Mechanical intelligent manufacturing steel CPS transmission and monitoring system based on large language model

Abstract

The invention provides an intelligent mechanical manufacturing steel CPS transmission and monitoring system based on a large language model, and the system comprises the steps: arranging a sensor and detection equipment in a production line to collect real-time data, and transmitting the real-time data to a rear-end service through an ECS; the back-end service preprocesses and standardizes the real-time data to obtain standardized data, obtains embedded data after semantic embedding and vectorization processing, stores the embedded data in a database, and calls a large language model to analyze the embedded data to obtain optimization suggestions and early warning information of the production line; the intelligent control system dynamically adjusts operation of the production line according to the optimization suggestion and the early warning information to obtain an operation result and feedback information, and records the operation result and the feedback information in a database to train a large language model; and the ESC calls a back-end service and a large language model to generate an analysis report, and provides decision support information. According to the invention, the problems of insufficient intelligence, slow response, real-time monitoring and prediction maintenance are solved, the system reliability and production continuity are improved, and the real-time performance and integrity of production data are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial automation. Specifically, it relates to a mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model, and in particular, a precise transmission and monitoring system for steel CPS based on a large language model. Background Technique

[0002] Industrial CPS systems are widely applied in many fields, including environmental monitoring, agricultural planting, intelligent transportation, smart home, etc., providing real-time data collection, monitoring, and analysis services for various industries, effectively improving production efficiency and decision-making scientificity. At the same time, with the continuous progress of technology, the demand for personalized customization is increasing day by day, and industrial CPS systems are gradually developing towards a more intelligent and refined direction. However, challenges such as high costs for sensor deployment and maintenance and data transmission stability have slowed down the pace of automation and intelligence development to a certain extent.

[0003] In the prior art, reference can be made to some existing industrial process control systems. Similar systems usually solve problems such as optimizing the control of comprehensive production indicators such as product quality, output, cost, and consumption in the face of market and raw material changes. However, the disadvantages of the prior art are that they often lack sufficient flexibility and dynamic reconfiguration capabilities and are difficult to adapt to the rapidly changing personalized customization requirements. Especially in the context of accelerating product replacement, complex manufacturing technologies, shortened product batch conversion times, and increasing consumer requirements for product quality and price, existing systems are difficult to quickly adapt to these changes.

[0004] The technology disclosed in the patent document "An Industrial CPS System and Implementation Method for Personalized Customization" (CN109933010A) targets the requirements of personalized customization production, the demand for reconfiguration capabilities in industrial production lines, and the requirement for quickly implementing production line control systems. It can accurately generate production processes dynamically according to different product requirements to adapt to the needs of different products and provides an implementation method for drag-and-drop programming of programmable logic controllers (PLCs). However, it lacks sufficient data processing capabilities and self-learning optimization mechanisms. Traditional systems often make adjustments based on fixed rules or simple algorithms and are difficult to cope with rapidly changing external environments and internal conditions.

[0005] The casting defect diagnosis and optimization system disclosed in the patent document "Casting Defect Diagnosis and Optimization System Based on Artificial Intelligence Large Language Model" (CN119126762A) mainly focuses on real-time monitoring, prediction, and process optimization design during the casting production process. Although the technologies used in it, such as large language models and Internet of Things sensors, are universal in multiple fields, the design and parameter settings are all centered around specific scenarios. Due to differences in specific application scenarios, standards and specifications, and the need for adjustment of technical details, etc., it cannot be fully applicable to all fields.

[0006] Therefore, there is an urgent need for a mechanical intelligent manufacturing steel CPS transmission and monitoring system that can adjust details such as system design and parameter configuration according to specific requirements to adapt to new scenarios. Summary of the Invention

[0007] Aiming at the defects in the prior art, the purpose of the present invention is to provide a mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model.

[0008] According to the mechanical intelligent manufacturing steel CPS transmission and monitoring method based on a large language model provided by the present invention, it includes:

[0009] Step S1: Arrange sensors and detection devices in the production line to collect real-time data, and transmit it to the backend service through ECS;

[0010] Step S2: The backend service preprocesses and standardizes the real-time data to obtain standardized data;

[0011] Step S3: The backend service performs semantic embedding and vectorization processing on the standardized data, and stores the obtained embedded data in the database;

[0012] Step S4: The backend service calls the large language model to analyze the embedded data to obtain optimization suggestions and warning information for the production line;

[0013] Step S5: The intelligent control system dynamically adjusts the operation of the production line according to the optimization suggestions and warning information to obtain operation results and feedback information;

[0014] Step S6: The intelligent control system records the operation results and feedback information in the database, and retrains the large language model;

[0015] Step S7: Call the backend service and the large language model through ESC to generate an analysis report and provide decision support information.

