Weight measurement data acquisition and traceability management method and system based on Internet of Things

Abstract

The invention provides a weight metering data acquisition and traceability management method and system based on the Internet of Things, and relates to the technical field of the Internet of Things, and the method comprises the steps: implanting an RFID electronic tag in a weight, collecting environment parameters and weight data, building a mapping relation between the environment parameters and the weight of the weight through combining with a deep learning algorithm, and carrying out the traceability management of the weight. And multi-dimensional environment compensation and thermal expansion coefficient correction are carried out, and a self-adaptive traceability chain structure is constructed by using a block chain technology, so that safe storage and whole-process traceability of measurement data are realized. According to the invention, the accuracy of weight metering is improved, credible traceability of metering data is realized, and the safety of the metering process is guaranteed.

Description

Technical Field

[0001] The present invention relates to Internet of Things technology, and in particular to a method and system for collecting and tracing management of weight measurement data based on the Internet of Things. Background Art

[0002] At present, electronic tags, sensors and other technical means have been adopted for weight measurement management to collect data, and information systems are used for data storage and management. However, the existing methods for collecting and managing weight measurement data still have the following deficiencies:

[0003] The existing measurement data collection system lacks the ability to comprehensively monitor and analyze environmental parameters, and cannot accurately evaluate the comprehensive influence of environmental factors such as temperature, humidity and air pressure on the weight measurement accuracy, resulting in the reliability of measurement results being affected.

[0004] Traditional data management methods are difficult to ensure the authenticity and integrity of measurement data, and it is easy to have situations such as data tampering or loss. A credible measurement traceability system cannot be established, which is not conducive to the quality traceability of measurement results and liability determination.

[0005] The evaluation of the service life of existing weights mainly relies on empirical judgment, lacking a scientific prediction method based on historical data, and unable to detect the problem of weight performance degradation in time, increasing the risk of failure of measuring instruments and affecting the reliability of measurement work. Summary of the Invention

[0006] Embodiments of the present invention provide a method and system for collecting and tracing management of weight measurement data based on the Internet of Things, which can solve the problems in the prior art.

[0007] In the first aspect of the embodiments of the present invention,

[0008] A method for collecting and tracing management of weight measurement data based on the Internet of Things is provided, including:

[0009] An RFID electronic tag storing weight identity information is implanted into the weight body, temperature parameters, humidity parameters and air pressure parameters are collected through an environmental parameter sensor, and weight data of the weight is collected through a weight data collection module provided with a high-precision strain gauge; the weight identity information, environmental parameters, weight data of the weight and metrological standard instrument information are timestamped and then composed into measurement raw data;

[0010] The metering raw data is sent to an edge computing server through an Internet of Things communication module with data encryption function for preprocessing of outlier detection and data smoothing, and the preprocessed data is transmitted to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and weight of weights based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data of weights according to the coupling relationship of temperature parameters, humidity parameters and air pressure parameters, and corrects the coefficient of thermal expansion in combination with the material characteristics of the weights to generate metering calibration data. At the same time, the wear rate of the weights is predicted based on historical metering data;

[0011] A secure isolation architecture is constructed, and the metering calibration data and the predicted wear rate are encrypted based on the secure isolation architecture. An adaptive traceability chain structure including a base chain and an extended chain is constructed, and the metering calibration data and the wear rate prediction result are written into a blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the whole metering process; An access control policy based on behavioral characteristics is established, and hierarchical response measures are triggered according to the anomaly score, and the traceability analysis result is provided to the user terminal.

[0012] In an optional implementation manner,

[0013] The steps for the cloud data processing server to establish a mapping relationship between environmental parameters and weight of weights based on a deep learning algorithm, perform multi-dimensional environmental compensation on the weight data of weights according to the coupling relationship of temperature parameters, humidity parameters and air pressure parameters, and correct the coefficient of thermal expansion in combination with the material characteristics of the weights to generate metering calibration data include:

[0014] A multi-branch attention network structure is constructed by using a deep learning algorithm. The multi-branch attention network structure includes a two-stream architecture of a time stream and a space stream, and adaptive fusion of multi-scale features is performed to obtain a mapping result;

[0015] Based on the mapping results, Kalman filtering and wavelet transform are used to perform spatio-temporal dimension compensation on environmental parameters, construct a non-linear coupling relationship of temperature parameters, humidity parameters and air pressure parameters, and combine material thermodynamics calculation and molecular dynamics simulation to perform physical dimension compensation; the multi-dimensional environmental compensation also includes intelligent dimension compensation: construct a heterogeneous graph neural network to model the interaction relationship of environmental parameters, use temperature parameters, humidity parameters and air pressure parameters as graph network nodes, capture the dynamic association between parameters through the attention edge convolution layer, and introduce skip connections to retain historical interaction information; design a deep reinforcement learning model with a dual reward mechanism to dynamically adjust the compensation strategy, and the dual reward mechanism includes environmental stability reward and measurement accuracy reward, and use the policy gradient algorithm based on the advantage function to optimize the compensation action; construct a hybrid architecture of transfer learning and multi-task learning, extract the general features of environmental parameters through the shared representation layer, use adaptive feature alignment to eliminate the distribution differences under different working conditions, and combine the multi-head attention mechanism to perform fine-grained expression of features to complete the optimization of the compensation strategy;

[0016] Perform multi-scale thermal expansion correction on the compensated data, establish a stress-strain tensor at the lattice scale to describe the thermal deformation characteristics and calculate the contribution of atomic vibration to volume change; at the microscale, determine the thermal expansion correlation characteristics through the kinetic simulation of grain boundary sliding effect and dislocation motion analysis; at the macroscale, establish a multi-scale thermodynamics equation set to couple and calculate elastic deformation and thermal expansion;

[0017] Input the corrected data into the calibration data generation module, perform outlier detection and filtering and denoising on the data, and perform adaptive weighted fusion on the deep learning prediction results, multi-dimensional environmental compensation results and thermal expansion correction results to generate metrological calibration data.

[0018] In an alternative embodiment,

[0019] Use a deep learning algorithm to construct a multi-branch attention network structure. The multi-branch attention network structure includes a time stream and a space stream dual-stream architecture. The steps of adaptively fusing multi-scale features to obtain mapping results include;

[0020] Use DenseNet as the backbone network to construct a multi-branch attention network structure. The multi-branch attention network structure includes a time stream branch and a space stream branch; the multi-branch attention network structure is provided with a branch interaction module. The branch interaction module performs information exchange between branches through a cross-attention mechanism, constructs a feature map similarity matrix to calculate the attention weight, and controls the information flow through a gating mechanism;

[0021] The time - flow branch extracts temporal features through causal convolution, sets a time - step attention mechanism to calculate weights for the temporal features, and models the temporal features through an LSTM layer; the time - flow branch sets a temporal - feature processing module, and the temporal - feature processing module constructs a temporal multi - scale feature pyramid and processes the features through a temporal attention pooling layer;

[0022] The space - flow branch extracts spatial features through a spatial attention convolutional layer and sets deformable convolution to deform the convolution kernel; the space - flow branch sets a spatial - relationship modeling module, and the spatial - relationship modeling module calculates the correlation relationship of the spatial features through a graph attention network and introduces position encoding to mark the spatial positions of the features;

[0023] Adaptive fusion of the features of the time - flow branch and the space - flow branch includes: unifying the feature dimensions through a feature transformation network, using a learnable up - sampling module to align the feature resolutions; constructing a dynamic weight prediction network to calculate the feature fusion weights, and obtaining the final mapping result through an attention - guided feature aggregation module;

[0024] The network training is optimized using a multi - task loss function with adaptive weights. The multi - task loss function includes the Focal L1 loss of the main task, as well as the feature consistency loss and temporal smoothness loss of the auxiliary tasks, and calculates the loss weights through an adaptive weight learning method based on task uncertainty; a staged training strategy is adopted, training the time - flow branch and the space - flow branch, the branch interaction module, and the feature fusion module in sequence, and finally performing end - to - end optimization.

