Blockchain-based wire and cable production data traceability method

Abstract

The application discloses a blockchain-based wire and cable production data tracing method, and relates to the technical field of credible storage and tracing of industrial production data; the method comprises the following steps: encapsulating production data into process data blocks at each process node and calculating the first hash value of the process data blocks; obtaining the hash value of the previous process, splicing the hash value with the current hash value to calculate a chain hash, and then adding the two to the current data block to form an enhanced process data block and calculating the second hash value of the enhanced process data block; submitting a storage transaction containing the second hash and the chain hash to a consensus storage of a consortium chain network; when tracing, obtaining a hash sequence from the chain and corresponding enhanced data blocks from off-chain, and verifying the data integrity and the correlation authenticity between processes by comparison. The application realizes the tamper-proofing and credible tracing of production full-process data by constructing a cryptographic hash chain between processes and storing a fingerprint in a consortium chain, and fundamentally improves the security of production data.

Description

Technical Field

[0001] This invention relates to the field of trusted storage and traceability technology for industrial production data, specifically to a blockchain-based method for tracing wire and cable production data. Background Technology

[0002] In continuous industrial production such as wire and cable manufacturing, the quality of the final product is determined by the process parameters of multiple processes (such as wire drawing, stranding, extrusion, and shearing). When product quality defects occur, accurately locating the problematic process and clarifying responsibility relies on reliable retrospective analysis of historical production data.

[0003] Currently, the centralized data management systems commonly used in the industry (such as MES and SCADA systems based on relational databases) suffer from the following problems caused by inherent defects in their technical architecture: Single point of trust and tampering risk in data storage: All production data is centrally stored on one or a few database servers controlled by the manufacturer. The entire path from data generation to storage, and the data itself after storage, depends on the ethics and system security capabilities of a single administrator. Insiders or external attackers can directly access the database to tamper with, delete, or insert forged data, and such operations are difficult to detect in post-audit. This means that the data on which traceability is based does not possess the property of self-proof.

[0004] The logical nature and vulnerability of data relationships: The sequence and hierarchical relationships between process data are usually maintained through foreign key constraints in database tables or application logic code. This relationship is logical, not physical. When an attacker or a erroneous operation tamperes with a data record, they can simultaneously tamper with or destroy these foreign keys and logic, thereby forging or severing the real connection between processes, and the system cannot independently verify the authenticity of this relationship.

[0005] Lack of data integrity verification: Existing systems lack a simple, unforgeable mechanism to prove that a data record has never been altered since its creation. Even with database logs, the logs themselves are still at risk of tampering. In quality disputes, the disputing parties cannot independently, independently of the data custodian (usually the manufacturer), perform cryptographic-level verification of the originality and integrity of the provided traceability data.

[0006] Existing technological solutions attempt to establish undeniable trust within a centrally controlled data environment, which is inherently contradictory. Their technical architecture dictates that the state of data and its associated information is mutable, thus rendering it unsuitable as credible evidence for arbitrating liability.

[0007] Therefore, there is an urgent need for a new technical solution that can build a data storage and traceability system that does not rely on the integrity of any single participant. Summary of the Invention

[0008] The purpose of this invention is to provide a blockchain-based method for tracing production data of wires and cables, which solves the key technical problem that under a centralized data management architecture, the production data of multiple processes of wires and cables is easily tampered with and the correlation is easily forged, resulting in the data being unable to serve as credible evidence during quality traceability.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A blockchain-based method for tracing wire and cable production data includes the following steps: Process data block generation steps: At each node of the production process, the production data and identification information of the corresponding process are encapsulated into a process data block, and the first hash value of the process data block is calculated; The process hash chain construction steps are as follows: Obtain the first hash value corresponding to the previous process data block; concatenate the first hash value of the previous process data block with the first hash value of the current process data block, and calculate the chain hash value based on the concatenation result; add the first hash value of the previous process data block, the chain hash value, and the first hash value of the current process data block to the current process data block to form an enhanced process data block, and calculate the second hash value of the enhanced process data block. Blockchain evidence storage steps: Submit the evidence storage transaction containing the second hash value and the chain hash value to the blockchain network so that the consensus nodes in the blockchain network can reach a consensus on the evidence storage transaction and record it in the distributed ledger; Trusted traceability verification steps: In response to the traceability request, obtain the second hash value and chain hash value related to the target batch from the distributed ledger, and obtain the corresponding enhanced process data block from the off-chain storage; by comparing the hash value obtained from the distributed ledger with the hash value obtained from the distributed ledger, verify the integrity of the off-chain data and the authenticity of the relationship between processes.

[0010] Furthermore, in the process data block generation step, the production data includes process parameters and equipment status data, and the identification information includes material batch identifier, timestamp, and node identity identifier; the process data block generation is driven by a preset triggering rule, which is to reach a preset time window or complete a specific production event.

[0011] Furthermore, in the hash chain construction step, the cryptographic hash function used to calculate the first hash value, the second hash value, and the chain hash value is the SHA-256 or SM3 algorithm.

