Plastic business data redundancy disaster recovery system
By introducing information entropy auditing and security consensus verification mechanisms into the plastic business data redundancy disaster recovery system, the problems of data logic tampering and recovery time lag in the existing technology are solved, and real-time data reconstruction and state consistency are realized in a heterogeneous network environment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHAANXI WANGJIN ENERGY TECHNOLOGY CO LTD
- Filing Date
- 2026-05-09
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies lack in-situ verification mechanisms for the intrinsic attributes of data when facing covert threats such as slow drift poisoning in plastic industrial production. This leads to logical distortions in disaster recovery nodes after they take over operations, and increasing the verification strength results in delayed recovery time, failing to meet the real-time requirements of continuous industrial production.
By employing a heterogeneous data offloading module, a high-frequency time-series synchronization module, a security consensus verification module, and an information entropy audit module, the distribution characteristics of differential data sequences are identified through probability features within a statistical sliding data window. Consistency verification credentials are generated to ensure the legality and time-series alignment of the data, thereby reconstructing the business status of the disaster recovery terminal.
It enables the identification and protection against logical tampering in non-ideal network environments, ensuring the state integrity and real-time performance of disaster recovery nodes when taking over production services, and reducing system computing power redundancy and communication bandwidth consumption.
Smart Images
Figure CN122437703A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of data integrity disaster recovery technology, and in particular relates to a data redundancy disaster recovery system for plastics business. Background Technology
[0002] Current mainstream data backup methods employ a heterogeneous traffic splitting architecture. The system divides production operations into high-frequency sensor condition streams and low-frequency recipe instruction streams based on data attributes, and establishes logically isolated redundant transmission links. To meet the real-time synchronization requirements of the production site, the high-frequency channel typically uses differential push and timestamp-based playback mechanisms. However, because the high-frequency disaster recovery channel is designed to prioritize low-latency transmission, the system's real-time auditing capabilities for payload content are insufficient. Industrial sensor data exhibits strong inertia during physical evolution, manifesting as a low-disorder distribution characteristic of differential sequences. When subjected to covert threats such as slow drift poisoning, attackers inject small-amplitude malicious offsets into the time-series stream. Due to the lack of in-situ verification mechanisms for the intrinsic attributes of the data in existing technologies, the poisoned data fragments directly enter the disaster recovery aggregation node and are persisted. When the primary production node fails and the disaster recovery node takes over the business, the reconstructed full business logic state undergoes topological distortion, causing the production node logic to fail.
[0003] Increasing verification strength to ensure integrity results in significant data lag, making it impossible to meet the recovery time target for continuous industrial production. Relying solely on boundary protection measures fails to identify semantic tampering within industrial protocols. However, existing redundancy architectures often focus on physical link connectivity and hardware node redundancy. Even with optimization of the physical layout of extrusion equipment units or sensor wiring to reduce environmental interference, if only hardware physical redundancy is addressed, software-level control methods and data auditing logic still have security blind spots. For example, Chinese invention patent CN106919473B discloses a data disaster recovery system and business processing method that deploys read... The underlying logic of this technology, which uses a database and a snapshot database to compensate for the state recovery when the main database fails, is based on the ideal assumption that synchronous data is trustworthy data. When dealing with hidden logical tampering such as slow drift poisoning commonly seen in plastic industrial production, the read database and the snapshot database will synchronously receive and solidify the poisoned differential data fragments. However, because this existing technology lacks an in-situ verification mechanism for the intrinsic distribution characteristics of the data, the disaster recovery node will still follow the distorted production logic after taking over the business, causing the production node logic to fail or even causing equipment damage. If integrity is guaranteed only by increasing the verification strength, there will be a significant transmission lag, making it impossible to meet the real-time requirements of continuous industrial production.
[0004] Therefore, the technical problem to be solved by this invention is how to provide a heterogeneous data flow scheme that can identify logical tampering without causing transmission delay, and achieve strong consistency alignment of states in non-ideal network environments. Summary of the Invention
[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A plastic business data redundancy disaster recovery system, the system comprising a heterogeneous data diversion module, a high-frequency time-series synchronization module, a security consensus verification module, an information entropy audit module, and a business state reconstruction module:
[0006] The heterogeneous data diversion module is used to process and classify the acquired business data messages, distribute the differential data sequence in the business data message to the high-frequency time synchronization module, or divert the production formula data in the business data message to the security consensus verification module.
