Intelligent linen management system and method based on Internet of Things and RFID technology

Abstract

The invention discloses an intelligent linen management system and method based on the Internet of Things and the RFID technology, and the method comprises the following steps: S1, binding an RFID electronic tag for each piece of linen, and building a linen digital identity; s2, deploying an RFID (Radio Frequency Identification Device) read-write device; s3, constructing a linen flow path graph, and predicting the flow trend of the linen by adopting a GraphSAGE graph neural network; s4, establishing a linen life cycle database; s5, using an improved Elastic Net sparse regression model to predict the residual life of the linen; s6, the linen type is recognized, and a linkage control system automatically matches corresponding washing program parameters; and S7, writing the linen related information into a block chain account book. The system integrates the Internet of Things, RFID and an intelligent algorithm, realizes automatic identification, scheduling and tracing of the whole process of the linen, and has the advantages of accurate identification, intelligent scheduling and traceable management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management, and particularly to a smart management system and method for linen based on Internet of Things and RFID technologies. Background Art

[0002] As fabric articles frequently used in scenarios such as hotels, hospitals, factories, high-speed rails, and airlines, linen has characteristics such as a wide variety of types, frequent turnover, easy confusion, and fast wear and tear. Traditional linen management mainly relies on manual registration, paper labels, or barcode identification, etc., which have problems such as low efficiency, high error rate, and lagging data update, and it is difficult to meet the current large institutions' requirements for refined, digital, and intelligent management of linen. With the expansion of the management scale and the increase in the circulation links of linen, how to achieve precise tracking, status perception, life assessment, and traceable management of the entire life cycle of linen has become an urgent problem in the industry.

[0003] In recent years, radio frequency identification technology has been initially applied in linen management due to its advantages such as non-contact, batch reading, and fast identification speed. By binding RFID electronic tags to each piece of linen, fast identification and information collection of linen in key links such as inbound and outbound, distribution, recycling, and washing can be achieved. However, in practical applications, due to the physical characteristics of linen such as flexible stacking, wet adhesion, and frequent use, RFID identification still faces problems such as signal occlusion, tag collision, and identification blind spots. Especially in scenarios where linen is stacked in batches or flowing at high speed, misreading and missing reading are prone to occur, seriously affecting the identification accuracy and data integrity. In addition, most existing RFID identification solutions adopt static antenna layouts and fixed parameter configurations, lacking the adaptive processing ability for factors such as environmental changes and tag response intensity, and it is difficult to dynamically optimize the identification performance.

[0004] In terms of linen scheduling and allocation, existing systems mostly use manual experience or static rules for area allocation and logistics planning, lacking data-driven prediction modeling capabilities, and it is difficult to achieve efficient scheduling and dynamic optimization of linen among multiple regions and batches. Especially in scenarios such as hospitals, different functional regions have different requirements for the types, quantities, and cleaning grades of linen. Existing systems cannot model and deduce the flow trend of linen, resulting in frequent situations of linen backlog in some regions and linen shortage in other regions, affecting the overall operation efficiency.

[0005] In terms of the evaluation of the lifespan of linen, traditional methods often rely on empirical means such as manual inspections and regular scrapping based on time. These methods can neither reflect the actual usage of individual linen nor provide a scientific basis for lifespan prediction, resulting in overuse or premature elimination of linen, wasting resources and posing safety hazards such as cross-infection. Although some systems introduce data analysis methods for lifespan judgment, most of them use linear models, which do not fully consider the non-linear relationships among multi-dimensional features such as linen material, usage frequency, and washing times, nor do they give differential weights to the chronological order of key behaviors. There is still much room for improvement in prediction accuracy and generalization ability.

[0006] In terms of system linkage control, although existing industrial washing equipment has the function of preset programs, most of them rely on manual input of linen types or selection of washing modes, lacking an automatic linkage mechanism with the linen identification system. This leads to problems such as incorrect linen classification, inaccurate program matching, and improper setting of cleaning parameters in actual operations, affecting washing quality and equipment energy consumption control.

[0007] In addition, in terms of linen resume recording and responsibility traceability, existing management systems generally adopt a centralized database structure. Although the usage status, transfer history, and disposal records of linen can be queried, due to complex data sources and easy tampering, and the lack of a credible data protection mechanism, it is difficult to meet the high-trust and high-transparency requirements for linen traceability in medical, regulatory, etc. In particular, in sensitive scenarios such as cross-use of linen, infection control, or investigation of cleaning accidents, the lack of a verifiable full-process resume severely restricts the credibility and supervision efficiency of the system.

[0008] In summary, the existing linen management technologies generally have the following deficiencies: First, the RFID tag recognition rate is not high in complex environments, and the batch reading stability is poor; second, the scheduling and allocation lack data modeling and trend prediction capabilities, and the resource allocation efficiency is not high; third, the lifespan prediction model lacks the integrated modeling of behavioral characteristics, time sensitivity, and linen differential attributes, and the judgment is not accurate enough; fourth, the washing system does not achieve automatic identification and program linkage control, affecting the processing efficiency and standardization level; fifth, the life cycle data recording mechanism is easy to be tampered with and cannot support the traceable management of linen. Summary of the Invention

[0009] An object of the present invention is to propose a smart management method for linen based on the Internet of Things and RFID technology. The present invention integrates the Internet of Things and RFID sensing technology, combines the GraphSAGE graph neural network with an improved Elastic Net sparse regression model, realizes high-precision identification of linen in complex environments, intelligent scheduling of cross-regional transfer, and proactive elimination management based on lifespan prediction. At the same time, the blockchain is introduced to build an immutable linen resume ledger to ensure the full-process traceability of data.

[0010] According to an embodiment of the present invention, a linen intelligent management method based on the Internet of Things and RFID technology includes the following steps:

[0011] S1. Bind an RFID electronic tag with a unique identifier to each piece of linen, and associate the electronic tag data with the identity data of the linen to establish a digital identity for the linen;

[0012] S2. Deploy RFID reading and writing equipment in each circulation link of linens, poll and enable different antenna combinations to read electronic tag data, and build a tag response priority sorting mechanism, upload the reading result data to the background management system in real time, and complete the collection and recording of linen status;

[0013] S3. Construct a linen flow path diagram, use the GraphSAGE graph neural network to generate an embedded representation of each node, predict the flow trend of linens between different areas, and generate a dynamic linen scheduling plan;

[0014] S4. Count the number of times the linens are used and washed based on the data read from the linen electronic tags, and establish a linen life cycle database;

[0015] S5. Construct an input feature vector based on the linen life cycle database, input it into the improved Elastic Net sparse regression model, predict the remaining life of the linen, and mark the linen as to be scrapped when the remaining life is lower than the set threshold;

[0016] S6, read the RFID electronic tag data, identify the type of linen, and automatically match the corresponding washing program parameters through the linkage control system;

[0017] S7. Write the identity information, timestamp and status information of the linens at each circulation node into the blockchain ledger to form an unalterable linen life cycle history and achieve traceable management of the entire process.

