corresponding to the target geometric parameters and process parameters.

[0082] It should be understood that during the training process, for example, the constraints in the target geometric parameters and the actual geometric parameters need to be quantified so that the machine learning model can accurately understand and process these features. One quantification method is to use binary representation, where 1 indicates the existence of a constraint and 0 indicates the absence of a constraint. Another quantification method is to use specific numbers to represent the number of constraints that exist and the location of the constraints. For example, for a three-dimensional geometric structure, a digital matrix can be used to represent the constraints, where each element of the matrix indicates whether there is a constraint at the corresponding position, and the specific value of the constraint indicates the degree of the constraint or the restriction.

[0083] 103. Calculate the deformation of the overhanging structure based on the actual geometric parameters and the target geometric parameters.

[0084] The comparison of geometric parameters is to compare the actual geometric parameters with the target geometric parameters, and calculate the difference between the two to obtain the specific change amount of the overhang structure, that is, the deformation amount, which characterizes the degree of deformation that will occur in the overhang structure during the printing process.

[0085] Figure 8 FIG. 2 shows a schematic diagram of the correction of the overhang angle provided by an embodiment of the present invention. Figure 8 As shown, for example, taking the overhang angle as an example, assuming that the actual overhang angle θ1 of the overhang structure predicted and output by the machine learning model is 23°, and it is known from the design (geometric topology model) that the target overhang angle θ2 of the overhang structure is 25°, then the difference between the target overhang angle θ2 and the actual overhang angle θ1 is calculated, that is, 25°-23°=2°, and the difference in the overhang angle is used as the deformation amount, that is, the overhang structure will deform by 2° clockwise in actual printing.

[0086] It should be understood that in practical applications, the deformation amount of the overhang structure is not limited to the overhang angle, but also includes changes in other geometric parameters. For example, the geometric dimensions such as the actual length, width, and thickness of the component can be compared with the designed target dimensions, and the corresponding deformation amount can be obtained by calculating the differences to comprehensively understand the deformation of the overhang structure during the printing process, which is convenient for more accurate compensation of the deformation.

[0087] In addition, considering the particularity of the overhang structure, a mathematical model or algorithm can be introduced to capture the complex deformation behavior of the overhang structure. For example, methods such as finite element analysis can be used to obtain accurate deformation amounts by simulating physical behaviors.

[0088] 104. Perform deformation compensation on the overhang structure according to the deformation amount to generate a geometric topology model after compensation and correction.

[0089] After obtaining the deformation amount of the overhang structure, it is necessary to convert it into specific geometric correction requirements. According to different geometric parameters, corresponding correction strategies are set. For example, for the correction of the overhang angle, the overhang structure can be rotated in the original geometric topology model to reach the target angle; for the correction of the length or width, stretching or compression can be performed in the corresponding directions, etc. Using the set correction strategies, the overhang structure is deformed, such as rotating, translating, scaling, etc., to achieve the compensation of the overhang structure. Finally, a geometric topology model after compensation and correction is generated.

[0090] For the deformation compensation of the overhang angle, a rotation correction strategy can be adopted. By applying an appropriate rotation transformation to the geometric elements containing the overhang structure, the overhang angle gradually approaches the target value. For example, continuing to refer to Figure 8 the example of the correction of the overhang angle shown, Figure 8 In the example, the difference between the calculated target overhang angle θ2 and the actual overhang angle θ1 is 2°. According to the set correction strategy, the overhang structure can be rotated counterclockwise in the original geometric topology model to increase its angle by 2°, that is, the target overhang angle θ2 is corrected from 25° to the compensated overhang angle θ3, which is 27°, to achieve the deformation compensation of the overhang structure. Thus, in actual printing, the overhang angle of the overhang structure is reduced by 2°, so that it is restored from 27° to the desired target overhang angle θ2, which is 25°.

