High-precision industrial quality control system based on AI large model and application method

Abstract

The invention discloses a high-precision industrial quality control system based on an AI large model and an application method, which are characterized by being applied to detection and measurement of industrial products. Comprising a product data acquisition module in an industrial production process, a data preprocessing module, an AI large model suitable for industrial quality control, a quality detection module, a result feedback and decision support module, a control module and a system management module, the application of the AI large model suitable for industrial quality control comprises the steps of constructing the AI large model suitable for industrial quality control, calling the AI large model suitable for industrial quality control, executing the control system, carrying out multi-modal data fusion, and carrying out multi-modal data fusion. The product data acquisition module in the industrial production process acquires data of multiple modes for quality detection and measurement, the AI large model suitable for industrial quality control adopts a data fusion algorithm, and the feedback module provides a visual detection result. And visually displaying the detected quality problem to a user in the form of an image and a chart.

Description

Technical Field

[0001] The present invention relates to the use of AI big models for automated inspection and quality control of industrial products, including but not limited to equipment fault prediction and maintenance, product quality control, production planning and scheduling optimization, industrial data analysis and mining, and more particularly to a high-precision industrial quality control system based on AI big models. Background Art

[0002] In the field of industrial production, ensuring product quality has always been a core issue of concern to manufacturers and consumers. Traditional quality control methods mainly rely on manual inspection and experience-based judgment. Although this method can detect obvious defects to a certain extent, it has many limitations and challenges:

[0003] Strong subjectivity: Relying on human inspection, factors such as the state, experience, and subjective judgment of the inspectors will greatly affect the accuracy and consistency of the test results. Different inspectors may have different quality judgments on the same product, resulting in insufficiently strict and accurate implementation of quality standards.

[0004] Inefficiency: Manual inspection is relatively slow, especially in large-scale production. It is difficult to conduct a detailed and comprehensive inspection of each product, which is prone to missed inspections and cannot meet the needs of efficient production.

[0005] Poor adaptability: Traditional automated testing equipment is unable to cope with complex, irregular product features or minor defects. Its fixed algorithms and preset rule working modes are difficult to adapt to product diversification and production changes, and cannot accurately identify atypical quality problems.

[0006] Lack of learning ability: Traditional equipment cannot learn and improve autonomously from large amounts of data like AI models. As production processes improve and products are updated, their detection accuracy and effectiveness will decrease, requiring continuous manual adjustments and upgrades.

[0007] Lack of data mining: Although industrial production has massive amounts of data, traditional quality control methods lack effective mining and analysis capabilities. A large amount of valuable information is idle, making it difficult to provide in-depth decision-making support for quality control and unable to discover potential quality problems and production optimization space.

[0008] Data island problem: Data from different production links and testing equipment are independent of each other, lacking effective integration and correlation analysis, making it difficult for quality control personnel to control product quality from a global perspective and unable to fully understand the impact of various factors on quality.

[0009] With the rapid development of artificial intelligence and machine learning technologies, high-precision industrial quality control systems based on AI big models have emerged. This system integrates advanced AI technology and machine learning algorithms to achieve automatic, accurate and efficient detection and control of industrial product quality. Summary of the invention

[0010] In order to solve the above-mentioned problems of the prior art, the present invention provides a high-precision industrial quality control system based on an AI big model, which is characterized in that it is applied to the detection and measurement of industrial products, including a product data acquisition module in the industrial production process, a data preprocessing module, an AI big model suitable for industrial quality control, a quality detection module, a result feedback and decision support module, a control module and a system management module. The product data acquisition module in the industrial production process is used to comprehensively collect product data in the industrial production process, and the data preprocessing module performs preprocessing operations such as data cleaning, data labeling, data conversion and normalization on the collected data in preparation for subsequent analysis and judgment. The AI ​​big model suitable for industrial quality control is used to preprocess the product data in the industrial production process and then analyze and train to generate high-precision industrial quality standards for products in the industrial production process. The AI ​​big model suitable for industrial quality control includes building an AI big model suitable for industrial quality control and calling an AI big model suitable for industrial quality control. The quality detection module calls the AI ​​big model suitable for industrial quality control to detect and measure the user's industrial products. The result feedback and decision support module feeds back the detection results to the user in real time so that they can understand the quality status of the product, and provides decision support for the user based on the detection results and historical data. The control module automatically adjusts the production line parameters or issues instructions according to the detection results to correct deviations and ensure product quality. The system management module stores and manages all relevant data, including historical data, detection results, model parameters, etc., for analysis and backtracking, monitors the system's operating status, promptly discovers and solves system failures and problems, and regularly maintains and upgrades the system to ensure system stability and reliability.

[0011] As an improvement of the high-precision industrial quality control system based on the AI ​​large model of the present invention, the product data acquisition module in the industrial production process is connected to various types of sensors (such as optical sensors, pressure sensors, displacement sensors, etc.) to collect various physical parameters of industrial products in the production process in real time, such as size, shape, weight, temperature, pressure and other data, and uses high-resolution image acquisition equipment, such as industrial cameras, scanners, etc., to perform comprehensive image acquisition of the product's appearance to detect appearance defects such as blemishes, scratches, color differences on the product surface.