[0018] Optionally, the circulation links include distribution, use, recycling, washing and storage links.

[0019] Optionally, the S2 specifically includes:

[0020] S21. Deploy RFID reading and writing equipment at each circulation link of the linen, wherein the RFID reading and writing equipment includes a UHF frequency band reader and an array antenna module;

[0021] S22, setting the working parameters of each circulation link, wherein the working parameters include transmission power, reading interval, tag anti-collision protocol type and antenna port enabling strategy, and writing the parameters through the system configuration interface;

[0022] S23. Define a dynamic scheduling period based on the regional recognition density and the passing rate of the linen, and perform cyclic polling on the antenna combination {A1, A2,..., A n};

[0023] S24. During each scheduling period, collect the RFID electronic tag data response data, extract the unique identifier of the electronic tag data, the signal strength value, the reading timestamp, and the antenna port number, and record them as a response event quadruple;

[0024] S25. Count the response times of each electronic tag data in the current and historical scheduling periods, estimate the relative distance between the antenna and the tag, and calculate the tag sorting score;

[0025] S26. Sort the priorities of all the read electronic tag data according to the tag sorting score, filter the duplicate electronic tag data, and output the deduplicated tag reading result using the time-series deduplication strategy;

[0026] S27. Upload the final reading result data to the background management system through the Internet of Things to complete the real-time collection and recording of the linen status.

[0027] Optionally, the reading result data includes electronic tag data, reading area code, reading time, antenna number, and tag status code.

[0028] Optionally, the specific steps of S3 are as follows:

[0029] S31. Based on the transfer records of the linen in each area, construct a linen transfer path graph G=(V, E), where V={v1, v2,..., v n} represents the linen management link nodes, and E={e ij} represents the directed edge of the linen from node v i to node v j , and the edge weight is represented by the number of linen transfers per unit time;

[0030] S32. Extract the node feature vector x i for each node v i . The node features include the cumulative quantity, average residence time, historical active times, and regional attribute code of the linen at node v i . Based on the node features, construct an initial node representation matrix X=[x1, x2,..., x n T , where x i represents the node feature vector of the i-th node, and T represents the transpose operation;

[0031] S33. Use the GraphSAGE graph neural network to perform node embedding learning on the linen transfer path graph G, and initialize the representation of each node as​ Perform aggregation update at the k-th layer:

[0032]

[0033] where, represents the embedding representation of node i in the k-th layer, σ represents the non-linear activation function, W (k) represents the weight matrix of the k-th layer, AGG (k) represents the aggregation function of the k-th layer, represents the embedding representation of node j in the (k - 1)-th layer, N(i) represents the set of adjacent nodes of node i, ∪ represents the "union" operation of sets, represents the embedding representation of node i in the (k - 1)-th layer;

[0034] S34. Repeatedly perform K-layer aggregation to obtain the final embedding representation of each node to form the node embedding representation matrix Z = [z1, z2,..., z n T , where z n represents the embedding representation of node i;

[0035] S35. Based on the node embedding representation matrix Z, calculate the trend score flowing between any two nodes:

[0036]

[0037] where, s i→j represents the trend score of the linen flowing from node i to node j, T represents the transpose operation, z j represents the embedding representation of node j, z i represents the embedding representation of node i;

[0038] S36. Construct a linen trend scoring graph according to all the trend scores, generate a trend score vector for each node, and combine the current distribution state and capacity constraints of the linen in each area to output a linen dynamic scheduling scheme for indicating the optimal transfer path of the linen.

[0039] Optionally, the S4 specifically includes:

[0040] S41. Receive the linen electronic tag data uploaded from each transfer link from the background management system, and extract the data records including the tag unique identifier, the reading timestamp, and the reading location code;

[0041] S42. Group the data records by the tag unique identifier and sort them in ascending order according to the timestamp to form a time series event for each piece of linen;

[0042] ​S43. Set the coding sets representing the usage locations of the linen and the coding sets representing the washing locations of the linen, and determine the set to which each record l belongs in the time series event;

[0043] S44. For each tag unique identifier, count the number of records in the event stream that belong to the coding set of the linen usage location as the usage times, and count the number of records that belong to the coding set of the linen washing location as the washing times;

[0044] S45. Construct linen life cycle data entries based on the tag unique identifier, usage times, washing times, and material type coding vector, write all linen life cycle data entries into the linen life cycle database, use the tag unique identifier as the primary key, establish a life cycle data table, and realize the continuous update and unified management of the linen life cycle information.

[0045] Optionally, the specific steps of S5 are as follows:

[0046] S51. Obtain the usage times, washing times, and material type coding of the linen from the life cycle database, and obtain the corresponding usage time series and washing time series;

[0047] S52. Calculate the decay usage times and decay washing times based on the time decay mechanism:

[0048]

[0049] Wherein, represents the decay usage times of linen i, represents the decay washing times of linen i, decay represents decay, u i represents the usage times of linen i, w i represents the washing times of linen i, T represents, represents the kth usage time, represents the kth washing time, δ represents the time decay coefficient, P represents the current time step;

[0050] S53. Construct the feature vector of the linen, and the feature vector includes the original times, square terms, cross terms, logarithmic transformation terms, time decay terms, and material type coding;

[0051] S54. Construct a loss function with a dynamic adjustment term, and train an improved Elastic Net sparse regression model. The improvement includes introducing a non-linear feature transformation and a time decay mechanism, and adopting a grouped network and introducing a dynamic regularization parameter mechanism in the loss function:

[0052]

[0053] Wherein, represents the loss function, m represents the number of linen samples, and y i represents the true life label of the linen, represents the predicted remaining life of the linen, λ1(m i ) represents the group regularization coefficient dynamically adjusted according to the material type code m i m i represents the material type code, G represents the total number of material groups, ||β g ||2 represents the regression parameter vector belonging to the g-th group, λ2(u i , w i ) represents the L1 regularization coefficient dynamically adjusted based on the usage times and washing times, u i represents the usage times of linen i, w i represents the washing times of linen i, ||β||1 represents the L1 norm of all regression coefficients;

[0054] S55. Input the feature vector into the improved Elastic Net sparse regression model to predict the remaining life of the linen:

[0055]

[0056] Among them, represents the predicted remaining life of the linen, β0 represents the bias term of the sparse regression model, β j represents the regression coefficient corresponding to the j-th feature vector, x ij represents the j-th feature vector of the linen, and d represents the total number of dimensions of the feature vector;

[0057] S56. Set the life threshold θ. If is satisfied, then mark the linen as the to-be-scrapped state and update the status field in the linen management system.

