Manufacturing industry product full life cycle decision analysis method based on large model

Abstract

The invention provides a manufacturing industry product full-life-cycle decision analysis method based on a large model, and the method comprises the steps: generating an sql query statement based on a pre-trained large model according to a product decision analysis demand of a user; on the basis of the sql query statement, querying corresponding structured data in a product full life cycle database; based on the pre-trained large model, analyzing the structured data, and determining a product operation decision suggestion; and outputting the structured data and the product operation decision suggestions to a user. According to the method, the pre-trained large model is introduced, the sql query statement is generated according to the product decision analysis requirement of the user, the structured data are called based on the sql query statement, the structured data are analyzed based on the pre-trained large model, and the product operation decision suggestions are determined. And the user can quickly make a product full-life-cycle decision according to the product operation decision suggestions.

Description

Technical Field

[0001] The present invention relates to the technical field of computer data processing, and in particular to a large-model-based full-life cycle decision analysis method for manufacturing products. Background Art

[0002] With the transformation and upgrading of the manufacturing industry, enterprises need to analyze the market and customer needs more accurately, so as to make product life cycle decisions that are more in line with market needs. Correspondingly, in order to more scientifically analyze the market position and sales of products and more accurately grasp the needs of the market and customers, it is often necessary to systematically process a large amount of data, integrate, screen and analyze through appropriate algorithms, so as to make rigorous decisions.

[0003] With the continuous development of big data technology, more and more companies have begun to realize the important role of data in decision-making. On this basis, processing data from different sources such as the market, customers, and the company itself to make reasonable decisions has become an important issue that companies need to solve.

[0004] Therefore, there is an urgent need for a data-based decision analysis method to assist enterprises in making rapid decisions on the entire life cycle of manufacturing products. Summary of the invention

[0005] The present invention provides a large-model-based decision analysis method for the entire life cycle of manufacturing products, aiming to solve at least one of the technical problems existing in the prior art.

[0006] In a first aspect, an embodiment of the present invention provides a manufacturing product life cycle decision analysis method based on a large model, comprising:

[0007] Based on the pre-trained large model, generate SQL query statements according to the user's product decision analysis needs;

[0008] Based on the SQL query statement, the corresponding structured data is queried in the product life cycle database;

[0009] Based on the pre-trained big model, the structured data is analyzed to determine product management decision suggestions;

[0010] The structured data and the product management decision suggestion are output to the user.

[0011] Optionally, the process of obtaining the pre-trained large model includes the following steps:

[0012] Obtain training samples; training samples are product data of a large number of different manufacturing products in various stages of research and development, production, sales, and after-sales;

[0013] Preprocessing the training samples to obtain preprocessing results;

[0014] Based on the preprocessing results, training the large model to be trained;

[0015] When the large model to be trained is trained to convergence based on the preprocessing result, the large model to be trained that has been trained to convergence is used as the pre-trained large model.

[0016] Optionally, the preprocessing of the training samples to obtain preprocessing results includes: performing the processing steps of deduplication, eliminating erroneous or invalid data, and standardization on the training samples in sequence.

[0017] Optionally, the pre-trained large model generates SQL query statements according to the user's product decision analysis requirements, including:

[0018] Inputting the user's product decision analysis requirements into the vector output component in the pre-trained large model to obtain the prompt words output by the vector output component;

[0019] Based on natural language processing technology, the prompt words are parsed to obtain query constraints and key information; wherein the query constraints at least include: data generation time range conditions, data type conditions, data relevance conditions, data sorting conditions and data filtering conditions; the key information at least includes: product name and product life cycle stage;

[0020] The query constraints and the key information are input into the pre-trained big model to obtain the SQL query statement output by the pre-trained big model.

[0021] Optionally, the process of training the large model to be trained based on the preprocessing result further includes:

[0022] Based on the model update mechanism, obtaining sample update data of the preprocessing result;

[0023] Based on the preprocessing result and the sample update data, jointly training the large model to be trained;

[0024] and / or,

[0025] Adjust the feedback mechanism based on the model and obtain usage feedback information;

[0026] Adjusting the preprocessing result based on the usage feedback information;

[0027] Based on the adjusted preprocessing results, the large model to be trained is trained.