[0012] As an improvement of the high-precision industrial quality control system based on the AI ​​large model of the present invention, in the manufacture of electronic chips, the product data acquisition module in the industrial production process acquires the circuit information on the chip surface by collecting the appearance image of the chip through an industrial camera.

[0013] As an improvement of the high-precision industrial quality control system based on the AI ​​large model of the present invention, the data preprocessing module preprocesses the collected data including the following three steps:

[0014] The first step is cleaning, removing noise, outliers, and duplicate data to ensure the accuracy and reliability of the data.

[0015] The second step is data labeling, which labels the cleaned data to provide accurate label information for model training. For image data, the labeler needs to mark the defect location and type of the product in the image; for numerical data, the corresponding product quality grade, qualified range and other information need to be labeled.

[0016] The third step is to convert the collected data in different formats and units into a unified format and unit, and normalize them so that the data can be effectively calculated and analyzed in the model.

[0017] As an improvement of the high-precision industrial quality control system based on the AI ​​big model of the present invention, the construction of the AI ​​big model suitable for industrial quality control includes the following four steps:

[0018] The first step is to select a deep learning model suitable for industrial quality control, where the deep learning model is a convolutional neural network (CNN) for image recognition and a recurrent neural network (RNN) for processing time series data, and build a deep learning model.

[0019] The second step is to combine machine learning algorithms, such as support vector machine (SVM), decision tree, etc., to improve the performance and accuracy of the model.

[0020] The third step is to use a large amount of labeled data to train and fine-tune the model, and continuously adjust the model parameters so that it can learn the complex relationship between product quality characteristics and qualification standards.

[0021] The fourth step is to evaluate the model. During the training process, the model is evaluated regularly, and the accuracy, recall rate, precision rate and other indicators of the model are tested using the validation set data. According to the evaluation results, the model is optimized, such as adjusting the structure of the model, increasing the amount of training data, adjusting the training parameters, etc., to improve the performance of the model.

[0022] As an improvement of the high-precision industrial quality control system based on the AI ​​big model of the present invention, the construction of the AI ​​big model suitable for industrial quality control includes selecting a large-scale language model based on the Transformer architecture suitable for industrial quality control, and the large-scale language model based on the Transformer architecture is adjusted and optimized for image and data analysis.

[0023] As an improvement of the high-precision industrial quality control system based on the AI ​​big model of the present invention, the construction of the AI ​​big model suitable for industrial quality control includes using a large amount of labeled data to train the AI ​​big model. These data come from historical production data, laboratory test data, and manually labeled samples. By continuously adjusting the parameters of the model, it can accurately identify product quality problems and defects. Distributed training and optimization algorithms are used to improve training efficiency and shorten model training time.

[0024] As an improvement of the high-precision industrial quality control system based on the AI ​​big model of the present invention, the calling of the AI ​​big model suitable for industrial quality control includes deploying the trained AI big model to the production line to perform real-time detection and analysis of the products being produced:

[0025] For image data, we use image recognition technology to detect product appearance defects, such as scratches, dents, and color differences, and use image processing technology to accurately measure product parameters such as size and angle to ensure that the product meets design requirements;

[0026] For sensor data, analyze whether the product's performance parameters are within the normal range, and use sensor data to perform dynamic measurements, such as vibration frequency, temperature changes, etc., to promptly identify potential quality problems.

[0027] Intelligent judgment and classification: Based on the test results, intelligent judgment and classification are carried out on products, and products are divided into different categories such as qualified products, unqualified products, and products that need further re-inspection, and the corresponding test data and results are recorded for subsequent traceability and analysis.

[0028] As an improvement of the high-precision industrial quality control system based on the AI ​​large model of the present invention, an abnormality detection and alarm module is also provided. When an abnormality or system failure is detected, the abnormality detection and alarm module can automatically or prompt the operator to handle it to minimize production interruption;

[0029] A report and statistics module is also provided, which generates quality control reports and statistical analysis to help management understand production quality status and system performance;

[0030] There is also a security and compliance module, which ensures that the system complies with industry security and compliance standards and protects data security and privacy;

[0031] An integration and compatibility module is also provided to ensure that the system can be seamlessly integrated with existing production line equipment and software, supporting multiple data formats and communication protocols.

[0032] The present invention provides an application method of a high-precision industrial quality control system based on an AI big model, characterized in that the high-precision industrial quality control system based on the AI ​​big model described in any one of claims 1 to 9 is executed, multimodal data fusion, the product data acquisition module in the industrial production process collects data of multiple modalities for quality inspection and measurement, the AI ​​big model suitable for industrial quality control adopts a data fusion algorithm, and the feedback module provides intuitive visual detection results, and intuitively displays the detected quality problems to users in the form of images and charts, so that users can quickly understand the quality status of the products.

[0033] As an improvement of the application method of the high-precision industrial quality control system based on the AI ​​large model of the present invention, the result feedback and decision support module workflow includes:

[0034] The first step is to send feedback information to the user's production system in a timely manner based on the test results. If quality problems are found, the alarm device can be automatically triggered to notify the user operator to handle it, or the parameters of the production equipment can be directly adjusted to achieve automatic control and avoid the continued production of problematic products.

