Supercritical foaming material AI anti-counterfeiting traceability management method for packaging

Abstract

The application provides an AI anti-counterfeiting traceability management method for a packaged supercritical foaming material, comprising the following steps: collecting pressure change and temperature fluctuation data in a supercritical foaming material production process in real time through a sensor array, and synchronously collecting micro-texture image data of a material surface to obtain a joint production environment data set containing multi-dimensional time sequence and spatial image; according to the collected multi-dimensional time sequence, a convolutional neural network is used to extract feature correlation between pressure change, temperature fluctuation and material structure, and to determine an embedded vector captured dynamically; and the application aims to solve the technical problem that dynamic thermodynamic environmental variables and micro-bubble textures of a physical entity of a packaging material in a supercritical foaming production process are difficult to accurately map, leading to the fact that a digital anti-counterfeiting ledger and a physical entity cannot be verified in a fault-tolerant and tamper-proof consistent closed loop.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to an AI-based anti-counterfeiting and traceability management method for supercritical foaming materials used in packaging. Background Technology

[0002] In the modern packaging industry, the importance of anti-counterfeiting and traceability technology is self-evident. It is not only a key means of ensuring product safety and consumer rights, but also a core support for enhancing brand trust. Especially when using supercritical foaming materials as packaging, how to ensure the uniqueness and traceability of these materials through technological means has become a major challenge that the industry urgently needs to overcome. Research in this area is directly related to the security of commodity circulation and the stability of market order.

[0003] However, existing anti-counterfeiting and traceability methods on the market often face some deep-seated challenges. Many solutions, when recording and verifying material production information, lack the precise capture of dynamic changes in the production process, making it difficult to closely link complex process environments with material properties. This deficiency makes anti-counterfeiting labels easy to copy or tamper with, and the authenticity of traceability information difficult to guarantee. Especially when dealing with high-tech supercritical foaming materials, existing methods often cannot cope with the complexity brought about by the interaction of multiple variables in production.

[0004] More specifically, the core technical challenge in this field lies in effectively mapping key environmental factors in the production process to the physical characteristics of the materials themselves. The production of supercritical foamed materials involves real-time fluctuations in pressure and temperature, which directly affect the material's microstructure and surface properties. Without accurately capturing and recording these fluctuations, a unique identifier corresponding to the production process cannot be created. Furthermore, this lack of recording makes accurate traceability of the material difficult in subsequent use, as there is a lack of reliable evidence reflecting the entire production process. For example, in actual business operations, a batch of packaging materials may differ from other batches due to subtle changes in a key parameter during production, but current technology cannot deduce the specific production conditions from the material's surface characteristics, leading to a break in the traceability chain.

[0005] Therefore, how to establish a precise correspondence between dynamically changing environmental factors such as pressure and temperature and the physical properties of the material surface in the production of supercritical foamed materials, and how to achieve anti-counterfeiting and traceability functions through this correspondence, has become a key problem that this research urgently needs to solve. Summary of the Invention

[0006] This invention provides an AI-based anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging. Its purpose is to solve the technical problem that it is difficult to accurately map the dynamic thermodynamic environmental variables in the supercritical foaming process with the microscopic bubble texture of the physical entity of the packaging material, resulting in the inability to perform fault-tolerant and tamper-proof consistency closed-loop verification between the digital anti-counterfeiting ledger and the physical entity.

[0007] This invention provides an AI-based anti-counterfeiting and traceability management method for supercritical foaming materials used in packaging, the method comprising:

[0008] The pressure changes and temperature fluctuations during the production of supercritical foamed materials are collected in real time by a sensor array, and micro-texture images of the material surface are collected simultaneously to obtain a joint production environment dataset containing multi-dimensional time-series sequences and spatial images. Based on the collected multi-dimensional time-series sequences, a convolutional neural network is used to extract the feature correlations between pressure changes and temperature fluctuations and the material structure to determine dynamically captured embedding vectors. From the determined embedding vectors, micro-texture patterns of surface properties are obtained. If the similarity of the embedding vectors exceeds a preset threshold, the pattern is mapped to a uniquely identified encoded sequence. For the uniquely identified encoded sequences, blockchain technology is used to record an immutable log of the environmental factors and their corresponding relationships in the production process, obtaining a traceable distributed log. The process involves: 1) Obtaining log data from the distributed ledger; 2) Verifying whether pressure changes and temperature fluctuations are consistent with the surface properties of the material structure using a hash function, obtaining a Boolean value for the verification result; 3) Extracting corresponding relationship data related to technical challenges from the distributed ledger to determine the completeness of dynamic capture of environmental factors and establish reliability indicators for anti-counterfeiting and traceability; 4) Using recurrent neural networks to predict the impact of potential temperature fluctuations on unique identifiers during production, obtaining an optimized correspondence model; 5) Obtaining surface property adjustment parameters for the material structure from the optimized correspondence model; 6) Updating the traceability log of the distributed ledger to obtain the final anti-counterfeiting and traceability chain if the adjustment parameters meet preset conditions.

[0009] In one aspect of the invention, the method involves real-time acquisition of pressure changes and temperature fluctuations during the production process of supercritical foamed materials via a sensor array, and simultaneous acquisition of microscopic texture image data of the material surface, to obtain a joint production environment dataset containing multidimensional time-series sequences and spatial images, including: The pressure changes and temperature fluctuations during the production process of supercritical foamed materials are collected in real time by a sensor array to construct a multi-dimensional time series. A pre-established data processing module is used to denoise the acquired multidimensional time series to obtain a smoothed time series signal. If a pressure change exceeding a preset threshold is detected in the smoothed time-series signal, the abnormal time point is marked by the anomaly detection module to determine the abnormal range. Based on temperature fluctuation data within the abnormal range, pattern matching is performed using preset logical rules to determine whether there are potential production environment risks. By using time window analysis, local features are extracted from multidimensional time series signals within abnormal intervals to obtain key fluctuation patterns. For key fluctuation patterns, the support vector machine algorithm is applied for classification to obtain the classification results of production process stability; Based on the classification results, corresponding production environment adjustment parameters are generated and output to the control system to complete the automated response.

[0010] In one aspect of the present invention, the step of extracting the feature correlation between pressure changes and temperature fluctuations and material structure using a convolutional neural network based on the acquired multidimensional time series, and determining the dynamically captured embedding vector, includes: By using multi-dimensional acquisition methods, time-series data containing pressure changes and temperature fluctuations are obtained from sensors, stored as an initial dataset, and a structured time-series record is obtained. For structured time-series records, a convolutional neural network is used to perform hierarchical processing of pressure changes and temperature fluctuations, extract correlation features related to material structure, and determine a preliminary feature matrix. Based on the preliminary feature matrix, the spatiotemporal distribution patterns of key changes are analyzed, the mapping relationship between features is constructed, and an intermediate vector representing the trend of change is obtained. After obtaining the intermediate vector, the vector is compressed using a dimensionality reduction method to retain the main change information, and a dynamically captured embedding vector is constructed to determine whether it meets the preset expression requirements. If the dynamically captured embedding vector meets the preset expression requirements, it will be used as the core representation for subsequent analysis. If it does not meet the requirements, a second feature extraction is performed on the intermediate vector to obtain the adjusted embedding representation; By adjusting the embedded representation and combining it with the physical properties of the material structure, the influence patterns of pressure changes and temperature fluctuations on the structure are analyzed, and the final correlation description results are determined.

[0011] In one aspect of the present invention, the step of obtaining the micro-texture pattern of surface characteristics from the determined embedding vectors, and determining that if the similarity of the embedding vectors exceeds a preset threshold, the pattern is mapped to a uniquely identified encoded sequence, includes: Microscopic texture information related to surface properties is obtained from the stored embedded vector data, and a preliminary set of texture patterns is determined by parsing the vector structure. For the initial texture pattern set, the cosine similarity algorithm is used to calculate the similarity value between each embedding vector, and the similarity result between each pair of vectors is obtained; Based on the similarity results, if the similarity value of a pair of embedded vectors exceeds a preset threshold, the texture pattern corresponding to the pair of vectors is marked as a highly relevant pattern, and a list of highly relevant patterns is determined. For the list of highly relevant patterns, each highly relevant pattern is mapped to a unique encoding sequence through pattern transformation rules, resulting in a set of corresponding encoding sequences; From the set of encoded sequences, obtain the unique identifier information of each encoded sequence. By comparing the sequences, determine whether there are duplicate codes. If there are duplicates, perform sequence deduplication to obtain the final set of encoded sequences. Based on the final set of encoded sequences, a unique identifier record corresponding to the surface characteristics is generated, and the record is saved through a data storage tool to complete the encoding and mapping process of the texture pattern.

[0012] In one aspect of the invention, the encoding sequence for the unique identifier, using blockchain technology to record an immutable log of environmental factors and their corresponding relationships in the production process, to obtain a traceable distributed ledger, includes: For unique identifiers and coding sequences, a pre-established mapping mechanism is used to bind each link in the production process with a specific code, obtain the corresponding identifier data, and determine the basis for unique tracking of each production unit; Based on the acquired identification data, blockchain technology is used to collect and record environmental factors in the production process in real time, forming tamper-proof log entries and obtaining distributed storage units. For distributed storage units, the consensus mechanism of blockchain is used to synchronize and update the distributed ledger for each log entry, obtain consistent data records, and determine the integrity of data across multiple nodes; If data records are inconsistent across multiple nodes, the log entries are compared using timestamp verification and hash verification to obtain the corrected unified log data. Based on the restored unified log data, the correspondence between environmental factors and production processes is structured and stored to obtain a traceable query index and determine the environmental impact records of each production link. For the query index, smart contract technology is used to automatically verify traceable data, obtain access permissions and query paths, and determine whether the historical environmental factors of a specific coded sequence meet preset standards. Based on the verified query path, data is extracted from the log system in the distributed ledger to obtain the complete production process traceability chain and determine the final tamper-proof record content.

