Metal additive manufacturing method and system, electronic equipment, medium and program product
In the process of laser powder bed melting (LPBF), the registration technology of training data and multiple sensor data is used to establish a defect detection model with higher accuracy, which solves the problem of low confidence in LPBF defect detection, real-time defect correction and quality control during the processing process, and reduces production costs.
Patent Information
- Application Number
- CN202510420119.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-07-18
AI Technical Summary
The existing laser powder bed melting (LPBF) defect detection confidence is low, traditional data analysis methods cannot correct defects during the manufacturing process, and multidimensional data cannot be accurately fused, resulting in low detection accuracy.
By acquiring training data, using defect detection models to register and analyze the three-dimensional image data during product processing, combining a variety of sensor data, such as powder laying, sintering, tomography and melt pool data, iterative nearest point algorithm and single threshold segmentation technology can improve the registration accuracy of image data and establish a defect detection model with stronger accuracy.
It realizes timely identification of defects and improvements during product processing, provides more comprehensive manufacturing process information, reduces production costs and improves the stability and accuracy of product quality.
Smart Images
Figure CN120339226A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of industrial part manufacturing, and in particular, to a metal additive manufacturing method, system, electronic device, medium, and program product. Background Art
[0002] In recent years, metal additive manufacturing has provided a new development direction for the geometric design and process optimization of industrial part production. However, process repeatability and quality control remain one of the biggest obstacles to the widespread adoption of this technology in the industry. Laser Powder Bed Fusion (LPBF) is considered to be one of the most widely used metal additive manufacturing technologies. It spreads metal particle materials on the surface of a building platform to form a powder bed, and then uses the information provided by a CAD file (Computer-Aided Design, a technical file that uses computer technology for design and drawing) to apply a high-power laser as a heat source to a specific area of the powder bed, melting the metal powder, and repeating this process for each printed layer to finally complete the manufacturing of a three-dimensional part. The LPBF method has flexible design and high resource utilization efficiency, and has been widely applied in industrial fields such as aviation and aerospace.
[0003] Due to its raw material limitations, LPBF is prone to forming various types of pores and defects and is prone to generating fatigue behavior. At the same time, the formation principle of LPBF and its material removal technology require an assessment of the microstructure and mechanical behavior of the material. To solve this problem, relevant researchers have explored various monitoring methods to identify LPBF defects. The internal voids of LPBF can be clearly identified under CT scanning (Computed Tomography). Industrial CT detection technology is well-known for its high precision and intuitive three-dimensional tomographic images. It can comprehensively detect and quantify the surface and internal conditions of a sample, including surface roughness, dimensional accuracy, and internal defects. This technology is particularly suitable for the detection of small and medium-sized structural parts in complex components, and especially shows its unique advantages in detecting internal pores and cracks.
[0004] However, traditional part evaluation and identification techniques, including computed tomography (CT), can only be used after manufacturing is completed, so defects cannot be corrected during the manufacturing process. Compared with post-manufacturing CT images, defect detection based on process sensing enables in-process remelting and correction of detected defects, that is, measurement of various process information. Such a process monitoring system helps in defect detection and process optimization, and can further enable the feedback control system to automatically and real-time correct detected defects, thereby improving the printing quality. However, traditional data analysis methods often rely on single-sensor data, but single-sensor data cannot comprehensively reflect the complexity of the manufacturing process. In addition, there are often differences in spatial and temporal dimensions between different types of data in the metal additive manufacturing (LPBF) process (for example, the molten pool monitoring data is one-dimensional electrical signal data in the time domain, while the powder spreading-sintering monitoring data is two-dimensional image data), and accurate fusion cannot be achieved between multi-dimensional data, resulting in low accuracy of defect detection. Summary of the Invention
[0005] The technical problem to be solved by the present disclosure is to provide a metal additive manufacturing method, system, electronic device, medium, and program product to overcome the defect of low confidence in defect detection in laser powder bed fusion (LPBF) in the prior art.
