An interactive construction method of flowchart and device state linkage method

Abstract

This application relates to the field of computer technology, and in particular to an interactive flowchart construction and device status linkage method. The method includes: receiving a device type and status name input by a user; generating corresponding prompt words according to the device type and status name using a pre-set prompt word generation rule; generating image data that conforms to the prompt word description based on the prompt words and a pre-trained Stable Diffusion model; and storing the image data according to a predetermined rule. The pre-trained Stable Diffusion model is obtained by training a Stable Diffusion model on a training dataset, which includes: pre-acquired device images and descriptive prompt words corresponding to the device images.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method for interactively constructing flowcharts and linking them with device status. Background Technology

[0002] In fields such as industrial control, equipment operation and maintenance management, and production process visualization, flowcharts have been widely used as an important means of describing the operating status of equipment and its logical relationships. Flowcharts can intuitively display equipment types, operating statuses, and the linkages between various nodes, thereby assisting operation and maintenance personnel in monitoring, analysis, and decision-making. For example, in power systems or oil and gas extraction scenarios, different types of equipment (such as pumps, valves, and sensors) typically need to be labeled with status and linked in the form of flowcharts to reflect changes in the operating status of the equipment.

[0003] However, most existing flowchart construction methods rely on manual configuration or pre-built icon libraries. Specifically, users typically need to manually select icons for the corresponding device types in the system and manually annotate or switch icons based on device status (such as running, stopped, or faulty). This method not only relies on human experience and is cumbersome to configure, but also struggles to generate semantically meaningful graphical content in a timely manner when faced with scenarios involving numerous device types and complex status changes, resulting in low flowchart construction efficiency. For example, patent application CN117875315A (classification G06F) provides a method and device for generating prompt words for engineering machinery, and engineering machinery; patent application CN119106143A (classification G06F) provides an AI large-scale model enterprise knowledge base system based on prompt words; and patent application CN119416761A (classification G06F) provides a method and device for generating security data, storage medium, and electronic device.

[0004] Furthermore, while some existing solutions offer automated drawing methods based on templates or rules, they still essentially rely on predefined icon resources or fixed style libraries. When new device types or new status expression requirements arise, additional design and expansion of icon resources are needed, resulting in poor system flexibility and difficulty in adapting to diverse application scenarios. At the same time, different device states often need to be distinguished by different visual styles, but traditional icon libraries struggle to support dynamic style generation and flexible switching, leading to limitations in the visual representation of flowcharts. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, this application provides an interactive flowchart construction and device status linkage method, which solves the technical problems of the prior art, such as flowchart construction relying on manual configuration, poor icon resource scalability, and difficulty in automatically generating corresponding image content according to device type and status semantics.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the main technical solutions adopted in this application include:

[0009] In a first aspect, embodiments of this application provide an interactive flowchart construction and device status linkage method, comprising: receiving a device type and status name input by a user; generating corresponding prompt words according to the device type and status name using pre-set prompt word generation rules; generating image data that conforms to the prompt word description based on the prompt words and a pre-trained Stable Diffusion model; and storing the image data according to predetermined rules; wherein the pre-trained Stable Diffusion model is obtained by training the Stable Diffusion model on a training dataset, the training dataset including: pre-acquired device images and descriptive prompt words corresponding to the device images.

[0010] Optionally, in some embodiments of this application, the step of generating image data that conforms to the description of the prompt word based on the prompt word and the pre-trained StableDiffusion model includes: processing the prompt word using the pre-trained StableDiffusion model to generate image data that conforms to the description of the prompt word.

[0011] Optionally, in some embodiments of this application, before processing the prompt word using a pre-trained Stable Diffusion model, the method further includes: receiving a style instruction triggered by a user, determining a LoRA style adapter corresponding to the style instruction; loading the LoRA style adapter into the pre-trained Stable Diffusion model; and using the Stable Diffusion model with the LoRA style adapter loaded to process the prompt word, generating image data that conforms to the prompt word description and has style features corresponding to the specified style instruction.

[0012] Optionally, in some embodiments of this application, before processing the prompt words using a pre-trained Stable Diffusion model, the method further includes: receiving multiple style instructions triggered by the user, and determining a LoRA style adapter corresponding to each style instruction; loading each LoRA style adapter into the pre-trained Stable Diffusion model; processing the prompt words using the Stable Diffusion model loaded with each LoRA style adapter, and generating image data that conforms to the prompt word description and integrates the style features corresponding to each style instruction, wherein the contribution of each LoRA style adapter in the process of generating image data is controlled by its corresponding pre-set weight.

[0013] Optionally, in some embodiments of this application, after storing the image data according to a predetermined rule, the method further includes: upon receiving an operation instruction from a user to drag the stored image data to a canvas area and perform a connection operation on the image data, generating a corresponding flowchart in response to the operation instruction.

[0014] Optionally, in some embodiments of this application, when the state of the device corresponding to the image data in the flowchart cannot be obtained, the method further includes: when the state of the device corresponding to the image data in the flowchart is obtained again, determining whether the current state of the device is consistent with the state name represented by the corresponding image data in the flowchart; if they are inconsistent, determining that the state of the device corresponding to the image data in the flowchart has changed.

[0015] Optionally, in some embodiments of this application, when it is determined that the state of the device corresponding to any image data in the flowchart has changed, the method further includes: replacing the image data corresponding to the device in the flowchart with new image data; wherein, the new image data is generated by generating corresponding prompt words according to the state name of the device after the state change and the device type using a pre-set prompt word generation rule, and generating image data based on the prompt words and a pre-trained Stable Diffusion model.

[0016] Optionally, in some embodiments of this application, the LoRA style adapter is obtained by fine-tuning the low-rank adaptation parameters of the Stable Diffusion model based on a style training dataset. The style training dataset includes image data with style features, descriptive cue words corresponding to the image data, and style labels for identifying style categories.

[0017] Secondly, embodiments of this application also provide a server, including: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the interactive construction and device state linkage method of the flowchart described in the first aspect.

