Association analysis method and system for industrial production and storage medium
By building a real-time industrial database and using an intent recognition module to perform correlation analysis with a large industrial model, the error problem caused by manual analysis was solved, and the efficiency of industrial production was improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-03
- Publication Date
- 2026-04-21
AI Technical Summary
In existing technologies, the reliance on manual data analysis in industrial production processes results in a high probability of errors, which affects production efficiency.
By acquiring real-time industrial production data, a real-time industrial database is built. The intention recognition module and large industrial model are used for correlation analysis to reduce the influence of subjective factors.
This reduces the probability of erroneous analysis and improves the overall efficiency of industrial production.
Smart Images
Figure CN121901310A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial production technology, and in particular to a correlation analysis method, system, and storage medium for industrial production. Background Technology
[0002] With the development of technology, a large number of sensors are set up to acquire data in industrial production processes. In the chemical industry, these sensor-acquired data play an important role in monitoring chemical processes. Currently, these data are generally analyzed by professionals to determine the status of chemical production and control equipment based on human experience. However, the process of adjusting equipment requires rehearsals to determine whether adjustments are necessary. But manual rehearsals or monitoring may have a high probability of error due to subjective perception, which can lead to a decrease in production efficiency.
[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention
[0004] The main objective of this invention is to provide a correlation analysis method, system, and storage medium for industrial production, aiming to improve the production efficiency of industrial production. To achieve the above objective, this invention provides a correlation analysis method for industrial production, which includes the following steps: Acquire real-time production data and user-inputted industrial production queries, and construct a real-time industrial database based on the real-time production data; The industrial business intent is determined based on the industrial production query statement and the intent recognition module. The correlation analysis results are determined based on the industrial business intent, the real-time industrial database, and the preset industrial big model. The real-time production data is real-time data during the industrial production process, the industrial database is a real-time retrieval enhanced generation RAG database, and the intent recognition module is a sub-module of the industrial big model.
[0005] Optionally, the real-time production data includes: sensor time-series data and equipment status logs, and the step of constructing a real-time industrial database based on the real-time production data includes: Sliding window feature extraction is performed on the time-series data of the sensor to obtain a time-series feature vector; The device status log is processed into a text structure to obtain a text feature vector; The time-series feature vector and the text feature vector are fused in a multimodal manner and stored in a vector database to construct the real-time industrial database; The sensor timing data and the device status log are acquired in real time through an industrial protocol gateway.
[0006] Optionally, the intent recognition module includes an intent classification model, and the steps of the intent recognition module in determining the industrial business intent include: The industrial production query statement is segmented using a preset industrial word segmenter to extract core industrial terms. The core industrial terms are input into the intent classification model, which outputs multiple candidate intents and their corresponding confidence scores. The candidate intent with the highest confidence score is selected as the industrial business intent. The industrial business intents include at least one of the following: equipment fault tracing intent, process parameter optimization intent, energy consumption anomaly analysis intent, and production scheduling suggestion intent.
[0007] Optionally, the step of determining the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial big data model further includes: Match and associate knowledge subgraphs in the knowledge graph module according to the industrial business intent; Based on the aforementioned associated knowledge subgraph, semantic retrieval is performed in the real-time industrial database to determine real-time associated data. The industrial production query statement, the associated knowledge subgraph, and the real-time associated data are input into the industrial big model for association analysis to obtain the association analysis results.
[0008] Optionally, the step of matching associated knowledge subgraphs in the knowledge graph module according to the industrial business intent includes: Determine the target entity type and target relationships based on the stated industrial business intent; Retrieve knowledge paths containing the target entity type and the target association in the knowledge graph module; Based on the knowledge path, relevant entity nodes, attribute information, and topological structure are extracted to determine the associated knowledge subgraph.
[0009] Optionally, the industrial big model is provided with a domain knowledge alignment layer, and the generation weight of the association analysis results is determined by the domain knowledge alignment layer according to the entity constraint relationship in the association knowledge subgraph; The domain knowledge alignment layer is a weighted correction of the attention mechanism matrix of the industrial big model based on the physical quantity constraint rules and logical evolution laws in the associated knowledge subgraph.
