A digital tracking system for hosiery processing materials

Abstract

The application discloses a kind of knitted fabric processing material digitization tracking system, it is related to textile processing information management technical field, the system aims at solving the problems such as physical label failure under knitted fabric processing environment, cross-process data island and production scheduling lag.The system includes: multi-modal environment adaptive perception module, material information is obtained by fusing physical carrier identification and fabric feature extraction;Distributed heterogeneous data fusion gateway realizes cross-process data aggregation and normalization processing;Full life cycle digital twin engine builds material digital image and maintains its space-time attributes;Space-time atlas correlation analysis unit identifies flow bottleneck through dynamic topological atlas;And intelligent scheduling collaborative execution terminal generates optimization instruction and drives material flow execution.The application realizes the accurate tracking and intelligent scheduling of knitted fabric full process, and significantly improves material turnover efficiency.

Description

Technical Field

[0001] This invention belongs to the field of information management in textile processing, and specifically relates to a digital tracking system for knitted fabric processing materials. Background Technology

[0002] As a crucial component of the traditional industrial system, the digital transformation and intelligent upgrading of the textile manufacturing process have become core drivers for improving production efficiency and resource utilization. In a highly integrated factory environment, accurate identification, real-time monitoring, and end-to-end traceability of production materials are fundamental supports for achieving flexible production and lean management, and also serve as a key link connecting underlying equipment with upper-level management systems.

[0003] The knitting process involves multiple complex steps, including yarn selection, weaving techniques, dyeing and finishing, and finished product inspection. This places extremely high demands on the efficiency and accuracy of material flow during the transformation of materials into different physical forms. The digital tracking system, through sensor, identification, and information processing technologies, aims to achieve seamless connection and status synchronization of knitted materials across workshop workstations, storage areas, and processing machines.

[0004] Existing technologies typically employ manual paper records or simple barcode scanning for material tracking. However, in the complex environments of high temperature, high humidity, and the presence of chemicals in knitted fabric processing, physical tags are easily damaged or detached, leading to reading failures. Furthermore, existing tracking systems often suffer from severe data silos when handling cross-process material flow, with insufficient information correlation between different production stages, failing to reflect the precise location and real-time quality status of materials in the dynamic processing path. In addition, for the high-volume, multi-variety knitted production model, existing systems lack intelligent analysis capabilities for complex material flows, resulting in delayed production scheduling optimization and frequent material accumulation and downtime due to material shortages. Summary of the Invention

[0005] The present invention provides a digital tracking system for knitted fabric processing materials to solve the technical problems mentioned in the background art, such as the physical tags in the knitted fabric processing environment being easily damaged by high temperature, high humidity and chemical agents, resulting in failure; data silos in cross-process material flow; lack of intelligent analysis capability for material flow; and production scheduling lag.

[0006] To solve the above problems, the technical solution adopted by the present invention is as follows: A digital tracking system for knitted fabric processing materials includes: The multimodal environment adaptive sensing module is used to acquire material identity information and physical state data in real time in the extreme environment of knitted fabric processing by integrating physical carrier recognition and bio-inspired fabric feature extraction technology. The distributed heterogeneous data fusion gateway is used to clean and normalize unstructured data collected by the multimodal environment adaptive sensing module, and to realize data aggregation across devices and processes based on a preset communication protocol. The full lifecycle digital twin engine is used to receive real-time data transmitted by a distributed heterogeneous data fusion gateway, build a digital image that corresponds one-to-one with the physical material, and maintain the spatiotemporal attribute tensors of the material at different processing stages. The spatiotemporal graph correlation analysis unit is used to construct a dynamic spatiotemporal topology graph of the entire production process based on the material status maintained by the full life cycle digital twin engine, and to identify bottleneck nodes and abnormal fluctuations in the material flow path. The intelligent scheduling and collaborative execution terminal is used to automatically generate material flow optimization instructions based on the output results of the spatiotemporal graph correlation analysis unit, and drive the logistics execution mechanism to complete the precise delivery and workstation flow of materials.

[0007] As one embodiment of the present invention, the multimodal environment adaptive sensing module includes a weather-resistant microelectronic identification unit, a structured feature visual capture unit, and a multispectral fabric fingerprint extraction unit. The weather-resistant microelectronic identification unit uses a radio frequency identification chip with acid and alkali corrosion resistant packaging, which is embedded in the seam edge of the knitted fabric or the carrier structure to achieve contactless reading of basic identity. The structured feature visual capture unit uses an industrial imaging component deployed above the workstation to collect images of the surface texture and structure of the knitted fabric in real time during transport. The multispectral fabric fingerprint extraction unit uses a light source of a specific wavelength to illuminate the fabric surface, extracts the unique spectral features formed by the differences in fiber composition and dye distribution, and converts them into a unique material fingerprint code. The multimodal environment adaptive sensing module ensures that even when the physical tag is damaged or missing, the system can still accurately determine the material identity through the physical properties of the fabric itself by weighted fusion of radio frequency identification information, texture structure features, and multispectral fingerprint codes.

[0008] Furthermore, the multimodal environment adaptive perception module also includes an environmental compensation factor calculation unit, which is used to monitor the temperature, humidity and light intensity of the processing site. The environmental compensation factor calculation unit dynamically compensates the imaging parameters of the structured feature visual capture unit according to the real-time environmental parameters. In environments where the humidity exceeds 85% and there is water vapor interference, it automatically increases the image contrast and enables infrared supplementary lighting to ensure the stability of image feature extraction.

