Method and device for automatically designing gas path diagram based on AI

Abstract

The invention discloses a method and device for automatically designing a gas path diagram based on AI. The method comprises the steps that based on feature points of a three-dimensional model, key point cloud of the three-dimensional model is generated through a point cloud processing algorithm, and based on the key point cloud, part category recognition is conducted through a neural network model; correspondingly binding the identified part category with a preset two-dimensional legend block to obtain part legend binding data; analyzing the gas path schematic diagram vector file, extracting graphic metadata in the gas path schematic diagram vector file, and generating part connection relation data based on the part legend binding data and the extracted graphic metadata; and based on the part connection relation data, connecting the corresponding two-dimensional legend blocks and adding feature attributes to generate a gas path diagram. According to the invention, the technical problems of inaccurate gas path diagram generation and low efficiency in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent design of automobiles, in particular, to a method and device for automatically designing a pneumatic circuit diagram based on AI. BACKGROUND

[0002] The competition in the automobile industry is becoming increasingly fierce, and emerging automobile enterprises are emerging like mushrooms after rain. With the growth of automobile production lines and the development of many new vehicle models, along with the design and modification of tooling fixtures for different vehicle body part assemblies and sub-assemblies, designers need to combine product processes to illustrate the principles of the fixtures in order to guide on-site technicians to assemble tooling fixtures according to the pneumatic circuit principle.

[0003] The design of a pneumatic circuit diagram requires air sources, valve rows, solenoid valves, air cylinders, suction cups, throttle valves, etc. Designers classify different air cylinders under a certain solenoid valve for control (collectively referred to as an action group) according to the process documents, and then connect the legends of each part according to certain standards or rules.

[0004] When designing a pneumatic circuit, the number of air cylinders in the action group is defined by the designer, which can easily result in missed or repeated selection and is difficult to detect. The connection relationship between different parts requires a high level of understanding of the pneumatic circuit principle of the entire fixture by the designer, which limits the design experience of the designer.

[0005] In view of the above problems, no effective solution has been proposed so far. SUMMARY

[0006] The embodiments of the present application provide a method and device for automatically designing a pneumatic circuit diagram based on AI, to at least solve the technical problems of inaccurate and low efficiency in generating a pneumatic circuit diagram in the prior art.

[0007] According to one aspect of an embodiment of the present application, a method for automatically designing a pneumatic circuit diagram based on AI is provided, comprising: generating a key point point cloud of a three-dimensional model based on feature points of the three-dimensional model using a point cloud processing algorithm, and performing part recognition using a neural network model based on the key point point cloud to obtain a part category; binding the recognized parts in the three-dimensional model with corresponding two-dimensional legend blocks based on the part category to obtain part legend binding data; parsing a pneumatic principle diagram vector file to extract graphic metadata therein, and generating part connection relationship data based on the part legend binding data and the extracted graphic metadata; and connecting and adding feature attributes to corresponding two-dimensional legend blocks based on the part connection relationship data to generate a pneumatic circuit diagram.

[0008] According to another aspect of the embodiment of the present application, there is also provided an apparatus for automatically designing a pneumatic circuit diagram based on AI, comprising: an identification module configured to generate a key point point cloud of a three-dimensional model based on feature points of the three-dimensional model, and perform part identification based on the key point point cloud by using a neural network model to obtain a part category; a binding module configured to bind the identified parts in the three-dimensional model with corresponding two-dimensional legend blocks based on the part category to obtain part legend binding data; a connection determination module configured to parse a pneumatic schematic diagram vector file, extract graphic metadata therein, and generate part connection relationship data based on the part legend binding data and the extracted graphic metadata; and a generation module configured to connect and add feature attributes to the corresponding two-dimensional legend blocks based on the part connection relationship data to generate a pneumatic circuit diagram.