[0010] Optionally, before the step of constructing a real-time industrial database based on the real-time production data, the method further includes: Acquire historical production case data and expert-annotated corpora from the industrial sector; The basic large model is pre-trained based on the historical production case data to obtain a preliminary industrial model; The preliminary industrial model is subjected to supervised fine-tuning based on the expert-annotated corpus to obtain the large industrial model.
[0011] This invention proposes a correlation analysis method for industrial production. This method acquires real-time production data and user-input industrial production queries, constructs a real-time industrial database based on the real-time production data, and determines industrial business intent based on the industrial production queries and an intent recognition module. Compared to manual analysis, determining the correlation analysis results based on the industrial business intent, the real-time industrial database, and a preset industrial big data model can effectively utilize historical data for effective deduction and analysis, avoiding subjective factors, thereby reducing the probability of erroneous analysis and improving overall industrial production efficiency. Attached Figure Description
[0012] Figure 1 This is a schematic diagram of the structure of the industrial production correlation analysis equipment of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the correlation analysis method for industrial production of the present invention; Figure 3 This is a flowchart illustrating the second embodiment of the correlation analysis method for industrial production of the present invention; Figure 4 This is a flowchart illustrating the fourth embodiment of the correlation analysis method for industrial production of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0013] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0014] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of an industrial production correlation analysis device for the hardware operating environment involved in the embodiments of the present invention.
[0015] like Figure 1As shown, the correlation analysis device for industrial production may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, an interactive device 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The interactive device 1003 may include a display screen and an input unit such as a keyboard. Optionally, the interactive device 1003 may also be connected to the communication bus via standard wired or wireless interfaces. The network interface 1004 may optionally include standard wired or wireless interfaces (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0016] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the correlation analysis equipment for industrial production and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0017] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an industrial production correlation analysis program.
[0018] exist Figure 1 In the industrial production correlation analysis device shown, the network interface 1004 is mainly used for data communication with other devices; the interactive device 1003 is mainly used for data interaction with users; the processor 1001 and memory 1005 in the industrial production correlation analysis device of the present invention can be set in the industrial production correlation analysis device, and the industrial production correlation analysis device calls the industrial production correlation analysis program stored in the memory 1005 through the processor 1001 and executes the industrial production correlation analysis method provided in the embodiment of the present invention.
[0019] This invention provides a correlation analysis method for industrial production, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of a correlation analysis method for industrial production according to the present invention.
[0020] In this embodiment, the correlation analysis method for industrial production includes: Step S1: Obtain real-time production data and user-inputted industrial production query statements, and construct a real-time industrial database based on the real-time production data; In this embodiment, industrial production can refer to chemical production. Typically, multiple sensors are used in chemical production processes. The type of sensor is not limited; common sensors include temperature sensors, pressure sensors, and flow sensors. The flow sensor is used to measure material flow rate to determine the current material usage or output. The industrial production query statement can be a continuous query statement or multiple statements, generally including data from multiple question-and-answer processes. All contextual data is used as the industrial production query statement. Optionally, only the most recently entered statement by the user can be used as the industrial production query statement. In this embodiment, the real-time industrial database needs to include real-time detection data at the current moment. Optionally, to facilitate data use and retrieval, the real-time industrial database may also include historical data recorded before the current time.
[0021] Step S2: Determine the industrial business intent based on the industrial production query statement and the intent recognition module; Specifically, the intent recognition module is a component used to parse and classify core user intents from user data. Optionally, this intent recognition module may include: a Prompt template, an intent definition library, and an output parser. The intent definition library includes a predefined list of intents, which may include: intent name, description, triggering keywords, associated functions, associated tools, etc. In this embodiment, the industrial production query statement is input into the intent recognition module, and the intent recognition module outputs industrial business intents. These industrial business intents may include: anomaly analysis, fault identification, maintenance planning, parameter adjustment simulation, etc. The parameter adjustment simulation, also known as parameter adjustment pre-simulation or control parameter pre-simulation, is mainly used to verify control and measurement strategies and simulate the dynamic response of the system under different parameter settings, such as overshoot, settling time, and oscillation, thereby avoiding fluctuations, quality problems, or safety risks caused by actual adjustments.
