A Method, Device and Medium for Local Characterization of Ceramic Microstructures in Laser Additive Manufacturing

Abstract

The present invention discloses a method, device, and medium for local characterization of ceramic microstructures in laser additive manufacturing. The method comprises: obtaining image data of an SLS ceramic slurry product, inputting the image data into a trained local porosity characterization model, and outputting porosity; the local porosity characterization model processes the input image data as follows: extracting features from the image data to obtain a feature map; identifying proposed regions on the feature map, and aligning the identified proposed regions to obtain an anchor map; obtaining an output image based on the anchor map, the output image including the original image, a detection frame, and a segmentation mask; calculating the area of the mask based on the segmentation mask, and calculating the ratio of the mask area to the original image area as the porosity. The present invention uses the local porosity characterization model to obtain each mask and calculate the porosity, enabling rapid ceramic porosity characterization. The present invention can be widely applied to the field of laser additive manufacturing of alumina ceramics.

Description

Technical Field

[0001] The present invention relates to the field of laser additive manufacturing of alumina ceramics, and in particular to a method, device and medium for local characterization of a microstructure of ceramics manufactured by laser additive manufacturing. Background Art

[0002] Advanced ceramic materials have been widely used in many fields such as aviation, biomedical implants, energy production systems, optoelectronics, etc. due to their excellent mechanical, thermodynamic and dielectric properties. Based on these applications of advanced ceramic materials, advanced ceramic materials need to be given different properties, such as complex geometric shapes, different degrees of densification, etc. However, it is difficult to meet this demand by relying solely on traditional mechanical processing methods. Laser processing is a new advanced manufacturing technology that is famous for its high precision, high processing efficiency and complex shape manufacturing. Combining the manufacturing needs of advanced ceramic products with the advantages of laser processing, laser processing is expected to become a new processing method for advanced ceramic products. For laser additive manufacturing of advanced ceramics, selective laser sintering (SLS) is an optimal solution and is widely used in laser processing of ceramics.

[0003] During the SLS process, the ceramic slurry to be processed is exposed to a CO2 laser. Due to the high absorption rate of the ceramic slurry to the CO2 laser, the high-intensity laser beam causes the irradiated area to heat up rapidly in a short period of time, thereby depositing and densifying the ceramic slurry to complete the ceramic processing. The entire process is very rapid and can be completed in seconds. However, since the spatial intensity distribution of the laser follows a Gaussian distribution, the temperature field of the laser acting on the ceramic slurry during the SLS process also follows a Gaussian distribution, which directly causes thermal imbalance during the processing.

[0004] Because of this thermal imbalance during the SLS process, the properties of the processed ceramic product will be affected to a certain extent, such as porosity (the ratio of the pore volume within the ceramic product to the overall volume) and grain size. Generally speaking, a higher effective processing temperature will improve the sintering effect, causing the grains of the ceramic product to grow, thereby reducing the porosity. Based on the above theory, the thermal imbalance during the SLS process will cause an imbalance in the properties of the ceramic product. Under different characteristics of ceramic products, their performance, such as mechanical properties and thermal resistance, will also be different. Therefore, in order to obtain the ceramic product that best meets the needs, it is necessary to characterize the distribution of its characteristics in order to optimize the entire processing process.

[0005] Existing ceramic porosity characterization techniques are mainly divided into the following two categories: 1. Direct porosity testing of bulk ceramic materials. 2. Using image analysis software to analyze SEM images of ceramic products and determine the porosity of the corresponding images.

[0006] The first type of technology directly tests the bulk material of the ceramic product, returning the porosity of the entire material and failing to provide local information. This has limitations for analyzing the porosity distribution of SLS-processed ceramic products.

[0007] For the second type of technology, using SEM images to analyze the local porosity of ceramic products can help us understand local information. However, the use of these image analysis software also has its corresponding limitations. First, mainstream image analysis software often uses manually set thresholds to classify different pixel sizes to segment and count the desired targets (for porosity, the target we need to count is the pores). This process is too subjective, and the segmentation performance may decrease accordingly when facing targets of different shapes, thereby causing a decrease in the accuracy of the analysis. Second, when faced with large amounts of data, using image analysis software for analysis will be very complicated and time-consuming. Summary of the Invention

[0008] In order to solve at least one of the technical problems existing in the prior art to a certain extent, the purpose of the present invention is to provide a method, device and medium for local characterization of ceramic microstructures in laser additive manufacturing.

