Industrial data secure transmission method based on block chain

Abstract

The invention belongs to the field of industrial data secure transmission, and particularly relates to a block chain-based industrial data secure transmission method, which comprises the following steps of: acquiring the rotating speed of a fan impeller, dividing the rotating speed into equal beat scales, and initializing a local number at an edge side; recording each alarm event, calculating a content fingerprint, binding a current beat scale and a local number, recording a current environment brief value, and generating an event record; in the off-network stage, a plurality of events are packaged into voucher packages by a packaging side according to start-stop beat scales and local numbers, and the number of packaging pieces, the first and last beat scales and the local numbers are attached; when the network is recovered, only the packet header of the voucher packet is uploaded, consistency verification and conflict detection are carried out on the packet header and the event record, and feedback treatment and tracing details are generated; according to the invention, under the condition of network disconnection or weak network, the communication cost can be obviously reduced, and the tamper-resistant and dispute-resistant rapid convergence capability can be improved.

Description

Technical Field

[0001] This invention belongs to the field of secure industrial data transmission, and specifically relates to a blockchain-based method for secure industrial data transmission. Background Technology

[0002] The Chinese patent application (CN202510162110.6) discloses a blockchain-based method for secure transmission of industrial data. This invention involves a server receiving industrial data requests, acquiring user behavior feature vectors, classifying and predicting them, and calculating the probability of legitimacy. The server sends a data key and encrypted industrial data to the user device to generate a user key. These two keys are combined to obtain a combined key, which is then used to re-encrypt the encrypted data, generating second encrypted industrial data. The requester and the proxy node verify the data using blockchain verification information. After verification, the requester queries the combined key and the second encrypted data on the proxy node. The requester and the user device obtain the data key through authorization information authentication, then use the combined key to decrypt the second encrypted data to obtain the first decrypted data, and finally use the data key to decrypt the original industrial data.

[0003] While this invention, in the field of industrial data security transmission technology, prevents the risk of illegal acquisition and tampering during data transmission by employing secondary encryption to ensure the security of industrial data transmission, it still has the following problems in the wind power operation and maintenance scenario: After being stored, wind power operation and maintenance data is used to record equipment operating information. Due to the need for multi-party collaboration in later maintenance, and the fact that wind power operation and maintenance areas are remote and prone to weak networks, critical alarms may become disordered, and the authenticity of remote control commands may be difficult to distinguish. In addition, after a network outage, data may experience packet loss and retransmission during data recovery, making it difficult to determine the accurate boundaries of responsibility during maintenance, thereby increasing maintenance and management costs.

[0004] To address the aforementioned issues, this invention proposes a blockchain-based method for secure industrial data transmission. This method constructs a closed loop consisting of a beat scale, local number, ternary tags, supporting evidence fragments, offline voucher packaging, and on-chain summary verification. It uses package summaries to form continuous small ledgers. After recovery, ternary tag reconciliation and supporting evidence continuity checks jointly identify replay insertions and tampering, thereby restoring a consistent timeline across entities and achieving reliable transmission with minimal on-chain traffic, without revealing operational details. Summary of the Invention

[0005] This invention provides a blockchain-based method for secure transmission of industrial data, aiming to solve the following problems: In wind power operation and maintenance scenarios, wind power operation and maintenance data is used to record equipment operation information after storage. Due to the need for multi-party collaboration in later maintenance, and the fact that wind power operation and maintenance areas are remote and prone to weak networks, critical alarms may become disordered, and the authenticity of remote control commands may be difficult to distinguish. In addition, after a network outage, data may experience packet loss and retransmission during data recovery, which may make it difficult to determine the accurate boundary of responsibility during maintenance, thereby increasing maintenance and management costs.

[0006] The technical solution adopted by this invention to solve the above problems is: a blockchain-based method for secure transmission of industrial data, comprising:

[0007] Step S1: Collect the fan impeller speed and divide it into equal parts of the beat scale, and initialize the local number on the edge side;

[0008] Step S2: Based on the beat scale and local number, record each alarm event and calculate the content fingerprint, bind the current beat scale and local number, record the current environmental summary value, and generate an event record;

[0009] Step S3: Based on the alarm event, during the network outage phase, the packaging side packages multiple events into a voucher package according to the start and end beat scale and local number, and attaches the number of packaged items and the start and end beat scale and local number.

[0010] Step S4: Based on the credential packet, during network recovery, only the packet header and packet-level digest of the credential packet are uploaded, and consistency verification and conflict detection are performed with the event log to generate feedback governance and traceability details.