[0025] In an alternative implementation,

[0026] The steps for predicting the wear rate of a weight based on historical metrological data include:

[0027] Pre - process the historical metrological data using a sliding median filter, extract features such as the mass change rate, environmental parameters, usage intensity, and surface roughness, and perform standardization processing to obtain a standardized feature sequence;

[0028] Input the standardized feature sequence into a bidirectional LSTM network, extract bidirectional features through forward propagation and backward propagation. The bidirectional LSTM network sets a time - step calculation module to quantify the degree of temporal correlation, and performs weighted combination of the bidirectional features and the degree of temporal correlation to obtain a deep feature representation of the weight wear;

[0029] Construct a multi-task learning framework based on the depth feature representation. The multi-task learning framework includes a wear pattern classification branch and a rate prediction branch. The wear pattern classification branch classifies wear into normal wear, accelerated wear, and abnormal wear through a fully connected network and is optimized using focal loss. The rate prediction branch predicts the wear rate in a future time period through a residual fully connected structure. A prediction reliability evaluation module is constructed in the multi-task learning framework, and the mean and variance of the prediction results are output through a neural network and optimized using a negative log-likelihood loss function. An adaptive correction mechanism is constructed based on the prediction reliability evaluation results. When the prediction uncertainty exceeds a set threshold, the data sampling density is increased and the prediction time span is shortened. The prediction deviation is calculated based on the latest measurement data, and the correction coefficient is updated by exponential moving average to perform online correction on the prediction results.

[0030] In an alternative implementation,

[0031] The steps of constructing a security isolation architecture, encrypting the measurement calibration data and the predicted wear rate based on the security isolation architecture, constructing an adaptive traceability chain structure including a base chain and an extended chain, and writing the measurement calibration data and the wear rate prediction results into the blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the entire measurement process include:

[0032] Construct a security isolation architecture including a hardware trusted root layer, a trusted execution environment layer, and a security processing layer. The hardware trusted root layer stores keys and measures integrity through a trusted platform module. The trusted execution environment layer divides the secure world and the ordinary world and performs key management. The security processing layer is provided with an independent data processing environment.

[0033] Input the measurement calibration data and the predicted wear rate into the security isolation architecture for encryption processing. Based on the encrypted data, construct an adaptive traceability chain structure. The adaptive traceability chain structure includes a base chain for storing measurement parameters and measurement calibration data and an extended chain for storing wear rate prediction results and measurement scenario data. The cross-chain data between the base chain and the extended chain is maintained for consistency through an inter-chain data reference mechanism.

[0034] Design a hierarchical consensus mechanism for the adaptive traceability chain structure. Through the introduction of a practical Byzantine fault tolerance algorithm with dynamic view switching, consensus processing is performed on the measurement traceability reference nodes. Data confirmation is performed on the measurement data collection nodes through proof-of-stake consensus, and a consensus switching mechanism is used to be compatible with different consensus algorithms.

[0035] Perform standardized conversion and quality assessment on the measurement calibration data, obtain the characteristic parameters of the measurement process using the spatio-temporal feature extraction algorithm, establish a measurement process correlation matrix by calculating the temporal correlation degree and spatial similarity between the characteristic parameters of the measurement process and the predicted wear rate, and generate traceability information based on the intelligent contract rule engine to perform data fusion operations;

[0036] Write the traceability information into the blockchain network, and establish a secure storage of the whole-process measurement information and a trustworthy traceability information chain based on the adaptive traceability chain structure and the hierarchical consensus mechanism.

[0037] In an optional implementation manner,

[0038] The steps of performing consensus processing on the measurement traceability reference nodes through the practical Byzantine fault tolerance algorithm with dynamic view switching, performing data confirmation on the measurement data collection nodes through the proof-of-stake consensus, and adopting a consensus switching mechanism to be compatible with different consensus algorithms include:

[0039] Design a practical Byzantine fault tolerance algorithm with dynamic view switching for the measurement traceability reference nodes. The practical Byzantine fault tolerance algorithm calculates the view liveness metric through the number of messages, consensus time, and view stability, triggers view switching and updates the primary node based on the view liveness metric, and makes the measurement traceability reference node enter a new view when receiving more than twice the fault tolerance limit plus one view switching message;

[0040] Calculate the stake value of the measurement data collection nodes based on data quality, response time, and storage capacity, and use the proof-of-stake consensus to determine the verification nodes according to the stake value and complete data confirmation;

[0041] Standardize and encode the status information of the practical Byzantine fault tolerance algorithm and the proof-of-stake consensus, organize the status data using the Merkle tree structure, and construct a unified status description format including block height, state root, and consensus proof; construct a cross-consensus message based on the unified status description format, use a multi-level message confirmation mechanism to transmit the cross-consensus message, perform retransmission control and timeout detection on the cross-consensus message, authenticate the source of the cross-consensus message, and verify the legality of the state mapping and the state transition proof;

[0042] Construct a state version chain and set checkpoints, monitor the consensus efficiency metric, node response status, and state consistency deviation. When the consensus efficiency metric is lower than the preset threshold, or the node response times out, or the state consistency deviation exceeds the limit, trigger consensus switching, broadcast the switching proposal to the nodes in the network and collect node confirmation information, pause the current consensus mechanism and activate the target consensus mechanism;

[0043] Use the consensus result of the metrological traceability reference node as the trust basis for the proof-of-stake consensus of the metrological data collection node through the unified status description format. Feed back the data confirmed by the metrological data collection node to the metrological traceability reference node through the cross-consensus message. Maintain the state consistency of heterogeneous consensus based on the state version chain, and adopt a consensus switching mechanism to adjust between the practical Byzantine fault tolerance algorithm and the proof-of-stake consensus.

[0044] In an alternative embodiment,

[0045] The steps of establishing an access control policy based on behavioral characteristics, triggering hierarchical response measures according to the anomaly score, and providing the traceability analysis result to the user terminal include:

[0046] Obtain the identity characteristics and behavioral characteristics of the user's access behavior. The identity characteristics include the user's static identifier, dynamic authentication information, and environmental attributes. The behavioral characteristics include temporal behavioral characteristics, operation behavioral characteristics, and data interaction characteristics. Use an autoencoder to reduce the dimension of the identity characteristics and behavioral characteristics to obtain a feature vector, and calculate the matching degree between the feature vector and the historical baseline based on the weighted cosine similarity to obtain a real-time trust score;

[0047] Divide the access control levels based on the real-time trust score. When the real-time trust score is in the first trust interval, execute the basic access control policy to allow access with the original permissions. When the real-time trust score is in the second trust interval, execute the enhanced access control policy for secondary authentication and limit the access frequency and data range. When the real-time trust score is in the third trust interval, execute the strict access control policy to only allow read-only operations and require multi-factor authentication;

[0048] Dynamically adjust the control intensity of the access control policy according to the change trend of the real-time trust score within a continuous observation period, and generate an access control result including the user identity information, access behavior characteristics, and control policy;

[0049] Calculate the anomaly score based on the access control result, construct an abnormal behavior hierarchical response mechanism, and trigger a warning-level response, a blocking-level response, or a locking-level response according to the anomaly score when an abnormal behavior occurs. The warning-level response includes recording the abnormal event and starting feature resampling. The blocking-level response includes suspending the session permission and requiring identity re-authentication. The locking-level response includes freezing the access permission and triggering a security audit;

[0050] Construct an association graph of the user's operating resources, record the occurrence time, abnormal type, response measures, and disposal results of the abnormal behavior, establish an abnormal behavior feature library, and push the traceability analysis results including metrological data query, calibration period reminder, wear warning, and maintenance suggestions to the user terminal in real time according to the level of the anomaly score.

[0051] In the second aspect of the embodiments of the present invention,

[0052] a weight measurement data acquisition and traceability management system based on the Internet of Things is provided, including:

[0053] A first unit for implanting an RFID electronic tag storing weight identity information into the weight body, collecting temperature parameters, humidity parameters and air pressure parameters through an environmental parameter sensor, and collecting weight data of the weight through a weight data acquisition module provided with a high-precision strain gauge; timestamping the weight identity information, environmental parameters, weight data of the weight and metrological standard instrument information to form metrological raw data;

[0054] A second unit for sending the metrological raw data to an edge computing server through an Internet of Things communication module with data encryption function for preprocessing of outlier detection and data smoothing, and transmitting the preprocessed data to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data of the weight according to the coupling relationship of the temperature parameters, humidity parameters and air pressure parameters, corrects the coefficient of thermal expansion in combination with the material characteristics of the weight, generates metrological calibration data, and simultaneously predicts the wear rate of the weight based on historical metrological data;

[0055] A third unit for constructing a security isolation architecture, encrypting the metrological calibration data and the predicted wear rate based on the security isolation architecture, constructing an adaptive traceability chain structure including a basic chain and an extended chain, writing the metrological calibration data and the wear rate prediction result into a blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the whole metrological process; establishing an access control policy based on behavioral characteristics, triggering hierarchical response measures according to an anomaly score, and providing a traceability analysis result to a user terminal.

