Method and system for producing a forged component of a tc6 material structure

Abstract

A method and system for preparing forged structural parts from TC6 material are disclosed. First, TC6 bar stock is sawn, and the β-phase transformation temperature (Tβ) of the sawn TC6 bar billet is measured. Then, the billet is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held to obtain a pre-treated TC6 bar billet. Next, the pre-treated TC6 bar billet undergoes a series of processes in sequence: small-end aggregation, flat forging, bending, flat forging, large-end aggregation, flat forging, large-end forming, flat forging, and intermediate billet treatment to obtain a secondary-treated TC6 bar billet. This secondary-treated TC6 bar billet is then heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held to obtain a billet to be forged. Finally, it is forged using an electric screw press, followed by heat treatment, isothermal annealing, and physicochemical testing to obtain the machined TC6 material structural part. Finally, the parts are inspected, and qualified parts are stored in the warehouse. This process yields qualified structural parts.

Description

Technical Field

[0001] This application relates to the field of intelligent manufacturing, and more specifically, to a method and system for preparing forgings of TC6 material structural parts. Background Technology

[0002] The traditional method for preparing TC6 complex structural parts with right-angle transitions and bosses at both ends is: pre-forging of large-size bar stock + removal of excess material by a large amount of machining + ordinary die forging.

[0003] This process method has the following disadvantages:

[0004] 1. Large forging equipment is required, with a minimum capacity of 2500t electric screw press.

[0005] 2. High material consumption and high forging costs;

[0006] 3. After forging, a large amount of excess material needs to be removed by machining, which results in high processing time and cost;

[0007] 4. When using this forging process, the forging is prone to localized temperature rise during the forging process, which leads to unqualified high-magnification microstructure and cannot guarantee the quality of the forging.

[0008] Therefore, an optimized forging preparation scheme for TC6 material structural parts is desired. Summary of the Invention

[0009] In view of this, this disclosure proposes a forging preparation method and system for TC6 material structural parts, which can improve the manufacturing quality and production efficiency of TC6 material structural parts.

[0010] According to one aspect of this disclosure, a method for forging a structural component made of TC6 material is provided, comprising:

[0011] The TC6 bar stock of a predetermined specification is sawn to a predetermined length to obtain the sawn TC6 bar billet, and the β phase transformation point (Tβ) of the sawn TC6 bar billet is measured.

[0012] The sawn TC6 billet is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held at that temperature to obtain a pretreated TC6 billet. The holding time is: the effective thickness of the sawn TC6 billet × 5~6min / 10mm.

[0013] The pretreated TC6 billet is subjected to the following processes in chronological order: small end aggregation, flat forging, bending, flat forging, large end aggregation, flat forging, large end forming, flat forging, and intermediate billet treatment to obtain a secondary-treated TC6 billet. The intermediate billet treatment includes grinding, sandblasting, and applying glass lubricant.

[0014] The TC6 billet after the secondary treatment is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held at that temperature to obtain a billet to be forged in the final stage. The holding time is: the effective thickness of the billet to be forged in the final stage × 5~6min / 10mm.

[0015] TC6 material structural parts are forged using a 630t-1000t electric screw press, followed by heat treatment, isothermal annealing, and physicochemical testing to obtain the finished product.

[0016] Inspect the processed TC6 material structural parts and put them into storage if they pass the inspection.

[0017] In one possible implementation, the processed TC6 material structural parts are inspected and, if qualified, are put into storage, including:

[0018] Obtain an image of the outer surface of a TC6 material structural component that has been machined and formed;

[0019] Extract the orientation gradient histogram of the outer surface image;

[0020] The directional gradient histogram and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image;

[0021] The multi-channel outer surface image is passed through a first convolutional neural network model, which serves as a feature extractor, to obtain a first depth outer surface feature map.

[0022] Spatial pyramid pooling is performed on the first depth outer surface feature map to obtain a pooled outer surface feature map;

[0023] The pooled outer surface feature map is processed by a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map; and

[0024] The classification feature map is passed through a classifier to obtain a classification result, which is used to indicate whether the surface quality of the TC6 material structural parts meets a predetermined standard.

[0025] In one possible implementation, the oriented gradient histogram and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image, including:

[0026] The directional gradient histogram and the outer surface image are aggregated along the channel dimension using the following cascade formula to obtain the multi-channel outer surface image;

[0027] The cascade formula is as follows:

[0028] F c =Concat[F1, F2]

[0029] Where F1 represents the directional gradient histogram, F2 represents the outer surface image, and Concat[·] represents the cascade function. c This represents the multi-channel outer surface image.

[0030] In one possible implementation, the multi-channel outer surface image is processed by a first convolutional neural network model, which acts as a feature extractor, to obtain a first depth outer surface feature map, including:

[0031] Each layer of the first convolutional neural network model, which acts as a feature extractor, processes the input data during the forward propagation of the layer as follows:

[0032] The input data is processed by convolution to obtain a convolutional feature map;

[0033] The convolutional feature map is subjected to mean pooling to obtain a pooled feature map; and

[0034] The pooled feature map is nonlinearly activated to obtain an activated feature map;

[0035] Wherein, the output of the last layer of the first convolutional neural network model, which serves as the feature extractor, is the first depth outer surface feature map, and the input of the first layer of the first convolutional neural network model, which serves as the feature extractor, is the multi-channel outer surface image.

[0036] In one possible implementation, the spatial pyramid pooling employs four different scales of max pooling operations: 13x13, 9x9, 5x5, and 1x1.

