An automatic drawing method and device for a digitized flowchart

Abstract

The application discloses an automatic drawing method and device for a digital flowchart, wherein the method comprises the following steps: step S1, obtaining drawing requirements input by a user, analyzing the drawing requirements, and obtaining device information and a topological relationship of the flowchart; step S2, based on the analysis result, generating a graph element layout scheme and a pipeline routing scheme of the flowchart by using a generative model; and step S3, matching corresponding device legends in a legend library according to the device information, establishing a connection relationship between the device legends according to the topological relationship, obtaining structured data of the digital flowchart, and realizing drawing of the digital flowchart. The application can realize automatic generation of the digital flowchart.

Description

Technical Field

[0001] This invention relates to the field of flowchart digitization technology, and in particular to an automatic drawing method and apparatus for digital flowcharts. Background Technology

[0002] Currently, the creation of digital flowcharts relies primarily on manual labor. Even with AI assistance, significant human intervention is still required, making it time-consuming and labor-intensive for complex drawings and repetitive tasks. In scenarios lacking original base maps, such as new system design or conceptual design, drafters must create flowcharts from scratch, resulting in low efficiency. Furthermore, inconsistent styles among different drafters make it difficult to guarantee consistency within the same series of drawings. Therefore, there is an urgent need for methods to automatically generate structured digital drawings, achieving end-to-end automated drafting to improve efficiency and consistency. Summary of the Invention

[0003] In view of this, it is necessary to provide an automatic drawing method and apparatus for digital flowcharts to effectively solve the technical problems of low automation and low consistency in flowchart drawing.

[0004] This invention provides an automatic drawing method for digital flowcharts, comprising the following steps: Step S1: Obtain the drawing requirements input by the user, parse the drawing requirements, and obtain the device information and topological relationships of the flowchart; Step S2: Based on the analysis results, a generative model is used to generate the element layout scheme and pipeline routing scheme of the flowchart. Step S3: Match the corresponding equipment legend in the legend library according to the equipment information, establish the connection relationship between each equipment legend according to the topology relationship, obtain the structured data of the digital flowchart, and realize the drawing of the digital flowchart.

[0005] Preferably, step S1 specifically includes: Multiple input methods are set up to obtain multimodal drawing requirements from users; these multimodal drawing requirements include hand-drawn sketches, text descriptions, reference drawings, and system templates. When a user inputs a hand-drawn sketch, a convolutional neural network is used to extract sketch features, identify equipment symbols, pipeline areas, and text annotations in the hand-drawn sketch, identify the corresponding equipment information by recognizing the equipment symbols, perform semantic segmentation on the pipeline areas to extract pipeline vector paths, and perform text recognition on the text annotations to obtain equipment codes, attributes, and annotations. When a user inputs text descriptions, natural language processing methods are used to parse the text descriptions to extract entities, relationships, and sequences. Based on the extracted entities, flowchart device information is obtained, and based on the extracted relationships and sequences, the topological relationships between devices are obtained. When a user inputs data using a reference drawing, the reference drawing is parsed using image processing methods to extract its legend and topological structure. When a user selects a preset system template from the template library, the system template automatically generates a legend and its initial layout to obtain a draft flowchart.

[0006] Preferably, the generation of the primitive layout scheme in step S2 specifically involves: Using the devices to be laid out as graph nodes and the connections between devices as edges, an initial topology graph is constructed. Based on the message passing mechanism of GNN, the feature information of adjacent nodes is aggregated among the graph nodes to learn the topological distance and relative positional relationship between devices. The output layer maps the hidden features of each node into two-dimensional coordinates. The GNN is optimized and trained using the derivative of the total length of the connection, the number of crossover penalties, and the space utilization weight as loss functions. The trained GNN model is then used to generate the primitive layout scheme.

[0007] Preferably, the pipeline route generation in step S2 specifically involves: The flowchart canvas is divided into a uniform grid. The grid where the pipeline starts is used as the search starting point and the grid where the pipeline ends is used as the search ending point. The A* algorithm is used to search for paths in the grid space. The cost function of the A* algorithm uses the Manhattan distance as a heuristic term. The searched grid paths are converted into vector pipelines and the node coordinates are recorded to obtain the pipeline routing scheme.

[0008] Preferably, step S2 further includes: Generate multiple different element layout schemes and pipeline routing schemes, calculate the layout quality index of different schemes, and push the scheme with the best layout quality index as the default recommendation to the user.

[0009] Preferably, the device legend matching in step S3 specifically involves: Match the corresponding legend type from the legend library based on the equipment type, select the legend version based on the reference drawing or user preference, automatically generate the corresponding equipment code according to the equipment coding rules, and obtain the corresponding equipment attributes from the equipment master data based on the equipment code, and generate annotations at the matching equipment legend.