[0028] Optionally, when the large model to be trained is trained to convergence based on the training sample, the large model to be trained that has been trained to convergence is used as the pre-trained large model, further comprising:

[0029] Based on the model verification mechanism, the large model to be trained that has been trained to convergence is verified to obtain the model verification result;

[0030] Based on the model verification result, the large model to be trained that has been trained to convergence is corrected, and the large model to be trained that has been trained to convergence after the model correction is used as the pre-trained large model.

[0031] In a second aspect, an embodiment of the present invention provides a manufacturing product life cycle decision analysis system based on a large model, including:

[0032] The generation module is used to generate SQL query statements based on the pre-trained large model and the user's product decision analysis needs;

[0033] A query module, used to query the corresponding structured data in the product life cycle database based on the SQL query statement;

[0034] A determination module, used to analyze the structured data based on the pre-trained large model to determine product management decision suggestions;

[0035] The output module is used to output the structured data and the product management decision suggestions to the user.

[0036] In a third aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising:

[0037] at least one processor; and

[0038] A memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned large model-based manufacturing product full life cycle decision analysis methods.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement any of the above-mentioned large-model-based manufacturing product full life cycle decision analysis methods when executed.

[0040] In a fifth aspect, an embodiment of the present invention provides a computer program product, which includes computer instructions. When executed by an electronic device, the electronic device executes the computer program code of any of the above-mentioned methods for decision analysis of the entire life cycle of manufacturing products based on large models.

[0041] This application has achieved the following beneficial effects:

[0042] A pre-trained big model is introduced to generate SQL query statements according to the user's product decision analysis needs, and structured data is retrieved based on the SQL query statements. The structured data is analyzed based on the pre-trained big model to determine product management decision recommendations. Users can quickly make decisions throughout the product life cycle based on the product management decision recommendations.

[0043] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0044] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0046] Figure 1 It is a flowchart of a method for decision analysis of the entire life cycle of manufacturing products based on a large model in an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of a specific application implementation of a large model-based manufacturing product full life cycle decision analysis method in an embodiment of the present invention;

[0048] Figure 3 It is another specific application implementation schematic diagram of a manufacturing product full life cycle decision analysis method based on a large model in an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of a manufacturing product full life cycle decision analysis system based on a large model in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0051] The embodiment of the present invention provides a manufacturing product life cycle decision analysis method based on a large model, such as Figure 1 As shown, including:

[0052] Step S1: Generate SQL query statements based on the pre-trained large model and the user's product decision analysis needs;

[0053] Step S2: Based on the SQL query statement, query the corresponding structured data in the product life cycle database;

[0054] Step S3: Analyze the structured data based on the pre-trained large model to determine product management decision suggestions;

[0055] Step S4: output the structured data and the product management decision suggestion to the user.

[0056] In the above technical solution:

[0057] The user's product decision analysis demand refers to the user's need to understand the business decision suggestions for manufacturing products at different stages, such as the need for business suggestions for a manufacturing product at the sales stage, etc.; using the pre-trained big model, generate SQL query statements based on the user's product decision analysis needs; the product life cycle database has the status data of various manufacturing products of the user's company at different life cycles; based on the SQL query statement, the basis for the pre-trained big model to make business decision suggestions, that is, structured data, can be retrieved from the product life cycle database. For example: the user's product decision analysis demand is the demand for business suggestions for a manufacturing product at the sales stage, then based on the SQL query statement, the corresponding structured data is queried from the product life cycle database, such as the current production status of the manufacturing product, product attribute information, etc.; the pre-trained big model can analyze the structured data to determine the product business decision suggestions; output structured data and product business decision suggestions to the user, and the user can perform corresponding manufacturing business management based on the product business decision suggestions.

[0058] This application introduces a pre-trained big model to generate SQL query statements based on the user's product decision analysis needs, and retrieves structured data based on the SQL query statements. The structured data is analyzed based on the pre-trained big model to determine product management decision recommendations. Users can quickly make product life cycle decisions based on the product management decision recommendations.

