Connection process quality analysis and parameter optimization system and method based on deep learning and large language model

Abstract

The invention discloses a connection process quality analysis and parameter optimization system and method based on deep learning and a large language model. An input parameter acquisition module acquires process parameters, material marks and material geometric dimension parameters of a connection process; the material database calling module obtains material attribute parameters according to the material trademark; the expert model prediction module predicts the connection quality of the connection process according to the process parameters, the material geometric dimension parameters and the material attribute parameters; the process standard comparison and analysis module compares a prediction result with a standard process requirement; when the contrastive analysis has no deviation, the big language model analysis module integrates the contrastive analysis results, and outputs a process analysis report and a quality analysis result; when the deviation exists through comparison and analysis, the intelligent optimization algorithm module optimizes and adjusts the parameters; the big language model optimization suggestion generation module integrates the optimal optimization parameter set to generate optimization suggestions; and automation and intellectualization of connection process quality analysis and optimization means are realized.

Description

Technical Field

[0001] The present invention belongs to the field of parameter optimization and quality analysis of connection processes, and specifically relates to a quality analysis and parameter optimization of connection processes based on deep learning and a large language model. The present invention can be widely used in quality control and optimization of various connection processes, including but not limited to welding, bonding, brazing, riveting and other connection technologies. Background Art

[0002] Connection processes, especially welding, brazing, riveting and bonding technologies, play a vital role in modern manufacturing, and their quality directly affects key performance of products such as strength, durability and service life. Therefore, in order to ensure product quality, welding engineers must accurately analyze and reasonably optimize various parameters during the design and control of the connection process, and companies also need to accurately predict the connection effect in product design and development to achieve product reliability. However, existing connection process quality analysis and optimization methods generally have problems with insufficient automation and intelligence, and are difficult to meet the efficient production needs of modern manufacturing.

[0003] At present, the quality analysis of connection process mostly relies on traditional prediction models for parameter analysis and quality prediction, which can usually only give some specific numerical indicators, such as joint strength, defect types, etc. After obtaining the prediction results, users need to find the relevant process standards by themselves, and manually compare the predicted values ​​with the standard values ​​to evaluate the process quality. This method is not only cumbersome and increases the workload of engineers, but also prone to errors due to human factors. At the same time, the results of the prediction model are limited to quantitative analysis, and no optimization suggestions can be provided for connectors that do not meet the quality standards. Even for qualified connectors, the system cannot provide targeted optimization and improvement directions for the process based on process standards and empirical knowledge. These deficiencies reduce the efficiency of connection process design in practical applications and also bring significant technical bottlenecks to welding engineers.

[0004] In addition, the design optimization process of the connection process often requires engineers to adjust parameters through multiple tests in order to find a suitable process solution in a complex connection process. This process relies on the experience accumulation of engineers, has strong subjective factors, high test costs, and a long time period, making large-scale automated production difficult. Since the existing technology lacks the function of optimizing recommendations for unqualified joints, it is difficult for companies to quickly and efficiently obtain the optimal parameter solution set from failed tests, resulting in repeated tests, waste of resources, and increased difficulty in quality control.

[0005] Moreover, from the perspective of the enterprise, the traditional connection process quality analysis and optimization methods rely heavily on the professional knowledge and rich field experience of welding engineers. This experience-intensive approach not only places high demands on the professional background of engineers, but also increases the recruitment and training costs of high-quality talents for enterprises. Especially when designing welding processes and selecting parameters, welding engineers need to have a comprehensive understanding of material properties, process standards, quality control, etc. in order to make effective decisions. This makes welding process design a time-consuming and complex link in the product development process, limiting the participation of non-professional welding engineers.

[0006] In view of the above shortcomings, if a more intelligent connection process quality analysis and parameter optimization method can be introduced, it can not only automatically collect and analyze various connection parameters, but also provide optimization suggestions for connectors that do not meet the requirements through interactive intelligent recommendation, combined with rich process standards and experience knowledge base, to assist engineers in quickly adjusting connection process parameters, thereby improving process quality. In particular, when the large language model is combined with the intelligent optimization algorithm, it can not only realize automatic process parameter recommendation by comparing process standards, but also intelligently interact with engineers based on rich experience knowledge, provide parameter optimization direction and personalized suggestions, realize deep interaction with users, and provide more precise guidance for connection processes.

