A process data integration and quality traceability management method and system for a badminton racket intelligent manufacturing line

Abstract

This invention belongs to the field of intelligent manufacturing technology and discloses a method and system for process data integration and quality traceability management in intelligent manufacturing production lines for badminton rackets. It unifies data standards and interaction rules by constructing a five-in-one semantic ontology model encompassing products, processes, etc.; it builds an adaptive multimodal acquisition system, integrating three types of data sources: standardized equipment, non-standard processes, and offline testing, and completing edge preprocessing; it achieves chain-based data storage through a consortium blockchain hybrid storage mode, ensuring data immutability and shareability, and achieving efficient storage of all data; it assigns a unique, lifelong identity to each racket, and combined with a full-process digital twin mirror, it visualizes the entire process data from raw materials to the final product and simulates the production process; the consortium blockchain storage ensures the authenticity and immutability of traceability data, and when defects occur in the finished product, it can accurately locate the core influencing processes, parameter deviations, and other root causes through deep correlation graphs.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, specifically a method and system for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets. Background Technology

[0002] The manufacturing process of carbon fiber composite badminton rackets spans four major areas: composite material processing, precision machining, surface treatment, and performance testing, encompassing nine core processes. It is characterized by long process flows, strong correlation of process parameters, a high proportion of non-standard manual and semi-automated processes, and complex quality-influencing factors, placing extremely high demands on the integration of data across the entire production line and full lifecycle quality traceability. Currently, badminton racket manufacturers have promoted intelligent transformation of their production lines, introducing basic tools such as SCADA and MES to achieve data collection and batch-level traceability for some standardized processes. However, significant technical deficiencies and pain points still exist. Due to differences in equipment manufacturers, communication protocols, and data formats, as well as the existence of numerous non-standard processes, existing technologies cannot achieve full standardization and integration of various types of data, resulting in broken data links between processes and the formation of data silos.

[0003] Existing technologies can only achieve batch-level coarse traceability, and cannot achieve fine-grained traceability of a single racket throughout its entire life cycle; moreover, centralized storage makes data easy to tamper with or lose, resulting in low reliability of traceability results and making it impossible to accurately pinpoint the root cause of quality problems.

[0004] The existing traceability is retrospective, which cannot predict risks or intercept defective products in real time based on process data, resulting in defective products flowing into the next process and wasting raw materials and time.

[0005] Existing process data is only used for storage and basic traceability. No deep correlation model between process parameters and quality indicators has been built, making it impossible to quantify the weight of parameter influence and making it difficult to use traceability data to feed back into process optimization and improve the intelligence level of the production line. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for process data integration and quality traceability management in intelligent manufacturing production lines for badminton rackets, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for process data integration and quality traceability management in a smart manufacturing production line for badminton rackets, comprising the following specific steps: Preferably, the ontology modeling stage outlines the entire process of intelligent manufacturing of badminton rackets, and for the nine core processes, it clarifies the core process parameters, key quality characteristics, equipment attributes, personnel attributes, and material attributes for each process. Construct a semantic ontology model that integrates products, processes, equipment, parameters, and quality indicators, and unify the semantic standards, data formats, and data interaction rules of heterogeneous data across the entire production line; Each badminton racket is assigned a laser-engraved QR code or RFID electronic tag as a lifelong identification mark, which is then bound to a product instance in the physical model, serving as a unique traceability unit throughout its entire lifecycle.

[0008] Preferably, the data acquisition stage is based on the semantic ontology model and the unique traceability unit of a single racket constructed in the ontology modeling stage. For the three core data sources of the badminton racket production line, an adaptive multimodal acquisition system is constructed. Through industrial protocols such as OPCUA, Modbus, and Profinet, the system realizes real-time adaptive acquisition of process sequence parameters and equipment operating status of standardized industrial equipment such as cutting machines, hot presses, and CNC drilling machines. For non-standard manual and semi-automated processes, a machine vision and intelligent sensing fusion acquisition scheme is adopted. Machine vision is used to collect data on layup angle, layup number, and carbon fiber overlap compliance. An array of tension sensors is used to collect dynamic tension and tension uniformity data of the stringing process. Protocol conversion gateways are deployed at the edge of the production line to achieve unified adaptation of different industrial protocols. OPCUA protocol devices are directly connected to the gateway, while Modbus and Profinet protocol devices are connected through industrial switches. The gateway converts data from each protocol into JSON format for interaction. The data acquisition frequency of standardized equipment is set according to the characteristics of the process. Discrete processes such as cutting and drilling are acquired at a frequency of 1 time / second, while continuous processes such as hot pressing and curing are acquired at a frequency of 5 times / second. Equipment operating status data is acquired through anomaly triggering to ensure the real-time and efficient acquisition of data.

[0009] The standardized interface enables automatic synchronization of offline laboratory test data such as fatigue life and torsional bending performance tests, and supports manual structured data entry. All collected data undergoes time alignment, outlier filtering, data deduplication, missing value completion, and standardization at the edge of the production line. The pre-processed data is bound to the unique identifier of the corresponding racket.