[0058] Optionally, the specific content of S6 includes:

[0059] S61. Before the linen enters the industrial washing process, read the electronic tag data of each piece of linen through the RFID reading and writing device set at the linen entrance to obtain the data record containing the unique identification information and type information;

[0060] S62. Send the read information to the linen type recognition module to determine the standard classification type to which the linen belongs according to the preset linen type mapping rule;

[0061] S63. Input the recognized linen type into the washing parameter configuration system to retrieve the corresponding washing program template of this type of linen in the preset program library. The washing program template includes the disinfection time, main washing time, washing temperature, dehydration speed, and detergent dosage;

[0062] S64. Automatically transfer the retrieved washing program template parameters to the control unit of the industrial washing equipment to achieve program linkage with the control system;

[0063] S65. If the system detects that the current linen to be washed contains different types, select the main type according to the proportion of the linen quantity, and generate a comprehensive washing program parameter set according to the preset parameter merging rule to cover the washing requirements of the mixed-type linen;

[0064] S66. Automatically load the finally determined washing program parameters to the execution module of the washing equipment to complete the automatic matching of the linen and the washing program, and at the same time upload the unique identification information of the linen, the identified linen type, the matching washing program number and the processing time to the background management system.

[0065] Optionally, the S7 specifically includes:

[0066] S71. When the linen completes any link of distribution, use, recycling, washing or storage, collect the unique identity identification, the current event type, the event occurrence time, the event location code, the on-site status information and the operator number corresponding to the linen to form complete linen event data;

[0067] S72. Package each piece of linen event data into a structured data format, and the structured data format includes an identity field, an event field, a status field and a time field to form a standard transaction data record;

[0068] S73. Perform processing operations on the standard transaction data record, and the processing operations include data integrity verification, time legality verification and uniqueness confirmation, and submit it to the blockchain system after completion;

[0069] S74. Call the data on-chain interface through the smart contract module of the blockchain system to perform identity verification, format verification and consensus mechanism confirmation on the standard transaction data record;

[0070] S75. Write the verified linen event data into the blockchain ledger system, record the event data, the timestamp and the previous block digest information in the current block to form a blockchain data structure;

[0071] S76. Generate an independent life cycle linked list for each piece of linen. The system links all the historical event blocks related to the linen in chronological order to form a tracking path, and establish a query index in the management system;

[0072] S77. Provide a traceability query interface based on the identity identification, event type and time range, support users to view the status records of any linen at each node in the whole life cycle, and realize the traceable and immutable management of the whole process of linen transfer.

[0073] A smart management system for linen based on Internet of Things and RFID technology according to an embodiment of the present invention includes:

[0074] A linen identity binding module, configured to bind an RFID electronic tag with a unique identifier to each piece of linen, establish an association with the linen identity data, and generate a linen digital identity;

[0075] A status acquisition module, configured to deploy RFID reading and writing devices in the processes of linen distribution, use, recycling, washing, and storage, read the electronic tag data and upload it to the background management system, and complete the real-time acquisition and recording of the linen status;

[0076] A transfer modeling and scheduling module, configured to construct a linen transfer path diagram, generate node embedding representations using a graph neural network algorithm, predict the flow trend of linen between regions, and generate a scheduling plan;

[0077] A life cycle statistics module, configured to count the usage times and washing times of linen, and establish a linen life cycle database;

[0078] A life prediction module, configured to construct a feature vector based on the usage times, washing times, and material attributes of linen, input it into an improved Elastic Net sparse regression model, predict the remaining life of linen, and determine whether the scrapping standard is reached;

[0079] A washing linkage module, configured to read the RFID tag information before the linen is put into washing, identify the linen type, and automatically match the washing program parameters to achieve linkage control;

[0080] A blockchain management module, configured to record the identity information, status information, and timestamps in the linen transfer process into the blockchain ledger, construct an immutable linen life cycle resume, and provide a traceability query service.

[0081] The beneficial effects of the present invention are:

[0082] First of all, the present invention binds an RFID electronic tag with a unique identifier to each piece of linen, and deploys RFID reading and writing devices supporting multi-antenna polling and tag priority sorting mechanisms in each transfer link, realizing high-stability identification of linen in complex environments such as stacking, humidity, and high interference, improving the accuracy and reading efficiency of batch identification. Compared with the traditional fixed antenna reading method, the present invention adopts a dynamic scheduling and temporal de-duplication strategy, which can effectively reduce the incidence of missed reading and duplicate reading, and ensure the real-time and integrity of linen status data.

[0083] Secondly, in terms of linen scheduling, the present invention introduces a method for modeling the transfer path and calculating node embeddings based on the GraphSAGE graph neural network, which can fully explore the flow patterns and trend rules of linen between different regions, automatically generate an optimized dynamic scheduling plan, thereby realizing the intelligent allocation and regulation of linen among multiple regions and multiple links. By aggregating node features and outputting trend scores, the rationality of the linen circulation path is effectively improved, the backlog or shortage of linen in local areas is alleviated, and the collaborative efficiency of overall resource scheduling is enhanced.

[0084] In addition, in terms of lifespan management, the present invention constructs an input vector by integrating features such as the usage frequency, washing times, and material type of linen, and introduces a time decay factor and non-linear feature transformation to construct an improved Elastic Net sparse regression model for lifespan prediction. The model simultaneously introduces grouped regularization and dynamic regularization coefficients, enhancing the adaptability to material category differences and frequent usage behaviors, achieving accurate prediction of the remaining lifespan of linen, and automatically marking the scrapped state according to the set threshold, thereby realizing proactive warning and scientific elimination, and avoiding resource waste and safety risks.

[0085] Finally, in terms of data management and traceability, the present invention introduces blockchain technology to construct an immutable data ledger for the entire lifecycle of linen, taking the identity information, event time, status records, etc. of linen at each transfer node as transaction data to be uploaded to the chain, forming a complete lifecycle resume, and providing a traceability query interface based on multi-dimensional conditions such as identity, time, and event type. This mechanism enhances the data credibility and transparency of the system, providing verifiable technical support for supervision and auditing, accident traceability, quality accountability, etc. in the process of linen management, and is particularly applicable to scenarios with high requirements for linen management such as medical, high-speed rail, and military, which require high credibility and traceability.