[0009] As one embodiment of the present invention, the distributed heterogeneous data fusion gateway adopts a two-layer architecture, including an edge preprocessing layer and a collaborative aggregation layer. The edge preprocessing layer is deployed near each processing unit in the production workshop and is responsible for filtering and reducing the dimensionality of features in the massive amount of sensing data on-site, eliminating redundant background noise. The collaborative aggregation layer connects each edge preprocessing layer through a deterministic industrial network and uses a time-sensitive network protocol to ensure that the transmission latency of key material location update data is less than 10 milliseconds. The distributed heterogeneous data fusion gateway establishes a unified data semantic dictionary to transform processing equipment data from different manufacturers and different communication protocols into material status descriptors that conform to the standard, thereby eliminating information semantic barriers between processes.

[0010] Furthermore, the distributed heterogeneous data fusion gateway also integrates a data consistency verification module. This module compares the material flow records of adjacent workstations and uses Markov logic networks to perform state consistency reasoning. When it is found that the output time of a certain material in the second process is later than its input time in the third process, it automatically triggers an abnormal alarm and starts the traceability logic to correct the data deviation caused by sensor misreading.

[0011] As one embodiment of the present invention, the full lifecycle digital twin engine includes a dynamic attribute tensor construction unit, a physical simulation mapping unit, and a historical evolution database. The dynamic attribute tensor construction unit establishes a multi-dimensional vector for each material entering the production system. This vector contains parameters such as the material's raw material composition, batch number, current process, historical processing temperature, tensile stress, and real-time moisture content. The physical simulation mapping unit uses a preset knitted fabric mechanical model, combined with real-time collected processing parameters, to simulate the physical deformation and quality evolution of the material in key processes such as shaping and finishing. The historical evolution database is responsible for storing the entire process data of the material from raw yarn to finished product, providing in-depth data support for subsequent quality traceability and process optimization.

[0012] Furthermore, the full lifecycle digital twin engine also includes a virtual sensing synchronizer, which is used to predictively fill the material status of the offline node based on the production plan schedule and the turnover rate of the same batch of materials when the physical sensor is temporarily offline due to equipment maintenance, thereby maintaining the continuity of the digital twin entity.

[0013] As one embodiment of the present invention, the spatiotemporal graph association analysis unit includes a dynamic topology construction module, a flow bottleneck identification module, and a path conflict detection module. The dynamic topology construction module defines workshop workstations as nodes and material flow paths as edges, and constructs a complex production network graph containing tens of thousands of nodes and edges in real time. The flow bottleneck identification module identifies redundant workstations that cause production backlogs by calculating the flow density and residence time distribution of each node in the graph. The path conflict detection module calculates the time window overlap rate of different material batches at intersections or shared equipment based on the predicted material flow direction, and warns of potential logistics collisions and work stoppages due to material shortages.

[0014] Furthermore, the spatiotemporal graph correlation analysis unit employs a deep learning framework based on graph convolutional networks to extract deep spatiotemporal features of material flow paths. This framework analyzes flow trajectory data from the past 360 consecutive hours to learn the flow rhythm of different types of knitted fabrics under specific process combinations, thereby improving the prediction accuracy of the material arrival time at the next process to over 95%.

[0015] As one embodiment of the present invention, the intelligent scheduling and collaborative execution terminal includes a multi-objective optimization decision-maker, a task distribution interface, and a field feedback controller. The multi-objective optimization decision-maker receives early warning information from the spatiotemporal graph correlation analysis unit and solves for the optimal scheduling scheme under the constraints of the shortest production cycle, the most balanced equipment load, and the least energy consumption. The task distribution interface converts the generated scheduling decision into instructions that can be recognized by each execution mechanism, driving the automated guided vehicle, industrial robot, and intelligent suspension system to complete the handling and sorting of materials. The field feedback controller monitors the execution status of the scheduling instructions in real time, and automatically starts the replanning logic when the execution deviation exceeds a preset 5% threshold.

[0016] Furthermore, the intelligent scheduling and collaborative execution terminal also has an autonomous learning module. This module uses a reinforcement learning algorithm to use the result of each successful or failed scheduling as a reward or penalty signal, continuously optimizing the parameter configuration of the scheduling strategy, so that the system can adapt to the extreme flexible production needs of product switching frequency increased by more than 3 times.

[0017] As one embodiment of the present invention, the system also includes a quality status collaborative diagnosis module, which associates the digital tracking information of the material with the quality data of the online detection system; when a batch of knitted fabric is detected to have color difference or weight deviation, the system automatically reverse-tracks its real-time temperature curve and chemical concentration record in the dyeing and finishing process, accurately locates the source of the fault and automatically isolates the affected related materials.

[0018] As one embodiment of the present invention, the system runs on a distributed computing platform based on a private cloud, and containerization technology is used to realize the independent deployment and elastic expansion of each functional module. The system adopts a dual-machine hot standby mechanism to ensure that when the main server fails, the backup server can take over all tracking services within 2 seconds, ensuring the 24-hour uninterrupted operation of digital tracking services.

[0019] As one embodiment of the present invention, the user interaction layer of the system adopts augmented reality technology, which overlays the digital twin information of the material onto the physical entity through a head-mounted display device or a handheld terminal; the operator only needs to scan the material through the terminal to view the process requirements, remaining processing time and directional guidance of the next workstation in real time in the field of view.