[0011] As a preferred embodiment, the specific steps for collecting the fan impeller speed and dividing it into equal intervals on the beat scale, and initializing the local number on the edge side are as follows:

[0012] The impeller speed is acquired by reading the A / B phase pulses of the incremental encoder on the main shaft and using the Z phase zero position as the reference per revolution. The number of pulses per revolution is used to calculate the mechanical angular velocity and angle. The mechanical angular velocity is quantized into equal-division clock cycles T according to the Circle per impeller revolution. The time-incrementing number ADD is initialized on the edge side, with the initial increment set to ADD0. After each event is generated, ADD0 is incremented by 1, and rollback is allowed but unrollback is prohibited, thus obtaining the initialized local number.

[0013] As a preferred embodiment, the specific steps for recording each alarm event and calculating the content fingerprint are as follows:

[0014] Calculate the content fingerprint TUN for the generated alarm event, and bind it to the current tick scale T and local number A of the current alarm event to obtain a ternary label. .

[0015] As a preferred embodiment, the specific steps of recording each alarm event and calculating the content fingerprint further include:

[0016] Record the brief environmental values ​​of the current alarm event. Using the beat scale T as the phase reference, perform phase alignment on the wind speed and output power before the alarm event and resample them into an equal-length sequence. Extract the scalarized feature vector STA from the equal-length sequence according to a fixed dimension to obtain the brief environmental values ​​used to characterize the statistical features of the environmental measurements. Generate an independent signature and store it along with the ternary record to generate the event record. .

[0017] As a preferred embodiment, the specific steps for the packaging side to package multiple events into a voucher package according to the start and end timescale and local number during the network disconnection phase are as follows:

[0018] The packetizing side packages the alarm events generated during the network outage phase into credential packets. During the packaging process, a comprehensive cost function is constructed, and thresholds for the number of entries t1, time t2, and bytes t3 are adaptively set. When any threshold is reached, the current packet is closed and the number of events in the previous packet summary, the first and last beat scale, and the local number are written. After calculating the packet summary, the packet is enqueued and awaits uploading. Reinforcement learning is used for offline pre-training to obtain a packet-level continuity and low overhead that can be maintained under different link conditions.

[0019] As a preferred embodiment, the specific steps of packaging multiple events into a voucher package by the packaging side according to the start and end timescale and local number during the network disconnection phase further include:

[0020] The steps for defining the comprehensive cost function are as follows:

[0021] R1 collects link metrics, event arrival rate, and current packet status to obtain the input for cost calculation;

[0022] R2 calculates the arrival times of the number of records threshold t1, time threshold t2, and byte threshold t3 based on the input volume. The formula for calculating the number of records threshold is:

[0023] ,

[0024] Where t1 is the threshold for the number of items. To truncate negative values, B is the adaptive number of records threshold, and t is the current decision time. This represents the current event arrival rate.

[0025] The formula for calculating the time threshold is:

[0026] ,

[0027] Where t2 is the time threshold. To truncate negative values, t is the current decision time. The time elapsed since the beginning;

[0028] The formula for calculating the byte threshold is:

[0029] ,

[0030] Where t3 is the byte threshold. To truncate negative values, t represents the current decision time. For cumulative bytes, Where D is the arrival rate of the current event and D is the size of the current event in bytes.

[0031] By calculating all the above thresholds, the trough time required to trigger the threshold and the predicted number of events and bytes at the time of shutdown are obtained, providing a definite input for subsequent risk and cost calculations.

[0032] R3, based on all the above thresholds, predicts the number of packet segments by dividing the data into segments according to the Maximum Transmission Unit (MTU), and forms a risk function using the link packet loss rate and jitter term. The formula for the risk function is as follows:

[0033] ,

[0034] Where DA represents the risk score. Here, denoted as the single-segment packet loss rate, DU is the number of segments after segmentation, and E is the weighting coefficient for jitter penalty. Let F be the standard deviation of the round-trip time delay, and let F be the mean of the round-trip time delay. It is the numerical stability constant;

[0035] By constructing a risk function, the quantified packet transmission failure tendency is obtained;

[0036] R4, based on all the above thresholds and risk functions, synthesizes a comprehensive cost function, the formula of which is:

[0037] ,

[0038] For the overall cost, The weighting factor for the delay in reimbursement is Dealy, where Dealy is the delay in reimbursement. Here, Lose represents the weighting coefficient for packet loss risk. is the weighting coefficient for chain overhead, and Spend is the chain overhead.

