A blockchain-based semiconductor film diffusion quality traceability method and system

Abstract

The application relates to the technical field of blockchains, in particular to a semiconductor film diffusion quality tracing method and system based on a blockchain. The method comprises the following steps: constructing a blockchain main network and a plurality of subnetworks associated with the main network; connecting the main network and the subnetworks through a trusted binding mechanism of node economy guarantee; deploying a standardized tracing template of a semiconductor film diffusion process in the blockchain main network; collecting process parameters and quality data through a sensor network deployed in a production line, pre-processing the parameters and the data through an edge computing node, and uploading the parameters and the data to corresponding subnetworks; deploying a quality anomaly processing smart contract in the main network, associating multi-source data based on the standardized template, and the execution process of the smart contract comprising tampering risk evaluation, local congestion risk evaluation and conflict possibility evaluation; and solving the problems of insufficient data credibility and security, low abnormal tracing efficiency and lack of multi-dimensional risk evaluation in the prior art.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, and in particular to a blockchain-based method and system for tracing the diffusion quality of semiconductor films. Background Technology

[0002] Semiconductor film diffusion is a core and critical process in semiconductor device manufacturing. By precisely controlling process parameters such as diffusion furnace temperature, dopant gas flow rate, and chamber pressure, it forms a semiconductor film of specific thickness and electrical properties on the wafer surface, directly determining the semiconductor device's conductivity, breakdown voltage, and other key performance indicators. As the precision requirements for film quality in semiconductor devices continue to increase, achieving quality traceability throughout the entire semiconductor film diffusion process is of great significance for locating the root cause of anomalies, ensuring device reliability, and standardizing supply chain management.

[0003] In existing technologies, traceability of semiconductor film diffusion quality mainly relies on centralized data management solutions. These solutions store process parameters and quality inspection data through enterprise manufacturing execution systems (MES), power supply systems (PLCs), or local relational databases, and aggregate the data through manual input or simple system integration. However, such solutions have the following prominent drawbacks in practical applications:

[0004] First, data credibility and security are insufficient. Centralized architectures rely on a single data center or server cluster, making them vulnerable to hardware failures, network attacks, human error, and tampering, leading to the loss or unavailability of traceability data across the entire process. Furthermore, semiconductor film diffusion involves multiple stakeholders, including wafer fabs, equipment suppliers, and quality inspection agencies. Data from each stage is stored in independent systems, requiring cross-system integration, which distorts traceability results and renders them unreliable for liability determination. Second, data correlation is low, resulting in inefficient anomaly traceability. When quality anomalies occur, manual comparison of data logs from different systems is necessary, making it difficult to quickly establish mapping relationships and leading to lengthy traceability cycles. Third, risk assessment is lacking, and data credibility is not guaranteed. This includes risks such as data tampering (e.g., sensor data being maliciously intercepted and modified during database transmission), local congestion (e.g., multiple devices uploading data simultaneously during peak production periods causing subnet node delays and disrupting data timing), and data flow conflicts (e.g., synchronization discrepancies between the main network and subnets leading to multiple versions of the same batch of data). This leads to a lack of credibility in traceability data, and may even mislead decision-making in handling anomalies, affecting trust in the supply chain.

[0005] Therefore, there is a need for a blockchain-based method and system for tracing the diffusion quality of semiconductor films to address the problems of insufficient data credibility and security, low efficiency in anomaly tracing, and lack of multi-dimensional risk assessment in existing technologies. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a blockchain-based method and system for tracing the diffusion quality of semiconductor films.

[0007] In a first aspect, the present invention provides a semiconductor film diffusion quality traceability method based on blockchain, which adopts the following technical solution:

[0008] A blockchain-based method for tracing the diffusion quality of semiconductor films includes the following steps:

[0009] S1. Construct a blockchain mainnet and multiple subnets associated with it, and connect the mainnet and subnets through a trusted binding mechanism of node economic guarantee. The subnets include a device data subnet for storing diffusion process data and a quality inspection subnet for storing quality performance testing data.

[0010] S2. Deploy a standardized traceability template for semiconductor film diffusion process in the blockchain mainnet;

[0011] S3. Collect process parameters and quality data through the sensor network deployed on the production line, and upload them to the corresponding subnet after preprocessing by the edge computing node;

[0012] S4. Through the quality anomaly handling smart contract deployed on the main network, based on the standardized traceability template and associated with multi-source data, quality anomalies are judged and graded for handling. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment and conflict probability assessment.

[0013] S5. Implement data access and control for multi-level users based on an enhanced role-based access control model.

[0014] Furthermore, the blockchain mainnet is used to store the digital identity information of the participants, the definition of standardized traceability templates, and the rules for determining quality responsibility;

[0015] The device data subnet is used to store sensor data on diffusion furnace temperature, doping gas flow rate, and chamber pressure.

[0016] The quality inspection subnet is used to store the quality inspection results and quality anomaly reports for the semiconductor film's thickness, sheet resistance, minority carrier lifetime.

[0017] Furthermore, the trusted binding mechanism for the node economic guarantee includes:

[0018] Before the subnet is connected to the mainnet, the authorized guarantor nodes in the mainnet pledge a preset amount of digital assets for the subnet to be connected, and the detailed information of the pledged preset amount of digital assets is recorded in the distributed ledger of the mainnet through the on-chain asset staking contract.

[0019] After the contract of the trusted binding mechanism of the subnet is invoked, a binding event with a cryptographic signature is generated, triggering the consensus mechanism within each subnet. After each subnet returns a consensus success message, the mainnet performs consensus consistency verification, completing the trusted binding and data communication between the mainnet and each subnet.

[0020] Furthermore, the amount of the pre-set quantity of digital assets pledged must satisfy the following formula:

[0021]

[0022] in, This is the actual amount pledged. Based on the amount of the pledged collateral, This is the coefficient for the number of nodes. This represents the total number of nodes waiting to be connected to the subnet.

[0023] Furthermore, the criteria for the mainnet to perform consensus verification are as follows: ,in, The number of nodes required for consensus to be reached. This represents the total number of nodes in the subnet. The preset ratio threshold is used; when the condition is met... > The mainnet and subnets must be bound together; otherwise, the consensus mechanism within each subnet must be retried.

[0024] Furthermore, the standardized traceability template includes a multi-dimensional structured field set, and is judged by inter-field association verification. Data that fails the inter-field association verification will be marked as "to be completed" and data upload will be suspended until the associated field data is completed and passes the re-verification.

[0025] Furthermore, the standardized traceability template includes:

[0026] Basic wafer information fields include wafer unique identifier, wafer material type, and physical dimensions;

[0027] The equipment identification information field includes a unique identifier for the diffusion furnace, a hardware feature code of the diffusion furnace generated based on the physical characteristics of the equipment, and a recent calibration timestamp;

[0028] Process parameter fields include diffusion temperature setpoint and actual value, diffusion time, dopant gas type and flow rate, and diffusion chamber pressure.

[0029] Quality inspection fields include semiconductor film thickness measurement, sheet resistance measurement, and minority carrier lifetime measurement;

[0030] The operator information field includes the operator's digital identity and the hash value of the operator's qualification certificate;

[0031] The exception handling record fields include the exception type code, the timestamp of the exception occurrence, and a description of the handling measures taken.

[0032] Furthermore, the step of preprocessing and classifying the data before uploading it to the corresponding subnet via edge computing nodes includes:

[0033] Edge computing nodes for raw data Collection timestamp and sensor identifiers Perform concatenation and calculate the hash value of the concatenated data: ;

[0034] The corresponding subnet recalculates the hash value after receiving the data. And verify the following two conditions:

[0035] Hash consistency: ;

[0036] Data freshness: the time difference between data collection and upload time. , The time difference threshold;

[0037] Data will only be stored if all of the above conditions are met; otherwise, storage will be refused and an exception notification will be returned.

