A Method and System for Traceability Management of Liquid Hydrogen Cylinders Based on RFID Tags

Abstract

The present invention relates to the technical field of liquid hydrogen cylinder management, and specifically to a traceability management method and system for liquid hydrogen cylinders based on RFID tags. It includes: First, obtain the historical RFID tag information of the liquid hydrogen cylinders. Then, through data clustering and training, establish an RFID tag information analysis model, and use classification and anomaly detection technologies to label and optimize abnormal data. During the model training process, use one-hot encoding to label data, extract sample features, generate an RFID tag information analysis vector, and further achieve the convergence of the model by optimizing the loss function. Finally, obtain the RFID information of the current liquid hydrogen cylinder and input it into the optimal analysis model to generate an analysis result. In addition, this method also combines planning information to construct a status curve and perform matching. When the prediction result deviates greatly from the actual situation, adjust the RFID tag information through an adaptive update mechanism to achieve real-time and accurate management.

Description

Technical Field

[0001] The present invention relates to the technical field of liquid hydrogen cylinder management, and particularly to a traceability management method and system for liquid hydrogen cylinders based on RFID tags. Background Art

[0002] Liquid hydrogen cylinders are the core equipment for liquid hydrogen storage and transportation, and their safety, reliability, and service life directly affect the storage and transportation process of liquid hydrogen. Therefore, the management of liquid hydrogen cylinders is crucial, especially in aspects such as cylinder traceability, maintenance, repair, and filling, where an effective and accurate management system is required to ensure safety and compliance.

[0003] In the management of liquid hydrogen cylinders, various methods have been adopted. For example, based on the traditional liquid hydrogen cylinder management method, manual records and inspections are used, but this method has problems such as inaccurate records, information lag, and cumbersome operations. Especially during the use of cylinders, it is impossible to track key information such as their usage status, maintenance records, and inspection cycles in real time.

[0004] With the gradual development of Internet of Things technology and automated management systems, especially RFID (Radio Frequency Identification) technology, as a technology that can achieve remote, non-contact information reading and automatic identification, has been widely used in many industries; by installing RFID tags on liquid hydrogen cylinders, the defects of the traditional liquid hydrogen cylinder management method can be well solved. According to the RFID tags, functions such as real-time tracking, status monitoring, and information recording of liquid hydrogen cylinders can be realized, greatly improving the degree of management intelligence.

[0005] Although the management of liquid hydrogen cylinders based on RFID tags has significant advantages, there are still some defects in actual applications. For example, the existing RFID tag-based liquid hydrogen cylinder management method lacks the ability to adapt to scene changes and cannot automatically adjust management strategies according to different environmental conditions and supply chain states, which will lead to inaccurate data and low management efficiency.

[0006] Therefore, the present invention proposes a traceability management method and system for liquid hydrogen cylinders based on RFID tags. Summary of the Invention

[0007] The object of the present invention is to provide a method and system for traceability management of liquid hydrogen cylinders based on RFID tags, which mainly realizes the accurate traceability and management of the whole life cycle of liquid hydrogen cylinders by constructing and optimizing an RFID tag information analysis model. The method and system collect and process the historical RFID tag information of liquid hydrogen cylinders, including environmental information (such as temperature, humidity, extrusion, vibration, etc.) and production and usage information (such as cylinder capacity, liquid hydrogen volume, status information, location, maintenance and fault information, etc.), annotate, deduplicate, standardize and process outliers for this information, and finally obtain a standard set of historical annotated data.

[0008] Furthermore, through cluster analysis and training the optimal RFID tag information analysis model, the system can realize the detection of abnormal data based on supply chain nodes and accurately predict the status of liquid hydrogen cylinders. When the actual RFID tag information of a liquid hydrogen cylinder does not match the status curve of the planned information, possible problems can be detected in a timely manner and adaptive updates can be made. Through the optimized RFID tag information, the system can provide accurate management support to the staff, thereby improving the management efficiency and safety of liquid hydrogen cylinders in the supply chain. The present invention makes the traceability management of liquid hydrogen cylinders more intelligent through in-depth analysis and processing of various types of data in the life cycle of liquid hydrogen cylinders.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A method for traceability management of liquid hydrogen cylinders based on RFID tags, comprising:

[0011] Obtaining the historical RFID tag information of multiple liquid hydrogen cylinders in the whole cycle; wherein, the historical RFID tag information includes: historical first tag information and historical second tag information;

[0012] The historical first tag information includes: temperature, humidity, extrusion and vibration; the historical second tag information includes: production information, cylinder capacity information, liquid hydrogen volume information, status information, usage information, location information, usage time information, maintenance information and fault information.

[0013] Furthermore, annotating the problems and supply chain nodes in the historical RFID tag information to obtain annotated RFID tag information;

[0014] Furthermore, processing the annotated RFID tag information to construct a standard set of annotated data;

[0015] Among them, the process of processing the labeled RFID tag information includes: removing duplicate data from the historical RFID tag information to obtain the first historical labeled data; performing format standardization processing on the first historical labeled data to obtain the second historical labeled data; performing data filling on the second historical labeled data to obtain the third historical labeled data; performing outlier processing on the third historical labeled data to obtain the fourth historical labeled data; performing data standardization processing on the fourth historical labeled data to obtain the standard historical labeled data; integrating multiple pieces of the standard historical labeled data to obtain the standard labeled data set.

