Information acquisition method and device of alloy material, electronic equipment and medium

Abstract

The invention discloses an information acquisition method and device of an alloy material, electronic equipment and a medium. The method comprises the following steps: acquiring an information query instruction of the alloy material; analyzing the information query instruction through a specified large model to determine a current information query task; the information query task comprises at least one of a general information query task, an alloy performance prediction task and an alloy reverse design task; according to the information query task, calling a target model from a pre-constructed alloy field model library; and executing the information query task through the target model to obtain a target query result. Therefore, the general understanding, reasoning and interaction capabilities of the specified large model and the professional prediction precision and reverse optimization capability of the alloy field model are effectively combined, so that the efficiency, accuracy and intelligent level of alloy material design and discovery are improved, and the rich knowledge requirements of users on alloy material design are met.

Description

Technical Field

[0001] This application relates to the field of materials science and technology, and in particular to a method, apparatus, electronic device and medium for acquiring information about alloy materials. Background Technology

[0002] Traditional alloy material design and development primarily rely on trial-and-error methods and experience-driven experimental approaches. The development process includes multiple stages such as composition design, smelting and preparation, heat treatment, mechanical property testing, and microstructure characterization. This traditional model has revealed numerous bottlenecks and challenges when facing increasingly complex and diverse material performance requirements, severely restricting the efficiency of new material development and industrialization.

[0003] For example, researchers often need to conduct numerous repeated experiments to screen potential alloy compositions and process parameters in order to complete the development of alloy materials. From initial design to performance verification, multiple iterations are usually required, resulting in a long development cycle, low efficiency, and difficulty in meeting the demand for rapid updates and replacements of high-performance materials.

[0004] Furthermore, knowledge in the field of alloy materials science is highly fragmented, with a large number of research results existing in unstructured form in a vast amount of academic literature, patent documents, and technical reports. This fragmentation of knowledge makes it difficult for researchers to quickly acquire rich knowledge about alloy materials, thus limiting the reuse and innovation of knowledge.

[0005] Therefore, how to achieve efficient and accurate alloy material design and meet users' extensive knowledge needs for alloy material design is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, one aspect of this application provides a method for obtaining information on alloy materials, the method comprising: Commands to retrieve information about alloy materials; The information query instruction is parsed by specifying a large model to determine the current information query task; the information query task includes at least one of a general information query task, an alloy performance prediction task, and an alloy reverse design task. Based on the information query task, the target model is retrieved from the pre-built alloy domain model library; The information query task is executed through the target model to obtain the target query result.

[0007] Optionally, when the information query task is the alloy performance prediction task, the target model is a machine learning model trained based on the XGBoost algorithm; the machine learning model includes a tensile strength prediction module and a yield strength prediction model.

[0008] Optionally, the step of executing the information query task through the target model to obtain the target query result includes: Extract the initial characteristic parameters of the alloy to be predicted in the alloy property prediction task; After preprocessing the initial feature parameters, the first feature parameters are obtained; the preprocessing includes at least one of parameter cleaning and parameter verification. The first feature parameter is converted into data in a two-dimensional array format to obtain the second feature parameter; The second feature parameter is input into the machine learning model to obtain the predicted tensile strength and predicted yield strength of the alloy to be predicted.

[0009] Optionally, when the information query task is the alloy reverse design task, the target model is an XGBoost regression model based on a genetic algorithm optimization framework.

[0010] Optionally, the step of executing the information query task through the target model to obtain the target query result includes: Extract the target performance parameters of the alloy to be designed in the alloy reverse design task; The target performance parameters are input into the XGBoost regression model to obtain the initial technical parameters for generating the alloy to be designed. Based on the initial technical parameters, the current initial population of the genetic algorithm optimization framework is generated; Determine whether the current genetic iteration meets the preset iteration conditions; If so, based on the XGBoost regression model, the target technical parameters are obtained from the population corresponding to the current genetic iteration; wherein, the technical parameters include at least one of alloy composition and process parameters; If not, update the population using the genetic operators and return to the step of determining whether the current genetic iteration has met the preset iteration conditions.

