Generation method, material, structural member, vehicle, device, medium and electronic equipment

Abstract

The invention belongs to the technical field of material science, and relates to a generation method of a material formula, a target material, an automobile structural part, a vehicle, a device, a medium and electronic equipment. The method comprises the following steps: acquiring target performance data of a target material to be prepared; and inputting the target performance data into a pre-trained material large model, so that the material large model outputs formula data for preparing the target material and corresponding process data. The formula data and the process data of the preparation method are reversely deduced according to the target performance data required by the target material by utilizing the large material model, the large model is introduced into the field of material science, and the generation method of the reverse material formula realizing'target performance driving 'is provided without relying on a large number of manual experiment trials and errors; and the advantages of strong generalization ability and reasoning ability of the large material model are exerted, so that the research and development efficiency and success rate of new materials are improved, and the application scenarios of the large material model are enriched.

Description

Technical Field

[0001] The present disclosure belongs to the field of material science and technology, and in particular relates to a method for generating a material formula, a target material, an automobile structural part, a vehicle, a device for generating a material formula, a computer-readable storage medium, and an electronic device. Background Art

[0002] With the rapid development of materials science, people's demand for new materials is increasing.

[0003] Traditional material formulation design methods typically rely on trial and error, which is time-consuming, costly, and inefficient. Even the application of machine learning models and small-scale neural networks can only predict the properties of existing materials, limiting their applicability. Summary of the Invention

[0004] To overcome the problems existing in the related art, the present disclosure provides a method for generating a material formula, a target material, an automotive structural part, a vehicle, a device for generating a material formula, a computer-readable storage medium, and an electronic device.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for generating a material formula is provided, comprising: Obtaining target performance data of the target material to be prepared; The target performance data is input into a pre-trained material model so that the material model outputs the recipe data and corresponding process data for preparing the target material.

[0006] Optionally, before inputting the target performance data into a pre-trained material model, the method further comprises: Acquiring knowledge data of the field to which the sample material belongs and sample performance data of the sample material; Inputting the knowledge data into a large material model to be trained, so that the large material model to be trained outputs predicted performance data; A target loss function is determined based on the sample performance data and the predicted performance data, so as to use the target loss function to train the large material model to be trained, and the target loss function includes a cross-entropy loss function and a material loss function.

[0007] Optionally, inputting the knowledge data into a large material model to be trained so that the large material model to be trained outputs predicted performance data includes: Extracting first structured data and first unstructured data from the knowledge data using a first large model; Performing element processing on the first unstructured data to obtain second structured data and second unstructured data, and performing data preprocessing on the first structured data and the second structured data to obtain third structured data; The third structured data and the second unstructured data are input into a large material model to be trained, so that the large material model to be trained outputs predicted performance data.

[0008] Optionally, the method further includes: When the recipe data includes multiple groups, similarity calculation is performed on the recipe data and the knowledge data to obtain a target similarity; Experimental data are determined from multiple groups of recipe data according to the target similarity, and material preparation processing is performed on the experimental data according to process data corresponding to the experimental data to obtain the target material and experimental performance data.

[0009] Optionally, the method further includes: The first structured data and the experimental performance data are used to perform parameter fine-tuning on the material macro model to obtain a fine-tuned material macro model.

[0010] Optionally, inputting the target performance data into a pre-trained material model includes: The target performance data is vector-converted to obtain a performance dictionary, so as to input the performance dictionary into a pre-trained material macro model.

[0011] According to a second aspect of an embodiment of the present disclosure, a target material is provided. The target material is prepared by the method for generating the material formula according to any one of claims 1 to 6.

[0012] Optionally, the target material includes steel, and the steel is composed of the following components: C 0.29%~0.42%, Si 0.30%~0.90%, Mn 0.30%~0.90%, P≤0.100%, S≤0.100%, Cr 0.01%~0.40%, B 0.001%~0.01%, Al 0.10%~ 0.40%, Nb 0.02%~0.05%, Cu 0.1%~0.3%, V 0.05%~0.20%, and the remaining components are Fe and unavoidable impurities.

[0013] According to a third aspect of an embodiment of the present disclosure, there is provided an automotive structural part, comprising a structure at least partially formed of a target material, wherein the target material is a target material prepared by any of the material formula generation methods described above or is any of the target materials described above.

[0014] According to a fourth aspect of an embodiment of the present disclosure, a vehicle is provided, comprising the automobile structural component described in the third aspect.