[0009] In the embodiment of the present application, a key point point cloud of a three-dimensional model is generated based on feature points of the three-dimensional model by using a point cloud processing algorithm, and part identification is performed based on the key point point cloud by using a neural network model to obtain a part category; the identified parts in the three-dimensional model are bound with corresponding two-dimensional legend blocks based on the part category to obtain part legend binding data; a pneumatic schematic diagram vector file is parsed, graphic metadata therein is extracted, and part connection relationship data is generated based on the part legend binding data and the extracted graphic metadata; and the corresponding two-dimensional legend blocks are connected and feature attributes are added based on the part connection relationship data to generate a pneumatic circuit diagram. The above-mentioned solution solves the technical problems of inaccurate and low-efficiency generation of a pneumatic circuit diagram in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0010] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0011] Figure 1 is a flowchart of an optional method for automatically designing a pneumatic circuit diagram based on AI according to an embodiment of the present application;

[0012] Figure 2 is a flowchart of another optional method for automatically designing a pneumatic circuit diagram based on AI according to an embodiment of the present application;

[0013] Figure 3 is an optional schematic diagram of key feature points of a cylinder according to an embodiment of the present application;

[0014] Figure 4 is an optional legend schematic diagram according to an embodiment of the present application;

[0015] Figure 5 is a schematic diagram of two different connection relationships and styles according to an embodiment of the present application;

[0016] Figure 6 is a schematic diagram of block names and variables according to an embodiment of the present application;

[0017] Figure 7 is a schematic diagram of layers according to an embodiment of the present application;

[0018] Figure 8 is a flowchart of another optional method for automatically designing a pneumatic circuit diagram based on AI according to an embodiment of the present application;

[0019] Figure 9 is a schematic diagram of a structure for automatically designing a pneumatic circuit diagram based on AI according to an embodiment of the present application;

[0020] Figure 10 shows a schematic diagram of a computer device suitable for implementing embodiments of the present disclosure. DETAILED DESCRIPTION

[0021] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative work should fall within the scope of protection of the present application.

[0022] It should be noted that the terms "first", "second", and the like in the description and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to include only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to the process, method, product or device.

[0023] According to an embodiment of the present application, a method for automatically designing a gas circuit diagram based on AI is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0024] Figure 1 A method for automatically designing a gas circuit diagram based on AI according to an embodiment of the present application, as shown in Figure 1 the method comprises the following steps:

[0025] In step S102, based on the feature points of the three-dimensional model, a key point point cloud of the three-dimensional model is generated using a point cloud processing algorithm, and based on the key point point cloud, a part recognition is performed using a neural network model to obtain a part category.

[0026] For example, the feature points of the three-dimensional model are extracted, wherein the feature points include: a center of gravity point, a maximum contour point, a protective cover mounting point, a pressure arm connecting point, an air inlet and outlet mounting point, a cylinder assembly point and a pressure arm rotating point; the feature points are combined to form the key point point cloud of the three-dimensional model using the point cloud processing algorithm with the center of gravity point as the reference point; based on the key point point cloud, a feature learning is performed using the neural network model to obtain spatial structure data of the three-dimensional model, and based on the spatial structure data, the part category of the three-dimensional model is determined.

[0027] In step S104, based on the part category, the parts recognized in the three-dimensional model are correspondingly bound with corresponding two-dimensional legend blocks to obtain part legend binding data.

[0028] The part category is matched with a mapping relationship table to determine the two-dimensional legend blocks corresponding to the part category, wherein the mapping relationship table stores the mapping relationship between parts and two-dimensional legend blocks; based on the part category and the corresponding two-dimensional legend blocks, a binding comparison relationship between parts and two-dimensional legend blocks is generated to obtain part binding mapping data; based on the part binding mapping data and the parts in the three-dimensional model, the part legend binding data is obtained.

[0029] In step S106, a gas circuit schematic diagram vector file is parsed to extract graphic metadata therefrom, and based on the part legend binding data and the extracted graphic metadata, part connection relationship data is generated.