[0007] Therefore, there is an urgent need for an intelligent quality analysis and parameter optimization method for the connection process to improve the intelligence level of the connection process, help welding engineers to perform process analysis and quality control more efficiently in design and research and development, and provide reliable technical support for enterprises, thereby effectively reducing the failure rate, improving production efficiency, and providing a systematic solution for the application and improvement of high-quality connection processes. Summary of the invention

[0008] In order to address the deficiencies in the prior art, the present invention proposes a connection process quality analysis and parameter optimization system and method based on deep learning and a large language model. The present invention performs intelligent analysis and optimization of material parameters, geometric dimensions, process conditions, etc. in the connection process, thereby ensuring the connection quality while improving the accuracy and reliability of the welding process. Unlike traditional welding process optimization methods that rely on manual experience and professional knowledge, the present invention combines the powerful predictive ability of deep learning models and the knowledge base of large language models to provide engineers with a comprehensive and intelligent welding quality optimization platform that can automatically recommend parameters and analyze quality, thereby greatly reducing manual intervention, improving process optimization efficiency, and meeting high-precision and high-efficiency industrial needs.

[0009] The technical solution adopted by the present invention is as follows:

[0010] The connection process quality analysis and parameter optimization system based on deep learning and big language model includes: input parameter acquisition module, material database call module, expert model prediction module, data experience library, process standard comparison and analysis module, big language model analysis module, intelligent optimization algorithm module and big language model optimization suggestion generation module;

[0011] The input parameter acquisition module acquires process parameters, material grades, and material geometric dimension parameters of the connection process;

[0012] The material database calling module obtains material property parameters corresponding to the material grade according to the collected material grade;

[0013] The expert model prediction module predicts the connection quality of the connection process according to the process parameters, material geometric size parameters and material property parameters, and outputs the prediction result;

[0014] The process standard comparison and analysis module compares and analyzes the prediction results with the connection process standard process requirements in the data experience library;

[0015] In the case where there is no deviation between the prediction result and the standard process requirement, the large language model analysis module integrates the comparative analysis results of the process standard comparative analysis module through the large language model, and outputs a process analysis report and a quality analysis result;

[0016] In case of deviation between the prediction results and the standard process requirements, the intelligent optimization algorithm module optimizes and adjusts the parameters to obtain the best optimization parameter set;

[0017] In the large language model optimization suggestion generation module, the large language model is used to integrate the best optimization parameter set to generate optimization suggestions.

[0018] Furthermore, the material database calling module includes a plurality of material property databases, namely, a material grade database, a material mechanical property database, a material chemical composition database, and a material thermal property database.

[0019] Furthermore, the expert model prediction module is constructed based on a deep learning model and is used to predict the connection quality of the connection process.

[0020] Furthermore, the method of constructing an expert model prediction module based on a deep learning model is:

[0021] S1. Build a deep learning model network structure, including input layer, hidden layer and output layer;

[0022] S2, collecting process parameters, material geometry parameters, material property parameters and connection quality of various connection processes, thereby constructing a data set, and using the data set to train the deep learning model network constructed in S1;

[0023] During the prediction process, process parameters, material geometric size parameters and material property parameters are used as inputs of the expert model prediction module, and the trained deep learning model network is used to predict the connection quality of the connection process.

[0024] Furthermore, the method of constructing an intelligent optimization algorithm module based on the optimization algorithm is:

[0025] S1. Taking the process parameters of the connection process, the geometric size parameters of the material, and the material property parameters as the optimization objects, the objective function is established based on the quality indicators of the connection process;

[0026] S2. Use the optimization algorithm to optimize the process parameters, material geometric size parameters, and material property parameters of the connection process to obtain the best optimization parameters.

[0027] Further, the method for generating optimization suggestions by the large language model optimization suggestion generating module is:

[0028] S1. Receive optimized process parameters and quality prediction results, and conduct intelligent analysis based on experience knowledge;

[0029] S2. Generate optimization suggestions on how to adjust process parameters and how to improve quality indicators through natural language generation technology, and output easy-to-understand text content.