[0015] According to a fifth aspect of an embodiment of the present disclosure, there is provided a device for generating a material formula, comprising: a performance determination module, configured to obtain target performance data of a target material to be prepared; The recipe generation module is configured to input the target performance data into a pre-trained material model so that the material model outputs the recipe data and corresponding process data for preparing the target material.

[0016] Optionally, the device further comprises: A sample acquisition module is configured to acquire knowledge data of the field to which the sample material belongs and sample performance data of the sample material; A model training module is configured to input the knowledge data into a large material model to be trained, so that the large material model to be trained outputs predicted performance data; The loss determination module is configured to determine a target loss function based on the sample performance data and the predicted performance data, so as to use the target loss function to train the large material model to be trained, and the target loss function includes a cross-entropy loss function and a material loss function.

[0017] Optionally, the model training module includes: a data extraction unit configured to extract first structured data and first unstructured data from the knowledge data using a first large model; a data processing unit configured to perform element processing on the first unstructured data to obtain second structured data, and perform data preprocessing on the first structured data and the second structured data to obtain third structured data and second unstructured data; The data input unit is configured to input the third structured data and the second unstructured data into the large material model to be trained, so that the large material model to be trained outputs predicted performance data.

[0018] According to a sixth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the method for generating a material formula provided in any one of the first aspects of the present disclosure are implemented.

[0019] According to a seventh aspect of the embodiments of the present disclosure, there is provided an electronic device, including: processor; a memory for storing processor-executable instructions; The processor is configured to: execute the executable instructions to implement the steps of any one of the material formula generation methods provided in the first aspect of the present disclosure.

[0020] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects: In the method and apparatus provided in the exemplary embodiments of the present disclosure, a large material model is used to reversely deduce the recipe data and process data of the preparation method based on the target performance data required by the target material, thereby introducing the large model into the field of materials science and providing a method for generating reverse material recipes that achieves "target performance drive". This method does not rely on a large number of manual experiments and trial and errors, and gives full play to the advantages of the large material model in generalization and reasoning ability, thereby greatly improving the research and development efficiency and success rate of new materials and enriching the application scenarios of the large material model.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0023] Figure 1 The following schematically shows a flow chart of a method for generating a material formula in an exemplary embodiment of the present disclosure; Figure 2 The following schematically illustrates a flow chart of a method for training a large material model in an exemplary embodiment of the present disclosure; Figure 3 Schematically illustrates a flow chart of a method for further training a large material model in an exemplary embodiment of the present disclosure; Figure 4 The following schematically shows a flow chart of a method for preparing a target material in an exemplary embodiment of the present disclosure; Figure 5 The following schematically illustrates a flow chart of a method for model training in an application scenario in an exemplary embodiment of the present disclosure; Figure 6 The following schematically illustrates a flow chart of a method for formula generation in an application scenario in an exemplary embodiment of the present disclosure; Figure 7 The following schematically illustrates a flow chart of a verification iteration method in an application scenario in an exemplary embodiment of the present disclosure; Figure 8 A schematic structural diagram of a device for generating a material formula in an exemplary embodiment of the present disclosure is schematically shown; Figure 9A schematic diagram schematically illustrates the structure of another device for generating a material formula in an exemplary embodiment of the present disclosure; Figure 10 The following schematically shows the structure of a device for generating another material formula in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0025] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the corresponding data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0026] With the rapid development of materials science, people's demand for new materials is increasing.

[0027] Traditional material formulation design usually relies on trial and error, which is time-consuming, costly and inefficient.

[0028] In recent years, the rapid development of artificial intelligence technology has provided new solutions for materials science. In particular, the application of large models in data-driven material property prediction and formula generation has shown great potential.

[0029] Large models have been widely used in fields like natural language processing and image recognition, but their application in materials science is still in its infancy, particularly in the area of ​​inverse material formulation generation. These technologies primarily rely on simple machine learning models or empirical formulas, which can lead to high computational resource requirements, low accuracy in generated formulations, and difficulty handling complex material systems.

[0030] According to research, most of the existing publicly available technologies are based on traditional machine learning or specialized small-scale neural networks to predict material properties, such as neural network prediction and optimization algorithms such as genetic algorithms, and then correct parameters through a large number of experiments during the exploration process.