[0035] The second step is decision support. Based on the test results and historical data, the system provides decision support for production personnel and managers. The system can analyze possible causes and make corresponding improvement suggestions, such as adjusting production process parameters, replacing raw materials, etc.

[0036] The third step is to automatically generate detailed quality reports, including information such as the number of products tested, the number of qualified products, the number of unqualified products, defect types and distribution. These reports can provide data support for the company's quality management and help the company formulate quality improvement plans.

[0037] As an improvement on the application method of the high-precision industrial quality control system based on the AI ​​large model of the present invention, the control module performs statistics and analysis on the detection results, provides data support for the optimization of the production process, analyzes the distribution and frequency of quality problems, identifies the weak links in the production process, and proposes improvement measures.

[0038] As an improvement on the application method of the high-precision industrial quality control system based on the AI ​​large model of the present invention, the system management module manages the users of the system, including user registration, login, permission setting, etc., and manages the data in the system, including data storage, backup, recovery, etc., to ensure the security and integrity of the data, prevent data loss and leakage, monitor the operating status of the system, promptly discover and resolve system faults and problems, and regularly maintain and upgrade the system to ensure the stability and reliability of the system.

[0039] The high-precision industrial quality control system and application method based on the AI ​​large model of the present invention have the following beneficial effects: greatly improving the accuracy and efficiency of quality inspection, whether it is the complex appearance defects of small household appliances such as electric kettles, toasters or coffee machines, or the minor abnormalities of parameters such as temperature, pressure, flow rate, etc. in the production process, they can be accurately and quickly identified, greatly reducing missed detection and false detection, and realizing real-time multimodal detection and intelligent judgment classification on the production line, ensuring the smooth circulation of qualified products, proper handling of repairable products, and timely isolation of unqualified products, effectively reducing the defective rate and improving the overall quality of the products. At the same time, the result feedback and decision support functions enable the management of the enterprise to quickly grasp the production quality status and trends, optimize the production process and adjust the supplier strategy based on data-driven decision suggestions, and ensure continuous improvement of production. In addition, the system management function guarantees the reasonable permissions of different users, the safe storage and efficient use of data, and the stable operation of the system, which helps manufacturing enterprises achieve excellent improvement in quality control, production efficiency, cost control and sustainable development from many aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of a preferred embodiment of a high-precision industrial quality control system and application method based on an AI large model of the present invention.

[0041] Figure 2 This is a flow chart of preprocessing the collected data by the data preprocessing module of the preferred embodiment of the high-precision industrial quality control system and application method based on the AI ​​large model of the present invention.

[0042] Figure 3 This is a flow chart of constructing an AI big model suitable for industrial quality control in the preferred embodiment of the high-precision industrial quality control system and application method based on the AI ​​big model of the present invention.

[0043] Figure 4 It is a flow chart of the application method of the preferred embodiment of the high-precision industrial quality control system and application method based on AI large model of the present invention. DETAILED DESCRIPTION

[0044] Below, combined with the attached Figure 1-4 As well as specific implementation methods, the present invention is further described. It should be noted that, under the premise of no conflict, the various technical features described below can be arbitrarily combined to form a new embodiment.

[0045] Reference Figure 1-4 In a preferred embodiment, the present invention is a high-precision industrial quality control system based on an AI large model. A high-precision industrial quality control system based on an AI large model is characterized in that it is applied to the detection and measurement of industrial products 1, including a product data acquisition module 2 in the industrial production process, a data preprocessing module 3, an AI large model suitable for industrial quality control 4, a quality detection module 5, a result feedback and decision support module 6, a control module 7 and a system management module 8. The product data acquisition module 2 in the industrial production process is used to comprehensively collect product data in the industrial production process, and the data preprocessing module 3 performs preprocessing operations such as data cleaning, data labeling, data conversion and normalization on the collected data to prepare for subsequent analysis and judgment. The AI ​​large model 4 suitable for industrial quality control is used to preprocess the product data in the industrial production process and then analyze and train to generate high-precision industrial quality standards for products in the industrial production process. , applying the AI ​​big model 4 suitable for industrial quality control includes constructing an AI big model 4 suitable for industrial quality control and calling the AI ​​big model 4 suitable for industrial quality control. The quality detection module 5 calls the AI ​​big model 4 suitable for industrial quality control to detect and measure the industrial product 1 of the user 9. The result feedback and decision support module 6 feeds back the detection results to the user 9 in real time so that they can understand the quality status of the product, and provides decision support for the user 9 based on the detection results and historical data. The control module 7 automatically adjusts the production line parameters or issues instructions according to the detection results to correct deviations and ensure product quality. The system management module 8 stores and manages all relevant data, including historical data, detection results, model parameters, etc., for analysis and backtracking, monitors the operating status of the system, promptly discovers and solves system faults and problems, and regularly maintains and upgrades the system to ensure the stability and reliability of the system.

[0046] In a preferred embodiment, the product data acquisition module 2 in the industrial production process is connected to various types of sensors (such as optical sensors, pressure sensors, displacement sensors, etc.) to collect various physical parameters of the industrial product 1 in the production process in real time, such as size, shape, weight, temperature, pressure and other data, and uses high-resolution image acquisition equipment, such as industrial cameras, scanners, etc., to perform comprehensive image acquisition of the product's appearance and detect appearance defects such as blemishes, scratches, color differences, etc. on the product surface.