[0030] The gas circuit principle diagram vector file is parsed to extract the graphic metadata, where the graphic metadata includes: legend blocks, layers, variable objects, connection relationships between different legend blocks, and text objects; connection point information between the legend blocks is determined based on the layers; annotation parameters related to the legend blocks are determined based on the variable objects, and text features associated with the legend blocks or the variable objects are determined based on the text objects; multiple cylinders are classified into the same electromagnetic valve for unified control according to a preset action group configuration rule, and action group configuration data is generated; and the part connection relationship data is generated based on the part legend binding data, the connection point information, the annotation parameters, the text features, the action group configuration data, and the connection relationships between the different legend blocks.

[0031] In step S108, the corresponding two-dimensional legend blocks are connected and feature attributes are added based on the part connection relationship data, and a gas circuit diagram is generated.

[0032] First, the line connection mode between the two-dimensional legend blocks is determined based on the part connection relationship data; connection lines are generated between the two-dimensional legend blocks, and connection topology data is formed based on the connection mode. In some embodiments, the connection topology data can also be optimized, for example, the current layout corresponding to the connection topology data is divided into a plurality of grid cells, the line intersection density of the plurality of grid cells is calculated, and based on the line intersection density, a conflict hotspot area is determined; for the conflict hotspot area, a candidate solution set is generated by adjusting the node position, inserting a virtual inflection point, or exchanging the positions of nodes on the same layer, and the adaptive generation value of each candidate solution in the candidate solution set is calculated; based on the adaptive generation value, the optimal candidate solution is determined from the candidate solutions, and the connection topology data is optimized based on the optimal candidate solution.

[0033] Then, feature attributes are added to the connection lines and the two-dimensional legend blocks based on the connection topology data, where the feature attributes include at least one of the following: line type, color, font, and font size; the two-dimensional legend blocks and the connection lines are rendered based on the feature attributes, and the gas circuit diagram is generated.

[0034] Figure 2is another AI-based automatic air circuit schematic design method according to an embodiment of the present application. This embodiment intelligently identifies the parts contained in each type of air circuit schematic under a certain category (or standard) through an AI model, such as air source, solenoid valve, air cylinder, etc., and the connection relationship between each part; then, the AI model further identifies the number and category of three-dimensional models contained in the fixture, such as the opening and closing states of the air cylinder model; finally, the identified three-dimensional models are bound to the corresponding legends in the air circuit schematic, so that the air circuit diagram can be automatically generated through the identification of the three-dimensional models.

[0035] Specifically, as shown in Figure 2 The method comprises the following steps:

[0036] Step S202, by creating a feature point cloud, the overall and local spatial structure is built to identify the category of the part.

[0037] In the design of the air circuit diagram, the parts of a certain category, such as air cylinder and suction cup, need to be automatically identified.

[0038] First, based on the feature points of the three-dimensional model, a key point cloud is generated. Figure 3 As shown in the figure, the key feature points of the air cylinder three-dimensional model are shown, wherein (a) is the maximum contour point, (b) is the protective cover mounting point, (c) is the pressure arm connecting point, (d) is the air inlet and outlet mounting point, (e) is the air cylinder assembly point, and (f) is the pressure arm rotating point. Taking the three-dimensional model gravity center as the reference point, using point cloud processing algorithms such as ISS or Harris3D algorithm, the above six categories of points are integrated to form the key point cloud of the specific model.

[0039] Specifically, using Graph Neural Networks (GNNs) or improved point cloud processing model PointNet++, the feature points in the three-dimensional model are respectively generated and defined according to the category, such as automatically generating the gravity point of the three-dimensional model in CATIA, and then defining this point as the reference point; the corresponding category points are created in turn according to the six categories of feature points shown in the above figure, and the created feature points can contain coordinates or the category features to which the feature points belong. Using PointNet++ as the core network architecture, the feature points are connected and processed to form the geometric contour and spatial structure data of the three-dimensional model; the local structures of different categories are learned in turn to obtain the part category.