[0013] In one aspect of the invention, the step of obtaining log data from the distributed ledger, verifying whether pressure changes and temperature fluctuations are consistent with the surface properties of the material structure using a hash function, and obtaining a Boolean value of the verification result includes: Log data is extracted from the distributed ledger, and the recorded environmental parameters are classified and processed to separate the datasets of pressure changes and temperature fluctuations, resulting in a preliminary set of environmental parameters. For the initially compiled set of environmental parameters, a hash function is used to encrypt the data on pressure changes and temperature fluctuations, generating a corresponding hash value dataset to ensure data integrity. By using hash value datasets, pre-stored characteristic data of material surfaces are obtained and compared with hash values ​​of pressure changes and temperature fluctuations to determine data matching. If the comparison results show that the hash values ​​of pressure changes and temperature fluctuations are consistent with the material surface property data, then the output boolean value is true, confirming that the matching verification has passed; If the comparison results show that the hash value of pressure change or temperature fluctuation is inconsistent with the material surface characteristic data, the output boolean value is false, indicating that the matching verification has failed. Based on the Boolean output, corresponding verification records are generated and stored in the distributed ledger, forming a closed-loop data processing flow.

[0014] In one aspect of the invention, if the Boolean value of the verification result is true, then extracting the corresponding relationship data related to the technical challenges from the distributed ledger, determining whether the dynamic capture of environmental factors is complete, and determining the reliability indicators of anti-counterfeiting and traceability, includes: If the verification result is true, the corresponding data related to the technical challenges is extracted from the distributed ledger, and the extracted data is structured using data parsing tools to obtain a pre-organized data set. For the initially processed dataset, a dynamic environmental factor capture method is adopted to record and compare environmental information in the data in real time to determine the completeness of environmental factor capture. Based on the completeness of the environmental factors captured, it is determined whether there is any missing or abnormal data. If there is any missing or abnormal data, backup data is obtained from the distributed ledger through a preset supplementation mechanism to obtain a complete environmental information dataset. By using a complete environmental information dataset and combining it with the logical association rules for anti-counterfeiting and traceability, the data is verified in multiple dimensions to determine its authenticity and consistency, and to obtain the verified data results. Based on the verified data results, the reliability level of anti-counterfeiting and traceability is determined through logical calculation and comparison using the evaluation criteria of reliability indicators. Based on the reliability level, corresponding technical challenge solution data is generated, and the data is updated to the distributed ledger using a preset storage mechanism to complete the data closed-loop processing.

[0015] In one aspect of the invention, the step of using a recurrent neural network to predict the impact of potential temperature fluctuations on unique identifiers during the production process, based on a determined reliability index, to obtain an optimized correspondence model includes: By collecting data records of temperature fluctuation values ​​and temperature changes from multiple data collection points during the production process, these data are stored as an initial dataset to obtain a complete temperature change sequence. For temperature change sequences, a recurrent neural network is used for time series analysis to process the correlation between temperature fluctuation values ​​and potential forces, and to determine the quantitative results of the fluctuation impact. Based on the quantification results of the impact of fluctuations, combined with the data characteristics of unique identifiers, a preliminary model for optimizing correspondence is constructed to obtain the mapping parameters between temperature change and unique identifiers. If the mapping parameters do not match the preset threshold of the reliability index, the model is iteratively adjusted through a recurrent neural network to handle the data deviation of temperature fluctuation values ​​and judge the accuracy of the adjusted model. Based on the adjusted model accuracy, the potential impact of temperature fluctuations on the unique identifier is analyzed, evaluation data on the impact of fluctuations is generated, and the final optimized corresponding model is obtained. By optimizing the corresponding model, the real-time temperature changes during the production process are processed, the potential impact of fluctuations is predicted, and the reliability indicators of the real-time data are determined.

[0016] In one aspect of the invention, the step of obtaining surface characteristic adjustment parameters of the material structure from the optimized correspondence model, determining whether the adjustment parameters meet preset conditions, and then updating the traceability log of the distributed ledger to obtain the final anti-counterfeiting traceability chain includes: By optimizing the model, surface property data of the material structure are extracted from the correspondence to generate an initial set of adjustment parameters; Based on the initial set of adjustment parameters, the surface characteristic data are compared and analyzed to determine whether the adjustment parameters meet the preset conditions. If they do, a parameter record that meets the standard is generated. Obtain parameter records that meet the standards, and combine them with the traceability mechanism of the distributed ledger to match the parameter records with historical data to determine the updated content of the traceability log. For updates to the traceable logs, distributed ledger technology is used to write the data, resulting in an updated log dataset. Extract key information for anti-counterfeiting and traceability from the updated log dataset, construct the structural framework of the anti-counterfeiting and traceability chain, determine the integrity of the chain, and generate preliminary chain data if the chain is complete. By associating preliminary chain data with the corresponding relationship between characteristic parameters and the final chain, complete anti-counterfeiting and traceability chain data is generated. Obtain complete anti-counterfeiting and traceability chain data, store it to a designated node in the distributed ledger, and confirm that the final chain data has been synchronized.

[0017] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses an integrated solution to address the challenges of dynamically capturing, tamper-proof recording, and reliable verification of the relationship between environmental factors and material surface properties during production. The invention uses a sensor array to collect pressure changes and temperature fluctuations in real time, employs a convolutional neural network to extract feature associations and generate embedding vectors, which are then mapped to unique identifier encoding sequences. A distributed ledger is constructed using blockchain technology to ensure the immutability of production environment data. Simultaneously, a hash function verifies data consistency, a recurrent neural network predicts the impact of temperature fluctuations, optimizes the correspondence model, dynamically adjusts surface characteristic parameters, and updates traceability logs. The core innovation of this invention lies in the deep integration of AI algorithms and blockchain technology, enabling end-to-end management from data collection to anti-counterfeiting and traceability. Ultimately, this improves the anti-counterfeiting reliability and traceability of supercritical foamed material packaging, providing the industry with an efficient and intelligent quality assurance method. Attached Figure Description

[0018] Figure 1 This is a flowchart of the AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to the present invention.

[0019] Figure 2 This is a schematic diagram of the AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to the present invention.

[0020] Figure 3 This is another schematic diagram of the AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to the present invention. Detailed Implementation

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figures 1-3 The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in this embodiment may specifically include: Step S1: Real-time data on pressure changes and temperature fluctuations during the production process of supercritical foamed materials are collected using a sensor array, and microscopic texture image data of the material surface are collected simultaneously to obtain a joint production environment dataset containing multidimensional time series and spatial images.

[0023] Specifically, the sensor array includes temperature sensors and melt pressure sensors deployed in the melt reaction section and die section of the extruder. Physical quantities during the mixing stage of the supercritical fluid and polymer melt are acquired at a sampling frequency of 100 Hz. The acquired raw signals are processed by a Gaussian sliding window filter to extract a smooth time-series signal reflecting the physical foaming process. The filtering formula is as follows:

[0024] in: Represents the smoothed time-series signal variable; The index value represents the current discrete time point; Represents the index value Smoothed time-series signal data points at the location; Represents the constant value of pi; The parameter representing the standard deviation of the Gaussian function is set to 2.0 in this embodiment; The index of the step size variable representing the summation operation; The radius of the sliding window is 5 in this embodiment; The physical quantity signal variable representing the original acquisition; Represents the index value The original pressure change or temperature fluctuation data points collected at the location.

[0025] If a pressure change is detected after filtering... If the temperature exceeds the preset upper limit of the normal supercritical foaming threshold of 15.5 MPa or falls below the lower limit of 14.5 MPa, the time period will be marked as an abnormal interval by the microprocessor system.

[0026] This step addresses the problem that traditional industrial data acquisition methods often directly record raw sensor signals, failing to effectively remove high-frequency noise and electromagnetic interference from equipment mechanical vibrations commonly found in industrial settings. This leads to distorted physical baseline data and a lack of accurate data sources for subsequent feature mapping. By introducing a dual noise reduction mechanism of Gaussian sliding window and Fast Fourier Transform (FFT) frequency domain filtering, environmental artifacts are precisely removed, and a pure time-series signal that truly reflects the supercritical phase transition dynamics is obtained. This provides a high signal-to-noise ratio and high-fidelity physical data foundation for subsequent AI model feature extraction.