[0016] Preferably, the sensors include temperature, vibration, pressure, speed, optoelectronic, and video monitoring;

[0017] The detection device is used to capture key parameters during the mechanical transmission process in real time, form real-time data with the data collected by sensors, and transmit it to the ECS through the industrial control network, and distribute data requests and transmission requirements to the backend services.

[0018] The ECS is an intermediate hub for data access and instruction scheduling.

[0019] The step S2 includes:

[0020] Step S2.1: Trigger the preprocessing logic using the backend service, call algorithms or microservices, and judge and eliminate extreme values or abnormal data in the data through statistical means such as z-score, mean, and standard deviation;

[0021] Step S2.2: Convert real-time data from different sources and types into a unified format;

[0022] Step S2.3: Align and correct the timestamps of real-time data from different sources through the backend service, and store the cleaned and standardized data in the database.

[0023] Preferably, the step S3 includes:

[0024] Step S3.1: The backend service interacts with the embedding model through the search module, performs feature extraction and semantic vectorization on the standardized data or text description data captured by the search model to obtain embedding data;

[0025] Step S3.2: Store the embedding data in the database, maintain and manage the database, and optimize the index and storage strategy.

[0026] The search module is a part of the backend service and is responsible for receiving input and calling appropriate processing logic.

[0027] The embedding model is used to convert text or sensor data into a numerical vector representation.

[0028] The process data stored in the database includes real-time data, standardized data, embedding data, optimization suggestions, warning information, operation results, and feedback information, and is stored and backed up at different locations set according to production requirements, and the data structure, index, and security strategy are maintained.

[0029] Preferably, the step S4 includes:

[0030] Fault mode recognition step: The large language model identifies and predicts the risk of the fault type of the production line based on the embedding data and general data;

[0031] Optimization steps for operating status: The large language model analyzes the conveyor path, conveyor speed, and equipment load parameters of the production line in combination with production experience and process rules, obtains basic operating data, and outputs optimization suggestions;

[0032] Trend analysis and warning steps: The large language model conducts trend analysis on the basic operating data, identifies potential anomalies, and issues warning messages.

[0033] The fault mode identification step includes:

[0034] Identification step: By analyzing the anomalies in the general data and known fault cases, construct a classifier or model for the fault mode, discover the anomalies of the equipment in different operating states according to the decision tree or neural network, and associate the anomalies with the fault types;

[0035] Risk prediction step: Based on the identified fault types and related general data, use statistical methods or time series analysis techniques to evaluate the risk of each fault type.

[0036] The potential anomalies include that the operating data shows a significant increase in the data fluctuation range, a sudden decrease or increase in the key performance indicators, or a changing trend of frequent occurrence of fault types within a specific time period.

[0037] The warning message includes a specific anomaly description and preliminary countermeasures recommended to be taken.

[0038] Preferably, in step S5, the optimization suggestions and warning messages are sent to the ECS through the backend service and forwarded to the intelligent control system to dynamically adjust the operation of the production line, record the operation results and feedback information, including;

[0039] Fault warning and shutdown protection step: When a potential anomaly is predicted, the ECS instructs the intelligent control system to execute load reduction, suspension of operation, or switching to a standby line;

[0040] Automation adjustment step: According to the optimization suggestions, adjust the speed, load distribution, and path planning transfer parameters of the production line in real time.

[0041] In step S6, the intelligent control system records the operation results and feedback information into the database through the ECS and the backend service, including:

[0042] Feedback loop optimization step: The operation results and feedback information are re-embedded through the embedded model to obtain new data and stored in the database;

[0043] Adaptive learning step: The large language model calls the new data for retraining or parameter fine-tuning according to the set time interval.

[0044] A mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model according to the present invention includes: a production line, sensors and monitoring devices, an ECS, a backend service, an embedded model, a database, a large language model, and an intelligent control system;

[0045] The backend service includes a search module.