[0011] Optionally, the genetic operators include selection, crossover, and mutation operations; The genetic algorithm optimization framework is embedded with pre-set constraints; the constraints include at least one of the following: all process parameters are greater than 0; the total content of specified chemical elements in the alloy composition is not greater than a preset percentage.

[0012] Optionally, after performing the information query task through the target model, the following steps are included: Determine whether the information query task was successful; If successful, output the target query result; If the query fails, an alternative query result is generated using the specified large model; the alternative query result includes at least one of the following: the analysis text of the query failure and the adjustment suggestion text of the information query instruction.

[0013] Another aspect of this application provides an information acquisition device for alloy materials, the device comprising: The instruction acquisition module is used to obtain information query instructions for alloy materials; The instruction parsing module is used to parse the information query instruction by specifying a large model in order to determine the current information query task; the information query task includes at least one of a general information query task, an alloy performance prediction task, and an alloy reverse design task; The model retrieval module is used to retrieve the target model from a pre-built alloy domain model library based on the information query task. The target query module is used to execute the information query task through the target model and obtain the target query result.

[0014] Another aspect of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor executes the computer program to implement the steps of a method for acquiring information about the alloy material.

[0015] Another aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for acquiring information about the alloy material.

[0016] The information acquisition method, apparatus, electronic device, and medium for alloy materials provided in this application have the following beneficial effects: Based on a specified large model, it can perform high-precision intent recognition on user-inputted information query commands, thereby determining different information query tasks such as general question answering, alloy performance prediction, or alloy reverse design. Furthermore, it selects the corresponding target model for querying, realizing a dynamic selection mechanism for different processing paths for different tasks. Thus, it effectively combines the general understanding, reasoning, and interaction capabilities of the specified large model with the professional prediction accuracy and reverse optimization capabilities of alloy domain models, thereby improving the efficiency, accuracy, and intelligence level of alloy material design and discovery, and meeting users' rich knowledge needs for alloy material design. Attached Figure Description

[0017] Figure 1 A flowchart illustrating an information acquisition method for alloy materials provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an information acquisition system for alloy materials provided in an embodiment of this application; Figure 3A schematic diagram illustrating the principle of an information acquisition method for alloy materials provided in an embodiment of this application; Figure 4 A schematic diagram illustrating the principle of an information acquisition method for alloy materials provided in another embodiment of this application; Figure 5 This is a schematic diagram of the structure of an information acquisition device for alloy materials provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0018] The reference numerals in the attached diagram are as follows: 60 is memory, 61 is processor, 62 is display screen, 63 is input / output interface, 64 is communication interface, 65 is power supply, 66 is communication bus, 601 is computer program, 602 is operating system, and 603 is data. Detailed Implementation

[0019] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0020] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0021] Figure 1 This is a flowchart illustrating a method for acquiring information about alloy materials provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes: S10: Command to retrieve information on alloy materials; Figure 2 This is a schematic diagram of the structure of an information acquisition system for alloy materials provided in an embodiment of this application. In an optional embodiment, the information acquisition method for alloy materials provided in this application can be applied to this information acquisition system, such as... Figure 2 As shown in the embodiments of this application, the information acquisition system for alloy materials includes a data interaction layer, an integration scheduling layer, and a domain expert layer.

[0022] The data interaction layer includes a user interface and a specified large model. The user interface is used to obtain information query instructions for alloy materials. These instructions can be multimodal data such as text, voice, and images, and this application does not limit their use. Furthermore, in addition to providing users with the ability to input queries via natural language, the user interface can also receive and display the query results corresponding to the information query instructions. In an optional embodiment, the display interface can be a web-based chat interface or an API interface.

[0023] The term "designated large model" refers to a large language model (LLM). In specific embodiments, the designated large model may include, but is not limited to, the GPT series, Tongyi Qianwen, DeepSeek, and LLaMA series. This application does not limit the scope of the designation.

[0024] Of course, in another alternative embodiment, in order to meet the rich and accurate knowledge requirements of the alloy materials field, a large amount of dialogue training datasets in the alloy field can be collected to train a pre-built large model and obtain a general large model in the alloy field. That is, the specified large model can also be a large model obtained by training the alloy field training dataset.