[0027] The upper and lower limits of the normal supercritical foaming threshold are determined based on a finite number of historical data sets, followed by a series of data processing steps. Specifically, the system pre-retrieves historical production logs of the equipment under normal yield conditions, extracts 10,000 pressure time-series data points in the stable foaming stage, calculates the mean and standard deviation of this sample set, and applies Laida's rule (i.e., 3... The criteria () set the upper limit as the mean plus three standard deviations and the lower limit as the mean minus three standard deviations, to cover 99.73% of the normal fluctuation range. For example, by analyzing the stable production history data of a certain type of extruder for one month, the mean melt pressure during this period was found to be 15.0 MPa and the standard deviation was 0.16 MPa. After calculating the range of fluctuation of the mean by about three standard deviations, the upper limit of the normal supercritical foaming threshold was finally set precisely at 15.5 MPa and the lower limit was set precisely at 14.5 MPa. Step S2: Based on the collected multidimensional time-series data, a convolutional neural network is used to extract the feature correlation between pressure changes and temperature fluctuations and the material structure, and to determine the embedding vector for dynamic capture.

[0028] After the supercritical foamed material is extruded, a laser confocal microscope in the sensor array simultaneously acquires microscopic bubble texture image data on the material surface. A convolutional neural network fuses multidimensional temporal sequence data with spatial image data. The convolutional neural network extracts feature associations through multiple convolutional operations; the calculation formula for its core convolutional layer is as follows:

[0029] in: The extracted associated feature map represents the spatial coordinates The activation value of the location; This represents a linear rectifier function, which is mathematically defined as taking the maximum value between the input value and 0. The width dimension represents the size of the convolution kernel; This represents the height dimension of the convolution kernel; Represents the traversal index along the width dimension; The index represents the traversal index along the height dimension; Represents coordinates Elements of the convolution kernel weight matrix at the specified location; The tensor representing the combination of multidimensional time-series data and image data input in coordinates Fragment value at; This represents the value of the bias term.

[0030] After dimensionality reduction using fully connected layers, a tensor containing 64-dimensional floating-point numbers is output, which is defined as a dynamically captured embedding vector. This step addresses the inherent modal gap between macroscopic thermodynamic time-series signals (one-dimensional time) and microscopic bubble geometry (two-dimensional space). Existing technologies struggle to establish a rigorous causal mathematical connection between the two, leading to a lack of physical basis for anti-counterfeiting identifiers. Based on a rheological spatiotemporal alignment mechanism, combined with a dual-branch CNN and a cross-attention network, this approach successfully overcomes the temporal and spatial modal barriers. It fuses thermodynamic nucleation kinetics and bubble morphology features within the same high-dimensional embedding space, quantifying the correlation between thermodynamic fluctuations and microscopic morphology. Step S3: Obtain the micro-texture pattern of surface characteristics from the determined embedding vectors. If the similarity of the embedding vectors exceeds a preset threshold, map the pattern to a unique identifier encoding sequence.

[0031] A subset of data representing the microscopic bubble pore size distribution and pore wall thickness characteristics is extracted from the embedding vectors. The cosine similarity algorithm is used to calculate the similarity of the batch-obtained embedding vectors; the calculation formula is as follows:

[0032] in: Represents two dynamic capture embedding vectors and Similarity calculation results between them; This represents the total number of dimensions of the embedded vectors; in this embodiment, the value is 64. The index representing the current dimension's traversal; Representative vector In the The specific component values ​​of the dimension; Representative vector In the The specific component values ​​of the dimension; If the similarity calculation result If the value is greater than or equal to a preset threshold of 0.95, the corresponding microtexture pattern is determined to be a highly correlated pattern. A locality-sensitive hashing algorithm is then used to map the multidimensional vector of this highly correlated pattern into a unique identifier sequence consisting of 256 hexadecimal characters.

[0033] This step addresses the issue that in large-scale continuous production, there is a very low probability that different batches of materials may have similar physical regions on their surfaces. If the anti-counterfeiting code is generated solely based on geometric features, it is easy to cause hash collisions, which would compromise the global uniqueness of the traceability system. After using cosine similarity for high-dimensional feature screening, the collision risk is assessed using Levenstein distance. Furthermore, the absolute geographic coordinate system that generates the micro-texture is introduced and XORed with a millisecond-level timestamp. By combining mathematical probability with physical spatiotemporal information, the uniqueness and anti-counterfeiting capability of the anti-counterfeiting code sequence are improved.

[0034] The preset threshold value is determined based on a limited number of historical data sets, followed by a series of data processing steps. Specifically, the system pre-collects 1000 sets of surface images of supercritical foamed materials known to be produced in the same and different batches. These images are then input into a feature extraction model to obtain corresponding embedding vectors. The system calculates the cosine similarity distribution characteristics between vectors from the same batch and between vectors from different batches. By plotting the normal distribution probability density curves of intra-class similarity and inter-class similarity, the value at the intersection of the two curves, or the value that minimizes the sum of the false rejection rate and the false acceptance rate, is set as the preset threshold. For example, in historical data testing, the mean cosine similarity of 500 sets of genuine samples from the same batch is concentrated above 0.98, while the mean similarity of 500 sets of counterfeit or unrelated samples from different batches is around 0.70. To ensure that the system has extremely high anti-counterfeiting recognition accuracy while also taking into account a certain degree of physical fault tolerance, the preset threshold is precisely set to 0.95 after statistical optimization calculations.

[0035] Step S4: For the uniquely identified coded sequence, blockchain technology is used to record the environmental factors and corresponding relationships of the production process in an immutable log, thereby obtaining a traceable distributed ledger.

[0036] The generated unique identifier's encoded sequence is used as the primary key. The corresponding pressure change data, temperature fluctuation data, production timestamp, and batch number are serialized into JSON format key-value data. A consortium blockchain network based on a practical Byzantine fault-tolerant consensus mechanism is used to write this key-value data into the block. The hash generation formula for the block header is as follows:

[0037] in: Represents the target hash value of the currently generated block; Represents a 256-bit operation function for a secure hash algorithm; Represents the hash value of the previous block in the blockchain network; The string concatenation operator; This represents the Merkle tree root node hash value, which contains data on all environmental factors and their corresponding relationships within the current block. The timestamp value representing the system's data packaging action; This represents a random value used to satisfy the proof-of-work requirement of the consensus mechanism.

[0038] Through this step, each individual packaging entity of the supercritical foam material generates a corresponding evidence log on the blockchain, forming a traceable distributed ledger.

[0039] This step addresses the problem that directly uploading massive amounts of physical time-series data generated by high-frequency industrial sensors to the blockchain would lead to severe network congestion and latency, generate significant storage space overhead, and severely slow down traceability and query efficiency. It employs statistical feature extraction combined with Merkle tree data structures for edge computing anchoring, achieving extremely high compression rates for massive physical data. While absolutely ensuring data immutability, it reduces the computational complexity of blockchain verification and terminal traceability from a linear level. Reduced to logarithmic level This improved the system's data processing efficiency.

[0040] Step S5: Obtain log data from the distributed ledger, verify whether the pressure changes and temperature fluctuations are consistent with the surface characteristics of the material structure through a hash function, and obtain the Boolean value of the verification result.

[0041] When a tracing request is initiated, the microprocessor system extracts pressure change and temperature fluctuation data corresponding to the encoded sequence of the unique identifier from the distributed ledger. A one-way secure hash function is used to verify and calculate the feature vector reconstructed from the data, as shown in the following formula:

[0042] in: Output variables representing the validation results; Represents the "true" state in Boolean logic; Represents the "false" state in Boolean logic; This represents the "if" condition in a conditional statement. This represents the "otherwise" condition in a conditional statement. This represents the function that calculates the Hamming distance between two strings. This represents a hash function that quantizes a continuous floating-point vector into discrete string features; This represents the actual embedding vector generated by scanning the surface texture of the physical packaging material using a confocal microscope. This means that historical pressure changes and temperature fluctuation data read from the distributed ledger will be input into the predictive embedding vector generated by the model; Represents the less than or equal to relational operators; This represents the tolerance threshold for the allowed Hamming distance set by the system; in this embodiment, the value is 3.

[0043] The preset tolerance threshold The value is obtained based on a limited number of historical data, followed by a series of data processing steps. Specifically, the system performs 2000 historical repeated scans on the same standard packaging material sample under different ambient light intensities, different tolerance parameters of optical scanning instruments, and different scanning angles. It obtains the local sensitive hash values ​​of these 2000 sets of actual embedded vectors containing physical reading noise, compares them with the standard predicted hash value, and calculates the Hamming distance. The maximum distance offset covering more than 98% of legitimate scanning error scenarios is used as the tolerance threshold. For example, in historical experiments simulating actual industrial field inspections, due to changes in dust and low light, the hash signature of the same material will have an average of 1 to 2 bits that jump in 256 bits. Historical data shows that the Hamming distance of 98.5% of legitimate repeated scan errors will not exceed 3. Therefore, in order to achieve a balance between anti-counterfeiting strictness and on-site inspection convenience, the preset tolerance threshold for the allowed Hamming distance is set to 3.

[0044] If the calculated Hamming distance is less than or equal to 3, the output Boolean value is true; otherwise, the output Boolean value is false.

[0045] This step addresses the problem that traditional strong avalanche effect cryptographic hash functions are extremely sensitive to input and cannot tolerate the measurement tolerances, angular deviations, and environmental noise inherent in the actual physical optical scanning process of terminal verification equipment, leading to a very high false negative rate where genuine products are mistakenly identified as counterfeits. It introduces a Locality Sensitive Hashing (LSH) algorithm based on random projection, which reduces continuous high-dimensional physical features to Boolean signatures that allow for specific Hamming distance tolerances. This approach is mathematically compatible with physical measurement errors, achieving entity-ledger consistency verification that combines high security with high physical robustness.