[0046] Sensors and detection devices are arranged in the production line to collect real-time data, which is transmitted to the backend service through the ECS;

[0047] The backend service preprocesses and standardizes the real-time data to obtain standardized data, interacts with the embedded model, performs semantic embedding and vectorization processing, stores the obtained embedded data in the database, and calls the large language model to analyze the embedded data to obtain optimization suggestions and warning information for the production line;

[0048] The intelligent control system dynamically adjusts the operation of the production line according to the optimization suggestions and warning information to obtain operation results and feedback information;

[0049] The intelligent control system records the operation results and feedback information in the database and retrains the large language model;

[0050] The ESC calls the backend service and the large language model to generate an analysis report and provide decision support information.

[0051] Preferably, the sensors include temperature, vibration, pressure, speed, optoelectronics, and video monitoring;

[0052] The detection device is used to capture key parameters during the mechanical transmission process in real time, form real-time data with the data collected by the sensors, transmit it to the ECS through the industrial control network, and distribute data requests and transmission requirements to the backend service.

[0053] The ECS is an intermediate hub for data access and instruction scheduling.

[0054] The backend service triggers the preprocessing logic, calls algorithms or microservices, and uses statistical means such as z-score, mean, and standard deviation to judge and eliminate extreme values or abnormal data in the data;

[0055] Convert real-time data from different sources and types into a unified format;

[0056] Align and correct the timestamps of real-time data from different sources, and store the cleaned and standardized data in the database.

[0057] Preferably, the backend service interacts with the embedded model through the search module, performs feature extraction and semantic vectorization on the standardized data or text description data captured by the search model to obtain embedded data;

[0058] Store the embedded data in a database, maintain and manage the database, and optimize the indexing and storage strategies.

[0059] The search module is part of the backend service and is responsible for receiving inputs and invoking appropriate processing logic.

[0060] The embedding model is used to convert text or sensor data into a numerical vector representation.

[0061] The process data stored in the database includes real-time data, standardized data, embedded data, optimization suggestions, warning information, operation results, and feedback information. According to production requirements, backups are stored at different set locations, and the data structure, indexing, and security policies are maintained.

[0062] Preferably, the large language model identifies the types of faults and predicts risks for the production line based on the embedded data and general data;

[0063] Analyze the transfer path, transfer speed, and equipment load parameters of the production line in combination with production experience and process rules to obtain basic operation data and output optimization suggestions;

[0064] Conduct trend analysis on the basic operation data, identify potential anomalies, and issue warning information.

[0065] By analyzing the anomalies and known fault cases in the general data, construct a classifier or model for the fault mode, discover anomalies in the equipment under different operating states based on decision trees or neural networks, and associate the anomalies with the types of faults;

[0066] Based on the identified types of faults and the relevant general data, use statistical methods or time series analysis techniques to evaluate the risk of each type of fault.

[0067] The potential anomalies include that the operation data shows a significant increase in the data fluctuation range, a sudden decrease or increase in key performance indicators, or a changing trend of frequent occurrence of fault types within a specific time period.

[0068] The warning information includes a specific anomaly description and preliminary countermeasures recommended to be taken.

[0069] Preferably, the optimization suggestions and warning information are sent to the ECS through the backend service for forwarding to the intelligent control system to dynamically adjust the operation of the production line and record the operation results and feedback information.

[0070] When a potential anomaly is predicted, the ECS instructs the intelligent control system to execute load reduction, suspension of operation, or switching to a standby line.

[0071] The intelligent control system adjusts the production line speed, load distribution, and path planning transfer parameters in real time according to the optimization suggestions;

[0072] The operation results and feedback information are re-embedded through the embedded model to obtain new data, which are recorded and stored in the database through ECS and the backend service.

[0073] The large language model calls the new data for retraining or parameter fine-tuning according to the set time interval.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1. By adopting the data analysis and processing structure based on the large language model, the present invention solves the problems of insufficient intelligence and slow response in the traditional CPS system when dealing with complex production data, thereby realizing efficient and accurate data transmission and monitoring.

[0076] 2. By adopting the modular design and the deep integration structure of the large language model, the present invention solves the problems of real-time monitoring and predictive maintenance in the steel manufacturing process, and improves the reliability of the system and the continuity of production.