[0006] The present disclosure solves the above technical problem through the following technical solutions:
[0007] The present disclosure provides a metal additive manufacturing method, and the metal additive manufacturing method includes:
[0008] Obtain training data; wherein, the training data includes first image data for characterizing the three-dimensional image of the processed product and the product defect result;
[0009] Use the training data to train a defect detection model to obtain a defect detection model for predicting product defects;
[0010] Obtain second image data registered with the first image data; wherein, the second image data is used to characterize the three-dimensional image data during the product processing;
[0011] Input the second image data into the defect detection model to output a defect detection result.
[0012] Preferably, the step of obtaining the second image data registered with the first image data includes:
[0013] Obtain the three-dimensional image data during the product processing;
[0014] Determine the transformation matrix between the first image data and the three-dimensional image data during the product processing through the iterative closest point algorithm;
[0015] Determine second image data registered with the first image data based on the transformation matrix and the first image data.
[0016] Preferably, the step of determining the transformation matrix between the first image data and the three-dimensional image data in the product processing process through the iterative closest point algorithm includes:
[0017] Convert the first image data into a first three-dimensional point cloud, and convert the three-dimensional image data in the product processing process into a second three-dimensional point cloud;
[0018] Determine the optimal rigid body transformation between the first three-dimensional point cloud and the second three-dimensional point cloud as the transformation matrix through the iterative closest point algorithm.
[0019] Preferably, the step of converting the first image data into a first three-dimensional point cloud and converting the three-dimensional image data in the product processing process into a second three-dimensional point cloud includes:
[0020] Convert the first image data into a first three-dimensional binary label map through single-threshold segmentation, and convert the three-dimensional image data in the product processing process into a second three-dimensional binary label map;
[0021] Convert the first three-dimensional binary label map into a first three-dimensional point cloud, and convert the second three-dimensional binary label map into a second three-dimensional point cloud.
[0022] Preferably, the step of obtaining the three-dimensional image data in the product processing process includes:
[0023] Extract the region of interest, crop and synthesize each layer of two-dimensional image data in the processing process, and calibrate the three-dimensional voxel size to obtain the three-dimensional image data in the product processing process.
[0024] Preferably, the first image data is obtained by computer tomography;
[0025] And / or,
[0026] The type of the second image data includes at least one of powder spreading data, sintering data, tomography data, and molten pool data;
[0027] And / or,
[0028] The second image data is determined according to a preset laser trajectory.
[0029] The present disclosure also provides a metal additive manufacturing system, and the metal additive manufacturing system includes:
[0030] A first acquisition module, configured to acquire training data; wherein, the training data includes first image data for characterizing a three-dimensional image of a processed product and product defect results;
[0031] A model training module, configured to train a defect detection model using the training data to obtain a defect detection model for predicting product defects;
[0032] A second acquisition module, configured to acquire second image data registered with the first image data; wherein, the second image data is used to characterize three-dimensional image data during the product processing;
[0033] A model application module, configured to input the second image data into the defect detection model to output a defect detection result.
[0034] The present disclosure further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and configured to run on the processor. When the processor executes the computer program, the above-mentioned metal additive manufacturing method is implemented.
[0035] The present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned metal additive manufacturing method is implemented.
[0036] The present disclosure further provides a computer program product, including a computer program. When the computer program is executed by a processor, the above-mentioned metal additive manufacturing method is implemented.
[0037] On the basis of conforming to common general knowledge in the art, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred examples of the present disclosure.