[0091] Exemplarily, taking the geometric configuration dimensions as an example again, assume that the actual dimension (length) of the overhang structure predicted and output by the machine learning model is 21 mm. From the design (geometric topology model), it is known that the target dimension (length) of this overhang structure is 20 mm. Then, calculate the difference between the target dimension and the actual dimension 1 as 1 mm. Take the 1 mm difference as the deformation amount, that is, the overhang structure will increase by 1 mm during actual printing. According to the set correction strategy, a scaling transformation can be performed on the overhang structure in the original geometric topology model to reduce the target dimension by 1 mm, that is, the target dimension is corrected from 20 mm to 19 mm to compensate for the overhang angle, so as to achieve the deformation compensation of the overhang structure. Thus, during actual printing, the actual dimension of the overhang structure increases by 1 mm, making it recover from 19 mm to the desired target dimension of 20 mm, ensuring that the geometric dimensions of the printed component are consistent with the design goals and improving the forming quality.

[0092] In practical applications, when the overhang structure involves multiple geometric parameters, collaborative correction needs to be carried out based on the deformation amount, that is, considering the mutual influence between different parameters to avoid conflicting or inconsistent correction effects. In addition, when designing the deformation strategy, adjustable parameters can be considered. For example, a weight coefficient can be introduced to adjust the correction intensity of different geometric parameters.

[0093] In one implementation of deforming and compensating the overhang structure according to the deformation amount, the geometric topology model can also be divided into triangular patches as units, and independent overhang structure deformation compensation can be performed on each triangular patch according to the deformation amount.

[0094] By decomposing the overall geometric topology model into many small triangular patches, each triangular patch is regarded as a local unit, which can more flexibly handle the deformation of the overhang structure and reduce the computational complexity during the processing. By independently performing deformation compensation on each triangular patch, the local shape of the overhang structure can be adjusted more precisely, thereby reducing the deformation of the overall structure.

[0095] Specifically, the entire geometric topology model can be segmented into several small triangular patches using a triangulation algorithm. Each triangular patch corresponds to a local area, and these areas are independent of each other. For each triangular patch, or for each triangular patch that overlaps with the area corresponding to the overhang structure, calculate the difference between the actual geometric parameter and the target geometric parameter to obtain the local deformation amount. Then, according to the deformation amount of each triangular patch, adopt appropriate deformation compensation strategies, such as rotation, translation, scaling, etc., to independently correct each local area. This process can be carried out in the local coordinate system, avoiding complex calculations for the overall structure. Finally, all the triangular patches that have undergone independent deformation compensation are re-integrated to form the final geometric topology model after compensation and correction.

[0096] 105. Use the geometric topology model after compensation and correction as the target printing model of the component to be formed for unsupported 3D printing.

[0097] Import the geometric topology model after compensation and correction as the target printing model of the component to be formed into the 3D printing system. The 3D printing system consists of a 3D printer and a computer control system. The computer control system controls the 3D printer to manufacture the component in an unsupported state according to the target printing model.

[0098] The computer control system receives and analyzes the geometric topology model after compensation and correction, that is, the target printing model, and converts it into instructions that the 3D printer can understand (it is necessary to use slicing software to slice the target printing model into two-dimensional cross-sections layer by layer and generate corresponding printing paths). The computer control system transmits the processed printing instructions to the 3D printer through the corresponding communication protocol, so that the 3D printer adds materials layer by layer to the building platform according to the instructions provided by the computer control system to gradually form the final component. The present invention preferably uses the LPBF (Laser Powder Bed Fusion) process to manufacture the component. In addition, according to the different characteristics of the component, other types of 3D printing processes can also be used. It should be understood that no support is added during the manufacturing process regardless of the selected 3D printing process.

[0099] In some embodiments, the 3D printer involved in the LPBF process mainly consists of a powder cylinder for storing powder, a powder bed formed on the building platform, a powder spreading device (such as a scraper or a roller) for transporting and spreading the powder overflowing from the powder cylinder onto the powder bed, and a laser for emitting a laser beam to melt the powder.