[0031] However, these methods are mainly limited to "formulation first, prediction later" and lack the "reverse recipe generation" function based on large-scale models. They also have shortcomings in data scale, model generalization ability and the scope of applicable material systems.

[0032] Therefore, related methods mostly remain at the prediction level and cannot reversely deduce material formulations from “target performance”; Related methods often rely on small-scale or specialized networks, which are insufficient for generalization across a broad spectrum of material systems. The accuracy and feasibility of the generated results are low, and their scope of application is relatively limited. Related methods rarely provide highly automated and scalable generation mechanisms, lack end-to-end automated processes, and are difficult to iterate quickly or collaborate with laboratory verification, resulting in long R&D cycles and low efficiency.

[0033] In response to the problems existing in the related art, the present disclosure provides a method for generating a material formula. Figure 1 is a flow chart of a method for generating a material formula according to an exemplary embodiment. Figure 1 As shown, the method may include at least the following steps: Step S110: Obtain target performance data of the target material to be prepared.

[0034] Step S120: Input the target performance data into the pre-trained material macro model, so that the material macro model outputs the recipe data and corresponding process data for preparing the target material.

[0035] In the exemplary embodiments of the present disclosure, a large material model is used to reversely deduce the recipe data and process data of the preparation method based on the target performance data required by the target material, and the large model is introduced into the field of materials science. This provides a method for generating reverse material formulas that achieves "target performance drive". It does not require reliance on a large number of manual experiments and trial and errors, and gives full play to the advantages of the large material model's strong generalization and reasoning capabilities, which greatly improves the efficiency and success rate of new material research and development, and enriches the application scenarios of the large material model.

[0036] The following is a detailed description of each step of the method for generating a material formula.

[0037] In step S110 , target performance data of the target material to be prepared is obtained.

[0038] In an exemplary embodiment of the present disclosure, a user or R&D personnel can input target performance data that characterizes the performance requirements of the target material through a human-computer interface. For example, the target performance data can be "elastic modulus ≥ 100 GPa, conductivity ≥ 5×10^5 S / m, high temperature resistance 400°C", or other data set according to actual needs. This exemplary embodiment does not specifically limit this.

[0039] In step S120, the target performance data is input into a pre-trained material macro model, so that the material macro model outputs the recipe data and corresponding process data for preparing the target material.

[0040] In an exemplary embodiment of the present disclosure, the pre-trained large material model may be obtained by training an open source pre-training model, such as Qwen2.5-70b, etc., and this exemplary embodiment does not impose any special limitation on this.

[0041] In an alternative embodiment, Figure 2 A flow chart of the method for training a large model of materials is shown, as Figure 2 As shown, the method may include at least the following steps: in step S210, knowledge data of the field to which the sample material belongs and sample performance data of the sample material are acquired.

[0042] The knowledge data may be data from a patent database, a literature database, and internal enterprise experimental data records. At the same time, performance indicator data extracted therefrom may also be used as sample performance data.

[0043] In step S220 , the knowledge data is input into the large material model to be trained, so that the large material model to be trained outputs predicted performance data.

[0044] In an alternative embodiment, Figure 3 A flow chart of a method for further training a large material model is shown, such as Figure 3 As shown, the method may at least include the following steps: in step S310, first structured data and first unstructured data in the knowledge data are extracted using a first large model.

[0045] Using a first-level model such as LiBai, we can extract data from knowledge data related to development objectives, such as raw material specifications, ratios, and preparation processes. The extracted data can be divided into two parts: first structured data and first unstructured data. This first unstructured data can include patent specifications, descriptive paragraphs in literature, and so on.

[0046] In step S320, the first unstructured data is element-processed to obtain second structured data and second unstructured data, and the first structured data and the second structured data are pre-processed to obtain third structured data.

[0047] For the first unstructured data, natural language processing technology, such as text cleaning, word segmentation, entity recognition and relationship extraction, can be used to perform element processing to obtain the corresponding second structured data and second unstructured data.

[0048] Furthermore, the extracted first structured data and the second structured data obtained by element processing are further cleaned and standardized, such as unit conversion, synonym merging, etc., to obtain corresponding third structured data.

[0049] In step S330 , the third structured data and the second unstructured data are input into the large material model to be trained, so that the large material model to be trained outputs predicted performance data.

[0050] For the initial parameters of the large-scale language model based on the Transformer architecture, an open source pre-training model, such as Qwen2.5-70b, is used, and then secondary pre-training is performed on knowledge data in the field of materials science.