[0016] Furthermore, clustering the data in the standard labeled data set to obtain multiple standard labeled data subsets;

[0017] Furthermore, training the RFID tag information analysis model to obtain the optimal RFID tag information analysis model;

[0018] Among them, the training process of the RFID tag information analysis model includes:

[0019] Inputting the sample data in multiple standard labeled data subsets through the input layer;

[0020] Furthermore, inputting the sample data of the input layer into the classification layer to obtain the corresponding anomaly data detector; among them, the classification layer classifies the sample data according to the supply chain node.

[0021] Furthermore, using the anomaly data detector to detect the sample data and label the anomaly data in the sample data; the detection process of the anomaly data detector includes:

[0022] Constructing a reasonable RFID tag information according to the node type corresponding to the supply chain node;

[0023] Performing feature transformation on the reasonable RFID tag information and constructing a reasonable RFID tag feature data matrix;

[0024] Performing label data impact change analysis on the reasonable RFID tag feature data matrix to obtain an RFID tag data change impact matrix;

[0025] Inputting the sample data and the RFID tag data change impact matrix into an anomaly detection function to obtain anomaly data;

[0026] Labeling the sample data according to the anomaly data using one-hot encoding;

[0027] Inputting the labeled sample data into the feature learning layer to obtain sample features;

[0028] Further, input the sample features into the information analysis layer to obtain an RFID tag information analysis vector;

[0029] Further, output the RFID tag information analysis vector through the output layer to obtain a problem prediction value and a supply chain node prediction value;

[0030] Further, calculate the loss between the problem prediction value and the labeled problem value of the sample data to obtain a problem prediction loss;

[0031] Further, calculate the loss between the supply chain node prediction value and the labeled supply chain node value of the sample data to obtain a supply chain node prediction loss;

[0032] Further, optimize the RFID tag information analysis model according to the problem prediction loss and the supply chain node prediction loss. When the problem prediction loss and the supply chain node prediction loss reach convergence, the training is completed, and the optimal RFID tag information analysis model is obtained.

[0033] Further, obtain the RFID tag information of the current liquid hydrogen cylinder. After processing, input it into the optimal RFID tag information analysis model to obtain an RFID tag information analysis result;

[0034] Further, obtain the planned information of the current liquid hydrogen cylinder;

[0035] Further, match the RFID tag information analysis result with the status curve of the planned information;

[0036] Among them, the construction process of the status curve includes:

[0037] Divide the planned information according to the supply chain nodes to obtain multiple planned information subsets;

[0038] Further, input the planned information subset into the status curve generation function to obtain a status sub-curve;

[0039] The acquisition formula of the status sub-curve is:

[0040] ;

[0041] Among them, is the status sub-curve of the supply chain node ; is the time variable; is the status curve generation function; is the supply chain node ; is the set of time information for supply chain nodes ; is the factor related to the state change;

[0042] Further, when the matching result does not match, adaptively update the RFID tag information. The specific process includes:

[0043] Calculate the deviation degree between the analysis result of the RFID tag information and each data point on the state curve, and establish a deviation degree vector;

[0044] Further, conduct statistical analysis on the deviation degree vector, evaluate the influence degree of the analysis result, and obtain an influence weight vector;

[0045] Further, obtain the instantaneous change difference of the RFID tag information;

[0046] Further, automatically update the RFID tag information according to the deviation degree vector, the influence weight vector, the instantaneous change difference, and the RFID tag data change influence matrix.

[0047] Further, send the updated RFID tag information to the staff for management.

[0048] A liquid hydrogen cylinder traceability management system based on RFID tags, comprising:

[0049] An information storage unit for storing the historical RFID tag information of the entire life cycle of the liquid hydrogen cylinder; wherein, the historical RFID tag information includes: historical first tag information and historical second tag information; the first tag information includes: temperature, humidity, extrusion, and vibration; the second tag information includes: production information, cylinder capacity information, liquid hydrogen quantity information, status information, usage information, location information, maintenance information, and fault information.

[0050] A problem recording unit for recording problems related to the liquid hydrogen cylinder at different supply chain nodes;

[0051] An information collection unit for collecting the RFID tag information of the liquid hydrogen cylinder;

[0052] An RFID tag information analysis unit for analyzing the obtained RFID tag information;

[0053] Among them, the RFID tag information analysis unit uses an RFID tag information analysis model for analysis, including:

[0054] The input layer receives the processed sample data of the RFID tag information;

[0055] Further, input the sample data into the classification layer to obtain the corresponding anomaly data detector; wherein, the anomaly data detector includes:

[0056] Construct rationalized RFID tag information according to the node type corresponding to the supply chain node;

[0057] Perform feature transformation on the rationalized RFID tag information and construct a rationalized RFID tag feature data matrix;

[0058] Perform label data impact change analysis on the rationalized RFID tag feature data matrix to obtain an RFID tag data change impact matrix;

[0059] Input the sample data and the RFID tag data change impact matrix into the anomaly detection function to obtain anomaly data;

[0060] Label the sample data using one-hot encoding according to the anomaly data.