[0100] The unsupported 3D printing method provided by the embodiments of the present invention realizes the accurate prediction of the actual geometric parameters of the overhanging structure through the feature recognition of the overhanging structure and the application of the machine learning model. Through the calculation and correction of the deformation amount, the deformation problem of the overhanging structure during the printing process is effectively solved. This method combines advanced 3D printing technology and intelligent machine learning means, provides high controllability for the manufacturing of components under unsupported conditions, and finally realizes high-quality and high-efficiency 3D printing in an unsupported state.

[0101] Figure 9 Another process schematic diagram of the unsupported 3D printing method provided by the embodiments of the present invention is shown. As Figure 9 shown, the unsupported 3D printing method 100 provided in some embodiments of the present invention further includes steps 301 and 302, etc. It should be understood that Figure 9 the shown unsupported 3D printing method 100 is based on Figure 4 the shown unsupported 3D printing method 100 (steps 101 - 105, etc.).

[0102] 301. At least use the target geometric parameters of the overhanging structure as input, and predict and output optimized process parameters that match the overhanging structure through a machine learning model. The machine learning model indicates the mapping relationship between the target geometric parameters, process parameters, actual geometric parameters, process variables, and forming quality of the overhanging structure.

[0103] 302. Use the optimized process parameters as the target process parameters of the overhanging structure for unsupported 3D printing.

[0104] In some embodiments, the process variables include at least one of the molten pool size and morphology, molten pool temperature distribution, and stress distribution, preferably including all of these variables.

[0105] The molten pool size and morphology refer to the shape and size of the molten pool during 3D printing, that is, the geometric shape of the liquid region of the molten material, which is affected by factors such as printing speed, temperature distribution, and material properties. Changes in the molten pool size and morphology will affect the melting and solidification of the material during printing, and thus affect the forming of the component.

[0106] The molten pool temperature distribution represents the distribution of the molten pool temperature during 3D printing. Among them, factors such as printing speed, laser power, and material melting point will affect the temperature distribution of the molten pool, and the temperature distribution of the molten pool affects the melting and solidification process of the material, which has a direct impact on the properties such as the density and strength of the component.

[0107] The stress distribution represents the stress state inside or on the surface of the component, including tensile stress, compressive stress, etc. Among them, factors such as the thermal expansion and contraction of the material and the shrinkage during solidification affect the stress distribution of the component. An unfavorable stress distribution may cause problems such as component deformation and cracking, affecting the quality of the final forming.

[0108] In some embodiments, the forming quality includes at least one of defect characteristics, roughness, and dimensional accuracy, preferably including all of these parameters.

[0109] Defect characteristics refer to defects on the surface or inside of the component, such as pores, cracks, etc. Factors such as temperature gradient, residual stress, and printing path may all cause defects in the component.

[0110] Roughness characterizes the roughness of the component surface, that is, the unevenness of the surface, which is affected by factors such as the printing layer interval, printing speed, and material fluidity.

[0111] Dimensional accuracy represents the deviation between the actual size and the designed size of the component, which is directly related to whether the component meets the design requirements and is particularly critical for applications that require high precision.

[0112] It should be understood that Figure 9The machine learning model in the example is Figure 4 the machine learning model in the example, except that in Figure 9 the application of the example, the machine learning model performs different inputs and outputs from Figure 4 the example. Of course, the machine learning model in the example also needs to be further trained to adapt to different inputs and outputs from Figure 9 the example. Figure 4 the example.

[0113] In some embodiments, in order to better apply the machine learning model, it can be divided into two sub-models, one for performing Figure 4 data prediction in the example, and the other for performing Figure 9 data prediction in the example.

[0114] Figure 10 shows a schematic application structure diagram of the neural network model provided by the embodiments of the present invention. As Figure 10 shown, taking this machine learning model as a neural network (Neural Networks, NN) model as an example, the neural network model can be composed of two sub-models, namely a first neural network sub-model for performing Figure 4 data prediction in the example and a second neural network sub-model for performing Figure 9 data prediction in the example. It should be understood that the described neural network model includes both Figure 7 the initial neural network model 202 and the applied neural network model 205 shown.