[0051] Specifically, multimodal data such as the third structured data and the second unstructured data are input into the large material model to be trained, and the large material model to be trained can output predictive performance data that characterizes the multidimensional understanding and contextual representation of the sample material.

[0052] In step S230, a target loss function is determined based on the sample performance data and the predicted performance data, so as to complete the training of the large material model to be trained using the target loss function. The target loss function includes a cross-entropy loss function and a material loss function.

[0053] During the training process, we can introduce upper loss functions, such as MLE (Maximum Likelihood Estimate), and material loss functions that characterize the loss of material property prediction, such as MAE (Mean Absolute Error) or MSE (Mean-Square Error), as target loss functions to jointly constrain the difference between the predicted performance data and the sample performance data. We can use the target loss function to train a pre-trained large material model, thereby improving the ability to understand and generate material properties.

[0054] In an optional embodiment, the target performance data is subjected to vector conversion processing to obtain a performance dictionary, so as to input the performance dictionary into a pre-trained material macro model.

[0055] Before calling the pre-trained material model, the target performance data representing the requirements can be converted into a performance dictionary represented by a vector (value) or a topic label (key), so as to input the performance dictionary into the pre-trained material model.

[0056] The pre-trained material model can infer the recipe data and corresponding process data for preparing the target material from the internal semantic representation combined with the knowledge data of the corresponding field and the generation strategy based on probability distribution.

[0057] The candidate recipe data may include the proportion of main components, selection of auxiliary materials, etc., and the process data may include the recommended parameters of the preparation, sintering or curing process, such as temperature, time and pressure.

[0058] In an alternative embodiment, Figure 4 A schematic flow chart of a method for preparing a target material is shown, Figure 4 As shown, the method may at least include the following steps: in step S410, when the recipe data includes multiple groups, similarity calculation is performed on the recipe data and the knowledge data to obtain a target similarity.

[0059] When generating recipe data for preparing a target material, the material macro model may generate multiple sets of recipe data and corresponding process data. In this case, the similarity between each of the generated sets of recipe data and the knowledge data can be calculated to obtain a target similarity. This similarity can be calculated using cosine similarity or other similarity, which is not specifically limited in this exemplary embodiment.

[0060] In step S420, experimental data are determined from multiple sets of recipe data according to target similarity, and material preparation processing is performed on the experimental data according to process data corresponding to the experimental data to obtain target materials and experimental performance data.

[0061] After obtaining the target similarity, the size of multiple groups of target similarities can be compared to determine the set of recipe data with the highest similarity as the experimental data and recommended to the user for subsequent experimental verification.

[0062] The target material is prepared in the laboratory using automated preparation equipment or by the R&D team according to the process data and experimental data corresponding to the experimental data. The target material is then tested for performance to obtain experimental performance data. Furthermore, experimental performance data, such as compliance and performance achievement, can be transmitted back.

[0063] In an optional embodiment, the first structured data and the experimental performance data are used to perform parameter fine-tuning on the material macro model to obtain a fine-tuned material macro model.

[0064] After obtaining the experimental performance data, the characteristic vector of the recipe data, ie, the first structured data, can be automatically retrieved to perform local parameter correction on the material macro model in combination with the experimental error analysis to obtain a fine-tuned material macro model.

[0065] Specifically, LoRA (Low-Rank Adaptation, a training technology for fine-tuning the Stable Diffusion model) can be used to fine-tune the large material model, or fine-tune it through other methods, which is not particularly limited in this exemplary embodiment.

[0066] In addition, in an exemplary embodiment of the present disclosure, a target material is also provided. The target material is prepared by the method for generating a material formula.

[0067] In an optional embodiment, the target material includes steel consisting of the following composition: C 0.29%~0.42%, Si 0.30%~0.90%, Mn 0.30%~0.90%, P≤0.100%, S≤0.100%, Cr 0.01%~0.40%, B 0.001%~0.01%, Al 0.10%~ 0.40%, Nb 0.02%~0.05%, Cu 0.1%~0.3%, V 0.05%~0.20%, and the remainder being Fe and unavoidable impurities.

[0068] The following is a detailed description of the method for generating a material formula in an embodiment of the present disclosure in conjunction with an application scenario.

[0069] Figure 5 A flow chart of the model training method in the application scenario is shown, as Figure 5 As shown, raw material specifications, ratios, preparation processes and corresponding performance index data are extracted from patent databases, literature databases and internal enterprise experimental records.