[0020] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention integrates radio frequency identification, visual features and multispectral fingerprint technology by constructing a multimodal environment adaptive perception module, which fundamentally solves the identification problem caused by the easy damage and failure of traditional physical tags in the high temperature, high humidity and strong acid and alkali environment of knitting processing; the system extracts the inherent physical features of the fabric itself as identity credentials, ensuring the non-lossability and identification certainty of materials in the whole process flow, and greatly improving the reliability of the underlying data.

[0021] (2) This invention breaks the long-standing problem of data silos between processes in textile enterprises by using a distributed heterogeneous data fusion gateway and a full life cycle digital twin engine. The system constructs a dynamically updated digital image for each bundle of yarn and each piece of greige fabric, realizing an end-to-end data loop from raw material warehousing, weaving, dyeing and finishing to finished product delivery. This deep digital mapping not only improves production transparency, but also provides second-level response capability for accurate quality traceability, reducing the time for locating quality anomalies from several hours to minutes.

[0022] (3) This invention utilizes a spatiotemporal graph correlation analysis unit and an intelligent scheduling collaborative execution terminal to upgrade material tracking from passive recording and querying to proactive prediction and decision-making; through real-time analysis of the topology graph of the entire production process, the system can identify flow bottlenecks in advance and automatically optimize paths, effectively reducing material accumulation and downtime due to material shortages between workstations; in practical applications, this system can improve material turnover efficiency while reducing manual recording costs, significantly enhancing the flexible response capability and lean production level of knitting factories under the parallel processing mode of large batches and multiple varieties.

[0023] (4) This invention introduces an environmental compensation mechanism, a predictive filling algorithm and a reinforcement learning scheduling strategy, which endows the system with strong robustness and self-evolution capability. The system can automatically adjust the sensing parameters according to changes in the production environment, maintain business continuity when the sensor is abnormal, and continuously optimize the scheduling performance as production experience is accumulated. This intelligent design enables the system to perfectly meet the needs of the traditional textile industry to transform into intelligent manufacturing, laying a solid technical foundation for realizing unmanned workshops and smart factories.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, embodiments of the present invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the overall technical architecture of the digital tracking system for knitted fabric processing materials proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the full lifecycle digital twin engine and spatiotemporal graph correlation analysis in this invention; Figure 3 This is a logical flowchart of the multimodal environment adaptive perception and distributed data fusion stage in this invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 some embodiments of the present invention, but not all embodiments. Example

[0028] Please refer to the attached document. Figure 1This embodiment discloses a digital tracking system for knitted fabric processing materials, which constructs a complete technical closed loop covering physical sensing, data aggregation, digital modeling, spatiotemporal analysis, and closed-loop execution at the global architecture level. The system is deployed on a distributed computing platform based on a private cloud, utilizing containerization technology to decompose each functional module into microservices, ensuring flexible horizontal scalability when facing large-scale knitting production lines. The high temperature, high humidity, and strong chemical corrosion characteristics commonly found in knitted fabric processing environments pose significant challenges to traditional barcodes or ordinary electronic tags. This embodiment achieves robust identification of materials through a multimodal environment adaptive sensing module.

[0029] Combined with appendix Figure 3 The multimodal environment adaptive sensing module is located at the bottom sensing layer of the system. This module integrates a weather-resistant microelectronic identification unit, a structured feature visual capture unit, a multispectral fabric fingerprint extraction unit, and an environmental compensation factor calculation unit. The weather-resistant microelectronic identification unit serves as the first layer of identity verification, and its core is a specially packaged radio frequency identification (RFID) chip. The chip's packaging material uses a composite structure of polyphenylene sulfide and special epoxy resin, capable of resisting long-term corrosion from high-temperature steam at 150 degrees Celsius and strong acid and alkaline dyeing and finishing solutions. This chip is embedded in the seam of knitted fabrics or in the protected structure of material transport vehicles, transmitting basic material identification information, such as batch number, yarn composition, and original order number, via non-contact radio frequency signals.

[0030] The structured feature visual capture unit is deployed above the key material handling port of the production line. Its core hardware consists of a high-resolution industrial camera and a fixed-focus lens. As the knitted fabric flows through the detection area at a speed of 30 to 120 meters per minute, this unit can acquire real-time images of the fabric's micro-texture. Because knitted fabrics are composed of interlocking loops, their loop density, tilt, and micro-hair distribution possess inherent randomness and uniqueness, forming physically unreplicable characteristics. To maintain imaging quality in extreme environments, the environmental compensation factor calculation unit plays a crucial role. This unit monitors the temperature, humidity, and light intensity data in real time. When the ambient humidity exceeds 85% and there is a large amount of water vapor, the environmental compensation factor calculation unit outputs a compensation command, driving the structured feature visual capture unit to automatically switch to infrared illumination mode and utilize image enhancement algorithms to improve contrast, filter out scattering noise caused by water vapor, and ensure the edge sharpness of the loop structure.

[0031] The multispectral fabric fingerprint extraction unit further enhances the uniqueness of identification. This unit illuminates the fabric using a specific light source covering ultraviolet, visible, and near-infrared bands. Different fiber components and their dyes exhibit significant differences in absorption and reflectivity for different wavelengths of light; these differences are transformed into a multidimensional spectral feature vector. The system generates a multimodal fusion-based unique material fingerprint code by fusing this vector with texture features extracted by the structured feature visual capture unit and combining it with the basic information from the RFID chip. Even if the physical RFID tag is physically damaged under extreme mechanical stress, the system can still accurately determine the identity in subsequent processes by recognizing the fabric's own microscopic physical fingerprint.