[0022] Step S3: Determine the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial big model; In this embodiment, the industrial business intent is input into a preset industrial big model, and the industrial big model retrieves data from a real-time industrial database based on the industrial business intent for analysis, thereby outputting the result as the correlation analysis result.
[0023] The real-time production data refers to real-time data during the industrial production process, the industrial database is a real-time retrieval enhanced generation RAG database, and the intent recognition module is a sub-module of the industrial big model.
[0024] In this embodiment, by acquiring real-time production data and user-inputted industrial production queries, and constructing a real-time industrial database based on the real-time production data, and determining industrial business intent based on the industrial production queries and intent recognition module, compared to manual analysis, determining correlation analysis results based on the industrial business intent, real-time industrial database, and preset industrial big model can effectively utilize historical data for effective deduction and analysis, avoid subjective factors, thereby reducing the probability of erroneous analysis and improving overall industrial production efficiency.
[0025] Furthermore, based on the first embodiment, a second embodiment of the correlation analysis method for industrial production of the present invention is proposed. In this embodiment, reference is made to... Figure 3 The real-time production data includes: sensor time-series data and equipment status logs. The step of constructing a real-time industrial database based on the real-time production data includes: Step S11: Perform sliding window feature extraction on the sensor time series data to obtain a time series feature vector; In this embodiment, the sliding window feature extraction can be an adaptive sliding window. The window size is dynamically adjusted according to the data volatility. Specifically, the volatility of each data point is obtained and calculated, and the time length of the sliding window is set accordingly. It should be noted that the time length increases with the increase of volatility, and in this embodiment, the interval between two adjacent sliding windows can be fixed. Preferably, the interval between adjacent sliding windows can be 0, but in other embodiments, it can also be set by the administrator. In this embodiment, by adjusting the window size, i.e., the time length of the sliding window, by the volatility of each data point, the proportion of invalid data can be effectively reduced, so that the time series feature vector contains more effective features. These effective features can be: mean, variance, and volatility frequency.
[0026] Step S12: Perform text structuring processing on the device status log to obtain text feature vectors; To perform text structuring on device status logs, firstly, natural language processing tools are used for entity recognition and sentiment analysis, breaking the logs down into key-value pairs, such as extraction, device ID, status type, and timestamp. Further, rule-based log filters can be set up to prioritize the removal of redundant entries and inject contextual tags, thereby obtaining a more refined text feature vector.
[0027] Step S13: Perform multimodal fusion of the time-series feature vector and the text feature vector, and store them in the vector database to construct the real-time industrial database; The sensor time-series data and the device status log are acquired in real time through an industrial protocol gateway. Specifically, time-series vectors and text vectors are mapped to the same space, and attention weights are calculated by an encoder to generate vectors corresponding to types defined in the database. This data is then used to construct the real-time industrial database.
[0028] In this embodiment, a time-series feature vector is obtained by performing sliding window feature extraction on the sensor time-series data, and a text feature vector is obtained by performing text structuring processing on the device status log. The time-series feature vector and the text feature vector are then fused in a multimodal manner and stored in a vector database to construct the real-time industrial database, thereby improving the effectiveness of the data in the real-time industrial database.
[0029] Furthermore, based on the first or second embodiment, a third embodiment of the industrial production correlation analysis method of the present invention is proposed, wherein the step of the intent recognition module determining the industrial business intent includes: The industrial production query statement is segmented using a preset industrial word segmenter to extract core industrial terms. Optionally, the industrial word segmenter here is a tool such as HanLP or pkuseg, which obtains vocabulary data corresponding to the current industrial type to segment industrial production queries. This allows for the acquisition of core industrial vocabulary.