[0061] Further, use the anomaly data detector to detect the sample data and label the anomaly data in the sample data;

[0062] Further, input the labeled sample data into the feature learning layer to obtain sample features;

[0063] Further, input the sample features into the information analysis layer to obtain an RFID tag information analysis vector;

[0064] Further, output the RFID tag information analysis vector through the output layer to obtain a problem prediction value and a supply chain node prediction value.

[0065] The RFID tag information matching unit is used to perform fitness matching on the RFID tag information and judge the status;

[0066] Wherein, the RFID tag information matching unit includes:

[0067] Obtain the planned information of the current liquid hydrogen cylinder;

[0068] Further, match the RFID tag information analysis result with the status curve of the planned information; wherein, the construction process of the status curve includes:

[0069] Divide the planned information according to the supply chain node to obtain multiple planned information subsets;

[0070] Input the planned information subset into the status curve generation function to obtain a status sub-curve;

[0071] The acquisition formula for the state sub - curve is as follows:

[0072] ;

[0073] Among them, is the state sub - curve of the supply chain node ; is the time variable; is the state - curve generation function; is the subset of the planned information of the supply chain node ; is the set of time information of the supply chain node ; is the factor related to the state change;

[0074] Optimize and merge each of the state sub - curves to obtain the state curve;

[0075] The RFID tag information adaptive update unit is used for updating the RFID tag information;

[0076] The feedback unit is used to feedback the RFID tag information to the staff;

[0077] The management unit is used for the staff to manage the liquid hydrogen cylinders in a timely manner.

[0078] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0079] 1. The present invention proposes an RFID tag information analysis model to analyze the RFID tag information; through the processing and analysis of the RFID tag information, this model can effectively identify abnormal data, and use rational feature transformation and analysis of the influence of tag data changes to construct an accurate anomaly detection mechanism. By annotating abnormal data and learning features for the sample data, the accuracy of information analysis is further improved, thus providing efficient support for the prediction of supply chain nodes and problem prediction. This model can provide comprehensive data support for the application of RFID tag information through in - depth analysis and processing, which helps to improve the intelligent level of the traceability management of liquid hydrogen cylinders.

[0080] 2. The present invention proposes a method for matching the RFID tag information of a current liquid hydrogen cylinder with the planned information; this method ensures a high degree of consistency between the tag information and the planned status by obtaining the current planned information of the liquid hydrogen cylinder and comparing it with the analysis result of the RFID tag information. Specifically, the construction process of the status curve includes dividing the planned information of the supply chain nodes to generate multiple subsets of planned information, and using the status curve generation function to combine the usage time information and relevant factors to obtain the corresponding status sub-curves. After optimization, all the status sub-curves are merged into a complete status curve to accurately describe the changes of the liquid hydrogen cylinder in the supply chain. This method effectively ensures the accuracy and flexibility of the management of liquid hydrogen cylinders through dynamic update and real-time matching.

[0081] 3. The present invention proposes a method for adaptive update of RFID tag information; this method realizes the automatic update of RFID tag information through the following steps: First, calculate the deviation degree between the analysis result of the RFID tag information and each data point on the status curve, and establish a deviation degree vector; then, conduct statistical analysis on this deviation degree vector, and evaluate the influence degree of each data point on the system according to the analysis result to obtain an influence weight vector; next, obtain the instantaneous change difference of the RFID tag information, which represents the change amplitude of the tag information in a short period of time; finally, combine the deviation degree vector, the influence weight vector, the instantaneous change difference, and the RFID tag data change influence matrix to automatically update the RFID tag information. This method can accurately update the RFID tag information in a dynamic environment, thereby improving the accuracy and real-time nature of the information, and effectively supporting the traceability management of liquid hydrogen cylinders to help the staff manage the liquid hydrogen cylinders more efficiently and safely. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a flowchart of a method for traceability management of a liquid hydrogen cylinder based on an RFID tag provided by an embodiment of the present invention;

[0083] Figure 2 It is a system structure diagram of a system for traceability management of a liquid hydrogen cylinder based on an RFID tag provided by an embodiment of the present invention;

[0084] Figure 3 It is a schematic diagram of various losses during the training of an RFID tag information analysis model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0085] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0086] A liquid hydrogen cylinder is a container for storing and transporting liquid hydrogen. In order to better manage liquid hydrogen cylinders, the prior art uses RFID tag information to record the status of cylinders at each time node. RFID (Radio Frequency Identification) technology, as a technology that can achieve remote, non-contact information reading and automatic identification, has been widely used in many industries;

[0087] RFID tag technology can realize functions such as real-time tracking, status monitoring, and information recording of liquid hydrogen cylinders, greatly improving the intelligent level of management. However, in this process, due to the inability of RFID tag technology to achieve adaptive adjustment with the change of scenarios, there are errors in recording RFID tags; that is, when the current supply chain node changes, the RFID tag device cannot respond in time, resulting in the inability to synchronize the recorded information, which is not conducive to the traceability management of liquid hydrogen cylinders. For this reason, the present invention proposes a method and system for traceability management of liquid hydrogen cylinders based on RFID tags to overcome the defects existing in the current prior art. The specific process is described through the following two embodiments.

