A blockchain-based dynamic optimization method for production quality of nickel-based alloy

Abstract

The application discloses a kind of based on blockchain nickel-base alloy production quality dynamic optimization method, the method constructs by raw material supply end, smelting end, rolling end, heat treatment end, detection end and manufacturing management end Composition alliance chain network, relevant data in production process are collected, after time synchronization, abnormality is excluded, standardization is packaged to carry out encryption and chain and off-chain association storage;According to data integrity, consistency, timeliness, deviation control ability and historical good product contribution degree establish node credit score model, and score is converted into modeling credit weight;Further construct nickel-base alloy production quality digital twin model, combine improved particle swarm optimization algorithm to carry out rolling optimization to key process parameters, and pass through smart contract to issue optimization result to production execution system, combine execution feedback and finished product detection result to continuously update model and credit score.The method can improve production data credibility, process adjustment real-time and product quality consistency.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and blockchain integration technology, specifically to a blockchain-based method for dynamic optimization of nickel-based alloy production quality. Background Technology

[0002] Nickel-based alloys are widely used in aero-engines, gas turbines, petrochemical equipment, and high-temperature components due to their excellent high-temperature strength, oxidation resistance, corrosion resistance, and structural stability. The final quality of nickel-based alloys is influenced by a variety of factors, including raw material composition, smelting process, rolling process, heat treatment process, equipment operating status, and testing and control level. Furthermore, there are significant temporal correlations and parameter coupling relationships between different production stages. Therefore, the reliability of data, the timeliness of process adjustments, and the accuracy of quality predictions directly affect the consistency of product performance and the yield rate of finished products.

[0003] In existing nickel-based alloy production processes, raw material supply, smelting, rolling, heat treatment, and testing typically store their respective process data separately. This results in fragmented data sources, inconsistent recording standards, and difficulties in collaboratively utilizing data across different entities. Under these circumstances, on the one hand, production data is prone to late reporting, omissions, outlier entries, and retroactive recording, leading to an incomplete quality traceability chain and insufficient data reliability. On the other hand, existing process optimizations often focus on adjustments based on local historical experience or single process parameters, making it difficult to achieve unified and reliable integration of data from multiple entities, processes, and types.

[0004] Therefore, it is necessary to provide a blockchain-based method for dynamic optimization of nickel-based alloy production quality. This method, based on the realization of credible storage and traceability management of multi-entity production data, further forms credit weights through node credit scoring and introduces these credit weights into the digital twin prediction and optimization decision-making process, thereby improving the dynamic control capability and product quality stability of the nickel-based alloy production process. Summary of the Invention

[0005] To address the aforementioned issues, this invention aims to propose a blockchain-based dynamic optimization method for nickel-based alloy production quality. This method addresses problems in existing nickel-based alloy production processes, such as insufficient reliability of multi-entity production data, susceptibility to interference from low-quality data in quality modeling, and difficulty in full-chain traceability of the production process.

[0006] To achieve the above-mentioned technical objectives, the technical solution adopted by this invention is as follows: In a first aspect, the present invention proposes a blockchain-based method for dynamic optimization of the production quality of nickel-based alloys, comprising the following steps: S1. Construct a consortium blockchain network jointly accessed by the raw material supply end, smelting end, rolling end, heat treatment end, testing end, and manufacturing management end; S2. Collect raw material composition data, smelting process data, rolling process data, heat treatment process data, equipment status data, and quality inspection data during the nickel-based alloy production process. Perform time synchronization, anomaly removal, and standardized packaging on the collected data to form a production event data package. S3. Perform hash digest generation, digital signature and permission encryption processing on the production event data packet, write the digest value, off-chain storage index, timestamp, node identifier and signature result into the blockchain, store the corresponding complete data in the off-chain database, and establish an on-chain and off-chain correlation index. S4. Based on the completeness, consistency, timeliness, deviation control capability, and historical contribution of good products of the data submitted by each participating node, construct a node credit scoring model, calculate the node credit score of each participating node, and normalize the node credit score into a credit weight. S5. Based on historical production data, real-time production data and credit weights linked on and off the chain, construct a digital twin model of nickel-based alloy production quality to predict the microstructure performance indicators and quality risk indicators of the current batch. S6. With the optimization objectives of minimizing the deviation of organizational performance indicators, minimizing quality risks, limiting energy consumption, and controlling process fluctuations, an improved particle swarm optimization algorithm is used to perform rolling optimization of melting temperature, holding time, rolling pressure, rolling speed, solution temperature, solution time, aging temperature, and aging time to generate a set of target process parameters. S7. The target process parameter set is validated by smart contracts for parameter boundaries, node permissions, and batch consistency. After the validation is passed, the target process parameter set is sent to the production execution system for execution. The actual execution parameters, execution deviations, alarm records, and finished product inspection results are fed back to the consortium blockchain to update node credit scores, correct digital twin models, and trigger the next round of dynamic optimization.