[0037] In one possible implementation, the second convolutional neural network model serving as a feature extractor includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and an output layer.

[0038] In one possible implementation, the first convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the second convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the third convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the first pooling layer uses 2×2 max pooling with a stride of 2; the second pooling layer uses 2×2 max pooling with a stride of 2; and the third pooling layer uses 2×2 max pooling with a stride of 2.

[0039] In one possible implementation, the pooled outer surface feature map is passed through a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map, including:

[0040] The pooled outer surface feature map is processed by a second convolutional neural network model, which acts as a feature extractor, to obtain a second depth outer surface feature map; and

[0041] The first depth outer surface feature map and the second depth outer surface feature map are fused to obtain the classification feature map.

[0042] In one possible implementation, fusing the first depth outer surface feature map and the second depth outer surface feature map to obtain the classification feature map includes:

[0043] The first depth outer surface feature map and the second depth outer surface feature map are fused using the following fusion formula to obtain the classification feature map;

[0044] The fusion formula is as follows:

[0045]

[0046] Among them, M 1i and M 2i These are the i-th feature matrices of the first depth outer surface feature map and the second depth outer surface feature map, respectively, and M ci It is the i-th feature matrix of the classification feature map, (·) T It is the transpose of a matrix. and These represent matrix multiplication and addition, respectively.

[0047] According to another aspect of this disclosure, a forging preparation system for TC6 material structural parts is provided, comprising:

[0048] The sawing module is used to saw TC6 bars of a predetermined specification to a predetermined length to obtain sawn TC6 bar blanks, and to measure the β phase transformation point (Tβ) of the sawn TC6 bar blanks.

[0049] A primary heating module is used to heat the sawn TC6 billet to a high temperature (Tβ-40) ±10℃ in an electric furnace and hold it at that temperature to obtain a pretreated TC6 billet. The holding time is: the effective thickness of the sawn TC6 billet × 5~6 min / 10 mm.

[0050] The secondary processing module is used to process the pre-treated TC6 billet in a time sequence, including small-end aggregation, flat forging, bending, flat forging, large-end aggregation, flat forging, large-end forming, flat forging and intermediate billet processing, to obtain the secondary-processed TC6 billet. The intermediate billet processing includes grinding, sandblasting and applying glass lubricant.

[0051] The secondary heating module is used to heat the secondary-treated TC6 billet to a high temperature (Tβ-40) ±10℃ in an electric furnace and hold it at that temperature to obtain the billet to be forged in the final stage. The holding time is: the effective thickness of the billet to be forged in the final stage × 5~6min / 10mm.

[0052] The forging module is used for forging with a 630t-1000t electric screw press, followed by heat treatment, isothermal annealing, and physicochemical testing to obtain machined TC6 material structural parts; and

[0053] The conformity inspection module is used to inspect the processed TC6 material structural parts, and those that pass the inspection are put into storage.

[0054] According to an embodiment of this disclosure, TC6 bar stock is first sawn, and the β phase transformation point (Tβ) of the sawn TC6 bar billet is measured. Then, it is heated to a high temperature (Tβ-40) ±10°C in an electric furnace and held to obtain a pre-treated TC6 bar billet. Next, the pre-treated TC6 bar billet undergoes a series of processes in chronological order: small-end aggregation, flat forging, bending, flat forging, large-end aggregation, flat forging, large-end forming, flat forging, and intermediate billet treatment to obtain a secondary-treated TC6 bar billet. Then, the secondary-treated TC6 bar billet is heated to a high temperature (Tβ-40) ±10°C in an electric furnace and held to obtain a billet to be forged. This billet is then forged using an electric screw press, followed by heat treatment, isothermal annealing, and physicochemical testing to obtain a machined TC6 material structural component. Finally, the component is inspected, and qualified components are stored in the warehouse. This process yields qualified structural components.

[0055] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0056] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.

[0057] Figure 1 A flowchart illustrating a method for forging a TC6 material structural component according to an embodiment of the present disclosure is shown.

[0058] Figure 2This diagram illustrates an application scenario of sub-step S160 of the forging method for a TC6 material structural component according to an embodiment of the present disclosure.

[0059] Figure 3 A flowchart illustrating sub-step S160 of a method for preparing a forging of a TC6 material structural component according to an embodiment of the present disclosure is shown.

[0060] Figure 4 A schematic diagram of sub-step S160 of a forging method for a TC6 material structural component according to an embodiment of the present disclosure is shown.

[0061] Figure 5 A flowchart illustrating sub-step S166 of a method for preparing a forging of a TC6 material structural component according to an embodiment of the present disclosure is shown.

[0062] Figure 6 A block diagram of a forging preparation system for TC6 material structural parts according to an embodiment of the present disclosure is shown.

[0063] Figure 7 A schematic diagram of small-head aggregation according to an embodiment of the present disclosure is shown.

[0064] Figure 8 A schematic diagram of bending according to an embodiment of the present disclosure is shown.

[0065] Figure 9 A schematic diagram of a large-head aggregation according to an embodiment of the present disclosure is shown.

[0066] Figure 10 A schematic diagram of a large head forming according to an embodiment of the present disclosure is shown.

[0067] Figure 11 A 2D schematic diagram of the final formed forging according to an embodiment of the present disclosure is shown.

[0068] Figure 12 A 3D schematic diagram of the final formed forging according to an embodiment of the present disclosure is shown. Detailed Implementation

[0069] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are also within the scope of protection of this application.