[0012] Furthermore, in the blockchain evidence storage step, the blockchain network is a consortium blockchain network, and its consensus nodes include at least two types of nodes: producer nodes, quality inspection nodes, and customer nodes.

[0013] Furthermore, the consensus node uses the Practical Byzantine Fault-Tolerant (PBFT) consensus algorithm or the Raft consensus algorithm to reach a consensus on the evidence storage transaction.

[0014] Furthermore, in the blockchain notarization step, the notarization transaction also includes a timestamp and a digital signature generated by the node's private key; the consensus node verifies the validity of the digital signature before reaching consensus.

[0015] Furthermore, in the trusted traceability verification step, verifying the integrity of off-chain data specifically includes: calculating the real-time hash value of each acquired enhanced process data block and comparing it with the corresponding second hash value obtained from the distributed ledger.

[0016] Furthermore, the verification of the authenticity of the inter-process relationship specifically includes: for the j-th enhanced process data block, where j≥2, recalculating the chain hash value based on the first hash value of the previous process data block stored in it and the first hash value stored in it, and comparing it with the j-th chain hash value obtained from the distributed ledger.

[0017] Furthermore, after the trusted traceability verification step is passed, the method further includes: performing a root cause analysis of defects based on the production data in the verified enhanced process data block.

[0018] Furthermore, following the trusted traceability verification step, a quality risk prediction step is also included: Based on the enhanced process data blocks of multiple historical production batches obtained from the distributed ledger and off-chain storage and verified for integrity, a deep time-series quality prediction model integrating multiple processes is constructed. The quality prediction model uses the real-time process parameter sequence of the first to kth processes completed in the production process for the target batch. Given the input, output the predicted probability that a specific quality defect will occur in the batch in subsequent processes or the finished product stage. And a contribution vector representing the magnitude of each completed process's contribution to the current predicted risk. ;in, This represents the actual sequence of process parameters for the completed processes in the target batch. Indicates the number of completed processes and , This represents the predicted defect probability value output by the model, and , This represents the contribution vector of each process to defect prediction, and .

[0019] Furthermore, it also includes a step for dynamic adjustment of online process parameters: The defect prediction probability output by the quality risk prediction step and contribution vector , and the currently ongoing first Real-time process parameters of each process They are all input into an online optimization decision model; The online optimization decision model has a dual objective of minimizing the predicted defect probability and the control parameter adjustment amplitude. It is solved under all process constraints and outputs the result for the [missing information]. Recommended optimal process parameter adjustment values ​​for the first process and subsequent processes to be performed ; The optimal process parameter adjustment suggestion value corresponding to the first Part of the process The data is sent in real time to the corresponding production equipment controller for dynamic adjustment. in, Indicates the first Real-time process parameter vectors for each process step Indicates from the first To the The optimal adjustment value vector for each process step. This represents the total number of processes in the entire production process, and .

[0020] Furthermore, following the quality risk prediction step, a step of quantifying the propagation of inter-process anomalies is also included: The quantification step of inter-process anomaly propagation is based on the process contribution vector output by the quality prediction model. And the first extracted from the validated enhanced process data block To the Real-time process parameters of each process Build and run a quantitative model of inter-process influence propagation. ; The inter-process influence propagation quantification model The input is the first To the Process parameter deviation vector of each process step ,in , , Indicates from the first The first step extracted from the verified enhanced process data block corresponding to the first process step The actual measured values ​​of each process parameter Indicates the first Step 1 Standard settings for each process parameter Indicates the first The total number of process parameters monitored in each process step; The inter-process influence propagation quantification model The output is a Real-time influence intensity matrix The real-time influence intensity matrix elements in Used to quantify the current production state, the first Abnormal process parameters in the first process affect the second process. The intensity of the real-time impact of the process status of each step, among which... and All are process indexes. and ; The real-time influence intensity matrix Used to locate the critical path of anomaly propagation from the source process to the current process.

[0021] Furthermore, the inter-process influence propagation quantification model The construction and operation process specifically includes: Quantitative model of inter-process influence propagation It is constructed using a hybrid architecture that integrates long short-term memory network modules and structural equation modules; The inter-process influence propagation quantification model The dynamic propagation process of anomalies in the process chain is simulated using the following state update formula: ; In the state update formula, This indicates the first time in the simulation of abnormal propagation. At the time step, the first The implicit state vector of the process, the implicit state vector Used to encode the The overall process status of a process after an abnormality in a preceding process. This refers to the Long Short-Term Memory (LSTM) network module. It is the set of trainable parameters for the Long Short-Term Memory (LSTM) network module; Indicates the first The simulated time step is input into the Long Short-Term Memory network module. The process parameter deviation vector of each process step; Indicates the first The simulation time step, dynamically calculated by the structural equation module, is from the [number]th simulation time step. Step 1 to the next The real-time impact weight of each process step; This represents the preset activation function; and It is a process index; The real-time impact weight The following formula is calculated by the structural equation module: ; In the formula for calculating the real-time impact weight, This represents the Sigmoid activation function; and It consists of the trainable parameter matrix and bias vector of the structural equation module; This represents a multilayer perceptron module. It is the set of trainable parameters for the multilayer perceptron module; This represents a vector concatenation operation; and They represent the first time. The simulation time step The deviation vector of each process is summed in Time step Deviation vector of each process step; The inter-process influence propagation quantification model is applied in real time. At that time, the process parameter deviation vector of the target production batch will be... As the initial input sequence, forward propagation iterative calculation is performed using the state update formula until the hidden state vector of each process is obtained. When the change is less than the preset convergence threshold, a stable final influence weight is obtained. ; The final influence weight The real-time influence intensity matrix is ​​generated by performing peer normalization. The specific formula for normalization is: ,in, It is the index of the summation operation, representing all operations related to the target process. The preceding process that has an impact.

