Large language model driven flow field visual analysis method and device

Abstract

The invention relates to the technical field of visual analysis, and discloses a flow field visual analysis method and device driven by a large language model, and the method comprises the steps: obtaining a to-be-analyzed flow field data set and a user instruction; performing semantic analysis on the user instruction through a large language model to obtain a structured instruction; based on the structured instruction, acquiring corresponding target data from the to-be-analyzed flow field data set through the large language model, and generating a descriptive text corresponding to the target data; based on the target data and the descriptive text, running a visual agent to obtain a visual result; and displaying the target data according to the data display mode indicated by the user instruction. According to the method, interaction based on the natural language is realized through the large language model, and the operation instruction or the text instruction of the user is converted into the structured instruction, so that the visual result and interpretation are presented in real time according to the user demand, and the user experience is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of visual analysis technology, and in particular to a flow field visual analysis method and device driven by a large language model. Background Art

[0002] Flow field data has extensive applications in visualization. The complexity of flow field data research is primarily due to its high dimensionality and multivariate nature. However, currently popular flow field visual analysis tools still suffer from poor usability and high learning curves. These tools often rely on complex query statements or parameter settings and are unable to process the user's natural language. Users not only need to possess relevant domain expertise but also must master the specific system operation procedures.

[0003] Therefore, there is an urgent need for a flow field visual analysis method that can interact with natural language and convert the user's natural language instructions into specific visualization tasks, greatly improving the user experience. Summary of the Invention

[0004] In view of this, the present application provides a flow field visual analysis method and device driven by a large language model, which can convert the user's natural language instructions into specific visualization tasks, greatly improving the user experience. The technical solution is as follows.

[0005] In a first aspect, the present invention provides a flow field visual analysis method driven by a large language model, the method comprising:

[0006] Obtain the flow field data set to be analyzed and user instructions;

[0007] Perform semantic analysis on the user's instructions through a large language model to obtain structured instructions;

[0008] Based on the structured instruction, corresponding target data is obtained from the flow field data set to be analyzed through the large language model, and descriptive text corresponding to the target data is generated;

[0009] Based on the target data and the descriptive text, running a visualization agent to obtain a visualization result;

[0010] The target data is displayed according to the data display method indicated by the user instruction.

[0011] In an optional embodiment, the semantic analysis of the user instruction by a large language model to obtain a structured instruction includes:

[0012] The user instruction is semantically analyzed through a large language model to obtain a task identifier corresponding to the user instruction; parameters of the task identifier are extracted through preset parsing rules, and structured conversion is performed to obtain a structured instruction.

[0013] In an optional implementation, the user instruction includes natural language, interface button operation instruction or mouse gesture instruction.

[0014] In an optional embodiment, obtaining corresponding target data from the flow field data set to be analyzed by using the large language model includes:

[0015] Extracting a corresponding instruction task from the structured instruction; confirming a corresponding proxy function in the large language model proxy layer based on the instruction task; and executing the instruction task through the proxy function to obtain corresponding target data from the flow field data set to be analyzed.

[0016] In an optional embodiment, generating a descriptive text corresponding to the target data includes:

[0017] The target data is analyzed by the large language model to generate descriptive text corresponding to the target data; the descriptive text includes: variable description, dimension, measurement unit or value range information.

[0018] In an optional embodiment, the target data is displayed according to the data display method indicated by the user instruction, including:

[0019] A corresponding data presentation method is extracted from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description, or a video animation; and the target data is presented according to the data presentation method.

[0020] The present invention provides a large language model driven flow field visual analysis method with the following advantages.

[0021] The large language model-driven flow field visual analysis method of the present invention first requires obtaining a flow field dataset to be analyzed and specific user instructions. User instructions can be generated by the user through natural language input, clicking an interface button, or performing a specific mouse gesture. The user instructions are then semantically analyzed using a pre-trained large language model to obtain a task identifier corresponding to the user instruction. Parameters are then extracted from the task identifier using pre-set parsing rules and then structured into a structured instruction to obtain the structured instruction corresponding to the user instruction. This structured instruction is input into the proxy layer of the large language model. The corresponding proxy function in the proxy layer executes the instruction task corresponding to the structured instruction. This instruction task then retrieves the corresponding target data from the flow field dataset to be analyzed. Simultaneously, the large language model generates a descriptive text message corresponding to the target data. This descriptive text includes variable descriptions, dimensions, units of measurement, and value range information to help users better understand the context and meaning of the target data. A visualization result is generated based on the target data and the corresponding descriptive text. Before displaying the visualization result, a visualization display method is obtained from the user instruction. This display method includes a chart, rendering, model, text description, or video animation. The target data is displayed according to the display method corresponding to the user's instructions. The user's specific instructions are recognized through a large language model to achieve natural language-based interaction. The user's operation instructions or text instructions can be converted into structured instructions to obtain the corresponding target data. The corresponding descriptive instructions are generated through the large language model, and the target data and the corresponding descriptive description are presented in real time to help users better understand the context and meaning of the target data, greatly improving the user experience.