[0009] The technical solution adopted in the present invention is:

[0010] A method for local characterization of a ceramic microstructure during laser additive manufacturing, comprising the following steps:

[0011] Obtain image data of SLS ceramic slurry products, input the image data into the trained local porosity characterization model, and output the porosity;

[0012] The local porosity characterization model processes the input image data as follows:

[0013] Performing feature extraction on the image data to obtain a feature map;

[0014] Identify proposed regions on the feature map and align the identified proposed regions to obtain an anchor frame map; the anchor frame map includes at least one detection frame;

[0015] Obtaining an output image according to the anchor frame image, wherein the output image includes the original image, the detection frame, and the segmentation mask;

[0016] The area of the mask is calculated based on the segmentation mask, and the ratio of the mask area to the original image area is calculated as the porosity.

[0017] Furthermore, the local porosity characterization model is constructed by the following steps:

[0018] Acquire images of SLS ceramic slurry products and process the images to obtain image datasets;

[0019] Performing annotation processing on the images in the image dataset;

[0020] Using the labeled image data to train the local porosity characterization model;

[0021] The labeling process includes: adjusting the color scale of the image and labeling the pores in the image.

[0022] Furthermore, the processing of the image to obtain the image data set includes:

[0023] Performing resolution enhancement processing on images of SLS ceramic slurry products at different powers to obtain a first image;

[0024] Cropping the first image according to a preset size requirement to obtain a second image;

[0025] Images are extracted from all the second images as an image dataset.

[0026] Furthermore, the image of the SLS ceramic slurry product is a SEM image, and the pores in the image are annotated using the following standards:

[0027] 1) The connected area of black pixels must be greater than or equal to n pixels;

[0028] 2) The pixel channel value of the black area is lower than m;

[0029] 3) Porosity occurs around the boundaries of grains;

[0030] Wherein, n and m are positive integers.

[0031] Furthermore, the local porosity characterization model uses the Mask R-CNN network as a basic network, and the local porosity characterization model includes a backbone network, a region proposal network, and a head network;

[0032] The backbone network is used to extract features from the input original image and output a feature map;

[0033] The region proposal network is used to extract the target according to the feature map to obtain the proposed region; the proposed region is aligned with the region on the feature map to obtain the anchor frame.

[0034] The head network includes a parallel classification module and a segmentation module. The classification module is used to classify the desired target and deploy the detection frame. The segmentation module segments the specific outline of the desired target.

[0035] The final detection frame is obtained through the classification module, and the segmentation mask is obtained through the segmentation module.

[0036] Furthermore, the structure of the backbone network is a ResNet101-FPN network structure.

[0037] Furthermore, the local porosity characterization model also includes a convolutional attention module, which is used to improve the feature extraction capability.

[0038] Furthermore, the structure of the backbone network is a ResNet-FPN network structure;

[0039] The convolutional attention module is set in the ResNet network; or,

[0040] The convolutional attention module is set in the FPN network.

[0041] Another technical solution adopted in the present invention is:

[0042] A device for local characterization of ceramic microstructures in laser additive manufacturing, comprising:

[0043] at least one processor;

[0044] at least one memory for storing at least one program;

[0045] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0046] Another technical solution adopted in the present invention is:

[0047] A computer-readable storage medium stores a program executable by a processor, wherein the program executable by the processor is used to perform the method described above when executed by the processor.