[0010] Preferably, the trusted integration stage is based on the process data and racket unique identifier binding results obtained in the data acquisition stage to build a consortium blockchain network for badminton racket production lines. The consensus nodes include production line processing units, quality inspection units, raw material suppliers, end customers, and regulatory authorities. The badminton racket production line consortium blockchain is built on the Fabric underlying technology framework. It adopts a containerized deployment method to configure independent node services and ledger copies for each consensus node. The nodes interact with each other through an encrypted communication protocol. The consensus mechanism adopts the Practical Byzantine Fault Tolerance (PBFT) mechanism. The consensus nodes complete the block consensus in four stages: proposal, endorsement, sorting, and submission. The block generation interval is set to 5 seconds to ensure the immutability of data while taking into account the real-time nature of production line data on the chain.

[0011] For each badminton racket, after each process is completed and passes quality inspection, the core metadata corresponding to the process is generated into an independent block. The block contains the hash value of the previous process block, forming a chain-like evidence storage structure for the entire process of a single racket. It adopts a hybrid on-chain and off-chain storage mode. On-chain only stores hash values ​​and core metadata, while off-chain uses a distributed database to store all data such as time-series process data, visual acquisition data, and inspection reports. The on-chain hash value corresponds one-to-one with the full off-chain data.

[0012] Preferably, the correlation modeling stage targets six core quality indicators of the finished badminton racket: bending stiffness, torsional stiffness, balance point accuracy, bounce rate, fatigue life, and coating adhesion. Based on data collected throughout the entire process, a grey relational analysis algorithm is used to quantify the influence weight of key process parameters of each process on each quality indicator, and to screen out the core control parameters that affect quality. Furthermore, by combining graph neural network technology, a deep correlation graph is constructed for a single racket, process, main control parameters, quality indicators, and defect types, clarifying the causal relationship between each process parameter and quality indicator; at the same time, the correlation graph is deeply bound to the semantic ontology model and digital twin model.

[0013] Preferably, the quality traceability stage is based on the deep association graph constructed in the association modeling stage, the semantic ontology model, and the chain-like evidence data in the trusted integration stage to construct a digital twin mirror of the entire badminton racket production line. By inputting the unique identifier of the racket, the system can visualize the entire lifecycle data of the corresponding racket from raw material warehousing, process parameters of each process, operators, equipment status, quality inspection results, finished product testing, packaging and warehousing, and sales flow. At the same time, the production process of key processes can be simulated and replayed through the digital twin model. When a quality defect occurs in the finished product, the corresponding core influencing process, process parameter deviation, equipment abnormality, and raw material batch can be located in reverse through the correlation map based on the defect type, thereby locking the root cause of the quality problem and retrieving the on-chain evidence data of the corresponding process. When a certain piece of equipment or a certain batch of raw materials malfunctions, it is possible to quickly retrieve all badminton racket products corresponding to that equipment and batch of raw materials.

[0014] Preferably, the quality early warning stage is based on the quality correlation graph constructed in the correlation modeling stage, and sets dynamic thresholds for the core control parameters of each process. The dynamic thresholds are dynamically adjusted in real time based on the parameter data of the preceding process of the racket, the optimal process parameters of the same model of product, and the influence weights in the correlation graph. During the production process, the edge device compares the real-time collected process parameters with dynamic thresholds in real time, triggering a graded early warning and control mechanism. The first-level warning is for minor parameter deviations, and the system pushes the warning information to the on-site operators in real time, prompting them to adjust the parameters. At the same time, the warning data is stored on the blockchain for evidence. The second-level warning is for serious parameter deviations, and the system directly triggers the production line interception mechanism to prevent the racket from flowing into the next process. At the same time, the abnormal information is pushed to the process engineer for handling. Only when the process parameters meet the dynamic threshold requirements and the quality inspection is qualified can the racket be allowed to enter the next process.

[0015] The specific execution method of the production line interception mechanism is as follows: after the system triggers a level-two warning, it immediately sends a stop command to the unloading and loading conveying equipment of the process. At the same time, it triggers the opening of the gate of the abnormal material temporary storage bin next to the process, pushing the abnormal racket to the dedicated temporary storage bin. The temporary storage bin is bound to the unique identifier of the racket, realizing the separate management of the abnormal racket. The resumption process after parameter rectification is as follows: after the process engineer completes the parameter adjustment, he submits a resumption application in the system. The system drives the equipment to conduct three trial productions. After all the process parameters of the trial production products meet the dynamic threshold and pass the quality inspection, the interception is automatically lifted, and the production line resumes normal production. The trial production data must be stored separately on the blockchain.

[0016] Preferably, the process optimization stage is based on the hierarchical early warning and control data of the quality early warning stage, the root cause localization results of the quality traceability stage, and the correlation graph of the correlation modeling stage. It constructs a closed-loop optimization system of data collection, correlation analysis, traceability and localization, early warning and control, and process optimization. The full-process data, quality defect data, early warning and handling data, and finished product inspection data accumulated during the production line operation are continuously input into the graph neural network model of the correlation graph for iterative training of the model. The correlation between process parameters and quality indicators is continuously optimized, and the optimal combination of process parameters for different models of badminton rackets is output. The feature inputs of the graph neural network model are standardized process control parameter values, actual quality index values, defect type codes, and equipment operating status parameters. All input features are imported into the model in vector form, with defect types converted using one-hot encoding. The model output is a set of optimal control parameter values ​​for all processes of each badminton racket model, presented in key-value pairs. The key is "process number - control parameter name", and the value is the optimal parameter value. The recommended fluctuation range of each parameter is also output to provide a basis for dynamic threshold updates.