[0022] In a second aspect, the present invention provides a flow field visual analysis device driven by a large language model, the device comprising:

[0023] The acquisition module is used to obtain the flow field data set to be analyzed and user instructions;

[0024] The instruction conversion module is used to perform semantic analysis on the user instruction through a large language model to obtain a structured instruction;

[0025] a target data acquisition module, configured to acquire corresponding target data from the flow field data set to be analyzed through the large language model based on the structured instruction, and generate descriptive text corresponding to the target data;

[0026] A result generation module is used to run a visualization agent based on the target data and the descriptive text to obtain a visualization result;

[0027] The result display module is used to display the target data according to the data display method indicated by the user instruction.

[0028] In an optional implementation, the instruction conversion module is specifically configured to:

[0029] The user instruction is semantically analyzed through a large language model to obtain a task identifier corresponding to the user instruction; parameters of the task identifier are extracted through preset parsing rules, and structured conversion is performed to obtain a structured instruction.

[0030] In an optional implementation, the target data acquisition module is specifically configured to:

[0031] Extracting a corresponding instruction task from the structured instruction; confirming a corresponding proxy function in the large language model proxy layer based on the instruction task; and executing the instruction task through the proxy function to obtain corresponding target data from the flow field data set to be analyzed.

[0032] In an optional implementation, the target data acquisition module is further configured to:

[0033] The target data is analyzed by the large language model to generate descriptive text corresponding to the target data; the descriptive text includes: variable description, dimension, measurement unit or value range information.

[0034] In an optional embodiment, the result display module is specifically used to:

[0035] A corresponding data presentation method is extracted from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description, or a video animation; and the target data is presented according to the data presentation method.

[0036] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the large language model-driven flow field visual analysis method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the large language model-driven flow field visual analysis method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0038] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for causing a computer to execute the large language model driven flow field visual analysis method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a method flow chart of a flow field visual analysis method driven by a large language model provided by an embodiment of the present invention.

[0041] Figure 2 The figure is a schematic diagram of a front-end interface of a progressive visual analysis system based on a large language model according to an exemplary embodiment.

[0042] Figure 3 The figure is a schematic diagram of a visual analysis process of a progressive visual analysis system based on a large language model according to an exemplary embodiment.

[0043] Figure 4 The figure is a schematic diagram of a back-end user instruction processing flow of a progressive visual analysis system based on a large language model according to an exemplary embodiment.

[0044] Figure 5 The figure is a schematic diagram of a plug-in visualization design of a progressive visual analysis system based on a large language model according to an exemplary embodiment.

[0045] Figure 6 The figure is a flow chart of visualization semantic enhancement of a progressive visual analysis system based on a large language model according to an exemplary embodiment.

[0046] Figure 7 The figure is a schematic diagram showing the global streamline distribution within a data set according to an exemplary embodiment.

[0047] Figure 8 The figure is a schematic diagram showing the global velocity distribution of a data set according to an exemplary embodiment.

[0048] Figure 9 It is a schematic diagram showing the velocity distribution in a local area according to an exemplary embodiment.

[0049] Figure 10 A system is shown generating a heat map and automatically describing a schematic diagram according to an exemplary embodiment.

[0050] Figure 11 It is a schematic diagram of longitudinal and latitudinal velocity distribution near the relevant sea area according to an exemplary embodiment.

[0051] Figure 12FIG. 4 is a schematic diagram showing underwater velocity distribution according to an exemplary embodiment.

[0052] Figure 13 It is a schematic diagram showing the comparison of temperature and salinity distribution in a local area according to an exemplary embodiment.

[0053] Figure 14 The figure is a schematic diagram showing the changing trends of regional average temperature and salinity according to an exemplary embodiment.

[0054] Figure 15 This is a structural diagram of a flow field visual analysis device driven by a large language model provided in an embodiment of the present application.

[0055] Figure 16 It is a structural diagram of a computer device provided by an optional embodiment of the present invention. DETAILED DESCRIPTION

[0056] To make the purpose, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of this application.

[0057] It should be understood that the "indication" mentioned in the embodiments of this application can be a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also mean that there is an association between A and B.

[0058] In the description of the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and being indicated, configuration and being configured, etc.

[0059] In an embodiment of the present application, "predefinition" can be achieved by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in a device (for example, including a terminal device and a network device). This application does not limit its specific implementation method.

[0060] Real-time analysis of large-scale flow field data faces challenges such as low efficiency and complex interactions. Streamline-based visualization provides a technical means for studying flow field data. As time spans and variable dimensions increase, the size of the collective simulation data increases exponentially. The traditional calculation-then-analysis model requires long wait times and prevents mid-process intervention. Furthermore, the large number of intermediate results creates pressure on storage and I / O.