[0048] The beneficial effects of the present invention are as follows: the present invention obtains each mask through a local porosity characterization model and calculates the porosity, which can quickly realize the ceramic porosity characterization and help improve the laser additive manufacturing process of alumina ceramics. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present invention or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0050] Figure 1 is a flow chart of steps for processing input image data by the local porosity characterization model according to an embodiment of the present invention;

[0051] Figure 2 It is a technical roadmap of a method for local characterization of ceramic microstructures in laser additive manufacturing according to an embodiment of the present invention;

[0052] Figure 3 is a schematic diagram of pore marking in an embodiment of the present invention; wherein, Figure 3 (a) is the original unprocessed image; Figure 3 (b) is the result after color scale adjustment and pore marking;

[0053] Figure 4 is a schematic diagram of a local porosity characterization model according to an embodiment of the present invention;

[0054] Figure 5 Schematic diagram of the internal structure of CBAM in an embodiment of the present invention;

[0055] Figure 6 Schematic diagram of the method of embedding CBAM into ResNet and FPN in the example of the present invention; wherein, Figure 6 (a) Schematic diagram of embedding CBAM into ResNet; Figure 6 (b) Schematic diagram of embedding CBAM into FPN. In the figure, “C*” represents the feature extraction stage, “P*” represents the feature fusion stage, and “2×Up” represents the 2x upsampling process.

[0056] Figure 7 1 is a flow chart of a reliability verification experiment of a local porosity characterization model according to an embodiment of the present invention;

[0057] Figure 8 is the verification result of the local porosity characterization model in the embodiment of the present invention; wherein, Figure 8 (a) is the detection result on test set 1; Figure 8 (b) is the detection result on test set 2. DETAILED DESCRIPTION

[0058] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. The step numbers in the following embodiments are provided for ease of explanation only and do not limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0059] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention.

[0060] In the description of the present invention, "several" means one or more, "many" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0061] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0062] Explanation of terms:

[0063] Laser additive manufacturing: Developed from additive manufacturing, it uses high-energy laser pulses as an energy source to perform additive manufacturing on materials. It can be used to manufacture high-performance products such as metals and ceramics.

[0064] Object detection: A method in deep learning that locates and classifies objects of interest in an image.

[0065] Instance segmentation: A new method based on target detection. Compared with traditional target detection that uses a detection box to locate the target of interest, the instance segmentation method proposes to segment the contour of the target of interest on this basis to obtain the contour of the target of interest.

[0066] Mask R-CNN: Mask Region-based Convolutional Neural Network is a deep learning algorithm that combines target detection and instance segmentation capabilities.

[0067] Supervised learning: The most common model training method in deep learning, in which all data in the training set are labeled so that the model can be trained from beginning to end.

[0068] Transfer learning: Another learning method in deep learning. It was proposed to improve training results as much as possible when data is insufficient, and to transfer knowledge across domains (datasets).

[0069] AP value: One of the most critical indicators to measure the performance of the target detection algorithm. The value range is 0 to 1. The higher the value, the better the performance.

[0070] SEM image: an image obtained by scanning electron microscope (SEM).

[0071] like Figure 1 As shown, this embodiment provides a method for local characterization of ceramic microstructures in laser additive manufacturing, comprising the following steps:

[0072] Obtain image data of SLS ceramic slurry products, input the image data into the trained local porosity characterization model, and output the porosity;

[0073] The local porosity characterization model performs the following processing on the input image data:

[0074] S1. Extract features from the image data to obtain a feature map;

[0075] S2. Identify the proposed regions on the feature map and align the identified proposed regions to obtain an anchor frame map; the anchor frame map includes at least one detection frame;

[0076] S3. Obtain an output image according to the anchor frame diagram, wherein the output image includes the original image, the detection frame, and the segmentation mask;

[0077] S4. Calculate the area of the mask according to the segmentation mask, and calculate the ratio of the mask area to the original image area as the porosity.

[0078] In this embodiment, the original image referred to in step S3 is a broad concept and can refer to either the input original image or the data of the input original image. That is, the output image includes the data of the original image. This embodiment combines deep learning methods with laser processing, specifically with the laser additive manufacturing process for alumina ceramics. By combining the advantages of deep learning methods, the laser additive manufacturing process for alumina ceramics can be improved, achieving innovation in cross-disciplinary fields.

[0079] For SEM images of alumina ceramic products manufactured using laser additive manufacturing, it is necessary to obtain the porosity of the image and locate each specific pore. Furthermore, contour extraction of each located pore is required to calculate the size of each pore. Therefore, the present invention combines deep learning methods in the areas of object detection and instance segmentation to develop a deep learning model. Figure 2 The overall technical route of the entire invention is shown. The overall technical route will be specifically described below.