[0070] For unstructured texts, such as patent specifications and descriptive paragraphs in literature, natural language processing technologies are used, such as text cleaning, word segmentation, entity recognition, relationship extraction and other factor processing methods, to form structured data.

[0071] Furthermore, the structured data is further cleaned and standardized, such as unit conversion and synonym merging.

[0072] A large-scale language model based on the Transformer architecture is used. The initial model parameters adopt the open source pre-trained model Qwen2.5-70b, and then secondary pre-training (fine-tuning) is performed on the materials field corpus and experimental database.

[0073] For this large material model, the input is multimodal information such as material formula text, process parameters, and performance description; the output is a multi-dimensional understanding and contextual representation of the target material performance.

[0074] The training data is characterized by multiple sources and types, including chemical formulas, text descriptions, experimental processes, temperature or pressure data, etc. For example, in steel development, there are about 9,000 records.

[0075] The feedback mechanism of this large material model is to introduce cross-entropy loss during the training process and combine it with material property prediction loss to jointly constrain it, thereby improving the ability to understand and generate material properties.

[0076] Among them, mapping the high-quality material recipe text and performance data required for training, and incorporating constraints on the rationality of the recipe into the loss function, are key steps in solving the problem of "generated solutions being unexecutable" in related technologies.

[0077] Figure 6 A flow chart of a method for generating a recipe in an application scenario is shown, Figure 6 As shown, users or R&D personnel input the target material performance requirements through the human-machine interface, such as "elastic modulus ≥ 100 GPa, conductivity ≥ 5×10^5S / m, high temperature resistance 400℃", etc.

[0078] Furthermore, these requirements can be converted into vector representations or topic labels to facilitate subsequent calls to large models.

[0079] When a large model that has been fine-tuned is called, after receiving the target performance requirements, the model combines knowledge data from its internal semantic representation to infer the required basic components, such as metal, alloy, ceramic or polymer substrates and modifying additives.

[0080] This large model outputs several candidate recipes based on the generation strategy of probability distribution, including the proportion of main components, auxiliary material selection, etc., and gives recommended preparation, sintering or curing process parameters such as temperature, time, pressure, etc.

[0081] Among them, the key step in formula generation is that the large model will automatically filter out unreasonable formulas according to the compatibility of the material matrix, modifier, and process feasibility constraints during the generation process, so as to meet the requirements of preparability and safety.

[0082] While generating candidate recipes, the model gives each solution a preliminary score. The scoring sources include estimated performance and the success rate of similar historical cases. Therefore, recipes with higher scores are recommended to users for subsequent experimental verification.

[0083] Figure 7 A flow chart of the verification iteration method in an application scenario is shown, as Figure 7 As shown, candidate formulas are prepared and performance tested on a laboratory scale by automated preparation equipment or by the R&D team, and the actual test results, such as compliance and performance achievement, are fed back to the system.

[0084] When the actual test results deviate from the expected ones, the system will automatically retrieve the characteristic vector of the formula and perform local parameter corrections on the model in combination with experimental error analysis to achieve fine-tuning of the large model.

[0085] When the coverage is wider, new data will be added to the training set for periodic iterative training, so that the model can continuously improve the accuracy of reverse recipe generation.

[0086] In this application scenario, data collection and preprocessing are the basis of the material formula generation method. If there is a lack of reliable data sources, the formula generated in subsequent steps may not be feasible.

[0087] In the fine-tuning process of large-scale language models, domain knowledge data and material performance-driven loss functions are introduced, which is the core difference in solving the lack of generalization and inoperability problems in related technologies.

[0088] In addition, the multiple-round iterative loop of experiments and model corrections is an important feature of this method to ensure accuracy and practicality, gradually approaching the target performance requirements in small steps.

[0089] When this method is applied to the formulation development of steel, the steel shown in Table 1 can be obtained:

[0090] Table 1 The composition comparison of the two steels prepared according to the material formulation generation method and the other two comparison steels is shown in Table 2:

[0091] Table 2 The martensitic transformation starting temperature Ms satisfies Ms≥335°C; where Ms=500-320[C]-50[Mn]-30[Cr] -5([Cu]+[Si]).

[0092] The mechanical properties and three-point bending test results of the two steels prepared according to the material formulation generation method and the other two comparison steels are shown in Table 3:

[0093] Table 3 It can be seen that when the material formula generation method is applied to the formulation development of steel, a low-cost formula that meets both strength and toughness requirements can be obtained.