[0038] The positive and progressive effects of the present disclosure are as follows:
[0039] The present disclosure trains a defect detection model with higher precision through the three-dimensional image of the processed product, and then inputs the three-dimensional image data during the product processing registered with the first image data into the defect detection model to timely determine product defects during the product processing and take corresponding improvement measures, providing more intuitive and comprehensive manufacturing process information and reducing the production cost of products. Description of the Drawings
[0040] Figure 1 It is a flowchart of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0041] Figure 2 It is a workflow block diagram of a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0042] Figure 3The laser trajectory, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0043] Figure 4 The solid filling diagram of the laser trajectory, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0044] Figure 5 The sintering monitoring image, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0045] Figure 6 The tomography monitoring image, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0046] Figure 7 The molten pool monitoring composite image, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0047] Figure 8 The CT scan tomogram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0048] Figure 9 The three-dimensional registration result of mapping the CT data to the tomography data, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0049] Figure 10 The ideal laser trajectory design schematic diagram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0050] Figure 11 The CT data schematic diagram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0051] Figure 12 The sintering data schematic diagram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0052] Figure 13 The tomography data schematic diagram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0053] Figure 14 The molten pool data schematic diagram, which is a specific example of a metal additive manufacturing method provided in Embodiment 1 of the present disclosure;
[0054] Figure 15 The structure diagram of a metal additive manufacturing system provided in Embodiment 2 of the present disclosure;
[0055] Figure 16 The structure diagram of an electronic device provided in Embodiment 3 of the present disclosure. Detailed Implementation Modes
[0056] The present disclosure will be further described below by way of examples, but the present disclosure is not limited to the scope of the described examples for this reason.
[0057] In the embodiments of the present disclosure, prefix words such as "first" and "second" are only used to distinguish different described objects, and have no limiting effect on the position, order, priority, quantity, content, etc. of the described objects. The use of ordinal numbers and other prefix words for distinguishing described objects in the embodiments of the present disclosure does not constitute a limitation on the described objects. For the description of the described objects, refer to the description in the context of the claims or the embodiments. There should be no redundant limitation due to the use of such prefix words. In addition, in the description of this embodiment, unless otherwise specified, the meaning of "a plurality" is two or more.
[0058] In the embodiments of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information and other processing all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.
[0059] Embodiment 1
[0060] This embodiment provides a metal additive manufacturing method. Refer to Figure 1 , the metal additive manufacturing method includes:
[0061] S1. Obtain training data.
[0062] Among them, the training data includes first image data for characterizing the three-dimensional image of the processed product and the product defect results.
[0063] In an alternative implementation mode, the first image data is obtained by computer tomography (CT).
[0064] S2. Use the training data to train a defect detection model to obtain a defect detection model for predicting product defects.
[0065] S3. Obtain second image data registered with the first image data.
[0066] Among them, the second image data is used to characterize the three-dimensional image data during the product processing.
[0067] Among them, the second image data is determined according to a preset laser trajectory. The preset laser trajectory file includes coordinate information on all laser paths during the manufacturing process generated by a 3D (three-dimensional) printer. In this embodiment, the coordinates are first read, and then an empty image (with all image pixel values being 0) is created. The laser trajectory in the ideal printing situation is drawn on the empty image according to the coordinates of the read points. Subsequently, methods such as the Flood Fill Algorithm and Boolean logic operations are performed on the ideal trajectory on the image to fill the solid part in the ideal printing situation on the image, so as to generate a tomographic slice image of metal additive manufacturing in the ideal situation.
[0068] Among them, the types of the second image data include powder spreading data, sintering data, tomography data, and molten pool data, etc. It should be noted that different types of the second image data are obtained by different sensors.
[0069] The powder spreading data refers to the image data of evenly spreading powder materials on the working platform during the metal additive manufacturing process.
[0070] The sintering data refers to the image data when powder particles are combined into a dense material through diffusion or chemical reaction at high temperature during the metal additive manufacturing process.
[0071] The tomography data refers to the image data for monitoring the quality and structural characteristics of each layer of three-dimensional printing through tomography imaging technology during metal additive manufacturing.
[0072] The molten pool data refers to the image data obtained by monitoring and analyzing the dynamic behavior of the local molten area formed when a laser or electron beam acts on powder or solid materials during the metal additive manufacturing process. Since the molten pool data is time-domain electrical signal data, additional preprocessing is required for it: the molten pool data (voltage signal - time) is drawn on an empty image according to the prior laser trajectory, with the pixel coordinates being the position of the laser trajectory at a certain time and the pixel value being the normalized voltage signal.
[0073] S4. Input the second image data into the defect detection model to output the defect detection result.