[0059] In one embodiment, the process of obtaining the pre-trained large model includes the following steps:

[0060] Obtain training samples; training samples are product data of a large number of different manufacturing products in various stages of research and development, production, sales, and after-sales;

[0061] Preprocessing the training samples to obtain preprocessing results;

[0062] Based on the preprocessing results, training the large model to be trained;

[0063] When the large model to be trained is trained to convergence based on the preprocessing result, the large model to be trained that has been trained to convergence is used as the pre-trained large model;

[0064] The preprocessing of the training samples to obtain preprocessing results includes: performing the processing steps of deduplication, eliminating erroneous or invalid data, and standardization on the training samples in sequence.

[0065] In the above technical solution:

[0066] When obtaining training samples, they are collected from enterprise resource planning systems, product lifecycle management systems, customer relationship management systems, and supply chain management systems; product data includes: product market demand, competitor conditions, supply chain information, customer feedback, etc.; in the preprocessing of training samples, deduplication refers to the removal of duplicate data in training samples, and elimination of erroneous or invalid data refers to the elimination of erroneous or invalid data in training samples; standardization refers to converting data from different sources, different formats, or different dimensions into a unified, comparable format or range; when training a large training model based on training samples, a deep learning algorithm is used.

[0067] In one embodiment, the pre-trained large model generates SQL query statements according to the user's product decision analysis needs, including:

[0068] Inputting the user's product decision analysis requirements into the vector output component in the pre-trained large model to obtain the prompt words output by the vector output component;

[0069] Based on natural language processing technology, the prompt words are parsed to obtain query constraints and key information; wherein the query constraints at least include: data generation time range conditions, data type conditions, data relevance conditions, data sorting conditions and data filtering conditions; the key information at least includes: product name and product life cycle stage;

[0070] The query constraints and the key information are input into the pre-trained big model to obtain the SQL query statement output by the pre-trained big model.

[0071] In the above technical solution:

[0072] The vector output component is a computer component that can output prompt words based on information, and can be set in advance by technical personnel according to actual needs; after the query constraints and key information are input into the pre-trained large model, the pre-trained large model parses out keywords (such as "select", "from", etc.), field names (such as "product name", "sales", etc.) and query conditions (such as time range, filtering conditions, etc.), and builds various parts of the SQL query statement based on the parsed keywords, field names and query conditions (including determining the data table to be retrieved, the columns to be queried, how to connect different data tables, and processing aggregate functions (functions that perform calculations on a set of values ​​and return a single value) or calculated fields, etc.), including: SELECT clause, FROM clause, and optimize (such as selecting the most effective index, adjusting the query execution plan, reducing unnecessary data scanning, etc.); generates SQL query statements for corresponding data queries to improve query efficiency or reduce resource consumption.

[0073] In one embodiment, the process of training the large model to be trained based on the preprocessing result further includes:

[0074] Based on the model update mechanism, obtaining sample update data of the preprocessing result;

[0075] Based on the preprocessing result and the sample update data, jointly training the large model to be trained;

[0076] and / or,

[0077] Adjust the feedback mechanism based on the model and obtain usage feedback information;

[0078] Adjusting the preprocessing result based on the usage feedback information;

[0079] Based on the adjusted preprocessing results, the large model to be trained is trained.

[0080] In the above technical solution:

[0081] As time goes by, product data in the manufacturing industry may change (such as improvements in product design, adjustments to production processes, etc.), which may cause the performance of the original model to deteriorate. Therefore, it is necessary to regularly collect new data, retrain and optimize the model to maintain the performance of the model and establish a model update mechanism; regularly collect new data to retrain and optimize the model;

[0082] Consider the decision maker's feedback when putting it into production. Based on the decision maker's feedback and suggestions, the model can be fine-tuned to improve the model's performance. This includes adjusting the model's parameters, adding new features, changing the model's architecture, etc., and incorporating feedback suggestions into the model fine-tuning content.

[0083] In one embodiment, when the large model to be trained is trained to convergence based on the training sample, the large model to be trained that has been trained to convergence is used as the pre-trained large model, further comprising:

[0084] Based on the model verification mechanism, the large model to be trained that has been trained to convergence is verified to obtain the model verification result;

[0085] Based on the model verification result, the large model to be trained that has been trained to convergence is corrected, and the large model to be trained that has been trained to convergence after the model correction is used as the pre-trained large model.