[0030] Furthermore, the content of the optimization suggestion is presented as follows:

[0031] (1) Specific suggestions for adjusting key parameters such as welding time, pressure, and temperature;

[0032] (2) Optimization suggestions on material selection, joint location, process sequence, etc.;

[0033] (3) Optimization strategies for improving welding process, reducing defects, and improving joint strength and interface bonding quality.

[0034] Furthermore, the connection process includes but is not limited to a welding process, a bonding process, a brazing process, and a mechanical connection process.

[0035] Furthermore, quality indicators of the connection process include, but are not limited to, joint strength, interface bonding quality, material fusion degree, and connection defect type.

[0036] The connection process quality analysis and parameter optimization method based on deep learning and large language model includes the following steps:

[0037] Step 1: Collect process parameters, material grades, and material geometric dimension parameters of the connection process; and obtain material property parameters corresponding to the material grades according to the obtained material grades;

[0038] Step 2: Input the process parameters, material geometry parameters and material property parameters of the connection process into the deep learning model to predict the connection quality of the connection process;

[0039] Step 3: Compare and analyze the prediction results of step 2 with the standard process requirements for the connection process to determine whether there is a deviation between the prediction results and the standard; if there is no deviation, integrate the analysis results through the large language model and output the process analysis report and quality analysis results;

[0040] If there is a deviation, the optimization algorithm is used to adjust the process parameters of the connection process to obtain the best optimization parameter set; the large language model is used to integrate the best optimization parameter set to generate optimization suggestions.

[0041] Beneficial effects of the present invention:

[0042] (1) The connection process quality analysis and parameter optimization system and method based on deep learning and large language model provided by the present invention can realize accurate prediction and automatic optimization of welding quality through intelligent analysis and optimization of input parameters in the connection process. The prediction of key quality indicators such as joint strength, interface bonding quality, and material fusion degree by deep learning model greatly improves the accuracy and efficiency of welding process, significantly reduces the dependence on manual experience in traditional methods, and helps engineers quickly identify and optimize process parameters.

[0043] (2) The connection process quality analysis and parameter optimization system and method based on deep learning and large language model provided by the present invention can provide more intelligent optimization suggestions and decision support during the welding process by combining the advantages of large language model and deep learning model. The large language model generates personalized optimization suggestions based on historical experience and knowledge in the database, helping engineers to quickly find the best adjustment plan when facing complex welding problems, and further improve welding quality and production efficiency. Compared with traditional methods, the present invention can achieve breakthroughs in automation and intelligence, greatly improving the design and adjustment efficiency of welding processes.

[0044] (3) The connection process quality analysis and parameter optimization system and method based on deep learning and large language model provided by the present invention adopts a combination of deep learning model and intelligent optimization algorithm for quality assessment and process optimization in welding process. By automatically analyzing the deviation between input parameters and process standards, the system can quickly provide optimization solutions for processes that do not meet the standards, avoiding the deficiency of relying on expert experience in traditional methods. Intelligent optimization algorithms can accurately adjust process parameters through multi-objective optimization to ensure the optimal welding quality, thereby improving the efficiency and quality control capabilities of the overall production process.

[0045] (4) The connection process quality analysis and parameter optimization system and method based on deep learning and large language models provided by the present invention enable non-professional welding engineers to participate in the product design and development process, reducing the reliance on traditional welding expertise. This innovation enables companies to perform more efficient quality control and process optimization during the product design phase, thereby improving product development efficiency, shortening the development cycle, and reducing the defect rate and cost caused by improper processes during the production process. By combining deep learning with a large language model, the present invention greatly improves the intelligence of the welding process and brings a new solution to the industrial field. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the connection process quality analysis and parameter optimization method based on deep learning and large language model of the present invention.

[0047] Figure 2 This is an example diagram of the input parameter collection and preprocessing module in the example.

[0048] Figure 3 This is the structure diagram of the fully connected deep neural network in the example.

[0049] Figure 4 This is a graph showing the prediction results of the deep learning model prediction module in the example.

[0050] Figure 5 This is a graph of optimization results based on the objective function of the intelligent optimization algorithm module in the example.

[0051] Figure 6 The analysis results of the optimization structure of the large language model optimization suggestion generation module in the example. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or identification steps closely related to the scheme of the present invention are shown in the specific embodiments, while other details that are not closely related to the present invention are omitted.

[0054] In addition, it should be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.