[0056] In the third aspect of the embodiments of the present invention,

[0057] an electronic device is provided, including:

[0058] a processor;

[0059] a memory for storing instructions executable by the processor;

[0060] wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0061] In the fourth aspect of the embodiments of the present invention,

[0062] a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0063] The present invention realizes the automatic collection and real-time monitoring of the weighing data of weights by implanting RFID electronic tags in the weights and combining multiple environmental sensors, improves the accuracy and efficiency of data collection, and ensures the authenticity and traceability of data through timestamp marking.

[0064] The present invention uses a deep learning algorithm to establish a mapping relationship between environmental parameters and the weight of weights, improves the accuracy of measurement data through multi-dimensional environmental compensation and thermal expansion coefficient correction, and can predict the wear rate of weights, providing a scientific basis for the maintenance and replacement of weights and extending the service life of weights.

[0065] The present invention constructs a secure isolation architecture and an adaptive traceability chain structure based on blockchain technology, ensures the security and immutability of measurement data through a hierarchical consensus mechanism, and establishes an access control strategy based on behavioral characteristics to achieve trusted traceability of the entire measurement process, improving the standardization and reliability of measurement management. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic flowchart of the method for collecting and tracing management of weighing data of weights based on the Internet of Things according to an embodiment of the present invention;

[0067] Figure 2 A multi-level access control flowchart based on behavioral characteristics;

[0068] Figure 3 It is a schematic structural diagram of the system for collecting and tracing management of weighing data of weights based on the Internet of Things according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0070] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0071] Figure 1 It is a schematic flowchart of the method for collecting and tracing management of weighing data of weights based on the Internet of Things according to an embodiment of the present invention, as Figure 1 shown, the method includes:

[0072] An RFID electronic tag storing the identity information of the weight is implanted into the weight body. Temperature parameters, humidity parameters, and air pressure parameters are collected through environmental parameter sensors, and the weight data of the weight is collected through a weight data acquisition module with high-precision strain gauges. After timestamp marking the weight identity information, environmental parameters, weight data of the weight, and metrological standard instrument information, the original metrological data is formed.

[0073] The original metrological data is sent to an edge computing server through an Internet of Things communication module with data encryption function for preprocessing of outlier detection and data smoothing, and the preprocessed data is transmitted to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data of the weight according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, corrects the coefficient of thermal expansion in combination with the material characteristics of the weight, generates metrological calibration data, and simultaneously predicts the wear rate of the weight based on historical metrological data.

[0074] A secure isolation architecture is constructed, and the metrological calibration data and the predicted wear rate are encrypted based on the secure isolation architecture. An adaptive traceability chain structure including a base chain and an extended chain is constructed, and the metrological calibration data and the wear rate prediction results are written into a blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the entire metrological process. An access control policy based on behavioral characteristics is established, and hierarchical response measures are triggered according to anomaly scores, and the traceability analysis results are provided to the user terminal.

[0075] In an alternative embodiment,

[0076] The steps of the cloud data processing server establishing a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performing multi-dimensional environmental compensation on the weight data of the weight according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, and correcting the coefficient of thermal expansion in combination with the material characteristics of the weight to generate metrological calibration data include:

[0077] A multi-branch attention network structure is constructed using a deep learning algorithm. The multi-branch attention network structure includes a time stream and a space stream dual-stream architecture, and adaptive fusion of multi-scale features is performed to obtain a mapping result.

[0078] Based on the mapping results, Kalman filtering and wavelet transform are used to perform spatio-temporal dimensional compensation on environmental parameters, construct non-linear coupling relationships of temperature parameters, humidity parameters and air pressure parameters, and combine material thermodynamics calculations and molecular dynamics simulations for physical dimensional compensation; the multi-dimensional environmental compensation also includes intelligent dimensional compensation: constructing a heterogeneous graph neural network to model the interaction relationships of environmental parameters, taking temperature parameters, humidity parameters and air pressure parameters as graph network nodes, capturing the dynamic correlations between parameters through an attention edge convolution layer, and introducing skip connections to retain historical interaction information; designing a deep reinforcement learning model with a dual reward mechanism to dynamically adjust the compensation strategy, the dual reward mechanism includes environmental stability reward and measurement accuracy reward, and using a policy gradient algorithm based on the advantage function to optimize the compensation actions; constructing a hybrid architecture of transfer learning and multi-task learning, extracting common features of environmental parameters through a shared representation layer, using adaptive feature alignment to eliminate distribution differences under different working conditions, and combining a multi-head attention mechanism for fine-grained expression of features to complete the optimization of the compensation strategy;

[0079] Perform multi-scale thermal expansion correction on the compensated data, establish stress-strain tensors at the lattice scale to describe thermal deformation characteristics and calculate the contribution of atomic vibrations to volume changes; at the microscale, determine the thermal expansion correlation characteristics through dynamic simulations of grain boundary sliding effects and dislocation motion analysis; at the macroscale, establish multi-scale thermodynamics equations to couple elastic deformation and thermal expansion for calculation;

[0080] Input the corrected data into the calibration data generation module, perform outlier detection and filtering denoising on the data, and perform adaptive weighted fusion on the deep learning prediction results, multi-dimensional environmental compensation results and thermal expansion correction results to generate metrological calibration data.

[0081] Exemplarily, the cloud data processing server first obtains real-time data of temperature parameters, humidity parameters and air pressure parameters in the weight measurement environment of the weight. A multi-branch attention network is constructed using a deep learning algorithm, and this network includes a dual-stream architecture of a time stream and a space stream. The time stream branch uses a long short-term memory network to process the temporal variation characteristics of environmental parameters, the number of network units is set to 128, and the time window length is 24 hours. The space stream branch uses a three-dimensional convolutional neural network to extract the spatial distribution characteristics of environmental parameters, the convolutional kernel size is 3×3×3, the stride is 1, and the padding method is SAME. The features of the two branches are fused through an adaptive weighting method, and the weight coefficients are dynamically updated through the backpropagation algorithm.

[0082] Perform multi-dimensional environmental compensation on the obtained mapping results. First, use a Kalman filter to denoise the environmental parameters, and set the diagonal elements of the observation noise covariance matrix to 0.01. Then, perform multi-scale decomposition using wavelet transform, select the db4 wavelet basis function, and the decomposition level is 3 layers. On this basis, construct a heterogeneous graph neural network, use temperature, humidity, and air pressure parameters as the nodes of the graph network respectively, and the node feature dimension is 64. Extract the dynamic correlation features between parameters through two layers of graph attention convolutional layers, set the number of attention heads to 8, and the output feature dimension of each layer is 32. Introduce residual connections to retain historical interaction information, and the stride of the skip connection is 2.

[0083] Design a deep reinforcement learning model based on a dual reward mechanism. The environmental stability reward is calculated according to the parameter fluctuation range, and the measurement accuracy reward is determined based on the deviation from the standard value. Adopt the Actor-Critic network architecture, where the Actor network outputs compensation actions and the Critic network evaluates the action value. The network uses a three-layer fully connected layer structure, and the number of neurons in the hidden layers is 256, 128, and 64 respectively, and the activation function is ReLU.

[0084] In the thermal expansion correction stage, first establish a stress-strain relationship model at the lattice scale. Take copper-based weight material as an example. Under the condition of room temperature 293K, calculate the influence of atomic vibration on volume change through molecular dynamics simulation, set the simulation time step to 1 femtosecond, and the total simulation duration is 100 picoseconds. On the microscopic scale, use the dislocation dynamics method to analyze the grain boundary slip effect and calculate the evolution law of dislocation density. The initial dislocation density is set to 10 8 / m 2 . Establish a multi-scale coupling model on the macroscopic scale to perform unified calculations of elastic deformation and thermal expansion.

[0085] Finally, in the calibration data generation link, use the isolation forest algorithm for outlier detection, and set the contamination rate threshold to 0.1. Use a Butterworth low-pass filter for signal denoising, and the cut-off frequency is 0.1Hz. Perform weighted fusion on the deep learning prediction results, environmental compensation results, and thermal expansion correction results. The weight coefficients are determined through cross-validation and are 0.4, 0.35, and 0.25 respectively.