[0007] In some embodiments, the raw material composition data in step S2 includes at least the mass percentages of Ni, Cr, Co, Mo, Al, Ti, Nb, W, C, and B; The smelting process data includes at least the charging batch, smelting temperature, smelting time, vacuum degree, refining time, and casting temperature; The rolling process data includes at least the initial rolling temperature, final rolling temperature, pass reduction, rolling pressure, and rolling speed. The heat treatment process data includes at least homogenization temperature, homogenization time, solution temperature, solution time, aging temperature, and aging time. The equipment status data includes at least the furnace temperature control deviation, vacuum system stability, mill vibration amplitude, motor load rate, and actuator response delay. The quality inspection data includes at least grain size, tensile strength, yield strength, elongation, hardness, inclusion grade, and compositional segregation index.

[0008] In some embodiments, when standardizing and encapsulating the collected data in step S2, the production event data packet is generated according to a fixed field structure of batch identifier, process identifier, equipment identifier, timestamp, parameter name, parameter value, parameter unit, collection source, and quality status code; The quality status code is determined by the missing data detection result, the range legality detection result, and the logical consistency detection result, and is used to characterize whether the corresponding data meets the admission conditions for on-chain evidence storage and model calculation.

[0009] In some embodiments, step S3 involves processing the i-th production event data packet. Calculate hash digest The node that submitted the production event data packet then digitally signed the hash digest using its own private key to obtain the signature result. ; The hash digest, off-chain storage index, timestamp, node identifier, and signature result are written into the blockchain as on-chain records. The plaintext and ciphertext data corresponding to the production event data packet are stored in the off-chain database. During data verification, the off-chain data is read and the hash digest is recalculated, and then compared with the hash digest in the on-chain record to verify the integrity and tamper-proof nature of the corresponding production data.

[0010] In some embodiments, the node credit scoring model in step S4 includes at least data integrity score, consistency score, timeliness score, deviation control score, and historical good product contribution score. Among them, the data integrity score is determined based on the submission completion rate of the preset mandatory fields; the consistency score is determined based on the matching degree of corresponding parameters across nodes in the same batch; the timeliness score is determined based on the time difference between the data submission time and the process completion time; the deviation control score is determined based on the degree of deviation between the data reported by the node and the verified data; and the historical good product contribution score is determined based on the historical pass rate and performance stability of the batches in which the node participates. The original credit score is obtained by weighting and summing the scores of each item according to the preset weights. The score is then adjusted by penalties based on the number of abnormal reports, the number of late reports, and the quality accountability records to obtain the final node credit score. The final node credit scores of each participating node are then normalized to obtain the credit weight.

[0011] In some embodiments, the digital twin model in step S5 includes a mechanism sub-model and a data-driven sub-model; The mechanistic sub-model is used to characterize the influence of smelting process parameters, rolling process parameters, and heat treatment process parameters on the evolution of the microstructure and properties of nickel-based alloys; the data-driven sub-model is used to characterize the nonlinear mapping relationship between raw material composition data, process parameter data, equipment status data, quality inspection data, and quality results. The output results of the mechanism sub-model and the data-driven sub-model are fused to obtain the predicted results of the organizational performance indicators and quality risk indicators for the current batch. When training the data-driven sub-model, the credit weights corresponding to the sample source nodes are introduced into the loss function to reduce the impact of low-credit data on model training and prediction results.

[0012] In some embodiments, the improved particle swarm optimization algorithm in step S6 adaptively adjusts the inertia weight according to the quality risk level of the current batch, corrects the data credibility coefficient in the fitness evaluation process according to the node credit weight, and performs constraint repair on candidate solutions that exceed the equipment capability boundary, process safety boundary and product standard boundary during particle iteration; when the change of the global optimum value in several consecutive iterations is less than a preset threshold, a local perturbation strategy is triggered to avoid the optimization process from getting trapped in a local optimum.

[0013] In some embodiments, the optimization objective function in step S6 consists of an organizational performance deviation term, a quality risk term, an energy consumption term, and a process fluctuation term. Among them, the microstructure performance deviation term is used to characterize the deviation between the predicted microstructure performance index and the target microstructure performance index; the quality risk term is used to characterize the risk of cracking, component segregation, and microstructure inhomogeneity; the energy consumption term is used to characterize the energy consumption per unit output in the smelting, rolling, and heat treatment processes under the current process parameters; and the process fluctuation term is used to characterize the change range of the target process parameter set between adjacent rolling optimization cycles.