[0088] Embodiment 1:

[0089] In the embodiment of the present application, the traceability management of liquid hydrogen cylinders is realized through Figure 1 and Figure 2 is realized; Figure 1 The method flow proposed by the present invention includes: S10. Obtain the historical RFID tag information of multiple liquid hydrogen cylinders in the full cycle; S20. Mark the problems and supply chain nodes in the historical RFID tag information to obtain the marked RFID tag information; S30. Process the marked RFID tag information to construct a standard marked data set; S40. Cluster the data in the standard marked data set to obtain multiple standard marked data subsets; S50. Train the RFID tag information analysis model to obtain the optimal RFID tag information analysis model; S60. Obtain the RFID tag information of the current liquid hydrogen cylinder, and after processing, input it into the optimal RFID tag information analysis model to obtain the RFID tag information analysis result; S70. Obtain the planned information of the current liquid hydrogen cylinder; S80. Match the RFID tag information analysis result with the status curve of the planned information; S90. When the matching result does not match, adaptively update the RFID tag information; S100. Send the updated RFID tag information to the staff for management.Figure 2 This is the system structure diagram of the present invention, including: an information storage unit, a problem recording unit, an information acquisition unit, an RFID tag information analysis unit, an RFID tag information matching unit, an RFID tag information adaptive update unit, a feedback unit, and a management unit. The specific process according to the above content is as follows:

[0090] Obtain the historical RFID tag information of multiple liquid hydrogen cylinders in the full cycle from the information storage unit, corresponding to step S10; in the embodiment of the present application, the full cycle refers to the usage cycle of the liquid hydrogen cylinder, including: the production stage, the usage stage, the transportation stage, the recycling stage, the maintenance stage, and the scrap disposal stage; among them, the historical RFID tag information includes: historical first tag information and historical second tag information;

[0091] The historical first tag information is external environment information, obtained by sensors installed on the liquid hydrogen cylinder, including: temperature, humidity, extrusion, and vibration; the historical second tag information includes: production information (cylinder ID, production date, expected usage duration, maximum liquid hydrogen storage capacity, and cylinder specifications), cylinder capacity information, liquid hydrogen quantity information, status information (usable and unusable), usage information (usage status and liquid hydrogen usage amount), location information, usage time information, maintenance information, and fault information.

[0092] In the embodiment of the present application, by obtaining the historical RFID tag information of multiple liquid hydrogen cylinders, the key data of the liquid hydrogen cylinder in the entire life cycle can be comprehensively recorded, so as to realize real-time tracking and analysis of the cylinder status and usage situation, which helps to improve safety and optimize maintenance management.

[0093] Furthermore, according to step S20 above, problems and supply chain node annotations are made on the historical tag RFID information in the information storage unit; among them, the annotation of the historical tag RFID information is based on the record log of the problems encountered by the staff on the liquid hydrogen cylinder at each supply chain node in the problem recording unit, and these information are annotated by an automated annotation method.

[0094] The problems recorded in the problem recording unit are such as, "Cylinder ID: BW7754; Date: 2024 / 12 / 01 13:30; Supply chain node: Transportation; Problem: Overheating"; "Cylinder ID: BW7263; Date: 2024 / 12 / 20 7:00; Supply chain node: Transportation; Problem: Supply link error", etc.;

[0095] Further, according to the content of step S30, the labeled RFID tag information is processed to construct a standard labeled data set; among them, the processing process includes: removing duplicate data from the historical RFID tag information to obtain the first historical labeled data; performing format standardization processing on the first historical labeled data to obtain the second historical labeled data; performing data filling on the second historical labeled data to obtain the third historical labeled data; performing outlier processing on the third historical labeled data to obtain the fourth historical labeled data; performing data standardization processing on the fourth historical labeled data to obtain the standard historical labeled data; integrating multiple standard historical labeled data to obtain the standard labeled data set.

[0096] In the embodiment of the present application, the processing of the labeled data is realized through multiple data processing methods, which can ensure the high quality and high consistency of the data, thus providing a solid foundation for the subsequent analysis of RFID tag information.

[0097] Further, according to the content of step S40, the data in the standard labeled data set is clustered to obtain multiple standard labeled data subsets;

[0098] Among them, in the embodiment of the application, clustering is achieved with the problems labeled in the data set and the supply chain node information as the targets, and the specific process includes:

[0099] Step 1: Select a suitable clustering algorithm according to the data characteristics, DBSCAN (density-based clustering);

[0100] Step 2: Select a distance metric or similarity metric, and use the Manhattan distance algorithm to calculate the distance between each piece of data;

[0101] Step 3: Select the number of clusters. By drawing the sum of squared errors (SSE) graph under different numbers of clusters K, select the K value that makes the decrease in SSE the smallest as the final number of clusters;

[0102] Step 4: Input the standard labeled data set into the DBSCAN algorithm;

[0103] Step 5: Evaluate the clustering effect by calculating the compactness within the cluster (such as the mean square error of the data within the cluster) to evaluate the clustering effect;

[0104] Step 6: Output the result.