[0086] The present invention realizes the accurate mapping of environmental parameters and weight of weights by constructing a multi-branch attention network. The dual-stream architecture can capture temporal change features and spatial distribution features simultaneously, improving the prediction accuracy of the model. The adaptive feature fusion mechanism can dynamically adjust feature weights according to different working conditions, enhancing the generalization ability of the model. The multi-dimensional environmental compensation strategy comprehensively considers the compensation effects in the spatio-temporal dimension and the intelligent dimension. The heterogeneous graph neural network effectively models the complex interaction relationships between environmental parameters. The reinforcement learning model with a dual reward mechanism can adaptively optimize the compensation strategy, improving the accuracy and robustness of environmental compensation. The multi-scale thermal expansion correction method realizes the full-scale analysis from the lattice scale to the macroscopic scale, accurately describing the thermal deformation characteristics of materials. Combining molecular dynamics simulation and dislocation dynamics analysis, it deeply reveals the microscopic mechanism, improving the scientificity and reliability of thermal expansion coefficient correction. Finally, through outlier detection and adaptive weighted fusion, the accuracy and stability of calibration data are ensured.

[0087] In an alternative embodiment,

[0088] A multi-branch attention network structure is constructed using a deep learning algorithm. The multi-branch attention network structure includes a dual-stream architecture of a time stream and a space stream. The steps of adaptively fusing multi-scale features to obtain a mapping result include:

[0089] DenseNet is used as the backbone network to construct the multi-branch attention network structure. The multi-branch attention network structure includes a time stream branch and a space stream branch. The multi-branch attention network structure is provided with a branch interaction module. The branch interaction module exchanges information between branches through a cross-attention mechanism, constructs a feature map similarity matrix to calculate attention weights, and controls the information flow through a gating mechanism.

[0090] The time stream branch extracts temporal features through causal convolution, sets a time step attention mechanism to calculate weights for the temporal features, and models the temporal features through an LSTM layer. The time stream branch is provided with a temporal feature processing module. The temporal feature processing module constructs a temporal multi-scale feature pyramid and processes the features through a temporal attention pooling layer.

[0091] The space stream branch extracts spatial features through a spatial attention convolutional layer and sets deformable convolution to deform the convolution kernel. The space stream branch is provided with a spatial relationship modeling module. The spatial relationship modeling module calculates the correlation relationship of spatial features through a graph attention network and introduces position encoding to mark the spatial positions of the features.

[0092] Adaptive fusion of the features of the time stream branch and the space stream branch includes: unifying the feature dimensions through a feature transformation network and aligning the feature resolutions using a learnable upsampling module; constructing a dynamic weight prediction network to calculate the feature fusion weights and obtaining the final mapping result through an attention-guided feature aggregation module;

[0093] Optimizing network training using a multi-task loss function with adaptive weights. The multi-task loss function includes the Focal L1 loss of the main task, as well as the feature consistency loss and the temporal smoothness loss of the auxiliary tasks, and calculates the loss weights through an adaptive weight learning method based on task uncertainty; adopting a phased training strategy, training the time stream branch and the space stream branch, the branch interaction module, and the feature fusion module in sequence, and finally performing end-to-end optimization.

[0094] Exemplarily, first construct the basic architecture of a multi-branch attention network structure. Select DenseNet-121 as the backbone network, which contains 4 dense blocks, and each dense block contains 6-layer, 12-layer, 24-layer, and 16-layer dense connection layers respectively. Construct the time stream branch and the space stream branch respectively on the basis of the backbone network.

[0095] In the time stream branch, temporal feature extraction is performed through causal convolution. The causal convolution uses a one-dimensional convolution kernel of 3×1, with a stride of 1 and a causal padding method. The input sequence length is set to 16 frames, and the feature dimension is 256. The time step attention mechanism performs weighting by calculating the importance scores of each time step. Specifically, the feature vectors of each time step pass through two fully connected layers to obtain the attention scores, and then pass through softmax normalization to obtain the weight coefficients. Subsequently, a double-layer LSTM network is used for temporal modeling, and the hidden layer dimension is 512.

[0096] The temporal feature processing module constructs a 3-layer temporal feature pyramid with scale ratios of 1:2:4 respectively. Each layer of features is processed through temporal attention pooling, and the pooling window sizes are 1, 2, and 4 respectively to obtain multi-scale temporal representations.

[0097] In the space stream branch, the spatial attention convolution layer uses a 3×3 convolution kernel with 256 channels. The offsets of the deformable convolution are predicted by an independent convolution branch. The spatial relationship modeling module first divides the feature map into a 7×7 grid, and each grid serves as a node in the graph. An 8-head attention mechanism is used to calculate the association strength between nodes, and the attention head dimension is 32. The position encoding adopts the sine-cosine encoding method, and the encoding dimension is the same as the feature dimension.

[0098] In the branch interaction module, the cross-attention mechanism calculates the similarity matrix between the feature maps of the two branches. Each element of the similarity matrix represents the matching degree between the corresponding positions of the source feature map and the target feature map. The gating mechanism is implemented by a two-layer perceptron, which adaptively adjusts the information flow ratio according to the input features.

[0099] In the feature fusion stage, the feature dimensions are first unified to 256 through 1×1 convolutions. The learnable upsampling module uses a transposed convolution layer with a stride of 2 and a kernel size of 4. The dynamic weight prediction network consists of 3 convolutional layers, and the number of output channels is the same as the number of features. The attention-guided feature aggregation module calculates the attention map between features and performs weighted fusion accordingly.

[0100] In terms of network training, the parameter alpha of the Focal L1 loss for the main task is set to 2, and gamma is set to 0.25. The feature consistency loss uses cosine similarity to measure the similarity of features between the two branches. The temporal smoothness loss calculates the difference between features at adjacent time steps. The adaptive weight learning determines the weight coefficient based on the gradient norm of each task.

[0101] The training strategy is divided into three stages: In the first stage, the time flow and space flow branches are pre-trained separately, with each branch trained for 50 epochs; in the second stage, the branch interaction module is added and trained for 30 epochs; in the third stage, the feature fusion module is introduced for end-to-end training for 20 epochs. The initial learning rate is set to 0.001, and the cosine annealing strategy is used for adjustment.

[0102] Through the design of the two-stream architecture and the multi-branch attention mechanism, the present invention can fully capture the temporal dynamic features and spatial structure features of the input data, improving the feature expression ability and model performance. The branch interaction module and the adaptive feature fusion strategy achieve the effective fusion of spatio-temporal features, enhancing the model's understanding ability of complex scenes; by adopting multi-scale feature extraction and adaptive fusion mechanisms, the adaptability of the model to targets of different scales is improved, and the robustness of feature expression is enhanced. The introduction of deformable convolutions and graph attention networks improves the model's modeling ability for spatial geometric transformations; the multi-task learning framework and the adaptive weight optimization strategy achieve stable convergence of model training, improve the model's generalization ability, and the staged training strategy ensures the full optimization of each module, ultimately achieving end-to-end performance improvement.

[0103] In an alternative embodiment,

[0104] The steps for predicting the wear rate of the weight based on historical metrological data include:

[0105] Preprocess the historical metrological data using sliding median filtering, extract features such as mass change rate, environmental parameters, usage intensity, and surface roughness, and perform standardization processing to obtain a standardized feature sequence;

[0106] Input the standardized feature sequence into a bidirectional LSTM network, extract bidirectional features through forward propagation and backward propagation. The bidirectional LSTM network is provided with a time step calculation module to quantify the degree of temporal correlation, and perform weighted combination of the bidirectional features and the degree of temporal correlation to obtain a depth feature representation of the weight wear.

[0107] Construct a multi-task learning framework based on the depth feature representation. The multi-task learning framework includes a wear mode classification branch and a rate prediction branch. The wear mode classification branch divides wear into normal wear, accelerated wear, and abnormal wear through a fully connected network, and is optimized using focal loss. The rate prediction branch predicts the wear rate in a future time period through a residual fully connected structure. A prediction reliability evaluation module is constructed in the multi-task learning framework, and the mean and variance of the prediction results are output through a neural network, and are optimized using a negative log-likelihood loss function. An adaptive correction mechanism is constructed based on the prediction reliability evaluation result. When the prediction uncertainty exceeds a set threshold, increase the data sampling density and shorten the prediction time span, calculate the prediction deviation based on the latest measurement data, and update the correction coefficient through exponential moving average to perform online correction on the prediction result.

[0108] Exemplarily, first preprocess the historical measurement data. Process the original quality data through a sliding median filtering method, and set the sliding window size to 7 data points, which can effectively remove outliers and noise. Extract features from the processed data, including the quality change rate (monthly percentage of quality loss), environmental parameters (temperature, humidity, air pressure, etc.), usage intensity (number of monthly usages), and surface roughness, etc. Standardize all features and normalize the values to the range of 0-1.

