Intelligent casting process design method based on similar process reuse

Abstract

The invention discloses an intelligent casting process design method based on similar process reuse, and belongs to the technical field of casting process design. The method comprises the steps that posture normalization processing is conducted on a three-dimensional model of a target casting; two-dimensional view pictures of three views and a section view of the casting three-dimensional model after attitude normalization processing are obtained, edge sharpening processing is carried out on the obtained pictures, and then shape features of the pictures after edge sharpening are obtained through a convolutional neural network; extracting process design characteristics of the casting three-dimensional model; splicing and fusing the extracted process design features of the casting three-dimensional model and the shape features of the three-dimensional model after edge sharpening, calculating the similarity between the fused features of the target casting and the features of each casting in a model library according to a cosine similarity calculation formula, and comparing and sorting according to a similarity calculation result; and screening the casting with the highest similarity as a similar casting, and determining the process route of the target casting based on the process route of the three-dimensional model of the similar casting.

Description

Technical Field

[0001] The present invention relates to the technical field of casting process design, and in particular to a casting process intelligent design method based on similar process reuse. Background Art

[0002] Casting is an indispensable process in the pillar industries such as aviation, aerospace, transportation, and shipbuilding, and is an important support and basic guarantee for the innovative development of many host products and high-end equipment. Among them, typical castings represented by satellite brackets, missile shells, casings, engine cases, aerospace engine blades, etc. have complex structures such as special-shaped curved surfaces, internal channels or structural holes, which leads to high complexity, difficulty and long cycle of process design.

[0003] In industrial manufacturing, more than 75% of product designs are case-based designs. Therefore, reusing and improving the mature processes of existing products can effectively shorten the product production cycle. The current process reuse relies on the method of manually retrieving casting process cards, which is inefficient and highly dependent on experience. With the advancement of Industry 4.0 and digital intelligent manufacturing, the method of process design through 3D models has been widely used in foundry companies. Therefore, accurate retrieval of 3D models is a key means to achieve process reuse.

[0004] Three-dimensional retrieval methods generally include two types: text-based and model-based. Compared with model-based retrieval, text annotation is difficult to fully and accurately describe model features, and text annotation requires a lot of manpower and material resources. Therefore, model-based retrieval methods are currently widely used. Model-based retrieval mainly performs automatic retrieval based on the actual content of the 3D model (geometric shape, topological structure, etc.), which can objectively and comprehensively represent the model compared to text. In addition, model-based retrieval is currently mainly divided into shape matching-based and data-driven methods. Among them, the shape matching-based method mainly compares the shape descriptors of the three-dimensional model for similarity calculation, such as D2 shape distribution, 3D Zernike matrix descriptor, Fourier descriptor, etc. These shape descriptors have translation and rotation invariance. However, for complex three-dimensional casting models, there is a loss of shape detail information in the process of extracting shape features, which reduces the accuracy of retrieval. Data-driven methods can be mainly divided into two methods: model-based and view-based. Among them, the model-based method represents the 3D model as voxels, point clouds and polygonal meshes, and then uses deep learning to extract features and perform similarity calculations. The shape features obtained based on deep learning show better performance than shape descriptors. However, the view-based method consumes a lot of memory in the calculation process for complex casting models due to high dimensionality and sparsity. The point cloud-based method uses the point coordinates in 3D space as input and does not fully consider the spatial structure (such as adjacency). The grid-based method relies on the resolution, regularity and uniformity of the grid. At the same time, these three methods have the limitation of high computational complexity.

[0005] Since the 2D views of 3D models are easy to obtain and highly distinguishable, and deep learning has achieved great success in extracting features from 2D images, 3D model retrieval based on 2D images has shown advanced performance and higher efficiency. However, the reuse of casting processes not only includes the similarity of part shapes, but more importantly, the similarity of process design features formed based on part shapes. Current 3D model retrieval methods are mainly based on shape similarity, ignoring the process design features contained in the 3D model, and only indirectly support process reusability between parts by combining more comprehensive shape semantics.

