Quality tracing and predictive maintenance analysis platform for intelligent manufacturing

Abstract

The invention discloses a quality tracing and predictive maintenance analysis platform for intelligent manufacturing, and the platform comprises a comprehensive platform which comprises a quality tracing module, a predictive maintenance module, a data analysis and decision support module, and a data storage module. Unified acquisition, management and analysis of product full life cycle data and equipment multi-source operation data are realized. According to the platform, a unified identification system is adopted to encode and analyze products and associated objects thereof, so that data of links such as raw material purchasing, production and processing, logistics storage and finished product inspection can be accurately associated, and the platform can perform modeling and judgment on an equipment operation trend by utilizing equipment real-time data, process data and environmental data; the data analysis and decision support module of the platform performs statistical analysis and associated mining on quality data and equipment data, so that an enterprise can identify key factors influencing quality and equipment states, and analysis limitation caused by data isolation is avoided.

Description

Technical Field

[0001] This application relates to the field of intelligent manufacturing technology, and in particular to a quality traceability and predictive maintenance analysis platform for intelligent manufacturing. Background Technology

[0002] With the rapid development of Industry 4.0 and intelligent manufacturing, manufacturing enterprises are placing higher demands on the digitalization, transparency, and controllability of their production processes. Quality management throughout the entire product lifecycle and the stable operation of production equipment have become crucial means for enterprises to enhance their core competitiveness. By collecting, analyzing, and tracing data across the entire process—from raw material procurement, production and processing, logistics and warehousing, finished product inspection to after-sales service—enterprises can promptly identify the root causes of quality problems and improve process optimization and management efficiency. Real-time monitoring and predictive maintenance of equipment operating status can effectively prevent unplanned downtime and ensure production continuity. Therefore, building a comprehensive intelligent manufacturing platform that simultaneously supports quality traceability and predictive maintenance is of great significance.

[0003] Currently, while some quality traceability systems and equipment maintenance solutions exist in the market, most have significant shortcomings. In terms of quality traceability, many systems limit data collection to the manufacturing process, failing to cover key nodes such as raw materials, logistics and warehousing, and finished product flow, resulting in incomplete traceability chains. Different enterprises or different stages often use inconsistent coding methods and lack standardized identification systems, making cross-enterprise and cross-stage data sharing difficult. Furthermore, data in most systems remains siloed, making it difficult to establish a complete lifecycle from raw materials to finished products. Regarding predictive maintenance, existing technologies often rely on single types of data (such as vibration or temperature data) for fault prediction, ignoring the combined influence of multi-source information such as equipment operation, historical maintenance, process parameters, and environmental conditions, leading to insufficient prediction accuracy. In addition, many systems only issue alarms when equipment exhibits obvious anomalies, making it difficult to identify potential faults in a timely manner; their limited data processing capabilities also prevent the extraction of deeper equipment status characteristics from massive amounts of data. These shortcomings of existing technologies result in low efficiency, poor accuracy, and delayed response in both quality management and equipment maintenance.

[0004] Therefore, in the context of intelligent manufacturing, incomplete quality traceability chains, insufficient data standardization, and limited ability to utilize multi-source equipment status information have become urgent problems that need to be solved. Summary of the Invention

[0005] This application provides a quality traceability and predictive maintenance analysis platform for intelligent manufacturing, aiming to solve the problems of incomplete quality traceability chains, insufficient data standardization, and limited utilization of multi-source equipment status information in existing technologies in intelligent manufacturing scenarios.

[0006] A quality traceability and predictive maintenance analysis platform for intelligent manufacturing, the platform comprising: The quality traceability module is used to collect and manage product quality information throughout its entire lifecycle and to perform product quality traceability. The predictive maintenance module is used to process multi-source data from equipment and build equipment health models for equipment failure prediction. The data analysis and decision support module is used to perform statistical analysis and correlation analysis on quality data and equipment data and generate decision information. The data storage module is used to store data from the quality traceability module, the predictive maintenance module, and the data analysis and decision support module. The quality traceability module identifies and resolves products and their associated objects based on a unified identification system; the predictive maintenance module evaluates equipment status based on a time-series neural network model; and the data analysis and decision support module generates decision information for production management based on the correlation between quality data and equipment operation data.

[0007] Optionally, in the above solution, the quality traceability module includes a data acquisition unit, which is used to collect data at each stage of the product's entire lifecycle. Specifically, the data at each stage of the product's entire lifecycle includes: Specifications, batch numbers, supplier information, and quality inspection reports of raw materials; Equipment operating parameters, production environment parameters, operator information, process data, and quality inspection results during the production process; Storage location, inbound and outbound times, and transportation information during the logistics warehousing process; Inspection data of finished products, sales area information, and customer feedback information.

[0008] Optionally, in the above scheme, the quality traceability module includes a coding and identification unit, which uses a coding method based on unified coding rules to generate unique identification codes for products, raw materials and components. The unified identifier system includes a multi-level identifier resolution structure, which includes a root node resolution server, industry node resolution servers, and enterprise node resolution servers. The resolution servers at each level are connected via a network. The root node resolution server is used to route the resolution request to the corresponding industry node resolution server based on the identifier code prefix information. The industry node resolution server is used to forward the resolution request to the corresponding enterprise node resolution server. The enterprise node resolution server is used to parse the corresponding identifier code and access the stored lifecycle data.

[0009] Optionally, in the above scheme, the quality traceability module includes a traceability execution unit, which is used to initiate a data query request to the multi-level identifier resolution structure when receiving the identifier code input by the user, and generate a product quality traceability report based on the parsed raw material data, production process data, logistics and warehousing data and finished product data.