[0061] The complexity of flow data research is primarily due to its high dimensionality and multivariate nature. Since flow data often originate from long-term observations or high-precision numerical simulations, they encompass not only multidimensional attributes such as velocity, temperature, and salinity, but can also span timescales ranging from weeks to years. In terms of spatial capacity, the intermediate data generated during computation and visualization imposes stringent requirements on data storage and I / O. Volume rendering, in particular, further exacerbates the data expansion. In practical visualization applications, the required data size is typically limited to GB or less. While directly expanding storage capacity may seem like a solution, it actually misses the core of the problem. Traditional visualization workflows typically employ a post-processing model, requiring all experiments or simulations to complete and generate all results before they can be input into the system for visualization and analysis. This results in significant delays in analytical feedback when conducting large-scale or complex computations. Analysis cannot begin until all data processing is complete, which not only slows down research progress but also misses opportunities to adjust experimental strategies, limiting the ability to dynamically explore data and identify potential issues.

[0062] With the growing demand for scientific data visualization, interactive operations enable users to more deeply analyze and explore flow field data, greatly facilitating their understanding and interpretation of complex datasets. However, currently popular flow field visual analysis tools still suffer from poor usability in practice. These interfaces often rely on complex query statements or parameter settings, which poses a major challenge for users without a technical background. Users not only need to possess relevant domain expertise but also must master the operation of specific systems, which undoubtedly increases the difficulty of use.

[0063] In recent years, natural language processing technology has experienced rapid development, especially the emergence of pre-trained models and large language models, which have provided powerful technical support for understanding and processing natural language. Integrating natural language processing and understanding technology into visual interactive interfaces is an innovative development direction.

[0064] Most current visualization systems use a fixed graphical user interface (GUI), which lacks flexibility in component layout and adjustment. When numerous visualization components are integrated into a single interface, users often have difficulty knowing where to start, which increases the burden of operation and exploration. Furthermore, fixed visualization designs hinder users from reviewing their operation history. When users switch to studying new cases, previous operations and configurations are overwritten and reset by the new operations, restricting users' ability to flexibly switch between different datasets or cases, affecting the consistency and efficiency of research. Finally, adjustments to system components (such as adding or removing chart support) can affect the layout of other components, leaving room for improvement in scalability.

[0065] Therefore, an embodiment of the present invention provides a flow field visual analysis method driven by a large language model, which uses a large language model to identify the intention of the user's specific instructions, realizes natural language-based interaction, and can convert the user's operation instructions or text instructions into structured instructions, thereby obtaining the corresponding target data, and generates corresponding descriptive descriptions through the large language model, and presents the target data and the corresponding descriptive descriptions in real time to help users better understand the background and meaning of the target data, greatly improving the user experience.

[0066] The flow field visual analysis method driven by a large language model provided by the embodiment of the present invention is as follows Figure 1 As shown, the following steps are included.

[0067] S101: Obtain a flow field dataset to be analyzed and user instructions.

[0068] Optionally, in the above steps, the flow field dataset may involve multiple technical fields, such as airflow data in the aerospace field, water flow data in the environmental engineering field, or ocean data. The user instruction may be generated by the user through natural language input, clicking an interface button, or performing a specific mouse gesture.

[0069] S102: Perform semantic analysis on the user instruction using a large language model to obtain a structured instruction.

[0070] Specifically, in step S102, semantic analysis is performed on the user instruction using a large language model to obtain a task identifier corresponding to the user instruction; parameters are extracted from the task identifier using preset parsing rules, and structured conversion is performed to obtain a structured instruction.

[0071] S103 : Based on the structured instruction, corresponding target data is acquired from the flow field data set to be analyzed through the large language model, and descriptive text corresponding to the target data is generated.

[0072] Specifically, in step S103, a corresponding instruction task is extracted from the structured instruction; based on the instruction task, a corresponding proxy function in the large language model proxy layer is confirmed; and the instruction task is executed through the proxy function to obtain corresponding target data from the flow field data set to be analyzed.

[0073] Optionally, in the above steps, the target data is analyzed by the large language model to generate descriptive text corresponding to the target data; the descriptive text includes: variable description, dimension, measurement unit or value range information.

[0074] S104: Running a visualization agent based on the target data and the descriptive text to obtain a visualization result.

[0075] Specifically, in step S104 , the visualization result includes the target data and corresponding descriptive text, which can help users quickly understand the background and meaning of the target data.

[0076] S105: Display the target data in a data display manner indicated by the user instruction.

[0077] Optionally, in the above steps, a corresponding data presentation method is extracted from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description or a video animation; and the target data is presented according to the data presentation method.

[0078] Based on the above embodiment, the embodiment of the present invention also provides a progressive visual analysis system based on a large language model. The system combines back-end processing with a front-end interface, including a visual interface, a large language model and a data set. The front-end interface is as follows: Figure 2 As shown, Figure 2 Among them, a is the management view; b1 is the conversation view content area; b2 is the conversation view input area; and c is the gallery view.