[0030] The core industrial terms are input into the intent classification model, which outputs multiple candidate intents and their corresponding confidence scores. Optionally, the core industry vocabulary is input into the intent classification model in sequence. This intent classification model can be a lightweight large language model, which can be a model obtained through fine-tuning. It should be noted that using a lightweight large language model is based on the fact that workers generally only use professional vocabulary or industry-specific common vocabulary for communication during work processes, and it does not require corpora from multiple domains. Therefore, it can be implemented relatively easily during the fine-tuning stage, thereby effectively reducing costs. In other embodiments, the intent classification model can also be a deep learning model.
[0031] The candidate intent with the highest confidence score is selected as the industrial business intent. After outputting multiple candidate intents and their corresponding confidence scores, one intent can be determined as the industrial business intent based on the confidence score. The industrial business intent includes at least one of the following: equipment fault tracing intent, process parameter optimization intent, energy consumption anomaly analysis intent, and production scheduling suggestion intent. Optionally, the equipment fault tracing intent aims to locate the cause of production equipment failure through data analysis; the process parameter optimization intent aims to adjust processing parameters to improve product quality and efficiency; the energy consumption anomaly analysis intent aims to identify abnormal energy consumption; and the production scheduling suggestion intent aims to optimize task allocation and resource allocation based on real-time production conditions to improve overall production efficiency.
[0032] In this embodiment, the industrial production query statement is segmented according to a preset industrial word segmenter to extract core industrial terms. These core terms are then input into an intent classification model, which outputs multiple candidate intents and their corresponding confidence scores. The candidate intent with the highest confidence score is selected as the industrial business intent, thereby accurately determining the user's needs and improving the accuracy of subsequent analysis targets, thus avoiding errors in the analysis targets.
[0033] Furthermore, based on any of the above embodiments, a fourth embodiment of the correlation analysis method for industrial production of the present invention is proposed, referring to... Figure 4 The step of determining the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial large model further includes: Step S31: Match related knowledge subgraphs in the knowledge graph module according to the industrial business intent; Optionally, the knowledge graph module here is a multi-layered entity relationship graph pre-constructed based on ontology engineering. This entity relationship graph covers entities such as equipment, processes, failure modes, and energy consumption indicators in the current industrial business domain, along with their causal, temporal, and upstream / downstream relationships. The knowledge graph module stores data using a triplet data structure, such as triples for: abnormal vibration, cause, bearing failure; bearing failure, cause, insufficient lubrication; etc. Since a production process may include multiple pieces of equipment and corresponding entity relationship graphs, a matching associated knowledge subgraph is determined based on the industrial business intent.
[0034] Step S32: Based on the associated knowledge subgraph, perform semantic retrieval in the real-time industrial database to determine real-time associated data; In this embodiment, entities and relationships in the subgraph are used as query extensions to perform semantic retrieval in the vector representation of the real-time industrial database. Cosine similarity and relation constraints are used for filtering to determine relevant time-series data and log fragments, avoiding interference from irrelevant data.
[0035] Step S33: Input the industrial production query statement, the associated knowledge subgraph, and the real-time associated data into the industrial big model for association analysis to obtain the association analysis results.
[0036] In this embodiment, the industrial big data model performs multiple rounds of causal reasoning and evidence chain construction, ultimately outputting association analysis results that include root cause localization, impact assessment, and optimization suggestions. This design significantly reduces the risk of illusion and makes the analysis results more traceable. The industrial big data model can be a finely tuned version of the Thousand Questions model.
[0037] In this embodiment, the industrial business intent is matched with the associated knowledge subgraph in the knowledge graph module. Based on the associated knowledge subgraph, semantic retrieval is performed in the real-time industrial database to determine real-time associated data. The industrial production query statement, the associated knowledge subgraph, and the real-time associated data are input into the industrial big model for association analysis to obtain the association analysis results, thereby improving the accuracy of the association analysis results.