[0070] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0071] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0072] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0073] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0074] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0075] This application provides a method for forging structural components made of TC6 material, such as... Figure 1As shown, the specific steps include: S110, sawing a predetermined TC6 bar stock to a predetermined length to obtain a sawn TC6 bar billet, and measuring the β phase transformation point (Tβ) of the sawn TC6 bar billet; S120, heating the sawn TC6 bar billet to a high temperature (Tβ-40) ±10℃ in an electric furnace and holding it at that temperature to obtain a pretreated TC6 bar billet, wherein the holding time is: the effective thickness of the sawn TC6 bar billet × 5~6 min / 10mm; S130, performing the following steps in chronological order on the pretreated TC6 bar billet: small end aggregation, flat forging, bending, flat forging, large end aggregation, flat forging, and large end forming. The process involves forging with a flat forging machine and processing an intermediate billet to obtain a secondary-processed TC6 bar billet. The intermediate billet processing includes grinding, sandblasting, and applying glass lubricant. S140: The secondary-processed TC6 bar billet is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held at that temperature to obtain a final-forged bar billet. The holding time is: the effective thickness of the final-forged bar billet × 5~6 min / 10 mm. S150: After forging with a 630t~1000t electric screw press, the billet undergoes heat treatment, isothermal annealing, and physical and chemical testing to obtain a machined TC6 material structural component. S160: The machined TC6 material structural component is inspected and, if qualified, placed into the warehouse.

[0076] Traditional surface quality inspection methods mainly rely on manual visual inspection, which has drawbacks such as strong subjectivity and low accuracy. In order to improve the efficiency and accuracy of surface quality inspection, the technical concept of this application is to use the external surface image of the processed TC6 material structural parts, and combine deep learning and artificial intelligence technologies to realize the automated, intelligent and high-precision inspection of the surface quality of the processed TC6 material structural parts, thereby improving the manufacturing quality and production efficiency of TC6 material structural parts.

[0077] Specifically, in the technical solution of this application, an image of the outer surface of a TC6 material structural component is first acquired, and the histogram of oriented gradients (HOG) of the outer surface image is extracted. The outer surface image reflects important information about the surface quality of the structural component; defects, breaks, deformations, and other surface defects will also be reflected in the outer surface image. Furthermore, the histogram of oriented gradients (HOG) can describe the edge or texture information of local areas in the image. HOG features have the following advantages: they are robust to changes in lighting, scale, and rotation; they can effectively capture structural and shape information in the image; and the computational cost is relatively low. Therefore, extracting the HOG features from the outer surface image can be used to characterize the surface quality of the TC6 material structural component, such as whether there are scratches, dents, oxidation, or other defects on the surface.

[0078] Next, the histogram of oriented gradients and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image. As mentioned earlier, the histogram of oriented gradients can reflect structural information such as edges and corners in an image, but it also lacks other important information from the source domain. In the technical solution of this application, aggregating the histogram of oriented gradients and the original outer surface image can enhance the expressive power of features. That is, the aggregated multi-channel outer surface image can retain the color, brightness, and other information of the original outer surface image while also preserving structural and texture information, thus avoiding information loss.

[0079] Then, the multi-channel outer surface image is passed through a first convolutional neural network model, which acts as a feature extractor, to obtain a first depth outer surface feature map. The first convolutional neural network model utilizes convolutional kernels to perform a sliding operation on the local receptive field of the multi-channel outer surface image. Notably, the first convolutional neural network model has fewer convolutional layers, enabling it to extract shallow features from the image, such as edges, corners, and textures.

[0080] Furthermore, spatial pyramid pooling is applied to the first depth outer surface feature map to obtain a pooled outer surface feature map. The spatial pyramid pooling structure employs four different scales of max pooling operations: 13x13, 9x9, 5x5, and 1x1 (no processing). This significantly increases the receptive field, thereby improving the network's ability to detect small targets. In practice, diverse training image sizes are far more likely to converge during training than a single size. Spatial pyramid pooling can improve image scale invariance and reduce overfitting.

[0081] Subsequently, the pooled outer surface feature map is passed through a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map. Here, the second convolutional neural network model performs a higher-level abstraction and representation of the pooled outer surface feature map, enabling the extraction of deeper feature information from the pooled outer surface feature map, i.e., more discriminative features.

[0082] In a specific example of this application, the second convolutional neural network model serving as a feature extractor includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and an output layer. The input layer receives the pooled outer surface feature map and has a size of 32×32×64. The first convolutional layer uses a 3×3 kernel with a stride of 1, same padding, 128 output channels, ReLU activation function, and an output size of 32×32×128. The first pooling layer uses 2×2 max pooling with a stride of 2 and an output size of 16×16×128. The second convolutional layer uses a 3×3 kernel with a stride of 1, same padding, and an output layer. The first convolutional layer has 256 output channels, uses ReLU activation, and has an output size of 16×16×256. The second pooling layer uses 2×2 max pooling with a stride of 2 and an output size of 8×8×256. The third convolutional layer uses a 3×3 kernel with a stride of 1 and same padding, has 512 output channels, uses ReLU activation, and has an output size of 8×8×512. The third pooling layer uses 2×2 max pooling with a stride of 2 and an output size of 4×4×512.