[0006] Therefore, there is an urgent need for a model retrieval method that can effectively retrieve similar castings while ensuring the similarity of process design. Summary of the invention

[0007] In order to solve some or all of the technical problems existing in the above-mentioned prior art, the present invention provides an intelligent design method for casting process based on similar process reuse, which uses deep learning that integrates process design knowledge to perform three-dimensional model retrieval. It can accurately and effectively retrieve similar castings based on the existing casting model database while ensuring the similarity of process design.

[0008] The technical solution of the present invention is as follows:

[0009] The present invention provides a casting process intelligent design method based on similar process reuse, comprising:

[0010] Performing posture normalization processing on the target casting three-dimensional model to obtain the casting three-dimensional model after posture normalization processing;

[0011] Obtaining two-dimensional view images of the main view, left view, top view and cross-sectional view of the three-dimensional model of the casting after posture normalization, and performing edge sharpening on the acquired images, and then using a convolutional neural network to obtain shape features of the edge-sharpened images;

[0012] Extract the process design features of the casting 3D model;

[0013] The extracted process design features of the casting 3D model are combined with the shape features of the 3D model after edge sharpening;

[0014] The similarity between the fused features of the target casting and the features of each casting in the model library is calculated according to the cosine similarity calculation formula, and the calculation results are compared and sorted;

[0015] The castings with the highest similarity are selected as similar castings, and the process route of the target casting is determined based on the process route of the 3D model of the similar castings.

[0016] Furthermore, in the above-mentioned intelligent design method of casting process based on similar process reuse, the posture normalization processing includes rotating or translating the target casting three-dimensional model so that the longest side of the target casting three-dimensional model is placed in a direction that coincides with or is parallel to the preset three-dimensional coordinate Z axis.

[0017] Furthermore, in the above-mentioned intelligent design method for casting process based on similar process reuse, shape feature extraction is performed on the casting three-dimensional model after posture normalization, including extracting three views and sectional views of the casting three-dimensional model.

[0018] Furthermore, in the above-mentioned intelligent design method of casting process based on similar process reuse, the sectional view includes a vertical plane in the positive direction of the preset three-dimensional coordinate axis X-axis as the sectional view, and the main view of the sectional view is taken as the sectional view, representing the internal structure of the casting three-dimensional model.

[0019] Furthermore, in the above-mentioned intelligent design method for casting process based on similar process reuse, sharpening the edges of the extracted shape features includes converting the two-dimensional views of the main view, left view, top view and cross-sectional view extracted from the three-dimensional model into corresponding line drawings.

[0020] Furthermore, in the above-mentioned intelligent design method of casting process based on similar process reuse, the process design features of the casting three-dimensional model are extracted, including the modulus, main wall thickness, average wall thickness, maximum wall thickness, minimum wall thickness, symmetry, aspect ratio, length-to-height ratio and width-to-height ratio of the casting three-dimensional model.

[0021] Furthermore, in the above-mentioned casting process intelligent design method based on similar process reuse, the cosine similarity calculation formula includes:

[0022]

[0023] Where A and B represent two n-dimensional vectors respectively.

[0024] Furthermore, in the above-mentioned casting process intelligent design method based on similar process reuse, the symmetry of the casting three-dimensional model is calculated by the following formula:

[0025] Sym=(Sym x +Sym Y +Sym Z ) / 3

[0026] Among them, Sym represents the symmetry of the three-dimensional model of the casting, Sym X Indicates the symmetry of the casting in the X-axis direction, which is obtained by calculating the similarity between the front view and the back view. Y Indicates the symmetry of the casting in the Y-axis direction, which is obtained by calculating the similarity between the left view and the rear view. Z Indicates the symmetry of the casting in the Z-axis direction, which is obtained by calculating the similarity between the top view and the bottom view.

[0027] Furthermore, in the above-mentioned casting process intelligent design method based on similar process reuse, the modulus is calculated by the following formula:

[0028]

[0029] In the above formula, M represents the modulus in centimeters (cm), and V represents the volume of the casting in cubic centimeters (cm 3 ), A represents the heat dissipation surface area of ​​the casting, in square centimeters (cm 2 ).