[0038] Furthermore, the quality anomaly handling smart contract is configured with a dynamic quality threshold library, which dynamically sets the upper and lower limits of the quality threshold based on the wafer material type, film layer type, and process specifications.

[0039] When the quality anomaly handling smart contract is executed, it automatically obtains the quality detection field data in the standardized traceability template, compares it with the corresponding dynamic threshold, and determines whether there is an anomaly.

[0040] If an anomaly is detected, the quality anomaly handling smart contract further correlates and analyzes the process parameters corresponding to the abnormal data points, identifies abnormal process parameters that deviate from the standard, determines the root cause of the quality anomaly based on predefined abnormal parameter mapping rules, generates a structured traceability report, and triggers an on-chain alarm event.

[0041] Furthermore, the determination and grading of quality anomalies are based on a dynamic quality threshold library, the threshold range of which is determined by the following model:

[0042] ,in, It is a wafer material type. For membrane type, Based on historical good product data statistical characteristics, For a threshold model trained on a gradient boosting tree, if the quality detection data... satisfy or If the error is not detected, it is considered an anomaly, and further correlation with process parameter fluctuations is used to distinguish the anomaly type.

[0043] Furthermore, the determination and grading of quality anomalies includes:

[0044] The contract calls the dynamic quality threshold library to obtain the value relative to the current wafer material type. Film type Matching quality threshold range And compare them;

[0045] Calculate the degree to which outlier data deviates from the threshold. And associate the fluctuation range of the corresponding process parameters in the equipment data subnet, and classify them according to the following rules:

[0046] Minor abnormalities: ,and ;

[0047] Moderate abnormality: ,or , ;

[0048] Serious abnormality: ,or ;

[0049] in, These are the actual data values ​​for quality inspection fields, such as the actual measured values ​​of semiconductor film thickness, sheet resistance, minority carrier lifetime, etc.

[0050] This measures the degree to which outlier data deviates from a threshold, used to quantify the actual value. With threshold range The magnitude of the deviation;

[0051] The fluctuation range of the diffusion temperature corresponds to the degree of deviation between the actual value and the set value of the diffusion temperature in the corresponding process.

[0052] The fluctuation range of the dopant gas flow rate corresponds to the degree of deviation between the actual and set values ​​of the dopant gas flow rate in the corresponding process.

[0053] For minor anomalies, the smart contract triggers a local re-inspection, instructing the quality inspection subnet to re-collect quality data for at least three samples in the batch. If the re-inspection data is qualified, the anomaly is marked as a misjudgment; otherwise, an anomaly alert is pushed to the production line operator.

[0054] In the case of a moderate anomaly, the smart contract triggers a subnet data freeze, suspends the uploading of new data to the data subnet of the anomaly-related device, and generates an anomaly tracing task, which is then pushed to the device supplier node.

[0055] In case of a serious anomaly, the smart contract triggers cross-subnet tracing and verification, simultaneously freezing the corresponding batch data of the device data subnet and the quality inspection subnet, calling the mainnet audit node to verify the consensus records and data flow logs of the abnormal period, generating a tracing report and recording it in the main chain block.

[0056] Furthermore, the tampering risk assessment includes:

[0057] Extract the data fragment set of abnormally associated batches from the equipment data subnet or quality inspection subnet. , where n is the total number of data fragments in this batch;

[0058] Obtain the hash checksum of each shard on different nodes in the subnet, and calculate the hash difference value of each shard. ,in , They are shards The hash value between any two nodes in the subnet;

[0059] Tampering with the risk index calculation formula: ,in For fragmentation The absolute value of the hash difference;

[0060] Preset tampering risk threshold ,like The risk of data tampering in this batch was determined to be low, and the data reliability met the traceability requirements; if If the risk of tampering is deemed high, the contract triggers a data retransmission verification, requiring subnet nodes to re-upload the original data for this batch and perform a second verification.

[0061] Furthermore, the local congestion risk assessment includes:

[0062] Extract the average verification latency of each node during the abnormal period from the node latency logs of the subnet associated with the mainnet synchronization anomaly. ,in This represents the total number of nodes in the subnet.

[0063] Calculate the overall average verification latency of the subnet: ;

[0064] Calculate the local congestion risk index: ,in For nodes The absolute difference between the delay and the average delay of the subnet;

[0065] Preset local congestion risk threshold ,like If the subnet is determined to have no local congestion, the data transmission reliability is high; If a local congestion risk is detected, the contract triggers a subnet load balancing instruction to distribute some of the data processing tasks of the abnormal node to nodes with lower load.

[0066] Furthermore, the conflict probability assessment includes:

[0067] Extract the abnormal data transfer operation records from the interaction logs between the main network and the abnormal associated subnet. There are a total of L transfer operations, and the time interval of each transfer is recorded. and consensus deviation , representing the difference between the maximum and minimum confirmation times at different nodes for the h-th transfer;

[0068] Calculate the average consensus bias across all transfer operations: ;

[0069] Calculate the conflict probability index: , where ε is a preset positive number used to prevent the denominator from being zero;

[0070] like The abnormal data was determined to have no flow conflicts, and the data credibility met the traceability requirements.

[0071] like If a risk of data transfer conflict is detected, the contract triggers a data transfer path verification. This involves comparing the actual data transfer path of the abnormal data with the preset compliant path, locating the conflict node, and recording it in the audit log. This indicates a preset threshold for the likelihood of conflict.

[0072] Furthermore, the asset pledge contract stipulates that if data tampering or forgery is discovered in the functional subnet node through the hash verification mechanism or subsequent audit, the digital assets pledged by the guarantor node will be automatically deducted according to a preset ratio ρ, and the value of the deducted assets will be transferred to the mainnet's quality assurance fund account to compensate for quality loss caused by inaccurate data.

[0073] Furthermore, the enhanced role-based access control model includes the following permission levels:

[0074] The wafer fab operator role is granted access to raw process and quality data across all subnets;

[0075] The equipment supplier role is only granted access to operational parameter data related to the equipment it provides and directly associated quality anomaly information.

[0076] Downstream customer roles are only granted access to the final quality inspection results summary and confirmed anomaly handling records;

[0077] The system records all query operations on the mainnet through a query log smart contract. The record includes the queryer's digital identity, query time, query data range, and query result hash.

[0078] Secondly, a blockchain-based semiconductor film diffusion quality traceability system includes:

[0079] The network binding module is configured to build a blockchain mainnet and multiple subnets associated with it. The mainnet and subnets are connected through a trusted binding mechanism with node economic guarantees. The subnets include a device data subnet for storing diffusion process data and a quality inspection subnet for storing quality performance testing data.

[0080] The standardized template deployment module is configured to deploy a standardized traceability template for semiconductor film diffusion processes in the blockchain mainnet;

[0081] The data processing module is configured to collect process parameters and quality data through the sensor network deployed on the production line, and then classify and upload them to the corresponding subnet after preprocessing by the edge computing nodes.

[0082] The risk assessment module is configured to use a quality anomaly handling smart contract deployed on the mainnet to determine and classify quality anomalies based on the standardized traceability template and associated multi-source data. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment and conflict probability assessment.

[0083] The access control module is configured to implement multi-level user data access and control based on an enhanced role-based access control model.