[0022] The present invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the blockchain-based data traceability method for wire and cable production as described above.

[0023] Compared with the prior art, the present invention has the following beneficial effects: This invention transforms the physical production process into a cryptographically verifiable digital evidence chain through local hashing of process data blocks and the construction of hash chains between processes. Unlike traditional centralized databases that only store data content, this solution mandates that each edge node immediately calculates its hash value when generating a process data block and actively embeds the hash value of the previous process to generate a chained hash. This design ensures that any subtle tampering with a single data block will cause a drastic change in its hash value, leading to the failure of all subsequent chained hash verifications, thus achieving tamper-proofing from both the data generation source and the association logic level. Its innovation lies in creatively applying the core idea of ​​blockchain hash pointers to a linear production process with strict temporal relationships, so that the order and attribution between processes no longer depend on fragile database foreign keys or application logic, but are guaranteed by the cryptographic algorithm itself to ensure its unforgeability.

[0024] This invention employs a collaborative architecture of consortium blockchain for fingerprint evidence storage and off-chain storage of complete data. While ensuring data trustworthiness, it also addresses the practical needs of industrial scenarios regarding data storage capacity and query performance. Only the data hash value and key metadata are stored on the consortium blockchain, resulting in a small data volume and high consensus efficiency, suitable for the characteristics of distributed ledgers. The complete process data blocks remain stored on a high-performance off-chain platform. When traceability verification is required, the authorized party can retrieve the data from off-chain and independently verify its integrity by comparing it with the hash value stored on-chain. This innovative design breaks away from the conventional mindset of full on-chain data storage, achieving separation of the trust anchor and the data carrier. It achieves decentralized trust while avoiding the limitations of blockchain storage bottlenecks on system usability.

[0025] The trusted data foundation established by this invention provides a solid prerequisite for advanced quality analysis. Since all traceability data has undergone cryptographic integrity verification, the conclusions drawn from defect analysis, liability determination, and even process optimization based on this data are highly authoritative and non-repudiable. This changes the previous situation where parties endlessly debated the authenticity of basic data during quality traceability, shifting the focus back from whether the data is trustworthy to the technical essence of how to analyze the data to solve problems, greatly improving the efficiency of quality traceability and the credibility of arbitration.

[0026] This invention, through a series of interconnected technical designs, has for the first time achieved end-to-end trustworthiness in the field of industrial manufacturing data traceability, from data generation and association to verification. Its effect is not a gradual improvement on existing systems, but rather the construction of a completely new data storage and traceability method that does not rely on centralized trust. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 This is an overall flowchart of the method described in this invention.

[0029] Figure 2 This is the core flowchart for production data traceability in this invention.

[0030] Figure 3 This is a flowchart illustrating the core traceability process for wire and cable production data in this invention. Detailed Implementation

[0031] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0032] The following is in conjunction with the appendix Figures 1-3 The embodiments of the present invention will be described in detail below.

[0033] Example 1: This example discloses a blockchain-based method for tracing production data of wires and cables, characterized by the following steps: Process data block generation steps: At each node of the production process, the production data and identification information of the corresponding process are encapsulated into a process data block, and the first hash value of the process data block is calculated; The process hash chain construction steps are as follows: Obtain the first hash value corresponding to the previous process data block; concatenate the first hash value of the previous process data block with the first hash value of the current process data block, and calculate the chain hash value based on the concatenation result; add the first hash value of the previous process data block, the chain hash value, and the first hash value of the current process data block to the current process data block to form an enhanced process data block, and calculate the second hash value of the enhanced process data block. Blockchain evidence storage steps: Submit the evidence storage transaction containing the second hash value and the chain hash value to the blockchain network so that the consensus nodes in the blockchain network can reach a consensus on the evidence storage transaction and record it in the distributed ledger; Trusted traceability verification steps: In response to the traceability request, obtain the second hash value and chain hash value related to the target batch from the distributed ledger, and obtain the corresponding enhanced process data block from the off-chain storage; by comparing the hash value obtained from the distributed ledger with the hash value obtained from the distributed ledger, verify the integrity of the off-chain data and the authenticity of the relationship between processes.