[0083] After obtaining the classification feature map, it is passed through a classifier to obtain a classification result. The classification result indicates whether the surface quality of the processed TC6 material structural parts meets predetermined standards. The classifier can automatically learn rules for judging different surface qualities based on the input feature map and output the corresponding classification result. In other words, by analyzing the classification feature map through the classifier, it is possible to effectively identify whether there are defects in the surface quality of the processed TC6 material structural parts. Compared with traditional manual visual inspection methods, this method has advantages such as strong objectivity, high accuracy, and high efficiency, and can improve the surface quality inspection level and quality control capability of processed TC6 material structural parts.

[0084] In the technical solution of this application, when the pooled outer surface feature map is processed by a second convolutional neural network model as a feature extractor to obtain a classification feature map, considering that the pooled outer surface feature map is obtained by spatial pyramid pooling of the first depth outer surface feature map, in order to fuse the image semantic features of the multi-channel outer surface images at different depths, it is preferable to obtain the classification feature map by fusing the second depth outer surface feature map obtained from the pooled outer surface feature map by the second convolutional neural network model as a feature extractor with the first depth outer surface feature map.

[0085] Furthermore, considering that both the first and second depth outer surface feature maps extract local image semantic association features based on the spatial scale of the convolutional kernel during the feature extraction process of the first and second convolutional neural network models, their feature distribution granularity is based on the convolutional kernel association scale of the convolutional neural network model. Therefore, given that the first and second depth outer surface feature maps each possess spatial image semantic association features at their respective scales, when fusing the first and second depth outer surface feature maps to obtain the classification feature map, it is necessary to consider the scale representation of the spatial image semantic association features of the first and second depth outer surface feature maps for fusion.

[0086] Based on this, the applicant of this application performs global context space correlation enrichment fusion on the first depth outer surface feature map and the second depth outer surface feature map, specifically as follows:

[0087]

[0088] M 1i and M 2i These are the i-th feature matrices of the first depth outer surface feature map and the second depth outer surface feature map, respectively, and M ci It is the i-th feature matrix of the classification feature map.

[0089] Here, in order to aggregate the contextual spatial association semantics between the local spatial semantics associated with the first and second depth outer surface feature maps, the global contextual spatial association enrichment fusion enriches the frame-level spatial semantic fusion expression of the feature matrix under the global receptive field by focusing on the explicit contextual relevance at the spatial frame-level represented by the feature matrix of the feature map. This achieves the assimilation fusion of cross-channel spatial shared contextual semantics between the first and second depth outer surface feature maps, thereby improving the fusion effect of the classification feature map on the first and second depth outer surface feature maps.

[0090] Figure 2 This diagram illustrates an application scenario of sub-step S160 of a forging method for a TC6 material structural component according to an embodiment of the present disclosure. (See diagram for example.) Figure 2 As shown, in this application scenario, firstly, a pre-formed TC6 material structural component is obtained (e.g., Figure 2 The outer surface image of N (as shown) (e.g., Figure 2As shown in D), the outer surface image is then input to a server that deploys a forging preparation algorithm for TC6 material structural parts (e.g., Figure 2 As shown in S), the server is able to process the outer surface image using the forging preparation algorithm of the TC6 material structural part to obtain a classification result indicating whether the surface quality of the machined TC6 material structural part meets a predetermined standard.

[0091] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0092] Figure 3 A flowchart illustrating sub-step S160 of a method for preparing a forging of a TC6 material structural component according to an embodiment of the present disclosure is shown. Figure 3 As shown, the forging preparation method of TC6 material structural parts according to the embodiments of this application, including the following steps, includes: S161, acquiring an image of the outer surface of the forged TC6 material structural parts; S162, extracting the histogram of directional gradients from the outer surface image; S163, aggregating the histogram of directional gradients and the outer surface image along the channel dimension to obtain a multi-channel outer surface image; S164, passing the multi-channel outer surface image through a first convolutional neural network model as a feature extractor to obtain a first depth outer surface feature map; S165, performing spatial pyramid pooling on the first depth outer surface feature map to obtain a pooled outer surface feature map; S166, passing the pooled outer surface feature map through a second convolutional neural network model as a feature extractor to obtain a classification feature map; and S167, passing the classification feature map through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the surface quality of the forged TC6 material structural parts meets a predetermined standard.

[0093] Figure 4 A schematic diagram of sub-step S160 of a forging method for a TC6 material structural component according to an embodiment of the present disclosure is shown. Figure 4As shown, in this network architecture, firstly, an image of the outer surface of the processed TC6 material structural component is acquired; then, the histogram of directional gradients of the outer surface image is extracted; next, the histogram of directional gradients and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image; then, the multi-channel outer surface image is passed through a first convolutional neural network model as a feature extractor to obtain a first depth outer surface feature map; then, the first depth outer surface feature map is subjected to spatial pyramid pooling to obtain a pooled outer surface feature map; next, the pooled outer surface feature map is passed through a second convolutional neural network model as a feature extractor to obtain a classification feature map; finally, the classification feature map is passed through a classifier to obtain a classification result, which is used to indicate whether the surface quality of the processed TC6 material structural component meets a predetermined standard.

[0094] More specifically, in step S161, an image of the outer surface of the TC6 material structural component is acquired. The outer surface image can reflect important information about the surface quality of the structural component; any surface defects such as defects, breakage, or deformation will also be shown in the outer surface image.

[0095] More specifically, in step S162, the histogram of oriented gradients (HOG) of the outer surface image is extracted. The histogram of oriented gradients (HOG) can describe the edge or texture information of local regions in an image. HOG features have the following advantages: they are robust to changes in illumination, scale, and rotation; they can effectively capture structural and shape information in an image; and they have relatively low computational cost.