[0017] The optimized process parameters are synchronously updated to the semantic ontology model and the production line digital twin model. The process simulation is verified through the digital twin model. After the verification is passed, the parameters are distributed to each processing unit on the production line. Based on the continuously accumulated quality defect data, the dynamic threshold and warning rules of the quality early warning model are optimized.

[0018] This invention also provides a process data integration and quality traceability management system for intelligent manufacturing production lines of badminton rackets, based on the above method, including: The ontology modeling module sorts out the entire process of intelligent manufacturing of badminton rackets, clarifies the core parameters and attributes of each process, constructs a five-in-one semantic ontology model, unifies data standards and interaction rules, and assigns a unique laser-engraved QR code or RFID electronic tag to each racket as an identity identifier. The data acquisition module constructs an adaptive multimodal acquisition system to collect data from standardized equipment, non-standard manual and semi-automated processes, and offline laboratory testing. Data preprocessing is completed at the edge and the data is bound to the racket identifier. The trusted integration module builds a production line alliance chain network, generates core metadata blocks for each process of each racket to form a chain-like evidence storage structure, and adopts a hybrid on-chain and off-chain storage mode. The correlation modeling module uses relevant algorithms to quantify the influence weight of process parameters on quality indicators, constructs a deep correlation graph, and binds it to relevant models. The quality traceability module constructs a digital twin mirror of the entire production line process, enabling forward traceability of individual rackets, reverse root cause localization of defects, and batch-level risk interception. The quality early warning module sets dynamic thresholds based on the correlation graph, compares process parameters in real time, and triggers hierarchical early warning control to achieve in-process interception of quality defects. The process optimization module integrates data from various modules to build a closed-loop optimization system, iteratively trains the model to output optimal process parameters, updates relevant models and production line parameters, and optimizes early warning rules.

[0019] The beneficial effects of this invention are as follows: 1. This invention unifies the data standards and interaction rules of different equipment and non-standard processes by constructing a semantic ontology model integrating products, processes, etc.; it builds an adaptive multimodal acquisition system, integrates three types of data sources—standardized equipment, non-standard processes, and offline testing—and completes edge preprocessing; and it achieves chain-based data storage through a consortium blockchain hybrid storage mode, ensuring that the data is tamper-proof and shareable, while also achieving efficient storage of all data, laying a high-quality and standardized data foundation for digital management of production lines.

[0020] 2. This invention assigns a unique, lifelong identity to each racket. Combined with a full-process digital twin mirror, it can visualize the entire process data of the racket from raw materials to the finished product and simulate and replay the production process. The consortium blockchain ensures that the traceability data is authentic and tamper-proof. When defects occur in the finished product, the core influencing processes, parameter deviations, and other root causes can be accurately located through deep correlation graphs. When equipment or raw materials are abnormal, batch risk products can also be quickly retrieved and intercepted, improving the accuracy and efficiency of handling quality problems.

[0021] 3. This invention sets dynamic thresholds based on quality correlation graphs as core parameters. Through a graded early warning mechanism, minor deviations are promptly alerted to operators for adjustment, while severe deviations directly trigger production line interception, avoiding resource waste caused by defective products. A closed-loop system from data acquisition to process optimization is constructed. All data, including early warning and traceability data, are input into the model for iterative training. The optimal process parameters are output and verified by digital twin simulation before being distributed to the production line. At the same time, the early warning rules are continuously optimized, allowing process data to feed back into process upgrades and promoting the continuous iteration of intelligent and standardized production lines. Attached Figure Description

[0022] Figure 1 This is a flowchart of the process data integration and quality traceability management method for intelligent manufacturing production lines of badminton rackets, as described in this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1 As shown, this embodiment of the invention provides a method for process data integration and quality traceability management in a smart manufacturing production line for badminton rackets, including: In the ontology modeling stage, the entire process of intelligent manufacturing of badminton rackets is sorted out. For the nine core processes, the core process parameters, key quality characteristics (CTQ), equipment attributes, personnel attributes, and material attributes of each process are clearly defined. The nine core processes of intelligent manufacturing of badminton rackets, in the order of production, are: carbon fiber prepreg cutting, carbon cloth laying, hot pressing and curing, blank shaping, CNC drilling, handle assembly, stringing, surface coating, and finished product performance testing. Each process is a continuous production process, fully covering the entire manufacturing process of badminton rackets from carbon fiber raw material processing to finished product performance testing. These are the core technological processes of intelligent manufacturing of badminton rackets.

[0025] Construct a semantic ontology model that integrates products, processes, equipment, parameters, and quality indicators, and unify the semantic standards, data formats, and data interaction rules of heterogeneous data across the entire production line; Each badminton racket is assigned a unique, non-replicable laser-engraved QR code / RFID electronic tag as a lifelong identification mark. The tag is bound to the product instance in the racket model, serving as a unique traceability unit throughout the entire life cycle, enabling a single racket to be uniquely anchored throughout the entire process from raw materials to the end customer.