[0079] Visual analysis process such as Figure 3 As shown, the user first initiates an analysis session in the visual interface and selects the flow field dataset to be studied. The user then generates specific commands through natural language input, clicking interface buttons, or performing specific mouse gestures. These commands are transmitted to the system backend via a WebSocket connection established between the frontend and backend. This connection establishes a persistent, bidirectional communication channel, allowing for real-time, continuous data exchange.

[0080] Figure 2In the chat management view, the conversation management view is located on the left side of the visual interface. Users can create new conversations using the buttons at the top of the interface. Each conversation is automatically assigned a default name and a unique identifier. Users can edit the name by double-clicking the conversation label to distinguish different discussion topics. Conversations not only record user actions but also preserve the conversation context, facilitating understanding and continuing the conversation flow. To facilitate multi-tasking analysis, the system allows users to switch between different conversations by clicking on the conversation label. When a conversation is selected, other conversations are blocked, and their contextual information is preserved. This allows users to focus on the currently active conversation while being able to return to other conversations at any time. Below the management view, there are "Load Chat," "Save Chat," and "Delete Chat" buttons. Users can load previously saved conversations to review past conversations or continue unfinished tasks. They can also save the current conversation for future review or continued discussion, or delete unnecessary conversations to keep the list tidy.

[0081] The conversation view input area integrates the functions of uploading and parsing data files. Users select one or more required data files for upload through the "Select File" button in the input area, and multiple file selection is supported, making it convenient for users to process data in batches. After selecting the file, the user needs to click the "Upload" button to complete the data upload process. Once the data file is uploaded successfully, the system automatically performs the loading and parsing operations. This process includes identifying and reading the file content, analyzing the data structure, and extracting and displaying the basic descriptive information of the data (dimensions, fields, data ranges, etc.), so that users can quickly grasp the basic situation of the data set. The input box in the center of the area serves as the main channel for users to interact with the system. Users can use natural language to enter instructions or queries in this input box and click the arrow button to send these instructions.

[0082] The conversation view content area provides a highly interactive and feature-rich general display area. This area can display a variety of visualizations, including images, dynamic canvases, models, and various charts. These visualizations are generated by different backend agents, and the frontend selects appropriate components for rendering and display based on the needs. The system provides interactive tools for each visualization component in the content area. For example, in the map canvas, users can use a box to circle an area of ​​interest on the map to generate a flow chart. For chart presentation, multiple statistical chart styles are supported, and chart display items, such as data column visibility and display range, can be freely adjusted. In the 3D view, users can rotate and zoom to observe the model's structure. Furthermore, during user interaction, domain-specific questions can be asked simply by text input, and the system provides suggestions based on a large language model. This design significantly reduces user operation complexity and the learning curve, making it easy for even non-expert users to master and utilize the system for data analysis.

[0083] The gallery view is used to save and record interesting visualization results in the session view. The gallery view allows users to pin important charts or images when performing data analysis and exploration. This is achieved through a "pin" button in the session content area. Users only need to click this button to place the selected visualization content in the gallery at an appropriate scale for easy reference and review later. If you need to remove a pinned item, you can cancel it by clicking the "unpin" button. In addition, the gallery view supports saving visualization results from different sessions. This feature facilitates comparative analysis of results between different cases or data sets, helping users gain insights into data from multiple perspectives.

[0084] On the backend, a classifier based on a large language model dispatches user commands, accurately classifies them, forwards them to a processing queue, and converts them into a structured input format. Next, the structured commands are fed into the LLM (Large Language Model) proxy layer, which consists of numerous proxy functions, each responsible for executing a specific task.

[0085] Once the instructions are classified and parsed, the agent starts the process of generating visualization results. This outputs various forms of visualization results, including images, various charts, three-dimensional models, etc., to present and analyze flow field data in multiple aspects. These visualization data are transmitted to the user interface in real time via a WebSocket connection for viewing and exploration. At the same time, based on the Large Vision-Language Model (LVLM), the visualization results are automatically understood and interpreted to generate corresponding descriptive text. Users can propose the next query instruction based on the visualization results, and the system will save and update the context information related to the session to ensure the consistency of the user's thinking. In addition, the system also provides a "gallery" function that allows users to store visualization results generated by different sessions for subsequent analysis.

[0086] The backend large language model processes user instructions as follows Figure 4 shown.

[0087] When receiving a user's original command, the system first performs intent recognition. Using a large language model, such as GPT-3.5, the system understands the command's semantics and obtains a corresponding task identifier. Subsequently, specific parsing rules are selected based on the task identifier to extract key parameters from the command, resulting in a JSON-formatted string. This JSON string is then automatically converted into a structured object using Pydantic specifications. This process leverages the natural language understanding capabilities of the large language model through prompt engineering. By assigning roles to the model in the prompts, providing task descriptions, data format specifications, and input / output examples, the system ensures the accuracy and standardization of the large language model's output. Furthermore, the system features a proxy layer, which serves as the core of the backend processing logic. This proxy layer comprises multiple independent, modular proxy functions (or "services"), each responsible for performing a specific set of tasks. Upon receiving structured user command information, the system automatically selects the appropriate proxy to complete the entire workflow, from data processing to final visualization output. These proxy functions are independent of each other, allowing for independent updates and optimizations without impacting other parts of the system, ensuring maintainability and scalability. In addition, the system provides an "adaptation layer" to enhance the versatility of the system. If a new data set needs to be adapted, the adaptation layer can be simply adjusted without modifying other details of the system.