[0094] Not only that, this method can also be further applied to various industrial and scientific research occasions with high demand for new materials, including but not limited to: chemical formula development, polymer material design, electronic packaging material formulation, metal alloy composition screening, biomedical materials, etc.

[0095] For example, in a new material R&D center, R&D personnel put forward several indicator requirements for the target material, such as elastic modulus, thermal conductivity, electrical conductivity, corrosion resistance, etc. The material formula generation method will use the trained large model to understand the required properties of the target material, and automatically generate several candidate material formulas and corresponding preparation processes through the extraction, matching and reasoning of multi-source data, including literature, patents, databases, etc., for R&D personnel to further evaluate or conduct small-scale tests.

[0096] In the exemplary embodiment of the present disclosure, the reverse material formula generation of "target performance driven" is truly realized without relying on a lot of manual trial and error. By combining large-scale language models with domain knowledge data, the adaptability of technical solutions to various material systems and multiple performance requirements has been significantly improved, significantly shortening the R&D cycle.

[0097] At the same time, it provides an end-to-end feasible process, including a closed-loop mechanism from data collection, model training, recipe generation to experimental verification, which greatly reduces errors and delays caused by human intervention.

[0098] The overall process supports multiple iterative optimizations, which can quickly screen out the most promising formula combinations in the early stages of R&D and reduce unnecessary waste of resources.

[0099] Furthermore, it can be seamlessly connected with automated preparation and testing platforms. Through real-time data feedback and dynamic adjustment of model parameters, the model accuracy and stability will continue to improve, resulting in a qualitative leap in the efficiency and success rate of new material research and development.

[0100] In addition, in an exemplary embodiment of the present disclosure, an automotive structural part is also provided, including a structure at least partially formed of a target material, wherein the target material is a target material prepared by any of the material formula generation methods or is any of the target materials.

[0101] Furthermore, in an exemplary embodiment of the present disclosure, a vehicle is provided, comprising the above-mentioned automobile structural component.

[0102] In addition, in an exemplary embodiment of the present disclosure, a device for generating a material formula is also provided. Figure 8 A schematic diagram of the structure of a device for generating a material formula is shown, Figure 8 As shown, the material formula generation device 800 may include: a performance determination module 810 and a formula generation module 820. The performance determination module 810 is configured to obtain target performance data of the target material to be prepared; The recipe generation module 820 is configured to input the target performance data into a pre-trained material model, so that the material model outputs the recipe data and corresponding process data for preparing the target material.

[0103] In some embodiments of the present disclosure, the material formula generating device 800 is further configured to: Acquiring knowledge data of the field to which the sample material belongs and sample performance data of the sample material; Inputting the knowledge data into a large material model to be trained, so that the large material model to be trained outputs predicted performance data; A target loss function is determined based on the sample performance data and the predicted performance data, so as to use the target loss function to train the large material model to be trained, and the target loss function includes a cross-entropy loss function and a material loss function.

[0104] In some embodiments of the present disclosure, the material formula generating device 800 is further configured to: Extracting first structured data and first unstructured data from the knowledge data using a first large model; Performing element processing on the first unstructured data to obtain second structured data and second unstructured data, and performing data preprocessing on the first structured data and the second structured data to obtain third structured data; The third structured data and the second unstructured data are input into a large material model to be trained, so that the large material model to be trained outputs predicted performance data.

[0105] In some embodiments of the present disclosure, the material formula generating device 800 is further configured to: When the recipe data includes multiple groups, similarity calculation is performed on the recipe data and the knowledge data to obtain a target similarity; Experimental data are determined from multiple groups of recipe data according to the target similarity, and material preparation processing is performed on the experimental data according to process data corresponding to the experimental data to obtain the target material and experimental performance data.

[0106] In some embodiments of the present disclosure, the material formula generating device 800 is further configured to: The first structured data and the experimental performance data are used to perform parameter fine-tuning on the material macro model to obtain a fine-tuned material macro model.

[0107] In some embodiments of the present disclosure, the recipe generation module 820 is configured to: The target performance data is vector-converted to obtain a performance dictionary, so as to input the performance dictionary into a pre-trained material macro model.

[0108] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0109] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method for generating a material formula provided by the present disclosure.