[0030] Furthermore, in the above-mentioned intelligent design method of casting process based on similar process reuse, after the two-dimensional view images of the main view, left view, top view and sectional view of the casting three-dimensional model are obtained after posture normalization, it also includes arranging them in a layout of two rows and two columns according to the left view, main view, top view and sectional view.

[0031] The main advantages of the technical solution of the present invention are as follows:

[0032] The intelligent design method of casting process based on similar process reuse of the present invention obtains two-dimensional view images of the main view, left view, top view and cross-sectional view of the three-dimensional model of the casting, sharpens the edges of the images, so that the contour shape of the two-dimensional view of the model is more obvious, and then adopts a convolutional neural network to obtain shape features, splices and fuses the extracted shape features and process design features, calculates the similarity between the target model and the model in the model library, determines the reuse process of the final target model, adopts a three-dimensional retrieval method based on two-dimensional images, and fuses the process design features of the three-dimensional model of the casting on the basis of the shape features of the three-dimensional model of the casting, so that the overall features of the model are more accurate and complete, and by fully considering the similar effects of process design, it is possible to find similar processes that can be reused while ensuring similar shapes, thereby improving the accuracy of the reused process and better ensuring the similarity of process design. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0034] Figure 1 It is a schematic flow chart of a casting three-dimensional model retrieval method of a casting process intelligent design method based on similar process reuse provided by an embodiment of the present invention;

[0035] Figure 2 A casting three-dimensional model library for similar process retrieval in a casting process intelligent design method based on similar process reuse provided by an embodiment of the present invention;

[0036] Figure 3 It is a schematic diagram of standardizing a casting three-dimensional model, extracting a two-dimensional view, and processing Sobel edges by using the casting process intelligent design method based on similar process reuse of the present invention;

[0037] Figure 4 It is a schematic diagram of wheel casting standardization, two-dimensional view extraction and Sobel edge processing using the casting process intelligent design method based on similar process reuse of the present invention;

[0038] Figure 5It is a schematic diagram of a flow chart of calculating symmetry in a casting process intelligent design method based on similar process reuse provided by an embodiment of the present invention;

[0039] Figure 6 The search results of the wheel hub casting in the casting model library by using the casting process intelligent design method based on similar process reuse provided by an embodiment of the present invention are shown in FIG. 1 , wherein the CLIP model represents the search based on shape features only, and the CLIP-PDK represents the search results after the process design features are introduced;

[0040] Figure 7 In the casting process intelligent design method based on similar process reuse provided by one embodiment of the present invention, the process of similar castings retrieved is applied to the numerical simulation of shrinkage distribution of the queried casting: (a) represents the process of the queried model; (b) represents the process of the CLIP-PDK_1 model; (c) represents the process of the CLIP-PDK_2 model; (d) represents the process of the CLIP_1 model;

[0041] Figure 8 The figure is a flow chart of a casting process intelligent design method based on similar process reuse according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] The technical solution provided by the embodiments of the present invention is described in detail below with reference to the accompanying drawings.

[0044] As attached Figure 1 -Attached Figure 8 As shown, an embodiment of the present invention provides a casting process intelligent design method based on similar process reuse, the method comprising:

[0045] The three-dimensional model of the target casting is subjected to posture normalization to obtain the three-dimensional model of the casting after posture normalization; the two-dimensional view images of the main view, left view, top view and cross-sectional view of the casting three-dimensional model are obtained, and the edges of the images are sharpened to make the contour shape of the two-dimensional view of the model more obvious, and then the shape features are obtained by using a convolutional neural network; the process design features of the casting three-dimensional model are extracted; the extracted process design features of the casting three-dimensional model are spliced ​​and fused with the shape features of the three-dimensional model after edge sharpening, and the similarity between the fused features of the target casting and the features of each casting in the model library is calculated according to the cosine similarity calculation formula, and the calculation results are compared and sorted; the casting with the highest similarity is selected as a similar casting, and the process route of the target casting is determined based on the process route of the three-dimensional model of the similar casting.

[0046] Therefore, the intelligent design method of casting process based on similar process reuse of the present invention adopts a three-dimensional retrieval method based on two-dimensional images, and integrates the process design features of the casting three-dimensional model on the basis of the shape features of the casting three-dimensional model, so that the overall features of the model are more accurate and complete. By fully considering the similar effects of process design, similar processes that can be reused can be found while ensuring similar shapes, thereby improving the accuracy of the reused process and better ensuring the similarity of process design.