[0088] At the same time, to reduce reliance on storage resources, the system has designed proxy functions that transmit the results to the front-end in real time after completing data processing and visualization tasks, while automatically releasing the intermediate results generated during the process. This mechanism avoids the need for long-term storage of large amounts of intermediate data, optimizes the use of storage resources, and speeds up data processing and result presentation. Taking streamline and trajectory rendering tasks as an example, this process usually includes multiple steps such as parameter setting, velocity field generation, curve integral calculation, and saving and loading models or snapshots. When designing the proxy layer, the complex details behind it are shielded. Users only need to issue relevant instructions, such as the selected area, the number of initial seed points, the curve integration step size, etc. The system will automatically parse these instructions into standardized inputs, triggering the corresponding proxy to execute the complete processing flow, and ultimately obtaining a two-dimensional image or three-dimensional model of the streamline, which is then transmitted to the front-end display via the data stream.

[0089] When visual results need to be presented, different analysis tasks usually require different forms of visual presentation, such as statistical charts, flat renderings, 3D models, text descriptions, video animations, etc.

[0090] Therefore, the system sets up plug-in visualization, which provides a general visualization display area that supports multiple components. This area can dynamically select and load the corresponding visualization component according to the type of visualization result, such as Figure 5 For example, if the task is to draw a streamline, the display area will load an image rendering component; if the task is statistical analysis, it will switch to a specific type of statistical chart component. This design allows the same area to flexibly display different visualization formats. Another advantage is that adding new components or removing existing ones will not affect other modules, improving the scalability and maintainability of the system.

[0091] The core of the system is the visualization semantic enhancement based on a large language model. Semantic enhancement requires collecting the meaning of user instructions, the visualization results being described, and the configuration information of the visualization components that support the results. Among them, the visualization result image is losslessly converted into text format through Base64 encoding. The instruction subject, chart type, and original data embedded in the chart are extracted from the return data of the backend to obtain the relevant configuration parameters of the visualization component. A prompt word generation template is formulated for each visualization component. Figure 6 As shown, the chart type, theme, data, and preset configuration can be integrated into a prompt text. Meanwhile, the Base64 encoding of the visual image is sent to Gemini as additional information. By combining the image content and the contextual information in the prompt text, Gemini can generate descriptive text related to the instruction and image.

[0092] The following is a specific example to illustrate the progressive visual analysis system based on a large language model of the above embodiment, taking an ocean dataset as an example. A high-precision ocean dataset generated by the Earth numerical simulation device "Huan" is used as the basis for analysis.

[0093] The dataset description contains four dimensions and seven variables. The dimensions include: lon (longitude, 780 data points); lat (latitude, 480 data points); depth (33 data points); and time (1 time point). The variables include: lon (longitude, in degrees east, ranging from 0 to 360 degrees); lat (latitude, in degrees north, ranging from -90 to 90 degrees); depth (depth, in meters, indicating different depth levels); time (time, indicating the time when the data was recorded); ssh (sea surface height, in meters); t (temperature, in degrees Celsius; s (salinity, in PSU); u (zonal velocity, in meters per second); v (meridional velocity, in meters per second).

[0094] This dataset covers daily average ocean data over many years and records in detail the key physical parameters of the global ocean, including the longitude, latitude and depth information of each geographic location, the corresponding temperature and salinity data, and the water flow rate in the longitude (longitude direction) and latitude (latitude direction), covering important ocean areas in the Indian Ocean and the Pacific Ocean.

[0095] Manually select an area in the visual analysis interface and enter 2000 seed points to generate streamlines, from 32 degrees south latitude to 52 degrees north latitude, and from 30 degrees east longitude to 180 degrees west longitude, to generate 2000 seed points, or scatter 2000 points across the entire area with a step size of 0.1. Based on the feedback from the dialogue area, the system can correctly identify and interpret each instruction and draw a global streamline distribution map as expected, such as Figure 7 shown.

[0096] right Figure 7Analysis revealed that the density of streamlines varies across regions. For example, streamlines are dense and nearly parallel near the Philippine Sea and the North Pacific. While elongated streamlines exist in the northern Indian Ocean, their distribution is uneven. Based on this preliminary analysis, the progressive visual analysis system was asked the following question: "Why are streamlines dense and parallel near the western North Pacific and related waters? While the streamline distribution in the northern Indian Ocean appears sparse and uneven?" The system's response to the streamline visualization results is as follows.