[0007] The information entropy audit module is used to acquire the differential data sequence transmitted by the high-frequency timing synchronization module and maintain a sliding data window with a fixed byte length in memory. By statistically analyzing the probability characteristics of the numerical distribution within the sliding data window, the sliding entropy characteristic value, which characterizes the distribution characteristics of the differential data sequence, is determined. The information entropy audit module is also used to send a verification command to the security consensus verification module when the sliding entropy characteristic value exceeds a preset security threshold.
[0008] The security consensus verification module is used to respond to verification commands by performing hash operations on the differential data sequence to generate a consistency verification certificate, and then transmit the consistency verification certificate and the global time-series beacon to the business state reconstruction module.
[0009] The business status reconstruction module is used to filter the legality of differential data sequence processing based on consistency verification credentials, and to reorganize the filtered data in conjunction with global time-series beacons to generate the business reconstruction status of the disaster recovery end.
[0010] Preferably, when the information entropy audit module processes the audit logic of the differential data sequence, it includes the following sub-steps: Step S11, collect the differential values entering the sliding data window and determine the statistical frequency of each value interval; Step S12, calculate the corresponding distribution probability based on the statistical frequency and use the distribution probability to determine the sliding entropy feature value; Step S13, monitor the offset gradient of the sliding entropy feature value in the time domain and update the safety threshold when the offset gradient reaches the preset offset step size.
[0011] Preferably, the information entropy audit module further includes a dynamic baseline calibration unit; the dynamic baseline calibration unit is used to obtain historical entropy samples of the business status reconstruction module within a preset operating cycle, and dynamically correct the safety threshold processing by calculating the variance envelope of the historical entropy samples.
[0012] Preferably, the security consensus verification module includes a fingerprint calculation unit and a credential encapsulation unit; the fingerprint calculation unit is used to perform a one-way hash operation on the differential data segments in the sliding data window after receiving the verification command, and extract a globally consistent fingerprint to characterize the data integrity.
[0013] Preferably, the credential encapsulation unit is used to associate and encapsulate the global consistency fingerprint with the current business recipe version number to generate a consistency verification credential.
[0014] Preferably, the high-frequency timing synchronization module adopts a unidirectional push logic architecture based on the UDP protocol to carry the timing condition differential sequence generated by the plastic production equipment.
[0015] Preferably, the business state reconstruction module includes a beacon alignment unit; the beacon alignment unit is used to parse global time-series beacons and, based on the logical timestamps carried by the global time-series beacons, to perform time-series rearrangement of the out-of-order differential data sequences caused by network transmission jitter.
[0016] Preferably, the business state reconstruction module is also used to directly map the differential data sequence to the memory image when the sliding entropy feature value is lower than the safety threshold, so as to complete the business state generation in the no-verification mode and maintain the response speed of the data link within 1ms.
[0017] Preferably, the heterogeneous data diversion module is also used to identify the header file identifier of the business data message, and when the header file identifier points to static production parameters, it directs the business data message to the security consensus verification module for processing and encrypted storage.
[0018] Compared with existing technologies, the plastic business data redundancy disaster recovery system of the present invention has the following advantages:
[0019] 1. In the redundancy disaster recovery of plastic business data, by setting up an information entropy audit module at the disaster recovery aggregation end, the low disorder characteristics of plastic production operation data during the smooth evolution process are used to identify differential offset anomalies generated in high-frequency incremental flow channels. This mechanism causes the slow drift poisoning behavior hidden under the characteristics of legitimate traffic to cause a sudden change in data entropy value, and triggers logical blocking before the differential sequence is replayed to local storage, eliminating the risk of logical distortion of the reconstruction state at the disaster recovery end. Since the audit process directly acts on the intrinsic statistical attributes of the data rather than relying on external signatures, the system realizes in-situ identification of unauthorized tampering, ensuring the state integrity when the disaster recovery node takes over the production business.