[0096] More specifically, in step S163, the histogram of oriented gradients and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image. The histogram of oriented gradients can reflect structural information such as edges and corners in the image, but it also lacks other important information from the source domain. The aggregated multi-channel outer surface image can retain the color, brightness, and other information of the original outer surface image while also containing structural and texture information, thus avoiding information loss.

[0097] Accordingly, in one possible implementation, aggregating the histogram of oriented gradients and the outer surface image along the channel dimension to obtain a multi-channel outer surface image includes: aggregating the histogram of oriented gradients and the outer surface image along the channel dimension using the following cascaded formula to obtain the multi-channel outer surface image; wherein the cascaded formula is:

[0098] F c =Concat[F1, F2]

[0099] Where F1 represents the directional gradient histogram, F2 represents the outer surface image, and Concat[·] represents the cascade function. c This represents the multi-channel outer surface image.

[0100] More specifically, in step S164, the multi-channel outer surface image is processed by a first convolutional neural network model, which acts as a feature extractor, to obtain a first depth outer surface feature map. The first convolutional neural network model can utilize convolutional kernels to perform a sliding operation on the local receptive field of the multi-channel outer surface image. It is worth noting that the first convolutional neural network model has fewer convolutional layers and can extract shallow features from the image, such as edges, corners, and textures.

[0101] As you can understand, a Convolutional Neural Network (CNN) is an artificial neural network with wide applications in fields such as image recognition. A CNN can consist of an input layer, hidden layers, and an output layer. Hidden layers can include convolutional layers, pooling layers, activation layers, and fully connected layers. Each layer performs calculations on the input data and outputs the results to the next layer. The initial input data undergoes multiple layers of computation to arrive at a final result.

[0102] Accordingly, in one possible implementation, the multi-channel outer surface image is processed by a first convolutional neural network model serving as a feature extractor to obtain a first depth outer surface feature map, comprising: each layer of the first convolutional neural network model serving as a feature extractor performing the following operations on the input data during the forward propagation of the layer: performing convolution processing on the input data to obtain a convolutional feature map; performing mean pooling processing on the convolutional feature map to obtain a pooled feature map; and performing nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network model serving as a feature extractor is the first depth outer surface feature map, and the input of the first layer of the first convolutional neural network model serving as a feature extractor is the multi-channel outer surface image.

[0103] More specifically, in step S165, spatial pyramid pooling is performed on the first depth outer surface feature map to obtain a pooled outer surface feature map. The spatial pyramid pooling structure employs four different scales of max pooling operations, which can greatly increase the receptive field and thus improve the network's ability to detect small targets. In practice, training images with diverse sizes are much easier to converge during training than those with a single size. Spatial pyramid pooling can improve image scale invariance and reduce overfitting problems.

[0104] Accordingly, in one possible implementation, the spatial pyramid pooling employs four different scales of max pooling operations, namely 13x13, 9x9, 5x5, and 1x1.

[0105] More specifically, in step S166, the pooled outer surface feature map is processed by a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map. The second convolutional neural network model performs a higher level of abstraction and representation of the pooled outer surface feature map, enabling the extraction of deeper feature information from the pooled outer surface feature map, i.e., more discriminative features.

[0106] Accordingly, in one possible implementation, the second convolutional neural network model serving as the feature extractor includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and an output layer. Specifically, the first convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation; the second convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation; the third convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation; the first pooling layer uses 2×2 max pooling with a stride of 2; the second pooling layer uses 2×2 max pooling with a stride of 2; and the third pooling layer uses 2×2 max pooling with a stride of 2.

[0107] Accordingly, in one possible implementation, such as Figure 5 As shown, the method of obtaining a classification feature map by passing the pooled outer surface feature map through a second convolutional neural network model as a feature extractor includes: S1661, passing the pooled outer surface feature map through a second convolutional neural network model as a feature extractor to obtain a second depth outer surface feature map; and S1662, fusing the first depth outer surface feature map and the second depth outer surface feature map to obtain the classification feature map.

[0108] In the technical solution of this application, when the pooled outer surface feature map is processed by a second convolutional neural network model as a feature extractor to obtain a classification feature map, considering that the pooled outer surface feature map is obtained by spatial pyramid pooling of the first depth outer surface feature map, in order to fuse the image semantic features of the multi-channel outer surface images at different depths, it is preferable to fuse the second depth outer surface feature map obtained by the second convolutional neural network model as a feature extractor from the pooled outer surface feature map with the first depth outer surface feature map to obtain the classification feature map. Furthermore, considering that both the first depth outer surface feature map and the second depth outer surface feature map are used for local image feature semantic association feature extraction based on the spatial scale of the convolution kernel during the feature extraction process of the first and second convolutional neural network models, they both use the convolution kernel association scale of the convolutional neural network model as the feature distribution granularity. Therefore, given that the first and second depth outer surface feature maps each possess spatial image semantic association features at their respective scales, when fusing the first and second depth outer surface feature maps to obtain the classification feature map, it is necessary to consider the scale representation of the spatial image semantic association features of the first and second depth outer surface feature maps for fusion. Based on this, the applicant of this application performs global contextual spatial association enrichment fusion on the first and second depth outer surface feature maps.