[0097] The dense and parallel arrangement of streamlines in the western North Pacific and surrounding waters is primarily due to the following factors: The monsoon, which significantly impacts this region, creates a relatively regular wind pattern on the ocean surface, forming relatively parallel currents. These winds push the water into streamlines. Ocean currents influence this region, as it is home to a confluence of important currents, such as the Japan Current and the Kuroshio Current. The direction and strength of these currents influence the density and arrangement of local currents. Topography, with its complex terrain, including numerous islands and seamounts, also influences ocean circulation and current alignment. In contrast, the sparse and uneven flow lines in the northern Indian Ocean may be due to factors such as the monsoon and the equatorial current. The monsoon is relatively influential in the northern Indian Ocean, and variability in the monsoon and the equatorial current can lead to uneven current distribution. Topography and inshore landforms, including the topography and seafloor topography of the northern Indian Ocean, also influence the stability of ocean circulation and the density of streamlines. The northern Indian Ocean is also affected by climate change, including temperature fluctuations and instability in the ocean circulation system, which can affect the density and arrangement of ocean currents. In summary, the differences in streamline density and arrangement between the western North Pacific and the northern Indian Ocean are primarily due to a combination of factors, including monsoon influences, ocean currents, topography, and geomorphology.

[0098] The system's responses indicate that the North Pacific is significantly influenced by the monsoon, with regular winds driving the waters into parallel currents. Furthermore, multiple currents converge near the relevant waters, and the Philippine coast boasts complex terrain, including mountain ranges and islands. This is consistent with the presence of straits, peaks, and capes along the coast of the relevant waters, and the presence of a circulation current in the North Pacific. The northern Indian Ocean, in addition to being influenced by monsoon currents, may also be affected by its open terrain and climate change.

[0099] In ocean data research, seawater velocity and heat and salinity distribution are crucial to the ocean dynamic system. The interaction of these elements affects the local climate, marine ecological balance and sustainable use of resources.

[0100] Through the above analysis, spiral streamlines were also observed in the Indian Ocean and the North Pacific Ocean. Most of these vortex streamlines are relatively dense. Since the density of streamlines is related to the flow velocity, the velocity distribution of all regions in the dataset was checked. Figure 8 As shown in the figure, we can see that the current speed is usually fast where there are vortices, especially near the Mozambique Channel and the Comoros Islands. In addition, there are also narrow areas with faster seawater flow speeds along the coast of the Philippines, the East China Sea and southern Japan. Use the zoom tool to zoom in on these two areas, as shown in the figure below. Figure 9 As shown, the left side is near the Mozambique Channel, and the right side is near the East China Sea and the Philippine Sea.

[0101] Based on the above analysis, the system was further questioned and provided with relevant feedback. The question to the system was: "We also discovered spiral streamlines, which we believe to be vortices. These streamlines are mostly densely packed. We found that areas with vortex cavities have faster flow rates, particularly near the Mozambique and Comoros Islands. Furthermore, we discovered elongated areas with faster currents along the coast of the Philippines, the East China Sea, and southern Japan. Please explain this." The system's explanation was: The presence of spiral streamlines generally indicates the presence of strong ocean vortices. Ocean vortices are large-scale ocean circulation structures formed by rotating water. The water within these vortices is highly velocitious, with strong horizontal and vertical motion. The dense density of streamlines in these vortexes may be due to the highly active water movement within them. The following is an explanation: The vortex regions near the Mozambique and Comoros flocks may be prone to the formation of stable vortex structures in the marine environment. These vortices may be influenced by factors such as topography and ocean currents. In the vortex area, the seawater flow speed is accelerated because the rotational motion inside the vortex drives the seawater to move, forming a high-speed flow situation. The fast seawater flow speed along the coast of the Philippines, the East China Sea and southern Japan may be related to the local topography, climate and circulation system. For example, the Philippines is located in the edge of the western Pacific warm pool. Affected by the monsoon, a strong ocean current is formed; the East China Sea is affected by warm and cold currents, forming a complex ocean circulation system; southern Japan is located at the intersection of the western Pacific warm current and the Black Lake, and the seawater flow speed is relatively fast. In summary, the fast seawater flow speed in the vortex area and certain specific areas may be affected by the combined influence of multiple factors such as topography, ocean currents, and climate, forming a faster flow situation in the marine environment.

[0102] The seawater inside the vortex is accompanied by strong horizontal and vertical motions, resulting in active seawater movement. Stable vortex structures are easily formed in the Mozambique region. The east coast of the Philippines is on the edge of the western Pacific warm pool, and is affected by the monsoon, forming a strong ocean current; the East China Sea is affected by warm and cold currents, forming a complex ocean circulation system; and southern Japan is located at the intersection of the northwest Pacific warm current and the Kuroshio Current, and the seawater speed is relatively fast. These complex geographical and climatic conditions may cause the seawater in these areas to flow faster. On this basis, the speed distribution in the local area was checked. The area near the Philippine Sea and the Pacific Ocean was delineated and an instruction to view the attributes was given to the system. The system automatically generated a heat map of the speed distribution in the area, and provided descriptive text based on the visualization elements of the chart, such as the numerical range and color mapping, such as Figure 10 As shown in the figure, we can see the color mapping method of the heat map, that is, the change from blue to white and then to red indicates that the speed changes from low to high.