[0047] In a preferred embodiment, in the manufacture of electronic chips, the product data acquisition module 2 in the industrial production process acquires the circuit information on the surface of the chip by acquiring the appearance image of the chip through an industrial camera.

[0048] Reference Figure 2 In a preferred embodiment, the data preprocessing module 3 performs preprocessing on the collected data including the following three steps:

[0049] 301. Cleaning: remove noise, outliers, and duplicate data to ensure data accuracy and reliability.

[0050] 302. Data labeling: label the cleaned data to provide accurate label information for model training. For image data, the labeler needs to mark the defect location, type and other information of the product in the image; for numerical data, the corresponding product quality grade, qualified range and other information need to be labeled.

[0051] 303. Data conversion and normalization: convert the collected data in different formats and units into a unified format and unit, and perform normalization so that the data can be effectively calculated and analyzed in the model.

[0052] The data preprocessing module 3 performs data cleaning, data labeling, data conversion and normalization on the real-time data. The data cleaning techniques adopted may include missing value processing (such as deletion method, filling method), outlier processing (based on statistical method or clustering method), and duplicate data processing (precise matching or fuzzy matching deduplication) technology; data labeling includes manual labeling assisted by professional labelers using labeling tools and semi-automatic labeling methods assisted by rules or models; data conversion and normalization include data type conversion (such as date time to timestamp, categorical variable numericalization), data format conversion (unified image and other data formats), as well as linear normalization (mapping data to the [0,1] interval) and standard deviation normalization (making the data mean 0 and standard deviation 1) and other technologies.

[0053] Reference Figure 3 In a preferred embodiment, the construction of the AI ​​big model 4 suitable for industrial quality control includes the following four steps:

[0054] 401. Select a deep learning model suitable for industrial quality control, wherein the deep learning model is a convolutional neural network (CNN) for image recognition and a recurrent neural network (RNN) for processing time series data, and construct a deep learning model.

[0055] The construction of the AI ​​large model suitable for industrial quality control 4 selects a deep learning model suitable for industrial quality control, a convolutional neural network (CNN) for image recognition, and a recurrent neural network (RNN) for processing time series data, combined with machine learning algorithms, support vector machines (SVM), decision trees, etc., to improve the performance and accuracy of the model, use a large amount of labeled data to train and fine-tune the model, and continuously adjust the model's parameters so that it can learn the complex relationship between product quality characteristics and qualified standards. During the training process, the model is regularly evaluated and optimized.

[0056] 402. Combine machine learning algorithms, such as support vector machine (SVM), decision tree, etc. to improve the performance and accuracy of the model.

[0057] 403. Use a large amount of labeled data to train and fine-tune the model, and continuously adjust the model parameters so that it can learn the complex relationship between product quality characteristics and qualification standards.

[0058] 404. Evaluate the model. During the training process, evaluate the model regularly and use the validation set data to test the model's accuracy, recall rate, precision and other indicators. Based on the evaluation results, optimize the model, such as adjusting the model's structure, increasing the amount of training data, adjusting training parameters, etc., to improve the model's performance.

[0059] In a preferred embodiment, the construction of the AI ​​large model 4 suitable for industrial quality control includes selecting a large-scale language model based on the Transformer architecture suitable for industrial quality control, and the large-scale language model based on the Transformer architecture is adjusted and optimized for image and data analysis.

[0060] In a preferred embodiment, the construction of an AI big model 4 suitable for industrial quality control includes using a large amount of labeled data to train the AI ​​big model. These data come from historical production data, laboratory test data, and manually labeled samples. By continuously adjusting the parameters of the model, it can accurately identify product quality problems and defects. Distributed training and optimization algorithms are used to improve training efficiency and shorten model training time.

[0061] In a preferred embodiment, the calling of the AI ​​big model 4 suitable for industrial quality control includes deploying the trained AI big model 4 suitable for industrial quality control to the production line to perform real-time detection and analysis of the products being produced:

[0062] For image data, image recognition technology is used to detect product appearance defects such as scratches, dents, and color differences, and image processing technology is used to accurately measure product parameters such as size and angle to ensure that the product meets design requirements.

[0063] For sensor data, analyze whether the product's performance parameters are within the normal range, and use sensor data to perform dynamic measurements, such as vibration frequency, temperature changes, etc., to promptly identify potential quality problems.

[0064] Intelligent judgment and classification: Based on the test results, intelligent judgment and classification are carried out on products, and products are divided into different categories such as qualified products, unqualified products, and products that need further re-inspection, and the corresponding test data and results are recorded for subsequent traceability and analysis.

[0065] The quality inspection module 5 deploys the trained AI large model 4 suitable for industrial quality control to the production line, performs real-time inspection and analysis on the products being produced, and uses its fast processing capabilities to quickly identify product defects, anomalies or non-compliance with quality standards, and issue timely alarms. For example, real-time monitoring of products on high-speed production lines. Multimodal inspection technology can combine appearance, physical parameters with other modal data such as sound and vibration to comprehensively evaluate quality. For example, in mechanical equipment inspection, the condition of internal components can be judged by analyzing operating sound and vibration. Intelligent judgment and classification technology accurately divides products into qualified, unqualified and re-inspection categories based on the inspection results, and records data for easy traceability and analysis, thereby achieving comprehensive, efficient and intelligent inspection of the quality of industrial products 1.