[0109] Accordingly, in one possible implementation, fusing the first depth outer surface feature map and the second depth outer surface feature map to obtain the classification feature map includes: fusing the first depth outer surface feature map and the second depth outer surface feature map using the following fusion formula to obtain the classification feature map; wherein, the fusion formula is:

[0110]

[0111] Among them, M 1i and M 2i These are the i-th feature matrices of the first depth outer surface feature map and the second depth outer surface feature map, respectively, and M ci It is the i-th feature matrix of the classification feature map, (·) T It is the transpose of a matrix. and These represent matrix multiplication and addition, respectively.

[0112] Here, in order to aggregate the contextual spatial association semantics between the local spatial semantics associated with the first and second depth outer surface feature maps, the global contextual spatial association enrichment fusion enriches the frame-level spatial semantic fusion expression of the feature matrix under the global receptive field by focusing on the explicit contextual relevance at the spatial frame level represented by the feature matrix of the feature map. This achieves the assimilation and fusion of cross-channel spatial shared contextual semantics between the first and second depth outer surface feature maps, thereby improving the fusion effect of the classification feature map on the first and second depth outer surface feature maps.

[0113] More specifically, in step S167, the classification feature map is processed by a classifier to obtain a classification result. The classification result indicates whether the surface quality of the processed TC6 material structural part meets a predetermined standard. The classifier can automatically learn rules for judging different surface qualities based on the input feature map and output the corresponding classification result. That is, by analyzing the classification feature map through the classifier, it is possible to effectively identify whether there are defects in the surface quality of the processed TC6 material structural part.

[0114] In other words, in the technical solution of this application, the classifier's labels include "the surface quality of the TC6 material structural parts meets a predetermined standard" (first label) and "the surface quality of the TC6 material structural parts does not meet the predetermined standard" (second label). The classifier determines which label the classification feature map belongs to using a soft-maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially defined concepts. In fact, during the training process, the computer model does not have the concept of "whether the surface quality of the TC6 material structural parts meets the predetermined standard." It simply has two classification labels and outputs the probability of the feature under these two labels, i.e., the sum of p1 and p2 is one. Therefore, the classification result of whether the surface quality of the TC6 material structural parts meets the predetermined standard is actually transformed into a binary probability distribution that conforms to natural laws through the classification labels. Essentially, it uses the physical meaning of the natural probability distribution of the labels, rather than the linguistic meaning of "whether the surface quality of the TC6 material structural parts meets the predetermined standard."

[0115] As you can understand, the role of a classifier is to learn classification rules and classifiers using given categories and known training data, and then classify (or predict) unknown data. Logistic regression and SVM are commonly used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are needed to form the multi-class classification. However, this is prone to errors and inefficient. A commonly used multi-class classification method is the Softmax classification function.

[0116] In summary, the forging method for TC6 material structural parts based on the embodiments of this application first acquires an image of the outer surface of the TC6 material structural part. Next, it extracts the histogram of directional gradients from the outer surface image. Then, it aggregates the histogram of directional gradients and the outer surface image along the channel dimension to obtain a multi-channel outer surface image. Next, it passes the multi-channel outer surface image through a first convolutional neural network model as a feature extractor to obtain a first depth outer surface feature map. Then, it performs spatial pyramid pooling on the first depth outer surface feature map to obtain a pooled outer surface feature map. Next, it passes the pooled outer surface feature map through a second convolutional neural network model as a feature extractor to obtain a classification feature map. Finally, it passes the classification feature map through a classifier to obtain a classification result indicating whether the surface quality of the TC6 material structural part meets a predetermined standard.

[0117] Figure 6 A block diagram of a forging preparation system 100 for TC6 material structural parts according to an embodiment of the present disclosure is shown. Figure 6As shown, the forging preparation system 100 for TC6 material structural parts according to an embodiment of this application includes: a sawing module 110, used to saw TC6 bars of predetermined specifications to a predetermined length to obtain sawn TC6 bar blanks, and to measure the β phase transformation point (Tβ) of the sawn TC6 bar blanks; a primary heating module 120, used to heat the sawn TC6 bar blanks to a high temperature (Tβ-40) ±10℃ in an electric furnace and hold the temperature to obtain pretreated TC6 bar blanks, wherein the holding time is: the effective thickness of the sawn TC6 bar blank × 5~6 min / 10mm; and a secondary processing module 130, used to process the pretreated TC6 bar blanks in a time sequence as follows: small end aggregation, flat forging, bending, flat forging, large end aggregation, ... The process includes flat forging, large-end forming, flat forging, and intermediate billet treatment to obtain a secondary-treated TC6 billet. The intermediate billet treatment includes grinding, sandblasting, and applying glass lubricant. A secondary heating module 140 is used to heat the secondary-treated TC6 billet to a high temperature (Tβ-40) ±10℃ in an electric furnace and hold it at that temperature to obtain a billet to be forged in the final stage. The holding time is: the effective thickness of the billet to be forged × 5~6min / 10mm. A forging module 150 is used to perform forging using a 630t~1000t electric screw press, followed by heat treatment, isothermal annealing, and physical and chemical testing to obtain a machined TC6 material structural part. Finally, a qualification inspection module 160 is used to inspect the machined TC6 material structural parts and put them into storage if they pass the inspection.