[0032] The raw data collected by the multimodal environment adaptive sensing module is processed through a distributed heterogeneous data fusion gateway. Please refer to the appendix for further details. Figure 3 The gateway employs a two-layer architecture, consisting of an edge preprocessing layer and a collaborative aggregation layer. The edge preprocessing layer is deployed on edge computing nodes in each processing unit of the workshop, directly connecting to sensing devices. Its primary task is to perform on-site filtering of massive amounts of unstructured visual data. By running a pre-defined lightweight feature reduction algorithm, the edge preprocessing layer transforms terabytes of raw image data into kilobytes of feature vectors, significantly reducing the transmission pressure on the backbone network. The collaborative aggregation layer connects via a deterministic industrial network, employing a time-sensitive network protocol. This protocol, through strict time-slicing of network traffic, ensures that the transmission latency of critical data packets, such as material location updates, remains below 10 milliseconds, providing underlying communication assurance for real-time tracking.

[0033] The distributed heterogeneous data fusion gateway also establishes a unified data semantic dictionary. In the knitting process, various devices from different manufacturers are involved, such as weaving machines, dyeing machines, and setting machines, whose communication protocols cover multiple industrial bus standards. The data semantic dictionary transforms this heterogeneous data into material state descriptors that conform to a unified standard. For example, the temperature parameters of a dyeing machine from one brand and the speed parameters of a setting machine from another brand are uniformly mapped to the heat treatment intensity component in the material state tensor, thereby eliminating information semantic barriers between processes.

[0034] To ensure absolute data accuracy, the logic network consistency verification module integrated within the distributed heterogeneous data fusion gateway plays a role in data correction. This module utilizes Markov logic networks for state consistency reasoning. Its core logic lies in the fact that physical materials must adhere to irreversible time and space constraints during production processes.

[0035]

[0036] The above formula describes the probability distribution of material state transitions. Wherein, Conditional probability, which is the probability of the material's current state occurring given that its state at the previous moment and its current position are known. For the first The state of a material at time t includes all digital attributes such as material identity, batch, current process, quality parameters, and processing progress; For the first The state of a material at time t−1, that is, the final state of the material in the previous process (such as the output time and completion quality of the previous process). For the first The physical spatial logical location of a material at time t corresponds to a unique number of the workshop workstation, processing equipment, buffer area, or logistics track. The weight coefficient of the kth logical constraint is obtained by training with historical production data. The higher the weight, the stronger the physical credibility of the constraint. Let k be the logical constraint function, which describes the physical rules that the material flow must meet (such as "the input time of the subsequent process must be later than the output time of the previous process"), and the output is the constraint satisfaction degree. This is a normalization constant. Using this formula, when the system detects that the output record time of a certain material in the second process is later than its input record time in the third process, the verification module will immediately determine that it is a sensor misread or data packet out of order, automatically trigger an abnormal alarm and start the traceability logic, and use the redundant data of the preceding and following processes to correct the deviation.

[0037] The standardized data, after being processed by the gateway, is injected into the full lifecycle digital twin engine. Please refer to the appendix. Figure 2 This engine is the core data modeling center of the entire system, consisting of dynamic attribute tensor construction units, physical simulation mapping units, historical evolution databases, and virtual perception synchronizers. The dynamic attribute tensor construction unit creates a high-dimensional vector for each material entering the production system. This vector not only contains static attributes such as raw material composition, batch number, and current process, but also records dynamic parameters during processing in real time, such as tensile stress on the knitted fabric, instantaneous moisture content, dye concentration, and historical temperature and humidity curves.

[0038] The physical simulation mapping unit utilizes a pre-set multi-scale mechanical model of knitted fabrics, combined with the aforementioned dynamic parameters, to simulate the physical evolution of materials during key processes such as setting and finishing. For example, in the heat setting process, the system calculates the stress relaxation state of the fibers inside the material based on real-time collected data on the fabric width tensile force and oven temperature, predicting its final weight and shrinkage rate. This simulation is not performed offline but is synchronized with physical production, thus achieving a real-time mirroring of the material's quality state in the digital space.

[0039] The full lifecycle digital twin engine also integrates a virtual sensing synchronizer to address data interruption issues caused by sensor failure. In complex industrial environments, sensors may go offline due to maintenance or sudden malfunctions. In such cases, the virtual sensing synchronizer uses a Gaussian process regression algorithm to predictively fill in the material status of offline nodes based on production scheduling and the flow rate of other materials in the same batch. This algorithm calculates the covariance between known and unknown points in space and time, providing the optimal estimate of material location and status, ensuring the continuity of the digital twin entity in the spatiotemporal dimension.

[0040] Based on the precise data provided by the full lifecycle digital twin engine, the spatiotemporal mapping correlation analysis unit conducts in-depth analysis of the entire production process. Combined with the attached... Figure 2 This unit includes a dynamic topology construction module, a flow bottleneck identification module, and a path conflict detection module. The dynamic topology construction module defines hundreds of workstations, thousands of storage units, and dozens of logistics tracks throughout the workshop as nodes in graph theory, and defines the flow paths of materials between nodes as edges. As materials move continuously, the system updates this complex dynamic topology graph containing tens of thousands of nodes and edges in real time.