[0066] In other embodiments, an abnormality detection and alarm module is also provided. When an abnormality or system failure is detected, it can automatically or prompt the operator to handle it to minimize production interruption. Its technical solutions include: abnormality detection technology monitors and analyzes a large amount of real-time data, and uses statistical analysis methods, machine learning algorithms or rule-based logical judgments to identify abnormal data and potential fault signs that deviate from the normal production mode. Data fusion technology integrates and correlates multi-source data from different sensors and different production links to improve the accuracy and comprehensiveness of abnormality detection. Alarm triggering technology quickly generates an alarm signal when an abnormality is detected based on preset thresholds, rules or model evaluation results, and promptly informs the operator through various methods such as sound and light alarm devices, SMS notifications, email reminders or system interface pop-ups. In addition, intelligent diagnosis technology can further analyze and diagnose abnormal situations, provide operators with detailed fault information and handling suggestions, and assist them in taking countermeasures quickly and accurately, effectively reducing production interruption time and ensuring the continuity and stability of the production process.

[0067] In other embodiments, a report and statistics module is also provided to generate quality control reports and statistical analysis to help management understand the production quality status and system performance. The technical solution is to use data analysis technology, apply statistical analysis methods such as mean, standard deviation, variance and other calculations, as well as data mining algorithms, to deeply analyze the data and mine various indicators of product quality, defect distribution patterns, quality trends and other important information. The next is report generation technology, which uses automated templates and programming languages ​​to efficiently generate detailed quality control reports in the form of clear and intuitive charts, tables and text descriptions according to preset formats and content requirements with the help of automated templates and programming languages. The last is visualization technology, which uses a graphical interface to present the report content in an intuitive and easy-to-understand manner, such as a bar chart showing the number of qualified products in different time periods, and a line chart showing the changing trend of quality indicators, etc., so that management can quickly and accurately grasp the production quality status and system performance.

[0068] In other embodiments, a security and compliance module is also provided to ensure that the system complies with industry security and compliance standards and protects data security and privacy. The technical solution is to use encryption technology to encrypt data, such as using symmetric encryption or asymmetric encryption algorithms, to ensure the confidentiality of data during storage and transmission and prevent data from being stolen or tampered with. Access control technology restricts unauthorized users' access to system resources and data through strict user 9 authentication, authorization and access permission management to ensure data security and integrity. Data desensitization technology can process sensitive data to hide or replace key information without affecting data availability, thereby reducing the risk of data leakage. Network security protection technologies such as firewalls and intrusion detection systems can monitor and prevent external network attacks in real time, prevent malicious intrusions, and ensure the network security of the system. In addition, there is security audit technology to record and audit system operations and data access, so as to promptly discover and trace potential security issues and ensure that the system always complies with industry security and compliance standards.

[0069] In other embodiments, an integration and compatibility module is also provided to ensure that the system can be seamlessly integrated with existing production line equipment and software and support multiple data formats and communication protocols. Its technologies include device interface technology, which realizes physical connection and data interaction between the system and the equipment by developing special interfaces adapted to different production line equipment; data format conversion technology, which uses data analysis, mapping and other methods to standardize the conversion of data in different formats such as text, images, and numerical values ​​so that they can be uniformly processed in the system; communication protocol adaptation technology, which ensures accurate and stable communication between the system and various devices and software by supporting and converting multiple common communication protocols such as TCP / IP, OPC, Profibus, etc.; the application of middleware technology plays the role of a bridge and a link, which can shield the differences between underlying equipment and software, provide a unified service interface, and simplify the complexity of system integration, thereby realizing a high degree of integration and good compatibility of the entire system and ensuring the coordinated operation of all links in the industrial production process.

[0070] The present invention provides an application method of a high-precision industrial quality control system based on an AI big model, characterized in that the high-precision industrial quality control system based on the AI ​​big model described in any one of claims 1-9 is executed, multimodal data fusion, the product data acquisition module 2 in the industrial production process collects data of multiple modalities for quality detection and measurement, the AI ​​big model 4 suitable for industrial quality control adopts a data fusion algorithm, and the feedback module provides intuitive visual detection results, and intuitively displays the detected quality problems to the user 9 in the form of images and charts, so that the user 9 can quickly understand the quality status of the product.

[0071] In a preferred embodiment, the workflow of the result feedback and decision support module 6 includes:

[0072] According to the test results, feedback information is sent to the user's production system in a timely manner. If quality problems are found, the alarm device can be automatically triggered to notify the user's operator to handle it, or the parameters of the production equipment can be directly adjusted to achieve automatic control and avoid the continued production of problematic products.

[0073] Decision support: Based on test results and historical data, it provides decision support for production personnel and managers. The system can analyze possible causes and make corresponding improvement suggestions, such as adjusting production process parameters, replacing raw materials, etc.