[0055] like Figure 1 As shown in the figure, the connection process quality analysis and parameter optimization system based on deep learning and large language model includes input parameter acquisition module, material database call module, expert model prediction module, data experience library, process standard comparison and analysis module, large language model analysis module, intelligent optimization algorithm module and large language model optimization suggestion generation module. The specific work of each module is as follows:

[0056] Input parameter acquisition module, this module can at least collect various types of parameters such as process parameters of the connection process, material brand, material geometric size parameters, etc., as input parameters of subsequent modules. In this embodiment, the process parameters include welding current, welding time, welding pressure, etc.; the material geometric size parameters include thickness and other information.

[0057] The material database calling module includes multiple material property databases, namely, material grade database, material mechanical property database, material chemical composition database, material thermal property database, etc. The material property parameters corresponding to the material grade, such as mechanical properties, thermal properties, chemical composition, etc., can be obtained from the material database calling module according to the collected material grade.

[0058] The expert model prediction module is constructed based on the deep learning model and is used to predict the connection quality of the connection process. Specifically, the process parameters collected by the input parameter acquisition module, the material geometric size parameters and the material property parameters obtained in the material database call module are used as the input of the expert model prediction module, and the expert model prediction module predicts the connection quality and outputs the prediction results. Depending on the connection process, the quality indicators include but are not limited to joint strength, interface bonding quality, material fusion degree, defect type, etc.

[0059] A data experience database, in which standard process requirement data of various connection processes are pre-stored. The standard process requirements can adopt national standard data of various quality indicators, industry standard data or enterprise-defined standard data.

[0060] The process standard comparison and analysis module calls the prediction result output by the expert model prediction module, and compares and analyzes it with the standard process requirements of the corresponding connection process in the data experience library to determine whether there is a deviation between the prediction result and the standard.

[0061] In the large language model analysis module, when the prediction results output by the process standard comparison analysis module have no deviation from the standard, the large language model analysis module integrates the analysis results of the process standard comparison analysis module through the large language model, and outputs a process analysis report and quality analysis results.

[0062] In the intelligent optimization algorithm module, if the prediction results output by the process standard comparison and analysis module deviate from the standard, the optimization algorithm is called to adjust the input parameters to obtain the best optimization parameter set.

[0063] In the large language model optimization suggestion generation module, the large language model is used to integrate the best optimization parameter set to generate optimization suggestions.

[0064] Furthermore, the method of constructing an expert model prediction module based on a deep learning model is:

[0065] S1. Construction Figure 3 The network structure of the deep learning model shown includes input layer, hidden layer and output layer;

[0066] S2, collect process parameters, material geometry parameters, material property parameters and connection quality of various connection processes, thereby constructing a data set, and using the data set to train the deep learning model network constructed in S1. The trained deep learning model network can be used to predict the connection quality of the connection process.

[0067] More specifically, the deep learning model uses machine learning algorithms to extract and learn features from the temporal and spatial data in the connection process.

[0068] More specifically, the deep learning model uses regression, classification, or multi-task learning to generate multiple output indicators such as joint strength, interface bonding quality, and material fusion degree;

[0069] Furthermore, the method of constructing an intelligent optimization algorithm module based on the optimization algorithm is:

[0070] S1. Taking the process parameters of the connection process, the geometric size parameters of the material, and the material property parameters as the optimization objects, the objective function is established based on the quality indicators of the connection process;

[0071] S2. Use the optimization algorithm to optimize the process parameters, material geometric size parameters, and material property parameters of the connection process to obtain the best optimization parameters.

[0072] More specifically, the optimization algorithm may include particle swarm optimization, genetic algorithm, simulated annealing, ant colony optimization, Bayesian optimization, and the like.

[0073] Further, the method for generating process analysis report and quality analysis results in the large language model analysis module is:

[0074] S1. Receive the prediction results provided by the expert model prediction module and perform deviation analysis based on the standard process requirements provided in the data experience library;

[0075] S2. Generate a connection quality deviation analysis report based on the deviation analysis results and empirical knowledge through natural language generation technology;

[0076] More specifically, the connection quality analysis report may include a judgment on whether the connection quality characteristics such as connection morphology, connection strength, and connection defects meet the standard process requirements. It aims to provide reference and guidance for testers and designers, shorten the product design cycle, improve the quality control of the connection process, and increase production efficiency.