[0109] Then construct a bidirectional LSTM network for feature extraction. The network contains two layers of bidirectional LSTM layers, and the number of neurons in each layer is 128. The forward LSTM layer extracts temporal features from the past to the present, and the backward LSTM layer extracts features from the present to the past. Set a time step calculation module and use an attention mechanism to calculate the correlation degree of data at different time steps. Specifically, when implementing, use the data of the past 12 months as the input sequence, calculate the correlation weight of each time step with the current prediction moment. Multiply and superimpose the bidirectional features and the correlation weights to obtain a depth feature representation of the weight wear.

[0110] Then, a multi-task learning framework is constructed. The wear mode classification branch adopts a three-layer fully connected network with the number of neurons being 256, 128, and 3 respectively, and the probabilities of three wear modes are output through the softmax function. The rate prediction branch adopts a residual structure, including two parallel fully connected layers, and the final prediction value is obtained by adding the output results. The prediction reliability evaluation module also adopts a fully connected network structure to output the mean and variance of the prediction results. When the prediction variance exceeds 0.1, the adaptive correction mechanism is triggered.

[0111] In practical applications, taking a group of 1 kg standard weights as an example. The original mass data shows that the mass of the weight has decreased by 0.5 mg within one year. After sliding median filtering, a smooth mass change curve is obtained. The environmental parameter records show that the usage environment temperature is 20 - 25 °C, the relative humidity is 45 - 55%, the atmospheric pressure is 98 - 102 kPa. The usage intensity is 50 times per month on average, and the surface roughness Ra value is 0.4 μm. Inputting these data into the trained model, the wear rate in the next 3 months can be predicted to be 0.03 mg / month, and the prediction variance is 0.08, indicating that the prediction result is reliable. The wear mode classification result shows that it belongs to the normal wear category with a confidence level of 0.92.

[0112] The present invention realizes high-quality preprocessing of data through sliding median filtering and feature standardization, effectively reduces noise interference, and improves the accuracy and stability of subsequent predictions; adopts a bidirectional LSTM network combined with a time step calculation module to fully utilize the temporal information of historical data, accurately captures the dynamic change law of weight wear, and enhances the feature extraction ability of the model; realizes the collaborative optimization of wear mode classification and rate prediction based on a multi-task learning framework, and introduces a prediction reliability evaluation and an adaptive correction mechanism, significantly improving the accuracy and reliability of the prediction results, and providing strong support for the evaluation of the service life of weights and maintenance decisions.

[0113] In an optional implementation manner,

[0114] The steps of constructing a security isolation architecture, encrypting the metrological calibration data and the predicted wear rate based on the security isolation architecture, constructing an adaptive traceability chain structure including a base chain and an extended chain, and writing the metrological calibration data and the wear rate prediction results into the blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the entire metrological process include:

[0115] Construct a security isolation architecture including a hardware trusted root layer, a trusted execution environment layer, and a security processing layer. The hardware trusted root layer stores keys and measures integrity through a trusted platform module. The trusted execution environment layer divides the secure world and the normal world and conducts key management. The security processing layer is provided with an independent data processing environment;

[0116] Input the metrological calibration data and the predicted wear rate into the secure isolation architecture for encryption processing, and construct an adaptive traceability chain structure based on the encrypted data. The adaptive traceability chain structure includes a basic chain for storing metrological parameters and metrological calibration data, and an extended chain for storing the wear rate prediction results and metrological scenario data. The cross-chain data between the basic chain and the extended chain is maintained consistent through an inter-chain data reference mechanism;

[0117] Design a hierarchical consensus mechanism for the adaptive traceability chain structure, perform consensus processing on the metrological traceability reference nodes by introducing a practical Byzantine fault tolerance algorithm with dynamic view switching, confirm the data of the metrological data collection nodes through proof-of-stake consensus, and adopt a consensus switching mechanism to be compatible with different consensus algorithms;

[0118] Perform standardization conversion and quality assessment on the metrological calibration data, use a spatio-temporal feature extraction algorithm to obtain the characteristic parameters of the metrological process, establish a metrological process association matrix by calculating the temporal correlation and spatial similarity between the characteristic parameters of the metrological process and the predicted wear rate, and execute data fusion operations based on the intelligent contract rule engine to generate traceability information;

[0119] Write the traceability information into the blockchain network, and establish a secure storage and trusted traceability information chain for the whole metrological process based on the adaptive traceability chain structure and the hierarchical consensus mechanism.

[0120] Exemplarily, first construct a three-layer secure isolation architecture. The hardware trusted root layer uses a trusted platform module (TPM) chip to store keys and realizes trusted boot through the measurement and guidance process and the integrity of the running environment. The trusted execution environment layer divides the secure world and the ordinary world based on the ARM TrustZone technology. The secure world is responsible for key management and sensitive data processing, and the ordinary world runs general applications. The secure processing layer constructs an independent data processing environment and uses memory isolation technology to ensure data processing security.

[0121] Perform encryption processing on the input metrological calibration data. Taking the calibration process of measuring instruments as an example, encrypt environmental parameters such as temperature and humidity and calibration data through the AES-256 algorithm and store them in the basic chain. For the predicted wear rate data, use the homomorphic encryption method for processing and store it in the extended chain. A link is established between the basic chain and the extended chain through a hash pointer to achieve cross-chain data consistency maintenance.

[0122] In the design of the consensus mechanism, the metrological traceability reference nodes adopt the dynamic view PBFT algorithm (practical Byzantine fault tolerance algorithm). Taking temperature measurement as an example, when a node anomaly is detected, fault tolerance is achieved by dynamically adjusting the view members. The metrological data collection nodes are based on the PoS consensus mechanism, and the block generation weight is determined according to the node reputation value. The two consensus mechanisms are made compatible through hierarchical design.

[0123] In terms of measurement data processing, the calibration data is normalized and the data quality is evaluated. Taking pressure measurement as an example, the time-series characteristic parameters such as pressure and temperature in the measurement process are extracted. The time-series correlation and spatial similarity between these characteristic parameters and the predicted wear rate are calculated to generate a correlation matrix. Based on the intelligent contract rule engine, the associated data is fused to form complete traceability information.

[0124] Finally, the traceability information is written into the blockchain network. The distributed storage technology is used to fragment and store the data on different nodes. Through the consensus mechanism, the data consistency is ensured, and the trustworthy traceability of the whole measurement process is realized.

[0125] The present invention realizes the secure storage and trustworthy processing of measurement data through a three-layer security isolation architecture, effectively preventing data leakage and tampering, and ensuring the integrity and confidentiality of measurement data; based on the adaptive traceability chain structure and hierarchical consensus mechanism, it realizes the trustworthy storage and efficient consensus of measurement data, improves the fault tolerance and scalability of the system, and ensures the reliability of traceability information; uses the intelligent contract rule engine to perform fusion analysis on measurement data, establishes the correlation between the characteristics of the measurement process and wear prediction, improves the accuracy and usability of measurement traceability, and provides a reliable basis for the management of measuring instruments.

[0126] In an alternative embodiment,

[0127] The steps of performing consensus processing on the measurement traceability reference node by introducing a practical Byzantine fault tolerance algorithm with dynamic view switching, performing data confirmation on the measurement data collection node by proof-of-stake consensus, and using a consensus switching mechanism to be compatible with different consensus algorithms include:

[0128] Design a practical Byzantine fault tolerance algorithm with dynamic view switching for the measurement traceability reference node. The practical Byzantine fault tolerance algorithm calculates the view liveness metric through the number of messages, consensus time, and view stability. Based on the view liveness metric, the view is triggered to switch and the primary node is updated. When more than twice the fault tolerance limit plus one view switching message is received, the measurement traceability reference node enters a new view;

[0129] Calculate the stake value of the measurement data collection node based on data quality, response time, and storage capacity, and use proof-of-stake consensus to determine the verification node according to the stake value and complete data confirmation;

[0130] Standardize the encoding of the state information of the Practical Byzantine Fault Tolerance algorithm and the Proof of Stake consensus, organize the state data using the Merkle tree structure, and construct a unified state description format that includes block height, state root, and consensus proof; construct a cross-consensus message based on the unified state description format, use a multi-level message confirmation mechanism to transmit the cross-consensus message, perform retransmission control and timeout detection on the cross-consensus message, authenticate the source of the cross-consensus message, and verify the legitimacy of the state mapping and the state transition proof;

[0131] Build a state version chain and set checkpoints, monitor consensus efficiency indicators, node response status, and state consistency deviation, trigger consensus switching when the consensus efficiency indicator is lower than the preset threshold or the node response times out or the state consistency deviation exceeds the limit, broadcast the switching proposal to the nodes in the network and collect node confirmation information, suspend the current consensus mechanism and activate the target consensus mechanism;

[0132] The consensus result of the measurement traceability reference node is used as the trust basis for the measurement data collection node to perform proof-of-stake consensus through the unified state description format, and the data confirmed by the measurement data collection node is fed back to the measurement traceability reference node through the cross-consensus message. The state consistency of the heterogeneous consensus is maintained based on the state version chain, and a consensus switching mechanism is used to adjust between the practical Byzantine fault-tolerant algorithm and the proof-of-stake consensus.