[0010] Optionally, in the above scheme, the predictive maintenance module includes a data fusion and processing unit, which is used to collect, clean, and preprocess multi-source data from the equipment. The multi-source data includes: Real-time data on equipment operation, including vibration acceleration, temperature, rotational speed, and energy consumption; Historical maintenance data, including fault type, repair time, and information on replaced parts; Production process data, including processing parameters and production cycle time; Environmental data, including workshop temperature, humidity, and dust concentration; The data fusion and processing unit is also used to perform feature extraction on the preprocessed data, and the feature extraction adopts wavelet transform and / or principal component analysis algorithms.

[0011] Optionally, in the above scheme, the predictive maintenance module includes a model building and training unit. The model building and training unit is used to build a device health model based on a long short-term memory neural network (LSTM), and to build training samples using historical operating data and fault data of the device. The training samples are divided into a training set and a validation set, and the LSTM model is trained based on the backpropagation algorithm to determine the model parameters.

[0012] Optionally, in the above scheme, the predictive maintenance module includes a fault prediction and early warning unit. The fault prediction and early warning unit is used to input real-time processed feature data into a trained LSTM model to obtain a device health status value, and compare the health status value with multiple preset health threshold intervals to generate early warning information of corresponding levels. The early warning levels include minor abnormality early warning level, potential fault early warning level, and fault early warning level. The fault prediction and early warning unit is also used to generate maintenance suggestions based on the early warning level and historical maintenance records.

[0013] Optionally, in the above scheme, the data analysis and decision support module includes a statistical analysis unit, which is used to perform statistical analysis on various types of data from the quality traceability module and the predictive maintenance module. The statistical analysis includes generating quality control charts, wherein the quality control charts include XR charts and P charts, and performing statistical analysis on product defect types, defect distribution, equipment failure types, and equipment failure frequency.

[0014] Optionally, in the above scheme, the data analysis and decision support module includes an association analysis unit. The association analysis unit uses an association rule mining algorithm to obtain the association relationship between quality data and equipment operation data. The association rule mining algorithm is the Apriori algorithm. The association analysis unit is used to determine equipment parameters and environmental parameters that may affect product quality based on the association relationship.

[0015] In the above scheme, optionally, the data storage module is a distributed database system, the distributed database system is an HBase database, the data storage module includes a data backup mechanism, the data backup mechanism includes a periodic full backup strategy, and the backup data is stored in an independent storage server.

[0016] Compared with the prior art, this application has at least the following beneficial effects: Based on further analysis and research of existing technical problems, this application recognizes the issues of incomplete quality traceability chains, insufficient data standardization, and limited utilization of multi-source equipment status information in intelligent manufacturing scenarios. By constructing a comprehensive platform including a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module, it achieves unified collection, management, and analysis of product lifecycle data and multi-source equipment operation data. The platform employs a unified identification system to encode and parse products and their associated objects, enabling accurate data linking across all stages, including raw material procurement, production and processing, logistics and warehousing, and finished product inspection. This overcomes the data inconsistencies caused by incomplete traceability chains and inconsistent coding in existing technologies. By introducing an equipment status assessment method based on a time-series neural network model, the platform can model and judge equipment operating trends using real-time equipment data, process data, and environmental data. This allows for early identification of potential faults, addressing the shortcomings of existing technologies that rely on single data points and struggle to make forward-looking predictions. The platform's data analysis and decision support module performs statistical analysis and correlation mining on quality and equipment data, enabling enterprises to identify key factors affecting quality and equipment status, avoiding the analytical limitations caused by isolated data. The data storage module uses a distributed storage architecture to ensure long-term preservation and high-concurrency access capabilities for massive amounts of data throughout the entire lifecycle.

[0017] Therefore, it can be seen that the present invention can simultaneously solve the core problems in the background technology, such as incomplete traceability, inconsistent data, inability to use multi-source data for equipment prediction, lack of systematic analysis capabilities, and ineffective data management. Attached Figure Description

[0018] Figure 1 A block diagram of the module architecture of a quality traceability and predictive maintenance analysis platform for intelligent manufacturing provided in one embodiment of this application; Figure 2 A schematic diagram illustrating the workflow of a quality traceability module provided in one embodiment of this application; Figure 3 This is a schematic diagram illustrating the workflow of a predictive maintenance module provided in one embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] In one embodiment, such as Figure 1 As shown, a quality traceability and predictive maintenance analysis platform for intelligent manufacturing is provided, the platform comprising: The quality traceability module is used to collect and manage product quality information throughout its entire lifecycle and to perform product quality traceability. The predictive maintenance module is used to process multi-source data from equipment and build equipment health models for equipment failure prediction. The data analysis and decision support module is used to perform statistical analysis and correlation analysis on quality data and equipment data and generate decision information. The data storage module is used to store data from the quality traceability module, the predictive maintenance module, and the data analysis and decision support module. The quality traceability module identifies and resolves products and their associated objects based on a unified identification system; the predictive maintenance module evaluates equipment status based on a time-series neural network model; and the data analysis and decision support module generates decision information for production management based on the correlation between quality data and equipment operation data.

[0021] In this embodiment, the quality traceability module includes a data acquisition unit, which is used to collect data at each stage of the product's entire lifecycle. Specifically, the data at each stage of the product's entire lifecycle includes: Specifications, batch numbers, supplier information, and quality inspection reports of raw materials; Equipment operating parameters, production environment parameters, operator information, process data, and quality inspection results during the production process; Storage location, inbound and outbound times, and transportation information during the logistics warehousing process; Inspection data of finished products, sales area information, and customer feedback information.