[0020] 2. Construct an asymmetric cross-validation architecture, which enables the system to maintain only statistical-level computational overhead during the regular transmission of high-frequency differential data streams. Only when the sliding Shannon entropy exceeds the preset security baseline is the hash operation resource of the low-frequency security consensus channel invoked on demand. This cross-channel dynamic collaborative logic breaks the linear constraint between strong consistency verification and extreme transmission timeliness. Without occupying additional communication bandwidth, it provides a verification dimension for massive high-frequency time-series data based on the same dimension as confidential recipe data. This solution effectively reduces the global computing power redundancy of the system while ensuring the real-time transmission of wide area networks.
[0021] 3. By using the low-frequency security consensus channel to inject a global logical clock beacon into the parallel high-frequency channel when generating verification blocks, the forced timing alignment of data packets on the heterogeneous transmission link is achieved. This logical clock anchor mechanism eliminates the arrival time drift between the differential operating condition sequence and the encrypted recipe instruction caused by asymmetric network jitter, ensuring that the disaster recovery aggregation node has strict business causal consistency when performing merge reconstruction. Through the constraint of clock beacon connection to release instruction, the system avoids the misalignment of recipe update and operating condition evolution in time topology from the bottom layer, ensuring the authenticity of the reconstructed data in physical logic. Attached Figure Description
[0022] Figure 1 This is a flowchart of the plastic business data diversion and disaster recovery status reconstruction process of the present invention;
[0023] Figure 2 This is a diagram illustrating the core logic and functional mechanism architecture of the plastic disaster recovery system of this invention. Detailed Implementation
[0024] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.
[0025] A data redundancy disaster recovery system for plastics business includes a heterogeneous data offloading module, a high-frequency time-series synchronization module, a security consensus verification module, an information entropy auditing module, and a business state reconstruction module.
[0026] The heterogeneous data diversion module is used to process and classify the acquired business data messages, distribute the differential data sequence in the business data message to the high-frequency time synchronization module, or divert the production formula data in the business data message to the security consensus verification module.
[0027] The information entropy audit module is used to acquire the differential data sequence transmitted by the high-frequency timing synchronization module and maintain a sliding data window with a fixed byte length in memory. By statistically analyzing the probability characteristics of the numerical distribution within the sliding data window, the sliding entropy characteristic value, which characterizes the distribution characteristics of the differential data sequence, is determined. The information entropy audit module is also used to send a verification command to the security consensus verification module when the sliding entropy characteristic value exceeds a preset security threshold.
[0028] The security consensus verification module is used to respond to verification commands by performing hash operations on the differential data sequence to generate a consistency verification certificate, and then transmit the consistency verification certificate and the global time-series beacon to the business state reconstruction module.
[0029] The business status reconstruction module is used to filter the legality of differential data sequence processing based on consistency verification credentials, and to reorganize the filtered data in conjunction with global time-series beacons to generate the business reconstruction status of the disaster recovery end.
[0030] Preferably, when the information entropy audit module processes the audit logic of the differential data sequence, it includes the following sub-steps: Step S11, collect the differential values entering the sliding data window and determine the statistical frequency of each value interval; Step S12, calculate the corresponding distribution probability based on the statistical frequency and use the distribution probability to determine the sliding entropy feature value; Step S13, monitor the offset gradient of the sliding entropy feature value in the time domain and update the safety threshold when the offset gradient reaches the preset offset step size.
[0031] Preferably, the information entropy audit module uses the following calculation formula when determining the sliding entropy characteristic value: ,in, For a range of values, This represents the probability distribution parameter of the difference values within the corresponding numerical interval. The eigenvalues of the sliding entropy are defined.
[0032] Preferably, the information entropy audit module further includes a dynamic baseline calibration unit; the dynamic baseline calibration unit is used to obtain historical entropy samples of the business status reconstruction module within a preset operating cycle, and dynamically correct the safety threshold processing by calculating the variance envelope of the historical entropy samples.