[0077] 3. By adopting the optimized data transmission strategy, the present invention solves the problems of possible delay and data loss in the traditional transmission path, and ensures the real-time nature and integrity of production data. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more obvious:

[0079] Figure 1 It is a schematic flow chart of the steel CPS transmission and monitoring method for mechanical intelligent manufacturing based on the large language model;

[0080] Figure 2 It is a schematic diagram of the steel CPS transmission and monitoring system for mechanical intelligent manufacturing based on the large language model. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those of ordinary skill in the art can make several changes and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0082] The present invention proposes a mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model. By integrating the large language model, it realizes the intelligence and precision of data transmission and monitoring in the steel manufacturing process, thereby improving production efficiency and product quality. Taking Figure 1 as an example, it specifically includes:

[0083] Step S1: Data collection and monitoring;

[0084] In the steel production line, various sensors (such as temperature, vibration, pressure, speed, photoelectric, video monitoring, etc.) and monitoring devices are distributed to capture key parameters in the mechanical transmission process in real time.

[0085] All the collected sensor data is transmitted to the front-end computing and service platform (ECS) through the industrial control network. The ECS distributes data requests and transmission requirements to the downstream back-end services (such as Lambda functions). This is the basis of the data transmission strategy to ensure the integrity of the data and its real-time aggregation.

[0086] At this time, the ECS, as the intermediate hub for data access and instruction scheduling, unifies and integrates various production data and prepares it for further processing. This ensures that comprehensive and real-time data can be obtained. In addition, the same data is stored at multiple locations to prevent data loss or damage.

[0087] Step S2: Data preprocessing and standardization;

[0088] The data in steel production may be affected by noise, interference, etc., resulting in inconsistent and inaccurate raw data. After the back-end services (such as Lambda functions) connected to the ECS receive the sensor data, they perform preliminary cleaning and standardization processing. The specific processing includes:

[0089] Step S2.1: Abnormal data elimination: Use Lambda to trigger the preprocessing logic and call relevant algorithms or microservices to eliminate extreme values or abnormal data in the data.

[0090] Specifically, by using Lambda to trigger the preprocessing logic to call relevant algorithms or microservices to identify and eliminate extreme values or abnormal points in the data. Select appropriate determination methods, such as statistical means like z-score, mean and standard deviation, and based on these methods, determine which data points are considered abnormal and remove them from the original dataset to ensure the accuracy and effectiveness of subsequent analysis.

[0091] Step S2.2: Data normalization and standardization: Convert data from different sources and types into a unified format to ensure the consistency and comparability of the input of subsequent analysis models (LLM models). Through compression and encoding technologies, reduce the data transmission volume and improve the transmission speed.

[0092] Step S2.3: Time Synchronization: Align and correct the timestamps of different sensor data through the backend service to ensure accurate synchronization of data in the time dimension.

[0093] The cleaned and standardized data is stored in the database to ensure data quality and identity for subsequent analysis and processing.

[0094] Step S3: Data Embedding and Feature Vectorization;

[0095] Before the data enters the LLM model for analysis, the system calls the "Embedding Model" service to perform semantic embedding processing on text, semi-structured, and unstructured data.

[0096] Step S3.1: The backend service (Lambda) interacts with the embedding model through the search module to perform feature extraction and semantic vectorization (Embedding) on the preprocessed sensor data or text description data, thereby improving data usability.

[0097] Specifically, the search module is part of the backend service, which is responsible for receiving inputs and invoking appropriate processing logic.

[0098] The embedding model is an independent component dedicated to converting text or sensor data into numerical vector representations (i.e., feature extraction and semantic vectorization).

[0099] The text description data is first sent to the backend service through the front-end interface or other data sources. These data are captured by the search module and transmitted to the corresponding processing logic for subsequent processing by the embedding model or other relevant components.

[0100] Step S3.2: The embedded data is stored in the database for subsequent fast retrieval and associative query.

[0101] At this time, the database administrator can maintain and manage the database, optimize the index and storage strategy. Implement data security protection, including measures such as data encryption and access control, to ensure data security during transmission.

[0102] Step S4: Data Analysis and Optimization Based on the Large Language Model;

[0103] In the backend system, the LLM model (Large Language Model) is invoked through the Lambda service. When there is an analysis request or optimization requirement, Lambda sends a query request (SQL or other protocols) to the LLM model and inputs the embedded data or structured data retrieved from the database into the LLM model for intelligent analysis.