[0022] In this embodiment, the quality traceability module includes an encoding and identification unit, which uses an encoding method based on unified encoding rules to generate unique identification codes for products, raw materials and components. The unified identifier system includes a multi-level identifier resolution structure, which includes a root node resolution server, industry node resolution servers, and enterprise node resolution servers. The resolution servers at each level are connected via a network. The root node resolution server is used to route the resolution request to the corresponding industry node resolution server based on the identifier code prefix information. The industry node resolution server is used to forward the resolution request to the corresponding enterprise node resolution server. The enterprise node resolution server is used to parse the corresponding identifier code and access the stored lifecycle data.

[0023] In this embodiment, the quality traceability module includes a traceability execution unit, which is used to initiate a data query request to the multi-level identifier resolution structure when it receives the identifier code input by the user, and generate a product quality traceability report based on the parsed raw material data, production process data, logistics and warehousing data and finished product data.

[0024] In this embodiment, the predictive maintenance module includes a data fusion and processing unit, which is used to collect, clean, and preprocess multi-source data from the equipment. The multi-source data includes: Real-time data on equipment operation, including vibration acceleration, temperature, rotational speed, and energy consumption; Historical maintenance data, including fault type, repair time, and information on replaced parts; Production process data, including processing parameters and production cycle time; Environmental data, including workshop temperature, humidity, and dust concentration; The data fusion and processing unit is also used to perform feature extraction on the preprocessed data, and the feature extraction adopts wavelet transform and / or principal component analysis algorithms.

[0025] In this embodiment, the predictive maintenance module includes a model building and training unit. The model building and training unit is used to build a device health model based on a long short-term memory neural network (LSTM), and to build training samples using historical device operation data and fault data. The training samples are divided into a training set and a validation set, and the LSTM model is trained based on the backpropagation algorithm to determine the model parameters.

[0026] In this embodiment, the predictive maintenance module includes a fault prediction and early warning unit. The fault prediction and early warning unit is used to input real-time processed feature data into a trained LSTM model to obtain a device health status value, and compare the health status value with multiple preset health threshold intervals to generate early warning information of corresponding levels. The early warning levels include a minor anomaly early warning level, a potential fault early warning level, and a fault early warning level. The fault prediction and early warning unit is also used to generate maintenance suggestions based on the early warning levels and historical maintenance records.

[0027] In this embodiment, the data analysis and decision support module includes a statistical analysis unit, which is used to perform statistical analysis on various types of data from the quality traceability module and the predictive maintenance module. The statistical analysis includes generating quality control charts, which include XR charts and P charts, as well as performing statistical analysis on product defect types, defect distribution, equipment failure types, and equipment failure frequency.

[0028] In this embodiment, the data analysis and decision support module includes an association analysis unit. The association analysis unit uses an association rule mining algorithm to obtain the association relationship between quality data and equipment operation data. The association rule mining algorithm is the Apriori algorithm. The association analysis unit is used to determine equipment parameters and environmental parameters that may affect product quality based on the association relationship.

[0029] In this embodiment, the data storage module is a distributed database system, the distributed database system is an HBase database, the data storage module includes a data backup mechanism, the data backup mechanism includes a periodic full backup strategy, and the backup data is stored in an independent storage server.

[0030] In one specific implementation, the quality traceability and predictive maintenance analysis platform for intelligent manufacturing includes a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module. These modules interact with each other via a unified system bus or enterprise intranet. The platform is deployed on an enterprise private server or industrial cloud environment and interconnects with field equipment control systems, enterprise ERP systems, MES systems, warehouse management systems, and other information systems through API interfaces.

[0031] In the quality traceability module, the platform first establishes a unified identification system for uniquely identifying products, raw materials, and components. This system generates identification codes based on unified coding rules. These codes can be in the form of barcodes, QR codes, or RFID electronic tags, and can be scanned and read at various stages within the enterprise. To enable the platform to obtain traceability data across enterprises and industries, the system further deploys a multi-level identification resolution structure, including a root node resolution server, industry node resolution servers, and enterprise node resolution servers. The enterprise node resolution server stores the data records corresponding to the identification codes within the enterprise; the industry node resolution servers store the index information of enterprise resolution nodes within the industry; and the root node resolution server stores the routing table of industry nodes. After a user enters the product identification code on the platform interface, the traceability request is routed step-by-step through the root node resolution server, industry node resolution server, and enterprise node resolution server, ultimately accessing the corresponding enterprise database to obtain product lifecycle data.

[0032] In the predictive maintenance module, the platform sets up a unified data acquisition interface to access data from various types of equipment operation monitoring sensors, environmental sensors, and the enterprise maintenance management system. This module employs a data fusion strategy, centralizing equipment operation data, historical maintenance data, process data, and environmental data into a single data structure. After cleaning and preprocessing, the data is input into a time-series neural network model. In this embodiment, the model uses a Long Short-Term Memory (LSTM) neural network, which can perform state analysis based on the temporal characteristics of equipment operation data. The equipment health value is output by the trained model, representing the equipment's operating status. The system sets different threshold ranges for the health value; when the health value falls within a different range, the platform generates different levels of warnings.

[0033] The data analysis and decision support module periodically reads various types of data generated by the quality traceability and predictive maintenance modules through scheduled tasks. This module includes statistical analysis and correlation analysis functions. The statistical analysis function performs statistical calculations on quality inspection data, equipment operation data, and defect data, generating quality control charts, fault statistics tables, and process fluctuation statistics. The correlation analysis function uses association rule mining algorithms (such as the Apriori algorithm) to analyze the potential relationships between quality data and equipment operating parameters, providing a basis for adjustment suggestions output by the subsequent decision support module. This analysis module does not modify the original data; it only provides the analysis results as supplementary information for the enterprise to view.