[0080] To acquire the dataset, SEM images of SLS ceramic slurry products were obtained at different power levels and processed to enhance their resolution, resulting in a final resolution of 2048 × 1536 pixels. After acquiring the SEM images, they were cropped to a fixed size to reduce the number of features (pores), with the cropped images being 256 × 256 pixels in size. Finally, a random sample of all the cropped images was taken, and 210 images were selected as the total dataset. Seventy percent of these images were divided into the training set, and the remaining images were divided into the validation set. After the dataset was divided, all images were annotated (a common process for establishing a target detection dataset, i.e., manually annotating the objects to be detected on the image). Because the pore features in the unprocessed SEM images were not obvious, all images in the dataset were preprocessed using GIMP software to adjust the color scale before being annotated. Figure 3 Schematic diagram of color gradation adjustment and pore annotation. For pore annotation, we established the following criteria based on SEM image acquisition principles and material properties: 1) the connected area of black pixels must be greater than or equal to 10 pixels; 2) the pixel channel value of the black area must be less than 10; and 3) pores must occur around grain boundaries.

[0081] After completing the establishment of the dataset, Mask R-CNN was selected as the basic network for modeling the representation model. Figure 4 The specific structure of the representation model is shown. Figure 4As shown in the figure, during detection and segmentation tasks, the input SEM image first passes through the backbone network for feature extraction to produce a feature map for subsequent use. The most commonly used backbone network architecture in Mask R-CNN is the ResNet-FPN architecture. ResNet (Residual Neural Network) is a convolutional neural network (CNN) used for feature extraction. The Feature Pyramid Network (FPN) is a proposed architecture for refining feature maps. The resulting feature map then enters the Region Proposal Network and undergoes RoI Align. The former is used to generate proposed regions (the objects the network focuses on), while the latter aligns the proposed regions on the feature map with those on the original image, helping to improve the accuracy of subsequent segmentation tasks. After these two processes, the final anchor map, or initial detection map, is obtained. Each object in the anchor map has at least one detection box, which is then further filtered by the head network to obtain the final detection box. The head network consists of two parallel structures: one for classifying the desired object and assigning detection boxes, and the other for segmenting the specific outline of the desired object. After a series of processing, the anchor frames input into these two structures respectively can obtain the final output image, which consists of the original image, the detection frame, and the segmentation mask. At this point, the entire detection and segmentation task is completed, but to obtain the characterization data of the porosity, it is not enough to rely solely on the default output of Mask R-CNN. To solve this problem, this embodiment modifies the program code of the output end of Mask R-CNN. As an optional implementation method, by calling the OpenCV interface in python, the mask contour data output by Mask R-CNN is extracted, and then the area (occupied pixels) of each output mask is calculated through the function in the interface to obtain the corresponding area it occupies. Finally, the porosity is calculated by calculating the ratio of the mask area in the output image to the size of the original image. In particular, if the actual scale is known, the true size of each pore can be obtained.

[0082] As an optional implementation, we propose fixing all architectures except the backbone network and performing supervised learning and transfer learning respectively, using different backbone networks. After comparing various training combinations, we found that supervised learning achieved the best model performance when using ResNet101-FPN as the backbone network. In this example, this model is named ResNet101-FPN-S.

[0083] As an optional implementation, we've improved the underlying architecture of Mask R-CNN to enhance its detection and segmentation capabilities. Specifically, we've incorporated an attention module into Mask R-CNN to improve the network's feature extraction capabilities, thereby enhancing the overall model performance. This added attention module is broadly similar to the convolutional block attention module (CBAM) proposed by Woo Sanghyun et al. Figure 5 The detailed structure of CBAM is shown. CBAM is proposed to refine feature maps. Therefore, its input is a feature map F, which can be extracted by a CNN or processed by FPN. After a series of operations and processing, the final feature map F with attention is formed by combining channel attention and spatial attention.