[0038] Furthermore, the step of matching associated knowledge subgraphs in the knowledge graph module according to the industrial business intent includes: Determine the target entity type and target relationships based on the stated industrial business intent; Retrieve knowledge paths containing the target entity type and the target association in the knowledge graph module; Based on the knowledge path, relevant entity nodes, attribute information, and topological structure are extracted to determine the associated knowledge subgraph.
[0039] In this embodiment, firstly, based on the industrial business intent, a pre-trained intent-entity mapping table is used to quickly determine the target entity type and target association, and generate a query embedding vector. Secondly, a two-layer index is used in the knowledge graph module: Neo4j-based Cypher query retrieves knowledge paths containing target entities and relationships, such as paths for abnormal vibration, bearing wear, and lubrication failure. GNN propagation calculates path confidence, prioritizing relevant paths with high confidence. Entity nodes, attribute information, and topology are extracted from these paths. The attribute information here can include: threshold values corresponding to indicators and historical failure rates of components.
[0040] Furthermore, the industrial big model is equipped with a domain knowledge alignment layer, and the generation weight of the association analysis results is determined by the domain knowledge alignment layer according to the entity constraint relationship in the association knowledge subgraph; The domain knowledge alignment layer is a weighted correction of the attention mechanism matrix of the industrial big model based on the physical quantity constraint rules and logical evolution laws in the associated knowledge subgraph.
[0041] Furthermore, based on any of the above embodiments, a fifth embodiment of the correlation analysis method for industrial production of the present invention is proposed, wherein before the step of constructing a real-time industrial database based on the real-time production data, the method further includes: Acquire historical production case data and expert-annotated corpora from the industrial sector; The basic large model is pre-trained based on the historical production case data to obtain a preliminary industrial model; The preliminary industrial model is subjected to supervised fine-tuning based on the expert-annotated corpus to obtain the large industrial model.
[0042] In this embodiment, it should be noted that the industrial big model here requires historical production case data and expert-annotated corpus before the current moment, and incremental pre-training is performed on the basic big model to obtain a preliminary industrial model and realize domain knowledge transfer; finally, supervised fine-tuning is performed based on the expert-annotated corpus, and preference alignment is performed through reinforcement learning mechanism (such as RLHF) to make the model output more in line with the judgment standards of industrial experts, and finally the industrial big model is obtained.
[0043] Furthermore, this invention also proposes an industrial production correlation analysis system, which includes: The acquisition module is used to acquire real-time production data of industrial production and industrial production query statements input by users, and to construct a real-time industrial database based on the real-time production data. The identification module is used to determine the industrial business intent based on the industrial production query statement and the intent identification module. The analysis module is used to determine the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial big model. The real-time production data refers to real-time data during the industrial production process, the industrial database is a real-time retrieval enhanced generation RAG database, and the intent recognition module is a sub-module of the industrial big model.
[0044] Furthermore, this invention also proposes an industrial production correlation analysis device, which includes: a memory, a processor, and an industrial production correlation analysis program stored in the memory and executable on the processor. The industrial production correlation analysis program is configured to implement the steps of the industrial production correlation analysis method described above.
[0045] Furthermore, embodiments of the present invention also propose a storage medium storing an industrial production correlation analysis program, wherein when the industrial production correlation analysis program is executed by a processor, it implements the steps of the industrial production correlation analysis method described above.
[0046] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.
[0047] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0048] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0049] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A correlation analysis method for industrial production, characterized in that, The correlation analysis method for industrial production includes the following steps: Acquire real-time production data and user-inputted industrial production queries, and construct a real-time industrial database based on the real-time production data; The industrial business intent is determined based on the industrial production query statement and the intent recognition module. The correlation analysis results are determined based on the industrial business intent, the real-time industrial database, and the preset industrial big model. The real-time production data is real-time data during the industrial production process, the industrial database is a real-time retrieval enhanced generation RAG database, and the intent recognition module is a sub-module of the industrial big model.