[0105] Further, according to the content of step S50, multiple standard labeled data subsets are used to train the RFID tag analysis model to obtain the optimal RFID tag analysis model;

[0106] Among them, the training process of the RFID tag analysis model includes:

[0107] Input the sample data in multiple standard annotation data subsets through the input layer;

[0108] Further, input the sample data of the input layer into the classification layer to obtain the corresponding anomaly data detector; among them, the classification layer classifies the sample data according to the supply chain nodes; the classification layer uses a trained CNN neural network model, and the corresponding classification labels include: production stage, usage stage, transportation stage, recycling stage, maintenance stage, and scrapping and disposal stage;

[0109] Further, use the anomaly data detector to detect the sample data and label the anomaly data in the sample data; the detection process of the anomaly data detector includes:

[0110] Further, construct rationalized RFID tag information according to the node type corresponding to the supply chain node; among them, the rationalized RFID tag information is composed of historical RFID tag data in the information storage unit, and this data comes from valid records without problems, reflecting the most ideal RFID tag information under the corresponding supply chain node.

[0111] Further, perform feature transformation on the rationalized RFID tag information and construct a rationalized RFID tag feature data matrix;

[0112] Further, perform label data impact change analysis on the rationalized RFID tag feature data matrix to obtain an RFID tag data change impact matrix;

[0113] Among them, an impact analysis model is used to obtain the RFID tag data change impact matrix, and this model uses a non-linear combination method for calculation. The specific formula is:

[0114] ;

[0115] Among them, is the mutual influence value between feature and feature ; is the influence weight of feature ; is the influence weight of feature ; , , and are hyperparameters obtained through training; is an exponential function, reflecting the non-linear effect of feature on the influence value; is a logarithmic function, reflecting the non-linear influence of feature ; is a hyperbolic tangent function, which can capture the feature The combined non-linear effect with the feature ;

[0116] Referring to Table 1, the impact analysis in the "transportation stage" is given in the embodiments of the present application;

[0117] Table 1 Impact Analysis Matrix (Transportation Stage)

[0118]

[0119] Furthermore, the sample feature vector and the RFID tag data change impact matrix are input into the anomaly detection function to obtain anomaly data; the formula of the anomaly detection function is: ; where is the anomaly detection function; is the sample feature; is the RFID tag data change impact matrix; is the distance function; is the th feature vector, and ; is the impact value of the th feature vector, and ; is the number of feature items; is the normal range threshold generated from the training data;

[0120] Among them, the distance function is defined as: ; where is the feature of the th feature vector in dimension ; is the dimension of the sample feature vector.

[0121] Furthermore, the sample data is labeled using one-hot encoding according to the anomaly data;

[0122] Furthermore, the labeled sample data is input into the feature learning layer to obtain sample features;

[0123] Furthermore, the sample features are input into the information analysis layer to obtain the RFID tag information analysis vector;

[0124] Furthermore, the RFID tag information analysis vector is output through the output layer to obtain the problem prediction value and the supply chain node prediction value;

[0125] Furthermore, the loss between the problem prediction value and the labeled problem value of the sample data is calculated to obtain the problem prediction loss;

[0126] Further, calculate the loss between the predicted value of the supply chain node and the labeled supply chain node value of the sample data to obtain the supply chain node prediction loss;

[0127] Further, optimize the RFID tag information analysis model according to the problem prediction loss and the supply chain node prediction loss; refer to Figure 3 , in Figure 3 shows the change of loss during multiple rounds of training of the RFID tag information analysis model.

[0128] When the problem prediction loss and the supply chain node prediction loss converge, the training is completed, and the optimal RFID tag information analysis model is obtained.

[0129] Refer to Table 2, which shows the parameter settings of the optimal RFID tag information analysis model obtained after multiple rounds of training, including: the number of training rounds, learning rate, loss function, and hyperparameter settings;

[0130] Table 2 Parameter Settings

[0131]

[0132] Further, optimize the RFID tag information analysis model according to the parameter content in Table 2; and randomly select 80 historical RFID tag information from the information storage unit as the validation set for validation; refer to Table 3, which shows the validation performance of the optimal RFID tag information analysis model.

[0133] Table 3 Model Validation Performance Description

[0134]

[0135] In the embodiments of the present application, the training process of the RFID tag information analysis model is described. Through the training of this model, abnormal data at different nodes in the supply chain (such as production, transportation, use, etc.) can be identified, and effectively labeled and traced. The training process includes inputting RFID tag data into a multi-layer neural network, combining with CNN (Convolutional Neural Network) for sample classification and anomaly detection, and finally optimizing the model to achieve more accurate tag analysis and anomaly prediction. During the training process, by analyzing the sample feature vector and the influence matrix of the change of RFID tag data, the accuracy and precision of anomaly detection are further improved. Finally, through the loss function for optimization, a model that can accurately identify abnormal data, predict problems and the status of supply chain nodes in practical applications is trained, which can track all links of the production, transportation, use, and maintenance of gas cylinders in real time to ensure the safety and compliance of liquid hydrogen gas cylinders.

[0136] Further, the RFID tag information of the current liquid hydrogen cylinder is collected by the information collection unit, and after being processed, it is input into the RFID tag analysis unit for analysis, corresponding to step S60; among them, the RFID tag analysis unit adopts an RFID tag information analysis model.