[0074] It can automatically generate detailed quality reports, including information such as the number of products tested, the number of qualified products, the number of unqualified products, defect types and distribution. These reports can provide data support for the company's quality management and help the company formulate quality improvement plans.

[0075] The result feedback and decision support module 6 uses result feedback technology to provide real-time feedback of the test results to production personnel and management personnel in various forms such as display screens, mobile phone text message notifications, and email notifications; uses data mining and analysis technology to mine the potential relationship between quality data and production factors based on test results and historical data, and provides a basis for decision support, such as analyzing the reasons for the increase in defect rate and proposing improvement suggestions; with the help of automated report generation technology, it can automatically generate a quality report containing detailed information such as the number of tested products, the number of qualified products, the number of unqualified products, defect types and distribution, etc., to help enterprises formulate quality improvement plans and realize intelligent quality control and management.

[0076] In a preferred embodiment, the control module 7 performs statistics and analysis on the test results, provides data support for the optimization of the production process, analyzes the distribution and frequency of quality problems, finds out the weak links in the production process, and proposes improvement measures.

[0077] The control module 7 automatically adjusts the production line parameters or issues instructions based on the test results to correct deviations and ensure product quality. Its technical solution is data communication technology, which can realize the rapid and accurate transmission of test result data between the quality control system and the production line equipment to ensure timely information exchange. The second is data analysis and processing technology, which conducts in-depth analysis of the test results, determines the deviation situation and corresponding adjustment strategy based on preset quality standards and rules, and the third is automatic control technology. According to the adjustment strategy obtained through analysis, precise instructions are sent to the controller or actuator of the production line through programming logic and interface protocol to automatically adjust key production parameters such as temperature, pressure, speed, etc., or trigger alarms, shutdowns and other operations, so as to correct deviations in time and ensure the stability and consistency of product quality.

[0078] In a preferred embodiment, the system management module 8 manages the users 9 of the system, including the registration, login, and permission setting of the users 9, and manages the data in the system, including the storage, backup, and recovery of the data, to ensure the security and integrity of the data and prevent data loss and leakage. The system's operating status is monitored to promptly detect and resolve system failures and problems, and the system is regularly maintained and upgraded to ensure the stability and reliability of the system.

[0079] The system management module 8 stores and manages all relevant data, including historical data, test results, model parameters, etc., for analysis and backtracking, monitors the system's operating status, promptly discovers and solves system failures and problems, and regularly maintains and upgrades the system. Its technical solution involves user 9 management technology. Through the management of user 9 registration, login and permission setting, it ensures that different users 9 have corresponding operating permissions to ensure the security and orderly operation of the system; uses data management technology to be responsible for data storage, backup and recovery, effectively ensures data integrity and security, and prevents data loss and leakage; uses system monitoring and maintenance technology to monitor the system's operating status in real time, promptly discovers and solves failure problems, and regularly maintains and upgrades the system to maintain the stability and reliability of the system, thereby providing strong support for the efficient operation of the entire high-precision industrial quality control system.

[0080] Taking the production workshop of a small household appliance manufacturing enterprise as an example, the present invention provides a high-precision industrial quality control system based on an AI large model and an application method as follows: sensors are installed at key points of the production process such as the inner tank stretching and outer shell assembly of the electric kettle, the heating wire installation and baking tray assembly of the toaster, the water pump installation of the coffee machine, and the vicinity of the grinder of the small household appliance production line to obtain data, and high-definition industrial cameras are installed along the production line to collect appearance data, such as photographing the appearance of the electric kettle from different angles, including the kettle body, lid, handle and other parts, to capture whether there are scratches, pits, uneven color and other defects on the surface, and at the same time check whether the connection between the spout and the kettle body is smooth. For the toaster, the camera photographs the flatness of its outer shell, the smoothness of the baking tray, and the installation of buttons and indicator lights to ensure that there are no defects in the appearance and the operation The buttons are flexible and the appearance is exquisite; in the data preprocessing stage, the collected data is cleaned, sensor outliers and image artifacts are removed, and the data is annotated according to quality standards. For example, for the pressure data of the inner tank of the electric kettle, the pressure range value that meets the quality requirements is marked; for the appearance image, the location and type of defects such as scratches and pits are accurately marked, and the temperature data of the toaster is marked with the temperature range suitable for baking. For the coffee machine, the normal flow range of the water pump and the normal vibration mode of the grinder are marked. The qualified quality range, the data collected by different sensors are uniformly converted into a format suitable for model processing, for example, the temperature data is converted from Celsius to Kelvin, and the pressure data is unified from different units to international standard units. In the model training stage, select Convolutional neural network (CNN) is selected to process appearance image data, and its powerful image feature extraction ability is used to identify tiny defects on the product surface. For time series data collected by sensors (such as changes in temperature and pressure over time), recurrent neural network structures such as long short-term memory network (LSTM) or gated recurrent unit (GRU) are used, combined with support vector machine (SVM) to optimize the model to distinguish the boundary conditions between qualified and unqualified quality and improve the accuracy of the model. For example, when judging the quality of the inner pot of an electric kettle, the judgment of CNN on appearance and the analysis results of SVM on pressure data are combined, and a large amount of labeled quality inspection data of small kitchen appliances from different batches and production periods, including normal products and products with various defects, are used. Product data with quality problems are used to train the model. By adjusting the model's weights and parameters, the model can learn the complex relationship between product quality features and qualified standards. For example, the model can learn the relationship between appearance defects and quality problems of different styles of electric kettles, and the relationship between abnormal toaster temperature and baking effects. The trained model is evaluated with the reserved validation set data, and the model's accuracy, recall rate, F1 value and other indicators are calculated. If the model's recognition accuracy for coffee machine grinder failure is low, it may be due to insufficient training data or unreasonable model structure. The model can be optimized by increasing training data with similar failures, adjusting the number of neural network layers or neurons, and optimizing the SVM kernel function.Real-time detection is performed in the quality inspection stage. The models of electric kettles, toasters, and coffee machines analyze sensor and image data in real time during production. If there is a problem, an alarm will be issued. Multi-modal detection is also used to evaluate the quality of multiple types of data, and the products are intelligently classified according to the results. In the result feedback and decision support stage, the display screen is used to display and notify relevant personnel of the real-time feedback results, and decision-making basis is provided based on detection and historical data and quality reports are generated. In the system management stage, user management is done well, different permissions are set for different positions, and data storage, backup, recovery and access audit data management are performed. The system status is monitored in real time, and problems are promptly maintained and upgraded. ;