[0014] In some embodiments, the smart contract in step S7 includes at least parameter boundary verification rules, node permission verification rules, batch consistency verification rules, and execution write-back rules. When the target process parameter set fails any of the checks in the parameter boundary verification rule, node permission verification rule, or batch consistency verification rule, the target process parameter set is prevented from being sent to the production execution system, and the corresponding blocking event is written to the blockchain. After the production execution system completes its execution, the actual execution parameters, execution deviations, alarm records, and finished product inspection results are written into the on-chain index and the off-chain data is updated synchronously. This is used to update node credit scores, correct digital twin models, and trigger the next round of dynamic optimization.

[0015] In a second aspect, the present invention provides a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the blockchain-based dynamic optimization method for nickel-based alloy production quality described in the first aspect.

[0016] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: 1. This invention connects data from the raw material supply, smelting, rolling, heat treatment, testing, and manufacturing management stages of nickel-based alloy production to a consortium blockchain network and uses hash digests, digital signatures, and on-chain / off-chain correlation indexes for trusted storage, thereby achieving verifiable, tamper-proof, and traceable management of production data and improving the credibility of quality traceability throughout the entire nickel-based alloy production process.

[0017] 2. This invention constructs a node credit scoring mechanism to quantify the completeness, consistency, timeliness, deviation control capability, and historical contribution of data from different nodes, and further normalizes them to form credit weights. This enables the credibility of data from different sources to participate in subsequent quality modeling and optimization calculations, thereby reducing the interference of low-quality data on prediction and optimization results.

[0018] 3. By constructing a digital twin model that includes a mechanism sub-model and a data-driven sub-model, and combining it with credit weights for quality prediction, this invention can more accurately reflect the microstructure and performance status and quality risk status of the current batch of nickel-based alloys, thereby improving the accuracy and robustness of quality prediction.

[0019] 4. This invention improves the particle swarm optimization algorithm to perform rolling optimization of key process parameters, and combines smart contracts to complete the verification and distribution of the optimized parameter set, execution write-back, and model correction, forming a closed-loop control mechanism of "trusted data acquisition - quality prediction - process optimization - execution feedback - dynamic update", which is conducive to improving the performance consistency of nickel-based alloy products, the finished product qualification rate, and the adaptability of the production process. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0021] Figure 1 The flowchart shows the blockchain-based dynamic optimization method for nickel-based alloy production quality according to the present invention. Figure 2 This is a schematic diagram of the production data acquisition and event data packet encapsulation process. Detailed Implementation

[0022] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention provides a blockchain-based dynamic optimization method for nickel-based alloy production quality. It is applicable to multi-entity collaborative manufacturing scenarios involving raw material supply, smelting, rolling, heat treatment, and inspection and acceptance in nickel-based alloy production. The method aims to address issues such as insufficient data reliability across multiple stages of the existing nickel-based alloy production process, complex coupling of quality-influencing factors, delayed process parameter adjustments, and inconsistent product performance. This method constructs a trusted collaborative network using a consortium blockchain to reliably collect, encrypt, and manage key data in the nickel-based alloy production process through associated indexes. Based on this, it establishes a node credit scoring model and a digital twin model, and combines an improved particle swarm optimization algorithm to continuously optimize key process parameters. Finally, the optimization results are fed back to the production execution system, achieving dynamic closed-loop optimization of nickel-based alloy production quality.

[0024] See attached document Figure 1 As shown, this embodiment provides a blockchain-based method for dynamic optimization of nickel-based alloy production quality, including the following steps: In this embodiment, a consortium blockchain network is first constructed, consisting of a raw material supply end, a smelting end, a rolling end, a heat treatment end, a testing end, and a manufacturing management end. The manufacturing management end, acting as a member management node, is used for registering the identities of consortium blockchain participants, issuing digital certificates, controlling access, and configuring on-chain rules. The raw material supply end uploads the composition testing results, batch information, and supply information of the raw materials required for nickel-based alloy production. The smelting end uploads charging parameters, smelting parameters, refining parameters, and casting parameters. The rolling end uploads parameters such as initial rolling temperature, final rolling temperature, pass reduction rate, rolling pressure, and rolling speed. The heat treatment end uploads the temperature and time parameters corresponding to homogenization treatment, solution treatment, and aging treatment. The testing end uploads the microstructure, mechanical properties, and defect detection results. The manufacturing management end also receives data reported by each node and schedules the execution of optimized results. Each node, after authorization, deploys a blockchain access module, a data signature module, a data encryption module, and an off-chain database interface module, thereby forming a consortium blockchain network environment that can be used collaboratively by multiple entities.