[0026] The laser-engraved QR code coding process occurs after the carbon fiber prepreg is cut and before the carbon cloth is laid. The RFID electronic tag uses embedded coding during the handle assembly process. Both types of tags include the racket's unique code, product model, and basic production batch information. The binding of the tags to the product instance of the main model is completed through the production line data management system, which maps the tag code to the product instance ID in the main model one-to-one and stores it on the blockchain. During the production process, the tags are identified in real time using industrial barcode scanners and RFID readers. The identification equipment is deployed at the loading and unloading ends of each process, and the identification data is synchronized to the edge in real time and bound to the process data.

[0027] The data acquisition phase is based on the semantic ontology model and the unique traceability unit of a single racket constructed in the ontology modeling phase. For the three core data sources of the badminton racket production line, an adaptive multimodal acquisition system is constructed. The system uses mainstream industrial protocols such as OPCUA, Modbus, and Profinet to achieve real-time adaptive acquisition of process sequence parameters and equipment operating status of standardized industrial equipment such as cutting machines, hot presses, and CNC drilling machines. For non-standard manual and semi-automated processes such as manual layup and manual stringing, a machine vision and high-precision intelligent sensing fusion acquisition scheme is adopted. Machine vision is used to collect data on layup angle, layup number, and carbon fiber overlap compliance. Array-type tension sensors are used to collect dynamic tension and tension uniformity data of the stringing process to achieve standardized acquisition of non-standard process data. The machine vision acquisition system is equipped with a 20-megapixel industrial camera and a telecentric lens. The visual recognition algorithm uses the YOLOv8 algorithm and a dedicated visual recognition model is trained for non-standard processes such as layup and threading. This enables real-time visual detection of layup angle, carbon fiber overlap compliance, and threading tension uniformity, with a detection frame rate of 30 frames per second. The intelligent sensing system uses an array of tension sensors and angle sensors. The sensor acquisition frequency is 10Hz, and the sensor data is transmitted to the edge of the production line via a 485 serial port with encryption. The trigger threshold of the sensors is pre-set according to the standard values ​​of the core parameters of the corresponding process.

[0028] The standardized interface enables automatic synchronization of offline laboratory test data such as fatigue life and torsional bending performance tests, and supports manual structured data entry. All collected data undergoes time alignment, outlier filtering, data deduplication, missing value completion, and standardization at the production line edge. The pre-processed data is bound to the unique identifier of the corresponding racket.

[0029] The synchronization of offline laboratory test data adopts a standardized TCP / IP industrial Ethernet interface to achieve direct data transmission between laboratory testing equipment and the edge of the production line. The transmission protocol is MQTT to ensure the real-time performance and stability of data transmission. Manual structured data entry must be completed on the fixed input interface of the production line data management system. The fields to be entered include the unique identifier of the racket, the test item, the test value, the test equipment, the test time, and the tester. The format of the entered data must be consistent with the automatically synchronized data. The entered data must be reviewed and confirmed by the quality inspection supervisor before it can be bound.

[0030] Time alignment uses the unified timestamp of the production line as a benchmark to calibrate the asynchronous data of each acquisition device and sensing module to the same time dimension; outlier filtering uses the 3σ principle to identify and remove abnormal data that deviates from the normal range; data deduplication is achieved by combining the unique racket identifier, process number and acquisition timestamp; missing value completion uses the mean of adjacent time points to complete continuous process parameters, and the mode of normal data in the same batch and process to complete discrete quality inspection data; data standardization uses normalization processing to map all parameter values ​​to the 0-1 range to ensure data dimension uniformity.

[0031] The trusted integration phase is based on the standardized, high-quality process data and racket unique identifier binding results obtained in the data collection phase. It builds a consortium blockchain network for badminton racket production lines. The consensus nodes include production line processing units, quality inspection units, raw material suppliers, end customers, and regulatory authorities, so as to achieve controllable and trusted data sharing. Hierarchical permissions are configured for each consensus node in the consortium blockchain. The production line processing unit has the right to write data for its own process and the right to read data for the entire process. The quality inspection unit has the right to write, modify, and query all quality inspection data for all processes, as well as the right to read process parameters. Raw material suppliers only have the right to write data for their own batches of raw materials and the right to read data for the corresponding racket processes. End customers only have the right to read data for the entire lifecycle of the rackets they purchased. Regulatory authorities have the right to query and audit all data, but no right to write or modify data. Data sharing between nodes is limited in scope based on the unique identifier of the racket. Only racket data related to their own business can be accessed. Access requires encrypted verification of the node's identity before the corresponding data can be obtained.

[0032] For each badminton racket, after each process is completed and passes quality inspection, the core metadata corresponding to that process, such as the racket's unique identifier, process number, process parameter hash value, operator, equipment number, quality inspection result, and timestamp, is generated into an independent block. The block contains the hash value of the previous process block, forming a chain-like evidence storage structure for the entire process of a single racket. The core metadata of the process is a set of structured fields. In addition to the fields mentioned above, it also includes raw material batch number, process production time, environmental temperature and humidity parameters, and quality inspector number. All fields are character or numerical structured data. The hash value of the process parameters is generated using the SHA256 algorithm, and the timestamp is in UTC+8 standard time format to ensure the integrity and uniqueness of the core metadata.