[0137] Further, according to the content of step S70, the planned information of the current liquid hydrogen is obtained;

[0138] Further, the RFID tag information matching unit is used to match the analysis result of the analyzed RFID tag information, corresponding to step S80; the specific process includes:

[0139] The planned information is divided according to the supply chain nodes to obtain multiple planned information subsets;

[0140] Further, the planned information subset is input into the state curve generation function to obtain a state sub-curve;

[0141] Among them, the acquisition formula of the state sub-curve is:

[0142] ;

[0143] Among them, is the supply chain node 's state sub-curve; is the time variable; is the state curve generation function; is the supply chain node 's planned information subset; is the supply chain node 's time information set; is a factor related to state change;

[0144] In the acquisition of the state sub-curve, the state curve generation function is defined as:

[0145] ;

[0146] Among them, is the initial state; is the decay constant; is the number of parameters; 、 、 and are state curve calculation parameters;

[0147] Further, each state sub-curve is optimized and merged to obtain a state curve; among them, the decay constant is dynamically adjusted according to the fluctuations of the actual supply chain.

[0148] Further, when the matching result does not match, the RFID tag information is adaptively updated, corresponding to step S90; among them, the RFID tag information adaptive update unit is used for the update, and the specific process includes:

[0149] Calculate the deviation degree between the analysis result of the RFID tag information and each data point on the state curve, and establish a deviation degree vector;

[0150] Further, perform statistical analysis on the deviation degree vector, evaluate the influence degree of the analysis result, and obtain an influence weight vector; among them, the statistical analysis of the deviation vector includes: mean value, variance, extreme value, and deviation trend;

[0151] According to the above calculation of the influence degree, the calculation formula is: ; among them, is the comprehensive influence weight of the th data point; is the influence weight of the mean value of the th data point; is the influence weight of the variance of the th data point; is the influence weight of the extreme value of the th data point; is the influence weight of the deviation trend of the th data point; , , and are the coefficients of each influence weight;

[0152] Further, obtain the instantaneous change difference of the RFID tag information;

[0153] Further, automatically update the RFID tag information according to the deviation degree vector, influence weight vector, instantaneous change difference, and RFID tag data change influence matrix; among them, the automatic update formula is used to update each item of information, and the formula is: ; among them, is the updated value of the RFID tag information of the th data point; is the original RFID tag information value of the th data point; is the instantaneous change difference of the th data point; is a compensation function.

[0154] In the example of this application, by calculating the deviation degree between the RFID tag information and the data points of the status curve, a deviation degree vector is established. Further, through statistical analysis of the deviation degree vector (including mean, variance, extreme value, and deviation trend), the influence weight vector of each data point is obtained. Through comprehensive analysis of each weight, the comprehensive influence weight of each data point is evaluated, and the RFID tag information is automatically updated according to the deviation degree, influence weight, instantaneous change difference, and the RFID tag data change influence matrix. This automated update process ensures the accuracy and timeliness of the RFID tag information, greatly improving the data processing efficiency and reliability. For the traceability management of liquid hydrogen cylinders, this method based on RFID tag information analysis and update has significant benefits. It can monitor the status changes of liquid hydrogen cylinders in real time. Through accurate data analysis, abnormal situations in the use of cylinders can be detected in a timely manner, such as the temperature, pressure, etc. deviating from the normal range, so as to realize the whole-process tracking and early warning of the cylinders.

[0155] Further, the feedback unit feeds back to the staff, and the staff uses the management unit for management, corresponding to step S100.

[0156] In the embodiment of this application, according to Figure 1 and Figure 2 the content realizes the adaptive update and traceability management of the RFID tag information of liquid hydrogen cylinders at each supply chain node; mainly including: mainly including: through comprehensive collection and annotation of the historical RFID tag information of liquid hydrogen cylinders, a standardized data set is constructed, and data processing and optimization are carried out through cluster analysis and anomaly detection models. Subsequently, based on the training of the optimal RFID tag information analysis model, combined with the real-time status information of liquid hydrogen cylinders, potential problems and the status of supply chain nodes can be accurately predicted, and through matching with the planned information, the adaptive update of RFID tags is realized. This process ensures that in the entire supply chain process, the status information of liquid hydrogen cylinders is always under precise control and real-time tracking, effectively improving the accuracy and reliability of data, reducing the risks brought by abnormal data at the same time, and optimizing supply chain management.

[0157] Embodiment 2:

[0158] In Embodiment 1, the traceability management of the current liquid hydrogen cylinder is realized by using RFID tags; in order to further illustrate the versatility of this application, it is described again in the embodiment of this application, and the specific process includes:

[0159] Obtain the historical RFID tag information of multiple liquid hydrogen cylinders in the whole life cycle;

[0160] Further, label the problems and supply chain nodes in the historical RFID tag information to obtain the labeled RFID tag information;

[0161] Furthermore, process the labeled RFID tag information to construct a standard labeled data set;

[0162] Furthermore, cluster the data in the standard labeled data set to obtain multiple standard labeled data subsets;

[0163] Furthermore, train the RFID tag information analysis model to obtain an optimal RFID tag information analysis model;

[0164] Furthermore, obtain the RFID tag information of the current liquid hydrogen cylinder, and after processing, input it into the optimal RFID tag information analysis model to obtain an RFID tag information analysis result;

[0165] Furthermore, match the RFID tag information analysis result with the state curve of the planned information;

[0166] Among them, the construction process of the state curve includes:

[0167] Divide the planned information according to the supply chain nodes to obtain multiple planned information subsets;

[0168] Furthermore, input the planned information subset into the state curve generation function to obtain a state sub-curve;

[0169] The acquisition formula of the state sub-curve is:

[0170] ;

[0171] Among them, is the state sub-curve of the supply chain node ; is the time variable; is the state curve generation function; is the planned information subset of the supply chain node ; is the time information set of the supply chain node ; is the factor related to the state change;

[0172] Furthermore, send the updated RFID tag information to the staff for management.