[0077] More specifically, the content of the connection quality analysis report may be presented in the following text content:

[0078] (1) Specific analysis results of welding morphology dimensions such as weld core diameter and width;

[0079] (2) Specific analysis results of the weld joint morphology and size, morphology defects and other joint morphology dimensions;

[0080] (3) Specific analysis results of mechanical properties such as tensile strength, shear strength, and tensile shear strength of welded joints;

[0081] (4) Specific analysis results of the process, such as whether process defects occurred during welding

[0082] Further, the method for generating optimization suggestions by the large language model optimization suggestion generating module is:

[0083] S1. Receive optimized process parameters and quality prediction results, and conduct intelligent analysis based on experience knowledge;

[0084] S2. Generate optimization suggestions on how to adjust process parameters and improve quality indicators through natural language generation technology, and output easy-to-understand text content;

[0085] More specifically, the optimization suggestions may include the adjustment direction, adjustment range, tuning strategy, etc. of the input parameters, aiming to further improve the quality control of the connection process, reduce the defect rate, and increase production efficiency.

[0086] More specifically, the content of the optimization suggestion may be presented in the following text content:

[0087] (1) Specific suggestions for adjusting key parameters such as welding time, pressure, and temperature;

[0088] (2) Optimization suggestions on material selection, joint location, process sequence, etc.;

[0089] (3) Optimization strategies for improving welding process, reducing defects, and improving joint strength and interface bonding quality.

[0090] In this embodiment, the above system can be applied to various types of connection processes, including but not limited to:

[0091] Welding processes, such as arc welding, laser welding, spot welding, etc.;

[0092] Bonding process, such as structural adhesive bonding, hot melt adhesive bonding, etc.;

[0093] Brazing process, such as soft soldering, hard soldering, etc.;

[0094] Other mechanical connection processes, such as bolting, riveting, etc.

[0095] In this embodiment, the quality indicators of the connection process include:

[0096] Joint strength, which is related to the tensile strength and shear strength of the connection joint;

[0097] Interface bonding quality, including bonding integrity, bonding area, bonding strength, etc. of the connection interface;

[0098] The degree of material fusion, specifically refers to the melting depth, diffusion depth, uniformity of the fusion zone, etc. of the material in the connection area;

[0099] Types of connection defects, including but not limited to pores, cracks, inclusions, spatter, shrinkage holes, etc.;

[0100] Other factors that affect the connection quality, such as heat affected zone, residual stress, deformation, etc.

[0101] In this embodiment, the material property database includes:

[0102] Material grade database, which contains the grade information of common materials required for various joining processes, and is used to retrieve the corresponding material attribute data according to the collected material grades;

[0103] Material mechanical properties database, including material tensile strength, shear strength, hardness, elastic modulus, elongation after fracture and other mechanical properties data, used to guide the design and quality analysis of the connection process;

[0104] Material chemical composition database, which records the chemical composition information of materials, such as element composition, alloy element ratio, etc., to help analyze the thermophysical properties of materials and their possible reactions during the joining process;

[0105] The material thermal properties database, including the material's melting point, thermal expansion coefficient, thermal conductivity and other thermal performance data, is used to evaluate the thermal behavior and melting characteristics of the material during the joining process.

[0106] The connection process quality analysis and parameter optimization method based on deep learning and large language model includes the following steps:

[0107] Step 1: Collect process parameters, material grades, and material geometric dimension parameters of the connection process; and obtain material property parameters corresponding to the material grades according to the obtained material grades;

[0108] Step 2: Input the process parameters, material geometry parameters and material property parameters of the connection process into the deep learning model to predict the connection quality of the connection process;

[0109] Step 3: Compare and analyze the prediction results of step 2 with the standard process requirements for the connection process to determine whether there is a deviation between the prediction results and the standard; if there is no deviation, integrate the comparison and analysis results through the large language model to output the process analysis report and quality analysis results;

[0110] If there is a deviation, the optimization algorithm is used to adjust the process parameters of the connection process to obtain the best optimization parameter set; the large language model is used to integrate the best optimization parameter set to generate optimization suggestions.