[0086] Generally speaking, through the collaborative innovation of the perception layer, analysis layer, decision layer, and evidence storage layer, the present invention constructs a closed-loop linen management system from data collection, intelligent analysis, automatic decision-making to trusted storage, realizing accurate identification, dynamic scheduling, lifespan prediction, program linkage, and traceable management of linen throughout the lifecycle, improving the operation efficiency, management accuracy, and usage safety of linen, and having good promotion value and application prospects. Description of the Drawings

[0087] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0088] Figure 1 is a flowchart of a linen intelligent management method based on the Internet of Things and RFID technology proposed by the present invention;

[0089] Figure 2 A schematic diagram of a GraphSAGE graph neural network for a linen intelligent management method based on the Internet of Things and RFID technology proposed in the present invention;

[0090] Figure 3 This is a structural block diagram of the improved Elastic Net sparse regression model prediction of the linen intelligent management method based on the Internet of Things and RFID technology proposed by the present invention. DETAILED DESCRIPTION

[0091] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0092] refer to Figures 1-3 , a linen intelligent management method based on the Internet of Things and RFID technology, comprising the following steps:

[0093] S1. Bind an RFID electronic tag with a unique identifier to each piece of linen, and associate the electronic tag data with the identity data of the linen to establish a digital identity for the linen;

[0094] S2. Deploy RFID reading and writing equipment in each circulation link of linens, poll and enable different antenna combinations to read electronic tag data, and build a tag response priority sorting mechanism, upload the reading result data to the background management system in real time, and complete the collection and recording of linen status;

[0095] S3. Construct a linen flow path diagram, use the GraphSAGE graph neural network to generate an embedded representation of each node, predict the flow trend of linens between different areas, and generate a dynamic linen scheduling plan;

[0096] S4. Count the number of times the linens are used and washed based on the data read from the linen electronic tags, and establish a linen life cycle database;

[0097] S5. Construct an input feature vector based on the linen life cycle database, input it into the improved Elastic Net sparse regression model, predict the remaining life of the linen, and mark the linen as to be scrapped when the remaining life is lower than the set threshold;

[0098] S6, read the RFID electronic tag data, identify the type of linen, and automatically match the corresponding washing program parameters through the linkage control system;

[0099] S7. Write the identity information, timestamp and status information of the linens at each circulation node into the blockchain ledger to form an unalterable linen life cycle history and achieve traceable management of the entire process.

[0100] The present invention realizes the automation and intelligence of the whole process from identification, scheduling to elimination and resume by constructing an intelligent management process for linen that covers RFID perception, graph neural network scheduling, life prediction and blockchain traceability, improving the management efficiency, accuracy and traceability of linen.

[0101] In this embodiment, the transfer links include the issuing, using, recycling, washing and storing links.

[0102] The present invention makes the system management logic structured and closed-loop, which helps to comprehensively cover the key state nodes of the linen life cycle, ensuring the integrity of data collection and the continuity of behavior monitoring.

[0103] In this embodiment, S2 specifically includes:

[0104] S21. Deploy RFID reading and writing devices at each transfer link of the linen. The RFID reading and writing devices include UHF band readers and array antenna modules;

[0105] S22. Set the working parameters of each transfer link. The working parameters include transmission power, reading interval, tag anti-collision protocol type and antenna port enabling strategy, and write the parameters through the system configuration interface;

[0106] S23. Define a dynamic scheduling period based on the regional recognition density and the linen passing rate, and perform cyclic polling on the antenna combination {A1, A2,, A n};

[0107] S24. During each scheduling period, collect the RFID electronic tag data response data, extract the unique identifier of the electronic tag data, signal strength value, reading timestamp and antenna port number, and record them as a response event quadruple;

[0108] S25. Count the response times of each electronic tag data in the current and historical scheduling periods, estimate the relative distance between the antenna and the tag, and calculate the tag sorting score;

[0109] S26. Sort the priority of all the read electronic tag data according to the tag sorting score, filter the duplicate electronic tag data, and output the deduplicated tag reading result by using the time series deduplication strategy;

[0110] S27. Upload the final read result data to the background management system through the Internet of Things to complete the real-time collection and recording of the linen status.

[0111] The present invention adopts a dynamic scheduling multi-antenna array mechanism and a priority sorting algorithm, effectively improving the RFID recognition accuracy and stability of batch linen in a high-interference or stacked environment, reducing the phenomenon of missed reading and misreading, and enhancing the system perception robustness.

[0112] In this embodiment, the read result data includes electronic tag data, read area code, read time, antenna number, and tag status code.

[0113] The present invention standardizes the read result data into a five-tuple including tag identification, time, location, antenna number, and status code, enhancing the structuring and usability of the collected data and providing high-quality input for subsequent flow modeling and life cycle analysis.

[0114] In this embodiment, S3 specifically includes:

[0115] S31. Based on the transfer records of the linen in each area, construct a linen transfer path graph G=(V, E), where V={v1, v2,..., v n} represents the linen management link nodes, and E={e ij} represents the directed edge of the linen from node v i to node v j , and the edge weight is represented by the number of linen transfers per unit time;

[0116] S32. Extract the node feature vector x i for each node v i . The node features include the cumulative quantity, average residence time, historical active times, and area attribute code of the linen at node v i . Based on the node features, construct an initial node representation matrix X=[x1, x2,..., x n T , where x i represents the node feature vector of the i-th node, and T represents the transpose operation;

[0117] S33. Use the GraphSAGE graph neural network to perform node embedding learning on the linen transfer path graph G. Initialize the representation of each node as Perform aggregation update at the k-th layer:

[0118]

[0119] where, represents the embedding representation of node i in the k-th layer, σ represents the non-linear activation function, W (k) represents the weight matrix in the k-th layer, AGG (k) represents the aggregation function in the k-th layer, represents the embedding representation of node j in the (k-1)-th layer, N(i) represents the set of adjacent nodes of node i, ∪ represents the "union" operation of sets, represents the embedding representation of node i in the (k-1)-th layer;

[0120] ​S34. Repeat the K-layer aggregation to obtain the final embedding representation of each node Construct the node embedding representation matrix Z = [z1, z2,..., z n T , where z n represents the embedding representation of node i;

[0121] S35. Based on the node embedding representation matrix Z, calculate the trend score flowing between any two nodes:

[0122]

[0123] where s i→j represents the trend score of the linen flowing from node i to node j, T represents the transpose operation, z j represents the embedding representation of node j, and z i represents the embedding representation of node i;

[0124] S36. Construct a linen trend scoring graph based on all trend scores, generate a trend score vector for each node, and combine the current distribution state and capacity constraints of the linen in each area to output a dynamic linen scheduling plan for indicating the optimal transfer path of the linen.