[0118] In one possible implementation, in the forging preparation system 100 for TC6 material structural parts described above, the qualification inspection module 160 includes: an image acquisition unit for acquiring an image of the outer surface of the TC6 material structural part; an orientation gradient histogram extraction unit for extracting the orientation gradient histogram of the outer surface image; an image aggregation unit for aggregating the orientation gradient histogram and the outer surface image along the channel dimension to obtain a multi-channel outer surface image; a first convolutional coding unit for passing the multi-channel outer surface image through a first convolutional neural network model as a feature extractor to obtain a first depth outer surface feature map; a spatial pyramid pooling unit for performing spatial pyramid pooling on the first depth outer surface feature map to obtain a pooled outer surface feature map; a second convolutional coding unit for passing the pooled outer surface feature map through a second convolutional neural network model as a feature extractor to obtain a classification feature map; and a classification unit for passing the classification feature map through a classifier to obtain a classification result, the classification result being used to indicate whether the surface quality of the TC6 material structural part meets a predetermined standard.

[0119] In one possible implementation, in the forging preparation system 100 for the TC6 material structural component described above, the image aggregation unit is used to: aggregate the directional gradient histogram and the outer surface image along the channel dimension using the following cascade formula to obtain the multi-channel outer surface image; wherein, the cascade formula is:

[0120] F c =Concat[F1, F2]

[0121] Where F1 represents the directional gradient histogram, F2 represents the outer surface image, and Concat[·] represents the cascade function. c This represents the multi-channel outer surface image.

[0122] In one possible implementation, in the forging preparation system 100 for the TC6 material structural component described above, the first convolutional coding unit is configured to: perform the following on the input data during the forward propagation of the layer of the first convolutional neural network model, which serves as a feature extractor: convolve the input data to obtain a convolutional feature map; perform mean pooling on the convolutional feature map to obtain a pooled feature map; and perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network model serving as a feature extractor is the first depth outer surface feature map, and the input of the first layer of the first convolutional neural network model serving as a feature extractor is the multi-channel outer surface image.

[0123] In one possible implementation, in the forging preparation system 100 for the above-mentioned TC6 material structural parts, the spatial pyramid pooling employs four different scales of max pooling operations, namely 13x13, 9x9, 5x5, and 1x1.

[0124] In one possible implementation, in the forging preparation system 100 for the above-mentioned TC6 material structural parts, the second convolutional neural network model serving as the feature extractor includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and an output layer.

[0125] In one possible implementation, in the forging preparation system 100 for the above-mentioned TC6 material structural parts, the first convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the second convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the third convolutional layer uses a 3×3 convolutional kernel with a stride of 1, same padding, and ReLU activation function; the first pooling layer uses 2×2 max pooling with a stride of 2; the second pooling layer uses 2×2 max pooling with a stride of 2; and the third pooling layer uses 2×2 max pooling with a stride of 2.

[0126] In one possible implementation, in the forging preparation system 100 for the above-mentioned TC6 material structural parts, the second convolutional coding unit is used to: pass the pooled outer surface feature map through a second convolutional neural network model as a feature extractor to obtain a second depth outer surface feature map; and fuse the first depth outer surface feature map and the second depth outer surface feature map to obtain the classification feature map.

[0127] In one possible implementation, in the forging preparation system 100 for the TC6 material structural component described above, fusing the first depth outer surface feature map and the second depth outer surface feature map to obtain the classification feature map includes: fusing the first depth outer surface feature map and the second depth outer surface feature map using the following fusion formula to obtain the classification feature map; wherein, the fusion formula is:

[0128]

[0129] Among them, M 1i and M 2i These are the i-th feature matrices of the first depth outer surface feature map and the second depth outer surface feature map, respectively, and M ci It is the i-th feature matrix of the classification feature map, (·) T It is the transpose of a matrix. and These represent matrix multiplication and addition, respectively.

[0130] Here, those skilled in the art will understand that the specific functions and operations of each unit and module in the forging preparation system 100 for TC6 material structural parts described above have been referenced above. Figures 1 to 5 The forging preparation method of TC6 material structural parts has been described in detail, and therefore, its repeated description will be omitted.

[0131] As described above, the TC6 material structural component forging preparation system 100 according to the embodiments of this application can be implemented in various wireless terminals, such as servers with TC6 material structural component forging preparation algorithms. In one possible implementation, the TC6 material structural component forging preparation system 100 according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the TC6 material structural component forging preparation system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the TC6 material structural component forging preparation system 100 can also be one of many hardware modules of the wireless terminal.

[0132] Alternatively, in another example, the forging preparation system 100 for the TC6 material structural component and the wireless terminal can also be separate devices, and the forging preparation system 100 for the TC6 material structural component can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0133] Furthermore, in the forging preparation method and system for TC6 material structural parts, the small end aggregates and is forged using a flat forging machine, according to... Figure 7 Perform; bend, forge using a flat forging machine, according to Figure 8 The large-head forging is carried out using a flat forging machine, according to... Figure 9 The large end is formed using a flat forging machine, according to... Figure 10 The forming process is then carried out. The final 2D drawing of the forging is shown below. Figure 11 As shown, the 3D image is as follows Figure 12 As shown.

[0134] This application uses small-diameter bars for upsetting and then small-tonnage equipment for final forging, solving the problem that large forging equipment must be used when using large-diameter bars for pre-forging + final forging (conventional forging); it solves the problem of high forging cost caused by high material consumption in conventional forging (saving 30%-50% of material per forging); it solves the problem of high processing costs and long processing cycles for removing excess material from forgings after conventional forging; and it solves the problem of scrapping forgings due to uncontrollable local deformation and forging strain dead zones caused by unqualified forging structure during conventional forging.