[0115] Similarly, referring to the environment 200 shown in Figure 7 , in some embodiments, the database 203 stores sample data for training the neural network, including target geometric parameters, process parameters, actual geometric parameters, process parameters, and forming quality. The combination of these sample data forms the training set of the initial neural network model 202. By learning the patterns and relationships of these sample data, the initial neural network model 202 can form the applied neural network model 205 after training, establish the mapping relationship between the target geometric parameters, process parameters, actual geometric parameters, process parameters, and forming quality of the overhang structure, and predict the input target geometric parameters, so as to more accurately predict the corresponding optimized process parameters. By applying the output optimized process parameters to the overhang structure, the process parameters can be automatically adjusted according to the designed geometric features during the 3D printing process, so as to better adapt to the shape and requirements of the structure, which helps to improve the forming quality and surface finish of the overhang structure, and ultimately promotes the successful realization of support-free 3D printing.

[0116] In specific applications, only the target geometric parameters of the overhang structure need to be provided, that is, the requirements for the desired shape, size, etc. of the component. The neural network model can understand the complex mapping relationship between the target geometric parameters and the optimal process parameters by learning historical data. Therefore, when the target geometric parameters are input, the neural network model can intelligently predict and output the optimized process parameters that match them. Moreover, the output optimized process parameters are intelligently adjusted by the model to adapt to the current design of the overhang structure, so as to achieve the best forming quality during 3D printing, including improving the surface finish of the structure, reducing defects, ensuring dimensional accuracy, etc.

[0117] In some embodiments, in addition to using the target geometric parameters of the overhang structure as input, the target geometric parameters of the overhang structure and the forming quality can also be used as input together to achieve the prediction and output of the optimal process parameters that match the desired forming quality of the user. That is to say, in addition to the target geometric parameters, by specifying the desired forming quality standards, such as surface finish, defect tolerance, dimensional accuracy, etc., the requirements for these forming qualities are used as additional inputs, so that the neural network model can intelligently predict the optimized process parameters that match them by learning the relationship between the target geometric parameters of the overhang structure and the desired forming quality, so as to meet the requirements of both the target geometric parameters and the desired forming quality at the same time.

[0118] Figure 11 A schematic diagram of a prediction process of process parameters provided by an embodiment of the present invention is shown. As Figure 11 shown, in some embodiments, step 301 includes 311 and 312.

[0119] 311. According to the feature recognition result of the overhang structure, divide the angle region to which it belongs, and use the divided angle region as at least one of the target geometric parameters of the overhang structure as input.

[0120] In step 311, first, the angle information (overhang angle) of the structure is obtained through the feature recognition result of the overhang structure. The angle region division of the overhang structure is to consider the characteristics of the structure more carefully. The overhang angle is divided into different regions, and the division of these angle regions will be used as one of the target geometric parameters of the overhang structure.

[0121] 312. Predict and output the optimized process parameters that match the angle region to which the overhang structure belongs through the machine learning model.

[0122] In step 312, a trained machine learning model is used. This machine learning model can intelligently predict corresponding optimized process parameters according to the angular region to which the overhang structure belongs. By learning a large amount of sample data, it can identify the relationship between the angular region of the overhang structure and the optimal process parameters. Therefore, the input is the angular region to which the overhang structure belongs, and the output is the optimized process parameters matching this region.

[0123] Exemplarily, in one example, the angular region is divided into three types, namely Q1, Q2, and Q3, where Q1 < 20°, 20° ≤ Q2 < 35°, and Q3 ≥ 35°.