[0034] The data storage module employs a distributed database system to simultaneously store multi-source data, traceability data, model training data, and analysis results. The database can be deployed in a cluster of multiple servers to support large-scale data storage needs. The module also includes a periodic backup program that replicates the database contents to a separate backup server at set intervals to prevent data loss due to system failures or environmental changes.

[0035] Through the aforementioned structure and implementation, this platform can fulfill multiple requirements that are unmet by existing technologies. By establishing a unified identification system and a multi-level identification resolution structure, the platform enables continuous tracking of products across multiple stages, including raw material procurement, production and processing, logistics and warehousing, and finished product inspection, achieving data consistency and traceability across systems and enterprises. The problems of "incomplete traceability chains" and "difficulty in data sharing due to inconsistent coding" mentioned in the background technology are solved through a unified coding and identification resolution mechanism. Raw material information, process data, quality inspection data, and warehousing data generated at different stages can all be associated based on identification codes, ensuring the continuity and accessibility of the traceability path.

[0036] The platform achieves in-depth analysis of multi-source equipment status data by introducing a time-series neural network model-based equipment health assessment method into the predictive maintenance module. The problem mentioned in the background technology of "relying on a single type of data and failing to reflect the actual operating conditions of the equipment" is improved in this solution because the platform not only integrates real-time data such as vibration, temperature, and energy consumption, but also utilizes historical maintenance records, process data, and environmental data as model inputs, making the analysis of equipment health status more comprehensive. The LSTM model has excellent modeling capabilities for time-series characteristics, thus identifying potential risks from subtle changes in equipment operating status, solving the problem of "lack of early prediction capability" in the background technology.

[0037] This platform, through statistical and correlation analysis functions, enables enterprises to extract structured information and relationships from large amounts of quality and equipment operation data. The problem of "data silos preventing system analysis" mentioned in the background technology is solved by the platform's data analysis and decision support module. This module can reveal the relationship between changes in equipment parameters, fluctuations in production processes, changes in environmental conditions, and product defects, providing enterprises with a basis for adjustment.

[0038] Finally, this platform stores all process data through a distributed data storage system, enabling it to support the massive amounts of information generated in intelligent manufacturing scenarios. The problem of "large data volume and limited storage capacity" mentioned in the background technology is solved in this solution. The system supports continuous writing of real-time device data, traceability data, and model training data, ensuring the long-term operational stability of the entire platform from the bottom layer.

[0039] In summary, this embodiment, through the combination of a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module, collaboratively solves the problems in the prior art regarding incomplete traceability chains, inconsistent coding, insufficient utilization of multi-source device data, lack of predictive capabilities, and insufficient data storage, enabling the entire platform to have complete, continuous, and scalable application capabilities.

[0040] In one embodiment, a quality traceability and predictive maintenance analysis platform for intelligent manufacturing is provided, including a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module. The quality traceability module is used to realize the quality traceability of products throughout their entire life cycle, from raw materials to finished products, and includes a data acquisition unit, a coding and identification unit, and a traceability execution unit. The predictive maintenance module is used to perform predictive maintenance analysis on production equipment, including a data fusion and processing unit, a model building and training unit, and a fault prediction and early warning unit. The data analysis and decision support module is used to analyze quality data and equipment operation data and generate decision suggestions, including a statistical analysis unit, a correlation analysis unit, and a decision support unit; The data storage module is used to store various types of data collected by the platform and to establish a data backup mechanism.

[0041] In this embodiment, the data acquisition unit of the quality traceability module collects raw material information, production process data, logistics and warehousing data, and finished product data through sensors, IoT devices deployed in the raw material procurement process, production workshop, logistics and warehousing center, and finished product inspection station, as well as data interfaces with the enterprise's ERP and MES systems. The raw material information includes the specifications, batch number, supplier information, and quality inspection report of the raw materials. The production process data includes equipment operating parameters, production environment temperature and humidity, operator information, processing data of each process, and quality inspection results. The logistics and warehousing data includes storage location, inbound and outbound time, and transportation information. The finished product data includes inspection data, sales area, and customer feedback.

[0042] In this embodiment, the coding and identification unit of the quality traceability module adopts a unified coding method based on the GS1 standard to generate unique identification codes for products, raw materials, and components, and establishes a tree-like identification resolution system. The identification resolution system includes a root node resolution server, an industry node resolution server, and an enterprise node resolution server. Nodes at all levels achieve data interaction and collaboration through the network. The root node resolution server is responsible for top-level routing, the industry node resolution server is responsible for resolution services within its industry, and the enterprise node resolution server is responsible for resolving specific identifiers within the enterprise.

[0043] In this embodiment, after receiving the product identification code input by the user, the traceability execution unit of the quality traceability module initiates a data query request to the identification resolution system. The root node resolution server routes the request to the corresponding industry node resolution server according to the identification code prefix information. The industry node resolution server forwards the request to the corresponding enterprise node resolution server. The enterprise node resolution server queries the local database and returns the obtained product lifecycle data level by level, finally forming a product traceability report.

[0044] In this embodiment, the data fusion and processing unit of the predictive maintenance module integrates real-time equipment operation data, historical maintenance records, production process data, and environmental data, and performs cleaning, preprocessing, and feature extraction on the multi-source data. The real-time equipment operation data includes vibration acceleration, temperature, rotational speed, and energy consumption. The historical maintenance records include fault type, repair time, and information on replaced parts. The production process data includes processing parameters and production cycle time. The environmental data includes workshop temperature and humidity and dust concentration. The feature extraction uses wavelet transform or principal component analysis algorithms.