[0046] Step S6: If the Boolean value of the verification result is true, extract the corresponding relationship data from the distributed ledger, determine whether the dynamic capture of environmental factors is complete, and determine the reliability index of anti-counterfeiting and traceability.

[0047] like When the status is true, the system extracts data containing the correspondence between sensor data packet loss rate and sensor calibration drift parameter. The system uses a multivariate reliability assessment formula to determine the reliability index of anti-counterfeiting and traceability, as follows:

[0048] in: The calculation result of the reliability index representing anti-counterfeiting and traceability has a value range between 0 and 1; The weighting coefficient variable representing data integrity is set to 0.6 in this embodiment; The weighting coefficient variable representing environmental noise control is set to 0.4 in this embodiment; This represents the number of sampling points lost in multidimensional time-series data due to transmission failures. This represents the total number of complete data points that should theoretically be collected within a set period. The scalar parameter representing the attenuation sensitivity of the system is set to 0.05 in this embodiment; This represents the combined variance of pressure changes and temperature fluctuations within the marked outlier range.

[0049] This step addresses the issue of uncontrollable anomalies such as brief data loss from sensors or equipment calibration drift frequently occurring in harsh industrial production environments. Directly accepting such flawed data would severely reduce the technical reliability and legal validity of the anti-counterfeiting and traceability chain. A multivariate evaluation model based on data integrity rate and composite noise variance is established to provide a rigorously quantified reliability confidence level for a single traceability data collection process. It automatically filters out batches of poor-quality data, providing a rigid mathematical criterion for whether subsequent AI prediction models should accept the historical data.

[0050] Step S7: Based on the determined reliability index, a recurrent neural network is used to predict the impact of potential temperature fluctuations on the unique identifier during the production process, and an optimized correspondence model is obtained.

[0051] A Long Short-Term Memory (LSTM) network is used as the specific execution unit of the recurrent neural network. When the value is greater than or equal to 0.85, the prediction task of the Long Short-Term Memory (LSTM) network is initiated. The formula for controlling the state update of the forget gate within the network is as follows:

[0052] in: Representing the Long Short-Term Memory network in discrete time steps The forget gate output vector matrix at time step; This represents the Sigmoid non-linear activation function, used to map the input to a closed interval between 0 and 1; This represents the parameter weight matrix corresponding to the forget gate; Represents the vector concatenation operator; Representative at the previous time step The hidden state feature vectors transmitted by the network at each time step; Represents the current time step The temperature fluctuation sampling data tensor input at any given time; This represents the constant vector of the bias term in the current forget gate structure.

[0053] After forward propagation and reverse weight update based on reliability indicators, the Long Short-Term Memory network outputs a mapping matrix for the temperature deviation and the corresponding parameters of microtexture porosity, thus obtaining an optimized correspondence model.

[0054] This step addresses the significant thermal inertia and hysteresis effects inherent in the supercritical foaming process of polymers. Traditional systems often produce a large number of irreversible defective products by the time they detect abnormal material textures caused by temperature deviations, lacking proactive quality intervention capabilities. This approach fully leverages the powerful long-term time-dependent learning capabilities of LSTM networks and introduces a reliability index as a regularization penalty to accurately capture minute drift trends in thermodynamic parameters. This model can overcome thermal inertia time differences and predict the adverse effects of temperature fluctuations on material porosity in future time windows, upgrading anti-counterfeiting quality control from post-event traceability to pre-event prediction.

[0055] The preset threshold for this reliability index is determined based on a limited number of historical data sets, followed by a series of data processing steps. Specifically, the system backtracked production and model prediction logs for up to six months, and performed Pearson correlation analysis on the reliability index scores calculated under different data integrity and noise levels with the mean square error (MSE) of the subsequent long short-term memory network (LSTM) prediction temperature deviation values. This analysis identified the minimum reliability score that the front-end data source must possess to meet the actual adjustment needs of the equipment when the accuracy of the prediction model is sufficient. For example, analysis of 500 historical modeling processes revealed that when the reliability index of the input data is below 0.85, the prediction error of the time-series network caused by packet loss and noise accumulation amplifies dramatically, with the MSE soaring from 0.02 to over 0.15. The resulting adjustment parameters often cause equipment oscillations. Therefore, based on the regression analysis results, the system sets the preset threshold for the reliability index used to start network training at 0.85.

[0056] In practical implementation, the system first obtains the fluctuation impact degree output in step S7. (i.e., the predicted porosity deviation percentage). To convert the digital spatial prediction results into control inputs for physical devices, this embodiment uses preset calibration coefficients. The predicted porosity deviation percentage is converted into an error function value representing the deviation of the microtexture from the standard deviation. The specific mathematical conversion relationship is as follows: After obtaining the error function value, the optimized correspondence model is used to calculate the control variables for the next cycle of the production equipment.

[0057] Step S8: Obtain the surface property adjustment parameters of the material structure from the optimized correspondence model. If the adjustment parameters meet the preset conditions, update the traceability log of the distributed ledger to obtain the final anti-counterfeiting traceability chain.

[0058] Using the optimized correspondence model, the control variables for the next cycle of the production equipment are calculated. Surface characteristic adjustment parameters are extracted, specifically the suggested screw speed increment of the extruder. The control logic calculation formula is as follows:

[0059] in: The result of the surface characteristic adjustment parameter calculation represents the output. The proportional gain constant in the proportional-integral-derivative control algorithm; This represents the integral gain constant in the proportional-integral-derivative control algorithm. The differential gain constant represents the proportional-integral-derivative (PI-DI) control algorithm. Represents the multiplication operator; The error function value representing the deviation of the micro-texture acquired at the current moment from the standard deviation; Represents a continuous-time variable; Represents the definite integral operator; The differential element representing the time variable; This represents the derivative operator with respect to the time variable.

[0060] Determine the output of the microprocessor If the absolute value is less than the limit safety value of the equipment's motor control system, and the preset condition is met, the parameter adjustment command will be sent to the production equipment for execution to stabilize production. Subsequently, the original environmental data characteristics, model prediction parameters, and actual adjustment parameters of the current batch will be merged and packaged, encapsulated through hash operation, and written to the blockchain distributed node. The completed node log is associated with the previous product static feature records to form the final anti-counterfeiting and traceability chain.

[0061] This step addresses the problem that most existing traceability systems are merely one-way "data loggers" (open-loop systems). These systems cannot use cloud-based analysis results from the digital space to guide the physical production processes in the workshop, causing anti-counterfeiting technology to remain at the inspection level and unable to be transformed into stable production control. This step breaks down the barrier between digital prediction and physical compensation, establishing a fully automated closed-loop control logic of environmental disturbance identification, impact prediction, and physical equipment compensation (such as sending a PLC to adjust screws or water valves). The system not only actively adjusts physical parameters to maintain the stable generation of anti-counterfeiting features, but also encrypts this anti-interference dynamic control action itself on the blockchain, forming a complete anti-counterfeiting closed loop forged by physical entities, a distributed ledger, and control logic, eliminating the possibility of counterfeiting.

[0062] In summary, this embodiment addresses the technical challenge of accurately binding environmental variables to physical entities in the anti-counterfeiting and traceability process of supercritical foamed materials for packaging. Step S1 first uses a high-frequency sensor array deployed on the extruder to collect real-time pressure and temperature time-series data during production at a sampling frequency of 100Hz, for example, monitoring standard pressure fluctuations of 15.0MPa. Considering that complex electromagnetic interference in actual industrial workshops often results in a large amount of high-frequency spike noise in the original collected data, this solution uses a Gaussian sliding window filtering model based on y[n] to smooth the time-series data. This calculation logic is adopted because the Gaussian distribution function can dynamically allocate smoothing weights according to the distance of time nodes from the center point, effectively eliminating noise caused by mechanical vibration. This solves the technical problem that distortion of the original signal easily masks the subtle characteristics of the actual thermodynamic phase transition. The beneficial effect is that it provides a high-fidelity multidimensional time-series sequence that truly reflects the physical foaming evolution process for subsequent analysis. After obtaining reliable basic data, step S2 further uses a dual-branch convolutional neural network to extract associated features. The core logic of this algorithm model is to perform a dot product between a specific weight matrix and the input tensor and apply the ReLU activation function to fuse one-dimensional time-series thermodynamic features with two-dimensional optical micro-texture images across modes. This operation solves the pain point of the difficulty in quantifying the complex mapping relationship between the dynamic fluctuations of the macro environment and the physical topology of the micro-bubbles in the material. The output 64-dimensional feature vector provides a robust digital description for anti-counterfeiting.

[0063] Based on the 64-dimensional feature vectors obtained above, step S3 introduces a cosine similarity algorithm to perform a measurement calculation of the vector space. In the actual packaging and processing scenarios of supercritical foamed materials, the materials inevitably undergo physical cutting or slight extrusion deformation. The calculation logic of cosine similarity (i.e., calculating the cosine value of the angle between two vectors) is adopted because it focuses more on the directional consistency of vectors in high-dimensional space rather than their absolute numerical magnitude, thus solving the problem of high misjudgment rate caused by slight physical changes in materials when using Euclidean distance for strict comparison. When the similarity reaches, for example, 0.95 or higher, the system uses locality-sensitive hashing to convert it into a 256-bit unique identifier encoding sequence, achieving the effect of generating a stable digital fingerprint with a certain degree of morphological fault tolerance for the physical entity. To prevent this traceability information from being tampered with by centralized institutions, step S4 packages the unique identifier along with its corresponding time-series feature sequences such as pressure and temperature and submits it to the blockchain network. By constructing a Merkle tree structure by combining data in pairs at different levels to calculate SHA-256 hash values ​​and writing information containing the Merkle root into the header of a new block, this cryptographic logic utilizes the one-way and collision-resistant properties of hash functions to solve the data trust crisis among multiple participants in the supply chain. It ensures the synchronous update of the correspondence log between environmental factors and unique identifiers among distributed nodes, and constructs a data-consistent underlying ledger system with anti-counterfeiting and traceability credibility.