[0041] The bottleneck identification module identifies material accumulation points caused by equipment failure or process speed mismatch by calculating the flow density of nodes in the graph in real time. For example, when the average dwell time of the dyeing process of a certain type of knitted fabric exceeds a preset threshold of 25%, the system will accurately locate the source of the bottleneck through connectivity analysis of the graph. The path conflict detection module focuses on predicting future logistics collision risks.

[0042]

[0043] The above formula illustrates the basic feature extraction layers of the graph convolutional network used in the spatiotemporal graph correlation analysis unit. Among them, The feature vector representing a node contains multi-dimensional attributes such as the workstation's flow density, average residence time, equipment load, and work-in-process batches. For the first l In the +1 layer graphical convolutional network, the number is... i The node corresponds to a specific workstation, equipment, or logistics node in the workshop; For the first l Nodes in a layered graph convolutional network i The neighboring node j, i.e., the workstation i Directly connected upstream and downstream workstations / equipment; Represents a node ; It is a non-linear activation function; The neighborhood normalization coefficient; This represents the learnable weight matrix of the l-th layer graph convolutional network. Through this hierarchical aggregation, the system can extract deep spatiotemporal patterns in the material flow path. By analyzing historical trajectories over the past 360 consecutive hours, this framework learns the flow rhythm under different product types and process combinations, achieving a material arrival time prediction accuracy of over 95.8%.

[0044] The results of prediction and analysis ultimately converge at the intelligent scheduling and collaborative execution terminal. This terminal consists of a multi-objective optimization decision-maker, a task distribution interface, a field feedback controller, and a self-learning module. The multi-objective optimization decision-maker is a complex planning solver whose optimization objectives include minimizing production cycles, maximizing equipment load balance, and minimizing energy consumption. Faced with extremely flexible demands involving diverse products and small batches, this decision-maker can find the optimal solution from millions of possible path combinations within seconds. The generated scheduling scheme is converted into standard industrial control commands through the task distribution interface.

[0045] These commands drive the automated guided vehicles, intelligent overhead conveyor systems, and industrial robots within the workshop. The field feedback controller monitors the real-time pose and execution status of each actuator at a frequency of 200 milliseconds. If material handling deviation exceeds a set threshold of 5% due to ground obstacles or mechanical jamming, the field feedback controller immediately halts the execution flow and requests a multi-objective optimization decision-maker to perform local replanning.

[0046] The intelligent scheduling and collaborative execution terminal also possesses powerful self-evolution capabilities. The autonomous learning module employs a deep reinforcement learning framework, using the success rate, execution time, and resource utilization rate of each scheduling decision as reward signals. During long-term operation, the system continuously adjusts the parameter weights of the multi-objective optimization decision-maker, enabling it to adapt to extreme production environments where product switching frequencies increase by more than three times.

[0047] The system also features a dedicated quality status collaborative diagnostic module. This module deeply integrates the digital tracking chain of materials with online quality inspection data on the production line. Once the online inspection unit detects significant color difference fluctuations in a piece of knitted fabric, the quality status collaborative diagnostic module immediately retrieves the entire lifecycle digital twin of the material, tracing back all real-time parameters during dyeing, washing, and other processes. By comparing with standard process curves, the system can accurately pinpoint the cause as flow pulsation in the No. 3 feed pump leading to auxiliary agent ratio deviations, and automatically calculate other related batches of materials affected by this malfunction, achieving early isolation of quality risks.

[0048] To improve human-machine collaboration efficiency, augmented reality technology has been introduced into the system's user interaction layer. When production site managers scan materials using hand-held smart terminals or head-mounted augmented reality devices, the system overlays twin information hidden in the digital space onto the physical entity in real time, based on the material's digital identity. Operators can intuitively see the material's current processing progress bar, estimated completion time, specific location guidance for the next process, and the historical quality pass rate of the batch, achieving seamless information integration between the physical and digital worlds.

[0049] The system's security and reliability are ensured by a dual-machine hot standby mechanism. The master and slave servers synchronize memory status and database logs in real time via Fibre Channel. Once the master server detects a hardware-level heartbeat interruption, the backup server will take over the business within 1.5 seconds. At this time, all running distributed heterogeneous data fusion gateways and execution terminals will automatically switch connections, ensuring uninterrupted digital tracking operations in a 24 / 7 continuous production environment, with data integrity exceeding 99.99%.

[0050] Example 2 This embodiment focuses on describing the specific implementation details of the digital tracking system for knitted fabric processing materials in a large-scale dyeing and finishing workshop environment, to demonstrate its adaptability in extreme chemical environments and high-frequency material sorting scenarios. During the dyeing and finishing process, knitted fabrics are in a highly humid state and saturated with various chemical auxiliaries, which places extremely high demands on the continuous tracking of material identification.

[0051] In this embodiment, the multimodal environment adaptive sensing module features specialized hardware enhancements. In addition to the aforementioned high-temperature resistant packaging, the weather-resistant microelectronic identification unit integrates a micro-storage sector within its RF chip for backing up the three most critical processing parameters. The structured feature visual capture unit employs a protective cover with an automatic cleaning function, using compressed air to periodically blow away condensation droplets from the lens surface. The environmental compensation factor calculation unit incorporates color temperature monitoring. Due to the complex lighting environment in the dyeing and finishing workshop, which is significantly affected by water vapor refraction, this unit calculates the color temperature shift of the ambient light in real time and dynamically adjusts the white balance parameters of the image acquisition system to ensure that the extracted fabric fingerprint code is not ambiguous due to color temperature deviation.