[0135] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory including computer program instructions that can be executed by a processing component of the device to perform the above-described method.

[0136] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0137] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0138] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0139] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0140] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0141] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0142] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0143] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0144] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method of producing a forging of a structural component of a TC6 material, characterized in that, include: The TC6 bar stock of a predetermined specification is sawn to a predetermined length to obtain the sawn TC6 bar billet, and the β phase transformation point (Tβ) of the sawn TC6 bar billet is measured. The sawn TC6 billet is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held at that temperature to obtain a pretreated TC6 billet. The holding time is: the effective thickness of the sawn TC6 billet × 5~6 min / 10 mm. The pretreated TC6 billet is subjected to the following processes in chronological order: small end aggregation, flat forging, bending, flat forging, large end aggregation, flat forging, large end forming, flat forging, and intermediate billet treatment to obtain a secondary-treated TC6 billet. The intermediate billet treatment includes grinding, sandblasting, and applying glass lubricant. The TC6 billet after the secondary treatment is heated to a high temperature (Tβ-40) ±10℃ in an electric furnace and held at that temperature to obtain the billet to be forged. The holding time is: the effective thickness of the billet to be forged × 5~6 min / 10 mm. TC6 material structural parts are forged using a 630t~1000t electric screw press, followed by heat treatment, isothermal annealing, and physicochemical testing to obtain the finished product. Inspect the processed TC6 material structural parts and put them into storage if they pass the inspection.

2. The forging preparation method of TC6 material structural parts according to claim 1, characterized in that, The TC6 material structural parts that have been processed and formed are inspected and, if qualified, are put into storage, including: Obtain an image of the outer surface of a TC6 material structural component that has been machined and formed; Extract the orientation gradient histogram of the outer surface image; The directional gradient histogram and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image; The multi-channel outer surface image is passed through a first convolutional neural network model, which serves as a feature extractor, to obtain a first depth outer surface feature map. Spatial pyramid pooling is performed on the first depth outer surface feature map to obtain a pooled outer surface feature map; The pooled outer surface feature map is processed by a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map; and The classification feature map is passed through a classifier to obtain a classification result, which is used to indicate whether the surface quality of the TC6 material structural parts meets a predetermined standard.

3. The forging preparation method of TC6 material structural parts according to claim 2, characterized in that, The directional gradient histogram and the outer surface image are aggregated along the channel dimension to obtain a multi-channel outer surface image, including: The directional gradient histogram and the outer surface image are aggregated along the channel dimension using the following cascade formula to obtain the multi-channel outer surface image; The cascade formula is as follows: in, This represents the directional gradient histogram. This represents the image of the outer surface. Represents a cascade function. This represents the multi-channel outer surface image.

4. The forging preparation method of TC6 material structural parts according to claim 3, characterized in that, The multi-channel outer surface image is processed through a first convolutional neural network model, which acts as a feature extractor, to obtain a first depth outer surface feature map, including: Each layer of the first convolutional neural network model, which acts as a feature extractor, processes the input data during the forward propagation of the layer as follows: The input data is processed by convolution to obtain a convolutional feature map; The convolutional feature map is subjected to mean pooling to obtain a pooled feature map; and The pooled feature map is nonlinearly activated to obtain an activated feature map; Wherein, the output of the last layer of the first convolutional neural network model, which serves as the feature extractor, is the first depth outer surface feature map, and the input of the first layer of the first convolutional neural network model, which serves as the feature extractor, is the multi-channel outer surface image.

5. The forging preparation method of the TC6 material structural component according to claim 4, characterized in that, The spatial pyramid pooling employs four different scales of max pooling operations: 13x13, 9x9, 5x5, and 1x1.

6. The forging preparation method of the TC6 material structural component according to claim 5, characterized in that, The second convolutional neural network model, which serves as a feature extractor, includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and an output layer.

7. The forging preparation method of TC6 material structural parts according to claim 6, characterized in that, The first convolutional layer uses a 3×3 kernel with a stride of 1, same padding, and ReLU activation function; the second convolutional layer uses a 3×3 kernel with a stride of 1, same padding, and ReLU activation function; the third convolutional layer uses a 3×3 kernel with a stride of 1, same padding, and ReLU activation function; the first pooling layer uses 2×2 max pooling with a stride of 2; the second pooling layer uses 2×2 max pooling with a stride of 2; the third pooling layer uses 2×2 max pooling with a stride of 2.

8. The forging preparation method of the TC6 material structural component according to claim 7, characterized in that, The pooled outer surface feature map is passed through a second convolutional neural network model, which acts as a feature extractor, to obtain a classification feature map, including: The pooled outer surface feature map is processed by a second convolutional neural network model, which acts as a feature extractor, to obtain a second depth outer surface feature map; and The first depth outer surface feature map and the second depth outer surface feature map are fused to obtain the classification feature map.

9. The forging preparation method of the TC6 material structural component according to claim 8, characterized in that, The classification feature map is obtained by fusing the first depth outer surface feature map and the second depth outer surface feature map, including: The first depth outer surface feature map and the second depth outer surface feature map are fused using the following fusion formula to obtain the classification feature map; The fusion formula is as follows: in, and These are the first depth outer surface feature map and the second depth outer surface feature map, respectively. characteristic matrices, and It is the first of the classification feature maps Each feature matrix It is the transpose of a matrix. and These represent matrix multiplication and addition, respectively.

Images