[0018] Thirdly, embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it executes the interactive construction and device state linkage method of the flowchart described in the first aspect.

[0019] (III) Beneficial Effects

[0020] The interactive flowchart construction and device status linkage method, server, and storage medium provided in this application introduce a prompt word generation mechanism based on device type and status name, and combine it with a pre-trained image generation model to automatically generate image content. It can dynamically generate device graphics that conform to semantic descriptions based on user input, reducing reliance on manual configuration and pre-set icon libraries. At the same time, by storing the generated image data according to predetermined rules, it achieves effective association and linkage between device status and graphic expression. This enables the rapid construction of consistent flowcharts even in scenarios with diverse device types and frequent status changes, thereby improving the automation level of flowchart construction, the flexibility of graphic expression, and the accuracy and intuitiveness of device status visualization. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an interactive construction and device status linkage method according to an embodiment of this application;

[0022] Figure 2 A flowchart illustrating the process of generating image data having style features corresponding to a specified style instruction, according to one embodiment of this application;

[0023] Figure 3 This is a flowchart illustrating the process of generating image data that integrates style features corresponding to various style instructions according to one embodiment of this application;

[0024] Figure 4 This is a schematic diagram of a server structure according to an embodiment of this application. Detailed Implementation

[0025] To better explain and facilitate understanding of this application, the following detailed description of the application is provided in conjunction with the accompanying drawings and specific embodiments.

[0026] In related technologies, the construction of flowcharts and visualization of device status can be mainly categorized into three types:

[0027] The first type is the flowchart construction scheme based on manual configuration and icon library. In this scheme, icon resources corresponding to different device types and their states are pre-designed. When constructing a flowchart, users manually select the corresponding device graphics from the icon library and combine them with status identifiers for splicing and annotation, so as to achieve the visual expression of the flowchart. However, this scheme highly depends on manual operations and pre-set resources. When the types of devices or states increase, it is necessary to continuously expand the icon library, resulting in relatively high maintenance costs. At the same time, the efficiency of manual configuration in complex industrial scenarios is low, and it is difficult to ensure the consistency of the configuration results of different personnel.

[0028] The second type is the automatic flowchart generation scheme based on templates or rules. In this scheme, flowchart templates or device state mapping rules are predefined, and corresponding graphic elements are automatically filled according to the device type and status information input by the user, so as to reduce manual drawing work. Although this method improves the efficiency of flowchart generation to a certain extent, its essence still depends on fixed templates or rule libraries, with poor flexibility and difficulty in covering the long-tail scenarios of device type and state combinations. When new devices or new state expression requirements appear, it is necessary to redesign or expand the rule system, and the system adaptability is limited.

[0029] The third type is the visual generation scheme based on image generation models. This scheme uses generative models to generate image content according to input descriptions to achieve relatively flexible visual expressions. However, existing schemes usually lack a structured prompt word construction mechanism for device type and state semantics, and the quality of prompt words highly depends on manual writing, resulting in unstable semantic expressions. At the same time, the model training data mostly comes from general image datasets, lacking targeted training related to industrial devices and their state expressions, making the generated results insufficient in device consistency and state recognizability, and lacking an effective data association and storage mechanism between the generated images and device states, which is not conducive to subsequent linkage calls and management.

[0030] Therefore, the interactive construction of flowcharts and the device state linkage method, server, and storage medium provided by the embodiments of this application, based on the user's input of device types and state names, automatically construct structured prompt words by using preset prompt word generation rules, and combine with a pre-trained StableDiffusion model trained based on device images and corresponding descriptive prompt words to generate image data, realizing the automatic mapping from device state semantics to image expressions. At the same time, the generated image data is stored according to a predetermined rule, establishing a traceable association relationship between the device state and the image data, so as to achieve the automatic generation of flowcharts and device state linkage visualization without a large amount of manual configuration of icon resources, not only improving the automation degree and flexibility of flowchart construction, but also enhancing the consistency and scalability of device state expressions.

[0031] To better understand the above technical solutions, exemplary embodiments of this application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application can be understood more clearly and thoroughly, and that the scope of this application can be fully conveyed to those skilled in the art.

[0032] Figure 1 This is a flowchart illustrating an interactive construction and device status linkage method according to an embodiment of this application. Figure 1 As shown, the interactive construction and device status linkage method of this flowchart includes:

[0033] Receive the device type and status name input by the user, and generate corresponding prompt words according to the device type and status name using a pre-set prompt word generation rule;

[0034] Based on the prompt words and the pre-trained Stable Diffusion model, image data that conforms to the description of the prompt words is generated, and the image data is stored according to predetermined rules;

[0035] The pre-trained Stable Diffusion model is obtained by training the Stable Diffusion model on a training dataset, which includes: pre-acquired device images and descriptive prompts corresponding to the device images.

[0036] For example, in some embodiments, the interactive flowchart construction and equipment status linkage method described in this application can be applied to the operation monitoring and visualization of rail transit equipment. For instance, in an urban rail transit system, a user inputs the equipment type "track signal" or "train traction system" on the platform interface, and inputs the status name "normal operation," "speed-limited operation," or "fault alarm." First, based on pre-set prompt word generation rules, the above input is semantically structured to generate corresponding prompt words, such as "flowchart icon of the rail transit signal in normal operation" or "visual graphic of the alarm status of the train traction system malfunction," thereby mapping the equipment type and operating status into a text description with clear semantic expression.

[0037] Subsequently, the prompt word is input into a pre-trained Stable Diffusion model, which generates corresponding image data based on the learned semantic features of rail transit equipment images. For example, for "normal signal operation status," the model can generate a standardized graphic representation of a green indicator light on and the signal open; for "signal malfunction status," it generates a red warning sign or anomaly prompt graphic; for "train traction system speed-limited operation status," it generates a graphic with a speed limit sign or a power reduction visual symbol, thus achieving a significant visual distinction between different operating states. The generated image data is stored according to predetermined rules, such as being indexed according to the structure of "equipment type-line number-status type-time information," so that it can be quickly retrieved in subsequent flowchart construction or status linkage display, realizing dynamic association management between equipment status and image resources.