[0084] Thirdly, the present invention provides a computer-readable storage medium storing a plurality of instructions adapted to be loaded and executed by a processor of a terminal device, the aforementioned blockchain-based semiconductor film diffusion quality traceability method.

[0085] Fourthly, the present invention provides a terminal device, including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor to provide a blockchain-based semiconductor film diffusion quality traceability method.

[0086] In summary, the present invention has the following beneficial technical effects:

[0087] 1. This invention proposes a blockchain-based method and system for tracing the diffusion quality of semiconductor films, which enhances the credibility and security of data exchange between the main network and subnets. Through an implemented node economic guarantee mechanism, it requires authorized nodes from the main network to pledge pre-set digital assets before subnet access, and records the pledge information through an on-chain asset staking contract. Simultaneously, it combines cryptographic signature binding events with internal consensus verification within the subnet to ensure a trustworthy binding between the main network and the device data subnet and quality inspection subnet. This approach not only forms economic constraints through digital asset staking to prevent data tampering or forgery by subnet nodes, but also ensures the stability of data exchange between the main network and subnets through consensus consistency verification. It effectively avoids the risks of "single point of failure" and data tampering under a centralized architecture, providing a decentralized, secure storage and trusted transmission environment for traceability data across all stages.

[0088] 2. This invention deploys a standardized traceability template for semiconductor film diffusion processes on the main network, establishing standardized association logic for data from multiple participants and stages. Furthermore, the field association verification mechanism can mark data that does not meet the association requirements as "to be completed" and suspend uploading, avoiding the "information silo" problem. Simultaneously, the template fields are associated with the corresponding subnet block identifiers, enabling rapid location of data storage and significantly improving the efficiency of associating process parameter quality data with the responsible stage when quality anomalies occur. This addresses the shortcomings of existing technologies, such as scattered data and long traceability cycles.

[0089] 3. The dynamic quality threshold library of this invention dynamically generates quality threshold ranges through a threshold model trained by gradient boosting trees, ensuring that quality anomaly judgment fits different process scenarios. At the same time, the three-level handling strategy of minor, moderate and severe based on the degree of anomaly deviation and the fluctuation range of process parameters can trigger differentiated operations such as "local re-inspection", "subnet data freezing" and "cross-subnet traceability verification" for different anomaly levels. This avoids minor anomalies occupying too many production resources and can quickly curb the spread of severe anomalies, reducing quality losses.

[0090] 4. This invention employs a multi-dimensional risk assessment mechanism to evaluate the risks of data tampering, local congestion, and flow conflicts. The three indices, combined with preset threshold judgment logic, can comprehensively identify the credibility risks of data during storage, transmission, and flow, and trigger countermeasures such as data retransmission verification, subnet load balancing, and flow path verification to ensure the credibility of traceability data throughout the entire process from collection to application, providing a basis for quality responsibility determination. Attached Figure Description

[0091] Figure 1 This is a flowchart of the semiconductor film diffusion quality traceability method based on blockchain according to Embodiment 1 of the present invention;

[0092] Figure 2 This is a schematic diagram of the blockchain mainnet and its associated subnets according to Embodiment 1 of the present invention;

[0093] Figure 3 This is a schematic diagram of the smart contract processing for quality anomaly handling in Embodiment 1 of the present invention;

[0094] Figure 4 This is a schematic diagram of a blockchain-based semiconductor film diffusion quality traceability system according to Embodiment 2 of the present invention; Detailed Implementation

[0095] The present invention will be further described in detail below with reference to the accompanying drawings.

[0096] Example 1

[0097] Reference Figure 1 and Figure 2 This embodiment of a blockchain-based semiconductor film diffusion quality traceability method includes the following steps:

[0098] S1. Construct a blockchain mainnet and multiple subnets associated with it, and connect the mainnet and subnets through a trusted binding mechanism of node economic guarantee. The subnets include a device data subnet for storing diffusion process data and a quality inspection subnet for storing quality performance testing data.

[0099] Mainnet Node Configuration: The blockchain mainnet includes three types of core nodes: consensus nodes (no fewer than 5, using the Practical Byzantine Fault Tolerance (PBFT) consensus algorithm), audit nodes (no fewer than 3, responsible for subsequent verification of abnormal data and accountability, requiring access to the mainnet quality assurance fund account), and identity management nodes (1, solely responsible for registering and updating the digital identities of participants, using the national cryptographic SM2 algorithm to generate node identity key pairs). The mainnet nodes adopt a "master-slave synchronization + timed verification" mechanism, synchronizing block data every 10 minutes and verifying the consistency of the entire ledger through hash comparison every hour.

[0100] The equipment data subnet nodes are deployed in shards according to diffusion furnace numbers, with each shard corresponding to one diffusion furnace. Data is stored in shards according to the rule of "diffusion furnace number-timestamp", with the shard size set to 1GB. New shards are automatically generated when the threshold is exceeded. Nodes are connected to temperature sensors, gas flow controllers, and chamber pressure sensors. The quality inspection subnet nodes are deployed in shards according to wafer batches, with each shard corresponding to one production batch (containing 500 to 1000 wafers). Nodes need to be connected to the standard data interfaces of thickness measuring instruments, four-probe testers, and minority carrier lifetime testers, and support real-time reception of structured data (JSON format) output by quality inspection equipment.

[0101] The mainnet and subnets interact through a "digest upload + raw data localization" model. Every 5 minutes, the subnet uploads a data digest to the mainnet, including the data hash value, data volume, and collection period. The mainnet stores the digest and creates an index. When the mainnet needs to access the raw data from the subnet, it initiates an authorization request through a smart contract. After the subnet verifies the validity of the mainnet's request signature, it transmits the raw data fragments through an AES-256 encrypted channel. The temporary transmission file is destroyed immediately after the transmission is completed.

[0102] The blockchain mainnet is used to store the digital identity information of the participants, the definition of standardized traceability templates, and the rules for determining quality responsibility; the equipment data subnet is used to store sensor data on diffusion furnace temperature, doping gas flow rate, and chamber pressure; and the quality inspection subnet is used to store quality inspection results and quality anomaly reports for semiconductor film thickness, sheet resistance, minority carrier lifetime, etc.

[0103] It includes four types of participants: wafer foundries (identity identifier prefix "WF-"), equipment suppliers (prefix "EQ-"), downstream customers (prefix "CU-"), and third-party quality inspection agencies (prefix "QT-"). The digital identity information of each type of participant includes a unique DID, public key, role and permission level, and qualification certificate hash. All information is written to the main network after being verified by the identity management node, and modifications require confirmation by ≥3 consensus nodes.

[0104] Standardized traceability templates are defined and stored in the mainnet contract account in JSON format. The template includes field identifiers, data types, required attributes, validation rules, and associated subnet identifiers. For example, the "actual value of diffusion temperature" field is identified as "T-actual", the data type is "floating-point (with 1 decimal place)", the required attribute is "yes", the validation rule is "must be within ±5℃ of the set value", and the associated subnet identifier is "device data subnet - shard 03". Modifications to the template require a mainnet proposal and must be approved by ≥2 / 3 of the consensus nodes before taking effect. The modification record is synchronously written to the mainnet block and a version number is generated.

[0105] Quality responsibility determination rules: Includes an algorithm for attributing responsibility for anomalies, based on a three-dimensional determination of "abnormal process parameter period - responsible node operation record - data tampering risk index". The formula is: Responsibility weight = 0.4 × parameter anomaly contribution + 0.3 × operation record matching degree + 0.3 × (1-TRI). Compensation ratio standards: 10%~20% of the loss is compensated by the responsible party for minor anomalies, 30%~50% for moderate anomalies, and 60%~100% for serious anomalies. Disclaimer clauses: For anomalies caused by force majeure, such as power outages caused by earthquakes, a force majeure certificate issued by a third-party institution must be provided. The rules are deployed on the main network in the form of smart contract code, and automatically call historical data from the relevant subnets when executed.