[0010] Preferably, the connection relationship construction in step S3 specifically involves: Based on the topology, establish the connection relationship between each of the device legends and align the pipeline endpoints with the preset connection points of the device legends; if the topology involves jump characters, create jump character elements and record the cross-drawing connection relationship; after the connection relationship is established, check whether all device legends are connected, and if not, mark the unconnected device legends as a reminder.

[0011] Preferably, the method further includes generating the digital flowchart based on the structured data: When a user inputs a hand-drawn sketch, an end-to-end generative adversarial network model is used to convert the sketch into the digital flowchart: based on the structured data, the hand-drawn sketch is encoded into a latent feature vector, the hand-drawn symbols in the sketch are replaced with corresponding equipment legends, the hand-drawn curves in the sketch are replaced with pipelines while maintaining the connection relationship, and the visualization attributes of the equipment legends and pipelines are added according to preset rules to realize the visualization generation of the digital flowchart; When a user inputs text descriptions, a sequence-to-sequence model is used to generate the digital flowchart: based on the structured data, device entities and their attributes are extracted from the text descriptions, the connection relationships between the device entities are identified, and an attention-based layout generator is used to arrange the visual device positions according to the number and relationships of the entities, thereby realizing the visual generation of the digital flowchart. When the user inputs data using a system template, the digital flowchart is generated based on a typical structure: the digital flowchart is generated visually based on the typical structure of the selected system template and the structured data.

[0012] Preferably, the method further includes manually adjusting the generated digital flowchart, recording the user's adjustment operations, and optimizing the corresponding model based on the adjustment operations.

[0013] The present invention also provides an automatic drawing device for digital flowcharts, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the automatic drawing method for the digital flowchart.

[0014] Compared with the prior art, the present invention has the following advantages: When drawing flowcharts, users only need to input the drawing requirements, and the invention automatically parses the requirements to obtain the element information of the flowchart, and then automatically generates the element layout and pipeline routes, and finally automatically generates structured data to complete the automatic digital drawing process of the flowchart, which greatly improves the drawing efficiency and significantly reduces the drawing threshold; at the same time, the automatically generated flowcharts are generated based on the same model, strictly follow the specifications and have a unified style, thereby ensuring the consistency of the flowcharts. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart of an embodiment of an automatic drawing method for digital flowcharts provided by the present invention. Detailed Implementation

[0016] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0017] Example 1

[0018] Please see Figure 1 The automatic drawing method for digital flowcharts in this embodiment specifically includes the following steps: Step S1: Obtain the drawing requirements input by the user, parse the drawing requirements, and obtain the device information and topological relationships of the flowchart; Step S2: Based on the analysis results, a generative model is used to generate the element layout scheme and pipeline routing scheme of the flowchart. Step S3: Match the corresponding equipment legend in the legend library according to the equipment information, establish the connection relationship between each equipment legend according to the topology relationship, obtain the structured data of the digital flowchart, and realize the drawing of the digital flowchart.

[0019] This embodiment uses deep learning technology to automatically generate structured digital drawings from sketches, text descriptions, or existing base maps, achieving end-to-end automated drawing and significantly improving drawing efficiency and consistency.

[0020] The following provides a detailed explanation of each step.

[0021] Step S1: Multimodal Input Parsing. Supports multiple input modes, automatically parsing user intent and generating drawing instructions. Sketch Input: Users hand-draw sketches via tablet, mouse, or uploaded images. The system uses convolutional neural networks, such as ResNet, to extract sketch features, recognizing equipment symbols, pipeline routes, and text annotations. Symbol recognition matches hand-drawn symbols with a standard legend library, outputting equipment type and confidence level. Pipeline extraction uses semantic segmentation to extract pipeline regions and convert them into vector paths. Text recognition uses OCR to recognize handwritten text and convert it into equipment codes or annotations. Text Description Input: Users input natural language descriptions, such as "RCV system, from water tank T001 to pump P001, through valve V001, and then to heat exchanger E001." The system uses natural language processing (NLP), such as BERT, for intent parsing, extracting entities, relationships, and sequences. Entities include equipment entities and system entities. Reference Drawing Input: Users provide one or more reference drawings, which can be images or PDFs. The system uses image recognition to understand the drawing style and layout, extracting legends and topological structures as templates for generating new drawings. Template Selection: Users select preset system templates from the template library, such as "Typical Pump Set" or "Valve Station," and the system automatically generates the basic layout.