[0025] To ensure that multi-source data in the nickel-based alloy production process can be processed uniformly, a unique batch identifier is established for each product at the start of each production batch, and process identifiers and equipment identifiers are established according to the process steps. Subsequently, corresponding data are collected from different production stages, including raw material composition data, smelting process data, rolling process data, heat treatment process data, equipment status data, and quality inspection data.

[0026] The raw material composition data should include at least the mass percentages of Ni, Cr, Co, Mo, Al, Ti, Nb, W, C, and B; the smelting process data should include at least the charging batch, smelting temperature, smelting time, vacuum degree, refining duration, and casting temperature; the rolling process data should include at least the initial rolling temperature, final rolling temperature, reduction rate for each pass, rolling pressure, and rolling speed; the heat treatment process data should include at least the homogenization temperature, homogenization time, solution temperature, solution time, aging temperature, and aging time; the equipment status data should include at least the furnace temperature control deviation, vacuum system stability, mill vibration amplitude, motor load rate, and actuator response delay; and the quality inspection data should include at least the grain size, tensile strength, yield strength, elongation, hardness, inclusion grade, and compositional segregation index.

[0027] See attached document Figure 2 As shown, considering the differences in time base, unit format, and quality status of multi-source heterogeneous data, the collected raw data needs to undergo time synchronization, anomaly removal, and standardized encapsulation processing. Time synchronization refers to calibrating the sampling time of different nodes according to a unified clock source, enabling data from different processes to be serialized based on the same time base. Anomaly removal involves eliminating obviously erroneous data based on parameter validity range, sensor status, and logical relationships, such as temperature parameters exceeding equipment capacity limits, rolling pressure values ​​being negative, and detection results inconsistent with the process sequence. Standardized encapsulation refers to generating production event data packets according to a fixed field structure, including batch identifier, process identifier, equipment identifier, timestamp, parameter name, parameter value, parameter unit, data collection source, and quality status code. For ease of formal description, a single production event data packet is defined as: in, Indicates batch identifier, Indicates process identification, Indicates equipment identification. Represents a timestamp. Indicates the parameter name. Indicates the parameter value. Indicates the unit of the parameter. The source of the data is indicated by 'Q', and the quality status code is determined by the results of missing data detection, range validity detection, and logical consistency detection. Specifically, if key fields are missing, parameter values ​​exceed the equipment's allowable range, or the current data is inconsistent with the logic of the preceding and following processes, the corresponding quality status code will be marked as not meeting the admission criteria. Only when the quality status code meets the preset admission criteria will the corresponding production event data packet enter the on-chain evidence storage and model calculation process.

[0028] After standardized encapsulation, to ensure data immutability and traceability, digest generation and digital signature processing are performed on each production event data packet. Specifically, for the first... Production event data packets Calculate the hash digest: ;in, This is the hash digest value of the corresponding data packet. Subsequently, the node submitting the data uses its own private key to digitally sign the hash digest value, obtaining the signature result. .

[0029] To avoid overburdening the blockchain with large volumes of raw process data directly uploaded to the blockchain, this embodiment employs a combination of on-chain and off-chain storage. Specifically, the complete plaintext and ciphertext data corresponding to the production event data packet are stored in an off-chain database, while the digest value, off-chain storage index, timestamp, node identifier, and signature result are written to the blockchain as on-chain records. Correspondingly, the i-th on-chain record can be defined as: ; in, This represents the associated index in the off-chain database. Indicates the recording timestamp, Indicates the submitting node identifier. This indicates the result of the digital signature.

[0030] Subsequently, when any production data needs to be verified or traced, it is only necessary to read the original off-chain data, recalculate the hash digest, and compare it with the digest value in the on-chain record to verify whether the data has been tampered with during storage and transmission. If the two match, it is determined that the production event data packet remains intact and has not been tampered with; if the two do not match, it is determined that the data has an abnormal change risk. Therefore, this solution does not simply distribute and store production data, but rather achieves trusted storage, rapid verification, and full-process traceability of original data through an on-chain and off-chain correlated index mechanism.

[0031] To further differentiate the credibility of data submitted by different nodes and avoid the adverse effects of low-quality data on quality prediction and process optimization, this embodiment introduces a node credit scoring model. This model quantitatively evaluates each node from five dimensions: data integrity, consistency, timeliness, deviation control capability, and historical contribution of good products, calculating the node credit score.

[0032] The data integrity score is determined based on the completeness rate of submission of preset mandatory fields; the consistency score is determined based on the matching degree of corresponding parameters across nodes in the same batch; the timeliness score is determined based on the time difference between data submission time and process completion time; the deviation control score is determined based on the degree of deviation between node-reported data and verification data; and the historical good product contribution score is determined based on the historical pass rate and performance stability of the batches involved by that node. Details are as follows: For any node First, calculate its data integrity score. : in, Represents a node The number of required fields actually submitted completely within the predetermined period. Represents a node Total number of required fields to be submitted. This score reflects the completeness of the data reported by the node.