[0081] The high-precision industrial quality control system and application method based on the AI ​​large model of the present invention have the following beneficial effects: the accuracy and efficiency of quality inspection are greatly improved, whether it is the complex appearance defects of small household appliances such as electric kettles, toasters or coffee machines, or the slight abnormalities of parameters such as temperature, pressure, flow rate in the production process, they can be accurately and quickly identified, greatly reducing missed detection and false detection; real-time multimodal detection and intelligent judgment classification are realized on the production line to ensure the smooth circulation of qualified products, the proper handling of repairable products, and the timely isolation of unqualified products, which effectively reduces the defective rate and improves the overall quality of the products; at the same time, the result feedback and decision support functions enable the enterprise management to quickly grasp the production quality status and trends, optimize the production process and adjust the supplier strategy based on data-driven decision suggestions, and ensure continuous improvement of production. In addition, the system management function guarantees the reasonable permissions of different users, the safe storage and efficient use of data, and the stable operation of the system, which helps manufacturing companies to achieve excellent improvement in quality control, production efficiency, cost control and sustainable development from many aspects.

[0082] The present invention has been described in detail above. The above description is only a preferred embodiment of the present invention and should not limit the scope of implementation of the present invention. That is, all equivalent changes and modifications made within the scope of this application should still fall within the scope of the present invention.

Claims

1. A high-precision industrial quality control system based on AI large model, characterized in that: Applied to the detection and measurement of industrial products, it includes a product data acquisition module in the industrial production process, a data preprocessing module, an AI big model suitable for industrial quality control, a quality detection module, a result feedback and decision support module, a control module and a system management module. The product data acquisition module in the industrial production process is used to comprehensively collect product data in the industrial production process. The data preprocessing module performs preprocessing operations such as data cleaning, data labeling, data conversion and normalization on the collected data to prepare for subsequent analysis and judgment. The AI ​​big model suitable for industrial quality control is used to preprocess the product data in the industrial production process and then analyze and train to generate high-precision industrial quality standards for products in the industrial production process. The application of the AI ​​big model suitable for industrial quality control includes building a system suitable for industrial quality control. The AI ​​big model for control and the AI ​​big model suitable for industrial quality control are called. The quality inspection module calls the AI ​​big model suitable for industrial quality control to inspect and measure the user's industrial products. The result feedback and decision support module feeds back the inspection results to the users in real time so that they can understand the quality status of the products, and provide decision support for users based on the inspection results and historical data. The control module automatically adjusts the production line parameters or issues instructions according to the inspection results to correct deviations and ensure product quality. The system management module stores and manages all relevant data, including historical data, inspection results, model parameters, etc., for easy analysis and backtracking, monitors the system's operating status, promptly discovers and solves system faults and problems, and regularly maintains and upgrades the system to ensure system stability and reliability.

2. According to claim 1, a high-precision industrial quality control system based on AI large model is characterized in that: The product data acquisition module in the industrial production process is connected to various types of sensors (such as optical sensors, pressure sensors, displacement sensors, etc.) to collect various physical parameters of industrial products in the production process in real time, such as size, shape, weight, temperature, pressure and other data, and uses high-resolution image acquisition equipment, such as industrial cameras, scanners, etc., to perform comprehensive image acquisition of the product's appearance and detect appearance defects such as blemishes, scratches, color differences, etc. on the product surface.

3. A high-precision industrial quality control system based on AI large model according to claim 2, characterized in that: In the manufacture of electronic chips, the product data acquisition module in the industrial production process acquires the circuit information on the chip surface by collecting the appearance image of the chip through an industrial camera.