[0125] The present invention realizes intelligent flow trend prediction and scheduling optimization, and improves the linen transfer efficiency and regional balance by introducing the GraphSAGE graph neural network to perform node representation learning on the linen transfer path and mining the linen migration law between regions.

[0126] In this embodiment, the S4 specifically includes:

[0127] S41. Receive the linen electronic tag data uploaded by each transfer link from the background management system, and extract the data records including the tag unique identifier, the reading timestamp, and the reading location code;

[0128] S42. Group the data records according to the tag unique identifier and sort them in ascending order according to the timestamp to form a time series event for each piece of linen;

[0129] S43. Set the coding set representing the linen usage location and the coding set representing the linen washing location, and determine the set to which each record l belongs in the time series event;

[0130] S44. For each tag unique identifier, count the number of records in the event stream that satisfy each record belonging to the coding set of the linen usage location as the usage times, and count the number of records belonging to the coding set of the linen washing location as the washing times;

[0131] ​S45. Construct linen life cycle data entries based on the label unique identifier, usage times, washing times, and material type coding vectors, write all linen life cycle data entries into the linen life cycle database, use the label unique identifier as the primary key, establish a life cycle data table, and achieve continuous update and unified management of linen life cycle information.

[0132] Through the construction of the linen life cycle database, the present invention realizes the continuous archival management of usage times, washing times, and time series events, provides data support for life prediction and management strategies, and improves the visibility and quantifiable ability of linen management.

[0133] In this embodiment, the S5 specifically includes:

[0134] S51. Obtain the usage times, washing times, and material type coding of the linen from the life cycle database, and obtain the corresponding usage time series and washing time series;

[0135] S52. Calculate the decay usage times and decay washing times based on the time decay mechanism:

[0136]

[0137] Among them, represents the decay usage times of linen i, represents the decay washing times of linen i, decay represents decay, u i represents the usage times of linen i, w i represents the washing times of linen i, T represents, represents the kth usage time, represents the kth washing time, δ represents the time decay coefficient, P represents the current time step;

[0138] S53. Construct the feature vector of the linen, and the feature vector includes the original times, square terms, cross terms, logarithmic transformation terms, time decay terms, and material type coding;

[0139] S54. Construct a loss function with a dynamic adjustment term, train an improved Elastic Net sparse regression model, and the improvement includes introducing a non-linear feature transformation and a time decay mechanism, and adopting a grouped network and introducing a dynamic regularization parameter mechanism in the loss function:

[0140]

[0141] Among them, represents the loss function, m represents the number of linen samples, y i represents the true life label of the linen, represents the predicted remaining life of the linen, λ1(m i) represents the encoding m according to the material type i The dynamically adjusted group regularization coefficient, m i represents the material type encoding, G represents the total number of material groups, ||β g ||2 represents the regression parameter vector belonging to the g-th group, λ2(u i , w i ) represents the L1 regularization coefficient dynamically adjusted based on the usage times and washing times, u i represents the usage times of the linen i, w i represents the washing times of the linen i, ||β||1 represents the L1 norm of all regression coefficients;

[0142] S55. Input the feature vector into the improved Elastic Net sparse regression model to predict the remaining life of the linen:

[0143]

[0144] Among them, represents the predicted remaining life of the linen, β0 represents the bias term of the sparse regression model, β j represents the regression coefficient corresponding to the j-th feature vector, x ij represents the j-th feature vector of the linen, d represents the total number of dimensions of the feature vector;

[0145] S56. Set the life threshold θ, if it satisfies then mark the linen as the to-be-scrapped state and update the status field in the linen management system.

[0146] The present invention introduces an improved Elastic Net sparse regression model, combines time decay, non-linear feature construction and dynamic regularization mechanism to realize personalized and accurate prediction of the remaining life of linen, and effectively improves the scientificity and forward-looking of the scrapping strategy.

[0147] In this embodiment, the S6 specifically includes:

[0148] S61. Before the linen enters the industrial washing process, read the electronic tag data of each piece of linen through the RFID reading and writing device set at the linen entrance to obtain the data record containing the unique identification information and type information;

[0149] S62. Send the read information to the linen type recognition module, and determine the standard classification type to which the linen belongs according to the preset linen type mapping rule;

[0150] S63. Input the identified linen type into the washing parameter configuration system, and retrieve the corresponding washing program template for this type of linen in the preset program library. The washing program template includes disinfection time, main washing time, washing temperature, dehydration speed, and detergent dosage.

[0151] S64. Automatically transfer the parameters of the retrieved washing program template to the control unit of the industrial washing equipment to achieve program linkage with the control system.

[0152] S65. If the system detects that the current linen to be washed contains different types, select the main type according to the proportion of linen quantity, and generate a comprehensive washing program parameter set according to the preset parameter merging rules to cover the washing requirements of the mixed-type linen.

[0153] S66. Automatically load the finally determined washing program parameters to the execution module of the washing equipment to complete the automatic matching of the linen and the washing program. At the same time, upload the unique identification information of the linen, the identified linen type, the matching washing program number, and the processing time to the background management system.

[0154] Through the linkage between RFID identification and the washing control system, the present invention automatically identifies the linen type and matches the corresponding washing program, avoiding manual operation errors, improving the accuracy and automation of washing parameter settings, and ensuring cleaning quality and energy-saving control.

[0155] In this embodiment, the specific steps of S7 are as follows:

[0156] S71. When the linen completes any link of distribution, use, recycling, washing, or storage, collect the unique identity identification, current event type, event occurrence time, event location code, on-site status information, and operator number corresponding to the linen to form complete linen event data.

[0157] S72. Package each piece of linen event data into a structured data format, which includes an identity field, an event field, a status field, and a time field, to form a standard transaction data record.

[0158] S73. Perform processing operations on the standard transaction data record, including data integrity verification, time legality verification, and uniqueness confirmation. After completion, submit it to the blockchain system.

[0159] S74. Call the data on-chain interface through the intelligent contract module of the blockchain system to perform identity verification, format verification, and consensus mechanism confirmation on the standard transaction data record.

[0160] S75. Write the verified linen event data into the blockchain ledger system, record the event data, timestamp, and the previous block digest information in the current block to form a blockchain data structure.