[0104] Step S4.1: Fault mode recognition: The LLM model identifies common fault types of the conveying equipment and predicts risks based on the embedded historical data and existing knowledge.

[0105] The existing knowledge, that is, the general data, is the common sense publicly available in this field.

[0106] Specifically, this process includes the following aspects:

[0107] Fault mode recognition: By analyzing the abnormal features and known fault cases in the general data, a classifier or model for fault modes is constructed. It involves machine learning algorithms (such as decision trees, neural networks, etc.) to automatically discover the abnormal performances of the equipment in different operating states and associate these abnormalities with predefined fault types.

[0108] Risk prediction: Based on the identified specific fault modes and relevant historical data, statistical methods or time series analysis techniques are used to evaluate the impact degree of each possible fault on the system performance. For example, the probability of a certain fault occurrence and its impact on the overall stability and safety of the equipment can be calculated, and corresponding risk scores or warning levels can be given accordingly.

[0109] By integrating historical information and real-time monitoring data, accurate identification of potential fault types of the conveying equipment and effective evaluation of their potential harm degrees are achieved, thus providing a guiding basis for subsequent maintenance work.

[0110] Step S4.2: Operating state optimization: The LLM model combines historical production experience, that is, the well-known production experience and process rules, analyzes parameters such as the conveying path, conveying speed, and equipment load, and outputs optimization suggestions, making intelligent decisions by fully utilizing the existing knowledge base and the latest production experience. Optimize the network structure, adjust the bandwidth allocation, and improve the efficiency of data transmission.

[0111] Step S4.3: Trend analysis and warning: The LLM model performs trend analysis on the data, identifies potential abnormal patterns, and issues warning messages in advance.

[0112] Specifically, when the operating data shows phenomena such as a significant increase in the data fluctuation range, a sudden decrease or increase in certain key performance indicators, or a frequent occurrence of a certain type of change trend within a specific time period, the system will consider that there are corresponding abnormal situations. For example: A certain component of the equipment may start to wear excessively; A certain subsystem shows signs of failure; External environmental factors affect and lead to the current abnormal phenomenon.

[0113] Based on the above analysis, the system will send warning messages to relevant responsible persons or maintenance teams (such as via email, SMS, or platform notifications), including specific anomaly descriptions and preliminary countermeasures recommended. This can promptly alert relevant personnel to pay attention to and handle potential problems, thereby avoiding possible major consequences.

[0114] Step S5: Real-time monitoring and automatic adjustment;

[0115] The optimization suggestions and warning messages generated by the LLM model are sent to the ECS via a Lambda function, and then forwarded by the ECS to the downstream intelligent control system for execution. The intelligent control system dynamically adjusts the mechanical transmission process based on the instructions output by the LLM model:

[0116] Step S5.1: Fault warning and shutdown protection: When a potential fault is predicted, the ECS instructs the control system to execute corresponding protection measures, such as reducing the load, pausing operation, or switching to an alternative line.

[0117] Step S5.2: Automatic adjustment: According to the LLM optimization suggestions, the transmission parameters (speed, load distribution, path planning) are adjusted in real time to adapt to the current production status, thereby achieving a more optimal operating state. This step is crucial for optimizing the data transmission strategy, ensuring the continuous optimization and improvement of the model.

[0118] Step S6: Feedback and self-learning optimization;

[0119] After the adjustment or intervention is executed, the intelligent control system records the new data (actual operation results and feedback information) into the database again through the ECS and Lambda, forming a feedback loop.

[0120] Step S6.1: Feedback loop optimization: The effect data after each adjustment is re-embedded (through an embedding model) and stored for evaluating the effectiveness of the LLM model suggestions.

[0121] Step S6.2: Adaptive learning: The LLM model regularly calls these new data for retraining or parameter fine-tuning, continuously enhancing its adaptability to real working conditions and improving the accuracy of fault prediction and optimization decisions.

[0122] The closed-loop learning mechanism not only improves the system's response speed but also enhances its self-optimization ability.

[0123] Step S7: Data storage and report generation;

[0124] All process data (raw, preprocessed, embedded, analysis results, and adjustment logs) are stored in the database for a long time. The database administrator can maintain the data structure, indexes, and security policies according to production requirements. By storing the same data at multiple locations, data loss or damage can be prevented. The system can call Lambda and LLM models through the ECS interface on demand to generate periodic or customized analysis reports, providing decision support information to production management.