[0086] A model validation mechanism is established through error analysis methods, including the use of holdout validation, cross-validation and other methods; the performance of the model is evaluated by dividing the data set into training set, validation set and test set, and the hyperparameters of the model are adjusted to ensure that the model can perform well on unseen data.

[0087] In practical applications, the technology of this application is Figure 1 As shown in the figure, firstly, data from R&D, production, sales, after-sales and other stages are collected and preprocessed, and then a large model is built using deep learning and other algorithms. In response to users' demand for data on the entire product life cycle, SQL statements are generated through the large model to query data in the database. Based on the structured data obtained from the query, the life cycle is predicted and analyzed in combination with product characteristics, market demand and competition. The large model will make reference decisions on R&D, production plans, sales strategies and after-sales services based on the prediction and structured data analysis results, and output them to decision makers together with the query data.

[0088] This method converts the user's text questions directly into SQL statements for query in the database, analyzes the queried structured data structure to obtain decision suggestions, and finally provides the decision suggestions and data query structure to decision makers at the same time, which helps to simplify the data query process, improve decision-making efficiency, optimize resource allocation, enhance competitiveness and promote digital transformation.

[0089] like Figure 2 and Figure 3As shown in the figure, firstly, the structured data related to the product life cycle, including product market demand, competitor situation, supply chain information, customer feedback, etc., are collected from the enterprise system, and the collected data are cleaned to remove duplicate, erroneous or invalid data, and the data from different sources are standardized. At the same time, the training data set is constructed, and the query demand description is aligned with the structured query language (sql) one by one. On the other hand, according to the characteristics of manufacturing product data, a suitable open source large language model such as ChatGLM2 is used as the basis, and fine-tuned in combination with the above training data set. For example, the LoRA fine-tuning technology can be used to adjust the network parameters of the model. And the model is fine-tuned for multiple rounds using historical data to improve the prediction and analysis capabilities of the model. At the same time, a model verification mechanism is established to improve the accuracy and reliability of the model through error analysis and other methods. Finally, the trained model is deployed to the production environment to realize real-time analysis and prediction of data. At the same time, a model update mechanism is established to regularly collect new data to retrain and optimize the model to adapt to market changes and industry development trends. And when it is put into production, the decision maker's feedback is considered, and the feedback suggestions are incorporated into the content of model fine-tuning to improve the model's ability to analyze and solve practical problems.

[0090] In actual use, according to the question text raised by the user, the user's question text and the actual database architecture are analyzed in the vector output component to obtain the prompt word (prompt) for the big model. It is input into the big model. After the big model receives the user's prompt word, it first needs to parse it to understand the specific content that the user wants to query and the query constraints. Natural language processing technology, such as word segmentation, part-of-speech tagging, named entity recognition, etc., is used to extract the key information in the prompt word. After that, the big model will further understand the user's query intention. Including determining the table and field that the user wants to query, as well as various specific filtering conditions or aggregation operations. During this period, the big model will use the knowledge learned during its training process and the current prompt word to infer the user's query intention. Based on the big model's understanding of the database structure, the big model will construct various parts of the structured query language statement, such as the SELECT clause, the FROM clause, etc., based on the parsed keywords, field names and conditions, combined with its internal templates or rules, and optimize them to improve query efficiency or reduce resource consumption. Then the structured query language statement provided by the big model is queried in the database to obtain the structured data required by the user. Then, we use the big model to deeply mine the structured data obtained from the query to discover the potential patterns and valuable information in the data. We present the data to decision makers through visualization tools, so that they can intuitively understand market trends, customer needs and other conditions. And based on the data analysis results, we provide decision makers with targeted decision support and suggestions. For example, we can adjust the production plan based on market demand forecasts and optimize product strategies based on the situation of competitors.

[0091] The embodiment of the present invention provides a manufacturing product life cycle decision analysis system based on a large model, such as Figure 4 As shown, including:

[0092] Generation module 1 is used to generate SQL query statements based on the pre-trained large model and the user's product decision analysis needs;

[0093] Query module 2, used to query the corresponding structured data in the product life cycle database based on the SQL query statement;

[0094] Determination module 3, used to analyze the structured data based on the pre-trained large model to determine product management decision suggestions;

[0095] The output module 4 is used to output the structured data and the product management decision suggestions to the user.