[0124] Specifically, in the specific application of this example, for a given overhang structure, its overhang angle information is first obtained through a feature recognition process. Then, according to the overhang angle information, it is judged which angular region this overhang structure belongs to. For example, if the overhang angle is less than 20°, it is determined as Q1; if the overhang angle is greater than or equal to 20° and less than 35°, it is determined as Q2; if the overhang angle is greater than or equal to 35°, it is determined as Q3. In this way, the judged angular region (Q1, Q2, or Q3) is used as a kind of information of the target geometric parameter of the overhang structure and input into the machine learning model. Through this machine learning model, the optimized process parameters matching the angular region to which the overhang structure belongs are predicted and output, such as including laser power, scanning strategy, spot diameter, etc., to ensure that within a specific angular region, the overhang structure can obtain the best forming quality.

[0125] Exemplarily, in another example, the angular region is divided into four types, namely Q1, Q2, Q3, and Q4, where 8° ≤ Q1 < 15°, 15° ≤ Q2 < 25°, 25° ≤ Q3 < 45°, and 45° ≤ Q4.

[0126] It should be understood that in this example, the division of each angular region (Q1, Q2, Q3, and Q4) is adjusted according to actual needs. Considering that the overhang structure may have different printing characteristics and process requirements in different angular ranges, therefore, for each region, it can be adjusted according to the specific application scenario and printing requirements to ensure that it can more accurately adapt to the characteristics of various overhang structures.

[0127] In some embodiments, a process library for unsupported 3D printing of overhang structures at specific angles can also be established according to the prediction results of the machine learning model to establish process parameters matching the overhang structures at specific angles.

[0128] Figure 12 Shows a schematic diagram of the matching between the overhang structure and process parameters at specific angles provided by the embodiments of the present invention. As Figure 12As shown, taking the example where the above-mentioned overhang structure is divided into four angular regions, Q1 - Q4 are respectively matched with processes A - D. For example, if the prediction result divides the overhang structure into the Q1 region, process A associated with it can be directly recommended or selected, so as to provide the best supportless 3D printing process parameters for the overhang structure at this angle, thereby improving the printing efficiency and optimizing the printing quality.

[0129] In some embodiments, the embodiments of the present invention further provide a component, which is manufactured (printed) according to the supportless 3D printing method 100 described above.

[0130] Figure 13 The schematic block diagram of the supportless 3D printing device provided by the embodiments of the present invention is shown. As Figure 13 shown, in some embodiments, the supportless 3D printing device 400 provided by the present invention includes a model correction unit 410 and a 3D printing unit 420. Among them, the model correction unit 410 includes a feature recognition unit 411, a data prediction unit 412, a deformation amount calculation unit 413, and a deformation compensation unit 414.

[0131] Among them, the feature recognition unit 411 is configured to perform feature recognition on the geometric topology model of the to - be - formed component, at least including the overhang structure, and generate the target geometric parameters of the overhang structure.

[0132] Among them, the data prediction unit 412 is configured to obtain the process parameters of the overhang structure and use them together with the target geometric parameters as inputs, and predict and output the actual geometric parameters matching them through a machine learning model, where the machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters, and actual geometric parameters of the overhang structure.

[0133] Among them, the deformation amount calculation unit 413 is configured to calculate the deformation amount of the overhang structure according to the actual geometric parameters and the target geometric parameters.

[0134] Among them, the deformation compensation unit 414 is configured to perform deformation compensation on the overhang structure according to the deformation amount, and generate a compensated and corrected geometric topology model.

[0135] Among them, the 3D printing unit 420 is configured to use the compensated and corrected geometric topology model as the target printing model of the to - be - formed component for supportless 3D printing.

[0136] In some embodiments, the data prediction unit 412 is further configured to obtain the target geometric parameters and forming quality of the overhang structure as inputs, and predict and output the optimized process parameters matching the overhang structure through a machine learning model.

[0137] In some embodiments, the data prediction unit 412 is further configured to divide the angle region of the overhang structure according to the feature recognition result thereof, and use the divided angle region as at least one of the target geometric parameters of the overhang structure, and as an input, predict and output optimized process parameters that match the angle region to which the overhang structure belongs through a machine learning model.