[0103] Through the above analysis, it can be seen that there is a striking red area along the southwest coast of the relevant sea area. Zooming in on the local area to view the specific values, it is found that the peak flow velocity in the relevant sea area is several times that of the Pacific Ocean on the east side (light yellow area) under the same unit. For the convenience of description, the above-mentioned Pacific Ocean is called area A and the relevant sea area is called area B. By reviewing the streamline map, the flow direction in area A is mainly along the latitude direction, while the flow direction in area B is mainly along the longitude direction. Therefore, it is necessary to further explore the velocity distribution in the longitudinal and latitudinal directions.

[0104] like Figure 11 As shown, area B not only has a higher latitudinal velocity, but also has a significant longitudinal velocity (the red and blue blocks in area B indicate opposite longitudinal velocities). In contrast, the longitudinal velocity in area A is almost zero. This also directly leads to the fact that the comprehensive flow velocity in area B is much higher than that in area A. Then, the velocity distribution of different depth layers was checked. The velocity images of the 5th, 10th and 15th layers underwater (corresponding to depths of 100 meters, 300 meters and 800 meters, respectively) were drawn as shown below. Figure 12 As shown. Figure 12From the analysis, we can see that area B still maintains a relatively high velocity in the 10th layer, but by the 15th layer, the entire area has only a relatively small velocity value. With the help of the auxiliary analysis of the large model, we know that the slowdown of seawater flow velocity is actually the inevitable result of the interaction of various physical properties of seawater (such as friction, viscosity, etc.), energy distribution, temperature and salt stratification, seabed topography, Coriolis force effect, etc. These complex factors act together on deep ocean currents to shape their unique dynamic characteristics. Then explore the performance of temperature and salinity at different times. Two time points with a half-year interval, such as data files on January 1, 2014 and July 1, 2014, were selected, and two independent sessions were established to avoid data interference in the analysis process. The visualization results were observed and compared in the gallery view. As shown Figure 13 As shown, the left side is January 1st, and the right side is July 1st. The sea temperature values ​​on January 1st are generally lower than those on July 1st, and this is more obvious around 20 degrees north latitude. However, the difference in salinity is small, and even if the numerical boundaries of the color map are narrowed, the salinity distributions of the two remain similar.

[0105] At this time, check the attribute change trend of the sea area during the entire time period. Figure 14 As shown, the horizontal axis represents time steps, and the vertical axis represents the average value of regional attributes, with January 1st representing the initial time step. It was found that the average temperature in this sea area varied by 3 degrees Celsius. The system description text indicates that the average temperature reached its lowest point at time step 50 (around late February) and reached its highest point near July. The average salinity, on the other hand, fluctuated slightly, remaining relatively stable over the six months.

[0106] In summary, the large language model-driven flow field visual analysis method provided by the embodiments of the present invention first requires obtaining the flow field dataset to be analyzed and specific user instructions. User instructions can be generated by the user through natural language input, clicking an interface button, or performing a specific mouse gesture. The pre-trained large language model then performs semantic analysis on the user instruction to obtain a task identifier corresponding to the user instruction. Parameters of the task identifier are extracted using pre-set parsing rules, and then structured transformation is performed to obtain a structured instruction corresponding to the user instruction. This structured instruction is input into the proxy layer of the large language model. The corresponding proxy function in the proxy layer executes the instruction task corresponding to the structured instruction. This instruction task then retrieves the corresponding target data from the flow field dataset to be analyzed. Simultaneously, the large language model generates a descriptive text message corresponding to the target data. This descriptive text includes variable descriptions, dimensions, units of measurement, and value range information to help users better understand the context and meaning of the target data. Based on the target data and the corresponding descriptive text, a visualization result is generated. Before displaying the visualization result, a visualization presentation method is obtained from the user instruction. This presentation method includes a chart, rendering, model, text description, or video animation. The target data is displayed according to the display method corresponding to the user's instructions. The user's specific instructions are recognized through a large language model to achieve natural language-based interaction. The user's operation instructions or text instructions can be converted into structured instructions to obtain the corresponding target data. The corresponding descriptive instructions are generated through the large language model, and the target data and the corresponding descriptive description are presented in real time to help users better understand the context and meaning of the target data, greatly improving the user experience.

[0107] In the embodiments of the present application, a large language model-driven flow field visual analysis device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and the details that have been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0108] The present application provides a flow field visual analysis device driven by a large language model. Figure 15 : is a schematic diagram of the structure of a flow field visual analysis device driven by a large language model provided in an embodiment of the present application, the device comprising:

[0109] Acquisition module 1501, used to acquire the flow field data set to be analyzed and user instructions;

[0110] The instruction conversion module 1502 is used to perform semantic analysis on the user instruction using a large language model to obtain a structured instruction;

[0111] The target data acquisition module 1503 is configured to acquire corresponding target data from the flow field dataset to be analyzed using the large language model based on the structured instruction, and generate descriptive text corresponding to the target data;

[0112] A result generating module 1504 is configured to run a visualization agent based on the target data and the descriptive text to obtain a visualization result;

[0113] The result display module 1505 is used to display the target data according to the data display method indicated by the user instruction.