[0022] Step S2: Layout Generation and Optimization. Based on the analysis results, a generative model, such as GAN or Transformer, is used to automatically generate the initial layout, which includes two parts: primitive layout and pipeline routing. Primitive Layout Generation: The devices to be laid out are treated as graph nodes, and the connections between devices are treated as edges, constructing an initial topology graph. GNN aggregates the feature information of adjacent nodes through a message passing mechanism, learning the topological distance and relative positional relationships between devices; the output layer maps the hidden features of each node to two-dimensional coordinates, and the loss function is optimized by weighting the reciprocal of the total connection length, the number of crossover penalties, and space utilization, so that the generated layout conforms to the process flow sequence, that is, the main equipment is arranged along the main path axis, and the branch equipment is distributed on both sides.

[0023] Pipeline route generation: The canvas is divided into a uniform grid. The grid containing the pipeline's start and end points is used as the search start and end points. The A* algorithm is used to search for paths in the grid space. Each step of the path is only allowed to move in the horizontal or vertical direction to ensure that the path is straight. In the cost function, the Manhattan distance is used as a heuristic term. Crossing the equipment area is given a high penalty, while sharing the path segment with other existing pipelines is given a low cost to encourage merging of routes. After the search is completed, the grid path is converted into a vector pipeline and the node coordinates are recorded.

[0024] Simultaneously, 2-3 different layout schemes are generated, such as compact, balanced, and hierarchical, for users to choose from. Layout quality assessment automatically calculates layout quality indicators and selects the optimal scheme as the default recommendation. Layout quality indicators include the number of line intersections, element overlap rate, and space utilization.

[0025] Step S3 includes intelligent legend matching and attribute filling. Based on the generated equipment type, the most suitable legend is automatically matched from the legend library. Based on type matching, the corresponding legend is selected according to the equipment type, such as "centrifugal pump"; then based on style matching, a legend version conforming to the specific power plant style is selected based on reference drawings or user preferences, such as Style 1 and Style 2; attribute auto-fill: a default equipment code is automatically generated according to the equipment coding rules, such as "RCV001PO". If matching attribute data exists in the equipment master database, attributes such as equipment name, model, and pressure rating are filled in from the equipment master data; intelligent annotation generation: equipment code annotations are automatically generated next to the equipment, with the position automatically optimized based on the surrounding blank space.

[0026] Step S3 also includes automatic connection relationship construction. Based on the parsed topology, connections between devices are automatically established. Connection point matching automatically aligns pipeline endpoints with device connection points according to preset connection point locations for each device type, such as pump inlet and outlet. Cross-drawing connection processing automatically creates jumper elements and records cross-drawing connection relationships if the generated drawings involve jump symbols. Connection integrity checks automatically detect whether all devices are connected after generation; unconnected devices are highlighted, and a one-click automatic connection option is provided.

[0027] Specifically, this embodiment also includes step S4, which generates the visualization-related element layout, pipeline routing, and visualization attributes of the digital flowchart based on structured data, thereby completing the visualization of the digital flowchart.

[0028] Drawing generation based on hand-drawn sketches. For sketch input, an end-to-end Generative Adversarial Network (GAN) model is used to directly convert sketches into structured digital drawings. Sketch encoding: The sketch image is encoded into latent feature vectors; Legend replacement: Hand-drawn symbols are replaced with standard legends, maintaining their position and rotation angles; Line straightening: Hand-drawn curves are straightened into horizontal and vertical pipes, maintaining their connectivity; Color filling: Colors are automatically added according to preset rules, such as different colors for pipes of different media types.

[0029] Drawing generation based on text descriptions. For text descriptions, a sequence-to-sequence (Seq2Seq) model is used to generate drawings. Entity recognition extracts equipment entities and their attributes from the text, such as "pump P001, centrifugal, flow rate 100m³". 3 / h”; Relationship extraction, identifying the connection relationships between devices, such as “P001 outlet connects to V001”; Layout generation, based on the number of entities and relationships, using an attention-based layout generator to automatically arrange the device positions; Iterative optimization, users can add descriptions, such as “add another filter before the pump”, the system incrementally updates the drawings, keeping the original layout as unchanged as possible.

[0030] Rapid generation based on system templates. A rich template library is provided, supporting one-click generation of typical structures. System templates, such as overall templates for "RCV System" and "RCP System," include all typical equipment and connection relationships of the system; functional templates, such as sub-functional templates for "Pump Group," "Valve Station," and "Heat Exchanger Group"; and area templates, such as area layout templates for "Inside Containment" and "Auxiliary Plant." After the user selects a template, the system automatically generates initial drawings, which the user can then adjust.