[0173] Furthermore, when the matching result does not match, adaptively update the RFID tag information;

[0174] Furthermore, send the updated RFID tag information to the staff for management.

[0175] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A liquid hydrogen cylinder traceability management method based on RFID tags, characterized in that: include: Obtain historical RFID tag information of multiple liquid hydrogen cylinders throughout their entire life cycle; Mark the problems and supply chain nodes in the historical RFID tag information to obtain the marked RFID tag information; Process the annotated RFID tag information and construct a standard annotated data set; Cluster the data in the standard labeled data set to obtain multiple standard labeled data subsets; The RFID tag information analysis model is trained to obtain an optimal RFID tag information analysis model, including: inputting sample data from multiple standard labeled data subsets into a classification layer through an input layer to obtain an abnormal data detector; using the abnormal data detector to detect and label the sample data; inputting the labeled sample data into a feature learning layer to obtain sample features; inputting the sample features into an information analysis layer to obtain an RFID tag information analysis vector; outputting the problem prediction value and the supply chain node prediction value of the RFID tag information analysis vector through an output layer; calculating the loss of the problem prediction value and the labeled problem value of the sample data to obtain the problem prediction loss; calculating the loss of the supply chain node prediction value and the labeled supply chain node value of the sample data to obtain the supply chain node prediction loss; optimizing the RFID tag information analysis model according to the problem prediction loss and the supply chain node prediction loss, completing the training when convergence, and obtaining the optimal RFID tag information analysis model; The RFID tag information of the current liquid hydrogen cylinder is obtained, and after processing, it is input into the optimal RFID tag information analysis model to obtain the RFID tag information analysis result; Get the current plan information of liquid hydrogen tanks; Match the RFID tag information analysis results with the status curve of the planning information; The process of constructing the state curve includes: Divide the planning information according to the supply chain nodes to obtain multiple planning information subsets; Input the planning information subset into the state curve generation function to obtain the state sub-curve; The formula for obtaining the state sub-curve is: ； in, Supply Chain Node The state sub-curve of is the time variable; Generate functions for state curves; Supply Chain Node A subset of the planning information; Supply Chain Node A collection of time information; Factors related to state changes; Optimize each state sub-curve and merge them to obtain the state curve; When the matching result does not match, the RFID tag information is adaptively updated, including: calculating the deviation between the RFID tag information analysis result and each data point on the state curve, establishing a deviation vector; performing statistical analysis on the deviation vector, evaluating the impact of the analysis result, and obtaining an impact weight vector; obtaining the instantaneous change difference of the RFID tag information; automatically updating the RFID tag information according to the deviation vector, the impact weight vector, the instantaneous change difference, and the RFID tag data change impact matrix; Send the updated RFID tag information to the staff for management.

2. The RFID tag-based liquid hydrogen cylinder traceability management method according to claim 1 is characterized in that: Wherein, the historical RFID tag information includes: historical first tag information and historical second tag information; The historical first label information includes: temperature, humidity, extrusion and vibration; the historical second label information includes: production information, gas cylinder capacity information, liquid hydrogen quantity information, status information, usage information, location information, usage time information, maintenance information and fault information.

3. The RFID tag-based liquid hydrogen cylinder traceability management method according to claim 1 is characterized in that: The process of processing the annotated RFID tag information includes: deduplicating the historical RFID tag information to obtain first historical annotated data; performing format standardization processing on the first historical annotated data to obtain second historical annotated data; performing data filling on the second historical annotated data to obtain third historical annotated data; performing outlier processing on the third historical annotated data to obtain fourth historical annotated data; performing data standardization processing on the fourth historical annotated data to obtain standard historical annotated data; integrating multiple pieces of the standard historical annotated data to obtain the standard annotated data set.

4. The RFID tag-based liquid hydrogen cylinder traceability management method according to claim 1 is characterized in that: The training process of the RFID tag information analysis model includes: Inputting sample data from a plurality of standard annotated data subsets through an input layer; Inputting the sample data of the input layer into the classification layer to obtain a corresponding abnormal data detector; wherein the classification layer classifies the sample data according to the supply chain node; The sample data is detected by using the abnormal data detector, and abnormal data in the sample data is marked; the detection process of the abnormal data detector includes: According to the node type corresponding to the supply chain node, rationalized RFID tag information is constructed; Performing feature conversion on the rationalized RFID tag information and constructing a rationalized RFID tag feature data matrix; Performing tag data influence change analysis on the rationalized RFID tag feature data matrix to obtain an RFID tag data change influence matrix; Inputting the sample data and the RFID tag data change impact matrix into an anomaly detection function to obtain abnormal data; Labeling the sample data using one-hot encoding according to the abnormal data; Inputting the labeled sample data into the feature learning layer to obtain sample features; Input the sample features into the information analysis layer to obtain the RFID tag information analysis vector; Output the RFID tag information analysis vector through the output layer to obtain the problem prediction value and the supply chain node prediction value; Calculate the loss between the problem prediction value and the labeled problem value of the sample data to obtain the problem prediction loss; Calculate the loss of the supply chain node prediction value and the labeled supply chain node value of the sample data to obtain the supply chain node prediction loss; The RFID tag information analysis model is optimized according to the problem prediction loss and the supply chain node prediction loss. When the problem prediction loss and the supply chain node prediction loss converge, the training is completed and the optimal RFID tag information analysis model is obtained.