[0111] The following is a detailed introduction to the technical solution of the present invention using the resistance spot welding process as an example. The working process of this system is as follows:

[0112] First, the input parameter acquisition module collects various process parameters, material grades, and material geometric size parameters in the resistance spot welding process, such as welding time, welding pressure, welding current, material grade, material thickness, etc. Figure 2 shown.

[0113] The material database calling module calls corresponding material properties from multiple material property databases according to the collected material grades. In this embodiment, the material properties include tensile strength, yield strength, carbon equivalent, silicon equivalent, manganese equivalent, etc.

[0114] The process parameters and material geometric size parameters obtained by the input parameter acquisition module and the corresponding material properties retrieved by the material database calling module are input into the expert model prediction module. The expert model prediction module uses the trained deep learning model to predict the weld nugget size and mechanical properties of the resistance spot welding joint to obtain quality indicators, including weld nugget diameter, weld nugget height, indentation depth, spatter, defect, tensile shear strength, etc. Figure 4 shown.

[0115] Subsequently, the predicted quality indicators are compared and analyzed with the standard process requirements in the resistance spot welding process data experience database stored in the data experience database to evaluate whether there is a deviation between the predicted results and the standard process requirements.

[0116] If the comparative analysis results meet the standard process requirements, the analysis results are further integrated and optimized through the large language model in the large language model analysis module to output the process analysis report and quality analysis results.

[0117] If the quality analysis results do not meet the standard process requirements, the optimization algorithm is called in the intelligent optimization algorithm module to adjust the input parameters based on the ideal target, calculate the optimal welding parameter combination, and generate the optimized process parameters, such as Figure 5 In this specific implementation scheme, the ideal goal is to maximize the weld nugget diameter, minimize the indentation depth, and avoid spatter.

[0118] In the large language model optimization suggestion generation module, the large language model is used to intelligently analyze the optimized welding process parameters and prediction results to generate targeted optimization suggestions, such as Figure 6 As shown, feasible parameter adjustment directions are output to further improve the quality of the resistance spot welding process.

[0119] In this embodiment, the working principle of the intelligent optimization algorithm module includes:

[0120] S1. Through the NSGA-II genetic optimization algorithm, based on the process standard and the objective function, the objective function is to maximize the weld core diameter, minimize the indentation depth and avoid spatter, and explore the optimal solution in the input parameter space. In this case, the optimal solution set of resistance spot welding process parameters is generated as follows Figure 5 As shown;

[0121] S2. The optimization objectives include the maximum weld core diameter, the minimum indentation depth and no spatter. By adjusting the process parameters such as welding time, temperature and pressure, the target quality indicators are close to the ideal optimal values.

[0122] S3. A multi-objective optimization strategy is used in the optimization process to optimize multiple quality indicators at the same time to ensure that the connection process is optimal in multiple dimensions.

[0123] The intelligent quality analysis and parameter optimization method of the connection process based on deep learning and large language model provided by the present invention can be applied to a variety of industrial environments. It can be run on terminal devices such as industrial computers, personal computers, and smart devices, and can also be deployed in servers or server clusters for large-scale data processing and model calculation. In addition, the method can also be implemented on a cloud platform, supporting remote collaboration and data sharing, and is suitable for welding workshops, R&D laboratories, and smart manufacturing environments.

[0124] After combining the intelligent optimization technology of the large language model with the present invention, the system can automatically make qualification judgments and provide optimization suggestions based on process standards and existing data. This kind of recommendation guidance reduces the reliance on personal experience and significantly improves the intelligent level of product design. At the same time, this method can provide enterprises with a set of systematic and standardized intelligent design solutions, helping enterprises to maintain high-quality control of connection processes while reducing labor costs, realizing the transformation from experience-driven to data-driven, and bringing higher efficiency and stronger competitiveness to production and R&D.

[0125] The present invention combines a large language model with deep learning technology, which can significantly reduce this knowledge dependence. The large language model has a built-in rich connection process experience and technical knowledge base, and can intelligently provide optimization suggestions based on the design parameters and quality requirements entered by the user. This allows even engineers without a professional welding background to complete the connection process design and analysis of the product based on the model's suggestions. This technical advantage not only reduces the difficulty of design and R&D, but also shortens the product development cycle, and can effectively improve the efficiency of product R&D. More importantly, the interactivity of the model allows engineers to obtain targeted suggestions on specific issues, further improving the flexibility and accuracy of the design.