[0096] An embodiment of the present invention provides an electronic device, the electronic device comprising:

[0097] at least one processor; and

[0098] A memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned large model-based manufacturing product full life cycle decision analysis methods.

[0099] An embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a processor to implement any of the above-mentioned large-model-based manufacturing product full life cycle decision analysis methods when executed.

[0100] An embodiment of the present invention provides a computer program product, which includes computer instructions. When executed by an electronic device, the electronic device executes the computer program code of any of the above-mentioned methods for decision analysis of the entire life cycle of manufacturing products based on large models.

[0101] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A large-scale model-based decision analysis method for the entire life cycle of manufacturing products, characterized in that: include: Based on the pre-trained large model, generate SQL query statements according to the user's product decision analysis needs; Based on the SQL query statement, the corresponding structured data is queried in the product life cycle database; Based on the pre-trained big model, the structured data is analyzed to determine product management decision suggestions; The structured data and the product management decision suggestion are output to the user.

2. The large-model-based manufacturing product life cycle decision analysis method according to claim 1, characterized in that: The acquisition process of the pre-trained large model includes the following steps: Obtain training samples; the training samples are product data of a large number of different manufacturing products at various stages of research and development, production, sales, and after-sales; Preprocessing the training samples to obtain preprocessing results; Based on the preprocessing results, training the large model to be trained; When the large model to be trained is trained to convergence based on the preprocessing result, the large model to be trained that has been trained to convergence is used as the pre-trained large model.

3. The large-model-based manufacturing product life cycle decision analysis method according to claim 2, characterized in that: The preprocessing of the training samples to obtain preprocessing results includes: performing the processing steps of deduplication, eliminating erroneous or invalid data, and standardization on the training samples in sequence.

4. The large-model-based manufacturing product life cycle decision analysis method according to claim 1, characterized in that: The pre-trained large model generates SQL query statements based on the user's product decision analysis needs, including: Inputting the user's product decision analysis requirements into the vector output component in the pre-trained large model to obtain the prompt words output by the vector output component; Based on natural language processing technology, the prompt words are parsed to obtain query constraints and key information; wherein the query constraints at least include: data generation time range conditions, data type conditions, data relevance conditions, data sorting conditions and data filtering conditions; the key information at least includes: product name and product life cycle stage; The query constraints and the key information are input into the pre-trained big model to obtain the SQL query statement output by the pre-trained big model.

5. The large-model-based manufacturing product life cycle decision analysis method according to claim 2, characterized in that: The process of training the large model to be trained based on the preprocessing result also includes: Based on the model update mechanism, obtaining sample update data of the preprocessing result; Based on the preprocessing result and the sample update data, jointly training the large model to be trained; and / or, Adjust the feedback mechanism based on the model and obtain usage feedback information; Adjusting the preprocessing result based on the usage feedback information; Based on the adjusted preprocessing results, the large model to be trained is trained.

6. The large-model-based manufacturing product life cycle decision analysis method according to claim 2, characterized in that: When the large model to be trained is trained to convergence based on the training samples, the large model to be trained that has been trained to convergence is used as the pre-trained large model, further comprising: Based on the model verification mechanism, the large model to be trained that has been trained to convergence is verified to obtain the model verification result; Based on the model verification result, the large model to be trained that has been trained to convergence is corrected, and the large model to be trained that has been trained to convergence after the model correction is used as the pre-trained large model.

7. A large-scale model-based decision analysis system for the entire life cycle of manufacturing products, characterized in that: include: The generation module is used to generate SQL query statements based on the pre-trained large model and the user's product decision analysis needs; A query module, used to query the corresponding structured data in the product life cycle database based on the SQL query statement; A determination module, configured to analyze the structured data based on the pre-trained large model to determine product management decision suggestions; The output module is used to output the structured data and the product management decision suggestions to the user.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and A memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the large model-based manufacturing product full life cycle decision analysis method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the large-model-based manufacturing product full life cycle decision analysis method described in any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product includes computer instructions, and when executed by an electronic device, the electronic device executes the computer program code of the large model-based manufacturing product full life cycle decision analysis method as described in any one of claims 1 to 6.

Images