[0034] In practical implementation, the architecture includes edge acquisition and preprocessing nodes deployed at various production processes such as wire drawing, stranding, and extrusion; a consortium blockchain network jointly maintained by the manufacturer (M), authoritative quality inspection agency (Q), and key customer (C) as consensus nodes; and an off-chain big data storage platform for storing complete process data blocks. Edge nodes are responsible for data encapsulation and hash calculation, the consortium blockchain is responsible for fingerprint evidence storage, and the off-chain platform is responsible for raw data storage. The three work together to achieve trusted traceability.

[0035] The specific steps are as follows: Step S101: Structured encapsulation and local hash generation of process data blocks.

[0036] At the edge computing nodes of the extrusion process (such as industrial gateways), the preset trigger rule is that each standard reel (such as 5000 meters) is produced.

[0037] When the trigger condition is met, the node collects key process parameters within the production window: the average time series value of the mold temperature. Temperature difference between cooling water inlet and outlet traction speed Measured wire diameter .

[0038] At the same time, the batch number of the raw materials for this batch, MAT-2023-001-A, and the start timestamp of this process are also linked. End timestamp The node also includes its unique identifier, ID_Node_Extruder01. The node serializes and encapsulates this data according to a predefined JSON format, forming a structured extrusion process data block. After encapsulation, the node calls the built-in cryptographic hash function (SHA-256 is used as an example in this embodiment) to calculate the hash value of the data block: The calculated hash value might be, for example, the hexadecimal string a7f3d9c1e5... This hash value With the original data block Associative storage is performed locally on the edge nodes.

[0039] Step S102: Construct a cryptographic hash chain between processes.

[0040] The extrusion process node needs to establish a trusted link with the upstream stranding process. It requests the first hash value of its most recently generated process data block from the edge node of the stranding process via the network. .get Then, the extrusion process node executes the hash chain construction algorithm: First, the first hash value of the data block from the previous process is... The first hash value of the current process data block Perform concatenation and calculate the chain hash: '||' represents string concatenation. Then, modify the current process data block. Add three fields to the structure: "previous_hash": "chain_hash": and "first_hash": The modified data block is denoted as the enhanced process data block. Finally, the enhanced process data block containing the previous_hash, chain_hash, and first_hash fields is computed. From the hash value, we obtain its second hash value: .

[0041] The core of this step lies in constructing a dual hash chain (an enhanced data block hash chain and an independent chained hash), making it possible to access historical data blocks from any source. Any alteration will result in a change to its first hash value. The change, in turn, led to its enhanced data block hashing. Subsequent chain hashing and subsequent enhanced data block hashes All failed verification, thus achieving tamper-proof solidification of the process sequence from both the data content and the relationship level.

[0042] Step S103: Consensus and evidence storage based on consortium blockchain.

[0043] Constructing a certificate-based transaction for each node in the extrusion process The data structure of this transaction is as follows: { "block_hash":H_extrusion', / / The final hash of the enhanced data block "chain_hash":CH_extrusion, / / chain hash "timestamp":t_e, / / Process completion timestamp “node_id”:“ID_Node_Extruder01”, "signature": Sig(H_extrusion', CH_extrusion, t_e) / / Node private key signature }

[0044] Nodes use their private keys to sign the core content of a transaction (block_hash, chain_hash, timestamp), generating a signature. The transaction is then broadcast to the consortium blockchain network. Upon receiving the transaction, the consensus nodes (M, Q, C) of the consortium blockchain first verify the validity of the digital signature using their public keys. After successful verification, the transaction enters the consensus pool. The consortium blockchain employs the Practical Byzantine Fault Tolerance (PBFT) consensus algorithm, with the master node initiating a consensus round every 10 seconds or after accumulating 100 transactions. Nodes conduct multiple rounds of communication and voting to determine the validity of transactions, ultimately approving a batch of transactions (including...) The order and content of the hash values ​​were agreed upon and packaged into a new block, Block #1045. After confirmation by a majority of nodes, this block was permanently appended to the distributed ledger replicas on all nodes. At this point, the hash value... and It has obtained decentralized and tamper-proof evidence.

[0045] Step S104: Integrity verification and traceability of on-chain and off-chain collaboration.

[0046] Suppose that quality inspection agency Q discovers that the finished cable in batch MAT-2023-001-A has an insulation layer adhesion defect, and traceability is required.

[0047] On-chain fingerprint extraction: Q, as an authorized node of the consortium blockchain, invokes a smart contract specifically deployed for traceability, passing in the target batch number MAT-2023-001-A. This smart contract retrieves all notarized transactions containing this batch identifier from the ledger, sorts them according to the transaction timestamps, and returns a final hash value sequence related to that batch, arranged in chronological order. and chained hash sequences .

[0048] Off-chain data acquisition and integrity verification: Q sends a formal request to manufacturer M, requesting complete enhanced data blocks corresponding to these 5 processes. M retrieves and provides the data from its off-chain big data storage platform. Q performs two-step verification locally: The first step is data integrity verification: for each received... Recalculate its SHA-256 hash value Compare the calculation results with those obtained from the chain. Compare them one by one.