[0106] The trusted binding mechanism for node economic guarantees includes: before the subnet is connected to the mainnet, the authorized guarantee nodes in the mainnet pledge a preset amount of digital assets for the subnet to be connected, and the detailed information of the pledged preset amount of digital assets is recorded in the distributed ledger of the mainnet through the on-chain asset staking contract.

[0107] After the contract of the trusted binding mechanism of the subnet is invoked, a binding event with a cryptographic signature is generated, triggering the consensus mechanism within each subnet. After each subnet returns a consensus success message, the mainnet performs consensus consistency verification, completing the trusted binding and data communication between the mainnet and each subnet.

[0108] The amount of digital assets pledged in the preset quantity must meet the following formula: ,

[0109] in, This is the actual amount pledged. Based on the amount of the pledged collateral, This is the coefficient for the number of nodes. This represents the total number of nodes waiting to be connected to the subnet.

[0110] The authorization conditions for guarantor nodes must meet three qualification requirements: First, they must have completed identity registration on the mainnet for at least 90 days and have no history of data tampering, delayed uploads, or other violations; second, they must have staked ≥1000 units of the mainnet's native token ("Semiconductor Traceability Coin") as underlying credit assets; and third, the node's online rate must be ≥95% (statistics for the past 30 days), and the data processing latency must be ≤500ms. The authorization process is as follows: guarantor node submits application → mainnet audit node verifies qualifications → ≥3 audit nodes sign off → "Guarantee Node Authorization Certificate" is generated and written to the mainnet.

[0111] Digital asset staking process: The guarantor node initiates a staking request, which includes the identifier of the subnet to be guaranteed, the total number of subnet nodes N, and the proposed staking amount A; the on-chain asset staking contract calls the identity management node interface to verify the authorization qualification of the guarantor node; the contract verifies whether A meets the requirements. If the conditions are met, A units of tokens in the guarantor node account will be locked, and a staking certificate (including staking ID, guarantor subnet ID, staking period (default 180 days, which can be extended to 360 days), and unlocking conditions) will be generated. The contract will write the staking information (staking ID, guarantor node DID, subnet ID, A, N, k, staking timestamp) into the mainnet distributed ledger and synchronize it to all mainnet nodes. The staking certificate will be issued to the guarantor node account in the form of NFTs and can be used for credit endorsement within the mainnet.

[0112] In addition to the amount, term, and guarantee relationship stated in the original text, the pledge information record should also include "the lock-up period of the pledged assets, the rules for deductions due to violations, and the returns on the pledged assets".

[0113] The conditions for the mainnet to perform consensus consistency verification are as follows: ,in, The number of nodes required for consensus to be reached. This represents the total number of nodes in the subnet. The preset ratio threshold is used; when the condition is met... > The mainnet and subnets must be bound together; otherwise, the consensus mechanism within each subnet must be retried.

[0114] Contract binding call process: The subnet administrator initiates a binding request, which includes the subnet ID, a list of subnet nodes, and the staking ID of the guarantor node; the mainnet calls the subnet binding contract to verify the request signature and the staking status of the guarantor node; the contract generates a binding event, the event content of which includes the mainnet identifier, subnet ID, binding request timestamp, and contract signature, and broadcasts it to all nodes in the subnet to be bound; after receiving the binding event, the subnet node triggers internal consensus, the consensus content of which is "whether to agree to access this mainnet", and each subnet node votes based on the mainnet's historical credit and the guarantor node's qualifications; after the subnet consensus is passed, a consensus success message is generated, the message including the signature list of all nodes in the subnet and the number of nodes that passed the consensus. Total number of nodes in the subnet Send it to the mainnet.

[0115] Mainnet and Subnet Data Interoperability Configuration: After binding, the mainnet allocates a dedicated communication port to the subnet and configures a data interaction whitelist. Only mainnet consensus nodes and audit nodes can initiate data query requests to the subnet, and only subnet consensus nodes can upload data to the mainnet. Data transmission encryption is enabled, using the TLS 1.3 protocol. The key is dynamically distributed by the mainnet identity management node and updated every 24 hours. At the same time, the mainnet adds a "Subnet Binding Root Hash" field to the block header to store the identifiers and hash values ​​of all bound subnets, realizing a unified index of subnets by the mainnet.

[0116] S2. Deploy a standardized traceability template for semiconductor film diffusion process in the blockchain mainnet;

[0117] The standardized traceability template includes a multi-dimensional structured field set, and the data is judged by the correlation between fields. Data that fails the correlation verification will be marked as "to be completed" and the data upload will be suspended until the correlation field data is completed and the verification is passed again.

[0118] Template deployment process: The mainnet governance node initiates a template deployment proposal, which includes the initial template version, field definition documents, and association rule descriptions. The mainnet calls the "template review smart contract" to perform formal verification on the proposal content. After the verification is passed, a mainnet consensus vote is triggered. The proposal can only take effect if ≥2 / 3 of the mainnet consensus nodes sign and agree. After taking effect, the template is stored in the "template storage area" of the mainnet contract account in a structured data format, and a unique template ID is generated. The mainnet simultaneously sends a "template deployment notification" to all bound subnets, including the template ID, download address, and effective time.

[0119] Template version management: The mainnet establishes a version traceability mechanism for templates, with version numbers following the rule of "V + year + month + revision number"; template revisions must meet one of three triggering conditions: ① semiconductor film diffusion process update; ② new data requirements from participating parties; ③ optimization of field association rules; the revision process is consistent with the initial deployment, and a "version change log" must be recorded in the mainnet block after revision. Old version templates are archived to the mainnet historical database and retained for at least 3 years for future reference; subnets are compatible with the first two historical versions of the mainnet to avoid incompatibility of existing data due to template upgrades.

[0120] Template retrieval mechanism: Before uploading data, subnet nodes need to obtain the latest template through the mainnet's "template query interface"; the interface returns the complete field set of the template, field data type constraints, and required field markers; subnet nodes organize data according to the template format. If a required field is missing or the data format is incorrect, the mainnet will refuse to accept the data and return a specific prompt of "field missing / format error".

[0121] The standardized traceability template includes a multi-dimensional set of structured fields, and the data is judged by the correlation between fields. Data that fails the correlation between fields will be marked as "to be completed" and the data upload will be suspended until the correlation field data is completed and the data passes the re-verification.

[0122] The core logic of the correlation verification is as follows: The template has a built-in "field correlation mapping table" that defines the mandatory correlation relationships of key fields. Specifically, these include: ① Process parameter fields must be correlated with equipment identity information fields (e.g., "actual diffusion temperature value" must be bound to "unique identifier of diffusion furnace" to ensure that temperature data belongs to a specific equipment and avoid data confusion); ② Quality inspection fields must be correlated with process parameter fields (e.g., "semiconductor film thickness measurement value" must be bound to "diffusion time" and "doping gas flow rate value" to ensure that quality data corresponds to a specific process time period); ③ Anomaly handling record fields must be correlated with quality inspection fields and equipment data fields (e.g., "film thickness exceeding standard" anomalies must be bound to "film thickness measurement value" and "diffusion furnace temperature fluctuation record" to provide data support for anomaly root cause analysis).