[0033] Preferably, the security consensus verification module includes a fingerprint calculation unit and a credential encapsulation unit; the fingerprint calculation unit is used to perform a one-way hash operation on the differential data segments in the sliding data window after receiving the verification command, and extract a globally consistent fingerprint to characterize the data integrity.
[0034] Preferably, the credential encapsulation unit is used to associate and encapsulate the global consistency fingerprint with the current business recipe version number to generate a consistency verification credential.
[0035] Preferably, the high-frequency timing synchronization module adopts a unidirectional push logic architecture based on the UDP protocol to carry the timing condition differential sequence generated by the plastic production equipment.
[0036] Preferably, the business state reconstruction module includes a beacon alignment unit; the beacon alignment unit is used to parse global time-series beacons and, based on the logical timestamps carried by the global time-series beacons, to perform time-series rearrangement of the out-of-order differential data sequences caused by network transmission jitter.
[0037] Preferably, the business state reconstruction module is also used to directly map the differential data sequence to the memory image when the sliding entropy feature value is lower than the safety threshold, so as to complete the business state generation in the no-verification mode and maintain the response speed of the data link within 1ms.
[0038] Preferably, the heterogeneous data diversion module is also used to identify the header file identifier of the business data message, and when the header file identifier points to static production parameters, it directs the business data message to the security consensus verification module for processing and encrypted storage.
[0039] Example 1: In a wide area network (WAN) disaster recovery environment operating a continuous extrusion equipment, the communication link between the main production node and the disaster recovery center faces network jitter and numerical offset injection. Business data packets contain low-frequency update production formula data and high-frequency time-series operating condition sensor data. The full-volume strong consistency disaster recovery framework causes WAN bandwidth congestion when synchronizing high-frequency operating condition data. The non-checked asynchronous push scheme cannot identify numerical offsets that are continuously injected into the differential data sequence without damaging the communication protocol infrastructure, leading to data logic distortion in the business reconstruction state when the disaster recovery node takes over. The heterogeneous data diversion module processes and classifies the acquired business data packets, separating the data from the data within the packets. The production formula data is diverted to the security consensus verification module, while the differential data sequence in the business data message is distributed to the high-frequency timing synchronization module. The high-frequency timing synchronization module transmits the differential data sequence to the information entropy audit module. The information entropy audit module maintains a sliding data window with a fixed byte length in memory and collects the differential values entering the sliding data window. The receiving end divides the data into a ring delay buffer queue based on first-in-first-out logic. The buffer queue data residence period is used to physically smooth out the message arrival time difference between the high-frequency synchronization link and the low-frequency consensus verification link. For the initial sampling sequence with continuous floating-point characteristics, the information entropy audit module applies a quantization procedure based on the signal quantization noise theory.
[0040] Under absolute no-load conditions, a preset periodic background time series is acquired and the background variance is calculated. The quantization step size used for probabilistic discretization is anchored to a constant multiple of the background variance. Based on this quantization step size, continuous input floating-point values are mapped and assigned to corresponding discretization state intervals. The statistical frequency of each independent difference value interval within the sliding data window is determined, and the statistical frequency is mapped to a probability distribution. The probability distribution is used to calculate and determine the sliding entropy characteristic value representing the distribution characteristics of the difference data sequence. The formula for calculating the sliding Shannon entropy is as follows: ,in, Characterizing independent difference value intervals, To characterize the distribution probability of the difference value interval, and to overcome the physical limitations of high-frequency floating-point operations on system state update latency, the information entropy audit module has a pre-calculated fixed-point logarithm lookup table with a base of 2 embedded in its memory. After statistically analyzing and normalizing the distribution probability of each independent interval, the system uses bit operations to directly convert the floating-point probability variable into an integer physical index address in the lookup table, extracts the preset fixed-point logarithm operation result, and performs basic bit-level accumulation operations only in the central processing unit. This eliminates the time-consuming logarithmic floating-point calculation instruction cycle, and from the underlying hardware computing power allocation mechanism, ensures that the high-order statistical feature calculation action of a single sliding data window is completed in a closed loop within a sub-millisecond physical time scale. The information entropy audit module monitors the numerical changes of the sliding entropy feature value.