[0125] According to the present invention, a mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model is provided, including: a production line, sensors and monitoring devices, an ECS, a backend service, an embedding model, a database, a large language model, and an intelligent control system;

[0126] The backend service includes a search module.

[0127] Sensors and detection devices are arranged in the production line to collect real-time data, which is transmitted to the backend service through the ECS;

[0128] The backend service preprocesses and standardizes the real-time data to obtain standardized data, interacts with the embedding model for semantic embedding and vectorization processing, stores the obtained embedded data in the database, and calls the large language model to analyze the embedded data to obtain optimization suggestions and warning information for the production line;

[0129] The intelligent control system dynamically adjusts the operation of the production line according to the optimization suggestions and warning information to obtain operation results and feedback information;

[0130] The intelligent control system records the operation results and feedback information in the database and retrains the large language model;

[0131] The ESC calls the backend service and the large language model to generate an analysis report, providing decision support information.

[0132] In more preferred examples, the sensors include temperature, vibration, pressure, speed, optoelectronics, and video monitoring;

[0133] The detection device is used to capture key parameters during the mechanical transmission process in real time, form real-time data with the data collected by the sensors, transmit it to the ECS through the industrial control network, and distribute data requests and transmission requirements to the backend service.

[0134] The ECS is an intermediate hub for data access and instruction scheduling.

[0135] The backend service triggers the preprocessing logic, calls algorithms or microservices, and uses statistical methods such as z-score, mean, and standard deviation to judge and eliminate extreme values or abnormal data in the data;

[0136] Convert real-time data from different sources and types into a unified format;

[0137] Align and correct the timestamps of real-time data from different sources, and store the cleaned and standardized data in the database.

[0138] In more preferred examples, the backend service interacts with the embedding model through the search module, extracts features and semantically vectorizes the standardized data or the text description data captured by the search model to obtain embedding data;

[0139] Store the embedding data in the database, maintain and manage the database, and optimize the index and storage strategy.

[0140] The search module is part of the backend service and is responsible for receiving inputs and invoking appropriate processing logic.

[0141] The embedding model is used to convert text or sensor data into a numerical vector representation.

[0142] The process data stored in the database includes real-time data, standardized data, embedding data, optimization suggestions, warning information, operation results, and feedback information. According to production requirements, backups are stored at different set locations, and the data structure, index, and security policy are maintained.

[0143] In more preferred examples, the large language model identifies and predicts the risk of fault types on the production line based on the embedding data and general data;

[0144] Analyze the transfer path, transfer speed, and equipment load parameters of the production line in combination with production experience and process rules to obtain basic operation data and output optimization suggestions;

[0145] Conduct trend analysis on the basic operation data, identify potential anomalies, and issue warning information.

[0146] By analyzing the anomalies and known fault cases in the general data, construct a classifier or model for fault modes, discover anomalies in different operating states of the equipment according to decision trees or neural networks, and associate the anomalies with fault types;

[0147] Based on the identified fault types and the relevant general data, use statistical methods or time series analysis techniques to evaluate the risk of each fault type.

[0148] The potential anomalies include that the operation data shows a significant increase in the data fluctuation range, a sudden decrease or increase in key performance indicators, or a changing trend of frequent occurrence of fault types within a specific time period.

[0149] The warning information includes specific anomaly descriptions and preliminary countermeasures recommended to be taken.

[0150] In more preferred examples, the optimization suggestions and warning information are sent to the ECS by the backend service and forwarded to the intelligent control system to dynamically adjust the operation of the production line, and record the operation results and feedback information.

[0151] When a potential anomaly is predicted, the ECS instructs the intelligent control system to execute load reduction, suspension of operation, or switching to an alternative line.

[0152] The intelligent control system adjusts the speed, load distribution, and path planning transfer parameters of the production line in real time according to the optimization suggestions;

[0153] The operation results and feedback information are re-embedded through the embedding model to obtain new data, and are recorded and stored in the database through the ECS and the backend service.

[0154] The large language model calls the new data at set time intervals for retraining or parameter fine-tuning.

[0155] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be arbitrarily combined with each other.