[0045] In this embodiment, the model building and training unit of the predictive maintenance module uses the LSTM neural network algorithm to build an equipment health model. The model input is processed multi-source data features, and the output is the health status value of the equipment. Historical operating data and fault data of the equipment are selected as training samples, and the samples are divided into training set and validation set. The model parameters are adjusted by backpropagation algorithm, and the model is evaluated by validation set. Training stops when the prediction error reaches a preset threshold.

[0046] In this embodiment, the fault prediction and early warning unit of the predictive maintenance module inputs the real-time processed feature data into the trained equipment health model. Based on the comparison between the health status value output by the model and the preset health threshold range, it performs graded early warning according to the severity of the fault and provides maintenance suggestions based on the fault type and historical maintenance data. The graded early warning includes a level 1 warning, a level 2 warning, and a level 3 warning. A level 1 warning indicates a minor abnormality and prompts for key monitoring; a level 2 warning indicates a potential fault risk and suggests scheduling planned maintenance; a level 3 warning indicates that a fault is about to occur or has already occurred, and immediately issues an alarm and prompts for emergency handling.

[0047] In this embodiment, the statistical analysis unit of the data analysis and decision support module performs routine statistical analysis on the data collected by the quality traceability module and the predictive maintenance module, and generates quality control charts, statistical product defect types and distribution, and statistical equipment failure related information; the quality control charts include XR charts and P charts.

[0048] In this embodiment, the association analysis unit of the data analysis and decision support module uses an association rule mining algorithm to mine potential associations between quality data and equipment operation data; the association rule mining algorithm is the Apriori algorithm.

[0049] In this embodiment, the data storage module uses a distributed database to store data, and the distributed database is HBase; the data backup mechanism is a periodic full backup, and the backup data is stored in an independent storage server.

[0050] The purpose of this embodiment is to overcome the problems in existing technologies, such as incomplete quality traceability chains, inconsistent coding, data disconnect, and limited, inaccurate, and untimely predictive maintenance data. It provides a quality traceability and predictive maintenance analysis platform for intelligent manufacturing. This platform enables quality traceability throughout the entire product lifecycle, from raw materials to finished products. Simultaneously, it integrates multi-source data to perform accurate predictive maintenance analysis of equipment, improving product quality control and equipment reliability, and reducing enterprise production costs.

[0051] To achieve the above objectives, this embodiment provides a quality traceability and predictive maintenance analysis platform for intelligent manufacturing, including a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module.

[0052] Quality traceability module: Data Acquisition Unit: This unit comprehensively collects data through sensors and IoT devices deployed in raw material procurement, production workshops, logistics and warehousing centers, and finished product inspection stations, as well as data interfaces with enterprise ERP and MES systems. Specifically, this includes: raw material specifications, batch numbers, supplier information, and quality inspection reports; equipment operating parameters, environmental temperature and humidity, operator information, processing data for each step, and quality inspection results during production; storage locations, inbound and outbound times, and transportation information during logistics and warehousing; and finished product inspection data, sales regions, and customer feedback.

[0053] Encoding and Identification Units: A unified encoding method based on the GS1 standard is adopted to generate unique identification codes for products, raw materials, components, etc., at each stage of their entire lifecycle. A tree-like identifier resolution system is established, including a root node resolution server, industry node resolution servers, and enterprise node resolution servers. The root node resolution server is responsible for top-level routing, the industry node resolution servers are responsible for resolution services within their respective industries, and the enterprise node resolution servers are responsible for resolving specific identifiers within their respective enterprises. Nodes at each level achieve data interaction and collaboration through the network.

[0054] Traceability Execution Unit: When quality traceability is required, the user enters the product's identification code, and the traceability execution unit initiates a data query request to the identification resolution system. The root node resolution server routes the request to the corresponding industry node resolution server based on the prefix information of the identification code. The industry node resolution server further forwards the request to the corresponding enterprise node resolution server. The enterprise node resolution server queries its local database to obtain the product's full lifecycle data corresponding to the identification code and returns the data to the user level by level according to the traceability process, ultimately generating a detailed product traceability report containing raw material information, production process data, logistics information, inspection data, etc.

[0055] Predictive maintenance module: The data fusion and processing unit integrates real-time equipment operation data (such as vibration acceleration, temperature, rotational speed, and energy consumption), historical maintenance records (including fault types, repair times, and replaced parts), production process data (such as processing parameters and production cycle time), and environmental data (such as workshop temperature and humidity, and dust concentration). It cleans the collected multi-source data, removing noise and filling in missing values; then it performs preprocessing, including data standardization and normalization; finally, it extracts key features from the data using feature extraction algorithms (such as wavelet transform and principal component analysis), providing high-quality data for the construction of equipment health models.

[0056] Model Construction and Training Unit: An LSTM neural network algorithm is used to construct the equipment health model. The model's input is processed multi-source data features, and the output is the equipment's health status value. Historical operating data and fault data of the equipment are selected as training samples, which are divided into training and validation sets. During training, appropriate parameters such as the learning rate and number of iterations are set, and the model parameters are continuously adjusted through the backpropagation algorithm. The model is evaluated using the validation set. When the model's prediction error reaches a preset threshold, training stops, resulting in the optimized equipment health model.