[0025] S11: Parse the information query instruction by specifying a large model to determine the current information query task; the information query task includes at least one of the following: general information query task, alloy performance prediction task, and alloy reverse design task; In a specific embodiment, a large model is designated as the core decision-making and understanding unit of the system, responsible for parsing user intent, generating responses, and intelligently calling external tools (including unit-limited calling instructions, parameter extraction, result integration, and feedback). Specifically, after obtaining the information query instruction in step S10, the large model parses the instruction to determine the user's query intent, that is, to determine the current information query task.

[0026] The user-input commands can include general queries about alloys, performance queries about alloys, and design queries about alloys. Correspondingly, the information query task includes at least one of the following: general information query task, alloy performance prediction task, and alloy reverse design task.

[0027] Figure 3 This is a schematic diagram illustrating the principle of an information acquisition method for alloy materials provided in an embodiment of this application, as shown below. Figure 3As shown, in a specific embodiment, different information query tasks correspond to different processing paths. Specifically, when a user inputs an information query command through the user interface, the query is first routed to the question classifier 1. The question classifier 1 performs a preliminary identification of the user's intent, first determining whether the current information query task belongs to a general answer or a professional answer. Here, a general answer refers to a non-professional, routine question answer, such as, "What is the definition of an alloy?"

[0028] according to Figure 3 It is understood that professional answers include alloy prediction and alloy design. Professional answers refer to specific professional questions concerning alloy materials, particularly the prediction or design of heat treatment parameters and tensile properties. In a specific embodiment, when the information query task is a professional answer, the question classifier 2 is invoked for identification to determine the specific data category of the current information query task.

[0029] In fact, correspondingly, problem classifier 1 and problem classifier 2 implement step S11. It can be seen that different information query tasks can be processed through different paths to ensure the accuracy and efficiency of information query.

[0030] S12: Based on the information query task, retrieve the target model from the pre-built alloy domain model library; S13: Perform the information query task through the target model to obtain the target query results.

[0031] like Figure 2 As shown, the information acquisition system integration scheduling layer provided in this application embodiment includes an integration scheduling module, which is the core of the system's workflow orchestration and is responsible for managing the connections and data flows between all nodes. In an optional embodiment, the integration scheduling module can be implemented based on the visual workflow orchestration platform Dify, which is an open-source large language model (LLM) application development platform. It integrates the concepts of Backend-as-a-Service (BaaS) and LLMops, aiming to provide developers with an intuitive interface for quickly building and deploying applications based on large language models.

[0032] The integrated scheduling module includes functions such as node management, conditional branching, data transformation, and tool invocation. Node management is used to define and configure various types of nodes, including but not limited to LLM nodes, code execution nodes, tool nodes, knowledge retrieval nodes, and direct response nodes. Conditional branching is used to implement multi-path conditional routing based on the results of the problem classifier.

[0033] Data transformation is used to standardize data formats between different modules through code execution nodes (e.g., Dify's code type nodes). Standardization transformation includes, but is not limited to, converting text parameters into numeric arrays or JSON format. For example, converting text parameters generated by LLM into numeric arrays or JSON structures required by the target model.

[0034] The tool call acts as a bridge between the LLM and external tool type nodes (i.e., models in the domain expert layer), handling API requests and responses. It should be noted that the tool node includes at least one domain expert model for alloy property prediction and one domain expert model for alloy reverse design.

[0035] The integrated scheduling model can invoke the target model from the domain expert layer to execute the corresponding query task based on the current information query task, thereby obtaining the target query result. During the process of the integrated scheduling module retrieving the target model, such as... Figure 2 As shown, the integrated scheduling module can pass parameters and also send the results back to a specified large model so that the target query results can be output to the user.

[0036] Therefore, the alloy material information acquisition method provided in this application, based on a specified large model, can perform high-precision intent recognition of user-inputted information query commands, thereby determining different information query tasks such as general question answering, alloy performance prediction, or alloy reverse design. It further selects the corresponding target model for querying, realizing a dynamic selection mechanism for different processing paths for different tasks. Thus, it effectively combines the general understanding, reasoning, and interaction capabilities of the specified large model with the professional prediction accuracy and reverse optimization capabilities of alloy domain models, thereby improving the efficiency, accuracy, and intelligence level of alloy material design and discovery, and meeting users' rich knowledge needs for alloy material design.