[0138] Figure 14 FIG. shows a schematic block diagram of an electronic device provided by an embodiment of the present invention. As Figure 14 shown, in some embodiments, the electronic device 500 includes a processor 501 and a memory 502 (wherein the number of the processor 501 and the memory 502 may be one or more). The memory 502 is coupled to the processor 501 and is configured to store instructions executed by the processor 501. When the instructions are executed by the processor 501, the electronic device 500 is caused to execute the unsupported 3D printing method described in any one of the foregoing.

[0139] The processor 501 communicates with the memory 502, and the memory 502 may include a read-only memory and a random access memory, and provides instructions and data to the processor 501. In addition, a part of the memory 502 may further include a non-volatile random access memory (NVRAM). In the memory 502, operation instructions, executable modules, data structures, or subsets thereof, or even extended sets thereof are stored. These operation instructions cover various operations for implementing various operations.

[0140] The unsupported 3D printing method described in the embodiments of the present invention may be applied to the processor 501 or implemented by the processor 501. The processor 501 may be any suitable computer processor, such as a central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), etc. In the embodiments of the present invention, the processor 501 is responsible for executing each step of the unsupported 3D printing method, including but not limited to functions such as data processing, prediction of a neural network model, and optimization of process parameters. The processor 501 may be a general computer processor or a dedicated embedded processor to implement the control and management of the unsupported 3D printing of the overhang structure.

[0141] In some embodiments, the present invention further provides a computer-readable storage medium storing a computer program, which implements the unsupported 3D printing method described in any one of the foregoing when the computer program is executed by a processor.

[0142] Among them, a computer-readable storage medium refers to a medium that can be read by a computer system, such as a hard disk, a solid-state drive, an optical disc, a flash drive, etc. In some embodiments of the present invention, the computer-readable storage medium stores a set of computer programs, and these programs are executed by a processor to implement each step and function described in the unsupported 3D printing method. These computer programs may include an operating system, embedded software, application programs, etc., for controlling and managing the unsupported 3D printing process of the overhang structure. By reading and executing the programs stored on the computer-readable storage medium, the computer system can effectively implement the unsupported 3D printing method described in the present invention.

[0143] In some embodiments, the present invention also provides a computer program product, which includes computer-executable instructions that, when executed by a processor, cause the computer to implement the unsupported 3D printing method described in any one of the foregoing.

[0144] Among them, a computer program product is a product that stores computer-executable instructions, and its purpose is to implement each step and function described in the unsupported 3D printing method when executed by the processor of a computer system. The computer-executable instructions may include an operating system, application programs, embedded software, etc., to control and manage the unsupported 3D printing process of the overhang structure. By using such a computer program product, a user can execute the unsupported 3D printing method of the present invention on a computer system, improve the forming quality and surface finish of the overhang structure, and ultimately achieve successful unsupported 3D printing.

[0145] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "exemplary" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0146] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. An unsupported 3D printing method with optimized overhang structure, characterized in that, The method includes: Performing feature recognition on the geometric topology model of the to-be-formed component, including at least overhanging structures, to generate target geometric parameters of the overhanging structures; Obtaining process parameters of the overhanging structures and using them together with the target geometric parameters as inputs, and predicting and outputting matching actual geometric parameters through a machine learning model, where the machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters, and actual geometric parameters of the overhanging structures, and the target geometric parameters and actual geometric parameters include at least one of geometric configuration dimensions, overhanging angles, aspect ratios, spans, and constraint conditions; Calculating the deformation amount of the overhanging structures based on the actual geometric parameters and target geometric parameters; Performing deformation compensation on the overhanging structures according to the deformation amount to generate a compensated and corrected geometric topology model; Using the compensated and corrected geometric topology model as the target printing model of the to-be-formed component for support-free 3D printing.

2. The method according to claim 1, wherein The performing feature recognition on the geometric topology model of the to-be-formed component, including at least overhanging structures, includes: Respectively extracting the contour and skeleton from the obtained geometric topology model of the to-be-formed component; Performing inverse fitting on the contour and skeleton; Obtaining features including at least the overhanging structures according to the inverse fitting result.