[0074] In this embodiment, a defect detection model with higher accuracy is trained through the three-dimensional images of the processed products. Then, the three-dimensional image data during the product processing registered with the first image data is input into the defect detection model to timely determine product defects during the product processing and take corresponding improvement measures, providing more intuitive and comprehensive manufacturing process information and reducing the production cost of the products. This embodiment also introduces various sensor data in the manufacturing process (i.e., powder spreading data, sintering data, tomography data, and molten pool data), and supplementary information content can be shared among various data, further improving the correlation with actual product defects and the confidence level of product defect detection.
[0075] In an alternative embodiment, step S3 includes:
[0076] S31. Obtain the three-dimensional image data during the product processing.
[0077] S32. Determine the transformation matrix between the first image data and the three-dimensional image data during the product processing through the Iterative Closest Point (ICP) algorithm.
[0078] In an alternative embodiment, step S32 includes:
[0079] S321. Convert the first image data into a first three-dimensional point cloud and convert the three-dimensional image data during the product processing into a second three-dimensional point cloud.
[0080] S322. Determine the optimal rigid body transformation between the first three-dimensional point cloud and the second three-dimensional point cloud as the transformation matrix through the Iterative Closest Point (ICP) algorithm.
[0081] Next, a specific method for determining the transformation matrix is introduced:
[0082] The core objective of the Iterative Closest Point (ICP) algorithm is to find the optimal rigid body transformation (including rotation and translation) between the two sets of point clouds, namely the first three-dimensional point cloud and the second three-dimensional point cloud, to minimize the difference between the two sets of point clouds. The principle and steps of the ICP algorithm are as follows:
[0083] Step 1. Determine the objective function. The objective function of the ICP algorithm is to minimize the distance between the first three-dimensional point cloud and the second three-dimensional point cloud. Among them, the objective function can be expressed as:
[0084]
[0085] Among them, J is used to represent the difference between the first three-dimensional point cloud and the second three-dimensional point cloud, p i is used to represent the i-th point in the first three-dimensional point cloud, q iIt is used to represent the i-th point in the second 3D point cloud, n is used to represent the total number of points in the first 3D point cloud, R is the rotation matrix, and t is the translation vector.
[0086] Step 2, calculate the centroids of the first 3D point cloud and the second 3D point cloud through the following formula:
[0087]
[0088] where, μ p is used to represent the centroid of the first 3D point cloud, n is used to represent the total number of points in the first 3D point cloud, and pi is used to represent the i-th point in the first 3D point cloud.
[0089]
[0090] where, μ q is used to represent the centroid of the first 3D point cloud, n is used to represent the total number of points in the first 3D point cloud, and qi is used to represent the i-th point in the first 3D point cloud.
[0091] Then, considering the offset of the centroid, rewrite the objective function:
[0092]
[0093] where, J’ is used to represent the difference between the rewritten first 3D point cloud and the second 3D point cloud, p i is used to represent the i-th point in the first 3D point cloud, q i is used to represent the i-th point in the second 3D point cloud, n is used to represent the total number of points in the first 3D point cloud, R is the rotation matrix, t is the translation vector, μ p is used to represent the centroid of the first 3D point cloud, μ q is used to represent the centroid of the first 3D point cloud.
[0094] In this step, the point cloud is decentralized by subtracting the centroids of the two groups of point clouds from the rewritten objective function J’, which simplifies the calculation.
[0095] Step 3, solve the rotation matrix.
[0096] Solve the rotation matrix R through singular value decomposition (SVD):
[0097] H = P′ T Q′;
[0098] where, H is the covariance matrix, which is used to represent the linear relationship between the first 3D point cloud and the second 3D point cloud, P′ is used to represent the point cloud matrix after decentralizing the first 3D point cloud, and Q′ is used to represent the point cloud matrix after decentralizing the second 3D point cloud.
[0099] Then perform SVD decomposition on H to obtain:
[0100] H = U∑V T ;
[0101] Among them, H is used to represent the covariance matrix, U is used to represent the orthogonal matrix, ∑ is used to represent the diagonal matrix, and V is used to represent the orthogonal matrix.