[0038] Furthermore, the pre-trained Stable Diffusion model in this application is trained on a dedicated training dataset for the rail transit domain. This dataset includes pre-collected images of rail transit equipment and their corresponding descriptive cues. For example, it includes "image of subway turnout equipment in a switching state" and its corresponding cue "visual representation of rail transit turnout switching track direction," and "image of train braking system operating normally" and its corresponding cue "graphical representation of train braking system in a stable braking state." Through this domain-oriented training, the model can more accurately learn the mapping relationship between the structural features of rail transit equipment and the semantics of its operating state.

[0039] To illustrate further, consider a user inputting "Metro Fan" as the device type and "Fault Status" as the status name. The prompt word engineering engine structures the input information according to built-in templated prompt word generation rules. Positive prompt word templates might be something like "{Device Type}, {Status Description}, Technical Primitive Symbols, Industrial Design Style, Vector Style, White Background, High Definition, Clear Edges". The generated positive prompt words could specifically be "Metro Fan, Fault Status, Red Flashing Alarm Icon, Technical Graphic Symbols, Industrial Design Style, Vector Style, White Background, High Quality, Clear Edges". Simultaneously, negative prompt words are generated, such as "Low Resolution, Text, Watermark, Blur, Distortion, Excessive Shadows, Non-Vector Style", to constrain the quality of the generated image and prevent image effects that do not meet engineering drawing requirements. Then, the positive and negative prompt words are input into a pre-trained image generation model, and call request parameters are constructed, such as setting the sampling steps, prompt word guidance intensity, and image output size, and sent to the image generation server via a network request. The server-side uses a diffusion denoising mechanism to iteratively optimize noise, ultimately generating image data that conforms to the semantic description of the prompt words. For example, in the scenario of "subway fan failure status," an industrial-style vector icon with a red alarm indicator can be generated to visually represent the abnormal state of the equipment; while in normal operation, a standard equipment graphic with a green or stable indicator status is generated, thus achieving visual differentiation between different states. Finally, the generated image data is structured and stored according to predetermined rules. For example, the storage path is determined based on the unique identifier of the equipment, and the image is stored in the corresponding directory with the status name as the filename, such as storing the image corresponding to "failure status" as "failure status image.png". If the target directory does not exist, the corresponding folder structure is automatically created, thus forming a mapping relationship between equipment and status, facilitating subsequent flowchart construction and dynamic retrieval.

[0040] This application's embodiments introduce a prompt word generation mechanism based on the type and operating status of rail transit equipment. This enables the flowchart image generation process to directly respond to semantic input from the professional domain, reducing the workload of manually drawing rail transit equipment graphics and status labels, and improving the automation level of flowchart construction. Simultaneously, by utilizing a generation model trained on rail transit equipment images and descriptive prompt words, the generated images can more accurately reflect the status characteristics of key equipment such as signals, switches, and traction systems, enhancing the consistency and recognizability of the flowchart in professional scenarios. Furthermore, by structurally storing the generated images, dynamic association and traceable management of equipment status and image resources are achieved, thereby improving the visualization and linkage capabilities of the rail transit system in complex operating environments.

[0041] Optionally, in some embodiments of this application, the step of generating image data that conforms to the description of the prompt word based on the prompt word and the pre-trained StableDiffusion model includes: processing the prompt word using the pre-trained StableDiffusion model to generate image data that conforms to the description of the prompt word.

[0042] In a specific example, within a rail transit monitoring system, the user inputs the device type as "metro signal" and the status name as "fault status." A structured prompt word generation mechanism generates prompt words such as "metro signal, fault status, red alarm indicator, technical primitive symbol, industrial design style, vector style, white background, clear edges." These prompt words are then fed as semantic input into a pre-trained Stable Diffusion model. The pre-trained Stable Diffusion model first semantically encodes the prompt words, mapping the text information into a latent semantic space. Then, it progressively performs diffusion denoising within this latent space. Taking "metro signal fault status" as an example, the model gradually strengthens semantic features such as "red alarm," "abnormal status indication," and "industrial vector icon" during multiple iterations of denoising, while suppressing irrelevant noise information. Ultimately, it generates an image of the device status with a clear semantic meaning. For instance, the generated result might show a clearly identifiable main structure of the signal, overlaid with flashing red or alarm symbols, to intuitively express that the device is in an abnormal state.

[0043] In another example, when the input is "normal operation status of subway signal", the model will generate an image with green indicator lights or normal release signs as the main visual features, thereby achieving a clear distinction between different states at the visual level and making the expression of equipment status in the flowchart more intuitive and consistent.

[0044] This application embodiment directly uses a pre-trained Stable Diffusion model to process prompt words, enabling device type and state semantics to be automatically mapped to corresponding image representations. This reduces the manual drawing or icon selection steps required in traditional flowchart construction and improves the automation level of flowchart generation.

[0045] Optionally, such as Figure 2 As shown, in some embodiments of this application, before processing the prompt words using a pre-trained StableDiffusion model, the method further includes:

[0046] Receive a style command triggered by the user and determine the LoRA style adapter corresponding to the style command;

[0047] Load the LoRA style adapter into the pre-trained Stable Diffusion model;

[0048] The Stable Diffusion model, after loading the LoRA style adapter, is used to process the prompt words to generate image data that conforms to the prompt word description and has style features corresponding to the specified style instruction.