[0047] Specifically, how to splice and fuse the extracted casting 3D model process design features and the edge-sharpened 3D model shape features is further described, specifically:

[0048] The three-view and cross-sectional views of the three-dimensional model of the casting after posture standardization are obtained, and the edges of these two-dimensional views are sharpened to obtain a clear casting shape contour. The shape feature is extracted using a convolutional neural network to obtain a shape feature vector A; the process design features of the three-dimensional model of the casting are extracted to obtain a process design feature B, and the shape feature vector A is spliced ​​with the process design feature B to obtain a shape and process fusion feature A||B. For example: if there are two vectors u=[1,2,3] and V=[4,5], their splicing is u||V=[1,2,3,4,5].

[0049] In some optional implementations of the present embodiment, the number of castings with the highest similarity screened can be set to one or more. When set to one, if the similarity is greater than or equal to 90%, the casting with the highest sorted similarity is selected as a similar casting, and the casting with the highest similarity is used as the reuse process of the target casting; if the similarity is greater than or equal to 80% and less than 90%, two castings with the highest similarity are selected as similar castings, and the two castings with the highest similarity are used as similar castings as the reuse process of the target casting; if the similarity is less than 80%, three castings with the highest similarity are selected as similar castings, and the processes of the three similar castings are combined as the reuse process of the target casting; thereby improving the accuracy and reuse efficiency of the reuse process of the target casting.

[0050] It can be seen that the intelligent design method of casting process based on similar process reuse of the present invention, in response to the defects of the prior art, provides a deep learning retrieval method for casting three-dimensional models that integrates process design knowledge (PDK). By adopting deep learning that integrates process design knowledge to perform three-dimensional model retrieval, similar castings can be accurately and effectively retrieved based on the existing casting model database, thereby achieving the purpose of simultaneously ensuring the similarity of process design.

[0051] Specifically, the posture normalization process includes rotating or translating the target casting three-dimensional model so that the longest side of the target casting three-dimensional model is placed in a direction that coincides with or is parallel to the preset three-dimensional coordinate Z axis.

[0052] Exemplarily, a three-dimensional model of a target casting is obtained, and an optional file format is STL. Considering that similar models usually have similar length, width and height, the three-dimensional model of the casting is normalized so that the longest side is placed in a direction parallel to the Z axis.

[0053] This setting makes the coordinate axes between the target casting and the existing castings in the model library consistent, and the acquired data parameters are based on the same reference coordinates, so that the obtained data is more referenceable. Only minor changes and adjustments are needed to achieve process reuse, which can better improve the efficiency and accuracy of the manufacturing process.

[0054] Specifically, shape feature extraction is performed on the three-dimensional model of the casting after posture normalization, including extracting three views and a cross-sectional view of the three-dimensional model of the casting.

[0055] Specifically, the shape features of the casting are obtained by extracting features from the two-dimensional view using a deep learning model. The specific steps are as follows: a two-dimensional view is obtained for the three-dimensional model of the casting after posture standardization, with the preset three-dimensional coordinates and the longest side of the three-dimensional model of the casting parallel to the Z axis or the direction coinciding with the Z axis as a reference. The main view of the three-dimensional model is in the opposite direction of the X axis, the left view is in the positive direction of the Y axis, and the top view is in the opposite direction of the Z axis.

[0056] For solid castings, the three views of the above three-dimensional model can usually clearly represent most of the features of the castings. Since castings include but are not limited to solid parts, in order to make the models of cylindrical parts whose internal structures are difficult to represent through three views comparable, the vertical plane at half the length of the X-axis dimension is selected as the section plane, and then the main view of the section is taken to represent the internal structure of the casting model. The obtained two-dimensional views are arranged in two rows and two columns according to the left view, main view, top view and section view.