3. The method according to claim 2, wherein The obtaining features including at least the overhanging structures according to the inverse fitting result includes: Obtaining structures with overhanging angles greater than a preset threshold from the inverse fitting result to confirm them as the overhanging structures.

4. The method according to claim 1, wherein The process parameters include at least one of laser power, scanning strategy, spot diameter, and powder spreading layer thickness.

5. The method according to claim 1, wherein The performing deformation compensation on the overhanging structures according to the deformation amount includes: Dividing the geometric topology model with triangular patches as units; Performing independent overhanging structure deformation compensation on each triangular patch according to the deformation amount.

6. The method according to claim 1, wherein The machine learning model indicates the mapping relationship between the target geometric parameters, process parameters, actual geometric parameters, process parameters, and forming quality of the overhanging structures.

7. The method according to claim 6, wherein The method further includes: Using at least the target geometric parameters of the overhanging structures as inputs, and predicting and outputting optimized process parameters matching the overhanging structures through the machine learning model; Using the optimized process parameters as the target process parameters of the overhanging structures for support-free 3D printing.

8. The method according to claim 7, wherein The using at least the target geometric parameters of the overhanging structures as inputs includes: Using the target geometric parameters and forming quality of the overhanging structures as inputs.

9. The method according to claim 7, wherein The predicting and outputting optimized process parameters matching the overhanging structures through the machine learning model includes: Performing division of the belonging angle regions on the overhanging structures according to the feature recognition result, and using the divided angle regions as at least one of the target geometric parameters of the overhanging structures as inputs; Predicting and outputting optimized process parameters matching the angle regions to which the overhanging structures belong through the machine learning model.

10. The method according to claim 9, wherein The angle regions are at least three types, namely Q1, Q2, and Q3.

11. The method according to claim 10, wherein, Q1 < 20°, 20° ≤ Q2 < 35°, Q3 ≥ 35°.

12. The method according to claim 6, wherein The process parameters include at least one of the molten pool size and morphology, molten pool temperature distribution, and stress distribution.

13. The method according to claim 6, characterized in that, The forming quality includes at least one of defect characteristics, roughness, and dimensional accuracy.

14. A component, characterized in that, The component is printed by the method according to any one of claims 1 to 13.

15. A supportless 3D printing device, characterized in that, It includes a model correction unit and a 3D printing unit; Among them, the model correction unit includes: A feature recognition unit configured to perform feature recognition on the geometric topology model of the component to be formed, including at least overhanging structures, and generate target geometric parameters of the overhanging structures; A data prediction unit configured to obtain the process parameters of the overhanging structures and use them together with the target geometric parameters as inputs, and predict and output the actual geometric parameters matching them through a machine learning model, where the machine learning model at least indicates the mapping relationship between the target geometric parameters, process parameters, and actual geometric parameters of the overhanging structures, and the target geometric parameters and actual geometric parameters include at least one of geometric configuration dimensions, overhanging angle, aspect ratio, span, and constraint conditions; A deformation amount calculation unit configured to calculate the deformation amount of the overhanging structures according to the actual geometric parameters and the target geometric parameters; A deformation compensation unit configured to perform deformation compensation on the overhanging structures according to the deformation amount to generate a compensated and corrected geometric topology model; Among them, the 3D printing unit is configured to use the compensated and corrected geometric topology model as the target printing model of the component to be formed for support-free 3D printing.

16. An electronic device, characterized in that, It includes: At least one processor; At least one memory, the at least one memory is coupled to the at least one processor and is used to store instructions executed by the at least one processor. When the instructions are executed by the at least one processor, the electronic device executes the method according to any one of claims 1 to 13.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the method according to any one of claims 1 to 13.

18. A computer program product, characterized in that, The computer program product includes computer-executable instructions, and when the computer-executable instructions are executed by a processor, the computer implements the method according to any one of claims 1 to 13.

Images