[0033] The blocks in the consortium blockchain adopt a fixed structure, which includes the block version number, the hash value of the previous block, the root hash of the core metadata of this block, the timestamp, the consensus node signature, and the random number. The hash value of the first process block of a single racket is set to an initial fixed value. Each subsequent process block extracts the complete hash value of the previous process block and associates it internally. During verification, the hash value of the previous block is compared with the actual hash value of the previous process block to determine whether the data has been tampered with. If the comparison is inconsistent, the data of the block is determined to be invalid.

[0034] Simultaneously, a hybrid on-chain and off-chain storage mode is adopted. On-chain storage only stores hash values ​​and core metadata to ensure that the data is tamper-proof and traceable. Off-chain storage uses a distributed database to store complete time-series process data, visual acquisition data, inspection reports, and other full data. The on-chain hash values ​​correspond one-to-one with the full off-chain data, ultimately achieving trusted, secure, and standardized integration of heterogeneous data across the entire process.

[0035] The off-chain distributed database uses HBase, with the racket's unique identifier and process number as the sharding key. The complete process data for a single racket is sharded and stored on different database nodes, and a replication synchronization mechanism between nodes ensures that no data is lost. On-chain and off-chain data are uniquely associated through the root hash of the core metadata of the process. When retrieving off-chain data, it is necessary to perform an accurate retrieval in the database using the on-chain root hash. At the same time, the hash value of the retrieved full data is recalculated and compared with the value stored on the chain. Only if they match can it be confirmed that the data has not been tampered with.

[0036] Among them, the correlation modeling stage is based on the trusted integration results of heterogeneous data of the whole process achieved in the trusted integration stage. For the six core quality indicators of badminton racket finished products, namely bending stiffness, torsional stiffness, balance point accuracy, ball rebound rate, fatigue life and coating adhesion, based on the data collected in the whole process, the grey relational analysis algorithm is used to quantify the influence weight of the key process parameters of each process on each quality indicator, and screen out the core main control parameters that affect the quality. The specific application of the grey relational analysis algorithm in the badminton racket production line is as follows: The industry standard values ​​of six core quality indicators of the finished badminton racket, such as bending stiffness and torsional stiffness, are set as the reference sequence, and the actual collected values ​​of key process parameters of each process are set as the comparison sequence. First, all sequences are initialized to eliminate differences in dimensions and orders of magnitude. The resolution coefficient is fixed at 0.5. The correlation coefficient and grey weighted correlation degree between each comparison sequence and the reference sequence are calculated in turn. The correlation degree greater than 0.6 is used as the screening threshold to determine the main control parameters that have a core impact on the quality indicators.

[0037] By combining graph neural network technology, a deep correlation graph is constructed for a single racket, process, main control parameters, quality indicators, and defect types. This clarifies the causal relationship between process parameters and quality indicators. For example, the influence weights of the heating rate, holding temperature, and holding time of the hot pressing curing process on the racket's fatigue life, and the influence weights of the carbon cloth type and layup angle of the layup process on the racket's torsional stiffness. At the same time, the correlation graph is deeply bound to the semantic ontology model and the digital twin model.

[0038] The deep association graph is built on the GCN graph convolutional neural network architecture. The graph has five core node types: single racket ID, process number, main control parameter name, quality indicator name, and defect type name. The relationship between nodes is represented by the edges of the graph. Among them, the process and the main control parameter are "containment" edges, the main control parameter and the quality indicator are "influence" edges, the quality indicator and the defect type are "cause" edges, and the racket ID and the process are "experience" edges. The weights of the edges are directly assigned by the parameter influence weights obtained from the grey relational analysis algorithm. The node feature extraction and deep learning of the association relationship are completed through two layers of graph convolutional layers.

[0039] The quality traceability stage is based on the deep association graph constructed in the association modeling stage, the semantic ontology model, and the chain-like evidence data in the trusted integration stage. It constructs a digital twin mirror of the entire badminton racket production line to achieve multi-dimensional and refined traceability. By inputting the racket's unique identifier, the entire lifecycle data of the racket can be visualized, from raw material warehousing, process parameters of each process, operators, equipment status, quality inspection results, finished product testing, packaging and warehousing, to sales flow. At the same time, the production process of key processes can be simulated and replayed through the digital twin model to intuitively restore the entire production process. The digital twin mirror is built on the Unity3D engine. First, a 1:1 3D solid model of the physical production line of badminton rackets is obtained through laser 3D scanning. Then, the semantic ontology model's process, equipment, and parameter data, as well as the chain-stored full process time-series data, are precisely bound to the 3D model. Motion logic and parameter association logic consistent with the physical production line are configured for each production equipment and process link. The digital twin simulation playback drives the digital twin model to reproduce the entire process of process operation, equipment operation, and parameter changes according to the actual production timeline by calling the full process time-series data stored in the off-chain distributed database. The playback speed supports free adjustment of 1x, 2x, and 10x speed, and can directly locate any process node to view the real-time process parameters and equipment operating status of that node.