[0039] As a preferred embodiment, the specific steps of packaging multiple events into a voucher package by the packaging side according to the start and end timescale and local number during the network disconnection phase further include:

[0040] Offline pre-training using reinforcement learning is employed. Standardized samples are formed by extracting observable states, executed thresholds (t1, t2, t3), and their effects from historical records, removing outliers and missing items, and solidifying field order and version numbers. Continuous policies are trained using historical decisions as supervisory signals, enabling them to output adaptive thresholds (t1, t2, t3) in an offline environment. Boundary and monotonic constraints are applied to the output to obtain usable initial policies. Conservative reinforcement learning is used to estimate the long-term state-action payoff, suppressing overestimation bias in out-of-distribution actions. Combined with policy enhancement using advantage weights, a more robust offline policy to weak network fluctuations is obtained. Soft constraints such as packet loss rate, latency jitter, and packet length limits are introduced during training and selection. Offline evaluation methods are used to set thresholds for timeliness, reliability, and cost; policies failing to meet these thresholds are not deployed to the online stage.

[0041] As a preferred implementation, the specific steps for uploading only the packet header and packet-level digest of the credential packet during network recovery, and performing consistency verification and conflict detection with the event log, are as follows:

[0042] During network recovery, the header (Head) and short packet digest of the uploaded credential packet are used as central anchors to obtain publicly available comparison records with low traffic and low leakage. Subsequently, consistency verification and conflict detection are performed, including: integrity checks, retrieving details from the credential packet as needed, serializing the locally recalculated digest according to specifications, and comparing the reconstructed header (RHead) with the header of the credential packet uploaded during network recovery. If RHead and Head match, the packet is considered complete; otherwise, it is marked as having a digest inconsistency or suspected tampering, and the affected credential packet and its adjacent packets are immediately frozen. For the frozen packet, details are retrieved, the header (RRHead) is reconstructed, and the digest is recalculated, while each event record is verified. Signature and ternary tag Consistency.

[0043] As a preferred implementation, the specific steps of uploading only the packet header and packet-level digest of the credential packet during network recovery, and performing consistency verification and conflict detection with the event log, further include:

[0044] By verifying whether adjacent packet boundaries are seamlessly connected and whether the local number A is monotonically non-backtracking, and sorting by <packet order, A, T> to remove duplicates, it was found that adjacent packet boundaries were not seamlessly connected and the local number A was not monotonically backtracking, and the packet number and event index were located. A physical matching test was performed to recover the impeller main frequency R based on the scale T. The feature vector STA was inversely transformed to reconstruct the wind speed and power sequence and verify the signature. The main frequency coherence degree SIM and the main frequency deviation degree were calculated and judged.

[0045] As a preferred implementation, the specific steps for generating feedback governance and traceability details are as follows:

[0046] The test results are categorized and output. The results of integrity, logical consistency verification and physical matching tests are summarized and output to generate governance and traceability details. This is used to automatically trigger repackaging, retransmission, signature review and replay window correction to obtain auditable disposal instructions and traceable evidence chains. After passing items are recorded and failing items are frozen for rectification, the chain continuity and cross-entity consistent timeline are restored.

[0047] The beneficial effects of this invention are as follows:

[0048] 1. Divide each revolution of the impeller into equal parts to obtain the beat scale, maintain incremental numbering on the edge side, and generate a ternary label for each event to synchronously record statistical features. This results in an event stream that can still be globally sorted by physical order, deduplicated, and fingerprint verified even in scenarios without unified time synchronization, weak network, or network outage. This can eliminate alarm disorder, suppress forgery and replay, and provide a machine-readable evidence chain for multi-party collaboration.

[0049] 2. During the network outage, the event chain is bound into a credential packet P. When the network is restored, only the packet header and packet-level digest are sent and consistency verification and conflict detection are performed. This results in cross-entity verification capabilities with low traffic and low leakage, enabling uninterrupted authentication during network outages, rapid location of packet loss and rollback after network restoration. This allows for quick clarification of responsibility boundaries, reduction of maintenance and management costs, and ensures consistent timelines and auditability of operation and maintenance data throughout the entire lifecycle. Attached Figure Description

[0050] Figure 1 This is a flowchart of the present invention.

[0051] Figure 2 These are technical effect diagrams of the present invention, wherein the black bar chart represents the present invention, and the gray bar chart represents the prior art. Detailed Implementation

[0052] To make the technical means, creative features, and achieved objectives and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention. Unless otherwise specified, the experimental methods in the following embodiments are conventional methods, and the materials and reagents used in the following embodiments are commercially available unless otherwise specified.