[0052] When processing dyeing and finishing process data, the distributed heterogeneous data fusion gateway enhances the detection capability of sensor drift through its edge preprocessing layer. Since chemical sensors are prone to zero-point drift at high temperatures, the gateway compares the spatial correlation of multiple sensors within the same process segment and uses differential filtering technology to eliminate abnormal observation noise. The collaborative aggregation layer here executes even more stringent data alignment logic, forcibly correlating the precise, second-level time of material entering the dyeing vat with the process curve of the dyeing vat control system, forming an immutable digital chain of process evidence.

[0053] When modeling the dyeing and finishing process, the full lifecycle digital twin engine incorporates a chemical kinetics simulation module into its physical simulation mapping unit. This module dynamically predicts the diffusion depth of dye within the fiber based on real-time monitoring of dye liquor concentration, temperature, and the number of material cycles within the vat. This prediction is written in real-time into a multi-dimensional vector maintained by the dynamic attribute tensor construction unit, providing parameter guidance for subsequent finishing processes. For example, if the system predicts that the color yield of the current batch is slightly lower than the standard value, the intelligent scheduling and collaborative execution terminal will automatically adjust the speed of the setting machine during the subsequent setting process, compensating for the slight deviation in color depth by extending the heating time.

[0054] In this embodiment, the spatiotemporal graph correlation analysis unit focuses on solving the complex path planning problem in the dyeing and finishing workshop. Dyeing and finishing processes typically involve multiple rounds of reflow and rework, resulting in highly nonlinear material flow paths. The dynamic topology construction module analyzes the occupancy status and queuing sequence of each dyeing vat to calculate the expected cost of each path in the graph in real time. The flow bottleneck identification module analyzes the residence time gradient of materials in each buffer zone to identify potential queue explosion risks and issues early warning commands to the intelligent scheduling and collaborative execution terminal in advance.

[0055] The intelligent scheduling and collaborative execution terminal has optimized its instruction set for specialized logistics equipment in the dyeing and finishing workshop. In addition to driving automated guided vehicles (AGVs), this terminal directly connects to the intelligent central batching system. Based on the calculations of the multi-objective optimization decision-maker, the system can automatically weigh and transport the next batch of dyes 30 minutes in advance, achieving proactive coordination between logistics and information flow. When the on-site feedback controller detects that the AGV is slipping due to wet ground, it automatically reduces the output torque of the drive motor, using a smooth start-stop strategy to ensure stable transport of material containers.

[0056] In this embodiment, the quality status collaborative diagnosis module plays a core role in end-to-end traceability. When the finished product inspection process discovers that a batch of knitted fabrics has damaged tensile strength, the system quickly identifies an abnormal mechanical tension fluctuation in the fifth drying oven zone of the fourth setting machine by using the tensile stress history curve recorded by the full lifecycle digital twin engine. This second-level traceability capability allows factory managers to shut down faulty equipment in the shortest possible time, preventing larger-scale quality incidents.

[0057] Example 3 This embodiment further illustrates the implementation plan of the system in the finished product processing stage of knitted garments. Unlike the bulk material flow in the dyeing and finishing stage, the finished product processing stage involves extremely complex tracking of small cut pieces and semi-finished products. The quantity of materials increases by two dimensions, and the flow frequency also increases significantly.

[0058] At the perception level, the structured feature visual capture unit of the multimodal environment adaptive perception module is integrated into the workstation of each smart sewing machine. When the sewing operator stitches two pieces of fabric together, the unit automatically identifies the texture features of the edges of the pieces and compares them with the original cutting plan. If the operator mistakenly takes pieces from a different batch, the multimodal environment adaptive perception module will issue an alert through the audible and visual alarm device integrated into the environmental compensation factor calculation unit, thus preventing errors in pattern and size from occurring at the source.

[0059] In this embodiment, the distributed heterogeneous data fusion gateway faces extremely high concurrency data injection pressure. The edge preprocessing layer adopts a more advanced streaming computing architecture, completing the parsing of the fabric piece identity within milliseconds. The collaborative aggregation layer achieves low-latency aggregation of data from thousands of workstations through a lightweight message queue protocol. The data consistency verification module uses Markov logic networks to perform real-time verification of complex workstation flow logic, ensuring that the digital record and physical location of each semi-finished product after dozens of processes such as garment sewing, buttonhole stitching, and button attaching are completely consistent.

[0060] The full lifecycle digital twin engine primarily focuses on the cumulative effects of material processing at this stage. Dynamic attribute tensor construction units record the workstations, operator IDs, and corresponding sewing times for each garment. This data forms the garment's electronic archive, providing absolutely objective data for subsequent piece-rate wage settlement and production efficiency evaluation. The historical evolution database employs distributed storage technology, supporting rapid indexing and multi-dimensional cross-queries of tens of millions of records.

[0061] The spatiotemporal graph correlation analysis unit has been deeply optimized for the overhead conveyor system in the garment workshop. The path conflict detection module predicts the density of vehicles on the overhead conveyor lines in real time. When a severe congestion is predicted on the main line within the next 2 minutes, the dynamic topology construction module automatically calculates an alternative branch line and drives the intelligent scheduling and collaborative execution terminal to automatically switch the switches. This predictive scheduling based on graph convolutional networks has improved the average operating speed of the overhead conveyor system by more than 22%.