[0037] In one optional embodiment, when the information query task is an alloy performance prediction task, the target model is a machine learning model trained based on the XGBoost (eXtreme Gradient Boosting) algorithm; the machine learning model includes an independent tensile strength prediction module and a yield strength prediction model. This machine learning model receives input vectors containing process parameters and chemical element content, performs parallel predictions of tensile strength and yield strength, and organizes the prediction results into a JSON array.

[0038] Based on the above embodiments, as an optional embodiment, the information query task is performed through the target model to obtain the target query result, including: Extract the initial characteristic parameters of the alloy to be predicted in the alloy property prediction task; After preprocessing the initial feature parameters, the first feature parameters are obtained; the preprocessing includes at least one of parameter cleaning and parameter verification. The first feature parameter is converted into data in a two-dimensional array format to obtain the second feature parameter; The second feature parameter is input into the machine learning model to obtain the predicted tensile strength and predicted yield strength of the alloy to be predicted.

[0039] In a specific embodiment, such as Figure 2 and Figure 3 As shown, if the current information query task is alloy prediction, Figure 2 The system will first return a processing prompt to the user, such as "Alloy prediction problem, requires prediction model collaboration, please wait." This informs the user that the system is currently calculating and processing.

[0040] In an optional embodiment, initial characteristic parameters of the alloy to be predicted can be extracted using a specific LLM, wherein the initial characteristic parameters include, but are not limited to, outer diameter, thickness, quenching temperature, and elemental content. Similarly, this application does not limit the specific LLM used.

[0041] Furthermore, the extracted unstructured initial feature parameters are passed to the code execution node 1. In an optional embodiment, the code execution node 1 executes preset Python code to preprocess the initial feature parameters. Specifically, the initial feature parameters are cleaned, validated, and formatted into a standardized two-dimensional array format required by the machine learning model API.

[0042] The standardized two-dimensional array format is then passed to the Tool_API tool node, which constructs and sends an HTTP POST request to the pre-configured target model service address, i.e., the service address of the machine learning model trained based on the XGBoost algorithm.

[0043] Simultaneously, parameters in a standardized two-dimensional array format are encapsulated in the JSON request body, specifying the API path to invoke the tensile strength prediction module and the yield strength prediction model. The machine learning model receives characteristic parameters including element content, thickness, outer diameter, and temperature as input. After calculation and processing by the machine learning module, the predicted tensile strength and predicted yield strength of the alloy to be predicted can be determined. Figure 3 As shown, the predicted results are transmitted to the LLM for integration into a JSON array, and then returned to... Figure 2 The user interface shown.

[0044] like Figure 3 As shown, in an optional embodiment, if the machine learning model trained based on the XGBoost algorithm successfully predicts the tensile strength and yield strength, then... Figure 3 As shown, the predicted results can be transmitted to the LLM to be integrated into a JSON array, and then integrated, interpreted, and formatted into a natural language (e.g., a Markdown table) that is easy for the user to understand before being fed back to the user interface.

[0045] Therefore, the alloy material information acquisition method provided in this application embodiment, through a machine learning model trained based on the XGBoost algorithm, can predict multiple key mechanical properties of the alloy, such as tensile strength and yield strength, in parallel with high precision based on the detailed composition and process parameters of the alloy. The prediction results are output in a structured JSON format, which is convenient for LLM to integrate and interpret later.

[0046] In an optional embodiment, when the information query task is an alloy reverse engineering task, the target model is an XGBoost regression model based on a genetic algorithm optimization framework. The core of this model is a genetic algorithm-based optimization framework designed to reverse engineer the combination of alloy composition and process parameters that best approximates the performance of the internal prediction model to the target value—that is, the target technical parameters. A pre-trained XGBoost regression model is integrated to evaluate the performance of the generated individual (i.e., a set of parameters) in each iteration and compare it with the user-defined target performance.