[0033] Then calculate the data consistency score. : in, This indicates the number of parameters involved in the consistency comparison. Represents a node The reported number Parameter values, Indicates the first The allowable deviation range of each parameter, This represents the reference value for this parameter within the same batch. The reference value for this parameter within the same batch is obtained by weighting the corresponding parameter data from high-credit nodes within the same batch, after consistency screening and outlier removal, according to the credit weight of each data source node. This data consistency score is used to characterize the degree of matching between the node data and the reference data within the same batch.

[0034] Next, the timeliness score is calculated. : in, Represents a node The delay between data submission time and process completion time. This represents the timeliness decay coefficient, typically ranging from 0.1 to 5. This score reflects the timeliness of node data reporting; the more timely the data submission, the more beneficial it is for subsequent rolling optimization, resulting in a higher score.

[0035] Then calculate the deviation control score. : in, This indicates the number of parameters involved in the review process. The number of parameters involved in the review process is determined based on the key process parameters and key testing parameters preset for the current process. Represents a node The reported number Parameter values, This indicates the reference review value provided during the review process. This represents the upper limit of the allowable verification deviation corresponding to the r-th parameter. In this embodiment, the upper limit of the allowable verification deviation is determined based on the measurement accuracy, equipment control accuracy, process fluctuation range, and product quality standard tolerance of the corresponding parameter. This calculated deviation control score is used to reflect the degree of deviation between the node-reported data and the authoritative verification data.

[0036] Further calculate the historical good product contribution score : in, Represents a node The historical pass rate of the batches involved. This represents the performance stability index of the corresponding batch of products. In this embodiment, the performance stability index of the corresponding batch of products is determined based on the degree of dispersion of the key performance test values ​​of each finished product in the batch participating in this node relative to the target performance range. The smaller the degree of dispersion, the higher the corresponding product performance stability index. and This is the contribution coefficient. This score reflects the degree to which this node contributes to the stable output of high-quality products throughout the historical production process.

[0037] After obtaining the above five basic scores, a weighted sum is calculated according to preset weights to obtain the original credit score: in, This represents the weight coefficient corresponding to each basic score, and the sum of all weights is 1. Considering that some nodes may have negative records such as abnormal reporting, late reporting, or quality accountability, a penalty item is also set. This is used to correct the original credit score, resulting in the final node credit score: Specifically, penalty items The results are determined by weighting and accumulating the number of abnormal reports, late reports, and quality accountability records of nodes within a preset statistical period, according to the preset deduction coefficients corresponding to various types of dishonest behaviors.

[0038] To ensure that the credit scores of each node can directly participate in subsequent modeling, the final credit scores of all nodes are normalized to obtain credit weights: in, This represents the total number of nodes in the consortium blockchain. Represents a node Credit weights are used in subsequent modeling and optimization. Through the above processing, the present invention transforms the credibility of data from a qualitative concept into a quantifiable, calculable weight factor that can participate in model training, thereby enabling the production optimization process to explicitly suppress interference caused by low-quality data.

[0039] After obtaining historical and real-time production data, along with corresponding credit weights, from both on-chain and off-chain sources, a digital twin model of nickel-based alloy production quality is constructed. This digital twin model consists of a mechanistic sub-model and a data-driven sub-model. The mechanistic sub-model characterizes the influence of smelting, rolling, and heat treatment process parameters on the evolution of the microstructure and properties of nickel-based alloys. The data-driven sub-model characterizes the nonlinear mapping relationship between raw material composition data, process parameter data, equipment status data, quality inspection data, and quality results. By integrating these two models, both mechanistic explanation and real-time prediction capabilities can be balanced.

[0040] Specifically, the input to the mechanistic sub-model is key process parameters, and the output is a microstructure and performance prediction vector, which can be represented as: in, This represents the tissue performance prediction vector output by the mechanistic sub-model. Indicates the melting temperature. Indicates the smelting and holding time. Indicates rolling pressure, Indicates rolling speed. Indicates the solution treatment temperature. Indicates the solution treatment time. Indicates the aging temperature. Indicates the time limit, This represents the set of parameters for the mechanistic model. This model is used to characterize the influence of variations in process parameters on microstructure properties such as grain size, strength, and hardness.