[0057] Considering that the two-dimensional view of the casting model mainly provides shape features, and image preprocessing has been proven to enable the convolutional neural network to extract more effective and reliable features, the Sobel operator is used to sharpen the edges of the two-dimensional view to highlight the shape contour of the three-dimensional model to ensure that the convolutional neural network extracts more obvious and distinguishable shape features. Sharpening the edges of the extracted shape features includes converting the two-dimensional view images of the main view, left view, top view and cross-sectional view extracted from the three-dimensional model into corresponding line drawings.

[0058] For example, the edge of the two-dimensional view is sharpened by the Sobel operator, and the results of edge sharpening of some three-dimensional models in the casting database (model library) are as follows: Figure 3 A wheel hub casting is used as the query model. The combined two-dimensional view of the wheel hub casting after the Sobel operator edge sharpening is shown in Figure 4 shown.

[0059] In some optional implementations of this embodiment, a contrastive language-image pretraining model (CLIP) is used to extract features from the two-dimensional view of the three-dimensional model to obtain the shape features of the three-dimensional model. Specifically, based on the acquired two-dimensional view, the contrastive language-image pretraining model CLIP is used to obtain shape semantic features. CLIP has learned general semantic knowledge and image feature representation by training with a contrastive learning method on a large-scale data set, and has zero-sample capability. There is no need for pre-training like other convolutional neural network models. The CLIP model can be directly used to obtain shape features from two-dimensional views.

[0060] Specifically, the process design features of the casting 3D model are extracted including the modulus, wall thickness (main wall thickness or mode of wall thickness), average wall thickness, maximum wall thickness, minimum wall thickness, symmetry, aspect ratio, length-to-height ratio and width-to-height ratio of the casting 3D model. Among the above-mentioned process design features of the 3D model, the modulus directly affects the solidification mode of the casting and possible defects. The calculation formula of the modulus is as follows:

[0061]

[0062] In the above formula, M represents the modulus in centimeters (cm), V represents the volume of the casting in cubic centimeters (cm3), and A represents the heat dissipation surface area of ​​the casting in square centimeters (cm 2 ). When calculating, the total volume and total surface area of ​​the three-dimensional model of the casting are used to calculate the overall modulus of the model to simplify the calculation.

[0063] It should be noted that the above-mentioned wall thickness refers to the main wall thickness. Since the wall thickness of the three-dimensional model has different wall thicknesses in different places, the wall thickness may have many values. Here, only the main wall thickness, that is, the mode of the wall thickness, is taken.

[0064] It should also be noted that the wall thickness has an important influence on the solidification process and final quality of the casting. Areas with thicker walls solidify more slowly, which may lead to greater shrinkage. Therefore, according to the size of the part and the wall thickness, risers or chills are set according to actual needs to compensate for shrinkage and prevent shrinkage cavities and shrinkage. For castings with uneven wall thickness, more complex solidification control strategies are set according to actual conditions, such as adjusting the solidification rate and shrinkage behavior through local cooling or heating.

[0065] Symmetry refers to the degree of symmetry of the casting in terms of its geometry. Symmetry also helps to simplify the prediction of cooling and shrinkage, because each part of the casting will respond to temperature changes in a similar manner. The calculation formula for symmetry in the process design feature of the three-dimensional model of the casting of the present invention is as follows:

[0066] Sym=(Sym X +Sym Y +Sym Z ) / 3;

[0067] Among them, Sym represents the symmetry of the three-dimensional model of the casting, Sym X Indicates the symmetry of the casting in the X-axis direction, which is obtained by calculating the symmetry of the front view and the back view. Y Indicates the symmetry of the casting in the Y-axis direction, which is obtained by calculating the similarity between the left view and the rear view. Z It represents the symmetry of the casting in the Z-axis direction, which is obtained by calculating the similarity between the top view and the bottom view, such as Figure 5 shown.

[0068] At the same time, the aspect ratio and length-to-height ratio of the casting will affect the flow characteristics of the molten metal entering the mold. A larger aspect ratio or length-to-height ratio may cause the molten metal to flow farther, thereby affecting its fluidity and filling capacity. For this type of casting, it is necessary to design a more effective pouring system, such as increasing the number of gates or changing the position of the gates to ensure that the molten metal can evenly and fully fill the cavity. In addition, the aspect ratio of the casting is also one of the factors that affect the complexity of the casting. In the present invention, it should be noted that the aspect ratio is the ratio of the length of the casting in the Y-axis direction to the length in the X-axis direction, the aspect ratio is the ratio of the length of the casting in the Y-axis direction to the length in the Z-axis direction, and the width-to-height ratio is the ratio of the length of the casting in the X-axis direction to the length in the Z-axis direction.