[0047] Figure 4 This is a schematic diagram illustrating the principle of an information acquisition method for alloy materials according to another embodiment of this application. Based on the above embodiment, as an optional embodiment, an information query task is performed through a target model to obtain the target query result, including: Extract the target performance parameters of the alloy to be designed in the reverse design task of the alloy; Input the target performance parameters into the XGBoost regression model to obtain the initial technical parameters for generating the alloy to be designed. Based on the initial technical parameters, the current initial population of the genetic algorithm optimization framework is generated; Determine whether the current genetic iteration meets the preset iteration conditions; If so, based on the XGBoost regression model, obtain the target technical parameters from the population corresponding to the current genetic iteration; wherein, the technical parameters include at least one of alloy composition and process parameters; If not, update the population using the genetic operators and return to the step of determining whether the current genetic iteration has met the preset iteration conditions.

[0048] like Figure 3As shown in the specific embodiment, when the information query task is alloy design, the target performance parameters of the alloy to be designed are extracted through a specific LLM large model. The target performance parameters refer to the performance parameters that the alloy to be designed is expected to achieve, and the target performance parameters may include, but are not limited to, the target yield strength and tensile strength.

[0049] In practice, Figure 2 The system shown will first send a processing prompt signal, such as "Reverse design problem, which requires collaboration of XGBoost regression model based on genetic algorithm optimization framework, which has a large computational load, please wait."

[0050] At this point, the LLM large model synchronously extracts the unstructured target performance parameters and passes them to another code execution node. The code execution node executes pre-defined Python code to format the target performance parameters into the standardized JSON format required by the XGBoost regression model API based on the genetic algorithm optimization framework.

[0051] Furthermore, the standardized target performance parameters are then passed to the Tool_API tool node. This tool node constructs and sends an HTTP POST request to the domain expert layer, specifying the API path, thereby invoking the XGBoost regression model based on the genetic algorithm optimization framework.

[0052] In a specific embodiment, after the target performance parameters are input into the XGBoost regression model, the genetic algorithm optimization process is initiated. Specifically, as shown below... Figure 4 As shown, the initial population is generated based on the initial technical parameters output by the XGBoost regression model. That is, the initial values ​​of the feature vectors are randomly generated within a predefined reasonable range to generate the initial population.

[0053] If the current iteration meets the preset iteration conditions, the optimal target technical parameters are selected from the population corresponding to the current genetic iteration based on the XGBoost regression model. It should be noted that the technical parameters include at least one of the alloy composition and process parameters.

[0054] In another embodiment, if the current iteration does not meet the preset iteration condition, the genetic algorithm population is updated, and iterative calculations are performed repeatedly. Specifically, classic genetic operators are used, including tournament selection, crossover, and mutation operations to generate new offspring individuals. Simultaneously, multi-core parallel computation is employed to evaluate the fitness function, thereby improving the computational efficiency for large-scale populations and multi-generational evolution.

[0055] Parallel computation for evaluating fitness functions refers to distributing the fitness calculations for each individual across multiple CPU cores or machines for simultaneous execution, rather than performing them sequentially one by one in the main process. This trades computational resources for time, allowing a generation of the population to complete the overall evaluation in almost the same timeframe as evaluating a single individual, thus significantly reducing the total runtime of the GA (General Achievement Test).

[0056] It should be noted that the preset iteration conditions can preset the number of iterations, which is not limited in this application. After iterative optimization with a set number of iterations, the XGBoost regression model returns several optimal recommended combinations of alloy composition and process parameters. Each combination is presented in dictionary form, listing the suggested values ​​for each feature.

[0057] Thus, the genetic algorithm iteratively generates and evaluates a series of candidate alloy composition and process parameter combinations, uses its internally integrated XGBoost regression model to predict the performance of each combination, and evaluates the fitness based on the deviation from the target performance.

[0058] In one alternative embodiment, in Figure 3 As shown, after the XGBoost regression model based on the genetic algorithm optimization framework is successfully designed, the target technical parameters can be obtained. Similarly, the target technical parameters can be integrated and formatted using LLM, and then interpreted into natural language or Markdown format and displayed to the user through the user interface.