[0040] This embodiment further builds the overall and local spatial structure by creating a key point cloud to identify the category of the part. After identifying the part category, through data feeding, the AI model can automatically identify the corresponding 2D legend block corresponding to different three-dimensional models, and in the subsequent automatic generation of the air circuit diagram, only the identification of the three-dimensional model is needed to recommend the corresponding 2D legend block.

[0041] Step S204: The AI ​​model automatically identifies features such as legend blocks, layers, and variables contained in the gas path schematic vector file, as well as the connection relationships between different legend blocks and text that is related to the legend blocks.

[0042] Specifically, during the data learning phase, the AI ​​model analyzes the gas path schematic vector file from the feed to determine which symbols are used and the connection relationships between symbols under this drawing standard. The AI ​​model directly parses the gas path schematic vector file (DWG, DXF, etc.) designed by CAD software, which directly contains object information such as blocks, lines, layers, variables, and text. Blocks are connected by lines, layers are used to define connection points, variables define annotation content and their relative positions to blocks, and text can be parsed for features such as font, size, and color. The schematic diagram is shown below. Figure 4 As shown, (a) is a diagram of a solenoid valve, (b) is a label for a solenoid valve, (c) is a diagram of a cylinder, and (d) is a label for a cylinder. The connection relationships and styles are as follows. Figure 5 As shown, where, Figure 5 (a) represents the connection between different legends, and (b) represents the connection between different types of joints. The block names and variables are as follows: Figure 6 As shown, the layers are as follows Figure 7 As shown.

[0043] Step S206: Based on the 2D legend blocks corresponding to the identified 3D model, and combined with the parsed gas path principle vector file, connect the legend blocks according to the parsed rules, and add the corresponding features.

[0044] Compared with existing technologies, this application simplifies the operation steps, greatly reduces the workload of designers, improves work efficiency, and lowers design costs. Furthermore, compared with existing technologies, designers do not need to consider the 2D legends corresponding to the current model, the connection forms between different legends in the final gas path diagram, the line type and color, the font, size, and color of the text, etc. The design process does not require the full participation of designers, greatly improving design efficiency.

[0045] Figure 8 This is yet another AI-based method for automatically designing gas path diagrams according to embodiments of the present invention, such as... Figure 8 As shown, the method includes the following steps:

[0046] Step S802: Identify the part categories of the 3D model based on the neural network model.

[0047] In this embodiment, first, feature point information is extracted from the three-dimensional model file of the tooling fixture. The feature points include seven types of feature points, i.e., a gravity center point, a maximum contour point, a protective cover mounting point, a pressure arm connecting point, an air inlet / outlet mounting point, a cylinder assembly point, and a pressure arm rotating point. A local coordinate system is established with the gravity center point of the three-dimensional model as the origin, and the normalized feature point set is input into a point cloud processing module to construct a key point point cloud based on spatial clustering of a KD tree.

[0048] In the feature recognition stage, the improved multi-scale attention network PointNet++-A structure is adopted in this embodiment. PointNet++-A introduces a spatial attention mechanism in the feature aggregation layer of the traditional PointNet++, and adaptively weights the importance of local point cloud clusters. Then, through joint optimization of back propagation and cross-entropy loss function, the neural network model outputs the corresponding part category label. In addition, in the training stage, a class hierarchy constraint term is added to the loss function to make similar parts (such as single-acting cylinders and double-acting cylinders) maintain a reasonable distance relationship in the feature space, thereby reducing the misclassification problem caused by similar categories.

[0049] Step S804, bind the part category with the two-dimensional legend block.

[0050] To realize automatic matching, the embodiment establishes a mapping relationship table of part categories and two-dimensional legend blocks, which is stored in the database in JSON format, and each part category corresponds to at least one standard legend template. According to the category label, the corresponding two-dimensional legend block is automatically retrieved.