[0041] The data-driven sub-model uses multi-source production feature vectors. As input, output the quality prediction result corresponding to this sample: in, Indicates the first Data from a single sample drives the prediction results. Representation learning model, This represents the model parameters. Input feature vector. It consists of raw material composition data, process parameter data, equipment status data, and process quality inspection data.

[0042] Finally, the digital twin model fuses the outputs of the mechanistic sub-model and the data-driven sub-model to obtain the predicted results of the tissue performance indicators and quality risk indicators for this batch, as shown in the following expressions: in, Indicates the first The digital twin model prediction results for each sample Represents the fusion coefficient, preferably The fused prediction results can be used to predict the microstructure and performance indicators of the current batch of products, as well as to infer quality risk indicators under the current process conditions, such as crack risk, component segregation risk, and microstructure inhomogeneity risk.

[0043] To ensure that the model training process reflects differences in data credibility, node credit weights are incorporated into the loss function during data-driven sub-model training. The weighted loss function can be expressed as: Where L represents the weighted loss function, and N represents the total number of training samples. Indicates the first The credit weight of each sample source node, This indicates the model's prediction results. This represents the corresponding true detection value. Through this weighting mechanism, data from high-credit nodes have a higher weight in model training, while the impact of data from low-credit nodes on model parameter updates is appropriately weakened, thereby improving the robustness of the prediction model.

[0044] After completing the construction of the digital twin model and obtaining the predicted results of the microstructure performance index and the quality risk index, with the optimization objectives of minimizing the deviation of the microstructure performance index, minimizing the quality risk, limiting energy consumption, and controlling process fluctuations, the improved particle swarm optimization algorithm is further used to perform rolling optimization of melting temperature, holding time, rolling pressure, rolling speed, solution temperature, solution time, aging temperature, and aging time to generate a set of target process parameters.

[0045] Specifically, the process variables to be optimized are constructed into an optimization variable vector: Among them, variables These represent, in order, the melting temperature, melting holding time, rolling pressure, rolling speed, solution temperature, solution time, aging temperature, and aging time. To balance product performance, quality risks, and production costs, a comprehensive optimization objective function is constructed: in, This represents the organizational performance deviation term, used to characterize the deviation between the predicted performance index and the target performance index; This indicates a quality risk item, used to characterize quality risks such as cracks, segregation, and uneven microstructure. This represents the energy consumption item, used to characterize the energy consumption per unit output of smelting, rolling, and heat treatment processes under current process parameters; This represents the process fluctuation term, used to characterize the magnitude of change in the target process parameter set between adjacent rolling optimization cycles; These are the weighting coefficients for each objective term. The objective function is minimized. This can reduce quality risks and production fluctuations while ensuring that performance indicators meet the standards.

[0046] In particle swarm optimization, the velocity and position of the k-th particle in the t-th iteration are updated as follows: in, This represents the position of the k-th particle in the t-th iteration. Represents particle velocity. This represents the optimal position of an individual particle. This represents the globally optimal position of the group. and As a learning factor, and As a random factor, For the first The inertia weight of the next iteration.

[0047] To enhance the adaptability of the algorithm in the dynamic optimization of nickel-based alloy quality, this embodiment makes the following improvements to the particle swarm optimization algorithm.

[0048] The inertia weight is adaptively adjusted based on the quality risk level of the current batch, and the expression is as follows: in, and These represent the minimum and maximum values ​​of the inertia weight, respectively. The adjustment coefficient is preferably set within the range of 0.1 to 3. This indicates the quality risk level of the current batch. In this embodiment, the quality risk level is determined based on the weighted comprehensive result of crack risk, component segregation risk, and microstructure inhomogeneity risk output by the digital twin model, combined with a preset grading threshold. The higher the risk level, the more the inertial weights tend to maintain a strong search capability for the particle swarm, in order to find a better combination of process parameters as quickly as possible.

[0049] Simultaneously, the data credibility coefficient in the fitness evaluation process is adjusted according to the credit weight, giving higher weight to predictions corresponding to high-credit data in the fitness evaluation. During particle iteration, constraint repair is performed on candidate solutions that exceed equipment capability boundaries, process safety boundaries, and product standard boundaries to ensure the executability of the optimization results. When the change in the global optimum value over several consecutive iterations is less than a preset threshold, a local perturbation strategy is triggered to avoid the optimization process getting stuck in local optima. Specifically, the local perturbation strategy refers to adding a preset magnitude of random perturbation to one or more dimensions of the current global optimum particle or some high-fitness particles when the change in the global optimum value over several consecutive iterations is less than a preset threshold, and then re-performing boundary constraint repair before continuing the iterative search to enhance the local neighborhood exploration capability.

[0050] Through the above optimization process, the target process parameter set within the current rolling cycle can be obtained.