[0110] Figure 9 FIG1 is a block diagram of another device 900 for generating a material recipe according to an exemplary embodiment. For example, the device 900 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0111] Reference Figure 9 , the apparatus 900 may include one or more of the following components: a processing component 902 , a memory 904 , a power component 906 , a multimedia component 908 , an audio component 910 , an input / output interface 912 , a sensor component 914 , and a communication component 916 .

[0112] The processing component 902 generally controls the overall operation of the device 900, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 902 may include one or more processors 920 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 902 may include one or more modules to facilitate interaction between the processing component 902 and other components. For example, the processing component 902 may include a multimedia module to facilitate interaction between the multimedia component 908 and the processing component 902.

[0113] The memory 904 is configured to store various types of data to support the operations of the device 900. Examples of such data include instructions for any application or method operating on the device 900, contact data, phone book data, messages, pictures, videos, etc. The memory 904 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0114] The power supply component 906 provides power to the various components of the device 900. The power supply component 906 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 900.

[0115] The multimedia component 908 includes a screen that provides an output interface between the device 900 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensors can not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide action. In some embodiments, the multimedia component 908 includes a front-facing camera and / or a rear-facing camera. When the device 900 is in an operating mode, such as a capture mode or a video mode, the front-facing camera and / or the rear-facing camera can receive external multimedia data. Each front-facing camera and the rear-facing camera can have a fixed optical lens system or have focal length and optical zoom capabilities.

[0116] The audio component 910 is configured to output and / or input audio signals. For example, the audio component 910 includes a microphone (MIC) that is configured to receive external audio signals when the device 900 is in an operating mode, such as a call mode, a recording mode, or a voice recognition mode. The received audio signals may be further stored in the memory 904 or transmitted via the communication component 916. In some embodiments, the audio component 910 also includes a speaker for outputting audio signals.

[0117] The input / output interface 912 provides an interface between the processing component 902 and peripheral interface modules, such as a keyboard, a click wheel, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0118] The sensor assembly 914 includes one or more sensors for providing various aspects of the status assessment of the device 900. For example, the sensor assembly 914 can detect the open / closed state of the device 900, the relative positioning of components, such as the display and keypad of the device 900. The sensor assembly 914 can also detect changes in the position of the device 900 or a component of the device 900, the presence or absence of user contact with the device 900, the orientation or acceleration / deceleration of the device 900, and temperature changes of the device 900. The sensor assembly 914 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 914 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 914 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0119] The communication component 916 is configured to facilitate wired or wireless communication between the apparatus 900 and other devices. The apparatus 900 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 916 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 916 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0120] In an exemplary embodiment, the apparatus 900 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0121] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 904 including instructions. The instructions can be executed by the processor 920 of the apparatus 900 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0122] In addition to being an independent electronic device, the above-mentioned device can also be part of an independent electronic device. For example, in one embodiment, the device can be an integrated circuit (IC) or a chip, where the integrated circuit can be a single IC or a collection of multiple ICs. The chip can include, but is not limited to, the following types: GPU (Graphics Processing Unit), CPU (Central Processing Unit), FPGA (Field Programmable Gate Array), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), SOC (System on Chip, SoC), etc. The above-mentioned integrated circuit or chip can be used to execute executable instructions (or code) to implement the above-mentioned method for generating a material formula. The executable instructions can be stored in the integrated circuit or chip, or obtained from other devices or equipment, such as the integrated circuit or chip including a processor, memory, and an interface for communicating with other devices. The executable instruction can be stored in the memory, and when the executable instruction is executed by the processor, the above-mentioned method for generating the material formula is implemented; alternatively, the integrated circuit or chip can receive the executable instruction through the interface and transmit it to the processor for execution to implement the above-mentioned method.

[0123] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program executable by a programmable device, and the computer program has a code portion for executing the above-mentioned method for generating a material formulation when executed by the programmable device.

[0124] Figure 10 FIG. 1 is a block diagram of another device 1000 for generating a material formula according to an exemplary embodiment. For example, the device 1000 may be provided as a server. Figure 10 The apparatus 1000 includes a processing component 1022, which further includes one or more processors and memory resources represented by a memory 1032 for storing instructions, such as an application, that can be executed by the processing component 1022. The application stored in the memory 1032 can include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1022 is configured to execute the instructions to perform the above-described method for generating a material recipe.