[0051] When there is no standard legend for a certain category in the rule library, the embodiment uses a graph convolution network (GCN) based legend similarity calculation module to perform similarity matching on the candidate legend set, and selects the most similar legend for dynamic binding.

[0052] Step S806, parse the pneumatic circuit schematic vector file to generate part connection relationship data.

[0053] The pneumatic circuit schematic vector file (such as DWG, DXF, etc.) is parsed to extract the legend block, layer, variable object, text object, and their connection relationship, and an initial graph topology structure is constructed.

[0054] To overcome the connection misjudgment problem caused by line segment offset, breakage, or layering in traditional CAD drawings, the embodiment adopts an adaptive connection recognition algorithm based on geometric-semantic dual domain fusion. The algorithm includes three stages:

[0055] Firstly, the spatial coordinates of the connection endpoints of each node are calculated, and a spatial hash table is established. When the distance between the endpoints of any two nodes is less than the dynamic threshold ε (adaptively adjusted based on the global average connection length and the layer density), it is determined that there is a potential connection.

[0056] Next, the part category and attribute text are converted into semantic vectors using the part semantic embedding module fine-tuned by the BERT model, and the cosine similarity between nodes is calculated. When the similarity is higher than the set threshold and the connection direction conforms to the pneumatic control logic (such as solenoid valve → air cylinder), it is confirmed as a semantically valid connection.

[0057] Finally, the identified connection relationships are input into a lightweight graph neural network (GNN) for self-supervised training to learn the pneumatic circuit connection patterns under different standards. By minimizing the topological consistency loss function, topological rule transfer and adaptive learning are achieved. For example, the identified connection relationships are composed into a high-dimensional feature vector set, including geometric distance, connection angle, semantic similarity, and historical connection probability. The lightweight graph neural network is trained through a self-supervised learning mechanism to learn the pneumatic circuit connection patterns under different design styles. During training, the topological consistency loss function is minimized:

[0058]

[0059] wherein, is the connection feature predicted by the current model, is the standard template connection feature, is the current topological degree, is the reference topological degree, is the balance coefficient, and i and j represent the node (part) numbers in the pneumatic circuit graph, such as solenoid valve, air cylinder, suction cup, etc. represents the edge set between all nodes. Through the above method, the model can automatically learn the structural rules of different pneumatic circuit templates, realizing adaptive topological analysis across projects. Finally, the output part connection relationship data not only contains the connection node pairs, but also can be accompanied by the confidence coefficient of each edge, the semantic weight, the connection type, and the topological level information. These data are stored as JSON structured files as part connection relationship data.

[0060] Compared with the traditional geometric matching method, the present embodiment can accurately restore the logical connection even if there are broken lines, offsets, or layering in the CAD graph through the adaptive threshold and semantic embedding mechanism; the confidence coefficient output can be used for visual prompts in subsequent graphic rendering; through the graph neural network self-learning stage, different manufacturers' pneumatic circuit standards can be automatically adapted, realizing cross-project model transfer.

[0061] Step S808: generating a pneumatic circuit graph based on the part connection relationship data.

[0062] Firstly, based on the part connection relationship data, the logical relationship between the air source, electromagnetic valve, cylinder, suction cup, sensor and other parts is analyzed, and the connection properties and connection direction of each pair of legend block are determined by combining with the action group configuration data. In order to ensure the logical consistency of the layout, the electromagnetic valve is taken as the center node and is taken as the core of the connection topology; the cylinders and suction cups connected with the electromagnetic valve are distributed on both sides of the electromagnetic valve, and the sensors are arranged on the extension line of the cylinder output end. The initial placement position of each part legend is obtained by converting the space topology of the three-dimensional model, so that the two-dimensional legend layout and the three-dimensional physical structure maintain a corresponding relationship.