[0049] In detail, within the rail transit operation and maintenance visualization system, after a user inputs the equipment type "metro fan" and the status "fault status," the system generates basic prompts such as "metro fan, fault status, red alarm sign, industrial design style, vector icon, white background." The user then triggers a style command, such as selecting "minimalist wireframe style" or "high-contrast industrial alarm style." Upon receiving this style command, the system first determines the corresponding LoRA style adapter from the style library. For example, when the user selects "minimalist wireframe style," the system matches a pre-trained "wireframe style LoRA adapter"; when the user selects "high-contrast industrial alarm style," it matches the corresponding "industrial alarm style LoRA adapter." These LoRA style adapters are specifically trained for different visual styles to inject specific style features into the basic model generation process. Subsequently, the selected LoRA style adapter is loaded into the pre-trained Stable Diffusion model, enabling the basic model to maintain its semantic understanding of the equipment while possessing the ability to generate images in specific styles. For example, after loading the "Minimalist Wireframe Style" adapter, the model-generated "Metro Fan Fault Status Image" will primarily use simplified lines and outlines, retaining only key structural information. However, after loading the "High-Contrast Industrial Alarm Style" adapter, it generates an image with a strong red warning effect and high-contrast visual impact, thus enhancing the warning message of the fault status. Finally, the Stable Diffusion model, loaded with the LoRA style adapter, processes the warning words to generate image data that both conforms to the semantic description of the equipment status and possesses specified style characteristics, thereby achieving controllable expression of the same equipment status under different visual styles.

[0050] This application's embodiments introduce style instructions and the LoRA style adapter mechanism, enabling the flowchart image generation process to move beyond a single fixed style and flexibly switch or enhance different visual expression effects according to user needs. Simultaneously, by efficiently loading the style adapter into the pre-trained model with parameters, style transfer can be achieved without retraining the base model, significantly reducing model training and deployment costs. Furthermore, this solution achieves controllable changes in image style while maintaining semantic consistency of device status, giving flowcharts stronger adaptability and expressive power in different application scenarios (such as operation and maintenance monitoring, presentation and reporting, or fault warning).

[0051] Optionally, see Figure 3 In some embodiments of this application, before processing the prompt words using a pre-trained Stable Diffusion model, the method further includes:

[0052] Receive multiple style commands triggered by the user, and determine the LoRA style adapter corresponding to each style command;

[0053] Load each LoRA style adapter into the pre-trained Stable Diffusion model;

[0054] The prompt words are processed using a Stable Diffusion model loaded with each LoRA style adapter to generate image data that conforms to the prompt word description and incorporates the style features corresponding to each style instruction. The contribution of each LoRA style adapter in the image data generation process is controlled by its corresponding pre-set weight.

[0055] Specifically, in the integrated monitoring platform for rail transit, after the user inputs the equipment type "metro fan" and the status "fault status," the platform generates basic prompts: "metro fan, fault status, red alarm icon, industrial vector icon, white background." At this point, the user can further trigger multiple style commands, such as simultaneously selecting "minimalist wireframe style" and "high-contrast alarm style," to meet the different visual requirements of "maintenance interface display" and "emergency alarm display," respectively. Upon receiving multiple style commands, the corresponding LoRA style adapters are determined. For example, the "minimalist wireframe style" is mapped to the first LoRA style adapter, and the "high-contrast alarm style" is mapped to the second LoRA style adapter. Subsequently, these multiple LoRA style adapters are simultaneously loaded into a pre-trained Stable Diffusion model, enabling the base model to maintain its semantic understanding capabilities while possessing the ability to generate multiple style features through superposition. During image generation, the contribution of different LoRA style adapters is controlled according to pre-set weights. For example, when the weight of "minimalist wireframe style" is set to 0.7 and the weight of "high-contrast alarm style" is set to 0.3, the "subway fan fault status image" generated by the model will primarily use a wireframe structure, while overlaying some red high-contrast alarm elements, thus forming a fusion style image that maintains structural clarity while also providing a certain level of warning effect. If the weight ratio is adjusted, for example, to 0.4 for the wireframe style and 0.6 for the alarm style, the generated result will be more inclined to enhance the visual impact of the fault warning, thereby meeting the display needs of different business scenarios. Finally, the Stable Diffusion model, which loads multiple LoRA style adapters, processes the prompt words to generate image data that integrates multiple style features and conforms to the semantic description of the device status, achieving adaptive presentation of the same device status under different visual expression requirements.

[0056] This application embodiment introduces a joint loading mechanism of multiple style instructions and corresponding LoRA style adapters, enabling the flowchart image generation process to simultaneously integrate multiple visual style features, breaking through the limitations of single style expression. At the same time, by introducing a weight control mechanism, the influence of different style adapters in the generation process can be adjusted and controlled, making the image style combination configurable and interpretable, thereby meeting the diverse visual needs of rail transit systems in different scenarios such as operation and maintenance monitoring, scheduling display, and emergency command.

[0057] Optionally, in some embodiments of this application, after storing the image data according to predetermined rules, the method further includes:

[0058] Upon receiving an operation instruction from a user to drag and drop stored image data onto the canvas area and perform a connection operation on the image data, the system generates a corresponding flowchart in response to the operation instruction.

[0059] For example, in a rail transit operation and maintenance visualization system, the system has already stored the generated image data according to equipment type and status, such as "Metro Fan - Normal Status Icon," "Metro Fan - Fault Status Icon," and "Metro Signal - Release Status Icon," and manages them in a structured directory. At this point, the user enters the flowchart editing interface, selects the "Metro Fan - Fault Status Icon" from the image resource list, and drags it to a specified position in the canvas area to represent a specific equipment node. After the image is dragged to the canvas area, the user further performs connection operations on the image, such as dragging a line from the "Metro Fan" node to the "Metro Signal" node, to represent the linkage path between devices in terms of operational logic or control relationships. Upon detecting this operation command, the system automatically identifies the node relationship corresponding to the dragging action and, based on preset flowchart generation rules, structurally combines the image nodes and connection relationships of each equipment, thereby generating the corresponding flowchart in real time. For example, in a subway ventilation and signal linkage scenario, the user can drag the "fan operation status icon" into the canvas and establish a connection with the "signal operation status icon" to automatically generate a flowchart structure that shows the impact of the ventilation system operation status on the signal control logic, so that the equipment status and control relationship can be presented intuitively in a visual way.