[0123] Verification Execution Process: ① When a subnet node uploads data, the mainnet's "Template Verification Smart Contract" automatically extracts the associated fields from the data and compares them with the "Field Association Mapping Table"; ② If there are any unassociated mandatory fields (such as only uploading "film thickness measurement value" without binding "diffusion time"), the contract marks the data as "to be completed," generates a "Completion List" (listing the missing associated fields), and pushes it to the data upload node; ③ After the upload node completes the associated fields, it recalculates the data hash value (updated in conjunction with the original data hash value) and initiates the upload again; ④ The contract re-verifies the consistency between the association relationship and the data hash, and only after both pass the verification is the data stored in the corresponding subnet allowed;

[0124] Example of verification failure: If a device data subnet node uploads "Actual diffusion temperature = 1050.2℃" but does not associate it with the "Unique Identifier for Diffusion Furnace", the contract determines that the association verification has failed and returns "To be completed - Missing associated field 'Unique Identifier for Diffusion Furnace' (Format requirements: Manufacturer code + 6-digit serial number + 4-digit production year, such as 'ASML-123456-2023')". The data can only be stored after the node completes the identifier and passes the verification.

[0125] The standardized traceability template includes:

[0126] Basic wafer information fields include wafer unique identifier, wafer material type, and physical dimensions;

[0127] The equipment identification information field includes a unique identifier for the diffusion furnace, a hardware feature code of the diffusion furnace generated based on the physical characteristics of the equipment, and a recent calibration timestamp;

[0128] Process parameter fields include diffusion temperature setpoint and actual value, diffusion time, dopant gas type and flow rate, and diffusion chamber pressure.

[0129] Quality inspection fields include semiconductor film thickness measurement, sheet resistance measurement, and minority carrier lifetime measurement;

[0130] The operator information field includes the operator's digital identity and the hash value of the operator's qualification certificate;

[0131] The exception handling record fields include the exception type code, the timestamp of the exception occurrence, and a description of the handling measures taken.

[0132] S3. Collect process parameters and quality data through the sensor network deployed on the production line, and upload them to the corresponding subnet after preprocessing by the edge computing node;

[0133] The process of preprocessing and classifying data before uploading to the corresponding subnet via edge computing nodes includes:

[0134] S31. Edge computing nodes access raw data Collection timestamp and sensor identifiers Perform concatenation and calculate the hash value of the concatenated data: ;

[0135] S32. The corresponding subnet recalculates the hash value after receiving data. And verify the following two conditions:

[0136] Hash consistency: ;

[0137] Data freshness: the time difference between data collection and upload time. , The time difference threshold;

[0138] S33. Data is stored only if all of the above conditions are met; otherwise, storage is rejected and an exception notification is returned.

[0139] Specific verification process:

[0140] Hash Consistency Verification: After receiving the "data upload packet", the subnet consensus node first verifies the validity of the edge node's signature. If the signature is valid, the original concatenated data in the packet is extracted.

[0141] The hash value was recalculated using the same concatenation rules and SHA-256 algorithm as the edge computing nodes. ;

[0142] Comparison and If the two are completely consistent, the hash verification is considered successful; if they are inconsistent, the data is considered to have been tampered with, the "data upload packet" is immediately discarded, and a "hash inconsistency anomaly" is recorded, including the anomaly timestamp, upload node ID, etc. and Difference.

[0143] Data freshness verification:

[0144] Subnet consensus nodes extract the collection timestamp from the "data upload packet" And record the data reception time. ;

[0145] Calculate the time difference (Unit: seconds) Dynamically set according to data type: ① Process parameters (temperature, gas flow rate, pressure) =30 seconds (to ensure real-time performance and avoid parameter lag affecting process monitoring); ② Quality data =300 seconds (quality inspection is a batch operation, and a moderate delay is allowed);

[0146] like If the data is deemed fresh, the data upload is considered to have passed the freshness verification; otherwise, the data is deemed to have timed out and expired, the "data upload package" is discarded, and a "data timeout exception" is recorded.

[0147] S33. Data is stored only if all the above conditions are met; otherwise, storage is rejected and an exception notification is returned.

[0148] After receiving an anomaly notification, the edge computing node automatically triggers retry logic. If the hash is inconsistent, it reconstructs and recalculates the hash and uploads it. If the data times out, it re-collects the latest data and uploads it. The maximum number of retries is 3. If the 3 retries still fail, the edge node lights up the local alarm indicator and sends an alarm message to the production line control system to remind maintenance personnel to intervene and investigate, such as sensor failure or network interruption.

[0149] Reference Figure 3 S4. Through the quality anomaly handling smart contract deployed on the main network, based on the standardized traceability template and associated with multi-source data, quality anomalies are judged and graded for handling. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment and conflict probability assessment.

[0150] The quality anomaly handling smart contract is configured with a dynamic quality threshold library, which dynamically sets the upper and lower limits of the quality threshold based on the wafer material type, film layer type and process specifications.

[0151] Threshold dynamic update logic: The contract automatically triggers a threshold update every 7 days, or a manual update is initiated by ≥3 audit nodes on the mainnet. The update process is as follows: ① Extract good product data from the equipment data subnet and quality inspection subnet for the same process scenario in the past 7 days; ② Calculate the threshold using the threshold calculation model. ③ Recalculate the threshold range; ④ Compare the deviation between the new threshold and the current threshold. If the deviation is ≤5%, keep the current threshold; if the deviation is >5%, generate a "threshold update proposal" with the data source and calculation process attached; ⑤ After being confirmed by ≥2 / 3 consensus nodes on the mainnet, update the threshold library and record the "threshold update log", including the old threshold, the new threshold, and the reason for the update.

[0152] Process specification adaptation mechanism: When the semiconductor film diffusion process specification is updated, the contract supports adding a "process specification-threshold" mapping relationship through mainnet governance proposals.

[0153] S41. When the quality anomaly handling smart contract is executed, the quality detection field data in the standardized traceability template is automatically obtained, and compared with the corresponding dynamic threshold to determine whether there is an anomaly;

[0154] S42. If an anomaly is determined, the quality anomaly handling smart contract further correlates and analyzes the process parameters corresponding to the anomaly data points, identifies the abnormal process parameters that deviate from the standard, determines the root cause of the quality anomaly based on the predefined anomaly parameter mapping rules, generates a structured traceability report, and triggers an on-chain alarm event.

[0155] S43. The determination and grading of quality anomalies are based on a dynamic quality threshold library, the threshold range of which is determined by the following model: ,in, It is a wafer material type. For membrane type, Based on historical good product data statistical characteristics, For a threshold model trained on a gradient boosting tree, if the quality detection data... satisfy or If the error is not detected, it is considered an anomaly, and further correlation with process parameter fluctuations is used to distinguish the anomaly type.

[0156] Threshold model F is a multi-feature fusion threshold prediction model based on Gradient Boosting Tree (GBT) to dynamically output the acceptable range of quality indicators. The specific model construction and calculation process is as follows:

[0157] S431. Model Input Feature Processing

[0158] Input parameters (Wafer material type) (Membrane type) (Historical good product data statistical characteristics) are transformed into structured feature vectors that the model can recognize. The specific handling method is as follows:

[0159] wafer material type : Uses tag-based encoding for quantization, such as encoding monocrystalline silicon as Polycrystalline silicon is coded as Silicon carbide is coded as The encoding rules are pre-stored in the standardized traceability template field dictionary of the main network;

[0160] Film type L: Quantized using one-thermal encoding, such as the feature vector corresponding to a boron diffusion film. Phosphorus diffusion membrane Arsenic diffusion membrane corresponding The encoding dimension is consistent with the total number of membrane types predefined in the main network;

[0161] Historical Good Product Data Statistical Characteristics Extracted using the same process over the past 6 months (and) , Calculate statistical eigenvectors from quality inspection data (such as film thickness, sheet resistance, minority carrier lifetime) that are matched with the data. ,in For the mean, For standard deviation, For the median, Interquartile range;

[0162] Feature vector concatenation: The final model input feature vector is:

[0163] ,in The average temperature of the same process environment over the past 30 days, The average humidity over the past 30 days in the same process environment is used, and the environmental parameters are extracted from the process environment field of the equipment data subnet.