[0063] Then, connection lines are generated between each two-dimensional legend block, and connection topology data is formed. Specifically, first, an initial layout framework is created. According to the action group relationship, the corresponding legend template is called, and the electromagnetic valve, cylinder, suction cup, sensor and the like are placed in the drawing coordinate system according to the logical level. The constraint conditions for all parts are established, including: they must maintain connectivity with each other and cannot have breakpoints; the boundary rectangles of any two legend blocks cannot overlap and need to maintain a safe distance; the length of a single air path cannot exceed the set threshold; the cylinders in the same action group are horizontally aligned, etc.

[0064] Then, the initial layout is scanned, and the number of cross connections, local area density and total pipe length are calculated. When the number of cross connections or the density exceeds the preset threshold, it is determined that there is a conflict, and the layout optimization process is automatically entered. In the layout optimization stage, hierarchical placement and topology adjustment operations are performed. The electromagnetic valves are uniformly distributed to the first layer, the cylinders and suction cups are distributed to the second layer, and the sensors are located in the third layer. According to the relationship between the inlet and outlet directions, the left and right positions are automatically adjusted to minimize the cross-layer connection. Hierarchical distribution can concentrate the main air path lines between layers and reduce invalid intersections.

[0065] In each layer, the center of gravity of the node is calculated, and the center of gravity coordinates are determined by the average position of the connected nodes in the upper layer; then the nodes are sorted according to the center of gravity result. When the sorting still causes intersection, a dynamic exchange strategy is adopted, that is, the positions of adjacent nodes are exchanged, and only when the number of intersections is reduced, the new arrangement is retained.

[0066] After determining the node order, an optimal routing method is selected for each connection path. Straight-line connections are preferred, and when a straight line causes a crossing, a broken-line connection is used, but the broken-line inflection point does not exceed two. If there is still a high-density crossing in the local area, the area is extracted as a local subgraph, and adaptive topology association and local iterative optimization are performed. Specifically, the entire current layout is divided into multiple grid units, and the wire crossing density of each grid unit is calculated; if the density of a unit exceeds twice the global average, the unit is identified as a conflict hotspot area. For the hotspot area, multiple candidate layouts are generated, and new feasible layout schemes are generated as a candidate solution set by adjusting the node position, inserting a virtual inflection point, or exchanging the positions of nodes on the same layer. Each candidate solution is evaluated according to the adaptive cost function. In some embodiments, the adaptive cost function can be generated based on the number of wire crossings, the total path length, the total number of broken-line inflection points, the global average wire density, and the local average wire density:

[0067]

[0068] where C is the adaptive cost value of the candidate layout, is the number of connection line crossings, is the local area wire density, is the global average wire density, is the total path length, is the total number of broken-line inflection points, is the broken-line reference threshold, is the global weight, which can be dynamically adjusted, is the local weighting coefficient. The adaptive cost function reduces the weight of the number of crossings in the early stage of optimization to reduce the number of crossings and quickly eliminate major conflicts; in the middle stage, the weight of path smoothness and length is gradually increased to increase the total path length; in the later stage, the weight of the number of broken-line inflection points is reduced, and the weight of the local density is provided to optimize the uniformity and overall aesthetics of the layout. If the candidate solution has a cost value that is less than 20% of the initial cost, it is directly accepted; otherwise, the probability function

[0069]

[0070] decides whether to keep. In other words, when the adaptive cost value of the candidate solution is less than a preset cost threshold (20% of the initial cost), the candidate solution is kept; otherwise, the probability value of the candidate solution is calculated, and when the probability value is higher than a preset probability threshold, for example, 0.3, the candidate solution is kept, and when the probability value is lower than 0.3, the candidate solution is discarded. Where T is a temperature parameter, represents the local conflict factor, represents the cost function. Through the above probability function, the The value is greater, and the suboptimal solution can be accepted in the local search to jump out of the local minimum. Finally, the optimal candidate solution is determined from the reserved candidate solutions according to the size of the adaptive generation value, and the connection topology data is optimized based on the optimal candidate solution.