[0051] After generating the target process parameter set, it is not directly distributed, but first automatically verified by a smart contract. The smart contract includes at least parameter boundary verification rules, node permission verification rules, batch consistency verification rules, and execution write-back rules.

[0052] Parameter boundary verification rules are used to determine whether the target process parameters fall within the executable range of the equipment and the process safety range; node permission verification rules are used to determine whether the node that initiated the distribution has the corresponding control permissions; batch consistency verification rules are used to determine whether the batch number and process number corresponding to the optimization result are consistent with the current execution object; execution write-back rules are used to stipulate that the actual execution parameters, execution deviations, alarm records and finished product test results after the execution is completed must be written back to the on-chain index and the off-chain database.

[0053] When the target process parameter set fails any of the checks, such as parameter boundary verification rules, node permission verification rules, or batch consistency verification rules, the smart contract prevents the target process parameter set from being issued and writes the corresponding blocking event to the blockchain to ensure the auditability of the control process.

[0054] Once the target process parameter set passes the smart contract verification, it is sent to the production execution system. The smelting control unit, rolling control unit, and heat treatment control unit then execute the corresponding processes according to the target parameters. During execution, actual execution parameters are collected in real time. and execution deviation ; The execution deviation satisfies: ;in, Indicates the actual set of execution parameters. This represents the target process parameter set. If the execution deviation exceeds a preset threshold, the execution system will trigger an alarm record and write the alarm information to both the on-chain index and the off-chain database.

[0055] After execution, the testing end performs organization and performance testing on the finished product and feeds the test results back to the consortium blockchain. Based on the actual execution parameters, execution deviations, alarm records, and finished product test results fed back to the consortium blockchain, the manufacturing management end updates the credit scores of each relevant node and uses new samples to correct the parameters of the digital twin model in order to trigger the next round of rolling optimization.

[0056] It should be noted that the consortium blockchain deployment method, scoring weight setting method, specific modeling form of digital twin model, and particle swarm optimization parameter settings adopted in the above embodiments can all be adaptively adjusted according to different nickel-based alloy grades, different production equipment conditions, and different performance control objectives. However, their core technical ideas are: to ensure the credibility of multi-entity production data through blockchain, to quantify data credibility and participate in modeling through node credit scoring, to achieve quality prediction through digital twins, to achieve rolling optimization of key process parameters through improved optimization algorithms, and to achieve continuous updated closed-loop control through execution feedback.

[0057] Furthermore, the functional units in the embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A blockchain-based method for dynamic optimization of nickel-based alloy production quality, characterized in that, Includes the following steps: S1. Construct a consortium blockchain network jointly accessed by the raw material supply end, smelting end, rolling end, heat treatment end, testing end, and manufacturing management end; S2. Collect raw material composition data, smelting process data, rolling process data, heat treatment process data, equipment status data, and quality inspection data during the nickel-based alloy production process. Perform time synchronization, anomaly removal, and standardized packaging on the collected data to form a production event data package. S3. Perform hash digest generation, digital signature and permission encryption processing on the production event data packet, write the digest value, off-chain storage index, timestamp, node identifier and signature result into the blockchain, store the corresponding complete data in the off-chain database, and establish an on-chain and off-chain correlation index. S4. Based on the completeness, consistency, timeliness, deviation control capability, and historical contribution of good products of the data submitted by each participating node, construct a node credit scoring model, calculate the node credit score of each participating node, and normalize the node credit score into a credit weight. S5. Based on historical production data, real-time production data and credit weights linked on and off the chain, construct a digital twin model of nickel-based alloy production quality to predict the microstructure performance indicators and quality risk indicators of the current batch. S6. With the optimization objectives of minimizing the deviation of organizational performance indicators, minimizing quality risks, limiting energy consumption, and controlling process fluctuations, an improved particle swarm optimization algorithm is used to perform rolling optimization of melting temperature, holding time, rolling pressure, rolling speed, solution temperature, solution time, aging temperature, and aging time to generate a set of target process parameters. S7. The target process parameter set is validated by smart contracts for parameter boundaries, node permissions, and batch consistency. After the validation is passed, the target process parameter set is sent to the production execution system for execution, and the actual execution parameters, execution deviations, alarm records, and finished product test results are fed back to the consortium blockchain.

2. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S2, the raw material composition data includes at least the mass percentages of Ni, Cr, Co, Mo, Al, Ti, Nb, W, C, and B; The smelting process data includes at least the charging batch, smelting temperature, smelting time, vacuum degree, refining time, and casting temperature; The rolling process data includes at least the initial rolling temperature, final rolling temperature, pass reduction, rolling pressure, and rolling speed. The heat treatment process data includes at least homogenization temperature, homogenization time, solution temperature, solution time, aging temperature, and aging time. The equipment status data includes at least the furnace temperature control deviation, vacuum system stability, mill vibration amplitude, motor load rate, and actuator response delay. The quality inspection data includes at least grain size, tensile strength, yield strength, elongation, hardness, inclusion grade, and compositional segregation index.

3. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S2, when standardizing and encapsulating the collected data, the production event data packet is generated according to a fixed field structure of batch identifier, process identifier, equipment identifier, timestamp, parameter name, parameter value, parameter unit, collection source, and quality status code. The quality status code is determined by the missing detection result, the range legality detection result, and the logical consistency detection result, and is used to characterize whether the corresponding data meets the admission conditions for on-chain evidence storage and model calculation.

4. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S3, the i-th production event data packet is processed. Calculate hash digest The node that submitted the production event data packet then digitally signed the hash digest using its own private key to obtain the signature result. ; The hash digest, off-chain storage index, timestamp, node identifier, and signature result are written into the blockchain as on-chain records. The plaintext and ciphertext data corresponding to the production event data packet are stored in the off-chain database. During data verification, the off-chain data is read and the hash digest is recalculated, and then compared with the hash digest in the on-chain record to verify the integrity and tamper-proof nature of the corresponding production data.

5. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S4, the node credit scoring model includes at least data integrity score, consistency score, timeliness score, deviation control score, and historical good product contribution score. Among them, the data integrity score is determined based on the submission completion rate of the preset mandatory fields; the consistency score is determined based on the matching degree of corresponding parameters across nodes in the same batch; the timeliness score is determined based on the time difference between the data submission time and the process completion time; the deviation control score is determined based on the degree of deviation between the data reported by the node and the verified data; and the historical good product contribution score is determined based on the historical pass rate and performance stability of the batches in which the node participates. The original credit score is obtained by weighting and summing the scores of each item according to the preset weights. The score is then adjusted by penalties based on the number of abnormal reports, the number of late reports, and the quality accountability records to obtain the final node credit score. Finally, the final node credit scores of each participating node are normalized to obtain the credit weight.

6. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S5, the digital twin model includes a mechanism sub-model and a data-driven sub-model; The mechanistic sub-model is used to characterize the influence of melting process parameters, rolling process parameters, and heat treatment process parameters on the evolution of the microstructure and properties of nickel-based alloys; The data-driven sub-model is used to characterize the nonlinear mapping relationship between raw material composition data, process parameter data, equipment status data, quality inspection data and quality results; The output results of the mechanism sub-model and the output results of the data-driven sub-model are fused to obtain the predicted results of tissue performance indicators and quality risk indicators for the current batch. Furthermore, when training the data-driven sub-model, the credit weights corresponding to the sample source nodes are introduced into the loss function to reduce the impact of low-credit data on model training and prediction results.

7. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S6, the improved particle swarm optimization algorithm adaptively adjusts the inertia weight according to the quality risk level of the current batch, corrects the data credibility coefficient in the fitness evaluation process according to the node credit weight, and performs constraint repair on candidate solutions that exceed the equipment capability boundary, process safety boundary and product standard boundary during particle iteration; when the change of the global optimum value in several consecutive iterations is less than the preset threshold, a local perturbation strategy is triggered to avoid the optimization process from getting trapped in local optima.

8. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, The optimization objective function in step S6 consists of organizational performance deviation, quality risk, energy consumption, and process fluctuation terms. Among them, the microstructure performance deviation term is used to characterize the deviation between the predicted microstructure performance index and the target microstructure performance index; the quality risk term is used to characterize the risk of cracking, component segregation, and microstructure inhomogeneity; the energy consumption term is used to characterize the energy consumption per unit output in the smelting, rolling, and heat treatment processes under the current process parameters; and the process fluctuation term is used to characterize the change range of the target process parameter set between adjacent rolling optimization cycles.

9. The method for dynamic optimization of nickel-based alloy production quality based on blockchain according to claim 1, characterized in that, In step S7, the smart contract includes at least parameter boundary verification rules, node permission verification rules, batch consistency verification rules, and execution write-back rules; When the target process parameter set fails any of the checks in the parameter boundary verification rule, node permission verification rule, or batch consistency verification rule, the target process parameter set is prevented from being sent to the production execution system, and the corresponding blocking event is written to the blockchain. After the production execution system completes its execution, the actual execution parameters, execution deviations, alarm records, and finished product inspection results are written into the on-chain index and the off-chain data is updated synchronously. This is used to update node credit scores, correct digital twin models, and trigger the next round of dynamic optimization.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the blockchain-based dynamic optimization method for nickel-based alloy production quality as described in any one of claims 1 to 9.

Images