[0102] Determine the optimal rotation matrix:
[0103] R = VU T ;
[0104] Among them, R is used to represent the optimal rotation matrix, U is used to represent the orthogonal matrix, and V is used to represent the orthogonal matrix.
[0105] This step is to find the optimal rotation matrix to minimize the difference between the transformed first 3D point cloud and the second 3D point cloud.
[0106] Step Four, solve for the translation vector.
[0107] The translation vector t can be calculated by the following formula:
[0108] t = μ q - Rμ p ;
[0109] Among them, t is used to represent the translation vector, μ p is used to represent the centroid of the first 3D point cloud, μ q is used to represent the centroid of the first 3D point cloud, and R is used to represent the optimal rotation matrix.
[0110] This step determines the optimal translation vector to align the centroid of the first 3D point cloud with the centroid of the second 3D point cloud.
[0111] Step Five, iterative update.
[0112] Apply the transformation (R, t) to the first 3D point cloud, then re - perform point - pair matching, and solve for the optimal transformation matrix until the convergence condition is met.
[0113] S33. Determine the second image data registered with the first image data through the transformation matrix and the first image data.
[0114] After registering the first 3D point cloud and the second 3D point cloud through the ICP algorithm, the optimal rotation matrix R and translation matrix t can be obtained. Applying the transformation matrix (R, t) to the first image data to be transformed can obtain the matching second image data, which is convenient for subsequent more accurate defect prediction and process adjustment.
[0115] In this embodiment, the optimal rotation matrix R and translation vector t are obtained through the Iterative Closest Point (ICP) algorithm, so that the first three-dimensional point cloud can be more accurately aligned with the second three-dimensional point cloud, thereby improving the registration accuracy of the first image data and the second image data.
[0116] In an alternative embodiment, step S321 includes:
[0117] S3211. Through single-threshold segmentation, convert the first image data into a first three-dimensional binary label map, and convert the three-dimensional image data during the product processing into a second three-dimensional binary label map.
[0118] S3212. Convert the first three-dimensional binary label map into a first three-dimensional point cloud, and convert the second three-dimensional binary label map into a second three-dimensional point cloud.
[0119] In this embodiment, since the products of LPBF are usually metal parts, and the pixel values in the first image data and the second image data are significantly different from the background, the first image data and the second image data can be denoised through single-threshold segmentation and respectively converted into a first three-dimensional binary label map and a second three-dimensional binary label map, thereby improving the accuracy of converting the image data into a three-dimensional binary label map.
[0120] In an alternative embodiment, step S31 includes:
[0121] Extract the region of interest and crop and synthesize the two-dimensional image data of each layer in the processing process, and calibrate the three-dimensional voxel size to obtain the three-dimensional image data during the product processing.
[0122] In this embodiment, for the second image data of types such as powder laying data, sintering data, tomography data, and molten pool data, the region of interest (ROI) can be extracted and the image can be cropped, and the data of each layer in the printing process can be synthesized into three-dimensional image data. Then, according to the actual data acquisition area size and layer thickness, the three-dimensional voxel size is calibrated to restore the true size of the product, laying a foundation for detecting product defects in the subsequent product manufacturing process.
[0123] The following introduces a specific example to illustrate the metal additive manufacturing method of this embodiment in detail. This example analyzes the data of the manufactured sample block. The designed length, width, and height of the sample block are 10 mm (millimeters) * 10 mm * 10 mm, and the material is 316L stainless steel. The laser scanning power is 275 W (watts), the scanning layer thickness is 0.1 mm, the scanning speed is 1250 mm / s (millimeters per second), the scanning spacing is 0.13 mm, and the energy density is 16.92307692 J / mm 3(Joules per cubic millimeter), the density of the sample block is 7.1759 g / cm 3 (grams per cubic centimeter), the densification rate is 89.9235589%.
[0124] Figure 2 This is the flowchart of the work operation process for 3D registration in this example.
[0125] S201. Process the design data of the laser trajectory.