[0126] The above embodiments are only used to illustrate the design ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. A connection process quality analysis and parameter optimization system based on deep learning and large language model, characterized by: The system includes: input parameter acquisition module, material database call module, expert model prediction module, data experience library, process standard comparison and analysis module, large language model analysis module, intelligent optimization algorithm module and large language model optimization suggestion generation module; The input parameter acquisition module acquires process parameters, material grades, and material geometric dimension parameters of the connection process; The material database calling module obtains material property parameters corresponding to the material grade according to the collected material grade; The expert model prediction module predicts the connection quality of the connection process according to the process parameters, material geometric size parameters and material property parameters, and outputs the prediction result; The process standard comparison and analysis module compares and analyzes the prediction results with the connection process standard process requirements in the data experience library; In the case where there is no deviation between the prediction result and the standard process requirement, the large language model analysis module integrates the comparative analysis results of the process standard comparative analysis module through the large language model, and outputs a process analysis report and a quality analysis result; In case of deviation between the prediction results and the standard process requirements, the intelligent optimization algorithm module optimizes and adjusts the parameters to obtain the best optimization parameter set; In the large language model optimization suggestion generation module, the large language model is used to integrate the best optimization parameter set to generate optimization suggestions.

2. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The material database calling module includes a plurality of material property databases, namely a material grade database, a material mechanical property database, a material chemical composition database, and a material thermal property database.

3. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The expert model prediction module is constructed based on a deep learning model and is used to predict the connection quality of the connection process.

4. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The method of building an expert model prediction module based on a deep learning model is: S1. Build a deep learning model network structure, including input layer, hidden layer and output layer; S2, collecting process parameters, material geometry parameters, material property parameters and connection quality of various connection processes, thereby constructing a data set, and using the data set to train the deep learning model network constructed in S1; During the prediction process, process parameters, material geometric size parameters and material property parameters are used as inputs of the expert model prediction module, and the trained deep learning model network is used to predict the connection quality of the connection process.

5. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The method of building an intelligent optimization algorithm module based on the optimization algorithm is: S1. Taking the process parameters of the connection process, the geometric size parameters of the material, and the material property parameters as the optimization objects, the objective function is established based on the quality indicators of the connection process; S2. Use the optimization algorithm to optimize the process parameters, material geometric size parameters, and material property parameters of the connection process to obtain the best optimization parameters.

6. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The method for generating optimization suggestions by the large language model optimization suggestion generation module is: S1. Receive optimized process parameters and quality prediction results, and conduct intelligent analysis based on experience knowledge; S2. Generate optimization suggestions on how to adjust process parameters and how to improve quality indicators through natural language generation technology, and output easy-to-understand text content.

7. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 6 is characterized in that: The content of the optimization suggestion is presented as follows: (1) Specific suggestions for adjusting key parameters such as welding time, pressure, and temperature; (2) Optimization suggestions on material selection, joint location, process sequence, etc.; (3) Optimization strategies for improving welding process, reducing defects, and improving joint strength and interface bonding quality.

8. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The connection process includes but is not limited to a welding process, a bonding process, a brazing process, and a mechanical connection process.

9. The connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1 is characterized in that: The quality indicators of the connection process include but are not limited to joint strength, interface bonding quality, material fusion degree, and connection defect type.

10. A method for connection process quality analysis and parameter optimization based on deep learning and large language model, characterized in that: Based on the connection process quality analysis and parameter optimization system based on deep learning and large language model according to claim 1, the method comprises the following steps: Step 1: Collect process parameters, material grades, and material geometric dimension parameters of the connection process; and obtain material property parameters corresponding to the material grades according to the obtained material grades; Step 2: Input the process parameters, material geometry parameters and material property parameters of the connection process into the deep learning model to predict the connection quality of the connection process; Step 3: Compare and analyze the prediction results of step 2 with the standard process requirements for the connection process to determine whether there is a deviation between the prediction results and the standard; If there is no deviation, the analysis results are integrated through the large language model to output the process analysis report and quality analysis results; If there is a deviation, the optimization algorithm is used to adjust the process parameters of the connection process to obtain the best optimization parameter set; The large language model is used to integrate the best optimization parameter set and generate optimization suggestions.

Images