[0060] Based on the above technical solutions, this application embodiment introduces an image drag-and-drop and connection interaction mechanism, transforming the flowchart construction method from the traditional static configuration mode to a visual interactive construction mode, significantly reducing the complexity of user operations. At the same time, by using stored device status images as basic flowchart construction units, a direct mapping relationship is formed between device status representation and flowchart nodes, improving the intuitiveness and consistency of flowchart construction. Furthermore, by automatically recognizing and structuring drag-and-drop and connection operations, the ability to generate and dynamically update flowcharts in real time is achieved.

[0061] Optionally, in some embodiments of this application, when the state of the device corresponding to the image data in the flowchart cannot be obtained, the method further includes:

[0062] When the state of the device corresponding to the image data in the flowchart is re-acquired, it is determined whether the current state of the device is consistent with the state name represented by the corresponding image data in the flowchart. If they are inconsistent, it is determined that the state of the device corresponding to the image data in the flowchart has changed.

[0063] For example, in a rail transit integrated monitoring system, the flowchart has already generated and displayed equipment status images such as "Metro Fan (Normal Status)" and "Metro Signal (Release Status)," and is linked with the real-time monitoring system for display. However, during a certain period, due to a communication link interruption or a short-term failure of the data acquisition module, it is impossible to continuously acquire real-time status data for some equipment. At this time, the flowchart still maintains an image display based on historical status. When communication is restored or data acquisition returns to normal, the system re-acquires the current status information corresponding to each device in the flowchart. For example, for equipment that was originally displayed as "Metro Fan in Normal Operation Status" in the flowchart, the system re-acquires its current status as "Fault Status"; for "Metro Signal Release Status," the re-acquired result is "Restricted Operation Status."

[0064] At this point, the system will automatically compare and determine the current status of the reacquired equipment with the status name represented by the image data in the flowchart. For example, if the image corresponding to the "Metro Fan" node in the flowchart is still a green operating status indicator, but the current acquisition status has changed to a red fault status, the system will determine that the equipment status has changed. Similarly, when the current status of the signal is inconsistent with the release status shown in the flowchart, it will also be determined that its status has been updated.

[0065] After determining that the device status has changed, the system can further trigger the flowchart linkage update mechanism, such as automatically replacing the corresponding device image, updating the status indicator color, or readjusting the status connection relationship in the flowchart, so that the flowchart always remains consistent with the actual operating status of the device.

[0066] This application embodiment retains the existing state display of the flowchart in the event of interruption or delay in data acquisition, and performs a consistency comparison between the current state of the device and the state in the flowchart after data recovery, thereby achieving the ability to automatically identify changes in the device state. This avoids frequent erroneous updates or information loss of the flowchart due to short-term data loss, and improves the accuracy and reliability of state change identification.

[0067] Optionally, in some embodiments of this application, when it is determined that the state of the device corresponding to any image data in the flowchart has changed, the method further includes:

[0068] Replace the image data corresponding to the device in the flowchart with the new image data;

[0069] The new image data is generated by using pre-defined prompt word generation rules to generate corresponding prompt words based on the device's status name after the status change and the device type, and then generating image data based on the prompt words and the pre-trained StableDiffusion model.

[0070] For example, in a rail transit integrated monitoring system, the image data corresponding to the "Metro Fan" node in the flowchart was originally represented as "normal operation status" and displayed with a green operation icon. During subsequent status comparison, the system identifies that the current status of the equipment has changed from "normal operation" to "fault status," thus determining that the equipment status has changed. In this case, firstly, based on the equipment type "Metro Fan" and the changed status name "fault status," the system regenerates the corresponding prompt words according to pre-set prompt word generation rules. For example, it generates structured descriptive information such as "Metro Fan, Fault Status, Red Alarm Icon, Industrial Vector Icon, White Background, High Definition, Clear Edges," ensuring that the new prompt words accurately express the semantic features of the equipment's current status. Subsequently, the new prompt words are input into a pre-trained StableDiffusion model for image generation. Based on the learned semantic mapping relationship between rail transit equipment images and status, the model generates new image data matching "Metro Fan Fault Status," such as generating an equipment icon with a flashing red alarm icon or anomaly indicator to visually express the abnormal operating status of the equipment. Finally, the new image data is used to replace the original image data in the flowchart, so that the "Metro Fan" node in the flowchart is automatically updated from a green normal state image to a red fault state image, thereby achieving dynamic synchronous updating of the flowchart.

[0071] This application embodiment automatically updates and replaces the device status in the flowchart by regenerating the corresponding prompt words and driving the image generation model to generate new image data after detecting a change in device status. This avoids the tedious operation of manually modifying flowchart nodes. At the same time, this method can dynamically generate corresponding images based on device type and the latest status semantics, ensuring that the updated images always remain consistent with the actual operating status of the device, thus improving the real-time performance and accuracy of the flowchart. Furthermore, by using a pre-trained image generation model to reconstruct the status images, the system can complete dynamic generation without pre-storing all status icon resources, significantly improving the system's flexibility and adaptability in the complex operating environment of rail transit.

[0072] Optionally, in some embodiments of this application, the LoRA style adapter is obtained by fine-tuning the low-rank adaptation parameters of the Stable Diffusion model based on a style training dataset. The style training dataset includes image data with style features, descriptive cue words corresponding to the image data, and style labels for identifying style categories.

[0073] In detail, when constructing a style model for visualizing rail transit flowcharts, a style training dataset is first established. This dataset includes equipment images with different visual style characteristics, along with their corresponding descriptive cues and style labels. The image data can include images of equipment such as "metro fan icons," "signal icons," and "switches" in different visual representations, such as minimalist line styles, industrial alarm styles, and high-contrast display styles.

[0074] Meanwhile, each image data point corresponds to a descriptive cue word. For example, for a subway fan image with a minimalist line style, the corresponding cue words could be "subway fan equipment icon, minimalist line expression, low-detail structure, flowchart style, white background, vector graphics"; for a signal light image with an industrial warning style, the corresponding cue words could be "track signal light, red warning status expression, high-contrast industrial style, warning visual elements, vector icon". In addition, the dataset includes style labels to identify style categories, such as "minimalist line style", "industrial warning style", and "standard flowchart style", used to classify and label different style samples. Based on this, by fine-tuning the low-rank adaptation parameters of the Stable Diffusion model, the image features, cue word semantics, and style labels from the style training dataset are mapped to the model's style generation capabilities, thereby training the corresponding LoRA style adapters. For example, the "minimalist line style LoRA adapter" enables the model to generate flowchart icons that are more inclined towards simplified structural expressions, while the "industrial warning style LoRA adapter" strengthens the red warning and high-contrast visual expression capabilities.