[0164] S432. Gradient Boosting Tree Model Structure Design

[0165] The threshold calculation model F adopts a gradient boosting tree structure with two output branches. The specific parameters and structure are as follows:

[0166] CART regression decision trees were used, with a number of trees T=100, a maximum depth of d=6 per tree, a minimum number of splits of 20 per sample, and a minimum number of leaf nodes of 10 per sample.

[0167] Set dynamic learning rate ,in The learning rate is gradually reduced as the number of training iterations increases to improve model stability.

[0168] The Huber loss function is used, and the formula is as follows:

[0169] ,in This represents the actual quality value based on historical good product data. These are the model's predicted values. , This is the average of historical good product data, used to reduce the impact of outliers on model training;

[0170] The model contains two parallel output branches, corresponding to... and ,in" "Branch" predicts the lower limit threshold of quality indicators. The "branch" predicts the upper limit threshold of the quality indicator.

[0171] S433. Model Training and Threshold Calculation Process

[0172] Step 1: Extract good product data from the mainnet historical database within the past 6 months under the same process scenario. Each data sample contains a feature vector. and the corresponding actual mass value For example, the actual measured value of the film thickness, to construct a training set. ,in Ensure that the sample size meets the model training requirements;

[0173] Step 2: For " "Branch" and " The branches initialize the constant prediction values ​​respectively, where , ,in, , The mean and standard deviation of the quality values ​​of the training set are the corresponding one-sided 90% confidence intervals.

[0174] Step 3: For each iteration t=1 to :

[0175] Calculate the residual: for " "Branch", residual ;right" "Branch", residual ;

[0176] Fitting decision trees: Fitting the residuals using CART regression trees. and , thus obtaining the t-th decision tree and ;

[0177] Calculate the step size: Determine the optimal step size through line search. and ;

[0178] That is, minimize the loss function:

[0179] and ;

[0180] Update the model: , ;

[0181] Step 4: After the iteration is complete, the model outputs the final threshold. , And through the formula , Environmental remediation is carried out, including This represents the standard deviation of the fluctuations in the current process environment parameters (temperature, humidity). The average standard deviation of historical environmental parameters. This represents the environmental impact coefficient, which is pre-stored in the mainnet smart contract to ensure that the threshold adapts to real-time environmental fluctuations.

[0182] The threshold calculation model F is updated periodically through the mainnet smart contract. The update cycle is synchronized with the dynamic quality threshold library. Specifically, the mainnet obtains the latest week's good product data from the device data subnet and the quality inspection subnet, retrains the model parameters according to steps 1-3, stores the new model parameter decision tree structure, residual fitting coefficients, and step size in the mainnet block, and synchronizes them to each functional subnet to ensure the timeliness and accuracy of the threshold calculation.

[0183] The process of determining and classifying quality anomalies includes:

[0184] S44. The contract calls the dynamic quality threshold library to obtain the value related to the current wafer material type. Film type Matching quality threshold range And compare them;

[0185] S45. Calculate the degree to which outlier data deviates from the threshold. And associate the fluctuation range of the corresponding process parameters in the equipment data subnet, and classify them according to the following rules:

[0186] Minor abnormalities: ,and ;

[0187] Moderate abnormality: ,or , ;

[0188] Serious abnormality: ,or ;

[0189] in, These are the actual data values ​​for quality inspection fields, such as the actual measured values ​​of semiconductor film thickness, sheet resistance, minority carrier lifetime, etc.

[0190] This measures the degree to which outlier data deviates from a threshold, used to quantify the actual value. With threshold range The magnitude of the deviation;

[0191] The fluctuation range of the diffusion temperature corresponds to the degree of deviation between the actual value and the set value of the diffusion temperature in the corresponding process.

[0192] The fluctuation range of the dopant gas flow rate corresponds to the degree of deviation between the actual and set values ​​of the dopant gas flow rate in the corresponding process.

[0193] Minor anomaly, the smart contract triggers a local re-inspection, and the instruction quality detection subnet re-collects the quality data of more than 3 samples in the current batch. If the re-inspection data is qualified, the anomaly is marked as a misjudgment; otherwise, an anomaly reminder is pushed to the production line operator. The contract sends a re-inspection instruction to the quality detection subnet, clarifying the sample selection rules, randomly selecting 3 to 5 wafers from the abnormal batch, and adding 2 detection points to each wafer. After the subnet completes the re-inspection, it returns the re-inspection data (including the detection point positions and device signatures). The contract compares the re-inspection data with the threshold. If ≥80% of the re-inspection data is qualified, the anomaly is marked as "misjudgment", the anomaly mark is deleted, and a log is recorded. If <80% is qualified, an anomaly reminder is pushed to the production line operator, and the reminder content includes: abnormal batch number, re-inspection pass rate, and recommended process parameter adjustments.

[0194] Moderate anomaly, the smart contract triggers subnet data freezing, pauses the new data upload of the device data subnet associated with the anomaly, and simultaneously generates an anomaly traceability task, which is pushed to the device supplier node. The contract sends a "data freezing instruction" to the device data subnet, clarifying the freezing scope, only the process parameter data corresponding to the abnormal batch, without affecting other batches. An anomaly traceability task is generated, including the snapshot hash of the process parameters of the abnormal batch and the device ID to be investigated, and is pushed to the device supplier node. The supplier node needs to complete the investigation within 24 hours and return an "equipment investigation report", including the test results and repair records. After the contract verifies the report, if it confirms a device failure, it triggers the device repair process. If it confirms an incorrect process parameter setting, it pushes parameter correction suggestions to the production line.

[0195] Severe anomaly, the smart contract triggers cross-subnet traceability verification, synchronously freezes the corresponding batch data of the device data subnet and the quality detection subnet, calls the main network audit node to verify the consensus records and data transfer logs during the abnormal period, generates a traceability report and records it in the main chain block. The contract synchronously sends a "batch data freezing instruction" to the device data subnet and the quality detection subnet, freezing the modification and deletion permissions of all data in this batch. Call ≥3 audit nodes on the main network to verify 3 types of data during the abnormal period: subnet consensus records (whether there are malicious votes by nodes), data transfer logs (whether there are traces of data tampering), and node pledge status (whether there is insufficient collateral for nodes). After the audit node completes the verification, it generates a "cross-subnet traceability report", which is submitted to the contract after being signed. The contract writes the report hash into the main network block and notifies the wafer factory to start the "abnormal batch isolation process" to prevent abnormal wafers from flowing into downstream links.

[0196] S46. The tampering risk assessment includes:

[0197] S461. Extract the data shard set of the abnormally associated batch from the device data subnet or the quality detection subnet , where n is the total number of shards of the data in this batch;

[0198] S462. Obtain the hash checksum of each shard at different nodes in the subnet, and calculate the hash difference value of each shard. ,in , They are shards The hash value between any two nodes in the subnet;

[0199] S463. Tampering with the risk index calculation formula: ,in For fragmentation The absolute value of the hash difference;

[0200] S464. Preset tampering risk threshold ,like The risk of data tampering in this batch was determined to be low, and the data reliability met the traceability requirements; if If the risk of tampering is deemed high, the contract triggers a data retransmission verification, requiring subnet nodes to re-upload the original data for this batch and perform a second verification.