[0049] If all comparison results are equal, it proves that the off-chain data provided by M has not been tampered with since its storage. The second step is correlation verification: For data blocks starting from the second step, the chain hash is recalculated using their `previous_hash` and `first_hash` fields. and stored on the chain Comparison.

[0050] Verification conditions: If all are consistent, it proves that the sequence and relationship between all processes are true and reliable, and have not been destroyed or forged.

[0051] Defect analysis based on reliable data: After completing the aforementioned cryptographic verification and confirming the absolute trustworthiness of the obtained data chain, Q (or M) can use this high-fidelity process data to apply various data analysis algorithms (e.g., time series analysis or pattern recognition of verified trustworthy parameters of the extrusion process) to precisely pinpoint the specific abnormal process parameters causing adhesion (such as a momentary excessively high cooling water temperature). The entire process, from data acquisition to analysis conclusions, is built upon the trustworthy foundation guaranteed by cryptography.

[0052] It should be noted that the sensor types, process parameter types, specific edge node hardware, consortium blockchain consensus algorithms (which can be replaced by Raft, Kafka, etc.), cryptographic hash functions (which can be replaced by SM3, etc.), and smart contract business logic involved in this embodiment can all be adjusted and replaced according to the actual production environment, security level requirements, and agreements among consortium members. These adjustments and replacements based on the same principles all fall within the protection scope of this invention.

[0053] Example 2: This example is a further optimization based on Example 1. This example is used to solve the problem that the traceability analysis in the original invention lags behind the occurrence of defects, and provides a proactive quality control scheme that predicts in advance and intervenes in the process.

[0054] This embodiment is built on the trusted data foundation constructed in Embodiment 1. It uses blockchain-provided, tamper-proof historical production data to train an intelligent model that can accurately predict quality risks. Based on the prediction results, it realizes online dynamic optimization of production parameters, forming a complete technical system of trusted data, intelligent analysis, and closed-loop control.

[0055] Existing traceability technologies are all reactive, only able to locate and analyze defects after they occur, failing to prevent their occurrence and resulting in wasted materials and time. Production parameters largely rely on engineer experience or statistical analysis of limited historical data, making it difficult to adapt to real-time disturbances such as raw material fluctuations and equipment status drift, causing process windows to deviate from optimal states. In wire and cable production, there are strong coupling relationships between process parameters of upstream and downstream processes (e.g., strand tension affects extrusion pressure). Traditional single-process optimization methods cannot handle such complex cross-process influences, making it difficult to achieve global optimization. From problem identification to cause analysis and parameter adjustment, traditional processes are time-consuming, and the same type of defect may intermittently recur in multiple batches.

[0056] In specific implementation, based on the architecture of Example 1, two core intelligent modules, a quality prediction engine and an online optimization engine, were added.

[0057] Specifically, the steps include the following: Step S201: Construct a multi-process integrated quality prediction model based on historical blockchain data. S201.1 High-Quality Training Dataset Preparation: The quality of the training data directly determines the model performance. This step extracts the training data verified in step S104 from the consortium blockchain distributed ledger and its corresponding off-chain storage platform. Historical production batch data (e.g.) For the first... Each batch, its data can be represented as: in, Indicates the first The batch in An enhanced process data block for a process (containing the previous_hash and chain_hash fields). This represents the total number of processes.

[0058] From each In the process, the time series matrix of key process parameters is extracted. .in, Indicates the first Number of sampling time points in each process step This indicates the number of process parameters monitored for that process step. For example, for the extrusion process, It could be 8, corresponding to 8 parameters such as mold temperature and melt pressure.

[0059] At the same time, obtain the final quality label for this batch. For example, define This indicates that the insulation adhesion strength of this batch of finished cables exceeds the standard threshold (e.g., This belongs to a defective batch; This indicates a qualified batch.

[0060] S201.2 Data Preprocessing and Standardization: To ensure stable model training and accelerate convergence, each process parameter is standardized. in, It is the standardized parameter sequence. It is the original parameter sequence. and They are the first Step 1 All parameters The mean and standard deviation calculated from each historical batch.

[0061] S201.3 Deep Time Series Prediction Model with Multi-Process Fusion Building Model It adopts a divide-and-conquer-integration design approach.

[0062] Process-level feature extraction sub-network: Standardized parameter sequence for each process Input an independent Temporal Convolutional Network (TCN). TCNs are particularly good at capturing long temporal dependencies due to their dilated convolutional structure. ,in It is the first The high-level feature vector extracted from the first process As a preset feature space dimension, It is the first Parameters of each TCN subnetwork.

[0063] Inter-process Attention Fusion Layer: To quantify the impact of each process on the final quality, an attention mechanism is introduced. First, the characteristics of each process are calculated. Attention score : ,in, , , These are trainable parameters. It is the attention hidden layer dimension. For process-level feature vectors The dimension is then normalized using the Softmax function to obtain the attention weights. Contribution level: ,in For the summation index, a weighted summation is used. Finally, the weighted summation yields a feature representation that incorporates global information. : .