[0161] S76. Generate an independent life cycle linked list for each piece of linen. The system links all historical event blocks related to the linen in chronological order to form a tracking path and creates a query index in the management system.

[0162] S77. Provide a traceability query interface based on identity identification, event type, and time range, supporting users to view the status records of any piece of linen at each node in its entire life cycle, and realizing the tamper-proof management of the entire process of linen circulation.

[0163] The present invention uses blockchain technology to record the status information and timestamps of linen in each link, constructs an immutable life cycle ledger, enhances the credibility and verifiability of linen data, and meets the requirements of high-standard scenarios for resume transparency and responsibility traceability.

[0164] A smart management system for linen based on Internet of Things and RFID technology, including:

[0165] A linen identity binding module, used to bind an RFID electronic tag with a unique identifier to each piece of linen, establish an association with the linen identity data, and generate a digital identity for the linen.

[0166] A status collection module, used to deploy RFID reading and writing devices in the processes of linen distribution, use, recycling, washing, and storage, read the electronic tag data and upload it to the background management system to complete the real-time collection and recording of the linen status.

[0167] A circulation modeling and scheduling module, used to construct a linen circulation path diagram, generate node embedding representations using a graph neural network algorithm, predict the flow trend of linen between regions, and generate a scheduling plan.

[0168] A life cycle statistics module, used to count the usage times and washing times of linen and establish a linen life cycle database.

[0169] A remaining life prediction module, used to construct a feature vector based on the usage times, washing times, and material properties of linen, input it into an improved Elastic Net sparse regression model, predict the remaining life of the linen, and determine whether it reaches the scrapping standard.

[0170] A washing linkage module, used to read the RFID tag information before the linen is put into washing, identify the linen type, and automatically match the washing program parameters to achieve linkage control.

[0171] A blockchain management module, used to record the identity information, status information, and timestamps in the process of linen circulation into the blockchain ledger, construct an immutable linen life cycle resume, and provide a traceability query service.

[0172] The present invention constructs a smart linen management system that integrates identity binding, status collection, intelligent scheduling, lifespan assessment, washing linkage, and blockchain traceability, improving the intelligent, automated, and controllable levels of management and meeting the efficient management requirements of large-scale and multi-region linen operation scenarios.

[0173] Example 1:

[0174] To verify the feasibility of the present invention in implementation, the present invention is applied to the linen management scenario of a washing center in a certain tertiary hospital. The hospital involves more than 30 types of linens in its daily operation, including bed sheets, pillowcases, surgical gowns, protective clothing, patient gowns, cleaning towels, etc. The daily usage frequency of linens is high, the turnover cycle is fast, and after recycling, there are generally situations of wet weight, stacking, and cross-stacking. RFID identification is severely interfered. In addition, the average lifespan of linens is difficult to quantify, and there are often management problems of overuse and premature elimination. The hospital management urgently needs an efficient, stable, intelligent, and traceable full-process linen management solution.

[0175] In this scenario, first, ultra-high-frequency RFID tags with unique IDs are embedded in all linens, and digital identity mappings are established between them and information such as linen type, material, and purchase batch through the linen identity binding module provided by the present invention. The hospital distributes the linens to 25 clinical departments for use. Each linen enters the washing center uniformly during the recycling process. To improve the identification efficiency, the hospital installs array-type RFID reading and writing devices at key nodes such as the recycling entrance, sorting line, pre-washing input port, and post-washing classification port, and cooperates with the multi-antenna dynamic polling and priority sorting algorithm proposed by the present invention to enable stable identification of linens in scenarios of stacking, humidity, and high-speed transmission.

[0176] The identified data is automatically uploaded to the linen background management system. The system predicts the next transfer trend of the linen and generates scheduling suggestions through the GraphSAGE graph neural network transfer modeling module of the present invention based on the real-time location, historical trajectory, and regional congestion situation of the linen. For example, the system can determine that a batch of recycled surgical gowns should be preferentially allocated to the spare area on the third floor of the surgical department to avoid the imbalance phenomenon of shortages in the surgical department and surpluses in other departments.

[0177] In terms of linen lifespan management, the hospital has generally used fixed washing times limits or manual visual inspection of aging for elimination in the past, which is prone to inconsistent standards. After applying the lifespan prediction module of the present invention, the system automatically collects the usage times, washing frequencies, usage durations, and materials of each linen, and constructs feature vectors by combining the time decay mechanism and non-linear feature expansion and inputs them into the Elastic Net sparse regression model to accurately predict the remaining lifespan of the linen. For example, a batch of pure cotton patient gowns is predicted by the system to have a remaining lifespan of less than 1.5 times after the 23rd washing, and is automatically marked as the status to be scrapped, triggering a warning for linen replenishment procurement, significantly reducing the deviation of manual judgment.

[0178] To verify the performance of the system in practice, core indicators such as linen management efficiency, recognition accuracy, and elimination rationality were compared before and after implementing the present invention during the 90-day operation in this hospital. During the experiment, the average daily turnover of linen managed was approximately 13,500 pieces.

[0179] Table 1 Comparison Table of Experimental Data

[0180]

[0181] In terms of recognition accuracy, the average daily linen recognition accuracy before implementing the system was 88.7%, while it increased to 98.9% after adopting the solution of the present invention, an increase of 10.2 percentage points. This improvement stems from the multi-antenna array dynamic scheduling mechanism and tag priority sorting algorithm adopted in the present invention, which can still effectively read RFID electronic tags in stacking, high-humidity, and moving scenarios, greatly reducing the phenomena of misreading and missing reading. The data further shows that the misreading rate of batch recognition decreased from the original 7.5% to 1.3%, and the missing reading rate decreased from 4.8% to 0.9%, fully demonstrating that the technological innovation of the present invention in recognition stability has significant effects.

[0182] In terms of linen scheduling, the present invention models and predicts the transfer path through the GraphSAGE graph neural network, making the cross-regional allocation of daily linen more efficient. The data shows that after implementing the present invention, the average time-consuming of linen scheduling decreased from the original 62 minutes to 21 minutes, a decrease of 66.1%. This optimization directly alleviates the problem of imbalance between the supply and demand of linen between departments. For example, the number of shortage alarms for surgical linen decreased from 27 times per month to 3 times, significantly improving the efficiency of medical resource allocation and emergency response capabilities.