[0057] Fault Prediction and Early Warning Unit: This unit inputs real-time collected and processed feature data from the equipment into a trained equipment health model. The model outputs the equipment's health status value. By comparing this health status value with a preset health threshold range, it determines whether the equipment has potential faults. When the health status value exceeds the normal range, a graded early warning is issued according to the severity of the fault: Level 1 warning indicates a minor anomaly, prompting focused monitoring; Level 2 warning indicates a potential fault risk, recommending planned maintenance; Level 3 warning indicates an impending or already occurring fault, immediately issuing an alarm and prompting emergency handling. Simultaneously, based on the fault type and historical maintenance data, corresponding maintenance suggestions are provided, such as replacement parts and repair procedures.

[0058] Data Analysis and Decision Support Module: Statistical Analysis Unit: Performs routine statistical analysis on data collected by the quality traceability module and predictive maintenance module. Generates quality control charts, such as XR charts and P charts, to monitor the stability of the production process and issue timely alerts when data points exceed control limits; statistically analyzes the types and distribution of product defects, calculates the incidence rate of different defects to provide direction for quality improvement; and statistically analyzes the number of equipment failures, failure types, and mean time between failures to assess equipment reliability.

[0059] Association Analysis Unit: This unit uses association rule mining algorithms (such as the Apriori algorithm) to uncover potential correlations between quality data and equipment operation data. For example, it analyzes the correlation between fluctuations in a certain equipment operating parameter and the occurrence of a certain product defect to identify key equipment parameters; it also studies the relationship between changes in production environment temperature and humidity and equipment failure rate to provide a basis for optimizing production environment and equipment maintenance strategies.

[0060] Decision Support Unit: Generates decision recommendations based on the results of statistical and correlation analyses. For example, if a batch of raw materials repeatedly causes product quality issues, it recommends changing the supplier or strengthening quality control of that supplier's raw materials; if a certain equipment parameter is highly correlated with product quality defects, it recommends adjusting that equipment parameter to its optimal range; and based on the predictive maintenance warning results of the equipment, it develops a reasonable equipment maintenance plan to avoid unplanned downtime.

[0061] Data storage module: A distributed database (such as HBase) is used to store the massive amounts of data collected by the quality traceability module and the predictive maintenance module, ensuring high data reliability and scalability. Simultaneously, a data backup mechanism is established to regularly back up the data and prevent data loss.

[0062] Based on the aforementioned platform, this embodiment provides a quality traceability and predictive maintenance analysis method for intelligent manufacturing, such as... Figure 2 and Figure 3 As shown, it includes the following steps: Step 1: The data acquisition unit of the quality traceability module comprehensively collects data from all stages of the product's entire lifecycle, while the coding and identification unit generates unique identification codes for relevant objects and establishes an identification resolution system.

[0063] Step 2: When quality traceability is required, the identification code is entered through the traceability execution unit, and the product's entire life cycle data is traced back using the identification resolution system to generate a traceability report.

[0064] Step 3: The data fusion and processing unit of the predictive maintenance module integrates relevant data from multiple sources of equipment and performs cleaning, preprocessing and feature extraction.

[0065] Step 4: Model Building and Training. The unit uses the processed data to train an LSTM neural network device health model.

[0066] Step 5: The fault prediction and early warning unit inputs real-time data into the model, predicts the equipment status, provides graded early warnings, and gives maintenance suggestions.

[0067] Step Six: The data analysis and decision support module performs statistical and correlation analysis on quality and equipment data to generate decision recommendations.

[0068] Step 7: All data is stored in the data storage module to ensure data security and reliability.

[0069] This embodiment has the following beneficial effects: Comprehensive and efficient quality traceability: By comprehensively collecting product lifecycle data and employing a unified coding and tree-structured identifier resolution system, data standardization and unique identification are achieved, ensuring the integrity and accuracy of quality traceability. This enables rapid tracing of the root causes of product quality issues, improving the efficiency of quality problem location and resolution, and enhancing enterprises' control over product quality.

[0070] Predictive maintenance is accurate and reliable: By integrating and analyzing data from multiple equipment sources and employing LSTM neural network algorithms to construct equipment health models, the accuracy of equipment failure prediction is improved. It can provide early warnings of potential faults and offer targeted maintenance recommendations, reducing unplanned downtime, lowering equipment maintenance costs, extending equipment lifespan, and ensuring production continuity.

[0071] Data analytics supports scientific decision-making: By statistically and correlationally analyzing quality and equipment operation data, potential relationships between data points are uncovered, providing enterprises with scientific decision support. This helps enterprises optimize production processes, adjust equipment parameters, rationally arrange maintenance plans, improve product quality and production efficiency, reduce production costs, and enhance their intelligent manufacturing capabilities and market competitiveness.

[0072] Data storage is secure and reliable: A distributed database is used to store massive amounts of data, and a data backup mechanism is established to ensure the security, integrity, and scalability of the data, providing a strong guarantee for the stable operation of the platform.

[0073] In one embodiment, a quality traceability and predictive maintenance analysis platform for intelligent manufacturing includes a quality traceability module, a predictive maintenance module, a data analysis and decision support module, and a data storage module.

[0074] Quality traceability module: Data acquisition unit: Deploy RFID identification equipment in the raw material warehouse to collect information such as batch and supplier of raw materials; install sensors on key equipment in the production workshop to collect equipment operating parameters, such as the speed and temperature of lathes; set up barcode scanners in each process, and operators scan the codes to record processing information; obtain product transportation information through interface with the logistics management system.

[0075] Encoding and Identification Units: GS1-128 code is used as the identification code to generate a unique code for each product and raw material batch. In the identification resolution system, the root node resolution server uses a high-performance server configured with a 16-core CPU and 64GB of memory; the industry node resolution server and enterprise node resolution server are selected with appropriate configurations according to the enterprise size. Servers at all levels are connected through a fiber optic network to ensure data transmission speed.