Claims

1. A mechanical intelligent manufacturing steel CPS transmission and monitoring method based on large language models, characterized in that, Including: Step S1: Arrange sensors and detection devices in the production line to collect real-time data, and transmit it to the backend service through ECS; Step S2: The backend service preprocesses and standardizes the real-time data to obtain standardized data; Step S3: The backend service performs semantic embedding and vectorization processing on the standardized data to obtain embedded data and stores it in the database; Step S4: The backend service calls a large language model to analyze the embedded data to obtain optimization suggestions and warning information for the production line; Step S5: The intelligent control system dynamically adjusts the operation of the production line according to the optimization suggestions and warning information to obtain operation results and feedback information; Step S6: The intelligent control system records the operation results and feedback information in the database and retrains the large language model; Step S7: Call the backend service and the large language model through ESC to generate an analysis report and provide decision support information.

2. The method for mechanical intelligent manufacturing steel CPS transmission and monitoring based on a large language model according to claim 1, characterized in that The sensors include temperature, vibration, pressure, speed, optoelectronics, and video monitoring; The detection device is used to capture key parameters during the mechanical transmission process in real time, form real-time data with the data collected by the sensors, transmit it to the ECS through the industrial control network, and distribute data requests and transmission requirements to the backend service; The ECS is an intermediate hub for data access and instruction scheduling; The said Step S2 includes: Step S2.1: Use the backend service to trigger the preprocessing logic, call algorithms or microservices, and judge and eliminate extreme values or abnormal data in the data by means of statistics such as z-score, mean, and standard deviation; Step S2.2: Convert real-time data from different sources and types into a unified format; Step S2.3: Align and correct the timestamps of real-time data from different sources through the backend service, and store the cleaned and standardized data in the database.

3. The method for steel CPS transmission and monitoring in mechanical intelligent manufacturing based on large language models according to claim 1, characterized in that, The said Step S3 includes: Step S3.1: The backend service interacts with the embedding model through the search module, extracts features and performs semantic vectorization on the standardized data or text description data captured by the search model to obtain embedded data; Step S3.2: Store the embedded data in the database, maintain and manage the database, and optimize the index and storage strategy; The search module is a part of the backend service and is responsible for receiving input and calling appropriate processing logic; The embedding model is used to convert text or sensor data into a numerical vector representation; The process data stored in the database includes real-time data, standardized data, embedded data, optimization suggestions, warning information, operation results, and feedback information. According to production requirements, backups are stored at different set locations, and the data structure, index, and security strategy are maintained.

4. The method for mechanical intelligent manufacturing steel CPS transmission and monitoring based on large language model according to claim 1, characterized in that, The said Step S4 includes: Fault mode identification step: The large language model identifies the fault types and predicts risks for the production line based on the embedded data and general data; Operation status optimization step: The large language model analyzes the transmission path, transmission speed, and equipment load parameters of the production line in combination with production experience and process rules to obtain basic operation data and output optimization suggestions; Trend analysis and warning step: The large language model performs trend analysis on the basic operation data, identifies potential anomalies, and issues warning information; The fault mode recognition steps include: Recognition step: By analyzing the anomalies in the general data and known fault cases, construct a classifier or model for the fault mode, discover the anomalies of the device in different operating states according to the decision tree or neural network, and associate the anomalies with the fault types; Risk prediction step: Based on the identified fault types and the relevant general data, adopt statistical methods or time series analysis techniques to evaluate the risk of each fault type; The potential anomalies include that the operation data shows a significant increase in the data fluctuation range, a sudden decrease or increase in the key performance indicators, or a change trend of frequent occurrence of fault types within a specific time period; The warning information includes specific anomaly descriptions and preliminary countermeasures recommended to be taken.

5. The method for mechanical intelligent manufacturing steel CPS transmission and monitoring based on large language model according to claim 1, characterized in that, In step S5, the optimization suggestions and warning information are sent to the ECS by the backend service and forwarded to the intelligent control system to dynamically adjust the operation of the production line, and record the operation results and feedback information, including; Fault warning and shutdown protection step: When a potential anomaly is predicted, the ECS instructs the intelligent control system to execute load reduction, suspension of operation, or switching to a standby line; Automation adjustment step: According to the optimization suggestions, adjust the speed, load distribution, and path planning transfer parameters of the production line in real time; In step S6, the intelligent control system records the operation results and feedback information into the database through the ECS and the backend service, including: Feedback loop optimization step: The operation results and feedback information are re-embedded through the embedded model to obtain new data and stored in the database; Adaptive learning step: The large language model calls the new data for retraining or parameter fine-tuning according to the set time interval.