[0071] Then, based on the generated connection topology data, feature attributes are added to the connection lines and each two-dimensional legend block. The feature attributes include parameters such as line type, color, font, and font size. For example, the high-pressure gas path is represented by a red solid line, the low-pressure gas path is represented by a green thin line, the control signal line is represented by a blue dashed line, and the label font is automatically adjusted according to the scale. The feature attributes are not only used for graphic rendering, but also can be used as attribute labels for subsequent gas path control systems to support automatic recognition and pneumatic simulation.

[0072] Finally, based on the above feature attributes, rendering operations are performed on all two-dimensional legend blocks and their connection lines to generate the final gas path diagram. During the rendering process, when the connection relationship is determined by high-confidence data, the line is displayed in a thick solid line; when the connection is inferred by fuzzy matching, it is presented in a light dashed line. In this way, designers can intuitively identify the reliability of the AI-generated results. The final generated gas path diagram is output in vector form and can be exported as a DWG, DXF, or SVG file.

[0073] In summary, the embodiment of the application combines point cloud recognition, legend binding, and adaptive topology association analysis algorithm to realize the full-process automatic generation from three-dimensional structure to two-dimensional gas path diagram. Especially, it can adaptively identify and correct the connection relationship in complex and multi-layer gas path structure, providing a tool for the intelligentization of gas path design.

[0074] The application also provides an AI-based automatic design gas path diagram device, as shown in Figure 9 The recognition module 92 is configured to generate key point point cloud of the three-dimensional model based on feature points of the three-dimensional model using a point cloud processing algorithm, and perform part category recognition based on the key point point cloud using a neural network model. The binding module 94 is configured to correspondingly bind the recognized part category with a preset two-dimensional legend block to obtain part legend binding data. The connection determination module 96 is configured to parse a gas path schematic diagram vector file, extract graphic metadata therefrom, and generate part connection relationship data based on the part legend binding data and the extracted graphic metadata. The generation module 98 is configured to connect and add feature attributes to the corresponding two-dimensional legend block based on the part connection relationship data to generate a gas path diagram.

[0075] It should be noted that the above embodiment provides the device for automatically designing the gas circuit diagram based on AI, and only the division of the above functional modules is exemplified, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device for automatically designing the gas circuit diagram based on AI provided by the above embodiment and the method embodiment for automatically designing the gas circuit diagram based on AI belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be described here.

[0076] Figure 10 The structural schematic diagram of a computer device suitable for implementing the embodiments of the present disclosure is shown. It should be noted that Figure 10 The computer device shown is only an example and should not impose any limitation on the functions and use range of the embodiments of the present disclosure.

[0077] As Figure 10 shown, the computer device includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage portion 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for system operation are also stored. The CPU 1001, the ROM 1002, and the RAM 1003 are connected to each other through a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0078] The following components are connected to the I / O interface 1005: an input portion 1006 including a keyboard, a mouse, and the like; an output portion 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker, and the like; a storage portion 1008 including a hard disk, and the like; and a communication portion 1009 including a network interface card such as a LAN card, a modem, and the like. The communication portion 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as necessary. A removable recording medium 1011 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 1010 as necessary, so that a computer program read therefrom is installed in the storage portion 1008 as necessary.

[0079] The above is only the preferred embodiment of the present application, and it should be noted that those skilled in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A method for automatically designing gas path diagrams based on AI, characterized in that, include: Based on the feature points of the 3D model, a point cloud of key points of the 3D model is generated using a point cloud processing algorithm, and based on the point cloud of key points, a neural network model is used to identify the part category. The identified part categories are bound to the corresponding two-dimensional legend blocks to obtain part legend binding data; Parse the gas path schematic vector file, extract the graphic metadata, and generate part connection relationship data based on the part legend binding data and the extracted graphic metadata; Based on the component connection relationship data, the corresponding two-dimensional legend blocks are connected and feature attributes are added to generate an air path diagram.