2. The correlation analysis method for industrial production as described in claim 1, characterized in that, The real-time production data includes: sensor time-series data and equipment status logs. The step of constructing a real-time industrial database based on the real-time production data includes: Sliding window feature extraction is performed on the time-series data of the sensor to obtain a time-series feature vector; The device status log is processed into a text structure to obtain a text feature vector; The time-series feature vector and the text feature vector are fused in a multimodal manner and stored in a vector database to construct the real-time industrial database; The sensor timing data and the device status log are acquired in real time through an industrial protocol gateway.
3. The correlation analysis method for industrial production as described in claim 1, characterized in that, The intent recognition module includes an intent classification model, and the steps by which the intent recognition module determines industrial business intent include: The industrial production query statement is segmented using a preset industrial word segmenter to extract core industrial terms. The core industrial terms are input into the intent classification model, which outputs multiple candidate intents and their corresponding confidence scores. The candidate intent with the highest confidence score is selected as the industrial business intent. The industrial business intents include at least one of the following: equipment fault tracing intent, process parameter optimization intent, energy consumption anomaly analysis intent, and production scheduling suggestion intent.
4. The correlation analysis method for industrial production as described in claim 1, characterized in that, The step of determining the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial big data model further includes: Match related knowledge subgraphs in the knowledge graph module according to the industrial business intent; Based on the aforementioned associated knowledge subgraph, semantic retrieval is performed in the real-time industrial database to determine real-time associated data. The industrial production query statement, the associated knowledge subgraph, and the real-time associated data are input into the industrial big model for association analysis to obtain the association analysis results.
5. The correlation analysis method for industrial production as described in claim 4, characterized in that, The step of matching related knowledge subgraphs in the knowledge graph module according to the industrial business intent includes: Determine the target entity type and target relationships based on the stated industrial business intent; Retrieve knowledge paths containing the target entity type and the target association in the knowledge graph module; Based on the knowledge path, relevant entity nodes, attribute information, and topological structure are extracted to determine the associated knowledge subgraph.
6. The correlation analysis method for industrial production as described in claim 5, characterized in that, The industrial big model is equipped with a domain knowledge alignment layer, and the generation weight of the association analysis results is determined by the domain knowledge alignment layer according to the entity constraint relationship in the association knowledge subgraph. The domain knowledge alignment layer is a weighted correction of the attention mechanism matrix of the industrial big model based on the physical quantity constraint rules and logical evolution laws in the associated knowledge subgraph.
7. The correlation analysis method for industrial production as described in any one of claims 1 to 6, characterized in that, Before the step of constructing a real-time industrial database based on the real-time production data, the method further includes: Acquire historical production case data and expert-annotated corpora from the industrial sector; The basic large model is pre-trained based on the historical production case data to obtain a preliminary industrial model; The preliminary industrial model is fine-tuned in a supervised manner based on the expert-annotated corpus, and preference alignment is performed to obtain the large industrial model.
8. A correlation analysis system for industrial production, characterized in that, The industrial production correlation analysis system includes: The acquisition module is used to acquire real-time production data of industrial production and industrial production query statements input by users, and to construct a real-time industrial database based on the real-time production data. The identification module is used to determine the industrial business intent based on the industrial production query statement and the intent identification module. The analysis module is used to determine the correlation analysis results based on the industrial business intent, the real-time industrial database, and the preset industrial big model. The real-time production data refers to real-time data during the industrial production process, the industrial database is a real-time retrieval enhanced generation RAG database, and the intent recognition module is a sub-module of the industrial big model.
9. A correlation analysis device for industrial production, characterized in that, The industrial production correlation analysis device includes: a memory, a processor, and an industrial production correlation analysis program stored in the memory and executable on the processor, wherein the industrial production correlation analysis program is configured to implement the steps of the industrial production correlation analysis method as described in any one of claims 1 to 7.
10. A storage medium, characterized in that, The storage medium stores an industrial production correlation analysis program, which, when executed by a processor, implements the steps of the industrial production correlation analysis method as described in any one of claims 1 to 7.