[0053] Example 1: This application proposes a blockchain-based method for secure industrial data transmission, combined with... Figure 1 As shown, a blockchain-based method for secure industrial data transmission includes:

[0054] Step S1: Collect the fan impeller speed and divide it into equal parts of the beat scale, and initialize the local number on the edge side;

[0055] Specifically, the impeller speed is acquired by reading the A / B phase pulses from the incremental encoder on the main shaft and using the Z phase zero position as the reference per revolution. The number of pulses per revolution is then used to calculate the mechanical angular velocity and angle, providing a stable physical anchor point for the quantization of the beat scale T and event sequencing. Furthermore, the mechanical angular velocity is quantized into equally divided beat scales T according to the Circle per impeller revolution, where the range of the beat scale is... This indicates the position of each beat increment on the time scale; the time-incrementing ADD is initialized on the edge side, specifically: the initial increment is set to ADD0, ADD0 is incremented by 1 after each event is generated, and permission is set. The process rewinds but prevents rewinding, obtaining an initialized local number for deduplication, replay prevention, and gap detection, and providing a scale for the first and last numbers of the voucher package and chain verification; the aforementioned beat scale and local number serve as a sortable physical sequence under network-less time synchronization conditions, providing basic coordinates for subsequent event record generation, voucher package consistency, and conflict detection.

[0056] Step S2: Based on the beat scale and local number, record each alarm event and calculate the content fingerprint, bind the current beat scale and local number, record the current environmental summary value, and generate an event record;

[0057] Specifically, alarm events are generated when generator overheating, gearbox oil temperature is high, pitch system failure, yaw system failure, brake failure, grid connection failure, communication interruption, and emergency stop occur. A content fingerprint (TUN) is calculated for each generated alarm event, and this fingerprint is bound to the current alarm event's tick scale (T) and local number (A) to obtain a ternary tag. The system records the brief environmental values ​​of the current alarm event. Specifically, using the beat scale T as the phase reference, the wind speed and output power measured 5 seconds before the alarm event are phase-aligned and resampled into an equal-length sequence. From this equal-length sequence, a scalarized feature vector STA is extracted according to a fixed dimension of 20. This yields the brief environmental values ​​used to characterize the statistical features of the environmental measurements over the past 5 seconds. An independent signature is then generated and stored along with the ternary record to generate the event record. It is used for subsequent consistency verification and detection of data to prevent data from being tampered with, and provides data for the calculation of the package-level digest of the credential package digest;

[0058] The calculation of the content fingerprint of the generated alarm event is based on the minimum business, clock tick T, local number A, motor equipment number and version number of the currently generated alarm event. The content fingerprint is serialized in a fixed order (ConciseBinary Object Representation, CBOR) to obtain the CBOR-encoded byte sequence CB for hash calculation. The CB is hashed using Blake2s-128 and the high 64 bits are taken as the content fingerprint TUN of the alarm event. This is used as the core anchor point of the credential binding and the hash input of the packet digest D_pack. It can form a self-checking and traceable small ledger during the network outage stage without exposing business details, thereby preventing data tampering.

[0059] Step S3: Based on the alarm event, during the network outage phase, the packaging side packages multiple events into a voucher package according to the start and end beat scale and local number, and attaches the number of packaged items and the start and end beat scale and local number.

[0060] Specifically, during the network outage or weak network phase, the edge unit independently determines and generates alarm events based on local thresholds and state machines. For example, when the current network outage state occurs, any of the following alarm events may occur: excessive engine room vibration, generator stator overheating, grid connection failure, etc. The packaging side packages the alarm events generated during the network outage phase into a voucher package, and attaches the number of alarm events in the current voucher package, the first and last beat scales, and the local number.

[0061] The alarm events generated during the network outage phase are packaged into credential packets by the packaging side. When packaging, the aggregation degree between packets needs to be considered to avoid the risk of packet loss and increased reconciliation latency and retransmission probability due to a single packet exceeding the preset upper limit, and the increase in the number of anchor records and on-chain overhead and fragmentation of continuity due to a single packet being less than the preset lower limit. Therefore, based on the comprehensive cost ALL = reconciliation latency Dealy + packet loss risk Lose + chain overhead Spend, the number of records threshold t1, time threshold t2, and byte threshold t3 are adaptively set. When any threshold is reached, the current packet is closed and the number of records in the previous packet summary event, the first and last beat scale, and the local number are written. After calculating the packet summary, it is enqueued and waits to be sent. Reinforcement learning is used for offline pre-training to obtain a packet-level continuity and low overhead under different link conditions.

[0062] The steps for defining the comprehensive cost function are as follows:

[0063] R1 collects link metrics, event arrival rate, and current packet status to obtain the input for cost calculation;

[0064] R2 calculates the arrival times of the number of records threshold t1, time threshold t2, and byte threshold t3 based on the input volume. The formula for calculating the number of records threshold is:

[0065] ,

[0066] Where t1 is the threshold for the number of items. To truncate negative values ​​and ensure that the waiting time is not negative, B is the adaptive number of records threshold, and t is the current decision time. This represents the current event arrival rate.