2. The method according to claim 1, characterized in that, Based on feature points of a 3D model, a point cloud of key points of the 3D model is generated using a point cloud processing algorithm. Then, based on the point cloud of key points, a neural network model is used to identify part categories, including: The feature points of the three-dimensional model are extracted, including: center of gravity, maximum contour point, protective cover mounting point, pressure arm connection point, air inlet and outlet mounting point, cylinder assembly point, and pressure arm rotation point. Using the centroid as a reference point, the feature points are combined using the point cloud processing algorithm to form the key point cloud of the three-dimensional model; Based on the key point cloud, feature recognition is performed using the neural network model to obtain the spatial structure data of the three-dimensional model, and based on the spatial structure data, the part category of the three-dimensional model is determined.

3. The method according to claim 1, characterized in that, The identified part categories are then bound to their corresponding 2D legend blocks to obtain part legend binding data, including: The part categories are matched with the mapping table to determine the corresponding two-dimensional legend blocks, wherein the mapping table stores the mapping relationship between part categories and two-dimensional legend blocks; Based on the part category and the corresponding two-dimensional legend block, a binding relationship between the part category and the two-dimensional legend block is generated, which serves as the part legend binding data.

4. The method according to claim 1, characterized in that, The gas path schematic vector file is parsed, and its graphic metadata is extracted. Based on the component legend binding data and the extracted graphic metadata, component connection relationship data is generated, including: The gas path schematic vector file is parsed to extract the graphic metadata, which includes: legend blocks, layers, variable objects, connection relationships between different legend blocks, and text objects. Based on the layer, determine the connection point information between the legend blocks; based on the variable object, determine the annotation parameters related to the legend block; based on the text object, determine the text features associated with the legend block or the variable object. Based on the part legend binding data, the connection point information, the annotation parameters, the text features, and the connection relationships between different legend blocks, the part connection relationship data is generated.

5. The method according to claim 1, characterized in that, Based on the component connection relationship data, the corresponding two-dimensional legend blocks are connected and feature attributes are added to generate a gas path diagram, including: Based on the component connection relationship data, determine the line connection method between the corresponding two-dimensional legend blocks; Connecting lines are generated between the two-dimensional legend blocks, and connection topology data is formed based on the connection method; Based on the connection topology data, feature attributes are added to the connection lines and the corresponding two-dimensional legend blocks, wherein the feature attributes include at least one of the following: line type, color, font, and font size; Based on the aforementioned feature attributes, the corresponding two-dimensional legend block and the connecting lines are rendered to generate the gas path diagram.

6. The method according to claim 5, characterized in that, After forming connection topology data based on the connection method, the method further includes: The current layout corresponding to the connection topology data is divided into multiple grid cells, the line intersection density of the multiple grid cells is calculated, and conflict hotspot areas are determined based on the line intersection density. For the conflict hotspot area, a candidate solution set is generated by adjusting the node position, inserting virtual inflection points, or exchanging the positions of nodes at the same level, and the adaptive cost value of each candidate solution in the candidate solution set is calculated. Based on the adaptive cost, an optimal candidate solution is determined from the candidate solutions, and the connection topology data is optimized based on the optimal candidate solution.

7. An AI-based device for automatically designing gas path diagrams, characterized in that, include: The recognition module is configured to generate a key point cloud of the 3D model based on feature points of the 3D model using a point cloud processing algorithm, and to perform part category recognition based on the key point cloud using a neural network model. The binding module is configured to bind the identified part categories to the corresponding two-dimensional legend blocks to obtain part legend binding data; The connection determination module is configured to parse the gas path schematic vector file, extract the graphic metadata therein, and generate part connection relationship data based on the part legend binding data and the extracted graphic metadata; The generation module is configured to connect the corresponding two-dimensional legend blocks and add feature attributes based on the part connection relationship data to generate an air path diagram.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 6.

9. A computer device, characterized in that, include: Memory and processor The memory stores computer programs; The processor is configured to execute a computer program stored in the memory, wherein when the computer program is executed, the processor performs the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images