[0041] When the sliding entropy characteristic value exceeds a pre-set security threshold, the information entropy audit module sends a verification command to the security consensus verification module. The security consensus verification module responds to the verification command by performing a hash operation on the differential data sequence to generate a consistency verification credential. This credential, along with a global time-series beacon, is then transmitted to the business state reconstruction module. The business state reconstruction module filters the differential data sequence processing based on the consistency verification credential and, combined with the global time-series beacon, reassembles the filtered data to generate the business reconstruction state for the disaster recovery end. During this reassembly, the complex hash calculation process causes the generation of the consistency verification credential and its accompanying beacon to lag behind the arrival time of the suspected high-frequency differential sequence itself at the reconstruction module. Therefore, the system uses a pre-set circular delay buffer queue to forcibly suspend the suspected differential sequence by clock damping, ensuring it arrives at the reconstruction module first. The corresponding high-frequency message at the module entry point remains silently in the queue until the global timing beacon generated by the subsequent calculation arrives at the business node and the dual-track clock merging and verification are completed. Only then is the differential sequence released to perform subsequent reassembly mapping. This queue-based dwelling mechanism physically bridges and repairs the causal inversion fault in the order of data operations. When the sliding entropy characteristic value does not exceed the safety threshold, the system exempts the differential data sequence from hash operation. The business state reconstruction module directly distributes and reassembles the data. The information entropy audit module is set as the logical condition for triggering verification. The system only starts the security consensus verification module when the probability characteristics of the numerical distribution change. Hash operation resources are scheduled according to the sliding entropy characteristic value to resolve the contradiction between the real-time transmission requirements and the computational power consumption for anti-tampering verification when synchronizing massive high-frequency working data. Under the premise of ensuring real-time transmission, the system enhances the state consistency and logical integrity of the disaster recovery end.
[0042] Example 2: In the cross-regional disaster recovery synchronization of industrial control networks, a time-series simulation platform is built based on a computational fluid dynamics simulation model. Differential data sequences are output from sensor nodes with a sampling rate of 10000Hz. Determining the fixed byte length of the sliding data window in memory requires balancing the statistical significance of the data probability distribution with the timeliness of the abrupt response. When the power spectral density of the ambient white noise increases, a small window length causes high-frequency disorder jitter to mask the true numerical deviation. The system expands the window length to accumulate statistical samples. In a test environment with a superimposed high-frequency random disturbance signal with a signal-to-noise ratio of 15dB, the system sets the fixed length of the sliding data window to 1024 bytes.
[0043] The experiment was set up with a control group using full hash synchronization logic and an experimental group using information entropy auditing logic. Background differential data sequences were continuously input into two wide area network transmission links, and non-random numerical offsets with progressively increasing amplitudes were continuously injected at preset time nodes. The information entropy auditing module in the experimental group extracted the differential values entering the sliding data window and determined the intervals of each independent differential value. The statistical frequency and its mapping to the probability distribution Calculate the sliding entropy eigenvalues When observing the original input data sequence, if the offset amplitude is less than 0.4%, the absolute value exceeding the limit detection mechanism cannot extract the abnormal state due to the 15dB background noise masking the abnormal state. The sliding entropy feature value output by the experimental group... It exhibits nonlinear response characteristics, with the injected offset amplitude increasing from 0.1% to 0.5%, and the sliding entropy eigenvalue... Maintaining a baseline fluctuation range of 3.21 to 3.28, when the offset amplitude exceeds the critical point of 0.65%, continuous unidirectional offsets disrupt the random distribution law of the original physical process, and the sliding entropy eigenvalue... It rose to 5.87 within 2.4ms.