[0075] In some other embodiments of this application, the interactive flowchart construction and equipment status linkage method described in this application can be applied to the operation monitoring and graphical operation and maintenance display scenarios of rail transit equipment, thereby realizing the automatic generation of equipment status elements, interactive construction of flowcharts, and real-time linkage update of equipment status.

[0076] Specifically, in the integrated monitoring system for rail transit, the user first inputs the equipment type and status name, for example, the equipment type is "metro fan" and the status name is "fault status". After receiving this input, the system performs structured semantic processing on the equipment type and status name according to pre-set prompt word generation rules, thereby generating prompt word information for the image generation model input. For example, it can generate positive prompt words such as "metro fan, fault status, red alarm sign, industrial vector icon, white background, high definition, clear edges", and negative prompt words to constrain image quality, such as "low resolution, text, watermark, blur, distortion, excessive shadows, non-vector style".

[0077] After generating the prompt words, the prompt words are input into a pre-trained Stable Diffusion model for image generation processing. This model is pre-trained based on a training dataset constructed from images of rail transit equipment and descriptive prompt words. This training dataset includes pre-collected images of rail transit equipment and their corresponding descriptive prompt words, enabling the model to learn the structural features of rail transit equipment and the semantics of its operational status.

[0078] During image generation, the model first encodes the prompt words into latent semantic vectors and maps them into the latent space for iterative diffusion denoising. For example, given the semantic input of "subway fan malfunction status," the model continuously strengthens semantic features such as "red alarm," "abnormal indication," and "industrial equipment icon" during the progressive denoising process, while suppressing irrelevant noise information, ultimately generating image data that conforms to the semantic description. This image can be specifically represented as a clearly identifiable fan structure, overlaid with a red alarm label or flashing abnormal symbol, thus intuitively expressing that the equipment is in a malfunction state.

[0079] The generated image data is then stored and managed in a structured manner according to predetermined rules. For example, storage paths are constructed based on device type, and filenames are generated by combining the status name. Images of different devices and different statuses are stored in corresponding directory structures, thereby forming a stable mapping relationship between device type and status images, which facilitates the subsequent construction and retrieval of flowcharts.

[0080] Furthermore, during the flowchart construction process, the system uses the previously stored equipment status images as flowchart node resources. Users can add corresponding equipment image nodes to the canvas area by dragging and dropping, and construct logical relationships between devices by connecting them, thereby forming a flowchart of rail transit equipment operation. For example, in a scenario where the subway ventilation system and signal control system are linked, users can drag the "Subway Fan (Fault Status)" image node into the canvas and establish a connection with the "Subway Signal (Operating Status)" node, thereby expressing the potential impact of ventilation equipment malfunctions on signal control logic and achieving a visual representation of equipment status and control topology.

[0081] Regarding equipment status linkage, when the system is unable to continuously acquire real-time status data of the equipment in the flowchart, after restoring data acquisition capabilities, the system re-acquires the current operating status of the corresponding equipment and compares it with the status name represented by the image data in the flowchart for consistency judgment. For example, if the flowchart displays "Metro fan in normal operation status", but the actual status acquired is "fault status", the system determines that the equipment status has changed.

[0082] When a change in device status is detected, the system will generate new prompts based on the device type and the updated status name, following the prompt generation rules. These new prompts will then be input into the pre-trained Stable Diffusion model for image generation, resulting in new device status image data. For example, the original green normal operation image can be replaced with a fault status image with a red alarm indicator, and this new image data can replace the corresponding original image nodes in the flowchart, thus achieving dynamic updating of the flowchart.

[0083] In addition, during the interactive flowchart construction process, the system allows users to drag and drop device image nodes to the canvas area and connect them. When this operation is detected, the system automatically generates the corresponding flowchart topology based on the connection relationship between the image nodes, thereby realizing the automated construction and structured expression of the flowchart.

[0084] Through the above technical solutions, the embodiments of this application can realize a direct mapping relationship between equipment status semantics and image generation models, enabling the operation status of rail transit equipment to be automatically generated in image form and participate in the flowchart construction process, significantly reducing the workload of manually drawing flowcharts and status icons. At the same time, by combining equipment status changes with an automatic image update mechanism, the flowchart can reflect changes in equipment operation status in real time, improving the accuracy and real-time performance of system status expression. In addition, by pre-training the generation model based on a training dataset in the rail transit field, the generated images can accurately reflect the structural features and operation semantics of rail transit equipment, thereby improving the consistency and recognizability of flowcharts in professional operation and maintenance scenarios, and enhancing the system's visualization and linkage capabilities in complex operating environments.

[0085] In one specific embodiment, this application enhances the device feature representation capability of the Stable Diffusion model through the embedding vector training method, so that the generated rail transit equipment primitives can not only have state semantic representation capability, but also accurately reflect the structural features and appearance details of the equipment.

[0086] Specifically, the first step is to collect screenshots or design drafts of the actual monitoring interfaces of various equipment from the rail transit monitoring system as training data. For example, images of different models of subway ventilation fans can be collected, including appearance diagrams of equipment with different structural forms such as axial flow fans and centrifugal fans. At the same time, standard design drawings or operation interface screenshots of key rail transit equipment such as signal lights, turnout equipment, and traction power supply equipment can also be collected, thereby forming an image data set covering multiple equipment types.