[0125] Device 1000 may also include a power supply component 1026 configured to perform power management of device 1000, a wired or wireless network interface 1050 configured to connect device 1000 to a network, and an input / output interface 1058. Device 1000 may operate based on an operating system stored in memory 1032.

[0126] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the present disclosure. This disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0127] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for generating a material formula, characterized in that: include: Obtaining target performance data of the target material to be prepared; The target performance data is input into a pre-trained material model so that the material model outputs the recipe data and corresponding process data for preparing the target material.

2. The method for generating a material formula according to claim 1, characterized in that: Before inputting the target performance data into the pre-trained material model, the method further includes: Acquiring knowledge data of the field to which the sample material belongs and sample performance data of the sample material; Inputting the knowledge data into a large material model to be trained, so that the large material model to be trained outputs predicted performance data; A target loss function is determined based on the sample performance data and the predicted performance data, so as to use the target loss function to train the large material model to be trained, and the target loss function includes a cross-entropy loss function and a material loss function.

3. The method for generating a material formula according to claim 2, characterized in that: The step of inputting the knowledge data into a large material model to be trained so that the large material model to be trained outputs predicted performance data comprises: Extracting first structured data and first unstructured data from the knowledge data using a first large model; Performing element processing on the first unstructured data to obtain second structured data and second unstructured data, and performing data preprocessing on the first structured data and the second structured data to obtain third structured data; The third structured data and the second unstructured data are input into a large material model to be trained, so that the large material model to be trained outputs predicted performance data.

4. The method for generating a material formula according to claim 3, wherein: The method further comprises: When the recipe data includes multiple groups, similarity calculation is performed on the recipe data and the knowledge data to obtain a target similarity; Experimental data are determined from multiple groups of recipe data according to the target similarity, and material preparation processing is performed on the experimental data according to process data corresponding to the experimental data to obtain the target material and experimental performance data.

5. The method for generating a material formula according to claim 4, characterized in that: The method further comprises: The first structured data and the experimental performance data are used to perform parameter fine-tuning on the material macro model to obtain a fine-tuned material macro model.

6. The method for generating a material formula according to claim 1, characterized in that: Inputting the target performance data into a pre-trained material model includes: The target performance data is vector-converted to obtain a performance dictionary, so as to input the performance dictionary into a pre-trained material macro model.

7. A target material, characterized in that The target material is prepared by the method for generating the material formula according to any one of claims 1 to 6.

8. The method for generating a material formula according to claim 7, characterized in that: The target material includes steel, which is composed of the following components: C 0.29%~0.42%, Si 0.30%~0.90%, Mn 0.30%~0.90%, P≤0.100%, S≤0.100%, Cr 0.01%~0.40%, B 0.001%~0.01%, Al 0.10%~ 0.40%, Nb 0.02%~0.05%, Cu0.1%~0.3%, V 0.05%~0.20%, and the remaining components are Fe and unavoidable impurities.

9. An automobile structural part, characterized in that: The invention comprises a structure at least partially formed by a target material, wherein the target material is a target material prepared by the method for generating the material formula according to any one of claims 1 to 6 or a target material according to any one of claims 7 to 8.

10. A vehicle, characterized in that: The automotive structural component comprises the automotive structural component according to claim 9.

11. A device for generating a material formula, characterized in that: include: a performance determination module, configured to obtain target performance data of a target material to be prepared; The recipe generation module is configured to input the target performance data into a pre-trained material model so that the material model outputs the recipe data and corresponding process data for preparing the target material.

12. The material formula generating device according to claim 11, characterized in that: The device further comprises: A sample acquisition module is configured to acquire knowledge data of the field to which the sample material belongs and sample performance data of the sample material; A model training module is configured to input the knowledge data into a large material model to be trained, so that the large material model to be trained outputs predicted performance data; The loss determination module is configured to determine a target loss function based on the sample performance data and the predicted performance data, so as to use the target loss function to train the large material model to be trained, and the target loss function includes a cross-entropy loss function and a material loss function.

13. The material formula generating device according to claim 12, characterized in that: The model training module includes: a data extraction unit configured to extract first structured data and first unstructured data from the knowledge data using a first large model; a data processing unit configured to perform element processing on the first unstructured data to obtain second structured data, and perform data preprocessing on the first structured data and the second structured data to obtain third structured data and second unstructured data; The data input unit is configured to input the third structured data and the second unstructured data into the large material model to be trained, so that the large material model to be trained outputs predicted performance data.

14. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

15. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 6.

Images