[0064] Defect probability prediction output layer: fuses features The defect probability is output through a fully connected layer and a sigmoid activation function. in, , These are the output layer parameters. It is the Sigmoid function. .

[0065] Model S201.4 training and validation used a prepared historical dataset, with binary cross-entropy as the loss function. The model parameters are optimized using the backpropagation algorithm and the Adam optimizer. Training is then conducted. During training, the dataset is divided into training, validation, and test sets in an 8:1:1 ratio to monitor the model's generalization performance and prevent overfitting. The large-scale, highly reliable data provided by blockchain is the fundamental guarantee for the model to achieve high accuracy.

[0066] Step S202, Online Real-time Quality Risk Prediction: When a new batch (e.g., batch ID: CABLE-2024-1001) reaches the [number]th production stage... Process (such as) Real-time prediction is initiated when the stranding process has just been completed.

[0067] Input data construction: Obtain the real-time process parameters of the completed processes (steps 1 to k) in this batch. And also use the historical average described in step S201.2. and standard deviation Standardize it to obtain For the first [event] that has not yet occurred To the Each step involves filling in a standardized zero vector (i.e., the expected value) to form a complete, fixed-dimensional model input tensor. .

[0068] Forward propagation computation: Input the pre-trained The model performs forward propagation computation.

[0069] Risk Output: The model outputs two key results: Defect prediction probability This indicates the likelihood that this batch will eventually develop defects if the current trend continues. The system sets a threshold (e.g., ...). ),when At that time, a high-level warning was triggered.

[0070] Process contribution vector It clearly indicates which process parameters anomalies among the completed processes contribute the most to the current high-risk prediction, thus pointing the way for subsequent optimization.

[0071] Step S203, Online dynamic optimization and execution of process parameters: If the risk is predicted If the warning threshold is exceeded, an online optimization process is immediately triggered, which aims to reverse quality risks by adjusting parameters of subsequent processes.

[0072] S203.1 The core of constructing real-time optimization problems is to find a parameter for subsequent processes (from the first...) To the Adjustment plan for process) This ensures that, while satisfying various process constraints, the final defect probability is minimized and the adjustment cost is reduced. This can be formalized as a constrained optimization problem: ; yes A lightweight surrogate model (such as a gradient boosting tree) is used to quickly evaluate different tuning schemes. The expected defect probability is given. The proxy model uses... Using historical input and output data for training is a computationally efficient approximation.

[0073] It is the first The original planning parameters for each process step.

[0074] It is the first The upper and lower limits of the process parameters for each step.

[0075] It is a weighting coefficient used to balance the two objectives of reducing risk and minimizing adjustments.

[0076] The optimization solution for S203.2 employs the Sequential Quadratic Programming (SQP) algorithm for real-time solution of this nonlinear constrained optimization problem. The SQP algorithm approximates the original problem by constructing a quadratic programming subproblem at the current iteration point and iteratively solving it until convergence. It is suitable for handling medium-sized engineering optimization problems requiring rapid solutions. The optimal parameter adjustment vector is obtained after the solution. .

[0077] S203.3 Optimization Results Execution and Feedback: Immediate Execution: The next immediately executed step (the one corresponding to the optimal solution) will be selected from the optimal solution. Adjustment amount of (Tao) The parameters are dynamically adjusted by sending them to the controller of the production equipment (such as an extruder) in real time through industrial communication protocols (such as OPCUA).

[0078] The subsequent process (the first) to Adjustment suggestions for the Tao The message is pushed to the production management interface to guide operators in making preparations in advance.

[0079] Closed-loop learning: The adjusted production data and the final actual quality results of this batch will be used as a new data sample and stored on the blockchain using the process described in Example 1. The system will periodically use this newly added, practice-verified data to... The model undergoes incremental learning and updates, enabling its predictive capabilities to continuously evolve over time, forming a data-driven decision-making loop where decisions generate new data, and the new data feeds back into the model to enhance its intelligence.

[0080] Implementation Effect Verification (Comparative Example): To objectively evaluate the effectiveness of this embodiment, a 90-day comparative test was conducted on three production lines of the same specification in a large cable factory. Comparative Example 1 (Traditional Method): Parameters were adjusted using a combination of statistical process control (SPC) charts and human experience.

[0081] Comparative Example 2 (Blockchain Traceability Only): Only the solution in Example 1 is implemented. Accurate traceability is achieved after a defect occurs, but parameter adjustment still relies on manual analysis.

[0082] Table 1: Comparison of Core Data from Test Results Table 1:

[0083] Together with Example 1, Example 2 constitutes a complete solution from trusted data generation to intelligent decision execution. It not only inherits the core advantages of blockchain technology—data tamper-proofing and traceability—but also further explores the deeper value of trusted data. Through artificial intelligence models, it achieves proactive quality prediction and dynamic protection, effectively addressing the long-standing pain points in the wire and cable manufacturing industry: lagging quality control and reliance on experience for optimization.