[0114] In an optional implementation, the instruction conversion module 1502 is specifically configured to:

[0115] The user instruction is semantically analyzed through a large language model to obtain a task identifier corresponding to the user instruction; parameters of the task identifier are extracted through preset parsing rules, and structured conversion is performed to obtain a structured instruction.

[0116] In an optional implementation, the target data acquisition module 1503 is specifically configured to:

[0117] Extracting a corresponding instruction task from the structured instruction; confirming a corresponding proxy function in the large language model proxy layer based on the instruction task; and executing the instruction task through the proxy function to obtain corresponding target data from the flow field data set to be analyzed.

[0118] In an optional implementation, the target data acquisition module 1503 is further configured to:

[0119] The target data is analyzed by the large language model to generate descriptive text corresponding to the target data; the descriptive text includes: variable description, dimension, measurement unit or value range information.

[0120] In an optional embodiment, the result display module 1505 is specifically configured to:

[0121] A corresponding data presentation method is extracted from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description, or a video animation; and the target data is presented according to the data presentation method.

[0122] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0123] The large language model-driven flow field visual analysis device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0124] The embodiment of the present invention also provides a computer device having the above Figure 15 The flow field visual analysis device driven by a large language model is shown.

[0125] See also Figure 16 , Figure 16 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 16 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information in the graphical user interface on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 16 A processor 10 is taken as an example.

[0126] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0127] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0128] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0129] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0130] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 16 The bus connection is taken as an example.

[0131] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0132] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0133] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A flow field visual analysis method driven by a large language model, characterized by: The method comprises: Obtain the flow field data set to be analyzed and user instructions; Performing semantic analysis on the user instructions through a large language model to obtain structured instructions; Based on the structured instruction, corresponding target data is acquired from the flow field data set to be analyzed through the large language model, and descriptive text corresponding to the target data is generated; Based on the target data and the descriptive text, running a visualization agent to obtain a visualization result; The target data is displayed in accordance with the data display method indicated by the user instruction.

2. The method according to claim 1, characterized in that The semantic analysis of the user instruction by the large language model to obtain a structured instruction includes: Performing semantic analysis on the user instruction using a large language model to obtain a task identifier corresponding to the user instruction; Parameters of the task identifier are extracted using preset parsing rules, and structured conversion is performed to obtain structured instructions.

3. The method according to claim 2, characterized in that The user instructions include natural language, interface button operation instructions or mouse gesture instructions.

4. The method according to claim 2, characterized in that The acquiring corresponding target data from the flow field data set to be analyzed by using the large language model includes: Extracting corresponding instruction tasks from the structured instructions; Based on the instruction task, determining a corresponding proxy function in the proxy layer of the large language model; The instruction task is executed by the proxy function to obtain corresponding target data and visualization results from the flow field data set to be analyzed.

5. The method according to claim 4, characterized in that Generating a descriptive text corresponding to the target data includes: Analyzing the target data using the large language model to generate descriptive text corresponding to the target data; The descriptive text includes: variable description, dimension, measurement unit or value range information.

6. The method according to claim 5, characterized in that The displaying of the target data according to the data display method indicated by the user instruction includes: Extracting a corresponding data presentation method from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description, or a video animation; The target data is displayed according to the data display method.

7. A flow field visual analysis device driven by a large language model, characterized in that: The device comprises: The acquisition module is used to obtain the flow field data set to be analyzed and user instructions; An instruction conversion module, configured to perform semantic analysis on the user instruction using a large language model to obtain a structured instruction; a target data acquisition module, configured to acquire corresponding target data from the flow field data set to be analyzed using the large language model based on the structured instruction, and generate descriptive text corresponding to the target data; A result generation module, configured to run a visualization agent based on the target data and the descriptive text to obtain a visualization result; The result display module is used to display the target data in a data display manner indicated by the user instruction.

8. The device according to claim 7, characterized in that The instruction conversion module is specifically used to: Performing semantic analysis on the user instruction using a large language model to obtain a task identifier corresponding to the user instruction; Parameters of the task identifier are extracted using preset parsing rules, and structured conversion is performed to obtain structured instructions.

9. The device according to claim 8, characterized in that The target data acquisition module is specifically used to: Extracting corresponding instruction tasks from the structured instructions; Based on the instruction task, determining a corresponding proxy function in the proxy layer of the large language model; The instruction task is executed by the proxy function to obtain corresponding target data and visualization results from the flow field data set to be analyzed.

10. The device according to claim 9, characterized in that The result display module is specifically used to: Extracting a corresponding data presentation method from the user instruction; the data presentation method includes a chart, a rendering, a model, a text description, or a video animation; The target data is displayed according to the data display method.

Images