[0087] During the initial training phase, the existing basic embedding vectors in the Stable Diffusion model are used as initial representations. For example, embedding vectors corresponding to the common concept of "fan" or similar equipment in the model are selected to give the model a preliminary semantic foundation for equipment. Based on this, corresponding training sample pairs are constructed for each equipment image, with each image accompanied by a descriptive cue word. For example, for an image of an axial fan, the corresponding cue word could be set as "axial flow ventilation fan, silver metal blade structure, industrial equipment appearance, precision mechanical structure expression"; for an image of a certain type of signal machine, the corresponding cue word could be "rail transit signal machine equipment, red and green indicator light structure, standardized industrial design appearance." A semantic mapping foundation is established through the one-to-one correspondence between images and text.

[0088] During the embedding vector optimization process, the U-Net network structure parameters and variational autoencoder parameters in the Stable Diffusion model are kept constant. Only the text embedding vectors are updated, allowing them to gradually learn the fine-grained feature representation capabilities of rail transit equipment. During training, noise is added to the equipment images and diffusion modeling is performed. The prediction error of the model against the noise is used as the optimization criterion, enabling the embedding vectors to be continuously adjusted to more accurately match the relationship between image content and text semantics, thereby achieving alignment between equipment structural features and semantic descriptions.

[0089] During training, the effectiveness of the current embedding vector is periodically verified. This involves generating a corresponding device image based on the current embedding vector and comparing it with the original training image. When the semantic similarity between the generated image and the real device image reaches a preset threshold, such as a similarity index of 0.85 or higher, the embedding vector is considered to have converged, and the training process is stopped. This ensures that the embedding vector has expressive power while avoiding overfitting.

[0090] After training, the obtained embedding vectors are persistently saved and mapped to the corresponding device types. For example, the embedding vector corresponding to "axial fan" is bound to the device type and stored so that it can be quickly retrieved in subsequent generation processes.

[0091] In the actual primitive generation process, when it is necessary to generate an image of a specific equipment type, such as generating a "subway fan fault status diagram," the corresponding embedding vector is first obtained according to the equipment type and injected into the prompt word encoding process. This ensures that the prompt word carries the structural feature information of the equipment before entering the Stable Diffusion model. Subsequently, during the diffusion denoising generation process, the model can simultaneously combine state semantic information and equipment structural embedding information to generate image data that both conforms to the equipment state description and has the true appearance features of the equipment.

[0092] Through the above methods, this application can achieve accurate learning and representation of fine-grained features of rail transit equipment without changing the main structure of the Stable Diffusion model. This allows the generated images to maintain semantic consistency while more closely resembling the appearance of real engineering equipment, thereby improving the professionalism and recognizability of primitives. Simultaneously, this method avoids full retraining of the basic model, significantly reducing training costs and enhancing the rapid adaptation and scalability between different equipment types.

[0093] In another specific embodiment of this application, after generating the corresponding flowchart, the system first abstracts the flowchart into a device association topology based on the logical relationship represented by the connection between each device node in the flowchart. Each device node corresponds to a specific device entity in the rail transit system, and the connection relationship between the nodes is used to represent logical associations such as power supply relationship, signal control relationship, ventilation linkage relationship or operation dependency relationship.

[0094] Based on this, when the state of any device node changes, for example from "normal operation state" to "fault state" or "restricted operation state", the system performs propagation calculations on the state change based on the device association topology to determine the set of directly associated nodes and the set of indirectly associated nodes of that device node in the topology. Directly associated nodes represent device nodes that have a direct control or dependency relationship with the current device, while indirectly associated nodes represent device nodes that are indirectly affected through intermediate nodes.

[0095] After automatically deducing the status linkage changes of other device nodes that are related to it, the system further performs hierarchical status reasoning processing on each related device node. That is, according to the association strength or topology level, different status influence weights are assigned to different device nodes, and the status change trend of each related device node is calculated based on the weights to determine whether it is necessary to synchronize the status update or issue an early warning.

[0096] Subsequently, based on the derived linkage change results, the system performs linkage identification processing on the corresponding device node status in the flowchart. For example, it marks potentially affected device nodes as "potentially abnormal state" or "affected state" and distinguishes them in the flowchart through visual identification to prompt the user that the device node may be affected by changes in the status of upstream or related devices.

[0097] Furthermore, for device nodes whose status needs to be updated after deduction, the system will trigger the corresponding device status re-acquisition and verification process, and combine real-time monitoring data to determine whether its current operating status has undergone substantial changes; if a change is confirmed, the image data and status identifier of the corresponding node in the flowchart will be updated synchronously to keep it consistent with the actual device operating status; if no actual change has occurred, only the linkage prompt status will be retained without modifying the basic status expression, thereby avoiding erroneous updates.

[0098] Through the above methods, this application can automatically realize cross-node propagation and linkage deduction of equipment status based on the flowchart structure. This not only improves the identification range of equipment status changes, but also reflects potential impact paths in advance. This makes the flowchart not only used for static display of equipment status, but also has the ability to dynamically deduce the operation relationship of the rail transit system.

[0099] In one specific implementation, when performing propagation calculations for state changes based on the device association topology, the device node experiencing the state change is first taken as the starting node. A graph traversal operation is then performed in the topology corresponding to the flowchart to determine the set of device nodes directly connected to the starting node, which is then considered as the directly associated node set. The device nodes in the directly associated node set are those directly connected to the starting node in the flowchart via first-order connections. These connections represent direct control relationships, signal interaction relationships, or operational dependencies between devices.

[0100] After determining the set of directly related nodes, the system further uses each device node in the set of directly related nodes as an extension node, and continues to recursively traverse or hierarchically expand the search along their corresponding connection relationships to determine the set of device nodes that have multi-hop path connections with the starting node, which is then used as the set of indirectly related nodes. The device nodes in the set of indirectly related nodes are those that are logically associated with the starting node indirectly through one or more intermediate device nodes, with a path length greater than one. These paths are defined by the connection structure in the flowchart and are used to characterize the indirect influence or cascading dependency relationships between devices.

[0101] After dividing the set of directly associated nodes into the set of indirectly associated nodes, the propagation range of state changes is expanded hierarchically based on different association levels, thereby realizing the step-by-step derivation of the impact range of equipment state changes.