[0059] It should be noted that when calling the target model, this application standardizes the interface by encapsulating all models in the alloy domain model library into tool-type nodes that can be called by LLM in the Dify workflow. For example, the tool_api tool, as a general HTTP request wrapper, is responsible for communicating with the small model service actually deployed on the backend.

[0060] When LLM calls a model from the alloy domain model library for decision-making, it passes two core parameters to the tool_api: the route parameter and the input parameter. The route parameter explicitly specifies the specific small model API endpoint to be called. The input parameter contains all the data to be processed that LLM extracts or generates based on the user query.

[0061] After receiving these parameters, the tool_api tool constructs an HTTP POST request and sends the input parameters to the address of the alloy domain model service deployed on the backend in the form of a JSON request body, thereby enabling LLM to seamlessly call heterogeneous alloy domain models.

[0062] Therefore, the alloy material information acquisition method provided in this application innovatively introduces alloy reverse design. Its core is an intelligent optimization framework based on a genetic algorithm (GA), which integrates a pre-trained XGBoost prediction model—that is, an XGBoost regression model based on the genetic algorithm optimization framework. Users only need to provide the target performance (e.g., the desired strength value), and GA can iteratively search and optimize the combination of alloy composition and heat treatment process parameters that best meet the target performance under preset constraints. By evaluating individual fitness in parallel and using genetic operators such as crossover, mutation, and selection, the discovery and screening process of new materials is significantly accelerated, transforming traditional trial and error into intelligent recommendation.

[0063] Furthermore, to overcome the inherent data format differences between the output of large-scale language models and the input of alloy domain models, the alloy material information acquisition system includes a code execution module. This module can automatically and intelligently clean, validate, and convert the unstructured parameters extracted by LLM. For example, it can accurately convert parameters described in text into two-dimensional numerical arrays required by alloy property prediction models, or encapsulate the target parameters of reverse engineering into standardized JSON objects required by small genetic algorithm models. This enables efficient and seamless data exchange and function calls between LLM and various types of alloy domain models.

[0064] Based on the above embodiments, as an optional embodiment, the genetic algorithm optimization framework is embedded with pre-set constraints; the constraints include at least one of the following: all process parameters are greater than 0; the total content of specified chemical elements in the alloy composition is not greater than a preset percentage.

[0065] In a specific embodiment, for process or formulation variables that truly need optimization, such as raw material ratios, reaction temperature and pressure, 0 or negative numbers are prohibited in the chromosomes of the genetic algorithm; that is, all process parameters are greater than 0.

[0066] In another alternative embodiment, for certain specified elements, such as nickel and carbon, the sum of their mass fractions in the formulation does not exceed a preset percentage (e.g., 50%). In a specific embodiment, if the total content of the specified chemical elements exceeds the preset percentage, the solution is deemed infeasible, and a penalty function can be used to minimize the fitness or discard the solution directly.

[0067] In one alternative embodiment, after performing the information query task through the target model, the process includes: Determine whether the information query task was successful; If successful, output the target query results; If the query fails, an alternative query result is generated by specifying a large model. The alternative query result includes at least one of the following: the analysis text of the failed query and the adjustment suggestion text of the information query instruction.

[0068] It is understandable that in specific embodiments, there may be situations where the target model cannot respond. Returning an empty value, i.e., failing to provide any relevant results, would significantly degrade the user experience. Therefore, to address the aforementioned technical problem, in one optional embodiment, such as... Figure 3 As shown, a fallback strategy is provided.

[0069] Specifically, when both alloy prediction and alloy design are successful, the results are integrated and output using an LLM model. However, when prediction or design fails (which can be determined via a fail-branch), the system triggers a prompt message, such as "The current target model calculation is unreasonable; an answer will be provided based on the knowledge base shortly; please wait," to inform the user of alternative queries.

[0070] Specifically, in one optional embodiment, alternative query results can be generated directly by specifying a large model, or alternative query results can be generated by searching the knowledge database to see if there are identical historical query records or data that meet the requirements. If they exist, they are retrieved directly; if they do not exist, alternative query results are generated by specifying a large model.

[0071] The alternative query results may include analytical text indicating query failure, as well as text suggesting adjustments to the information query instructions; this application does not limit the scope of such adjustments.