[0067] The formula for calculating the time threshold is:

[0068] ,

[0069] Where t2 is the time threshold. To truncate negative values ​​and ensure the waiting time is not negative, let t be the current decision time. The time elapsed since the beginning

[0070] The formula for calculating the byte threshold is:

[0071] ,

[0072] Where t3 is the byte threshold. To truncate negative values ​​and ensure that the waiting time is not negative, let t be the current decision time. For cumulative bytes, Where D is the arrival rate of the current event and D is the size of the current event in bytes.

[0073] By calculating the above thresholds, we can obtain the trough time required to trigger the threshold and the predicted number of events and bytes at the time of shutdown, and provide definite input for subsequent risk and cost calculations.

[0074] R3, based on the threshold calculated above, predicts the number of packet segments by dividing the data into segments according to the Maximum Transmission Unit (MTU), and forms a risk function using the link packet loss rate and jitter term. The formula for the risk function is as follows:

[0075] ,

[0076] Where DA represents the risk score. Here, denoted as the single-segment packet loss rate, DU is the number of segments after segmentation, and E is the weighting coefficient for jitter penalty. Let F be the standard deviation of the round-trip time delay, and let F be the mean of the round-trip time delay. It is the numerical stability constant;

[0077] By constructing a risk function, the quantified packet transmission failure tendency is obtained, thereby suppressing the policy deviation of retransmission timeout caused by packets exceeding the threshold.

[0078] R4, based on the threshold and risk function calculated above, synthesizes a comprehensive cost function, the formula of which is:

[0079] ,

[0080] For the overall cost, The weighting factor for the delay in reimbursement is Dealy, where Dealy is the delay in reimbursement. Here, Lose represents the weighting coefficient for packet loss risk. The weight coefficients for chain overhead are denoted by , and Spend is the chain overhead.

[0081] After constructing the comprehensive cost function, in order to determine which choice to make at that time, such as immediately closing the packet or continuing to wait, a comprehensive cost threshold is set. Compare with the currently calculated comprehensive cost function score, This indicates that the overall cost has exceeded the set threshold, and the aggregation needs to be stopped immediately. The header of the current aggregated credential packet is recorded, the packet-level digest short value of the credential packet is calculated, and it is enqueued. Otherwise, the aggregation continues. This approach achieves an ideal choice between stopping the aggregation immediately and continuing to wait, thereby increasing the aggregation degree in weak networks, shortening the waiting time in good networks, and maintaining continuous authentication and low overhead with minimal leakage.

[0082] The offline pre-training process employs reinforcement learning. It extracts observable states, executed thresholds (t1, t2, t3), and their effects from historical records to form standardized samples. Anomalies and missing items are removed, and the field order and version number are fixed. Continuous policies are trained using historical decisions as supervisory signals, enabling them to output adaptive thresholds (t1, t2, t3) in an offline environment. Boundary and monotonic constraints are applied to the output to obtain usable initial policies. Conservative reinforcement learning is used to estimate the long-term state-action payoff, suppressing overestimation bias in out-of-distribution actions. Combined with policy enhancement using advantage weights, a more robust offline policy to weak network fluctuations is obtained. Soft constraints such as packet loss rate, latency jitter, and packet length limits are introduced during training and selection. Offline evaluation methods are used to set thresholds for timeliness, reliability, and cost; policies failing to meet these thresholds are not deployed to the online stage.

[0083] The historical decision refers to the threshold configuration (t1, t2, t3) actually executed in the historical operation under a given state and its corresponding effect, which is used as a supervision sample for offline training;

[0084] The continuous strategy is a parameterized decision rule that maps states to continuous action vectors and is subject to preset boundaries and monotonic constraints.

[0085] The method employs conservative reinforcement learning to estimate the long-term reward of a state-action sequence. It constructs a state ST based on historical network outage or weak network logs, including link quality, event arrival rate, buffer usage, and packet opening duration, along with samples of actions ACT. A value function V(ST,ACT) is trained according to a comprehensive cost function, and a conservative regularization term is introduced into the objective to suppress overestimation of out-of-distribution actions. This outputs a long-term reward estimate for any V(ST,ACT) used for threshold filtering and policy updates. This approach provides a robust packet closing decision basis under weak network fluctuations, reducing packet loss and latency risks caused by overactive actions and ensuring the policy meets safety boundaries.