[0084] Among them, according to the way Mask R-CNN processes feature maps, CBAM can be set in the ResNet network or in the FPN network. Figure 6 The methods of embedding CBAM into ResNet and FPN are shown respectively. Figure 6 (a), for embedding CBAM into ResNet, we add CBAM between each convolution block of ResNet and before each residual connection; according to Figure 6 (b) Description, for embedding CBAM into FPN, we add CBAM after each feature fusion process of FPN, that is, after each XOR operation on different feature maps.

[0085] To verify the effectiveness of adding the attention module, we conducted comparative experiments, measuring model performance using the AP (Average Precision) metric. The results are shown in Table 1. By comparing the AP values, we can see that adding CBAM to Mask R-CNN does improve network performance to a certain extent. Furthermore, adding the attention module at different locations yields different results. In particular, adding the attention module to ResNet yields the best performance improvement. Furthermore, the increase in parameter size after adding the attention module is within an ideal range, which does not significantly increase the computing power consumption during model training.

[0086] Table 1. Comparison of results of adding CBAM to the characterization model

[0087]

[0088] Through the above technical route, we obtained the optimal model, namely ResNet101-FPN-S with attention module. In addition, to verify the cross-dataset deployment capability of this model, we designed an experiment to verify it. The flowchart of the entire verification experiment is as follows Figure 7 As shown. First, a random seed of 0 is used to generate the dataset. Based on the characteristics of our SEM image dataset, the simulated pictures we generated consist of randomly generated black pixels and a white background to restore the real data as much as possible. The database of this verification experiment includes a total of 50 pictures, 70% of which are used for training and the rest are used for verification. In order to verify the model's cross-dataset deployment capability, that is, transfer learning capability, we used this lightweight dataset to fine-tune the ResNet101-FPN-S model that was trained when the aforementioned model was established, and shortened the training time to one-third of the original. After training, the obtained model achieved an AP value of 0.987. Secondly, two test sets containing 500 pictures were generated using different random seeds 1 and random seeds 2 (different random seeds are used to ensure that the datasets come from different distributions, and the simulated porosity of each picture in the dataset has been calculated in advance and used as the true value). Finally, the two test sets were input into the trained models to obtain test results 1 and 2, as shown in Figure 8 As shown. Figure 8 The results show that the model's detection results are nearly identical to the true values, with an average error of only approximately 1.73% across both test sets. This demonstrates that a porosity characterization model trained on a real-world pore dataset can be transferred to similar datasets with high accuracy. Furthermore, we provide a usable pre-trained porosity characterization model, enabling its deployment in diverse scenarios without the labor-intensive construction of large datasets.

[0089] The above is the technical solution for the entire invention. Ultimately, the model can calculate the porosity of a ceramic product by inputting an SEM image, outputting the corresponding porosity within the SEM image, and implementing efficient and intelligent localized characterization of the porosity of alumina ceramics in laser additive manufacturing, thereby optimizing the entire laser processing process for ceramics.

[0090] It should be noted that the above examples primarily use SEM images for illustration, but the method of this embodiment is not limited to SEM images and can also be used with other images. For example, electron microscope images of laser-processed glass can be used to train a model to detect and segment bubbles in glass. Alternatively, images of surface defects on certain products can be analyzed to determine the defect outline and specific size. It should be noted that this method can be used with any image that has specific features.

[0091] In summary, this embodiment has the following advantages and beneficial effects compared to the prior art:

[0092] (1) Compared with existing ceramic porosity detection technologies such as ultrasonic testing, thermal imaging, and CT imaging, the present invention has the advantage of using SEM images for analysis to characterize the local porosity of ceramic products, thereby determining the surface porosity heterogeneity of laser additively manufactured alumina ceramic products. This local characterization method can optimize the entire process to control product quality.

[0093] (2) Compared to existing local characterization techniques, which analyze SEM images using mainstream image processing software, the present invention has the following advantages: 1) It does not rely solely on pixel size for classification; it can utilize machine learning capabilities to classify different pores, thereby improving characterization accuracy. 2) When faced with large amounts of data, the analysis process using image processing software is very cumbersome. In contrast, the present invention accepts large amounts of data input, greatly reducing manual operations.