6. A mechanical intelligent manufacturing steel CPS transmission and monitoring system based on a large language model, characterized in that, Including: Production line, sensors and monitoring devices, ECS, backend service, embedded model, database, large language model, and intelligent control system; The backend service includes a search module; Arrange sensors and detection devices in the production line to collect real-time data, and transmit it to the backend service through the ECS; The backend service preprocesses and standardizes the real-time data to obtain standardized data, interacts with the embedded model for semantic embedding and vectorization processing, obtains the embedded data and stores it in the database, and calls the large language model to analyze the embedded data to obtain the optimization suggestions and warning information for the production line; The intelligent control system dynamically adjusts the operation of the production line according to the optimization suggestions and warning information to obtain the operation results and feedback information; The intelligent control system records the operation results and feedback information into the database and retrains the large language model; The ESC calls the backend service and the large language model to generate an analysis report and provide decision support information.

7. The mechanical intelligent manufacturing steel CPS transmission and monitoring system based on the large language model according to claim 6, characterized in that, The sensors include temperature, vibration, pressure, speed, optoelectronic, and video monitoring; The detection device is used to capture the key parameters in the mechanical transmission process in real time, form real-time data with the data collected by the sensors, transmit it to the ECS through the industrial control network, and distribute the data requests and transmission requirements to the backend service; The ECS is an intermediate hub for data access and instruction scheduling; The backend service triggers the preprocessing logic, calls algorithms or microservices, and judges and eliminates extreme values or abnormal data in the data by means of statistics such as z-score, mean, and standard deviation; Convert real-time data from different sources and types into a unified format; Align and correct the timestamps of real-time data from different sources, and store the cleaned and standardized data in the database.

8. The mechanical intelligent manufacturing steel CPS transmission and monitoring system based on the large language model according to claim 6, wherein The backend service interacts with the embedding model through the search module, extracts features and semantically vectorizes the standardized data or the text description data captured by the search model to obtain embedding data; Store the embedding data in the database, maintain and manage the database, and optimize the index and storage strategy; The search module is part of the backend service and is responsible for receiving inputs and invoking appropriate processing logic; The embedding model is used to convert text or sensor data into a numerical vector representation; The process data stored in the database includes real-time data, standardized data, embedding data, optimization suggestions, warning information, operation results, and feedback information. According to production requirements, backups are stored at different locations, and the data structure, index, and security policy are maintained.

9. The mechanical intelligent manufacturing steel CPS transmission and monitoring system based on the large language model according to claim 6, characterized in that, The large language model identifies the fault types and predicts risks for the production line based on the embedding data and general data; Analyze the transfer path, transfer speed, and equipment load parameters of the production line in combination with production experience and process rules to obtain basic operation data and output optimization suggestions; Conduct trend analysis on the basic operation data, identify potential anomalies, and issue warning information; By analyzing the anomalies and known fault cases in the general data, construct a classifier or model for the fault mode, discover anomalies of the equipment in different operating states according to the decision tree or neural network, and associate the anomalies with the fault types; Based on the identified fault types and the relevant general data, use statistical methods or time series analysis techniques to evaluate the risk of each fault type; The potential anomalies include that the operation data shows a significant increase in the data fluctuation range, a sudden decrease or increase in the key performance indicators, or a change trend of frequent occurrence of fault types within a specific time period; The warning information includes specific anomaly descriptions and preliminary countermeasures recommended to be taken.

10. The mechanical intelligent manufacturing steel CPS transmission and monitoring system based on the large language model according to claim 6, characterized in that, The optimization suggestions and warning information are sent to the ECS through the backend service for forwarding to the intelligent control system, dynamically adjust the operation of the production line, and record the operation results and feedback information; When potential anomalies are predicted, the ECS instructs the intelligent control system to execute load reduction, suspension of operation, or switching to an alternative line; The intelligent control system adjusts the transfer parameters such as the speed, load distribution, and path planning of the production line in real time according to the optimization suggestions; Re-embed the operation results and feedback information through the embedding model to obtain new data, and record and store them in the database through the ECS and the backend service; The large language model calls the new data at a set time interval for retraining or parameter fine-tuning.

Images