[0126] The design data of the laser trajectory includes the coordinate information on all laser paths during the manufacturing process generated by the 3D printer. In this step, its coordinates (unit: millimeter mm) are read, and then an empty image (with all image pixel values being 0) is created. The laser trajectory under the ideal printing condition is drawn on the empty image according to the read coordinates. Figure 3 This is the laser trajectory in this example.
[0127] S202. Solid filling of the laser trajectory. Figure 4 This is the solid filling diagram of the laser trajectory in this example.
[0128] In this step, methods such as the Flood Fill Algorithm and Boolean logic operations are run on the ideal trajectory on the image. The initial seed point for flood filling is set as the background point at the edge of the image. The solid part is filled on the image, and finally, the ideal situation slice diagram under the pre - manufacturing design is generated.
[0129] S203. Three - dimensional reconstruction of the three - dimensional point cloud model of the design data of the laser trajectory.
[0130] S204. Obtain the process monitoring data during the product processing (i.e., the three - dimensional image data during the product processing). The process monitoring data includes molten pool monitoring data, tomography monitoring images, powder spreading monitoring images, sintering monitoring images, and molten pool monitoring composite images. Figure 5 This is the sintering monitoring image, Figure 6 This is the tomography monitoring image, Figure 7 This is the molten pool monitoring composite image.
[0131] The size of the processed molten pool data used in this example is 2699mm * 2700mm, the size of the tomography monitoring image is 2560mm * 2160mm, and the sizes of the powder spreading and sintering images are 2588mm * 2579mm. The area of the 21st sample block in the data is extracted. The size of the sintering area is 174mm * 174mm, the pixel size is 0.058mm * 0.058mm, the size of the molten pool area is 190mm * 190mm, the pixel size is 0.053mm * 0.053mm, the size of the tomography area is 80mm * 80mm, and the pixel size is 0.125mm * 0.125mm.
[0132] S205. 3D point cloud model of the 3D reconstruction process monitoring data.
[0133] Specifically: perform adaptive threshold segmentation on the preprocessed process monitoring data to generate a binary image. Stack the binary images into a 3D image with a layer thickness consistent with the manufacturing design data, which is 0.1 mm, to generate a 3D model and convert it into a 3D point cloud.
[0134] S206. Obtain CT scan data (i.e., the first image data of the 3D image of the processed product). Figure 8 This is the CT scan tomogram of this example.
[0135] S207. 3D point cloud model of the 3D reconstruction of CT scan data.
[0136] Through the ICP registration algorithm, the 3D point cloud models of the design data of the 3D laser trajectory, the 3D point cloud model of the process monitoring data, and the 3D point cloud model of the CT scan data can be 3D registered, and the transformation parameters between any two of them can be output.
[0137] Take the 3D registration of CT data and tomography data as an example. Let the CT data be the registration data (moving data), and the tomography data be the reference data (fixed data). Run the ICP registration algorithm. Finally, the transformation matrix Transformation Matrix from CT data to tomography data is:
[0138]
[0139] According to the calculation, the final registration error (RMS) of the iteration is 0.093. Figure 9 This is the 3D registration result of mapping CT data to tomography data.
[0140] This example is implemented based on software, adopting an automated image registration and data analysis process, reducing manual intervention, improving the efficiency of data analysis, making data analysis faster and more accurate, and helping to quickly respond to problems in the manufacturing process. This example also uses 3D image registration technology to accurately align the image data at different stages in the metal additive manufacturing process with the CT data after manufacturing, enabling more accurate analysis of the minor changes and defects of the product during the manufacturing process, thereby optimizing the process. This method can reduce misjudgments caused by image deviations, improve the stability of the manufacturing process and product quality, and ensure the dimensional accuracy and surface quality of the final product.
[0141] The ideal laser trajectory design of the same layer of data after 3D registration is as Figure 10 shown, the CT data is as Figure 11 shown, the sintering data is as Figure 12 shown, and the tomography data is as Figure 13As shown, the molten pool data is as Figure 14 shown.
[0142] Example 2
[0143] Corresponding to the foregoing embodiments of the metal additive manufacturing method, the present disclosure also provides an embodiment of a metal additive manufacturing system.