[0133] Exemplarily, the practical Byzantine fault tolerance algorithm for dynamic view switching first realizes the consensus of the metering traceability benchmark node by calculating the view activity metric. The view activity metric includes three dimensions: message quantity statistics, consensus time calculation, and view stability evaluation. The message quantity statistics are measured by recording the total number of preparation messages and confirmation messages transmitted between nodes in each view. When the number of messages exceeds the preset threshold, it indicates that the current view communication overhead is too large. The consensus time calculation is to count the time interval from the generation of the proposal to the consensus. If it exceeds 3 seconds, the consensus efficiency is considered to be low. The view stability evaluation is based on the view switching frequency. When the number of view switches is greater than 5 times within 10 minutes, the network state is considered to be unstable. Based on the comprehensive evaluation value of these indicators, the view switching process is triggered when it is lower than 0.6. When the view switches, the node broadcasts a switching request with a digital signature, and enters the new view after receiving a view switching request that exceeds twice the fault tolerance limit plus one.

[0134] The rights and interests proof consensus mechanism manages the measurement data collection nodes. First, calculate the node rights and interests value, including three dimensions: data quality score, average response time, and available storage capacity. The data quality score is based on the results of data integrity checks and outlier detection, with a value range of 0 - 100. The response time records the average delay of the node in processing requests, with a requirement of not exceeding 200 milliseconds. The storage capacity examines the data storage ability of the node, with a requirement of reserving at least 100GB of space. The scores of the three dimensions are weighted to obtain the comprehensive rights and interests value, with weights of 0.4, 0.3, and 0.3 respectively. The nodes with the top 30% of the rights and interests values are selected as verification nodes to participate in the consensus.

[0135] To achieve the interoperability of heterogeneous consensus, a unified state description format is adopted. The consensus state information is encoded into a byte array with a fixed length, including the block height, state root hash, and consensus proof signature. The Merkle tree is used to store the state data, with the leaf nodes saving the specific state values and the intermediate nodes storing the hash values of the child nodes. The cross-consensus message adopts a three-stage confirmation mechanism during transmission: the sender first broadcasts a preparatory message, sends a commit message after receiving more than half of the confirmations, and finally waits for the completion confirmation from the receiver. During the message transmission process, timeout retransmission is performed through heartbeat detection, and the timeout threshold is set to 1 second.

[0136] A state version chain is constructed to track the consensus switching process. A state version is generated for each consensus cycle, recording information such as the consensus type, participating nodes, and state snapshot. The checkpoint interval is set to 100 blocks, and the state consistency is verified at the checkpoint. Three types of metrics are monitored: consensus efficiency (number of transactions processed per second), node response time, and state consistency deviation value. When the consensus efficiency is lower than 1000 TPS, or more than 20% of the nodes have response timeouts, or the state consistency deviation exceeds 5%, the consensus switch is triggered. During the switch, the current consensus is paused for 30 seconds, and then the new consensus mechanism is started and runs stably.

[0137] The consensus result of the measurement traceability reference node is transmitted to the measurement data collection node through the unified state format, serving as the trusted basis for its rights and interests proof consensus. The measurement data confirmed by the data collection node is also fed back to the reference node through the cross-consensus message. The state version chain records the state evolution during the whole process, ensuring the state consistency between heterogeneous consensuses. The consensus switching mechanism realizes the dynamic adjustment of the two consensus algorithms.

[0138] Through a two - layer consensus architecture that combines dynamic view switching and proof - of - stake, the present invention improves the credibility and consensus efficiency of measurement data, and solves the problem that it is difficult for traditional single - consensus mechanisms to balance security and performance; by adopting a unified state description format and a cross - consensus message mechanism, it realizes interoperability and state synchronization between heterogeneous consensuses, and overcomes the obstacle of information interaction between different consensus algorithms; based on the design of a state version chain and a consensus switching mechanism, the system can dynamically adjust the consensus strategy according to the running state, enhancing the adaptability and reliability of the system, and ensuring the continuity and consistency of measurement data.

[0139] In an alternative embodiment,

[0140] The steps of establishing an access control policy based on behavioral characteristics, triggering hierarchical response measures according to an anomaly score, and providing a traceability analysis result to the user terminal include:

[0141] Obtain the identity characteristics and behavioral characteristics of the user's access behavior. The identity characteristics include the user's static identifier, dynamic authentication information, and environmental attributes. The behavioral characteristics include temporal behavioral characteristics, operation behavioral characteristics, and data interaction characteristics. Use an auto - encoder to reduce the dimension of the identity characteristics and behavioral characteristics to obtain a feature vector, and calculate the matching degree between the feature vector and the historical baseline based on weighted cosine similarity to obtain a real - time trust score;

[0142] Based on the real - time trust score, divide the access control levels. When the real - time trust score is in the first trust interval, execute the basic access control policy to allow access with the original permissions. When the real - time trust score is in the second trust interval, execute the enhanced access control policy for secondary authentication and limit the access frequency and data range. When the real - time trust score is in the third trust interval, execute the strict access control policy to only allow read - only operations and require multi - factor authentication;

[0143] According to the change trend of the real - time trust score within a continuous observation period, dynamically adjust the control intensity of the access control policy, and generate an access control result including user identity information, access behavior characteristics, and control policies;

[0144] Calculate an anomaly score based on the access control result, construct an anomaly behavior hierarchical response mechanism, and trigger a warning - level response, a blocking - level response, or a locking - level response according to the anomaly score when an abnormal behavior occurs. The warning - level response includes recording the abnormal event and starting feature resampling. The blocking - level response includes suspending the session permission and requiring re - authentication of the identity. The locking - level response includes freezing the access permission and triggering a security audit;

[0145] Construct an association graph of user operation resources, record the occurrence time, abnormal type, response measures and disposal results of abnormal behaviors, establish an abnormal behavior feature library, and push real-time to the user terminal traceability analysis results including metering data query, calibration cycle reminder, wear warning and maintenance suggestions according to the level of the abnormal score.

[0146] Exemplarily, such as Figure 2 As shown in the multi-level access control flow chart based on behavior characteristics, first, it is necessary to collect user access behavior data. Identity feature collection includes static identification information such as user accounts, login credentials, IP addresses, device fingerprints, etc., authentication information such as dynamic passwords, biometric features, and environmental attributes such as time and geographical location. Behavior feature collection covers temporal features such as the time interval and operation sequence of user operations, operation features such as command types and parameter values, and interaction features such as data access volume and data flow direction.

[0147] In the feature processing stage, an autoencoder is used for feature dimensionality reduction. The input layer receives the original feature vector, compresses the high-dimensional features into a low-dimensional latent space through multiple layers of encoders, and then reconstructs the features through the decoder. Taking the user login scenario as an example, the original input features include 20 dimensions such as login time, location, and device, and are compressed into an 8-dimensional feature vector through an encoder structure of 64-32-16-8. Based on the weighted cosine similarity, the matching degree between the feature vector and the user's historical behavior baseline is calculated to obtain a real-time trust score between 0 and 1.

[0148] The access control policy performs hierarchical responses based on the trust score. When the trust score is in the range of 0.8 - 1.0, basic access control is executed to allow users to access system resources normally. When the score is in the range of 0.5 - 0.8, enhanced access control is triggered, requiring users to perform secondary authentication such as SMS verification codes, and at the same time limiting the access frequency to no more than 10 times per minute and the data query range to the user's affiliated department. When the score is below 0.5, strict access control is started, only allowing read-only data operations, and requiring multi-factor authentication such as face recognition.

[0149] The system continuously monitors the change of the trust score within a 30-minute observation period. If the score shows a downward trend, the access control intensity is correspondingly increased; if the score stabilizes and rebounds, the restrictions are appropriately relaxed. Finally, an access control result including user identification, behavior characteristics, and control policies is generated.