[0076] Traceability Execution Unit: Developed using Java, users initiate traceability requests by entering an identification code through a web interface. The system automatically parses and queries the data, and the generated traceability report is displayed in web page format.

[0077] Predictive maintenance module: Data fusion and processing unit: Collects real-time equipment data via OPC UA protocol at a sampling frequency of 1 time / second; reads historical maintenance data from the equipment management system database; and collects temperature and humidity data through workshop environmental sensors. Python programming language is used for data processing, Pandas library for data cleaning and preprocessing, and Scikit-learn library for feature extraction.

[0078] Model Building and Training Unit: An LSTM neural network model was built based on the TensorFlow framework. The model contains 3 LSTM layers and 1 fully connected layer. The training data consisted of equipment operation data and failure data from the past 3 years, divided into training and validation sets in a 7:3 ratio. The learning rate was set to 0.001, and the number of iterations was 1000. Training was stopped when the mean squared error of the validation set was less than 0.01.

[0079] Fault prediction and early warning unit: Receives and processes equipment data in real time, inputs it into the model, and the model outputs a health status value. The preset health threshold range is 0-1, with 0.8-1 representing a normal state, 0.5-0.8 indicating a Level 1 warning, 0.3-0.5 indicating a Level 2 warning, and less than 0.3 indicating a Level 3 warning. Warning information is pushed to relevant personnel via platform interface pop-ups, SMS, and email.

[0080] Data Analysis and Decision Support Module: Statistical Analysis Unit: Uses the Matplotlib library to draw quality control charts, defect distribution histograms, etc.; uses SQL statements to perform statistical queries on data in the database and generate various statistical reports.

[0081] Association analysis unit: Using the Apriori algorithm, with a minimum support of 0.2 and a minimum confidence of 0.7, association rules between quality data and equipment operation data are mined.

[0082] Decision Support Unit: According to the analysis results, it automatically generates a decision recommendation report, including problem description, cause analysis, recommended measures, etc. Users can view and download it on the platform.

[0083] Data Storage Module: It uses the HBase distributed database, which is deployed on a cluster composed of 3 servers. Each server is configured with an 8-core CPU, 32G of memory, and a 2TB hard disk. It adopts a timed backup strategy to perform a full backup of the data at 2 am every day, and the backup data is stored in an independent storage server.

[0084] The application process of the platform in this embodiment is as follows: Taking the production of float glass by a certain glass manufacturing enterprise as an example, the application process of this platform is described.

[0085] Quality Traceability Process: When a large area of bubble defects appears during the installation of a certain batch of float glass by customers, the enterprise initiates a traceability through the platform. The user inputs the identification code of this batch of glass in the quality traceability module, and the traceability execution unit sends a request to the identification resolution system. The root node resolution server routes the request to the building materials industry node resolution server, and then forwards it to the glass enterprise node resolution server.

[0086] The enterprise node resolution server queries the database to obtain the full life cycle data of this batch of glass: Raw Material Information: The quartz sand supplier is Company A, with batch number Q20230512. The test report of its silica content shows compliance; the soda ash is purchased from Manufacturer B, batch C678, with a purity of 99.8%.

[0087] Production Process Data: The temperature curve of the melting furnace (1560°C ± 5°C, meeting the process standards), the flow rate of the protective gas in the tin bath (nitrogen purity 99.99%), the temperature distribution in each zone of the annealing furnace (the measured temperature in the 3rd zone is 8°C lower than the set value), the operator is Engineer Li (employee number 0125), and the quality inspection record of this process shows "≤ 2 bubbles / m² at the edge".

[0088] Logistics Information: It is transported by Logistics Company C, and the license plate number of the transport vehicle is Beijing A XXXXX, the temperature in the carriage is maintained at 25°C ± 3°C during transportation, and it is stored on the 3rd row of shelves in the 5th zone of the enterprise's finished product warehouse.

[0089] Through the traceability report, the enterprise discovers that the abnormal temperature in the 3rd zone of the annealing furnace is the key cause of the bubble defects. Immediately, it adjusts the temperature control system of the annealing furnace and conducts a secondary quality inspection on the remaining glass of this batch.

[0090] Predictive Maintenance Process: The platform real-time collects the operation data of key equipment on the float glass production line: Furnace burner: Gas pressure (5.2 kPa), flame temperature (1620℃), valve opening (75%) Tin bath drive roller: rotational speed (1.2m / min), vibration value (0.08mm / s), bearing temperature (42℃) Cooling fan: air pressure (300Pa), air volume (5000m³ / h), motor current (15A) It also integrates historical maintenance records (the furnace burner nozzles have been replaced 3 times in the past year), production process data (glass thickness 3.2mm, drawing speed 1.5m / min) and workshop environmental data (humidity 65%, dust concentration 0.5mg / m³).

[0091] The data fusion and processing unit cleaned (removing one abnormal wind pressure value caused by sensor failure), standardized the data, extracted 128-dimensional features through wavelet transform, and input them into the trained LSTM device health model. The model output a health status value of 0.38 for the tin bath drive roller, which is lower than the level 2 warning threshold (0.5). The platform issued a level 2 warning, indicating that "there is a risk of wear on the drive roller bearing" and suggested "arranging a shutdown inspection within 48 hours, prioritizing the replacement of bearing model 32016".

[0092] Maintenance personnel conducted an inspection based on the recommendations and found minor scratches on the bearing balls. After timely replacement, they prevented unplanned downtime of the production line caused by roller jamming.