[0064] After establishing an immutable distributed ledger, step S5 specifically designs a verification algorithm based on a dimensionality reduction hashing mechanism to meet the verification requirements of real-world business scenarios that demand both high efficiency and physical tolerance. When end users or inspection agencies need to verify the authenticity of a batch of packaging materials, the system extracts the texture feature vector obtained from the current actual scan, as well as the feature vector reconstructed by retrieving historical environmental parameters from the ledger and inputting them into the model for prediction. To complete consistency verification without directly comparing high-dimensional floating-point data or revealing core neural network parameters, this scheme introduces the SimHash random projection algorithm to map high-dimensional features into binary hash signatures and calculate the Hamming distance between them. This verification logic of dimensionality reduction projection and calculating the difference in bits (Hamming distance) is adopted because traditional strict cryptographic hashing has a strong avalanche effect; even extremely small optical scanning tolerances at the input end can lead to drastically altered output values. By allowing a tolerance space for the Hamming distance, for example, with a threshold of 3, this computational design resolves the technical contradiction between the strict matching mechanism of digital information systems and the objective existence of measurement errors in physical entities from a mathematical perspective. Under the premise of reasonably tolerating normal equipment sampling noise, it can accurately determine whether the historical records of pressure and temperature fluctuations match the surface characteristics of materials. The result is that it improves the practicality and robustness of on-site anti-counterfeiting verification.

[0065] To evaluate the data quality foundation of the aforementioned traceability link, step S6 constructs a multivariate reliability assessment formula that includes data loss rate and environmental noise variance. In actual production, sensors occasionally experience network outages or data packet loss. This formula employs a comprehensive evaluation logic, assigning a weight of 0.6 to the packet loss rate and a weight of 0.4 to the composite variance, which reflects the degree of environmental interference. This addresses the problem of previous traceability systems blindly trusting underlying data while ignoring the health status of the acquisition channels. When the calculated reliability index reaches, for example, 0.85 or higher, it indicates that the current time-series record has high reference value. Next, step S7 utilizes a Long Short-Term Memory (LSTM) network to perform time-series modeling analysis on the highly reliable historical temperature data. In this network model, the forget gate formula outputs a state vector between 0 and 1 through the Sigmoid activation function. Its internal calculation logic adaptively determines which historical temperature fluctuation information needs to be retained and which noise information should be discarded based on the current input and the hidden state of the previous time step. This gating mechanism effectively solves the gradient vanishing phenomenon that traditional time series models are prone to when dealing with ultra-long physical foaming processes. It enables the model to accurately predict the quantitative impact of potential small temperature drifts on the pore structure and microstructure of supercritical foamed materials. The result is that it provides a forward-looking optimized correspondence model with a certain time lead for subsequent physical closed-loop regulation.

[0066] The final closed loop of the entire anti-counterfeiting and traceability chain is implemented in the specific physical equipment adjustment and on-chain record solidification in step S8. The system utilizes the optimized correspondence model output from the above steps to proactively calculate the physical error range that may cause micro-texture deviations from standard features in the future. To address this error, this solution employs a proportional-integral-derivative (PID) control formula to calculate the adjustment parameters of the material structure surface characteristics that need compensation, such as the screw speed increment or back pressure compensation value of the extruder. The PID algorithm uses a calculation logic that combines the current error term, the integral term of the accumulated error, and the derivative term of the error change rate. This solves the problem of system response lag or overshoot oscillation that can easily occur with a single control method, and can quickly and smoothly offset environmental deviation factors in the initial stage of physical foaming. When it is determined that the calculated adjustment parameters are within the safety-allowed threshold range of the equipment, the system will issue a compensation command for execution. Simultaneously, it extracts the predicted compensation action for temperature fluctuations, environmental data, and the micro-identification code of this batch of physical materials, uses a hash algorithm to correlate them, and records them in a distributed ledger. This end-to-end design overcomes the limitations of traditional anti-counterfeiting technologies, which are limited to passive post-event coding and recording and cannot intervene in production quality maintenance. The beneficial effect is that it integrates the static physical characteristics of the product with the dynamic anti-interference control history of the production line, forming a more rigorous traceability management system.

[0067] In some embodiments, the sub-steps of step S1 are described in detail in this embodiment, specifically as follows: A multi-dimensional time series is constructed by real-time acquisition of pressure changes and temperature fluctuations during the production process of supercritical foamed materials using a sensor array. This assumes a continuous... From the production batch, obtained Each data sampling point. A pre-established data processing module is used to denoise the acquired multidimensional time series, obtaining a smoothed time series signal. In addition to Gaussian filtering, a Fast Fourier Transform (FFT) is introduced to implement low-pass filtering for the high-frequency noise of periodic mechanical vibration caused by the equipment motor.

[0068] in: The representative signal, after transformation, is in the frequency domain at the 1st... The complex number representation of the value at each frequency index; Represents the total number of sampling points for a discrete-time signal; Represents the index variable of discrete sampling points in the time domain; Represents in the index Discrete data of pressure changes or temperature fluctuations in the time domain; It represents the basic unit of the imaginary number in the complex number field; Represents the index variable of discrete frequency components in the frequency domain.

[0069] After setting the coefficients of the high-frequency components above 20 Hz in the frequency domain to 0, the signal is restored to the time domain signal by inverse fast Fourier transform.

[0070] If the pressure change detected in the smoothed time series signal exceeds the preset threshold (set to an upper limit of 1), Then, the anomaly detection module marks the abnormal time points to determine the abnormal range. By employing time window analysis, local feature extraction is performed on multidimensional time-series signals within abnormal intervals to obtain key fluctuation patterns. Wavelet transform is used to extract abrupt change features.

[0071] in: This represents the wavelet coefficient values ​​calculated under a specific scaling and translation factor. The scaling factor scalar represents the scaling width of the wavelet function; The scalar represents the shift factor that controls the position of the wavelet function on the time axis; The generalized integral operator represents the range from negative infinity to positive infinity; This represents the time-series signal function within the abnormal interval after filtering in the preceding steps. Represents a continuous-time variable; This represents the complex conjugate form of the selected mother wavelet function; An integral infinitesimal element representing a time variable.

[0072] For the wavelet coefficient feature set of key fluctuation modes, the support vector machine (SVM) algorithm is applied for classification processing to separate categories such as "normal supercritical phase transition fluctuation", "nucleating agent agglomeration fluctuation" and "melt fracture risk fluctuation", and finally obtain the stability classification results of the production process. The logs corresponding to the high-risk classification are prepared to be uploaded to the blockchain.

[0073] Step S2 specifically includes the following sub-steps: First, multidimensional time-series data containing pressure changes and temperature fluctuations are acquired through multidimensional acquisition methods. Because time-series data (time...) ) and optical scanning of micro-texture (space) Since they exist in different dimensions, the system needs to be spatiotemporally aligned based on the material flow velocity principles of the extruder. The average extrusion velocity is determined according to rheology. Place the optical scanner at the location Images captured at that location correspond to a time window The time-series data within the dataset includes the estimated time delay as follows:

[0074] Aligned time-series-spatial image data pairs are formed and stored as the initial dataset, resulting in structured records.

[0075] For structured records, the system uses a two-branch convolutional neural network for hierarchical processing: Temporal feature extraction branch: A one-dimensional convolutional neural network (1D-CNN) is used to process the aligned pressure change and temperature fluctuation sequences. Extract eigenvectors characterizing nucleation dynamics :

[0076] Spatial feature extraction branch: A two-dimensional convolutional neural network (2D-CNN) is used to process the corresponding micro-texture images. Extract feature vectors representing cell morphology. :

[0077] After extracting the associated features related to the material structure, a mapping relationship between the features is constructed. Cross-Attention or bilinear pooling is used to calculate the interaction features of the two modalities, obtaining an intermediate vector representing the changing trend. :

[0078] in It is a learnable bilinear weight matrix.

[0079] After obtaining the intermediate vectors, dimensionality reduction methods such as principal component analysis (PCA) or fully connected layer truncation are used to further refine the vectors. After compression, the output is a 64-dimensional dynamic capture embedding vector. To determine whether it meets the preset expression requirements, the system calculates the information entropy of the vector. To measure its feature richness:

[0080] in The first digit after Softmax normalization of the embedding vector 3D probability distribution.

[0081] like (The preset expression requirement threshold) indicates that the feature vector has sufficient discriminative power, and it can be used as the core representation for subsequent anti-counterfeiting hash mapping analysis; If the feature fusion is insufficient at that network layer, the system introduces edge features from shallow layers through residual connections and performs secondary feature extraction on the intermediate vectors to obtain an adjusted embedding representation. The final output, dynamically captured embedding vector, mathematically achieves a precise mapping description from macroscopic thermodynamic fluctuations to microscopic geometric forms.