[0150] The abnormal behavior hierarchical response mechanism calculates an abnormal score based on the access control result. When the abnormal score is in the range of 0.3 - 0.5, a warning-level response is triggered, recording the abnormal event and starting feature resampling; when the abnormal score is in the range of 0.5 - 0.8, a blocking-level response is triggered, suspending the current session and requiring re-identification of the identity; when the abnormal score exceeds 0.8, a locking-level response is triggered, freezing the account and conducting a security audit.

[0151] The traceability analysis is based on the construction of an association graph of user-operation-resource. Information such as the occurrence time, type, and response measures of abnormal behaviors is recorded, and an abnormal feature library is established for similar case matching. According to the abnormal scoring level, the analysis results are pushed to the user terminal, including the time track, influence scope, and handling suggestions of abnormal operations, etc.

[0152] The present invention improves the accuracy and timeliness of abnormal behavior detection through multi-dimensional feature analysis and adaptive access control, effectively preventing internal threats and data leakage risks; realizing precise control based on a hierarchical response mechanism, maximizing the system availability while ensuring security, and improving the user experience and work efficiency; the real-time feedback of the traceability analysis results helps users understand and improve their operation behaviors, and at the same time provides decision-making support for security management personnel, promoting the improvement of security awareness and the optimization of management efficiency.

[0153] Figure 3 This is a schematic structural diagram of the weight measurement data acquisition and traceability management system based on the Internet of Things according to an embodiment of the present invention. As Figure 3 shown, the system includes:

[0154] The first unit is used to implant an RFID electronic tag storing weight identity information into the weight body, collect temperature parameters, humidity parameters, and air pressure parameters through an environmental parameter sensor, and collect weight data of the weight through a weight data acquisition module provided with a high-precision strain gauge; perform timestamp marking on the weight identity information, environmental parameters, weight data of the weight, and metrological standard instrument information to form metrological raw data;

[0155] The second unit is used to send the metrological raw data to an edge computing server through an Internet of Things communication module with data encryption function for preprocessing of outlier detection and data smoothing, and transmit the preprocessed data to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data of the weight according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, corrects the coefficient of thermal expansion in combination with the material characteristics of the weight, generates metrological calibration data, and at the same time predicts the wear rate of the weight based on historical metrological data;

[0156] The third unit is used to construct a security isolation architecture, encrypt the metrological calibration data and the predicted wear rate based on the security isolation architecture, construct an adaptive traceability chain structure including a basic chain and an extended chain, write the metrological calibration data and the wear rate prediction result into a blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the whole process of metrology; establish an access control policy based on behavior characteristics, trigger hierarchical response measures according to abnormal scoring, and provide traceability analysis results to the user terminal.

[0157] In the third aspect of the embodiments of the present invention,

[0158] a kind of electronic device is provided, including:

[0159] a processor;

[0160] a memory for storing instructions executable by the processor;

[0161] wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0162] In the fourth aspect of the embodiments of the present invention,

[0163] a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0164] The present invention may be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are loaded.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collecting and tracing management of weight measurement data based on the Internet of Things, characterized in that, Including: An RFID electronic tag storing weight identity information is implanted into the weight body, and temperature parameters, humidity parameters, and air pressure parameters are collected through an environmental parameter sensor to collect weight data of the weight; The weight identity information, environmental parameters, weight data, and metrological standard instrument information are timestamped and then combined into original metrological data; The original metrological data is sent to an edge computing server through an Internet of Things communication module for data smoothing preprocessing, and the preprocessed data is transmitted to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, corrects the coefficient of thermal expansion in combination with the material characteristics of the weight to generate metrological calibration data, and simultaneously predicts the wear rate of the weight based on historical metrological data; A secure isolation architecture is constructed, and the metrological calibration data and the predicted wear rate are encrypted based on the secure isolation architecture. An adaptive traceability chain structure including a basic chain and an extended chain is constructed, and the metrological calibration data and the wear rate prediction result are written into a blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the entire metrological process; An access control strategy based on behavioral characteristics is established, and hierarchical response measures are triggered according to the anomaly score, and the traceability analysis result is provided to the user terminal.

2. The method according to claim 1, characterized in that, The steps for the cloud data processing server to establish a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, perform multi-dimensional environmental compensation on the weight data according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, and correct the coefficient of thermal expansion in combination with the material characteristics of the weight to generate metrological calibration data include: A multi-branch attention network structure is constructed using a deep learning algorithm. The multi-branch attention network structure includes a time stream and a space stream dual-stream architecture, and multi-scale features are adaptively fused to obtain a mapping result; Based on the mapping result, Kalman filtering and wavelet transform are used to perform spatio-temporal dimension compensation on environmental parameters, and a non-linear coupling relationship of temperature parameters, humidity parameters, and air pressure parameters is constructed and combined with material thermodynamics calculation and molecular dynamics simulation for physical dimension compensation; The multi-dimensional environmental compensation also includes intelligent dimension compensation: constructing a heterogeneous graph neural network to model the interaction relationship of environmental parameters, taking temperature parameters, humidity parameters, and air pressure parameters as graph network nodes, capturing the dynamic association between parameters through an attention edge convolution layer, and introducing a skip connection to retain historical interaction information; Designing a deep reinforcement learning model with a dual reward mechanism to dynamically adjust the compensation strategy. The dual reward mechanism includes an environmental stability reward and a measurement accuracy reward, and the policy gradient algorithm based on the advantage function is used to optimize the compensation action; Constructing a hybrid architecture of transfer learning and multi-task learning, extracting the general features of environmental parameters through a shared representation layer, eliminating the distribution differences under different working conditions through adaptive feature alignment, and combining a multi-head attention mechanism for fine-grained expression of features to complete the optimization of the compensation strategy; Perform multi-scale thermal expansion correction on the compensated data, establish a stress-strain tensor at the lattice scale to describe the thermal deformation characteristics and calculate the contribution of atomic vibration to volume change; determine the thermal expansion correlation characteristics through dynamic simulation of grain boundary sliding effect and dislocation motion analysis at the microscale; establish a multi-scale thermodynamics equation set at the macroscale to couple and calculate elastic deformation and thermal expansion. Input the corrected data into the calibration data generation module, perform outlier detection and filter denoising on the data, and perform adaptive weighted fusion on the deep learning prediction results, multi-dimensional environment compensation results, and thermal expansion correction results to generate metrological calibration data.

3. The method according to claim 2, characterized in that Use a deep learning algorithm to construct a multi-branch attention network structure. The multi-branch attention network structure includes a dual-stream architecture of time flow and space flow. The steps of adaptively fusing multi-scale features to obtain a mapping result include: Use DenseNet as the backbone network to construct a multi-branch attention network structure. The multi-branch attention network structure includes a time flow branch and a space flow branch. The multi-branch attention network structure is provided with a branch interaction module. The branch interaction module performs information exchange between branches through a cross-attention mechanism, constructs a feature map similarity matrix to calculate attention weights, and controls information flow through a gating mechanism. The time flow branch extracts temporal features through causal convolution, sets a time step attention mechanism to calculate weights for the temporal features, and models the temporal features through an LSTM layer; the time flow branch is provided with a temporal feature processing module. The temporal feature processing module constructs a temporal multi-scale feature pyramid and processes the features through a temporal attention pooling layer. The space flow branch extracts spatial features through a spatial attention convolutional layer and sets deformable convolution to deform the convolution kernel; the space flow branch is provided with a spatial relationship modeling module. The spatial relationship modeling module calculates the correlation relationship of spatial features through a graph attention network and introduces position encoding to mark the spatial positions of the features. Adaptive fusion of the features of the time flow branch and the space flow branch includes: unifying the feature dimensions through a feature transformation network and aligning the feature resolutions using a learnable upsampling module. Construct a dynamic weight prediction network to calculate the feature fusion weights and obtain the final mapping result through an attention-guided feature aggregation module. Optimize network training using a multi-task loss function with adaptive weights. The multi-task loss function includes the Focal L1 loss of the main task, as well as the feature consistency loss and temporal smoothness loss of the auxiliary tasks. Calculate the loss weights through an adaptive weight learning method based on task uncertainty; adopt a staged training strategy, train the time flow branch and the space flow branch, the branch interaction module, and the feature fusion module in sequence, and finally perform end-to-end optimization.