[0093] Data analysis and decision support process: The statistical analysis unit analyzes the glass quality data of the past 3 months and generates a P control chart showing that the bubble defect rate went out of control in the second week (exceeding the upper control line), and "point bubbles" accounted for 68% of the defect types.

[0094] The correlation analysis unit, using the Apriori algorithm (minimum support 0.2, minimum confidence 0.7), found that when the furnace gas pressure fluctuation is > ±0.3 kPa, the glass bubble defect rate increases by 3.2 times (confidence 0.85).

[0095] The decision support unit generates a recommendation report based on this information: The furnace gas pressure is stably controlled within the range of 5.0±0.1 kPa, and a pressure compensation device is installed. Perform ultrasonic cleaning on the burner nozzles weekly to extend their service life. Adjust the temperature setting of zone 3 of the annealing furnace, increasing it by 5°C to eliminate the risk of air bubbles. After the company adopted the recommendations, the incidence of bubble defects dropped from 3.5% to 0.8% for one month, and production efficiency increased by 12%.

[0096] As can be seen from the above embodiments, this embodiment can effectively realize product quality traceability and predictive maintenance analysis of equipment, provide scientific decision support for enterprises, and improve the enterprise's intelligent manufacturing level and market competitiveness.

Claims

1. A quality traceability and predictive maintenance analysis platform for smart manufacturing, characterized in that, The platform comprises: a quality traceability module for collecting and managing product life cycle quality information and performing product quality traceability; a predictive maintenance module for processing equipment multi-source data and constructing equipment health models for equipment failure prediction; a data analysis and decision support module for statistical analysis, correlation analysis of quality data and equipment data, and generation of decision information; a data storage module for storing data of the quality traceability module, the predictive maintenance module and the data analysis and decision support module; wherein the quality traceability module identifies products and their associated objects based on a unified identification system, the predictive maintenance module evaluates equipment status based on a time series neural network model, and the data analysis and decision support module generates decision information for production management based on the correlation between quality data and equipment operation data.

2. The platform of claim 1, wherein, The quality traceability module comprises a data collection unit for collecting data at each stage of the product life cycle, specifically including: specifications, batch numbers, supplier information and quality inspection reports of raw materials; equipment operation parameters, production environment parameters, operator information, process data and quality detection results during the production process; storage location, warehouse entry and exit time, and transportation information during the logistics and storage process; inspection data, sales area information and customer feedback information of finished products.

3. The platform of claim 1, wherein, The quality traceability module comprises a coding and identification unit that generates unique identification codes for products, raw materials and components using a coding method based on a unified coding rule; The unified identification system comprises a multi-level identification analysis structure, which includes a root node analysis server, an industry node analysis server and an enterprise node analysis server, and the servers at each level are connected through a network, wherein the root node analysis server is used to route the analysis request to the corresponding industry node analysis server according to the prefix information of the identification code, the industry node analysis server is used to forward the analysis request to the corresponding enterprise node analysis server, and the enterprise node analysis server is used to analyze the corresponding identification code and access the stored life cycle data.

4. The platform of claim 3, wherein, The quality traceability module comprises a traceability execution unit for initiating a data query request to the multi-level identification analysis structure when receiving a user-input identification code, and generating a product quality traceability report based on the raw material data, production process data, logistics and storage data, and finished product data obtained by analysis.

5. The platform of claim 1, wherein, The predictive maintenance module comprises a data fusion and processing unit for collecting, cleaning and preprocessing equipment multi-source data, including: real-time data of equipment operation, including vibration acceleration, temperature, speed and energy consumption; historical maintenance data, including fault type, repair time and replacement part information; production process data, including processing parameters and production rhythm; environmental data, including workshop temperature and humidity and dust concentration; The data fusion and processing unit is further configured to perform feature extraction on the preprocessed data, the feature extraction employing a wavelet transform and / or a principal component analysis algorithm.

6. The platform of claim 1, wherein, The predictive maintenance module comprises a model building and training unit configured to build a long short-term memory neural network (LSTM) based equipment health model, and to build training samples using historical operation data and failure data of the equipment, divide the training samples into a training set and a validation set, and train the LSTM model based on a back propagation algorithm to determine model parameters.

7. The platform of claim 1, wherein, The predictive maintenance module comprises a failure prediction and early warning unit configured to input real-time processed feature data into the trained LSTM model to obtain an equipment health state value, compare the health state value with a plurality of preset health threshold intervals to generate early warning information of corresponding levels, the early warning levels including a slight abnormality early warning level, a potential failure early warning level, and a failure early warning level, and generate maintenance suggestions based on the early warning levels and historical maintenance records.

8. The platform of claim 1, wherein, The data analysis and decision support module comprises a statistical analysis unit configured to perform statistical analysis on various types of data of the quality traceability module and the predictive maintenance module, the statistical analysis including generating a quality control chart, wherein the quality control chart comprises an X-R chart and a P chart, and performing statistical analysis on product defect types, defect distribution, equipment failure types, and equipment failure times.

9. The platform of claim 1, wherein, The data analysis and decision support module comprises an association analysis unit configured to obtain an association relationship between quality data and equipment operation data using an association rule mining algorithm, the association rule mining algorithm being an Apriori algorithm, and to determine equipment parameters and environmental parameters that may affect product quality based on the association relationship.

10. The platform of claim 1, wherein, The data storage module is a distributed database system, the distributed database system being an HBase database, the data storage module comprising a data backup mechanism, the data backup mechanism comprising a periodic full backup strategy, and backup data being stored in a separate storage server.

Images