4. According to claim 1, a high-precision industrial quality control system based on AI large model is characterized in that: The data preprocessing module preprocesses the collected data in the following three steps: The first step is cleaning, removing noise, outliers, and duplicate data to ensure the accuracy and reliability of the data. The second step is data labeling, which labels the cleaned data to provide accurate label information for model training. For image data, the labeler needs to mark the defect location and type of the product in the image; for numerical data, the corresponding product quality grade, qualified range and other information need to be labeled. The third step is to convert the collected data in different formats and units into a unified format and unit, and normalize them so that the data can be effectively calculated and analyzed in the model.

5. According to claim 1, a high-precision industrial quality control system based on AI large model is characterized in that: The construction of an AI big model suitable for industrial quality control includes: The first step is to select a deep learning model suitable for industrial quality control, where the deep learning model is a convolutional neural network (CNN) for image recognition and a recurrent neural network (RNN) for processing time series data, and build a deep learning model. The second step is to combine machine learning algorithms, such as support vector machine (SVM), decision tree, etc., to improve the performance and accuracy of the model. The third step is to use a large amount of labeled data to train and fine-tune the model, and continuously adjust the model parameters so that it can learn the complex relationship between product quality characteristics and qualification standards. The fourth step is to evaluate the model. During the training process, the model is evaluated regularly, and the accuracy, recall rate, precision rate and other indicators of the model are tested using the validation set data. According to the evaluation results, the model is optimized, such as adjusting the structure of the model, increasing the amount of training data, adjusting the training parameters, etc., to improve the performance of the model.

6. According to claim 1, a high-precision industrial quality control system based on AI large model is characterized in that: The construction of the AI ​​large model suitable for industrial quality control includes selecting a large-scale language model based on the Transformer architecture suitable for industrial quality control, and the large-scale language model based on the Transformer architecture is adjusted and optimized for image and data analysis.

7. A high-precision industrial quality control system based on AI large model according to claim 5 or 6, characterized in that: The construction of an AI big model suitable for industrial quality control includes using a large amount of labeled data to train the AI ​​big model. These data come from historical production data, laboratory test data, and manually labeled samples. By continuously adjusting the parameters of the model, it can accurately identify product quality problems and defects. Distributed training and optimization algorithms are used to improve training efficiency and shorten model training time.

8. A high-precision industrial quality control system based on AI large model according to claim 1 or 5, characterized in that: The calling of the AI ​​big model suitable for industrial quality control includes deploying the trained AI big model to the production line to perform real-time detection and analysis of the products being produced: For image data, we use image recognition technology to detect product appearance defects, such as scratches, dents, and color differences, and use image processing technology to accurately measure product parameters such as size and angle to ensure that the product meets design requirements; For sensor data, analyze whether the product's performance parameters are within the normal range, and use sensor data to perform dynamic measurements, such as vibration frequency, temperature changes, etc., to promptly identify potential quality problems. Intelligent judgment and classification: Based on the test results, intelligent judgment and classification are carried out on products, and products are divided into different categories such as qualified products, unqualified products, and products that need further re-inspection, and the corresponding test data and results are recorded for subsequent traceability and analysis.

9. The high-precision industrial quality control system based on AI large model according to claim 1 is characterized by: There is also an abnormality detection and alarm module. When an abnormality or system failure is detected, the abnormality detection and alarm module can automatically or prompt the operator to handle it to minimize production interruption; A report and statistics module is also provided, which generates quality control reports and statistical analysis to help management understand production quality status and system performance; There is also a security and compliance module, which ensures that the system complies with industry security and compliance standards and protects data security and privacy; An integration and compatibility module is also provided to ensure that the system can be seamlessly integrated with existing production line equipment and software, supporting multiple data formats and communication protocols.

10. An application method of a high-precision industrial quality control system based on an AI large model, characterized in that: Execute the high-precision industrial quality control system based on the AI ​​big model as described in any one of claims 1-9, multi-modal data fusion, the product data acquisition module in the industrial production process collects data of multiple modes for quality detection and measurement, the AI ​​big model suitable for industrial quality control adopts a data fusion algorithm, and the feedback module provides intuitive visual detection results, and intuitively displays the detected quality problems to the user in the form of images and charts, so that the user can quickly understand the quality status of the product; The workflow of the result feedback and decision support module includes: The first step is to send feedback information to the user's production system in a timely manner based on the test results. If quality problems are found, the alarm device can be automatically triggered to notify the user's operator to handle it, or directly adjust the parameters of the production equipment to achieve automatic control and avoid the continued production of problematic products; The second step is decision support. Based on the test results and historical data, the system provides decision support for production personnel and managers. The system can analyze the possible causes and make corresponding improvement suggestions, such as adjusting production process parameters and replacing raw materials. The third step is to automatically generate detailed quality reports, including the number of products tested, the number of qualified products, the number of unqualified products, defect types and distribution, etc. These reports can provide data support for the company's quality management and help the company formulate quality improvement plans; The control module counts and analyzes the test results, provides data support for the optimization of the production process, analyzes the distribution and frequency of quality problems, finds out the weak links in the production process, and proposes improvement measures; The system management module manages the users of the system, including user registration, login, permission setting, etc., and manages the data in the system, including data storage, backup, recovery, etc., to ensure the security and integrity of the data, prevent data loss and leakage, monitor the operating status of the system, promptly discover and resolve system failures and problems, and regularly maintain and upgrade the system to ensure the stability and reliability of the system.

Images