4. The method according to claim 1, wherein The steps for predicting the wear rate of weights based on historical metrological data include: Preprocess the historical metrological data using sliding median filtering, extract features such as mass change rate, environmental parameters, usage intensity, and surface roughness, and perform standardization processing to obtain a standardized feature sequence. Input the standardized feature sequence into a bidirectional LSTM network, extract bidirectional features through forward propagation and backward propagation. The bidirectional LSTM network is provided with a time step calculation module to quantify the degree of temporal correlation, and perform weighted combination of the bidirectional features and the degree of temporal correlation to obtain a deep feature representation of the weight wear. Construct a multi-task learning framework based on the deep feature representation. The multi-task learning framework includes a wear pattern classification branch and a rate prediction branch. The wear pattern classification branch classifies wear into normal wear, accelerated wear, and abnormal wear through a fully connected network and is optimized using focal loss. The rate prediction branch predicts the wear rate in a future time period through a residual fully connected structure. A prediction reliability evaluation module is constructed in the multi-task learning framework, and the mean and variance of the prediction results are output through a neural network and optimized using a negative log-likelihood loss function. An adaptive correction mechanism is constructed based on the prediction reliability evaluation result. When the prediction uncertainty exceeds a set threshold, increase the data sampling density and shorten the prediction time span, calculate the prediction deviation based on the latest measurement data, and update the correction coefficient through exponential moving average to perform online correction on the prediction result.

5. The method according to claim 1, characterized in that, Construct a security isolation architecture, encrypt the measurement calibration data and the predicted wear rate based on the security isolation architecture, and construct an adaptive traceability chain structure including a base chain and an extended chain. The steps of writing the measurement calibration data and the wear rate prediction result into the blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the entire measurement process include: Construct a security isolation architecture including a hardware trusted root layer, a trusted execution environment layer, and a security processing layer. The hardware trusted root layer stores keys and performs integrity measurement through a trusted platform module. The trusted execution environment layer divides the secure world and the ordinary world and performs key management. The security processing layer is provided with an independent data processing environment. Input the measurement calibration data and the predicted wear rate into the security isolation architecture for encryption processing, and construct an adaptive traceability chain structure based on the encrypted data. The adaptive traceability chain structure includes a base chain for storing measurement parameters and measurement calibration data and an extended chain for storing wear rate prediction results and measurement scenario data. Consistency maintenance of cross-chain data between the base chain and the extended chain is performed through an inter-chain data reference mechanism. Design a hierarchical consensus mechanism for the adaptive traceability chain structure, perform consensus processing on the measurement traceability reference nodes through a practical Byzantine fault tolerance algorithm with dynamic view switching, perform data confirmation on the measurement data collection nodes through proof-of-stake consensus, and adopt a consensus switching mechanism to be compatible with different consensus algorithms. Perform standardized conversion and quality evaluation on the measurement calibration data, use a spatio-temporal feature extraction algorithm to obtain measurement process characteristic parameters, establish a measurement process association matrix by calculating the temporal correlation degree and spatial similarity between the measurement process characteristic parameters and the predicted wear rate, and generate traceability information by executing data fusion operations based on the intelligent contract rule engine. Write the traceability information into the blockchain network, and establish a secure storage of the whole-process measurement information and a trustworthy traceability information chain based on the adaptive traceability chain structure and the hierarchical consensus mechanism.

6. The method according to claim 5, wherein The steps of performing consensus processing on the measurement traceability reference node by introducing the practical Byzantine fault tolerance algorithm with dynamic view switching, performing data confirmation on the measurement data collection node through proof-of-stake consensus, and adopting a consensus switching mechanism to perform compatibility processing on different consensus algorithms include: Design a practical Byzantine fault tolerance algorithm with dynamic view switching for the measurement traceability reference node. The practical Byzantine fault tolerance algorithm calculates the view liveness metric through the number of messages, consensus time, and view stability, triggers view switching and updates the primary node based on the view liveness metric, and makes the measurement traceability reference node enter a new view when receiving more than twice the fault tolerance limit plus one view switching message; Calculate the stake value of the measurement data collection node based on data quality, response time, and storage capacity, and use proof-of-stake consensus to determine the verification node according to the stake value and complete data confirmation; Standardize and encode the status information of the practical Byzantine fault tolerance algorithm and proof-of-stake consensus, organize the status data using a Merkle tree structure, and construct a unified status description format including block height, state root, and consensus proof; construct a cross-consensus message based on the unified status description format, transmit the cross-consensus message using a multi-level message confirmation mechanism, perform retransmission control and timeout detection on the cross-consensus message, authenticate the source of the cross-consensus message, and verify the legality of the state mapping and the state transition proof; Construct a state version chain and set checkpoints, monitor the consensus efficiency metric, node response status, and state consistency deviation. When the consensus efficiency metric is lower than the preset threshold, or the node response times out, or the state consistency deviation exceeds the limit, trigger consensus switching, broadcast a switching proposal to the nodes in the network and collect node confirmation information, pause the current consensus mechanism and activate the target consensus mechanism; Use the consensus result of the measurement traceability reference node as the trust basis for the measurement data collection node to perform proof-of-stake consensus through the unified status description format, feedback the data confirmed by the measurement data collection node to the measurement traceability reference node through the cross-consensus message, maintain the state consistency of heterogeneous consensus based on the state version chain, and adopt a consensus switching mechanism to adjust between the practical Byzantine fault tolerance algorithm and proof-of-stake consensus.

7. The method according to claim 1, characterized in that, The steps of establishing an access control policy based on behavioral characteristics, triggering hierarchical response measures according to the anomaly score, and providing the traceability analysis result to the user terminal include: Obtain the identity characteristics and behavioral characteristics of the user's access behavior. The identity characteristics include the user's static identifier, dynamic authentication information, and environmental attributes. The behavioral characteristics include sequential behavioral characteristics, operation behavioral characteristics, and data interaction characteristics. Use an autoencoder to reduce the dimensionality of the identity characteristics and behavioral characteristics to obtain a feature vector, and calculate the matching degree between the feature vector and the historical baseline based on the weighted cosine similarity to obtain the real-time trust score; Based on the real-time trust score, access control levels are divided. When the real-time trust score is in the first trust interval, the basic access control policy is executed to allow the original permissions for access. When the real-time trust score is in the second trust interval, the enhanced access control policy is executed for secondary authentication and to limit the access frequency and data scope. When the real-time trust score is in the third trust interval, the strict access control policy is executed to only allow read-only operations and require multi-factor authentication; According to the change trend of the real-time trust score within a continuous observation period, the control intensity of the access control policy is dynamically adjusted to generate an access control result including user identity information, access behavior characteristics, and control policies; Based on the access control result, an anomaly score is calculated, and an abnormal behavior hierarchical response mechanism is constructed. When an abnormal behavior occurs, a warning-level response, a blocking-level response, or a locking-level response is triggered according to the anomaly score. The warning-level response includes recording the abnormal event and starting feature resampling. The blocking-level response includes suspending the session permissions and requiring re-authentication of the identity. The locking-level response includes freezing the access permissions and triggering a security audit; An association graph of user operation resources is constructed, recording the occurrence time, abnormal type, response measures, and disposal results of abnormal behaviors, establishing an abnormal behavior feature library, and real-time pushing traceability analysis results including metering data query, calibration period reminder, wear warning, and maintenance suggestions to the user terminal according to the level of the anomaly score.

8. An IoT-based weight measurement data acquisition and traceability management system for implementing the method described in any one of the foregoing claims 1-7, characterized in that It includes: The first unit is used to implant an RFID electronic tag storing the weight identity information into the weight body, collect temperature parameters, humidity parameters, and air pressure parameters through an environmental parameter sensor, and collect weight data of the weight through a weight data acquisition module with a high-precision strain gauge; timestamp the weight identity information, environmental parameters, weight data of the weight, and metrological standard instrument information to form metrological raw data; The second unit is used to send the metrological raw data to an edge computing server through an Internet of Things communication module with data encryption function for preprocessing of outlier detection and data smoothing, and transmit the preprocessed data to a cloud data processing server. The cloud data processing server establishes a mapping relationship between environmental parameters and the weight of the weight based on a deep learning algorithm, performs multi-dimensional environmental compensation on the weight data of the weight according to the coupling relationship of temperature parameters, humidity parameters, and air pressure parameters, corrects the coefficient of thermal expansion in combination with the material characteristics of the weight, generates metrological calibration data, and predicts the wear rate of the weight based on historical metrological data; The third unit is used to construct a security isolation architecture, encrypt the metrological calibration data and the predicted wear rate based on the security isolation architecture, construct an adaptive traceability chain structure including a basic chain and an extended chain, and write the metrological calibration data and the wear rate prediction result into the blockchain network through a hierarchical consensus mechanism to form a traceability information chain for the whole metrological process; establish an access control policy based on behavior characteristics, trigger hierarchical response measures according to the anomaly score, and provide traceability analysis results to the user terminal.

9. An electronic device, characterized in that, It includes: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images