[0086] Step S4: Based on the credential packet, during network recovery, only the packet header and packet-level digest of the credential packet are uploaded, and consistency verification and conflict detection are performed with the event log to generate feedback governance and traceability details;

[0087] Specifically, during network recovery, the header (Head) and short packet digest of the uploaded credential packet are used as central anchors to obtain publicly available comparison records with low traffic and low leakage, establishing a unified anchor for subsequent offline retrieval of details and rapid verification. Consistency verification and conflict detection are then performed, including: ensuring integrity by retrieving details from the credential packet as needed, serializing and recalculating the local digest according to specifications, and comparing the reconstructed header (RHead) with the header of the credential packet uploaded during network recovery. If RHead and Head match, the packet is considered complete; otherwise, it is marked as having a digest inconsistency or suspected tampering, and the affected credential packet and its adjacent packets are immediately frozen. For the frozen packets, details are retrieved, the header (RRHead) is reconstructed, and the digest is recalculated, while each event record is verified. Signature and ternary tag For consistency, if the first and last beat scales and local number are the same, and the reconstructed packet header RHead is different from the credential packet header but RRHead and RHead are the same, then consistency is achieved. This indicates that the packet header was corrupted during transmission, triggering a packet header re-retrieval, correcting the packet header according to the detail fields, and recalculating the digest. If RHead is different from the credential packet header and RRHead and RHead are different, then consistency is lost. This indicates that the content has been modified, and the packet is rejected, frozen, and the source end is required to review it. At the same time, the replacement anchor point and reason code are registered.

[0088] Logical consistency checks are performed by verifying whether adjacent packet boundaries are seamlessly connected and whether the local number A is monotonically non-backward. The duplicate content fingerprint TUN and gap intervals are discovered by sorting <packet order, A, T> to detect that adjacent packet boundaries are not seamlessly connected and that the local number A is not monotonically non-backward, and the packet number and event index are located.

[0089] The physical compatibility test is performed to recover the impeller main frequency R based on the scale T. The eigenvector STA is inversely transformed to reconstruct the wind speed and power sequence and verify the signature. The main frequency coherence degree SIM and the main frequency deviation degree are calculated. If min(SIM)≥Q and max(R)≤W, it is judged as passing. Otherwise, it is marked as physically inconsistent. Here, min is to find the minimum value, Q is the main frequency coherence degree threshold, max is to find the maximum value, and W is the main frequency deviation degree threshold.

[0090] The test results are categorized and output. The results of integrity, logical consistency verification and physical matching tests are summarized and output to generate governance and traceability details. This is used to automatically trigger repackaging, retransmission, signature review and replay window correction to obtain auditable disposal instructions and traceable evidence chains. After passing items are recorded and failing items are frozen for rectification, the chain continuity and cross-entity consistent timeline are restored.

[0091] like Figure 2The technical effects of the present invention are compared in the following figures: the black bars represent the technical effects of the present invention, and the gray bars represent the technical effects of the prior art. As can be seen from the figures, the present invention is superior to the prior art.

[0092] Example 2 is a blockchain-based industrial data security transmission method based on Example 1, and is used in conjunction with Example 1. The specific solution is as follows:

[0093] Step 1: Generate the record tag from the instruction issuing side;

[0094] Specifically, when issuing remote commands, the message mes is obtained by serializing the command minimum digest with the current tick scale T, local issuance number A, issuance side identifier, and version in a fixed field order according to the CBOR specification, and then calculating the content fingerprint to obtain the issuance record meta tag. It obtains a publicly verifiable and irreversible instruction identifier, which can prevent forgery of the source and parameter replacement, and provides a unique anchor point for subsequent closed-loop reconciliation.

[0095] Step 2: Generate ternary labels from the feedback test;

[0096] Specifically, after the execution side receives the instruction, it will send the received record metadata tag. The original data is copied back to the receipt; simultaneously, the local tick mark T and local number A at the moment of execution are read, the minimum summary of the execution result is serialized using CBOR and the result fingerprint Rmes is calculated, thereby generating a ternary tag for the receipt. This process generates a receipt body, resulting in a closed-loop record of sending, receiving, and executing the same result fingerprint. This allows for the determination of whether the data was received, when it was executed, and the execution result without the need for a unified clock.

[0097] Step 3: Align the summaries on the chain across subject anchor points;

[0098] Specifically, the issuing side and the executing side respectively load the issuing record and the execution receipt as event entries into their respective credential packages during the network outage phase (see S3 of Example 1), and only upload the packet header and packet-level digest of the credential package when the network is restored (see S4 of Example 1), thereby obtaining a bidirectional, cross-subject consistent anchor point, which can complete the closed-loop evidence registration in a low-traffic, low-leakage manner.