[0201] S47. The local congestion risk assessment includes:

[0202] S471. Extract the average verification delay time of each node during the abnormal period from the node delay logs of the subnet associated with the main network synchronization anomaly. ,in This represents the total number of nodes in the subnet.

[0203] S472. Calculate the overall average verification delay of the subnet: ;

[0204] S473. Calculate the local congestion risk index: ,in For nodes The absolute difference between the delay and the average delay of the subnet;

[0205] S474. Preset local congestion risk threshold ,like If the subnet is determined to have no local congestion, the data transmission reliability is high; If a local congestion risk is detected, the contract triggers a subnet load balancing instruction to distribute some of the data processing tasks of the abnormal node to nodes with lower load.

[0206] S48. The conflict probability assessment includes:

[0207] S481. Extract the abnormal data transfer operation records from the interaction logs between the main network and the abnormal associated subnet, totaling L transfer operations, and record the time interval of each transfer. and consensus deviation , representing the difference between the maximum and minimum confirmation times at different nodes for the h-th transfer;

[0208] S482. Calculate the average consensus bias across all transfer operations: ;

[0209] S483. Calculate the conflict probability index: , where ε is a preset positive number used to prevent the denominator from being zero;

[0210] like The abnormal data was determined to have no flow conflicts, and the data credibility met the traceability requirements.

[0211] like If a risk of data transfer conflict is detected, the contract triggers a data transfer path verification. This involves comparing the actual data transfer path of the abnormal data with the preset compliant path, locating the conflict node, and recording it in the audit log. This indicates a preset threshold for the likelihood of conflict.

[0212] S49. The asset pledge contract stipulates that if the hash verification mechanism or subsequent audit reveals that the functional subnet node has engaged in data tampering or forgery, the digital assets pledged by the guarantor node will be automatically deducted according to a preset ratio ρ, and the value of the deducted assets will be transferred to the mainnet's quality assurance fund account to compensate for the quality loss caused by inaccurate data.

[0213] Quality Assurance Fund Management: The fund is only used to compensate for "quality losses caused by inaccurate data". Compensation applications must be submitted by the affected party (such as downstream customers) and accompanied by proof of quality loss (third-party testing report, loss amount calculation table). The mainnet governance committee reviews compensation applications quarterly and allocates compensation funds from the fund upon approval. When the fund balance is insufficient, a mainnet proposal is triggered to adjust the basic pledge amount and supplement the fund source.

[0214] S5. Implement data access and control for multi-level users based on an enhanced role-based access control model.

[0215] The enhanced role-based access control model includes the following permission levels:

[0216] The wafer fab operator role is granted access to raw process data and quality data across all functional subnets.

[0217] Data access permissions: Access to all raw process data in the equipment data subnet, including real-time temperature curves of diffusion furnaces, historical parameter reports (last 12 months), and sensor calibration records; raw data of single wafers with multiple inspection points in the quality inspection subnet, including inspection equipment IDs, inspection personnel signatures, inspection timestamps, batch pass rate statistics, and heat maps of abnormal wafer distribution; access to the digital identity information of participants stored in the mainnet, excluding sensitive fields such as private keys, contact information, and historical versions of standardized traceability templates.

[0218] Operation permissions: Can perform data export operations, initiate quality anomaly re-inspection applications, and change its own operation password; cannot perform data deletion or tampering operations, and all data modifications must be confirmed by the signatures of ≥3 audit nodes on the main network.

[0219] Access restrictions: Operators can only access production line data within their own wafer fab and cannot access data across wafer fabs. Data isolation is achieved through the "Fab Identifier" field on the main network. For example, if the operator's DID contains the identifier "WF001", only the subnet data with the fab identifier "WF001" will be matched.

[0220] The equipment supplier role is only granted access to operational parameter data related to the equipment it provides and directly associated quality anomaly information.

[0221] Data access permissions: Access is only permitted to the equipment data subnet connected to the equipment provider, including: equipment operating parameters for the past 3 months, temperature fluctuation records, gas flow deviation statistics, chamber pressure stability reports, and equipment fault alarm records. Abnormalities must be labeled with "equipment-related" reasons, such as "abnormal heating tube resistance leading to temperature runaway." Access is only permitted to quality abnormality information directly associated with the equipment, such as "film thickness exceeding standards in this equipment's production batch." Access is not permitted to data from other brands and models of equipment, or abnormal records not caused by equipment issues, such as operator error or incorrect process parameter settings.

[0222] Operation permissions: Users can query equipment operating parameter trend charts, download equipment fault reports, and submit equipment maintenance suggestions; they cannot perform operations such as data export or re-inspection applications. All data access behaviors must be synchronized to the wafer fab monitoring node in real time.

[0223] Access control mechanism: The device vendor's DID must include "device manufacturer code + device serial number prefix". The main network access control module automatically filters data of unrelated devices by matching the DID with the prefix of the device's unique identifier, thus preventing unauthorized access.

[0224] Downstream customer roles are only granted access to the final quality inspection results summary and confirmed anomaly handling records;

[0225] Data access permissions: Access is only allowed for the purchased wafer batches, and the quality inspection result summary matched by the purchase contract number. The summary fields include: wafer batch number, average film thickness, average sheet resistance, average minority carrier lifetime, batch pass rate, and anomaly handling conclusion. Access is only allowed for anomaly handling records confirmed by the main network audit node. No process parameters, such as diffusion temperature, gas flow rate, or other customer purchase batch data, are allowed.

[0226] Operation permissions: Users can view quality inspection result summaries online and print quality qualification certificates with the main network signature; they cannot perform operations such as data downloading or querying historical batches. For purchase batches older than one year, an application must be submitted to the wafer fab.

[0227] Data anonymization rules: In the quality inspection result summary, all fields involving the testing equipment ID and quality inspector information must be anonymized. For example, the testing equipment ID is displayed as "Thickness-Tester-XXX", and the last 3 digits of the serial number are hidden to avoid leaking wafer fab production equipment information.

[0228] The system records all query operations on the mainnet through a query log smart contract. The record includes the queryer's digital identity, query time, query data range, and query result hash.

[0229] Example 2

[0230] Reference Figure 4 This embodiment provides a blockchain-based semiconductor film diffusion quality traceability system, including:

[0231] The network binding module is configured to build a blockchain mainnet and multiple subnets associated with it. The mainnet and subnets are connected through a trusted binding mechanism with node economic guarantees. The subnets include a device data subnet for storing process data of diffusion processes and a quality inspection subnet for storing quality performance testing data.

[0232] The standardized template deployment module is configured to deploy a standardized traceability template for semiconductor film diffusion processes in the blockchain mainnet;

[0233] The data processing module is configured to collect process parameters and quality data through the sensor network deployed on the production line, and then classify and upload them to the corresponding subnet after preprocessing by the edge computing nodes.

[0234] The risk assessment module is configured to use a quality anomaly handling smart contract deployed on the mainnet to determine and classify quality anomalies based on the standardized traceability template and associated multi-source data. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment and conflict probability assessment.

[0235] The access control module is configured to implement multi-level user data access and control based on an enhanced role-based access control model.

[0236] A computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device, the aforementioned blockchain-based semiconductor film diffusion quality traceability method.

[0237] A terminal device includes a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor to provide a blockchain-based semiconductor film diffusion quality traceability method.