[0069] In some optional implementations of the present embodiment, the main wall thickness, average wall thickness, maximum wall thickness, minimum wall thickness, aspect ratio, length-to-height ratio and width-to-height ratio in the process design can be directly obtained through the parameters of the three-dimensional model of the casting or indirectly calculated, and can also be obtained through actual measurement as needed during use of the three-dimensional model software.

[0070] In summary, the present invention aims at the problem of similar process reuse in the field of casting process design in the prior art, and proposes a deep learning retrieval method for casting three-dimensional models containing process-related features. The features such as modulus, main wall thickness, symmetry, aspect ratio and the like related to process design are considered, and similar processes that can be reused can be found while ensuring similar shapes. Based on the three-view and cross-sectional views of the casting model, the shape features are extracted using the CLIP model, and the three-view and cross-sectional views can provide external and internal shape features. The shape features obtained based on the two-dimensional view are more efficient than traditional shape descriptors, point clouds, voxels, etc., and can handle castings with complex shapes. Then, the shape features are fused with process design features such as modulus, main wall thickness, symmetry, aspect ratio, and the similarity between models is calculated by cosine similarity. The results show that by introducing process design features, the similarity of the processes can be ensured while ensuring the similarity of the shapes, and the process design cycle can be shortened.

[0071] For example, in order to further verify the influence of process design features on the accuracy of 3D casting model retrieval, a further search and comparison of wheel hubs was conducted. The search results are as follows: Figure 6 As shown in Table 1 below, Figure 6The corresponding target hub (i.e., the query hub) and the process parameters of the castings with higher similarity in the search results are obtained, and the similarity between the query results and the process parameters of the target hub itself is calculated, thereby verifying the effectiveness of the intelligent design method for casting processes based on similar process reuse of the present invention. The specific calculation method is: Euclidean distance is used to calculate the similarity between processes, and the Euclidean distance is normalized to between 0 and 1 to represent the similarity. Before calculating the Euclidean distance, standardization is first performed to remove the influence between dimensions. The calculation formulas for standardization, Euclidean distance, and normalization are shown below in order:

[0072]

[0073] In the above formula, x represents the original value of the vector, μ represents the mean of the vector, σ represents the standard deviation of the vector, z represents the standardized value, and x i and i represents the i-th value of vectors x and y, N represents the value normalized to between 0 and 1, representing the similarity, d Euclidean represents the Euclidean distance mentioned above, that is, the distance between two vectors, such as the distance between vector x1 and vector x2, the distance between vector y1 and vector y2, max(d Euclidean ) represents the maximum distance between two vectors, min(d Euclidean ) represents the minimum distance between two vectors.

[0074] Table 1: The casting process intelligent design method based on similar process reuse provided by an embodiment of the present invention queries the process parameter values ​​of the casting models with higher similarity in the target hub model and CLIP and CLIP-PDK model retrieval results, as well as the similarity with the process parameters of the queried target casting.

[0075]

[0076] Combined with the above analysis and Table 1, the process parameters of the CLIP-PDK_1 hub model have the highest similarity with the query model, which is 41.68%, while the process parameters of the CLIP_1 hub without PDK are 11.43% similar to the query model. By introducing the process design feature PDK, the process similarity is further improved while ensuring the shape similarity, which is 30.25% higher than that without PDK, indicating that this method can ensure the similarity of process parameters while maintaining shape similarity and improve reusability.

[0077] like Figure 7As shown in the figure, the casting process simulation results of the target casting and the retrieved castings were carried out using the casting numerical simulation software EasyCast. The material composition of the wheel hub castings is ZL101A, and the casting method is low-pressure casting. The shrinkage cavities generated by the simulation of the process of the model retrieved by CLIP-PDK are less than those without the introduction of process design knowledge. This shows that by introducing process design knowledge, the process similarity is further improved while ensuring the shape similarity.