[0102] The interactive flowchart construction and device status linkage method provided in this application receives user input of device type and status name, and generates corresponding prompt words according to pre-set prompt word generation rules based on the device type and status name. This enables standardized semantic expression of device status, thereby improving the structure and computability of device information. Based on the prompt words and a pre-trained Stable Diffusion model, image data conforming to the prompt word description is generated and stored according to predetermined rules. This achieves automatic generation and structured management of device status from image data, allowing device status to be expressed in a unified visual form, thus reducing the workload of manual primitive design. By introducing a LoRA style adapter, the generated image data can integrate style features corresponding to different style instructions, and multi-style controllable fusion is achieved through weight control, thereby improving the flexibility and adaptability of flowchart image expression. The flowchart generation method through image data drag-and-drop and connection operations transforms flowchart construction from manual drawing to interactive construction, thereby improving flowchart construction efficiency and reducing operational complexity. By re-acquiring and comparing the device status in the flowchart, the automatic identification of device status changes is achieved. By replacing the corresponding image data in the flowchart with new image data, the flowchart can maintain dynamic consistency with the actual operating status of the device.

[0103] This application calculates the propagation of state changes based on the logical relationships between device nodes in the flowchart, enabling the linkage derivation of device state changes between related nodes. This allows the flowchart to reflect the cascading influence of the system, improving the completeness and consistency of the expression of the operating status of rail transit equipment.

[0104] See Figure 4 This application also provides a server, including a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the interactive construction and device state linkage method of the flowchart described in the above embodiments.

[0105] This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the interactive construction and device state linkage method of the flowchart described in the above embodiments.

[0106] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0107] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0108] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for interactively constructing flowcharts and linking them with device status, characterized in that, include: Receive the device type and status name input by the user, and generate corresponding prompt words according to the device type and status name using a pre-set prompt word generation rule; Based on the prompt words and the pre-trained Stable Diffusion model, image data that conforms to the description of the prompt words is generated, and the image data is stored according to predetermined rules; The pre-trained Stable Diffusion model is obtained by training the Stable Diffusion model on a training dataset, which includes: pre-acquired device images and descriptive prompts corresponding to the device images; After storing the image data according to a predetermined rule, the method further includes: upon receiving an operation instruction from a user to drag the stored image data to a canvas area and perform a connection operation on the image data, generating a corresponding flowchart in response to the operation instruction; If the state of the device corresponding to the image data in the flowchart cannot be obtained, the method further includes: if the state of the device corresponding to the image data in the flowchart is obtained again, determining whether the current state of the device is consistent with the state name represented by the corresponding image data in the flowchart; if they are inconsistent, determining that the state of the device corresponding to the image data in the flowchart has changed. When it is determined that the state of the device corresponding to any image data in the flowchart has changed, the method further includes: Based on the logical relationships represented by the connections between the device nodes in the flowchart, the flowchart is abstracted into a device association topology, where each device node corresponds to a device entity, and the connection relationships between nodes are used to represent the control relationship, signal interaction relationship, operation dependency relationship or linkage relationship between devices. Based on the device association topology, the propagation calculation of state changes is performed to determine the set of directly associated nodes and the set of indirectly associated nodes corresponding to the device node whose state has changed. The set of directly associated nodes includes device nodes that have a direct connection with the device node whose state has changed, and the set of indirectly associated nodes includes device nodes that have established an association with the device node whose state has changed through one or more intermediate device nodes. Based on the association strength or topological level between each associated device node and the device node whose state has changed, a corresponding state influence weight is assigned to each associated device node, and the state change trend of each associated device node is calculated based on the state influence weight to determine whether to update the state of the corresponding associated device node or issue an early warning.

2. The interactive construction and device status linkage method for flowcharts according to claim 1, characterized in that, The step of generating image data that matches the description of the prompt words based on the prompt words and the pre-trained Stable Diffusion model includes: processing the prompt words using the pre-trained Stable Diffusion model to generate image data that matches the description of the prompt words.

3. The interactive construction and device status linkage method for flowcharts according to claim 2, characterized in that, Before processing the prompt words using a pre-trained Stable Diffusion model, the method further includes: Receive a style command triggered by the user and determine the LoRA style adapter corresponding to the style command; Load the LoRA style adapter into the pre-trained Stable Diffusion model; The Stable Diffusion model, after loading the LoRA style adapter, is used to process the prompt words to generate image data that conforms to the prompt word description and has style features corresponding to the specified style instruction.

4. The interactive construction and device status linkage method for flowcharts according to claim 2, characterized in that, Before processing the prompt words using a pre-trained Stable Diffusion model, the method further includes: Receive multiple style commands triggered by the user, and determine the LoRA style adapter corresponding to each style command; Load each LoRA style adapter into the pre-trained Stable Diffusion model; The prompt words are processed using a Stable Diffusion model loaded with each LoRA style adapter to generate image data that conforms to the prompt word description and incorporates the style features corresponding to each style instruction. The contribution of each LoRA style adapter in the image data generation process is controlled by its corresponding pre-set weight.

5. The interactive construction and device status linkage method for flowcharts according to any one of claims 3-4, characterized in that, When it is determined that the state of the device corresponding to any image data in the flowchart has changed, the method further includes: Replace the image data corresponding to the device in the flowchart with the new image data; The new image data is generated by using pre-defined prompt word generation rules to generate corresponding prompt words based on the device's status name after the status change and the device type, and then generating image data based on the prompt words and a pre-trained Stable Diffusion model.

6. The interactive construction and device status linkage method for flowcharts according to claim 5, characterized in that, The LoRA style adapter is obtained by fine-tuning the low-rank adaptation parameters of the Stable Diffusion model based on a style training dataset. The style training dataset includes image data with style features, descriptive cue words corresponding to the image data, and style labels used to identify style categories.

7. A server, characterized in that, include: A memory and a processor are interconnected and communicate with each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the interactive construction and device status linkage method of the flowchart according to any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the interactive construction and device status linkage method of the flowchart according to any one of claims 1-6.

Images