[0062] The intelligent scheduling and collaborative execution terminal acts as an intelligent production commander in the garment workshop. The multi-objective optimization decision-maker continuously balances the load of each production line. When a production line experiences a decrease in capacity due to employee leave, the system automatically reallocates material flow, distributing subsequent tasks to other workstations with surplus capacity. The self-learning module continuously collects the actual output cycle time of each workstation and constantly adjusts the standard time parameters for different processes, ensuring that the scheduling plan achievement rate remains above 98%.

[0063] The system's user interface layer provides workshop team leaders with a real-time, transparent management perspective via handheld terminals. When a team leader inspects a workstation, the handheld terminal automatically obtains the material information for that workstation using Bluetooth Low Energy positioning technology and displays a digital twin image of the garments in that batch. Team leaders can clearly view current production progress comparisons, pass rate statistics, and risk warnings regarding estimated delivery dates.

[0064] In summary, the digital tracking system for knitted fabric processing materials disclosed in this embodiment completely solves the long-standing technical bottlenecks in the knitting processing field, such as difficulty in identification, interrupted tracking, and delayed scheduling, through deep fusion of multimodal perception, efficient aggregation of distributed gateways, accurate modeling by a digital twin engine, and intelligent analysis of spatiotemporal maps. In practical applications, the system not only significantly improves material turnover efficiency but also provides solid technical support for the digital transformation of knitting enterprises through full-cycle quality traceability and self-evolving scheduling logic. The successful implementation of this system marks a significant leap in knitted fabric processing material management from traditional experience-driven to data-driven approaches.

[0065] The collaborative interaction between the various units in the system forms an organic whole. From the multimodal environment adaptive sensing module's ultimate capture of physical characteristics, to the distributed heterogeneous data fusion gateway's standardization and refinement of multi-source data, to the full lifecycle digital twin engine's digital cloning of physical entities, and the spatiotemporal mapping correlation analysis unit's deep insight into flow patterns, the intelligent scheduling and collaborative execution terminal ultimately completes precise intervention in the production process. This end-to-end technological innovation ensures that materials remain in a controlled, perceptible, and predictable state in the complex knitting processing environment.

[0066] Furthermore, the embodiments of this invention fully consider the engineering feasibility and fault tolerance of the system. Environmental compensation mechanisms combat harsh physical environments, virtual sensing synchronizers supplement missing sensor data, logical network verification corrects data deviations, and reinforcement learning algorithms continuously evolve the scheduling strategy. These design details collectively ensure the system's high robustness in practical industrial applications.

[0067] The application of this system is not limited to the knitting industry; its core architecture and algorithm logic can also be extended to other discrete manufacturing industries with similarly complex processing environments and high-precision material tracking requirements. For example, in the fields of papermaking, dyeing and printing, or high-performance composite material processing, the multimodal perception and digital twin linkage framework proposed in this invention also has extremely high application value. Through the disclosure of this embodiment, those skilled in the art can clearly understand the technical concept of this invention and, based on this, achieve digital and intelligent control of the entire knitted fabric processing process, significantly improving the industry's production efficiency and quality control level.

Claims

1. A digital tracking system for knitted fabric processing materials, characterized in that, include: The multimodal environment adaptive sensing module is used to acquire material identity information and physical state data in real time in the knitted fabric processing environment by integrating physical carrier recognition and bio-inspired fabric feature extraction technology. The distributed heterogeneous data fusion gateway is used to clean and normalize unstructured data collected by the multimodal environment adaptive sensing module, and to realize cross-device and cross-process data aggregation based on the preset communication protocol, establish a unified data semantic dictionary, and transform heterogeneous equipment data into material status descriptors that conform to the standard. The full lifecycle digital twin engine is used to receive real-time data transmitted by a distributed heterogeneous data fusion gateway, build a digital image that corresponds one-to-one with the physical material, and maintain the multi-dimensional spatiotemporal attribute tensor of the material at different processing stages. The spatiotemporal graph correlation analysis unit is used to construct a dynamic spatiotemporal topology graph of the entire production process based on the material status maintained by the full life cycle digital twin engine, identify bottleneck nodes and abnormal fluctuations in the material flow path, and use a deep learning framework based on graph convolutional networks to extract the spatiotemporal features of the material flow path in order to predict the time when the material arrives at the next process. The intelligent scheduling and collaborative execution terminal is used to generate material flow optimization instructions based on the output results of the spatiotemporal graph correlation analysis unit, with the constraints of the shortest production cycle, the most balanced equipment load, and the least energy consumption, and to drive the logistics execution mechanism to complete the distribution of materials and the flow of workstations. The multimodal environment adaptive sensing module includes a weather-resistant microelectronic identification unit, a structured feature visual capture unit, and a multispectral fabric fingerprint extraction unit. The multimodal environment adaptive sensing module determines the material identity by weighted fusion of radio frequency identification information, texture structure features, and multispectral fingerprint encoding, and determines the material identity based on the physical properties of the fabric itself when the physical tag is damaged.

2. The digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, The weather-resistant microelectronic identification unit includes a radio frequency identification (RFID) chip with acid and alkali corrosion resistant encapsulation. The RFID chip is embedded in the seam edge of the knitted fabric or in the carrier structure to achieve contactless reading of basic identity. The structured feature visual capture unit includes an industrial imaging component deployed above the workstation to acquire images of the surface texture and structure of the knitted fabric in real time during transport. The multispectral fabric fingerprint extraction unit uses a light source of a specific wavelength to illuminate the fabric surface, extracts the spectral features formed by differences in fiber composition and dye distribution, and converts them into material fingerprint codes.