[0078] In summary, the present invention aims at the defect that the model retrieval method in the prior art is mainly based on the shape similarity of the casting. By introducing the process design features, the effects of the similarity of the process design are fully considered, and the process design features are integrated with the shape features of the casting, so that the overall features of the model are more accurate and complete, especially for the model containing internal features, the structure and internal features of the model can be fully reflected, and the similarity is higher when compared with the existing models in the model library, thereby improving the accuracy of the reuse process and better ensuring the similarity of the process design. At the same time, compared with shape descriptors, point clouds, voxels and other methods, the combination of three-view and sectional views can comprehensively consider the internal and external shape features of the casting model, and the model based on two-dimensional views has better accuracy and efficiency.

[0079] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In addition, "front", "back", "left", "right", "upper" and "lower" in this article are all referenced to the placement state shown in the accompanying drawings.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent casting process design method based on similar process reuse, It is characterized in that include: Performing posture normalization processing on the target casting three-dimensional model to obtain the casting three-dimensional model after posture normalization processing; Obtaining two-dimensional view images of the main view, left view, top view and cross-sectional view of the three-dimensional model of the casting after posture normalization, and performing edge sharpening on the acquired images, and then using a convolutional neural network to obtain shape features of the edge-sharpened images; Extract the process design features of the casting 3D model; The extracted process design features of the casting 3D model are combined with the shape features of the 3D model after edge sharpening; The similarity between the fused features of the target casting and the features of each casting in the model library is calculated according to the cosine similarity calculation formula, and the calculation results are compared and sorted; The castings with the highest similarity are selected as similar castings, and the process route of the target casting is determined based on the process route of the 3D model of the similar castings.

2. According to claim 1, the casting process intelligent design method based on similar process reuse, It is characterized in that The posture normalization process includes rotating or translating the target casting three-dimensional model so that the longest side of the target casting three-dimensional model is placed in a direction that coincides with or is parallel to the preset three-dimensional coordinate Z axis.

3. According to claim 1, the casting process intelligent design method based on similar process reuse, It is characterized in that The cross-sectional view includes a vertical plane in the positive direction of the preset three-dimensional coordinate axis X-axis as a cutting plane, and a main view of the cutting plane is taken as a cross-sectional view, which represents the internal structure of the three-dimensional model of the casting.

4. The casting process intelligent design method based on similar process reuse according to claim 1, It is characterized in that Sharpening the edges of the extracted shape features includes converting the two-dimensional views of the front view, left view, top view and cross-sectional view extracted from the three-dimensional model into corresponding line drawings.

5. The casting process intelligent design method based on similar process reuse according to claim 1, It is characterized in that The process design features of the casting three-dimensional model are extracted, including the modulus, mode of wall thickness, average wall thickness, maximum wall thickness, minimum wall thickness, symmetry, aspect ratio, length-to-height ratio and width-to-height ratio of the casting three-dimensional model.

6. The casting process intelligent design method based on similar process reuse according to claim 1, It is characterized in that The cosine similarity calculation formula includes: Where A and B represent two n-dimensional vectors respectively.

7. The casting process intelligent design method based on similar process reuse according to claim 5, It is characterized in that The symmetry of the casting 3D model is calculated by the following formula: Sym=(Sym X +Sym Y +Sym Z ) / 3 Among them, Sym represents the symmetry of the three-dimensional model of the casting, Sym X Indicates the symmetry of the casting in the X-axis direction, which is obtained by calculating the similarity between the front view and the back view. Y Indicates the symmetry of the casting in the Y-axis direction, which is obtained by calculating the similarity between the left view and the rear view. z Indicates the symmetry of the casting in the Z-axis direction, which is obtained by calculating the similarity between the top view and the bottom view.

8. The casting process intelligent design method based on similar process reuse according to claim 1, It is characterized in that After acquiring the two-dimensional view images of the main view, left view, top view and sectional view of the casting three-dimensional model after posture normalization, it also includes arranging them in two rows and two columns in sequence according to the left view, main view, top view and sectional view.

Images