[0084] It can accurately predict the risk probability of defects several processes (an average of 2.3 processes) before they actually occur, providing a valuable time window for proactive intervention and transforming the quality control model from reactive firefighting to proactive fire prevention. Based on real-time risk predictions, it can automatically calculate and execute optimal parameter adjustment schemes, ensuring the production process is constantly dynamically adjusted around the optimal process window, improving quality stability and consistency. Through deep neural network models, it automatically learns and quantifies the impact weight (contribution) of each process parameter on the final quality, providing data-driven insights into understanding complex process interactions and supporting global collaborative optimization. The results of each production and adjustment are recorded by the blockchain system and used as new samples for model updates, allowing prediction and optimization capabilities to continuously evolve with the accumulation of production data. By reducing defect rates, scrap, shortening quality traceability time, and reducing unplanned downtime, it brings direct cost savings and efficiency improvements to enterprises.

[0085] Example 3: This example is mainly used to solve the problem of how, based on the confirmed reliability of the data, not only to locate the problematic process, but also to dynamically and quantitatively reveal the specific propagation path and impact intensity of quality abnormalities among multiple processes, thereby identifying the root cause and guiding precise intervention.

[0086] This embodiment achieves a leap from knowing where the problem is to understanding how the problem arises and propagates by introducing a quantitative model of inter-process influence propagation.

[0087] The specific steps are as follows: S601. Technical Principles and Input Preparation for Model Construction: The inter-process influence propagation quantification model The core principle is to view the production process as a dynamic system, where the output state (which may include anomalies) of the preceding process becomes the input to the subsequent process and influences its operational state. The model uses machine learning methods to learn the quantitative patterns of this state influence from historical, reliable data. Its construction and operation rely on two key inputs, both of which are directly derived from the reliable traceability system established in Example 1: Process contribution vector : Derived from the quality prediction model The predicted output for the current target batch. This vector indicates the relative contribution of each completed process to the current predicted quality risk, serving as a trigger signal and focus pointer for initiating deep propagation analysis.

[0088] Real-time process parameter deviation vector This is the core input of this model.

[0089] First, based on the traceability request, the notarized hashes of the relevant processes for the target batch are obtained from the blockchain distributed ledger, and the corresponding enhanced process data blocks are retrieved from the off-chain storage system. Cryptographic verification of the integrity of these off-chain data blocks and the authenticity of the inter-process relationships is then performed.

[0090] Only data blocks that pass verification can have their internal process parameters extracted. Only then is it considered credible. Subsequently, each parameter... Compared with the standard setting value retrieved from the process knowledge base By taking the difference, we obtain the process parameter deviation vector for each process step. . Each element in Both represent the degree to which the process deviates from the normal state in this parameter dimension.

[0091] S602. Detailed explanation of model architecture and training process: Quantitative model of inter-process influence propagation A hybrid architecture is adopted, which combines the advantages of long short-term memory networks and structural equation modeling: Long Short-Term Memory (LSTM) network module: Used to simulate the internal state evolution of a single process. This module can remember long-term dependencies, making it ideal for characterizing the continuous impact of process parameter deviations and the cumulative changes in the process's own state.

[0092] Structural Equation Module: Used for dynamically inferring real-time influence weights between processes. This module utilizes a lightweight neural network (multilayer perceptron). Analyze the deviation vectors between adjacent processes and output a weight value representing the strength of their influence. .

[0093] The offline training process of the model is as follows: collect a large amount of (e.g.) (Number) trusted data from historical production batches (i.e., enhanced process data blocks stored and verified via blockchain and their final quality labels). For each training sample, its processes... Sequence input model. During training, the model calculates the hidden states of each step through forward propagation. And finally, based on the state of the last process step, the quality risk is predicted. Through backpropagation, the loss function (such as cross-entropy loss) between the predicted risk and the actual quality label is minimized, thereby optimizing all trainable parameters in the model, including... , , and This process enables the model to automatically identify and quantify inter-process influence patterns from deviation data.

[0094] S603. Complete calculation process for real-time application of the model: Taking a specific production batch (product model: 0.6 / 1kV cross-linked polyethylene insulated power cable) as an example, this batch has completed the wire drawing process ( ), stranded wire ( ), internal shielding ( Three processes. The quality prediction model issues an early warning, and the contribution vector... Show the stranding process ( (They) made the greatest contribution.

[0095] Data acquisition and deviation calculation: After blockchain verification, trusted parameters are acquired and the deviation vector is calculated.

[0096] Wire drawing process ( ): .

[0097] stranding process ( ): .

[0098] Inner shielding process ( ): .

[0099] Model forward propagation iteration: As initial input ( (At that time), substitute into the model formula for iterative calculation.

[0100] initialization It is a zero vector.

[0101] First iteration ( Example: calculate : deal with and ,through , and Output, assuming .

[0102] renew : by and As input, output the new state. .

[0103] Similarly, calculate , And update .

[0104] Repeat the iteration until the state change converges (e.g., (Time). Assume stable weights are obtained: , , .

[0105] Generate the influence intensity matrix: Calculated according to the normalization formula: .