[0031] Specifically, this embodiment also includes manual interaction and fine-tuning. Automatically generated drawings support manual fine-tuning. Drag-and-drop adjustment allows users to drag equipment positions, and the system automatically reroutes pipelines. Attribute editing allows double-clicking on equipment to modify attributes, and the system automatically updates relevant annotations. Layout locking allows users to lock the positions of some equipment, allowing AI to optimize the rest. Style switching allows for one-click switching of the legend style, such as switching from the Qinshan style to the Tianwan style.

[0032] Specifically, this embodiment also includes self-learning and model optimization. User adjustments to the automatically generated results are recorded for continuous model optimization. Correction records are kept of user actions such as adjusting device location, modifying attributes, and reconnecting; model fine-tuning involves periodically collecting correction data to fine-tune the generated model, making subsequent generated results more aligned with user preferences; personalized models are trained for high-frequency users to adapt to their drawing habits.

[0033] Compared with existing technologies, this embodiment has the following advantages: Zero-based drawing: Users only need to provide sketches or text descriptions to automatically generate professional-grade digital drawings, significantly lowering the drawing threshold; Improved efficiency: The time to draw a complex system diagram from scratch is reduced from several days to several minutes, shortening the design iteration cycle; Consistency assurance: Automatically generated drawings strictly follow specifications, maintain a unified style, and avoid differences caused by different personnel drawing; Rapid comparison of multiple solutions: Multiple layout schemes are generated simultaneously to assist design decisions; Knowledge reuse: Template libraries and self-learning mechanisms solidify excellent design patterns, enabling rapid startup of new projects.

[0034] The following section uses the design of a new system in a nuclear power plant as an example to provide a more detailed explanation of this embodiment.

[0035] Scenario 1: Generate drawings from sketches.

[0036] The engineer hand-drew a sketch of the RCV system on a tablet, including the water tank, pumps, valves, and heat exchangers, and marked the flow direction with arrows. After uploading the sketch, the system automatically identified: water tank, centrifugal pump, gate valve, and plate heat exchanger, with a confidence level of 98% for the water tank, 95% for the centrifugal pump, 92% for the gate valve, and 90% for the plate heat exchanger. The hand-drawn pipelines were extracted as vector paths. The system replaced the hand-drawn symbols with standard symbols, straightened the pipelines to be horizontal and vertical, and automatically generated equipment codes, such as water tank T001 and pump P001. After generating the layout, the engineer checked and found that the pump position was too far to the left. After manually dragging and adjusting, the system automatically rerouted the pipelines. The total time was 5 minutes to complete the first draft.

[0037] Scenario 2: Generating drawings from text descriptions.

[0038] The designer input: "RCV system, water tank T001 outlet connects to pump P001, pump outlet connects to valve V001, V001 then splits into two paths: one connects to heat exchanger E001, the other connects to valve V002." The system parses the entities: T001, P001, V001, E001, V002, with series and parallel relationships. The system automatically generates the layout: water tank on the left, pump on the right, heat exchanger on top, valve V002 on the bottom, and pipelines automatically connected. After generation, the designer found a missing pressure gauge and added the description: "Add pressure gauge PI001 to pump outlet." The system automatically inserts the pressure gauge into the pump outlet pipeline and adjusts the pipeline.

[0039] Scenario 3: Quickly generate from templates.

[0040] A standard pump set needs to be designed. Select the "Centrifugal Pump Set Template" from the template library to generate the pump, inlet valve, outlet valve, pressure gauge, and check valve with one click, and automatically connect them. Modify the pump model to the specific model according to actual needs, and the system will automatically update the legend and related attributes.

[0041] Scenario 4: Comparison of multiple options.

[0042] For complex layouts, the system generates three options: compact, balanced, and hierarchical. The compact option features dense equipment and short pipelines; the balanced option features evenly distributed equipment; and the hierarchical option is divided by function. Designers compare these options and select the balanced option, then make further fine-tuning.

[0043] Example 2

[0044] This embodiment provides an automatic drawing device for digital flowcharts, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the automatic drawing method for digital flowcharts described in Embodiment 1.

[0045] The automatic drawing device for digital flowcharts provided in this embodiment is used to implement the automatic drawing method for digital flowcharts. Therefore, the automatic drawing device for digital flowcharts also possesses the technical effects of the automatic drawing method for digital flowcharts, which will not be described in detail here.

[0046] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of the present invention.

Claims

1. An automatic drawing method for digital flowcharts, characterized in that, Includes the following steps: Step S1: Obtain the drawing requirements input by the user, parse the drawing requirements, and obtain the device information and topological relationships of the flowchart; Step S2: Based on the analysis results, a generative model is used to generate the element layout scheme and pipeline routing scheme of the flowchart. Step S3: Match the corresponding equipment legend in the legend library according to the equipment information, establish the connection relationship between each equipment legend according to the topology relationship, obtain the structured data of the digital flowchart, and realize the drawing of the digital flowchart.

2. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, Step S1 specifically involves: Multiple input methods are set up to obtain multimodal drawing requirements from users; these multimodal drawing requirements include hand-drawn sketches, text descriptions, reference drawings, and system templates. When a user inputs a hand-drawn sketch, a convolutional neural network is used to extract sketch features, identify equipment symbols, pipeline areas, and text annotations in the hand-drawn sketch, identify the corresponding equipment information by recognizing the equipment symbols, perform semantic segmentation on the pipeline areas to extract pipeline vector paths, and perform text recognition on the text annotations to obtain equipment codes, attributes, and annotations. When a user inputs text descriptions, natural language processing methods are used to parse the text descriptions to extract entities, relationships, and sequences. Based on the extracted entities, flowchart device information is obtained, and based on the extracted relationships and sequences, the topological relationships between devices are obtained. When a user inputs data using a reference drawing, the reference drawing is parsed using image processing methods to extract its legend and topological structure. When a user selects a preset system template from the template library, the system template automatically generates a legend and its initial layout to obtain a draft flowchart.

3. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, The generation of the primitive layout scheme in step S2 is specifically as follows: Using the devices to be laid out as graph nodes and the connections between devices as edges, an initial topology graph is constructed. Based on the message passing mechanism of GNN, the feature information of adjacent nodes is aggregated among the graph nodes to learn the topological distance and relative positional relationship between devices. The output layer maps the hidden features of each node into two-dimensional coordinates. The GNN is optimized and trained using the derivative of the total length of the connection, the number of crossover penalties, and the space utilization weight as loss functions. The trained GNN model is then used to generate the primitive layout scheme.

4. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, The pipeline route generation process in step S2 is specifically as follows: The flowchart canvas is divided into a uniform grid. The grid where the pipeline starts is used as the search starting point and the grid where the pipeline ends is used as the search ending point. The A* algorithm is used to search for paths in the grid space. The cost function of the A* algorithm uses the Manhattan distance as a heuristic term. The searched grid paths are converted into vector pipelines and the node coordinates are recorded to obtain the pipeline routing scheme.

5. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, Step S2 further includes: Generate multiple different element layout schemes and pipeline routing schemes, calculate the layout quality index of different schemes, and push the scheme with the best layout quality index as the default recommendation to the user.

6. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, The specific steps of device legend matching in step S3 are as follows: Match the corresponding legend type from the legend library based on the equipment type, select the legend version based on the reference drawing or user preference, automatically generate the corresponding equipment code according to the equipment coding rules, and obtain the corresponding equipment attributes from the equipment master data based on the equipment code, and generate annotations at the matching equipment legend.

7. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, The connection relationship construction in step S3 is specifically as follows: Establish the connection relationship between each of the equipment legends based on the topology relationship, and align the pipeline endpoints with the preset connection points of the equipment legends; if the topology relationship involves jump characters, create jump character elements and record the cross-drawing connection relationship; After the connection relationship is established, check whether all device legends are connected. If not, mark the unconnected device legends as a reminder.

8. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, It also includes generating the digital flowchart based on the structured data: When a user inputs a hand-drawn sketch, an end-to-end generative adversarial network model is used to convert the sketch into the digital flowchart: based on the structured data, the hand-drawn sketch is encoded into a latent feature vector, the hand-drawn symbols in the sketch are replaced with corresponding equipment legends, the hand-drawn curves in the sketch are replaced with pipelines while maintaining the connection relationship, and the visualization attributes of the equipment legends and pipelines are added according to preset rules to realize the visualization generation of the digital flowchart; When a user inputs text descriptions, a sequence-to-sequence model is used to generate the digital flowchart: based on the structured data, device entities and their attributes are extracted from the text descriptions, the connection relationships between the device entities are identified, and an attention-based layout generator is used to arrange the visual device positions according to the number and relationships of the entities, thereby realizing the visual generation of the digital flowchart. When the user inputs data using a system template, the digital flowchart is generated based on a typical structure: the digital flowchart is generated visually based on the typical structure of the selected system template and the structured data.

9. The automatic drawing method for digital flowcharts according to claim 1, characterized in that, It also includes manually adjusting the generated digital flowchart, recording the user's adjustment operations, and optimizing the corresponding model based on the adjustment operations.

10. An automatic drawing device for digital flowcharts, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the automatic drawing method for digital flowcharts as described in any one of claims 1-9.

Images