[0094] This embodiment also provides a device for local characterization of ceramic microstructures in laser additive manufacturing, comprising:

[0095] at least one processor;

[0096] at least one memory for storing at least one program;

[0097] When the at least one program is executed by the at least one processor, the at least one processor implements Figure 1 The method shown.

[0098] A laser additive manufacturing ceramic microstructure local characterization device of this embodiment can execute a laser additive manufacturing ceramic microstructure local characterization method provided by the method embodiment of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0099] The present application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs Figure 1 The method shown.

[0100] This embodiment also provides a storage medium storing instructions or programs that can execute a method for local characterization of ceramic microstructures in laser additive manufacturing provided by an embodiment of the method of the present invention. When the instructions or program are run, any combination of implementation steps of the method embodiment can be executed, and the corresponding functions and beneficial effects of the method can be obtained.

[0101] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0102] Furthermore, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise indicated, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It will also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skill of an engineer. Therefore, a person skilled in the art using ordinary skill will be able to implement the present invention set forth in the claims without undue experimentation. It will also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0103] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0104] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0105] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0106] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0107] In the above description of this specification, reference to the terms "one embodiment / example," "another embodiment / example," or "certain embodiments / examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0108] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0109] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for local characterization of ceramic microstructures in laser additive manufacturing, characterized in that: The following steps are involved: Obtain image data of SLS ceramic slurry products, input the image data into the trained local porosity characterization model, and output the porosity; The local porosity characterization model processes the input image data as follows: Performing feature extraction on the image data to obtain a feature map; Identify the proposed regions on the feature map and align the identified proposed regions to obtain the anchor frame map; The anchor frame diagram includes at least one detection frame; Obtaining an output image according to the anchor frame image, wherein the output image includes the original image, the detection frame, and the segmentation mask; The area of the mask is calculated based on the segmentation mask, and the ratio of the mask area to the original image area is calculated as the porosity; The local porosity characterization model is constructed by the following steps: Acquire images of SLS ceramic slurry products and process the images to obtain image datasets; Performing annotation processing on the images in the image dataset; Using the labeled image data to train the local porosity characterization model; The annotation processing includes: adjusting the color scale of the image and annotating the pores in the image; The step of processing the image to obtain the image data set includes: Performing resolution enhancement processing on images of SLS ceramic slurry products at different powers to obtain a first image; Cropping the first image according to a preset size requirement to obtain a second image; extracting images from all the second images as an image dataset; The image of the SLS ceramic slurry product is a SEM image, and the pores in the image are annotated using the following standards: 1) The connected area of black pixels must be greater than or equal to n pixels; 2) The pixel channel value of the black area is lower than m; 3) Porosity occurs around the boundaries of grains; Wherein, n and m are positive integers; The local porosity characterization model uses the Mask R-CNN network as the basic network, and the local porosity characterization model includes a backbone network, a region proposal network, and a head network; The backbone network is used to extract features from the input original image and output a feature map; The region proposal network is used to extract the target according to the feature map to obtain the proposed region; the proposed region is aligned with the region on the feature map to obtain the anchor frame. The head network includes a parallel classification module and a segmentation module. The classification module is used to classify the desired target and deploy the detection frame. The segmentation module segments the specific outline of the desired target. The final detection frame is obtained through the classification module, and the segmentation mask is obtained through the segmentation module.

2. The method for local characterization of ceramic microstructures in laser additive manufacturing according to claim 1, characterized in that: The structure of the backbone network is the ResNet101-FPN network structure.

3. The method for local characterization of ceramic microstructures in laser additive manufacturing according to claim 1, characterized in that: The local porosity characterization model also includes a convolutional attention module, which is used to improve the feature extraction capability.

4. The method for local characterization of ceramic microstructures in laser additive manufacturing according to claim 3, characterized in that: The structure of the backbone network is a ResNet-FPN network structure; The convolutional attention module is set in the ResNet network; or, The convolutional attention module is set in the FPN network.

5. A device for local characterization of ceramic microstructures in laser additive manufacturing, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 4.

6. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 4 when executed by the processor.

Images