[0082] The preset expression requirement threshold is determined based on a limited number of historical data sets, followed by a series of data processing steps. Specifically, the system extracts 800 sets of intermediate layer embedding vectors generated during the iterative training of the deep learning network, calculates the information entropy value of each vector, and performs joint clustering analysis on these information entropy values ​​and their correct recognition rate in the downstream anti-counterfeiting identification task to find the minimum information entropy inflection point that can support a high confidence classification output of over 99%. For example, data processing on historical feature datasets shows that when the information entropy of the embedding vector is below 3.5, the information content of the vector is too singular, which will lead to a large number of duplicate codes in the downstream hash mapping. Only when the information entropy distribution reaches 3.8 or above can the feature richness meet the unique expression requirement. Based on this statistical performance, the system sets the preset expression requirement threshold for measuring feature richness to 3.8.

[0083] In some embodiments, the detailed process of performing deduplication verification of the uniquely identified encoded sequence in the sub-step of step S3 is as follows: After mapping highly correlated patterns to encoded sequences, unique identifiers of the encoded sequences are obtained, and sequence traversal and comparison are performed. The Levenstein distance formula is used to measure the difference between two encoded sequence strings, as follows:

[0084] in: Represents the encoded sequence string Take before Characters and encoded sequence string Take before The edit distance value calculated between each character; The string variable representing the first unique identifier encoded sequence of the comparison; The second unique identifier encoded sequence string variable that is being compared; Represents a string The index of the current substring character length; Represents a string The index of the current substring character length; This represents a function that extracts the maximum value from a given set of values. This represents the "if" condition in a conditional statement. A function that extracts the minimum value from a given set of values; This represents a conditional indicator function, specifically defined as: when the string... In the Characters at each position With strings In the Characters at each position When the signs are not equal, the function outputs 1; otherwise, when the signs are equal, the function outputs 0. Represents a string In the index position A single character; The not-equals relational operator; Represents a string In the index position A single character; It represents the "otherwise" clause in conditional judgment logic.

[0085] The microprocessor system traverses the database and performs the aforementioned distance calculation. When it finds a distance calculated for the encoded sequences of two different physical entities... When the value is less than a preset distance threshold, the system determines that there is a duplicate encoding. At this point, the microprocessor system extracts the two-dimensional coordinate data and a timestamp value with millisecond-level precision that generated the microtexture. The system converts the extracted two-dimensional coordinate data and timestamp value into a binary stream and performs a bitwise XOR operation with the original encoded sequence string. The new binary string output by the XOR operation is stored in the distributed ledger as the final unique identifier to complete the deduplication process of the encoding mapping.

[0086] In some embodiments, this embodiment further illustrates how to efficiently and securely map high-frequency collected physical environment data to a distributed ledger to solve the network congestion and storage redundancy problems caused by uploading massive amounts of industrial IoT data to the blockchain.

[0087] In step S4, for the uniquely identified encoded sequence, the system does not include all the original high-frequency data (such as...) Instead of directly uploading the sampled pressure and temperature data to the blockchain, a Merkle tree-based data compression and anchoring technique is used. This involves the following sub-steps: For a preset time window (e.g.) Within a production unit, an edge computing node performs statistical feature extraction on all pressure changes and temperature fluctuations within that time window, generating an environmental feature vector. ,in The mean, Standard deviation This represents the accumulated energy within the abnormal range.

[0088] The system will match the encoded sequence of the unique identifier with the corresponding environmental feature vector. The data is then combined to form the basic transaction data. Each transaction is then hashed to generate leaf nodes. The hash values ​​of any two adjacent nodes are concatenated and hashed again until a unique MerkleRoot is generated.

[0089] in: The Merklegen hash value representing the current block serves as the core credential for the immutable log of environmental factors in the production process of this batch. Represents the SHA-3 cryptographic hash function; Representing the The hash value of the basic transaction data of each production unit; This represents a byte string concatenation operation.

[0090] Using blockchain technology, The block height and timestamp are packaged to generate a new block. Through distributed storage units and a consensus mechanism, the distributed ledger is synchronized and updated. When tracing is required, only a hash path (MerklePath) of a specific encoded sequence needs to be provided. It completes the integrity and tamper-proof verification of data across multiple nodes within a time complexity, and obtains consistent data records.

[0091] In step S5, log data from the distributed ledger is obtained, and the pressure changes and temperature fluctuations are verified using a hash function to determine whether they are consistent with the surface characteristics of the material structure. This step includes the following sub-steps: Extract the environmental feature vectors and historical multidimensional time series sequences of the target batch from the distributed ledger. Input this sequence into the inference copy of the convolutional neural network trained in step S2 to generate predicted embedding vectors. (64-dimensional floating-point number).

[0092] Simultaneously, the terminal inspection equipment scans the material entity and extracts the actual embedding vector. To convert floating-point vectors into comparable hash boolean values, a random projection hash map (SimHash) is introduced: Construct a dimension as Gaussian random projection matrix For a given embedding vector (Whether it's an actual scan or a predicted one), its Bit hash signature The calculation process is as follows:

[0093]

[0094] in: This represents the intermediate vector after projection; Represents the first of the vectors One component; The first character representing the final hash signature string Bit Boolean value.

[0095] Then, the actual signature is calculated. With predictive signature Hamming distance between If the comparison results show:

[0096] in This represents a preset tolerance threshold (e.g., set to) based on the dynamic calibration of the optical scanner's accuracy. If the physical generation process and the digital ledger record are not matched, the output Boolean value will be True, confirming that the verification is successful. If not, the output Boolean value will be False. This fundamentally establishes an equivalent verification path between macroscopic thermodynamic parameters and microscopic geometric features in the digital space.

[0097] In some embodiments, the solubility and diffusion coefficient of supercritical fluids are extremely sensitive to temperature. To ensure the quality stability of anti-counterfeiting marks (i.e., microtextures) in continuous production, potential temperature fluctuations must be predicted and corrected before they cause physical deformations to exceed tolerances.

[0098] In step S7, based on the determined reliability index, a recurrent neural network is used to predict the impact of potential temperature fluctuations during the production process on the unique identifier. This specifically includes the following sub-steps: Obtain the complete temperature change sequence Construct a Long Short-Term Memory (LSTM) network and analyze its internal cell states. The updates not only rely on historical temperatures, but also introduce reliability indicators characterizing anti-counterfeiting and traceability. As a regularization term, the core state update formula for the forward propagation of LSTM is as follows:

[0099]

[0100] in: and These represent the cell states at the current and previous moments, respectively, and are used to preserve long-term memory. These represent the activation vectors for the forget gate, input gate, and output gate, respectively. Weight matrix and bias representing cell state; Represents the Hadamard product (element-wise matrix multiplication). The output represents the hidden state at the current moment, which contains the trend characteristics of temperature fluctuations.

[0101] Will A fully connected regression layer is used to map the result to the quantification of the impact of volatility. (i.e., the predicted percentage deviation of material porosity):

[0102] Constructing a system that includes a custom loss function The optimized model. This loss function not only minimizes the temperature prediction error, but also penalizes... Behaviors exceeding the anti-counterfeiting label recognition threshold:

[0103] in: and These are the predicted temperature and the actual temperature, respectively. This represents the penalty weight hyperparameter; This represents the maximum permissible physical deviation (preset physical boundary) that allows the SimHash algorithm to reliably identify micro-textures.

[0104] The optimized correspondence model is obtained by iteratively adjusting the network through backpropagation through time (BPTT) and the Adam optimizer until the network converges. This model can predict the network path in advance. The time window predicts temperature drift, providing a decision margin for closed-loop control.

[0105] In some embodiments, this embodiment addresses the endpoint of the entire system logic, namely, how to use predictive models to maintain the generation quality of the next batch of anti-counterfeiting labels and solidify the final traceability chain.

[0106] Step S8 specifically includes the following sub-steps: Using the optimized correspondence model, we can predict the impact of fluctuations at future times. Approaching the threshold The system will extract surface property adjustment parameters of the material structure from the correspondence to offset the physical deviation. In the supercritical foaming process, these adjustment parameters are specifically converted into the screw speed increment of the extruder. Or the opening degree change of the cooling water flow control valve .

[0107] The system derives the adjustment command vector through calculation. If all indicators of the vector are within the equipment's limit safety constraints (meeting preset conditions), then the data is sent to the PLC (Programmable Logic Controller) to perform automated compensation.

[0108] Once the compensation command is executed, the system extracts the entire process record of "anti-counterfeiting quality early warning - parameter adjustment calculation - physical equipment compensation" and generates a control event log. The log should be linked to the unique microtexture identifier of this batch of products and the environmental Merkelgen. Perform association matching:

[0109] in This represents the final digital anchor point in the anti-counterfeiting and traceability chain. Using distributed ledger technology, this anchor point is broadcast across nodes and written into blocks.

[0110] Thus, the anti-counterfeiting label generated by this invention is no longer a traditional physical label, but a physical characteristic that the material itself cannot replicate. Its traceability system not only records the thermodynamic processes the material underwent, but also records the dynamic parameter adjustments implemented by the system to ensure the uniqueness of the material's characteristics. This effectively prevents the possibility of counterfeiting packaging materials by tampering with a single environmental parameter or imitating surface texture, resulting in a final anti-counterfeiting traceability chain that is both physically and digitally locked.