[0072] exist Figure 2 In the system embodiment shown, the `tool_api` tool is responsible for receiving the HTTP response returned by the alloy domain model service. Normally, the alloy domain model calculation results will be sent back to the LLM for further processing. If the alloy domain model service returns an error or the request fails, the fail-branch mechanism of the Dify workflow will be triggered, guiding the process into the exception handling path.

[0073] The exception handling path refers to the preset backup path that is triggered when the tool node fails to execute. The backup path includes, but is not limited to, searching the knowledge database and generating alternative query results through a specified large model.

[0074] Therefore, the alloy material information acquisition method provided in this application includes a fault handling process. When the alloy domain model (whether for alloy prediction or alloy reverse design) fails to return a valid result due to external reasons (such as calculation failure or input exceeding the valid range), the system can automatically trigger a preset failure branch. This provides immediate user feedback, subsequently activating a knowledge database retrieval to obtain relevant information from the existing materials knowledge base. A large language model then generates explanatory or alternative responses based on this supplementary knowledge, greatly improving system reliability and user experience, and avoiding service interruptions caused by the failure of a single model.

[0075] In the above embodiments, the method for acquiring information about alloy materials has been described in detail. This application also provides an embodiment of an apparatus for acquiring information about alloy materials.

[0076] Figure 5 This is a schematic diagram of the structure of an information acquisition device for alloy materials provided in an embodiment of this application, as shown below. Figure 5 As shown, the device includes: Instruction acquisition module 50 is used to acquire information query instructions for alloy materials; The instruction parsing module 51 is used to parse the information query instruction by specifying a large model in order to determine the current information query task; the information query task includes at least one of the following: general information query task, alloy performance prediction task, and alloy reverse design task; The model retrieval module 52 is used to retrieve the target model from a pre-built alloy domain model library based on the information query task. The target query module 53 is used to perform information query tasks through the target model and obtain target query results.

[0077] Furthermore, the alloy material information acquisition device provided in this application embodiment also includes: The initial feature parameter extraction module is used to extract the initial feature parameters of the alloy to be predicted in the alloy property prediction task; The preprocessing module is used to preprocess the initial feature parameters to obtain the first feature parameters; the preprocessing includes at least one of parameter cleaning and parameter verification; The format conversion module is used to convert the first feature parameter into data in a two-dimensional array format to obtain the second feature parameter. The prediction module is used to input the second feature parameter into the machine learning model to obtain the predicted tensile strength and predicted yield strength of the alloy to be predicted.

[0078] The target performance parameter extraction module is used to extract the target performance parameters of the alloy to be designed in the reverse design task of the alloy. The initial technical parameter acquisition module is used to input the target performance parameters into the XGBoost regression model to obtain the initial technical parameters for generating the alloy to be designed. The current initial population generation module is used to generate the current initial population of the genetic algorithm optimization framework based on the initial technical parameters; The iterative loop module is used to determine whether the current genetic iteration has met the preset iteration conditions. If so, based on the XGBoost regression model, the target technical parameters are obtained from the population corresponding to the current genetic iteration. The technical parameters include at least one of alloy composition and process parameters. If not, the population is updated through genetic operators, and the step of determining whether the current genetic iteration has met the preset iteration conditions is returned.

[0079] The target processing module is used to determine whether the information query task is successful. If successful, the target query result is output. If unsuccessful, an alternative query result is generated by specifying a large model. The alternative query result includes at least one of the following: the analysis text of the query failure and the adjustment suggestion text of the information query instruction.

[0080] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 6 As shown, the electronic device includes: a memory 60 for storing computer programs; The processor 61 is used to execute a computer program to implement the steps of the alloy material information acquisition method mentioned in the above embodiments.

[0081] The electronic devices provided in this embodiment may include, but are not limited to, laptops or desktop computers.

[0082] The processor 61 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 61 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 61 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 61 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 61 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0083] The memory 60 may include one or more computer-readable storage media, which may be non-transitory. The memory 60 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 60 is used to store at least the following computer program 601, which, after being loaded and executed by the processor 61, is capable of implementing the relevant steps of the alloy material information acquisition method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 60 may also include an operating system 602 and data 603, and the storage method may be temporary or permanent storage. The operating system 602 may include Windows, Unix, Linux, etc. The data 603 may include, but is not limited to, the relevant data involved in the alloy material information acquisition method.