[0099] Step 4: Perform closed-loop consistency verification and output the results by category;

[0100] Specifically, retrieve the package details as needed and perform four verification steps:

[0101] (1) Integrity: Locally recalculated Lpack is compared with the on-chain CLpack; if Lpack ≠ CLpack but the Rpack recalculated with the detailed reconstructed Rhead is consistent with Lpack, it is determined that the packet header transmission is corrupted and is re-corrected;

[0102] (2) Logical consistency: Construct a total order on both sides using <package order, A, T>, and check that the content fingerprint of the same event exists on both sides and that A is monotonically non-backtracking;

[0103] (3) Physical matching: If the receipt contains environmental feature EN, the impeller main frequency ER is restored according to T, the EN is inversely transformed and reconstructed and the signature is checked. The main frequency coherence and frequency deviation are calculated. If min(EN)≥Q and max(ER)≤W, it is judged as passing. Otherwise, it is marked as physically inconsistent. Where min is to find the minimum value, Q is the main frequency coherence threshold, max is to find the maximum value, and W is the main frequency deviation threshold. (4) Classification output: The verification results are summarized into governance and traceability details, so as to obtain the closed-loop status and auditable evidence chain. It can quickly determine whether it is timely, whether it has been tampered with, and whether it has been executed. After the passed items are recorded and the failed items are frozen and rectified, the chain-like continuous and cross-entity consistent timeline is restored.

[0104] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based method for secure transmission of industrial data, characterized in that: Step S1: Collect the fan impeller speed and divide it into equal parts of the beat scale, and initialize the local number on the edge side; Step S2: Based on the beat scale and local number, record each alarm event and calculate the content fingerprint, bind the current beat scale and local number, record the current environmental summary value, and generate an event record; Step S3: Based on the alarm event, during the network outage phase, the packaging side packages multiple events into a voucher package according to the start and end beat scale and local number, and attaches the number of packaged items and the start and end beat scale and local number. Step S4: Based on the credential packet, during network recovery, only the packet header and packet-level digest of the credential packet are uploaded, and consistency verification and conflict detection are performed with the event log to generate feedback governance and traceability details.

2. The blockchain-based industrial data secure transmission method according to claim 1, characterized in that: The specific steps for collecting the fan impeller speed and dividing it into equal intervals on the beat scale, and initializing the local number on the edge side are as follows: The impeller speed is acquired by reading the A / B phase pulses of the incremental encoder on the main shaft and using the Z phase zero position as the reference per revolution. The number of pulses per revolution is used to calculate the mechanical angular velocity and angle. The mechanical angular velocity is quantized into equal-division clock cycles T according to the Circle per impeller revolution. The time-incrementing number ADD is initialized on the edge side, with the initial increment set to ADD0. After each event is generated, ADD0 is incremented by 1, and rollback is allowed but unrollback is prohibited, thus obtaining the initialized local number.

3. The blockchain-based industrial data secure transmission method according to claim 1, characterized in that: The specific steps for recording each alarm event and calculating the content fingerprint are as follows: Calculate the content fingerprint TUN for the generated alarm event, and bind it to the current tick scale T and local number A of the current alarm event to obtain a ternary label. .

4. The blockchain-based industrial data secure transmission method according to claim 3, characterized in that: The specific steps for recording each alarm event and calculating the content fingerprint also include: Record the brief environmental values ​​of the current alarm event. Using the beat scale T as the phase reference, perform phase alignment and resampling of the wind speed and output power before the alarm event, and extract a scalarized feature vector STA from the scalar sequence according to a fixed dimension to obtain the brief environmental values ​​used to characterize the statistical features of the environmental measurements. Generate an independent signature and store it together with the ternary record to generate the event record. .

5. The blockchain-based industrial data secure transmission method according to claim 1, characterized in that: The specific steps for the packaging side to package multiple events into a voucher package according to the start and end time scale and local number during the network outage phase are as follows: The packetizing side packages the alarm events generated during the network outage phase into credential packets. During the packaging process, a comprehensive cost function is constructed, and thresholds for the number of entries t1, time t2, and bytes t3 are adaptively set. When any threshold is reached, the current packet is closed and the number of events in the previous packet summary, the first and last beat scale, and the local number are written. After calculating the packet summary, the packet is enqueued and awaits uploading. Reinforcement learning is used for offline pre-training to obtain a packet-level continuity and low overhead that can be maintained under different link conditions.