[0183] In terms of linen life management, the improved Elastic Net regression model introduced in the present invention combines time decay, non-linear transformation, and group regularization, improving the scientific nature of linen life prediction. The results show that after implementation, the misjudgment rate of linen scrapping is controlled within 2.6%, and the proportion of linen overused due to manual experience or fixed washing times decreased sharply from 18.2% to 1.1%, greatly reducing the risk of cross-infection or hygiene hazards of linen, and also avoiding resource waste caused by premature scrapping.

[0184] In addition, in the washing process, the present invention automatically links the washing program matching through RFID recognition, greatly reducing the incidence of human setting errors. The number of times of incorrect washing parameter matching decreased from 51 times per month to 2 times, a reduction of 96.1%, effectively ensuring the standardization of washing quality and the processing safety of linen.

[0185] In terms of labor saving, thanks to the function substitution of automatic identification and intelligent scheduling, the average daily manual inspection time has been reduced by about 8 person-hours per hour, effectively alleviating the management pressure caused by labor shortage in linen operation and maintenance.

[0186] In terms of data traceability, the traceability rate before the system was launched was 42.3%. It mainly relied on manual records and a centralized database, with problems such as missed records, incorrect writing, and unverifiable data. After introducing blockchain, 100% of the linen transfer process has been registered on the chain. Each record has a timestamp and the property of being tamper-proof. The system has a fast query response and a complete chain of responsibility, greatly improving the regulatory transparency and compliance.

[0187] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered within the protection scope of the present invention.

Claims

1. An intelligent management method for linen based on Internet of Things and RFID technology, characterized in that, The steps include: S1. Bind an RFID electronic tag with a unique identifier to each piece of linen, and associate the electronic tag data with the identity data of the linen to establish a digital identity for the linen; S2. Deploy RFID reading and writing equipment in each circulation link of linens, poll and enable different antenna combinations to read electronic tag data, and build a tag response priority sorting mechanism, upload the reading result data to the background management system in real time, and complete the collection and recording of linen status; S3. Construct a linen flow path diagram, use the GraphSAGE graph neural network to generate an embedded representation of each node, predict the flow trend of linens between different areas, and generate a dynamic linen scheduling plan; S4. Count the number of times the linens are used and washed based on the data read from the linen electronic tags, and establish a linen life cycle database; S5. Construct an input feature vector based on the linen life cycle database, input it into the improved Elastic Net sparse regression model, predict the remaining life of the linen, and mark the linen as to be scrapped when the remaining life is lower than the set threshold; S6, read the RFID electronic tag data, identify the type of linen, and automatically match the corresponding washing program parameters through the linkage control system; S7. Write the identity information, timestamp and status information of the linens at each circulation node into the blockchain ledger to form an unalterable linen life cycle history and achieve traceable management of the entire process.

2. The intelligent management method for linen based on Internet of Things and RFID technology according to claim 1, characterized in that, The circulation links include issuance, use, recycling, washing and storage.

3. The intelligent management method for linen based on Internet of Things and RFID technology according to claim 1, wherein, The S2 specifically includes: S21. Deploy RFID reading and writing equipment at each circulation link of the linen, wherein the RFID reading and writing equipment includes a UHF frequency band reader and an array antenna module; S22, setting the working parameters of each circulation link, wherein the working parameters include transmission power, reading interval, tag anti-collision protocol type and antenna port enabling strategy, and writing the parameters through the system configuration interface; S23. Define a dynamic scheduling period based on the regional recognition density and the passing rate of linen, and perform cyclic polling on the antenna combination {A1, A2,, A n}}; S24. In each scheduling cycle, collect RFID electronic tag data response data, extract the electronic tag data unique identifier, signal strength value, read timestamp and antenna port number, and record them as a response event quadruple; S25, counting the number of responses of each electronic tag data in the current and historical scheduling cycles, estimating the relative distance between the antenna and the tag, and calculating the tag ranking score; S26, prioritizing all read electronic tag data according to the tag sorting scores, filtering duplicate electronic tag data, and using a time sequence deduplication strategy to output deduplicated tag reading results; S27. Upload the final reading result data to the background management system through the Internet of Things to complete the real-time collection and recording of the linen status.

4. A smart management method for linen based on Internet of Things and RFID technology according to claim 1, characterized in that, The reading result data includes electronic tag data, reading area code, reading time, antenna number and tag status code.

5. The intelligent management method for linen based on Internet of Things and RFID technology according to claim 1, characterized in that The S3 specifically includes: S31. Based on the transfer records of the linen in each area, construct a linen transfer path graph G=(V, E), where V={v1, v2,..., v n} represents the linen management link nodes, and E={e ij} represents the directed edge of the linen from node v i to node v j . The edge weight is represented by the number of linen transfers per unit time; S32. For each node v i Extract the node feature vector x i , where the node features include the cumulative quantity, average residence time, historical active times, and regional attribute encoding of the linen at node v i . Based on the node features, construct the initial node representation matrix X = [x1, x2,..., x n T , where x i represents the node feature vector of the i-th node, and T represents the transpose operation;​ S33. Use the GraphSAGE graph neural network to perform node embedding learning on the linen flow path graph G, and initialize the representation of each node as Perform aggregation update at the k-th layer: Among them, represents the embedding representation of node i in the k-th layer, σ represents the non-linear activation function, and W (k) represents the weight matrix of the k-th layer, and AGG (k) represents the aggregation function of the k-th layer. represents the embedding representation of node j in the (k - 1)-th layer, N(i) represents the set of adjacent nodes of node i, and ∪ represents the "union" operation of sets. represents the embedding representation of node i in the (k - 1)-th layer; S34. Repeat the K-layer aggregation to obtain the final embedding representation of each node Construct the node embedding representation matrix Z = [z1, z2,..., z n T , where z n represents the embedding representation of node i;​ S35. Based on the node embedding representation matrix Z, calculate the trend score of the flow between any two nodes: Among them, s i→j represents the trend score of the linen flowing from node i to node j, T represents the transpose operation, z j represents the embedded representation of node j, z i represents the embedded representation of node i; S36. Construct a linen trend scoring graph based on all trend scores, generate a trend score vector for each node, and combine the current distribution status and capacity constraints of linen in each area to output a dynamic scheduling plan for linen, which is used to indicate the optimal transfer path of linen.