[0144] As Figure 15 shown, the metal additive manufacturing system includes:
[0145] A first acquisition module 1 for acquiring training data. Among them, the training data includes first image data for characterizing the three-dimensional image of the processed product and the product defect result.
[0146] Among them, the first image data is obtained by computer tomography.
[0147] A model training module 2 for training a defect detection model using the training data to obtain a defect detection model for predicting product defects.
[0148] A second acquisition module 3 for acquiring second image data registered with the first image data. Among them, the second image data is used to characterize the three-dimensional image data during the product processing.
[0149] Among them, the second image data type includes at least one of powder spreading data, sintering data, tomography data, and molten pool data.
[0150] Among them, the second image data is determined according to a preset laser trajectory.
[0151] A model application module 4 for inputting the second image data into the defect detection model to output a defect detection result.
[0152] In an optional embodiment, the second acquisition module 3 is further configured to acquire three-dimensional image data during the product processing.
[0153] As Figure 15 shown, the metal additive manufacturing system further includes:
[0154] A determination module 5 for determining a transformation matrix between the first image data and the three-dimensional image data during the product processing by the iterative closest point algorithm; and further for determining the second image data registered with the first image data through the transformation matrix and the first image data.
[0155] In an optional embodiment, as Figure 15 shown, the metal additive manufacturing system further includes:
[0156] A conversion module 6 is configured to convert the first image data into a first three-dimensional point cloud, and is further configured to convert the three-dimensional image data during the product processing into a second three-dimensional point cloud.
[0157] A determination module 5 is further configured to determine the optimal rigid body transformation between the first three-dimensional point cloud and the second three-dimensional point cloud as a transformation matrix through the iterative closest point algorithm.
[0158] In an alternative embodiment, as Figure 15 shown, the conversion module 6 is further configured to convert the first image data into a first three-dimensional binary label map through single-threshold segmentation; is further configured to convert the three-dimensional image data during the product processing into a second three-dimensional binary label map; is further configured to convert the first three-dimensional binary label map into a first three-dimensional point cloud; and is further configured to convert the second three-dimensional binary label map into a second three-dimensional point cloud.
[0159] In an alternative embodiment, the second acquisition module 3 is further configured to obtain the three-dimensional image data during the product processing by extracting the region of interest, cropping and synthesizing each layer of two-dimensional image data of the processing process, and calibrating the three-dimensional voxel size.
[0160] For the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution.
[0161] Embodiment 3
[0162] Figure 16 The structure diagram of an electronic device shown in an exemplary embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and configured to run on the processor. When the processor executes the computer program, it implements the metal additive manufacturing method of any of the above embodiments. Figure 16 The displayed electronic device 160 is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.
[0163] As Figure 16 shown, the electronic device 160 may be presented in the form of a general computing device, for example, it may be a server device. The components of the electronic device 160 may include, but are not limited to: at least one of the above processors 161, at least one of the above memories 162, and a bus 163 connecting different system components (including the memory 162 and the processor 161).
[0164] The bus 163 includes a data bus, an address bus, and a control bus.
[0165] The memory 162 may include volatile memory, such as random access memory (RAM) 1621 and / or cache memory 1622, and may further include read-only memory (ROM) 1623.
[0166] The memory 162 may also include a program tool 1625 (or utility) having a set (at least one) of program modules 1624. Such program modules 1624 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.
[0167] The processor 161 executes various functional applications and data processing by running computer programs stored in the memory 162, such as the metal additive manufacturing method provided in any of the above embodiments.
[0168] The electronic device 160 may also communicate with one or more external devices 164 (such as a keyboard, a pointing device, etc.). Such communication may be carried out through an input / output (I / O) interface 165. And, the electronic device 160 may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 166. As shown in the figure, the network adapter 166 communicates with other modules of the electronic device 160 through the bus 163. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in combination with the electronic device 160, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.
[0169] It should be noted that, although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, such a division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above may be embodied in one unit / modules. Conversely, the features and functions of one unit / modules described above may be further divided and embodied by multiple units / modules.