5. A liquid hydrogen cylinder traceability management system based on RFID tags, characterized in that: The system includes: an information storage unit for storing historical RFID tag information of the entire cycle of the liquid hydrogen cylinder; a problem recording unit for recording problems related to the liquid hydrogen cylinder at different supply chain nodes; an information collection unit for collecting RFID tag information of the liquid hydrogen cylinder; an RFID tag information analysis unit for analyzing the acquired RFID tag information using an RFID tag information analysis model; an RFID tag information matching unit for matching the RFID tag information for fit and judging the status; an RFID tag information adaptive updating unit for adaptively updating the RFID tag information; a feedback unit for feeding back RFID tag information to the staff; and a management unit for the staff to manage the liquid hydrogen cylinder in a timely manner; The RFID tag information analysis model training process includes: inputting sample data from multiple standard annotated data subsets into the classification layer through the input layer to obtain an abnormal data detector; using the abnormal data detector to detect and annotate the sample data; inputting the annotated sample data into the feature learning layer to obtain sample features; inputting the sample features into the information analysis layer to obtain an RFID tag information analysis vector; outputting the problem prediction value and the supply chain node prediction value of the RFID tag information analysis vector through the output layer; calculating the loss between the problem prediction value and the annotated problem value of the sample data to obtain the problem prediction loss; calculating the loss between the supply chain node prediction value and the annotated supply chain node value of the sample data to obtain the supply chain node prediction loss; optimizing the RFID tag information analysis model according to the problem prediction loss and the supply chain node prediction loss, completing the training when convergence, and obtaining the optimal RFID tag information analysis model; The adaptive update of RFID tag information includes: calculating the deviation between the RFID tag information analysis results and each data point on the state curve, and establishing a deviation vector; performing statistical analysis on the deviation vector, performing an impact assessment on the analysis results, and obtaining an impact weight vector; obtaining the instantaneous change difference of the RFID tag information; and automatically updating the RFID tag information according to the deviation vector, the impact weight vector, the instantaneous change difference, and the RFID tag data change impact matrix.

6. The RFID tag-based liquid hydrogen cylinder traceability management system according to claim 5 is characterized in that: The historical RFID tag information includes: historical first tag information and historical second tag information; the first tag information includes: temperature, humidity, extrusion and vibration; the second tag information includes: production information, gas cylinder capacity information, liquid hydrogen quantity information, status information, usage information, location information, maintenance information and fault information.

7. The RFID tag-based liquid hydrogen cylinder traceability management system according to claim 5 is characterized in that: The RFID tag information analysis unit uses an RFID tag information analysis model to perform analysis, including: The input layer receives sample data of the processed RFID tag information; Input the sample data into the classification layer to obtain the corresponding abnormal data detector; Detecting the sample data using the abnormal data detector, and marking abnormal data in the sample data; Inputting the labeled sample data into the feature learning layer to obtain sample features; Input the sample features into the information analysis layer to obtain the RFID tag information analysis vector; The RFID tag information analysis vector is output through the output layer to obtain the problem prediction value and the supply chain node prediction value.

8. The RFID tag-based liquid hydrogen cylinder traceability management system according to claim 7 is characterized in that: The abnormal data detector comprises: According to the node type corresponding to the supply chain node, rationalized RFID tag information is constructed; Performing feature conversion on the rationalized RFID tag information and constructing a rationalized RFID tag feature data matrix; Performing tag data influence change analysis on the rationalized RFID tag feature data matrix to obtain an RFID tag data change influence matrix; Inputting the sample features and the RFID tag data change impact matrix into an anomaly detection function to obtain abnormal data; The sample data is labeled using one-hot encoding according to the abnormal data.

9. The RFID tag-based liquid hydrogen cylinder traceability management system according to claim 5, characterized in that: The RFID tag information matching unit includes: Get the current plan information of liquid hydrogen tanks; Matching the RFID tag information analysis result with the status curve of the plan information; wherein the construction process of the status curve includes: Dividing the plan information according to the supply chain nodes to obtain multiple plan information subsets; Inputting the plan information subset into a state curve generating function to obtain a state sub-curve; The formula for obtaining the state sub-curve is: ； in, Supply Chain Node The state sub-curve of is the time variable; Generate functions for state curves; Supply Chain Node A subset of the planning information; Supply Chain Node A collection of time information; Factors related to state changes; Each of the state sub-curves is optimized and combined to obtain the state curve.

Images