3. The digital tracking system for knitted fabric processing materials according to claim 2, characterized in that, The multimodal environment adaptive sensing module also includes an environment compensation factor calculation unit; the environment compensation factor calculation unit is used to monitor the temperature, humidity and light intensity of the processing site, and dynamically compensate the imaging parameters of the structured feature visual capture unit according to the real-time environmental parameters; in an environment where the humidity exceeds 85% and there is water vapor interference, the environment compensation factor calculation unit increases the image contrast and enables infrared supplementary light to maintain the stability of image feature extraction.

4. The digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, The distributed heterogeneous data fusion gateway includes an edge preprocessing layer and a collaborative aggregation layer; the edge preprocessing layer is deployed near the processing unit in the production workshop and is responsible for performing on-site filtering and feature dimensionality reduction on the sensed data; The collaborative aggregation layer connects each edge preprocessing layer through a deterministic industrial network and uses a time-sensitive network protocol to control the transmission latency of material location update data to within 10 milliseconds; the data semantic dictionary maps processing equipment data from different manufacturers and with different communication protocols to components in the material state tensor, eliminating information semantic barriers between processes.

5. A digital tracking system for knitted fabric processing materials according to claim 4, characterized in that, The distributed heterogeneous data fusion gateway also includes a data consistency verification module. The data consistency verification module compares the material flow records of adjacent workstations and uses Markov logic networks to perform state consistency reasoning. The state consistency reasoning process is as follows: based on the state of the material at a specific moment, its physical spatial logical location, preset logical constraint weights, and constraint functions, the probability distribution of material state transition is calculated. When it is determined that the output time of the material in the later process is later than its input time in the previous process, an abnormal alarm is triggered and the traceability logic is started. Redundant data from the preceding and following processes are used to correct the data deviation caused by sensor misreading.

6. The digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, The full lifecycle digital twin engine includes a dynamic attribute tensor construction unit, a physical simulation mapping unit, and a historical evolution database. The dynamic attribute tensor construction unit establishes a multi-dimensional vector for materials entering the production system. The multi-dimensional vector includes the material's raw material composition, batch number, current process, historical processing temperature, tensile stress, and real-time moisture content parameters. The physical simulation mapping unit uses a preset knitted fabric mechanical model combined with real-time collected processing parameters to simulate the physical deformation and quality evolution of materials in key processes online. The historical evolution database stores the entire process data of materials from raw yarn to finished product.

7. A digital tracking system for knitted fabric processing materials according to claim 6, characterized in that, The full lifecycle digital twin engine also includes a virtual sensing synchronizer; the virtual sensing synchronizer is used to predictively fill the material status of offline nodes based on the production plan scheduling and the turnover rate of materials in the same batch when physical sensors are offline, using a Gaussian process regression algorithm; the Gaussian process regression algorithm generates estimated values ​​of material position and status by calculating the covariance between known spatiotemporal points and unknown points, so as to maintain the continuity of the digital twin entity.

8. The digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, The spatiotemporal graph association analysis unit includes a dynamic topology construction module, a flow bottleneck identification module, and a path conflict detection module. The dynamic topology construction module defines workshop workstations as nodes and material flow paths as edges, constructing a dynamic production network graph in real time. The flow bottleneck identification module identifies redundant workstations that cause production backlogs by calculating the flow density and residence time distribution of each node in the graph. The path conflict detection module calculates the time window overlap rate of different material batches on shared equipment based on the predicted material flow direction, providing early warning of logistics collisions and the risk of work stoppages due to material shortages. The deep learning framework based on graph convolutional networks extracts the spatiotemporal features of material flow paths by hierarchically aggregating node feature vectors, weight matrices, and node neighborhood information.

9. A digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, The intelligent scheduling and collaborative execution terminal includes a multi-objective optimization decision-maker, a task distribution interface, a field feedback controller, and an autonomous learning module. The multi-objective optimization decision-maker solves the scheduling scheme based on the early warning information from the spatiotemporal graph correlation analysis unit. The task distribution interface converts the scheduling decision into recognizable instructions for the execution mechanism, driving the automated guided vehicle, industrial robot, and intelligent suspension system. The field feedback controller monitors the execution status of the scheduling instructions in real time and initiates replanning logic when the execution deviation exceeds a 5% threshold. The autonomous learning module uses a reinforcement learning algorithm to use the scheduling result as a reward signal or a penalty signal to optimize the parameter configuration of the scheduling strategy.

10. A digital tracking system for knitted fabric processing materials according to claim 1, characterized in that, Also includes: The quality status collaborative diagnosis module is used to associate the digital tracking information of materials with the quality data of the online detection system; When a quality deviation is detected in the material, the system automatically traces the real-time temperature curve and chemical concentration records of the material during the processing steps to locate the source of the fault and isolate the affected related materials. The user interaction layer uses augmented reality technology to overlay the digital twin information of the material onto the physical entity. Operators can scan the material through the terminal and view the process requirements, remaining processing time, and directional guidance to the next workstation in real time within their field of view. The distributed computing platform is built on a private cloud and uses containerization technology to achieve independent deployment of each functional module. It is also equipped with a dual-machine hot standby mechanism. In the event of a hardware failure on the main server, the backup server takes over all tracking operations within 2 seconds.

Images