[0170] Embodiment 4
[0171] The embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the metal additive manufacturing method provided in any of the above embodiments is implemented.
[0172] Among them, the readable storage medium can more specifically include, but is not limited to: portable disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, optical storage devices, magnetic storage devices, or any suitable combination of the above.
[0173] Embodiment 5
[0174] The embodiments of the present disclosure also provide a computer program product, including a computer program, which implements the metal additive manufacturing method according to any one of the above when executed by a processor.
[0175] Among them, the program code for executing the computer program product of the present disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on the user device, partially on the user device, executed as an independent software package, partially on the user device and partially on a remote device, or entirely on a remote device.
[0176] Although the specific implementation manners of the present disclosure have been described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Without departing from the principles and essence of the present disclosure, those skilled in the art can make various changes or modifications to these implementation manners, but these changes and modifications all fall within the protection scope of the present disclosure.
Claims
1. A metal additive manufacturing method, characterized in that, The metal additive manufacturing method includes: Obtaining training data; wherein, the training data includes first image data for characterizing the three-dimensional image of the processed product and product defect results; Training a defect detection model using the training data to obtain a defect detection model for predicting product defects; Obtaining second image data registered with the first image data; wherein, the second image data is used to characterize the three-dimensional image data during the product processing; Inputting the second image data into the defect detection model to output a defect detection result.
2. The metal additive manufacturing method according to claim 1, characterized in that, The step of obtaining second image data registered with the first image data includes: Obtaining three-dimensional image data during the product processing; Determining a transformation matrix between the first image data and the three-dimensional image data during the product processing through the iterative closest point algorithm; Determining second image data registered with the first image data through the transformation matrix and the first image data.
3. The metal additive manufacturing method according to claim 2, characterized in that, The step of determining a transformation matrix between the first image data and the three-dimensional image data during the product processing through the iterative closest point algorithm includes: Converting the first image data into a first three-dimensional point cloud and converting the three-dimensional image data during the product processing into a second three-dimensional point cloud; Determining the optimal rigid body transformation between the first three-dimensional point cloud and the second three-dimensional point cloud as the transformation matrix through the iterative closest point algorithm.
4. The metal additive manufacturing method according to claim 3, characterized in that, The step of converting the first image data into a first three-dimensional point cloud and converting the three-dimensional image data during the product processing into a second three-dimensional point cloud includes: Converting the first image data into a first three-dimensional binary label map through single-threshold segmentation, and converting the three-dimensional image data during the product processing into a second three-dimensional binary label map; Converting the first three-dimensional binary label map into a first three-dimensional point cloud and converting the second three-dimensional binary label map into a second three-dimensional point cloud.
5. The metal additive manufacturing method according to claim 2, wherein The step of obtaining three-dimensional image data during the product processing includes: Extracting a region of interest, cropping and synthesizing two-dimensional image data of each layer of the processing process, and calibrating the three-dimensional voxel size to obtain three-dimensional image data during the product processing.
6. The metal additive manufacturing method according to claim 1, wherein, The first image data is obtained through computed tomography; And / or The type of the second image data includes at least one of powder spreading data, sintering data, tomography data, and molten pool data; And / or The second image data is determined according to a preset laser trajectory.
7. A metal additive manufacturing system, characterized in that, The metal additive manufacturing system includes: A first acquisition module for obtaining training data; wherein, the training data includes first image data for characterizing the three-dimensional image of the processed product and product defect results; A model training module for training a defect detection model using the training data to obtain a defect detection model for predicting product defects; A second acquisition module for obtaining second image data registered with the first image data; wherein, the second image data is used to characterize the three-dimensional image data during the product processing; The model application module is used to input the second image data into the defect detection model to output a defect detection result.
8. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and adapted to run on the processor, characterized in that, When the processor executes the computer program, it implements the metal additive manufacturing method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the metal additive manufacturing method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the metal additive manufacturing method as claimed in any one of claims 1 - 6.
Citation Information
Cited By
VR-assisted fatigue fracture feature recognition and life prediction method and device
CN121600308A