[0106] .

[0107] .

[0108] This yields the real-time influence intensity matrix. It precisely quantifies the intensity of anomaly propagation.

[0109] This embodiment directly addresses the problem of opaque anomaly propagation paths by transforming fuzzy correlation guesses into precise quantization matrices. With reliable process data, this method enables quantitative analysis and visualization of the dynamic paths and intensity of anomaly propagation in multi-process production flows, overcoming the limitations of traditional traceability methods that can only perform qualitative and static correlation analysis. As shown in Example 3, combined with... and This clearly identifies the wire drawing process as the root cause, and its impact is fully transmitted to the stranding process with 100% intensity. This allows corrective measures to directly target the problem, avoiding ineffective or inefficient adjustments in intermediate processes (stranding). Matrix The study clearly demonstrates the primary propagation path as "filament drawing → stranding → inner shielding," and quantifies the impact intensity of each segment (1.00, 0.80), providing data-driven insights into the formation mechanism of quality issues. In subsequent processes (such as extrusion), adjustments can be made based on... and The proportion of these parameters is used to assess the residual impact of upstream anomalies, thereby enabling the development of more scientific and forward-looking compensatory process parameters and improving the success rate of adjustments.

[0110] In a practical application at a cable factory, the traceability and handling effects of the "insulation layer pinhole defect" were compared before and after using this embodiment (90 days each): the root cause location accuracy increased from 68% to 92%. After correcting the located causes, the recurrence rate of the same type of defect decreased from 31% to 9%. The average decision-making time for traceability analysis of a single quality event was shortened from 4.5 hours to 1.2 hours.

[0111] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A blockchain-based method for tracing data in wire and cable production, characterized in that: Includes the following steps: Process data block generation steps: At each node of the production process, the production data and identification information of the corresponding process are encapsulated into a process data block, and the first hash value of the process data block is calculated; The process hash chain construction steps are as follows: Obtain the first hash value corresponding to the previous process data block; concatenate the first hash value of the previous process data block with the first hash value of the current process data block, and calculate the chain hash value based on the concatenation result; add the first hash value of the previous process data block, the chain hash value, and the first hash value of the current process data block to the current process data block to form an enhanced process data block, and calculate the second hash value of the enhanced process data block. Blockchain evidence storage steps: Submit the evidence storage transaction containing the second hash value and the chain hash value to the blockchain network so that the consensus nodes in the blockchain network can reach a consensus on the evidence storage transaction and record it in the distributed ledger; Trusted traceability verification steps: In response to the traceability request, obtain the second hash value and chain hash value related to the target batch from the distributed ledger, and obtain the corresponding enhanced process data block from the off-chain storage; by comparing the hash value obtained from the distributed ledger with the hash value obtained from the distributed ledger, verify the integrity of the off-chain data and the authenticity of the relationship between processes.

2. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, In the process data block generation step, the production data includes process parameters and equipment status data, and the identification information includes material batch identifier, timestamp, and node identity identifier; the process data block generation is driven by a preset trigger rule, which is to reach a preset time window or complete a specific production event.

3. The blockchain-based method for tracing wire and cable production data according to claim 1 or 2, characterized in that, In the hash chain construction step of the process, the cryptographic hash function used to calculate the first hash value, the second hash value, and the chain hash value is the SHA-256 or SM3 algorithm.

4. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, In the blockchain evidence storage step, the blockchain network is a consortium blockchain network, and its consensus nodes include at least two types of nodes: producer nodes, quality inspection nodes, and customer nodes.

5. The blockchain-based method for tracing wire and cable production data according to claim 4, characterized in that, The consensus nodes use either the Practical Byzantine Fault-Tolerant (PBFT) consensus algorithm or the Raft consensus algorithm to reach a consensus on the evidence storage transaction.

6. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, In the blockchain notarization step, the notarization transaction also includes a timestamp and a digital signature generated by the node's private key; the consensus node verifies the validity of the digital signature before reaching consensus.

7. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, In the trusted traceability verification step, verifying the integrity of off-chain data specifically includes: calculating the real-time hash value of each enhanced process data block obtained, and comparing it with the corresponding second hash value obtained from the distributed ledger.

8. The blockchain-based method for tracing wire and cable production data according to claim 7, characterized in that, The verification of the authenticity of the inter-process relationship specifically includes: for the j-th enhanced process data block, where j≥2, recalculate the chain hash value based on the first hash value of the previous process data block stored in it and the first hash value stored in it, and compare it with the j-th chain hash value obtained from the distributed ledger.

9. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, After the trusted traceability verification step is passed, the method further includes: performing a root cause analysis of defects based on the production data in the verified enhanced process data block.

10. The blockchain-based method for tracing wire and cable production data according to claim 1, characterized in that, Following the trusted traceability verification step, a quality risk prediction step is also included: Based on enhanced process data blocks of multiple historical production batches obtained from the distributed ledger and off-chain storage and verified for integrity, a deep time-series quality prediction model integrating multiple processes is constructed.

Images