[0238] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A blockchain-based method for tracing the diffusion quality of semiconductor films, characterized in that, include: A blockchain mainnet and multiple subnets associated with it are constructed. The mainnet and subnets are connected through a trusted binding mechanism with node economic guarantees. The subnets include an equipment data subnet for storing process data of diffusion processes and a quality inspection subnet for storing quality performance testing data. A standardized traceability template for semiconductor film diffusion process is deployed in the blockchain mainnet. The standardized traceability template includes a multi-dimensional structured field set and is judged by the correlation verification between fields. Data that fails the correlation verification will be marked as to be completed and data upload will be suspended until the correlation field data is completed and re-verified. Process parameters and quality data are collected by a sensor network deployed on the production line, and then classified and uploaded to the corresponding subnet after preprocessing by edge computing nodes. By deploying a quality anomaly handling smart contract on the mainnet, and associating multi-source data based on the standardized traceability template, quality anomalies are judged and graded for handling. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment, and conflict probability assessment. The quality anomaly handling smart contract is configured with a dynamic quality threshold library. This library dynamically sets upper and lower limits for quality thresholds based on wafer material type, film layer type, and process specifications. The judgment and tiered handling of quality anomalies are based on this dynamic quality threshold library, and the threshold range of the dynamic threshold library is determined by the following model: ,in, It is a wafer material type. For membrane type, Based on historical good product data statistical characteristics, For a threshold model trained on a gradient boosting tree, if the quality detection data... satisfy or If the threshold is not found, it is determined to be abnormal, and further correlation with process parameter fluctuations is used to distinguish the type of abnormality. The threshold model adopts a gradient boosting tree structure with dual output branches. When the quality anomaly handling smart contract is executed, it automatically obtains the quality detection field data in the standardized traceability template, compares it with the corresponding dynamic threshold, and determines whether there is an anomaly. If an anomaly is detected, the quality anomaly handling smart contract further correlates and analyzes the process parameters corresponding to the abnormal data points, identifies abnormal process parameters that deviate from the standard, determines the root cause of the quality anomaly based on predefined abnormal parameter mapping rules, generates a structured traceability report, and triggers on-chain alarm events. This system implements data access and control for multi-level users based on an enhanced role-based access control model.

2. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The trusted binding mechanism for node economic guarantees includes: Before a subnet is connected to the mainnet, authorized guarantor nodes in the mainnet pledge a preset amount of digital assets for the subnet to be connected, and the detailed information of the preset amount of digital assets pledged is recorded in the distributed ledger of the mainnet through an on-chain asset staking contract. After the contract of the trusted binding mechanism of the subnet is invoked, a binding event with a cryptographic signature is generated, triggering the consensus mechanism within each subnet. After each subnet returns a consensus success message, the mainnet performs consensus consistency verification, completing the trusted binding and data communication between the mainnet and each subnet.

3. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The process of preprocessing and classifying data before uploading to the corresponding subnet via edge computing nodes includes: Edge computing nodes for raw data Collection timestamp and sensor identifiers Perform concatenation and calculate the hash value of the concatenated data: ; The corresponding subnet recalculates the hash value after receiving the data. And verify the following two conditions: Hash consistency: ; Data freshness: the time difference between data collection and upload time. , The time difference threshold; Data will only be stored if all of the above conditions are met; otherwise, storage will be refused and an exception notification will be returned.

4. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The process of determining and classifying quality anomalies includes: The contract calls the dynamic quality threshold library to obtain the value relative to the current wafer material type. Film type Matching quality threshold range And compare them; For minor anomalies, the smart contract triggers a local re-inspection, instructing the quality inspection subnet to re-collect quality data from at least three samples of the current batch. If the re-inspection data is qualified, the anomaly is marked as a misjudgment; otherwise, an anomaly alert is pushed to the production line operator. In the case of a moderate anomaly, the smart contract triggers a subnet data freeze, suspends the uploading of new data to the data subnet of the anomaly-related device, and generates an anomaly tracing task, which is then pushed to the device supplier node. In case of a serious anomaly, the smart contract triggers cross-subnet tracing and verification, simultaneously freezing the corresponding batch data of the device data subnet and the quality inspection subnet, calling the mainnet audit node to verify the consensus records and data flow logs of the abnormal period, generating a tracing report and recording it in the main chain block.

5. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The tampering risk assessment includes: Extract the data fragment set of abnormally associated batches from the equipment data subnet or quality inspection subnet. , where n is the total number of data fragments in this batch; Obtain the hash checksum of each shard on different nodes in the subnet, and calculate the hash difference value of each shard. ,in , They are shards The hash value between any two nodes in the subnet; Tampering with the risk index calculation formula: ,in For fragmentation The absolute value of the hash difference; Preset tampering risk threshold ,like The risk of data tampering in this batch was determined to be low, and the data reliability met the traceability requirements; if If the risk of tampering is deemed high, the contract triggers a data retransmission verification, requiring subnet nodes to re-upload the original data for this batch and perform a second verification.

6. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The local congestion risk assessment includes: Extract the average verification latency of each node during the abnormal period from the node latency logs of the subnet associated with the mainnet synchronization anomaly. ,in This represents the total number of nodes in the subnet. Calculate the overall average verification latency of the subnet: ; Calculate the local congestion risk index: ,in For nodes The absolute difference between the delay and the average delay of the subnet; Preset local congestion risk threshold ,like If the subnet is determined to have no local congestion, the data transmission reliability is high; If a local congestion risk is detected, the contract triggers a subnet load balancing instruction to distribute some of the data processing tasks of the abnormal node to nodes with lower load.

7. The semiconductor film diffusion quality traceability method based on blockchain according to claim 1, characterized in that, The conflict probability assessment includes: Extract the abnormal data transfer operation records from the interaction logs between the main network and the abnormal associated subnet. There are a total of L transfer operations, and the time interval of each transfer is recorded. and consensus deviation , representing the difference between the maximum and minimum confirmation times at different nodes for the h-th transfer; Calculate the average consensus bias across all transfer operations: ; Calculate the conflict probability index: , where ε is a preset positive number used to prevent the denominator from being zero; like The abnormal data was determined to have no flow conflicts, and the data credibility met the traceability requirements. like If a risk of data transfer conflict is detected, the contract triggers a data transfer path verification. This involves comparing the actual data transfer path of the abnormal data with the preset compliant path, locating the conflict node, and recording it in the audit log. This indicates a preset threshold for the likelihood of conflict.

8. A blockchain-based semiconductor film diffusion quality traceability system, characterized in that, A blockchain-based semiconductor film diffusion quality traceability method according to any one of claims 1-7 includes: The network binding module is configured to build a blockchain mainnet and multiple subnets associated with it. The mainnet and subnets are connected through a trusted binding mechanism with node economic guarantees. The subnets include a device data subnet for storing process data of diffusion processes and a quality inspection subnet for storing quality performance testing data. The standardized template deployment module is configured to deploy a standardized traceability template for semiconductor film diffusion processes in the blockchain mainnet; The data processing module is configured to collect process parameters and quality data through the sensor network deployed on the production line, and then classify and upload them to the corresponding subnet after preprocessing by the edge computing nodes. The risk assessment module is configured to use a quality anomaly handling smart contract deployed on the mainnet to determine and classify quality anomalies based on the standardized traceability template and associated multi-source data. The execution process of the smart contract includes multi-dimensional risk assessment of the data, including tampering risk assessment, local congestion risk assessment and conflict probability assessment. The access control module is configured to implement multi-level user data access and control based on an enhanced role-based access control model.

Images