[0084] In some embodiments, the electronic device may further include a display screen 62, an input / output interface 63, a communication interface 64, a power supply 65, and a communication bus 66.

[0085] Those skilled in the art will understand that Figure 6 The structures shown do not constitute a limitation on electronic devices and may include more or fewer components than those shown.

[0086] The electronic device provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the information acquisition method for alloy materials described in the above embodiments.

[0087] It should be noted that although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

Claims

1. A method for acquiring information about alloy materials, characterized in that, The method includes: Commands to retrieve information about alloy materials; The information query instruction is parsed by specifying a large model to determine the current information query task; the information query task includes at least one of a general information query task, an alloy performance prediction task, and an alloy reverse design task. Based on the information query task, the target model is retrieved from the pre-built alloy domain model library; The information query task is executed through the target model to obtain the target query result.

2. The method for acquiring information about alloy materials as described in claim 1, characterized in that, When the information query task is the alloy performance prediction task, the target model is a machine learning model trained based on the XGBoost algorithm; the machine learning model includes a tensile strength prediction module and a yield strength prediction model.

3. The method for acquiring information about alloy materials as described in claim 2, characterized in that, The step of executing the information query task through the target model to obtain the target query result includes: Extract the initial characteristic parameters of the alloy to be predicted in the alloy property prediction task; After preprocessing the initial feature parameters, the first feature parameters are obtained; the preprocessing includes at least one of parameter cleaning and parameter verification. The first feature parameter is converted into data in a two-dimensional array format to obtain the second feature parameter; The second feature parameter is input into the machine learning model to obtain the predicted tensile strength and predicted yield strength of the alloy to be predicted.

4. The method for acquiring information about alloy materials as described in claim 1, characterized in that, When the information query task is the alloy reverse design task, the target model is the XGBoost regression model based on the genetic algorithm optimization framework.

5. The method for acquiring information about alloy materials as described in claim 4, characterized in that, The step of executing the information query task through the target model to obtain the target query result includes: Extract the target performance parameters of the alloy to be designed in the alloy reverse design task; The target performance parameters are input into the XGBoost regression model to obtain the initial technical parameters for generating the alloy to be designed. Based on the initial technical parameters, the current initial population of the genetic algorithm optimization framework is generated; Determine whether the current genetic iteration meets the preset iteration conditions; If so, based on the XGBoost regression model, the target technical parameters are obtained from the population corresponding to the current genetic iteration; wherein, the technical parameters include at least one of alloy composition and process parameters; If not, update the population using the genetic operators and return to the step of determining whether the current genetic iteration has met the preset iteration conditions.

6. The method for acquiring information about alloy materials as described in claim 5, characterized in that, The genetic operators include selection, crossover, and mutation operations; The genetic algorithm optimization framework is embedded with pre-set constraints; the constraints include at least one of the following: all process parameters are greater than 0; the total content of specified chemical elements in the alloy composition is not greater than a preset percentage.

7. The method for acquiring information about alloy materials as described in claim 1, characterized in that, After the information query task is executed through the target model, the following steps are included: Determine whether the information query task was successful; If successful, output the target query result; If the query fails, an alternative query result is generated using the specified large model; the alternative query result includes at least one of the following: the analysis text of the query failure and the adjustment suggestion text of the information query instruction.

8. An information acquisition device for alloy materials, characterized in that, The device includes: The instruction acquisition module is used to obtain information query instructions for alloy materials; The instruction parsing module is used to parse the information query instruction by specifying a large model in order to determine the current information query task; the information query task includes at least one of a general information query task, an alloy performance prediction task, and an alloy reverse design task; The model retrieval module is used to retrieve the target model from a pre-built alloy domain model library based on the information query task. The target query module is used to execute the information query task through the target model and obtain the target query result.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the information acquisition method for alloy materials according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the information acquisition method for alloy materials according to any one of claims 1 to 7.

Images