6. The intelligent management method for linen based on Internet of Things and RFID technology according to claim 1, characterized in that, The specific steps of S4 are as follows: S41. Receive the linen electronic tag data uploaded by each transfer link from the background management system, and extract the data records containing the tag unique identifier, reading timestamp, and reading location code. S42. Group the data records by the tag unique identifier and sort them in ascending order according to the timestamp to form a time series event for each piece of linen. S43. Set the coding set representing the linen usage location and the coding set representing the linen washing location, and determine the belonging set of each record l in the time series event. S44. For each tag unique identifier, count the number of times that each record belongs to the coding set of the linen usage location in the event stream as the usage times, and count the number of times that each record belongs to the coding set of the linen washing location as the washing times. S45. Construct linen life cycle data entries based on the tag unique identifier, usage times, washing times, and material type coding vector, write all linen life cycle data entries into the linen life cycle database, use the tag unique identifier as the primary key, establish a life cycle data table, and realize the continuous update and unified management of linen life cycle information.

7. The intelligent management method for linen based on the Internet of Things and RFID technology according to claim 1, characterized in that The specific steps of S5 are as follows: S51. Obtain the usage times, washing times, and material type coding of the linen from the life cycle database, and obtain the corresponding usage time series and washing time series. S52. Calculate the decay usage times and decay washing times based on the time decay mechanism. Among them, represents the attenuation usage times of linen i, represents the attenuation washing times of linen i, decay represents attenuation, u i represents the usage times of linen i, w i represents the washing times of linen i, T represents, represents the k-th usage time, represents the k-th washing time, δ represents the time decay coefficient, P represents the current time step; S53. Construct a feature vector of the linen, and the feature vector includes the original times, square term, cross term, logarithmic transformation term, time decay term, and material type coding. S54. Construct a loss function with a dynamic adjustment term, train an improved Elastic Net sparse regression model, and the improvement includes introducing a non-linear feature transformation and a time decay mechanism, and adopting a grouped network and introducing a dynamic regularization parameter mechanism in the loss function. Among them, represents the loss function, m represents the number of linen samples, and y i represents the true life label of the linen, represents the predicted remaining life of the linen, and λ1(m i ) represents the group regularization coefficient dynamically adjusted according to the material type code m i m i represents the material type code, G represents the total number of material groups, and ||β g ||2 represents the regression parameter vector belonging to the g-th group, and λ2(u i , w i ) represents the L1 regularization coefficient dynamically adjusted based on the usage times and washing times, u i represents the usage times of the i-th linen, and w i represents the washing times of the i-th linen, and ||β||1 represents the L1 norm of all regression coefficients; S55. Input the feature vector into the improved Elastic Net sparse regression model to predict the remaining life of the linen. Among them, represents the predicted remaining life of the linen, β0 represents the bias term of the sparse regression model, and β j represents the regression coefficient corresponding to the j-th eigenvector, and x ij represents the j-th eigenvector of the linen, and d represents the total number of dimensions of the eigenvectors; S56. Set the service life threshold θ. If the following condition is met mark the linen as to-be-scrapped status and update the status field in the linen management system.

8. The intelligent management method for linen based on the Internet of Things and RFID technology according to claim 1, characterized in that The specific steps of S6 are as follows: S61. Before the linen enters the industrial washing link, read the electronic tag data of each piece of linen through the RFID reading and writing device set at the linen entrance, and obtain the data record containing the unique identification information and type information. S62. Send the read information to the linen type recognition module, and determine the standard classification type to which the linen belongs according to the preset linen type mapping rules. S63. Input the recognized linen type into the washing parameter configuration system, and retrieve the corresponding washing program template of this type of linen in the preset program library. The washing program template includes the disinfection time, main washing time, washing temperature, dehydration speed, and detergent dosage. S64. Automatically transmit the parameters of the retrieved washing program template to the control unit of the industrial washing equipment to realize the program linkage with the control system. S65. If the system detects that the current linen to be washed contains different types, select the main type according to the proportion of the linen quantity, and generate a comprehensive washing program parameter set according to the preset parameter merging rule to cover the washing requirements of the mixed-type linen; S66. Automatically load the finally determined washing program parameters into the execution module of the washing equipment to complete the automatic matching of the linen and the washing program. At the same time, upload the unique identification information of the linen, the identified linen type, the matching washing program number, and the processing time to the background management system.

9. The intelligent management method for linen based on the Internet of Things and RFID technology according to claim 1, characterized in that, The specific content of S7 includes: S71. When the linen completes any link of distribution, use, recycling, washing, or storage, collect the unique identity identification, the current event type, the event occurrence time, the event location code, the on-site status information, and the operator number corresponding to the linen to form complete linen event data; S72. Package each piece of linen event data into a structured data format, and the structured data format includes an identity field, an event field, a status field, and a time field to form a standard transaction data record; S73. Perform processing operations on the standard transaction data record, and the processing operations include data integrity verification, time legality verification, and uniqueness confirmation. After completion, submit it to the blockchain system; S74. Call the data on-chain interface through the intelligent contract module of the blockchain system to perform identity verification, format verification, and consensus mechanism confirmation on the standard transaction data record; S75. Write the verified linen event data into the blockchain ledger system, record the event data, timestamp, and the previous block digest information in the current block to form a blockchain data structure; S76. Generate an independent life cycle linked list for each piece of linen. The system links all historical event blocks related to the linen in chronological order to form a tracking path and establish a query index in the management system; S77. Provide a traceability query interface based on the identity identification, event type, and time range, support users to view the status records of any linen at each node in the whole life cycle, and realize the fully traceable and non-tamperable management of the linen circulation.

10. A smart management system for linen based on the Internet of Things and RFID technology, which executes the smart management method for linen based on the Internet of Things and RFID technology according to any one of claims 1 to 9, characterized in that, It includes: The linen identity binding module is used to bind an RFID electronic tag with a unique identifier to each piece of linen, establish an association with the linen identity data, and generate a linen digital identity; The status acquisition module is used to deploy RFID reading and writing devices in the links of linen distribution, use, recycling, washing, and storage, read the electronic tag data and upload it to the background management system to complete the real-time acquisition and recording of the linen status; The circulation modeling and scheduling module is used to construct a linen circulation path diagram, generate node embedding representations using a graph neural network algorithm, predict the flow trend of the linen between regions, and generate a scheduling plan; The life cycle statistics module is used to count the usage times and washing times of the linen and establish a linen life cycle database; The remaining life prediction module is used to construct a feature vector according to the usage times, washing times, and material properties of the linen, input it into an improved Elastic Net sparse regression model, predict the remaining life of the linen, and judge whether it reaches the scrapping standard; The washing linkage module is used to read the RFID tag information before the linen is put into washing, identify the type of linen, and automatically match the washing program parameters to achieve linkage control; The blockchain management module is used to record the identity information, status information, and timestamp in the linen transfer process into the blockchain ledger, construct an immutable resume of the linen life cycle, and provide traceability query services.

Images