[0040] When a quality defect occurs in the finished product, the corresponding core influencing process, process parameter deviation, equipment abnormality, and raw material batch can be located in reverse through the correlation map based on the defect type. The root cause of the quality problem can be accurately identified, and the on-chain evidence data of the corresponding process can be retrieved to ensure the immutability and reliability of the traceability results. The quality defects of badminton rackets are classified into six categories according to the six core quality indicators: insufficient bending stiffness, insufficient torsional stiffness, excessive balance point accuracy, insufficient rebound rate, low fatigue life, and coating adhesion failure. Each category of defects is further divided into three subcategories according to the degree of deviation: minor, moderate, and severe. The matching rule for the defect type and depth correlation map is as follows: first, the actual defect type is mapped to the corresponding quality indicator node in the map; then, the associated main control parameter node is traced down through the "influence" edge of the node; finally, the corresponding process node is traced through the "containment" edge of the main control parameter node, thus achieving a precise match between defects and processes and parameters.

[0041] When a certain piece of equipment or a certain batch of raw materials malfunctions, all badminton racket products corresponding to that equipment or batch of raw materials can be quickly retrieved, enabling rapid location and full-chain interception of risky products and preventing defective products from entering the market.

[0042] The quality early warning stage is based on the quality correlation map constructed in the correlation modeling stage. It sets dynamic thresholds for the core control parameters of each process. These dynamic thresholds are not fixed values, but are dynamically adjusted in real time based on the parameter data of the previous process of the racket, the optimal process parameters of the same model product, and the influence weights in the correlation map. For example, when there is a slight deviation in the previous layer angle, the system automatically and dynamically adjusts the parameter thresholds of the subsequent hot pressing and curing process to compensate for the previous deviation through process collaboration. The dynamic threshold calculation logic for the core control parameters of the process is as follows: Dynamic threshold = Optimal process parameter of the same model product × 60% + Average qualified value of the preceding process parameter × 20% + (Parameter influence weight in the correlation diagram × Parameter industry standard value) × 20%; If a first-level warning occurs in the preceding process, the weight of the average qualified value of the preceding process parameter is adjusted to 30%, and the weight of the optimal process parameter of the same model product is adjusted to 50%. Through dynamic weight adaptation, process coordination compensation for the deviation of the preceding process parameters is achieved.

[0043] During the production process, the edge device compares the real-time collected process parameters with dynamic thresholds in real time, triggering a graded early warning and control mechanism. The first-level warning is for minor parameter deviations, and the system pushes the warning information to the on-site operators in real time, prompting them to adjust the parameters. At the same time, the warning data is stored on the blockchain for evidence. The second-level warning is for serious parameter deviations, and the system directly triggers the production line interception mechanism to prevent the racket from flowing into the next process. At the same time, the abnormal information is pushed to the process engineer for handling. Only when the process parameters meet the dynamic threshold requirements and the quality inspection is qualified can the racket be allowed to enter the next process.

[0044] The determination of parameter deviation is based on the dynamic threshold of the core control parameters of each process. When the deviation ratio between the actual parameter value and the dynamic threshold is within ±5%, it is judged as a slight parameter deviation and triggers a first-level warning. When the deviation ratio between the actual parameter value and the dynamic threshold exceeds ±5%, it is judged as a serious parameter deviation and triggers a second-level warning. The deviation ratio is calculated based on the dynamic threshold.

[0045] The process optimization stage is based on the hierarchical early warning and control data from the quality early warning stage, the root cause localization results from the quality traceability stage, and the correlation graph from the correlation modeling stage. It constructs a closed-loop optimization system for the entire process, including data collection, correlation analysis, traceability and localization, early warning and control, and process optimization. The full-process data, quality defect data, early warning and handling data, and finished product inspection data accumulated during the production line operation are continuously input into the graph neural network model of the correlation graph for iterative training of the model. This continuously optimizes the correlation between process parameters and quality indicators and outputs the optimal combination of process parameters for different models of badminton rackets. The batch size for iterative training of the graph neural network model was set to 64, and the learning rate adopted an adaptive adjustment mode with an initial learning rate of 0.001, which decreased by 0.1 after every 100 rounds of training. The termination condition for model iteration was that after 50 consecutive rounds of training, the average error rate between the predicted and actual quality indicators was less than 3%. The evaluation criteria for the optimal combination of process parameters were that, under this parameter combination, the compliance rate of the six core quality indicators, including the bending stiffness and torsional stiffness of badminton rackets, was 100%, and the production efficiency of a single racket on the production line was increased by ≥5% compared with the current process parameters.

[0046] The data flow path of the closed-loop optimization system is as follows: The data acquisition module synchronizes the preprocessed full data to the trusted integration module for storage, and then pushes it to the correlation modeling module, quality traceability module, and quality early warning module respectively; the hierarchical early warning data of the quality early warning module and the root cause localization results of the quality traceability module are fed back to the correlation modeling module in real time to provide data support for model iteration; the optimized parameters of the correlation modeling module are pushed to the process optimization module, and after digital twin simulation verification, the process optimization module distributes them to the equipment control system of each processing unit on the production line. At the same time, the process optimization module feeds back the optimization rules to the quality early warning module and the ontology modeling module to complete the update of the data and rules of the entire system. All data flow is completed through the production line industrial bus, and the execution node is the core data interaction server of each module.

[0047] The optimized process parameters are synchronously updated to the semantic ontology model and the production line digital twin model. The process simulation is verified through the digital twin model. After the verification is passed, the parameters are distributed to each processing unit of the production line to achieve standardized iterative upgrades of the production line process.