[0111] In some embodiments, in order to enable those skilled in the art to fully understand and implement the technical solutions described in this specification, the specific implementation principle of the above technical solutions will be explained in detail below with reference to a specific supercritical foaming production scenario for thermoplastic polyurethane (TPU) for packaging. Assume that a production line is producing anti-counterfeiting packaging material with batch number TPU-20231105.

[0112] In stage S1, the sensor array acquires pressure and temperature data from the extruder die section at a frequency of 100Hz. to Within the 5-second abnormal interval, 500 raw pressure data points were collected. The system uses Gaussian filtering (window radius). Standard deviation ), for example in At that point, the original pressure jumps to Substitute into the filtering formula:

[0113] Combining the data before and after, the smoothed Revised to This effectively filters out electromagnetic spike interference. Simultaneously, it utilizes Fast Fourier Transform:

[0114] The signal was found to contain The system filters out high-frequency mechanical vibration noise and performs an inverse transform back to the time domain. Then, the wavelet transform formula is used:

[0115] In scale factor Capture local high energy coefficients (such as Based on this, the SVM model determined that the fluctuation was a "normal supercritical phase transition fluctuation".

[0116] In stages S2 and S3, an optical scanner acquires images of the material surface produced in this interval, which are then fused with the temporal data by a CNN model. In a certain convolutional layer, assuming the input tensor... The local area is Matrix (e.g.) ), corresponding convolution kernel Also (like ), bias Substitute into the formula:

[0117] The calculation yields:

[0118] After adding the bias, we get Output after ReLU activation function The final fully connected layer outputs a 64-dimensional embedding vector. (For example: If compared with another batch of vectors... Substituting into the cosine similarity formula:

[0119] The molecular dot product was calculated as follows: The product of the denominator and the modulus is The similarity was calculated as follows: This value is greater than The preset threshold needs to be deduplicated.

[0120] In the deduplication verification, assume that the Local Sensitive Hash (LSH) encoded sequence initially generated by the system for these two similar textures is as follows: and .

[0121] Substituting into the Levenstein distance formula Perform recursive calculation: the cost is 0 if the last two digits are the same, and the cost is 0 if the second to last digit 'A' and 'B' are different. After dynamic programming matrix calculation, the following results were obtained. Because the distance is too small, the system judges it as a high-risk duplicate.

[0122] To solve this problem, the system extracts the generated... timestamp (e.g.) Perform an XOR operation to generate a final 256-bit unique identifier sequence such as "A9C4...8B", ensuring the absolute uniqueness of the anti-counterfeiting code.

[0123] In stages S4 and S5, this batch of data is packaged into blocks. The block header hash uses:

[0124] The calculation assumes the previous hash is "0000abc...", the Merkle root hash is "def123...", the timestamp is "1700000005", and the random number is "48291". These are concatenated to generate the current block hash "00007f9c...". When a quality inspector randomly checks this batch, they scan to obtain the actual feature vector. Its SimHash quantization value is "11010010". The system reads the prediction vector from the chain. The SimHash value is "11000011". Substitute this into the verification formula. Calculate the Hamming distance: $\text{Hamming}("11010010","11000011")=2$. Because (Preset tolerance) ),so Output True to verify that the material's physical characteristics are completely consistent with the on-chain environment record.

[0125] In phase S6, the system evaluates the reliability of the anti-counterfeiting and traceability system for this batch. The total number of data points to be collected is known. Actual lost data points (packet loss rate) The composite variance within the outlier intervals was calculated as follows: Substituting into the reliability formula:

[0126] Assuming attenuation coefficient ,have to Although the data is accurate, the indicators are lower than expected due to significant ambient noise in the sensor environment.

[0127] Based on the relatively low reliability, predictions and adjustments are made in stages S7 and S8. The LSTM network reads historical hidden states. (like ) and current temperature fluctuation input Substituting into the forgetting gate formula:

[0128]

[0129] Assume weights Product plus bias The result is After mapping by the Sigmoid function, we get Based on this, the network decides to retain some historical deviation memory. The final model predicts that the next batch will exhibit porosity deviation, and the system calculates the current microtexture deviation error. Error change rate The integral term is Set PID parameters Substitute into the adjustment formula:

[0130] The system then issues a reduction order to the extruder. The back pressure command successfully enabled closed-loop feedback of the production line based on historical anti-counterfeiting data, and recorded this adjustment on the blockchain.

[0131] The preset lower limit threshold value is determined based on a finite number of historical data sets, followed by a series of data processing steps. Specifically, the system collects 5,000 hash code sequences generated by different physical entities from the historical production database, calculates the Lewinstein distance between each pair, and statistically determines the lower boundary of the minimum edit distance distribution between unrelated code sequences. Then, a security bias value at the lower boundary of this distribution is selected as the critical reference standard for judging whether a hash collision has occurred. For example, when cross-comparing 5,000 historical normal anti-counterfeiting codes, it was found that the Lewinstein distances of codes generated by different batches of physical entities were mostly distributed between 15 and 30, with the smallest distance in history being 8. Considering the fault tolerance in extreme and accidental cases, the system truncates the distance distribution map to the left and ultimately sets the preset lower limit threshold value for judging whether there is a high risk of code duplication to 3.

[0132] The ultimate safety limit of this equipment is determined based on a limited number of historical data points, followed by a series of data processing steps. Specifically, the system consults and extracts the physical load limit parameters from the mechanical design specifications of the extrusion molding equipment, and combines this with records of critical operating conditions that triggered protective shutdowns in the equipment's high-load operation logs over the past two years. By calculating the redundancy between the normal operating range and the triggered safety alarm state, the maximum instantaneous parameter change that the equipment can withstand during a single feedback adjustment is derived. For example, according to the historical data of a certain model of equipment, its rated operating back pressure is 15.0 MPa. Historically, when a single control command required a pressure drop exceeding 1.5 MPa, it was highly likely to cause melt rupture or motor overload shutdown. After deducting the safety margin, the system analyzes the process steady-state window and limits the maximum safe change allowed for a single adjustment to within 1.0 MPa, thus setting the specific value of the ultimate safety limit of the equipment used for judgment to 1.0 MPa.

[0133] If the technical solution of this application involves the collection, processing, or application of personal information, the relevant products have strictly complied with the requirements of the "Personal Information Protection Law of the People's Republic of China" and other laws and regulations before implementing any personal information processing activities, clearly and explicitly informing individuals of the rules for personal information processing and obtaining their independent and voluntary authorization and consent. Specifically, if the information involved is sensitive personal information, the product has not only obtained the individual's separate consent before processing, but this consent is also an explicit consent made on the basis of full knowledge. For example, in areas where personal information collection devices such as cameras are deployed, prominent and eye-catching signs have been set up to clearly inform users that entering the area is considered as consenting to the collection of their personal information; or, on the personal information processing interface (such as applications, web pages, etc.), through pop-ups, checkboxes, or active uploads, the user is required to actively authorize the process after clearly displaying key rules such as the identity of the personal information processor, the purpose of processing, the processing method, and the types of information involved.

[0134] The above description of the embodiments is only for the purpose of helping to understand the technical solutions and core ideas of this application; those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. An AI-based anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging, characterized in that, The method includes: The pressure change and temperature fluctuation data during the production process of supercritical foamed materials are collected in real time by a sensor array, and the micro-texture image data of the material surface are collected simultaneously to obtain a joint production environment dataset containing multi-dimensional time series and spatial images. Based on the collected multidimensional time series, a convolutional neural network is used to extract the feature correlation between pressure changes and temperature fluctuations and material structure, and to determine the embedding vector for dynamic capture. From the given embedding vectors, obtain the micro-texture pattern of surface characteristics. If the similarity of the embedding vectors exceeds a preset threshold, then map the pattern to a unique identifier encoding sequence. For the uniquely identified coded sequence, blockchain technology is used to record the environmental factors and corresponding relationships in the production process in an immutable log, thereby obtaining a traceable distributed ledger; Obtain log data from the distributed ledger, verify whether pressure changes and temperature fluctuations are consistent with the surface characteristics of the material structure through a hash function, and obtain a Boolean value for the verification result; If the Boolean value of the verification result is true, then extract the corresponding data related to the technical challenges from the distributed ledger, determine whether the dynamic capture of environmental factors is complete, and determine the reliability indicators of anti-counterfeiting and traceability. Based on the determined reliability indicators, a recurrent neural network is used to predict the impact of potential temperature fluctuations on unique identifiers during the production process, and an optimized correspondence model is obtained. From the optimized correspondence model, the surface property adjustment parameters of the material structure are obtained. If the adjustment parameters meet the preset conditions, the traceability log of the distributed ledger is updated to obtain the final anti-counterfeiting traceability chain.

2. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The process involves real-time acquisition of pressure changes and temperature fluctuations during the production of supercritical foamed materials using a sensor array, along with simultaneous acquisition of microscopic texture images of the material surface. This results in a joint production environment dataset containing multidimensional time-series sequences and spatial images, including: The pressure changes and temperature fluctuations during the production process of supercritical foamed materials are collected in real time by a sensor array to construct a multi-dimensional time series. A pre-established data processing module is used to denoise the acquired multidimensional time series to obtain a smoothed time series signal. If a pressure change exceeding a preset threshold is detected in the smoothed time-series signal, the abnormal time point is marked by the anomaly detection module to determine the abnormal range. Based on temperature fluctuation data within the abnormal range, pattern matching is performed using preset logical rules to determine whether there are potential production environment risks. By using time window analysis, local features are extracted from multidimensional time series signals within abnormal intervals to obtain key fluctuation patterns. For key fluctuation patterns, the support vector machine algorithm is applied for classification to obtain the classification results of production process stability; Based on the classification results, corresponding production environment adjustment parameters are generated and output to the control system to complete the automated response.

3. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The step involves using a convolutional neural network to extract the feature correlations between pressure changes, temperature fluctuations, and material structure based on the acquired multidimensional time-series sequences, and determining the dynamically captured embedding vector, including: By using multi-dimensional acquisition methods, time-series data containing pressure changes and temperature fluctuations are obtained from sensors, stored as an initial dataset, and a structured time-series record is obtained. For structured time-series records, a convolutional neural network is used to perform hierarchical processing of pressure changes and temperature fluctuations, extract correlation features related to material structure, and determine a preliminary feature matrix. Based on the preliminary feature matrix, the spatiotemporal distribution patterns of key changes are analyzed, the mapping relationship between features is constructed, and an intermediate vector representing the trend of change is obtained. After obtaining the intermediate vector, the vector is compressed using a dimensionality reduction method to retain the main change information, and a dynamically captured embedding vector is constructed to determine whether it meets the preset expression requirements. If the dynamically captured embedding vector meets the preset expression requirements, it will be used as the core representation for subsequent analysis. If it does not meet the requirements, a second feature extraction is performed on the intermediate vector to obtain the adjusted embedding representation; By adjusting the embedded representation and combining it with the physical properties of the material structure, the influence patterns of pressure changes and temperature fluctuations on the structure are analyzed, and the final correlation description results are determined.

4. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The step of obtaining the micro-texture pattern of surface characteristics from the determined embedding vectors, and determining whether the similarity of the embedding vectors exceeds a preset threshold, and then mapping the pattern to a uniquely identified encoded sequence, includes: Microscopic texture information related to surface properties is obtained from the stored embedded vector data, and a preliminary set of texture patterns is determined by parsing the vector structure. For the initial texture pattern set, the cosine similarity algorithm is used to calculate the similarity value between each embedding vector, and the similarity result between each pair of vectors is obtained; Based on the similarity results, if the similarity value of a pair of embedded vectors exceeds a preset threshold, the texture pattern corresponding to the pair of vectors is marked as a highly relevant pattern, and a list of highly relevant patterns is determined. For the list of highly relevant patterns, each highly relevant pattern is mapped to a unique encoding sequence through pattern transformation rules, resulting in a set of corresponding encoding sequences; From the set of encoded sequences, obtain the unique identifier information of each encoded sequence. By comparing the sequences, determine whether there are duplicate codes. If there are duplicates, perform sequence deduplication to obtain the final set of encoded sequences. Based on the final set of encoded sequences, a unique identifier record corresponding to the surface characteristics is generated, and the record is saved through a data storage tool to complete the encoding and mapping process of the texture pattern.

5. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The encoded sequence for the unique identifier uses blockchain technology to record immutable logs of environmental factors and their corresponding relationships in the production process, resulting in a traceable distributed ledger, including: For unique identifiers and coding sequences, a pre-established mapping mechanism is used to bind each link in the production process with a specific code, obtain the corresponding identifier data, and determine the basis for unique tracking of each production unit; Based on the acquired identification data, blockchain technology is used to collect and record environmental factors in the production process in real time, forming tamper-proof log entries and obtaining distributed storage units. For distributed storage units, the consensus mechanism of blockchain is used to synchronize and update the distributed ledger for each log entry, obtain consistent data records, and determine the integrity of data across multiple nodes; If data records are inconsistent across multiple nodes, the log entries are compared using timestamp verification and hash verification to obtain the corrected unified log data. Based on the restored unified log data, the correspondence between environmental factors and production processes is structured and stored to obtain a traceable query index and determine the environmental impact records of each production link. For the query index, smart contract technology is used to automatically verify traceable data, obtain access permissions and query paths, and determine whether the historical environmental factors of a specific coded sequence meet preset standards. Based on the verified query path, data is extracted from the log system in the distributed ledger to obtain the complete production process traceability chain and determine the final tamper-proof record content.

6. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The process involves obtaining log data from the distributed ledger, verifying whether pressure changes and temperature fluctuations are consistent with the surface properties of the material structure using a hash function, and obtaining a Boolean value for the verification result, including: Log data is extracted from the distributed ledger, and the recorded environmental parameters are classified and processed to separate the datasets of pressure changes and temperature fluctuations, resulting in a preliminary set of environmental parameters. For the initially compiled set of environmental parameters, a hash function is used to encrypt the data on pressure changes and temperature fluctuations, generating a corresponding hash value dataset to ensure data integrity. By using hash value datasets, pre-stored characteristic data of material surfaces are obtained and compared with hash values ​​of pressure changes and temperature fluctuations to determine data matching. If the comparison results show that the hash values ​​of pressure changes and temperature fluctuations are consistent with the material surface property data, then the output boolean value is true, confirming that the matching verification has passed; If the comparison results show that the hash value of pressure change or temperature fluctuation is inconsistent with the material surface characteristic data, the output boolean value is false, indicating that the matching verification has failed. Based on the Boolean output, corresponding verification records are generated and stored in the distributed ledger, forming a closed-loop data processing flow.

7. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, If the verification result is true (Boolean value), then the corresponding relationship data related to the technical challenges is extracted from the distributed ledger to determine whether the dynamic capture of environmental factors is complete, and to determine the reliability indicators of anti-counterfeiting and traceability, including: If the verification result is true, the corresponding data related to the technical challenges is extracted from the distributed ledger, and the extracted data is structured using data parsing tools to obtain a pre-organized data set. For the initially processed dataset, a dynamic environmental factor capture method is adopted to record and compare environmental information in the data in real time to determine the completeness of environmental factor capture. Based on the completeness of the environmental factors captured, it is determined whether there is any missing or abnormal data. If there is any missing or abnormal data, backup data is obtained from the distributed ledger through a preset supplementation mechanism to obtain a complete environmental information dataset. By using a complete environmental information dataset and combining it with the logical association rules for anti-counterfeiting and traceability, the data is verified in multiple dimensions to determine its authenticity and consistency, and to obtain the verified data results. Based on the verified data results, the reliability level of anti-counterfeiting and traceability is determined through logical calculation and comparison using the evaluation criteria of reliability indicators. Based on the reliability level, corresponding technical challenge solution data is generated, and the data is updated to the distributed ledger using a preset storage mechanism to complete the data closed-loop processing.

8. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The step of using a recurrent neural network to predict the impact of potential temperature fluctuations on unique identifiers during the production process, based on determined reliability indicators, to obtain an optimized correspondence model includes: By collecting data records of temperature fluctuation values ​​and temperature changes from multiple data collection points during the production process, these data are stored as an initial dataset to obtain a complete temperature change sequence. For temperature change sequences, a recurrent neural network is used for time series analysis to process the correlation between temperature fluctuation values ​​and potential forces, and to determine the quantitative results of the fluctuation impact. Based on the quantification results of the impact of fluctuations, combined with the data characteristics of unique identifiers, a preliminary model for optimizing correspondence is constructed to obtain the mapping parameters between temperature change and unique identifiers. If the mapping parameters do not match the preset threshold of the reliability index, the model is iteratively adjusted through a recurrent neural network to handle the data deviation of temperature fluctuation values ​​and judge the accuracy of the adjusted model. Based on the adjusted model accuracy, the potential impact of temperature fluctuations on the unique identifier is analyzed, evaluation data on the impact of fluctuations is generated, and the final optimized corresponding model is obtained. By optimizing the corresponding model, the real-time temperature changes during the production process are processed, the potential impact of fluctuations is predicted, and the reliability indicators of the real-time data are determined.

9. The AI ​​anti-counterfeiting and traceability management method for supercritical foamed materials used in packaging according to claim 1, characterized in that, The process involves obtaining surface property adjustment parameters for the material structure from the optimized correspondence model, determining if the adjustment parameters meet preset conditions, updating the traceability log of the distributed ledger, and obtaining the final anti-counterfeiting and traceability chain, including: By optimizing the model, surface property data of the material structure are extracted from the correspondence to generate an initial set of adjustment parameters; Based on the initial set of adjustment parameters, the surface characteristic data are compared and analyzed to determine whether the adjustment parameters meet the preset conditions. If they do, a parameter record that meets the standard is generated. Obtain parameter records that meet the standards, and combine them with the traceability mechanism of the distributed ledger to match the parameter records with historical data to determine the updated content of the traceability log. For updates to the traceable logs, distributed ledger technology is used to write the data, resulting in an updated log dataset. Extract key information for anti-counterfeiting and traceability from the updated log dataset, construct the structural framework of the anti-counterfeiting and traceability chain, determine the integrity of the chain, and generate preliminary chain data if the chain is complete. By associating preliminary chain data with the corresponding relationship between characteristic parameters and the final chain, complete anti-counterfeiting and traceability chain data is generated. Obtain complete anti-counterfeiting and traceability chain data, store it to a designated node in the distributed ledger, and confirm that the final chain data has been synchronized.

Images