6. The blockchain-based industrial data secure transmission method according to claim 5, characterized in that: The specific steps for the packaging side to package multiple events into a voucher package according to the start and end timescale and local number during the network disconnection phase also include: The steps for defining the comprehensive cost function are as follows: R1 collects link metrics, event arrival rate, and current packet status to obtain the input for cost calculation; R2 calculates the arrival times of the number of records threshold t1, time threshold t2, and byte threshold t3 based on the input volume. The formula for calculating the number of records threshold is: , Where t1 is the threshold for the number of items. To truncate negative values, B is the adaptive number of records threshold, and t is the current decision time. This represents the current event arrival rate. The formula for calculating the time threshold is: ， Where t2 is the time threshold. To truncate negative values, t is the current decision time. The time elapsed since the beginning; The formula for calculating the byte threshold is: ， Where t3 is the byte threshold. To truncate negative values, t represents the current decision time. For cumulative bytes, Where D is the arrival rate of the current event and D is the size of the current event in bytes. By calculating all the above thresholds, the trough time required to trigger the threshold and the predicted number of events and bytes at the time of shutdown are obtained, providing a definite input for subsequent risk and cost calculations. R3, based on all the above thresholds, predicts the number of packet segments by dividing the data into segments according to the Maximum Transmission Unit (MTU), and forms a risk function using the link packet loss rate and jitter term. The formula for the risk function is as follows: ， Where DA represents the risk score. Here, denoted as the single-segment packet loss rate, DU is the number of segments after segmentation, and E is the weighting coefficient for jitter penalty. Let F be the standard deviation of the round-trip time delay, and let F be the mean of the round-trip time delay. It is the numerical stability constant; By constructing a risk function, the quantified packet transmission failure tendency is obtained; R4, based on all the above thresholds and risk functions, synthesizes a comprehensive cost function, the formula of which is: ， For the overall cost, The weighting factor for the delay in reimbursement is Dealy, where Dealy is the delay in reimbursement. Here, Lose represents the weighting coefficient for packet loss risk. is the weighting coefficient for chain overhead, and Spend is the chain overhead.

7. The blockchain-based industrial data secure transmission method according to claim 5, characterized in that: The specific steps for the packaging side to package multiple events into a voucher package according to the start and end timescale and local number during the network disconnection phase also include: Offline pre-training using reinforcement learning is employed. Standardized samples are formed by extracting observable states, executed thresholds (t1, t2, t3), and their effects from historical records, removing outliers and missing items, and solidifying field order and version numbers. Continuous policies are trained using historical decisions as supervisory signals, enabling them to output adaptive thresholds (t1, t2, t3) in an offline environment. Boundary and monotonic constraints are applied to the output to obtain usable initial policies. Conservative reinforcement learning is used to estimate the long-term state-action payoff, suppressing overestimation bias in out-of-distribution actions. Combined with policy enhancement using advantage weights, a more robust offline policy to weak network fluctuations is obtained. Soft constraints such as packet loss rate, latency jitter, and packet length limits are introduced during training and selection. Offline evaluation methods are used to set thresholds for timeliness, reliability, and cost; policies failing to meet these thresholds are not deployed to the online stage.

8. The blockchain-based industrial data secure transmission method according to claim 1, characterized in that: The specific steps for uploading only the packet header and packet-level digest of the credential packet during network recovery, and performing consistency verification and conflict detection with the event log, are as follows: During network recovery, the header (Head) and short packet digest of the uploaded credential packet are used as central anchors to obtain publicly available comparison records with low traffic and low leakage. Subsequently, consistency verification and conflict detection are performed, including: integrity checks, retrieving details from the credential packet as needed, serializing the locally recalculated digest according to specifications, and comparing the reconstructed header (RHead) with the header of the credential packet uploaded during network recovery. If RHead and Head match, the packet is considered complete; otherwise, it is marked as having a digest inconsistency or suspected tampering, and the affected credential packet and its adjacent packets are immediately frozen. For the frozen packet, details are retrieved, the header (RRHead) is reconstructed, and the digest is recalculated, while each event record is verified. Signature and ternary tag Consistency.

9. A blockchain-based method for secure transmission of industrial data according to claim 8, characterized in that: The specific steps for uploading only the packet header and packet-level digest of the credential packet during network recovery, and performing consistency verification and conflict detection with the event log, also include: By verifying whether adjacent packet boundaries are seamlessly connected and whether the local number A is monotonically non-backtracking, and sorting by <packet order, A, T> to remove duplicates, it was found that adjacent packet boundaries were not seamlessly connected and the local number A was not monotonically backtracking, and the packet number and event index were located. A physical matching test was performed to recover the impeller main frequency R based on the scale T. The feature vector STA was inversely transformed to reconstruct the wind speed and power sequence and verify the signature. The main frequency coherence degree SIM and the main frequency deviation degree were calculated and judged.

10. A blockchain-based method for secure transmission of industrial data according to claim 1, characterized in that: The specific steps for generating feedback governance and tracing details are as follows: The test results are categorized and output. The results of integrity, logical consistency verification and physical matching tests are summarized and output to generate governance and traceability details. This is used to automatically trigger repackaging, retransmission, signature review and replay window correction to obtain auditable disposal instructions and traceable evidence chains. After passing items are recorded and failing items are frozen for rectification, the chain continuity and cross-entity consistent timeline are restored.

Images