[0048] The specific process of digital twin process simulation verification is as follows: First, the optimized process parameters are imported into the digital twin model to build a simulated production scenario consistent with the physical production line. Then, the model is driven to complete the virtual production of a single racket throughout the entire process. Simultaneously, process parameters, equipment operating status, and virtual inspection data of finished products are collected during the simulation process. The core verification indicators are the compliance of six core quality indicators, the stability of process parameters, and the equipment operating load. The criteria for passing the verification are that all six core quality indicators of the simulated finished product meet the standards, the fluctuation range of process parameters is within the dynamic threshold range, the equipment operating load is within the reasonable range of 80%-90%, and the yield rate of virtual production is 100%.

[0049] Based on continuously accumulated quality defect data, the dynamic threshold and warning rules of the quality early warning model are optimized to improve the accuracy of early warning and realize intelligent manufacturing of the entire badminton racket production line.

[0050] This invention also provides a process data integration and quality traceability management system for intelligent manufacturing production lines of badminton rackets, based on the above method, including: The ontology modeling module outlines the entire process of intelligent manufacturing of badminton rackets, clarifies the core parameters and attributes of each process, constructs a five-in-one semantic ontology model, unifies data standards and interaction rules, and assigns a unique laser-engraved QR code or RFID electronic tag to each racket as an identification identifier, achieving unique anchoring of a single racket throughout its entire life cycle.

[0051] The data acquisition module, based on the results of the ontology modeling module, constructs an adaptive multimodal acquisition system to collect data from standardized equipment, non-standard manual and semi-automated processes, and offline laboratory testing. Data preprocessing is completed at the edge and bound to racket identification, providing a high-quality data foundation.

[0052] The trusted integration module builds a production line alliance chain network, generating core metadata blocks for each process of each racket to form a chain-like evidence storage structure. It adopts a hybrid on-chain and off-chain storage mode to achieve trusted, secure, and standardized integration of heterogeneous data across the entire process.

[0053] The correlation modeling module, based on trusted integrated data, uses relevant algorithms to quantify the influence weight of process parameters on quality indicators, constructs a deep correlation map and binds it to relevant models, providing data support for subsequent functions.

[0054] The quality traceability module constructs a digital twin mirror of the entire production line process, enabling forward traceability of individual rackets, reverse root cause localization of defects, and batch-level risk interception, ensuring reliable and accurate traceability results.

[0055] The quality early warning module sets dynamic thresholds based on the correlation graph, compares process parameters in real time, and triggers hierarchical early warning and control to achieve in-process interception of quality defects.

[0056] The process optimization module integrates data from various modules to build a closed-loop optimization system, iteratively trains the model to output optimal process parameters, updates relevant models and production line parameters, optimizes early warning rules, and achieves continuous improvement in production line intelligence and intelligent manufacturing throughout the entire process.

[0057] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for process data integration and quality traceability management in a smart manufacturing production line for badminton rackets, characterized in that, The specific steps include the following: In the ontology modeling stage, the entire process of intelligent manufacturing of badminton rackets is sorted out, the process parameters, quality characteristics and equipment, personnel and material related attributes of each core process are clarified, a semantic ontology model including products, processes, equipment, parameters and quality indicators is constructed, the heterogeneous data standards and interaction rules of the entire production line are unified, and a unique lifelong identity is assigned to each racket and bound to the product instance of the ontology model. During the data acquisition phase, a multimodal data acquisition system is constructed to collect various process and testing data from the production line and complete preprocessing. The preprocessed data is then bound to the unique identifier of the racket. In the trusted integration phase, a badminton racket production line alliance chain network is built to generate core metadata blocks for each process of each racket and form a chain-based evidence storage structure, using a hybrid on-chain and off-chain storage mode. In the correlation modeling stage, for the core quality indicators of the finished badminton racket, the correlation algorithm is used to quantify the influence weight of each process parameter on the quality indicators, screen the core control parameters and construct a deep correlation map to clarify the correlation between parameters and quality indicators. During the quality traceability phase, a digital twin mirror of the entire production line process is constructed to achieve forward traceability of the entire life cycle of a single racket, reverse root cause localization of quality defects, and batch-level interception of risky products. During the quality early warning stage, dynamic thresholds are set for the core parameters of each process, and the process parameters are compared with the thresholds in real time to trigger hierarchical early warning and control, thereby achieving in-process interception of quality defects. During the process optimization phase, a closed-loop optimization system is constructed to optimize the relationship between process parameters and quality indicators through iterative training of the model. The optimal process parameters are output and the relevant models are updated simultaneously, and the early warning rules are optimized.

2. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 1, characterized in that, The ontology modeling stage specifically includes: The nine core processes of intelligent manufacturing of badminton rackets are analyzed, and the core process parameters, key quality characteristics, equipment attributes, personnel attributes and material attributes of each process are clearly defined. Construct a semantic ontology model that integrates products, processes, equipment, parameters, and quality indicators, and unify the semantic standards, data formats, and interaction rules of heterogeneous data across the entire production line; Each badminton racket is assigned a laser-engraved QR code or RFID electronic tag as a lifelong identification, and the tag is bound to the product instance in the model.

3. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 2, characterized in that, The data acquisition phase specifically includes: Based on a semantic ontology model and a unique racket identifier, an adaptive multimodal acquisition system is built for three types of core data sources in the production line. Through industrial protocols including OPCUA, Modbus, and Profinet, real-time acquisition of process timing parameters and equipment operating status is achieved for standardized industrial equipment, including cutting machines, hot presses, and CNC drilling machines. For non-standard manual and semi-automated processes, machine vision and intelligent sensing are combined to collect data on process compliance and dynamic processes. Offline laboratory test data is synchronized through a standardized interface and supports manual structured data entry. All collected data undergoes time-series alignment, outlier filtering, data deduplication, missing value completion, and standardization at the production line edge before being bound to the corresponding racket identifier.

4. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 3, characterized in that, The trusted integration phase specifically includes: Build a consortium blockchain network for badminton racket production lines, with consensus nodes covering production line processing units, quality inspection units, raw material suppliers, end customers, and regulatory authorities; After each racket passes quality inspection at each stage, the corresponding core metadata is generated into an independent block. The block contains the hash value of the previous stage block, forming a chain-like evidence storage structure for the entire process of a single racket. It adopts a hybrid on-chain and off-chain storage mode. The hash value and core metadata are stored on the chain, while the full process and testing data are stored off-chain through a distributed database. The on-chain hash value corresponds one-to-one with the full off-chain data.

5. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 4, characterized in that, The association modeling stage specifically includes: For six core quality indicators of badminton racket finished products, namely bending stiffness, torsional stiffness, balance point accuracy, ball rebound rate, fatigue life, and coating adhesion, a grey relational analysis algorithm is used based on data collected throughout the entire process to quantify the influence weight of key process parameters of each process on the quality indicators and to screen the core control parameters. By combining graph neural network technology, a deep correlation graph is constructed for a single racket, process, main control parameters, quality indicators, and defect types. The causal relationship between parameters and quality indicators is clarified, and the correlation graph is deeply bound to a semantic ontology model and a digital twin model.

6. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 5, characterized in that, The quality traceability stage specifically includes: Based on deep association graphs, semantic ontology models, and chain-based evidence data, a digital twin mirror of the entire production line process is constructed. By entering the racket's unique identifier, you can visually display the racket's entire lifecycle data and simulate and replay the key production processes. When a quality defect occurs in the finished product, the core influencing process, parameter deviation, equipment abnormality and raw material batch are located in reverse through the correlation map based on the defect type, the root cause is identified and the on-chain evidence data is retrieved. When there is an abnormality in the batch of equipment or raw materials, quickly retrieve all related racket products.

7. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 6, characterized in that, The quality early warning stage specifically refers to: Based on the quality correlation graph, dynamic thresholds are set for the core control parameters of each process. The thresholds are dynamically adjusted in real time according to the data of the racket's previous processes, the optimal process parameters of the same model, and the influence weight of the parameters. The edge device compares process parameters with dynamic thresholds in real time, triggering tiered early warning and control: Level 1 warning indicates minor parameter deviation, the system pushes warning information and adjustment direction, and records the data on the blockchain for evidence; Level 2 warning indicates severe parameter deviation, the system triggers production line interception, prohibits rackets from flowing into the next process, and pushes abnormal information to the process engineer; once the process parameters meet the standards and pass quality inspection, the racket can proceed to the next process.

8. The method for process data integration and quality traceability management for intelligent manufacturing production lines of badminton rackets according to claim 7, characterized in that, The process optimization stage specifically includes: Based on hierarchical early warning data, root cause localization results, and correlation maps, a closed-loop optimization system is constructed, encompassing data collection, correlation analysis, traceability and localization, early warning and control, and process optimization. The entire production line data is continuously input into the graph neural network model for iterative training to optimize the relationship between process parameters and quality indicators, and output the optimal combination of process parameters for different racket models. The optimized parameters are synchronously updated to the semantic ontology model and digital twin model, and after simulation verification, they are sent to the production line processing unit. The dynamic threshold and early warning rules are optimized based on the quality defect data.

9. A process data integration and quality traceability management system for intelligent manufacturing production lines of badminton rackets, based on the method described in any one of claims 1-8, characterized in that, include: The ontology modeling module is used to sort out the entire process of intelligent manufacturing of badminton rackets, clarify the core parameters and attributes of each process, construct a five-in-one semantic ontology model, unify data standards and interaction rules, and assign a unique identity to each racket. The data acquisition module is used to build an adaptive multimodal acquisition system, collect various process and inspection data of the production line, and complete data preprocessing and binding with racket identification at the edge. The trusted integration module is used to build a production line alliance chain network, generate core metadata blocks for each process of each racket and form a chain-like evidence storage structure, and adopt a hybrid on-chain and off-chain storage mode. The correlation modeling module is used to quantify the influence weight of process parameters on quality indicators using correlation algorithms, construct a deep correlation graph and bind it to relevant models. The quality traceability module is used to build a digital twin mirror of the entire production line process, enabling forward traceability of a single racket, reverse root cause localization of defects, and batch-level risk interception. The quality early warning module is used to set dynamic thresholds based on the correlation graph, compare process parameters in real time, and trigger hierarchical early warning and control to achieve in-process interception of quality defects. The process optimization module is used to integrate data from various modules to build a closed-loop optimization system, iteratively train the model to output optimal process parameters, update relevant model and production line parameters, and optimize early warning rules.

Images