[0044] Based on the aforementioned nonlinear rising critical point, the system establishes a safety threshold of 4.50. In this simulation and verification procedure, the system follows a set baseline multiple quantization step size, passively truncating all continuous floating-point states and distributing them into 256 uniformly divided independent discrete physical state intervals. Therefore, the theoretical limit of the system's Shannon entropy objectively converges to 8.0. In the normal operating trajectory, the inertia of the extrusion equipment's rotating structure induces strong probability density aggregation, lowering the baseline entropy value and confining it to the aforementioned low-chaos fluctuation band. Attack injection behaviors exceeding the tolerance range deconstruct the physical aggregation constraint, causing the access frequency of each state interval to passively diverge towards equalization, resulting in an irreversible steep increase in the sliding entropy towards the theoretical upper limit of 8.0 after breaking the critical point. When the sliding entropy eigenvalue... When the value is greater than 4.50, the security consensus verification module is triggered to extract data segments within the current sliding data window, perform hash calculations, and generate consistency verification credentials. In a 72-hour continuous high-frequency data stream injection test, the control group performed hash calculations on all data frames, resulting in an average latency of 14.2ms for WAN disaster recovery synchronization, accompanied by bandwidth congestion. The test group, in the normal transmission phase without intervention, eliminated hash calculations, and the average synchronization latency remained at 0.85ms, only decreasing when the sliding entropy characteristic value was reached. The instantaneous consumption of computing power greater than 4.50 generates consistency verification credentials. While maintaining sub-millisecond disaster recovery response speed, the test group achieved an interception rate of 99.4% against numerical offset injection. The technical contradiction between the real-time synchronization of massive time-series data and the consumption of verification computing power is balanced by the information entropy perception mechanism.
[0045] Example 3: In the wide area network disaster recovery synchronization operation of the plastic extrusion process, the environmental electromagnetic interference fluctuates suddenly due to the periodic start and stop of large power equipment, causing dynamic drift of the baseline information entropy of the differential data sequence. Using a fixed security threshold to evaluate the sliding entropy characteristic value causes invalid consumption of hash computing power or missed numerical offsets. The heterogeneous data diversion module parses the protocol header identifier of the business data packet and compares the protocol header identifier with the pre-stored static attribute feature code. The heterogeneous data diversion module diverts the business data packet that matches the static attribute feature code as production formula data to the security consensus verification module. It strips the protocol header of the unmatched packet to generate a differential data sequence and distributes it to the high-frequency time-series synchronization module. The high-frequency time-series synchronization module transmits the differential data sequence to the information entropy auditing module. The information entropy auditing module allocates a calibration time window covering 5000 discrete sampling points during the system startup phase and continuously collects the sliding entropy characteristic value within the calibration time window. The information entropy auditing module calculates the expected mean based on the collected feature value sequence. with standard deviation The information entropy audit module is based on the formula Determine dynamic security thresholds ,in, Refers to the dynamic security threshold. Refers to the eigenvalues of the sliding entropy The expected mean, Refers to the eigenvalues of the sliding entropy standard deviation This refers to the dimensionless tolerance coefficient, with a value of 3.
[0046] In the parallel logic for calculating the aforementioned standard deviation, to accurately capture the nonlinear fluctuations of the baseline, the information entropy audit module uses the sliding extreme value method to extract the local variance extreme points of historical samples in the time domain within a preset operating period. It then uses a cubic spline interpolation function to numerically smooth adjacent discrete variance extreme points, constructing a variance envelope covering the entire time axis. This serves as the physical benchmark for quantifying the gradual change trend of random thermal noise in the environment. The information entropy audit module refreshes the sample sequence within the calibration time window at 300-second intervals, recalculating and covering the dynamic safety threshold of the previous period. Before performing the overwrite operation, the information entropy audit module retrieves the absolute fixed mean value recorded in memory during the system's initial power-on or maintenance calibration state, and calculates the current expected mean value. If the relative offset between the current sample sequence and the absolute fixed mean exceeds the preset historical tolerance constraint limit, the information entropy audit module discards the current sample sequence and updates the calculation result, forcibly changing the expected mean. Anchoring and reverting to the absolutely fixed mean blocks the passive rise of the adaptive baseline caused by extremely slow and continuous numerical injection; when the sliding entropy eigenvalue Greater than the dynamic security threshold of the current period At that time, the information entropy audit module transmits a verification instruction carrying an abnormal timestamp to the security consensus verification module. The security consensus verification module extracts the data segment to be verified from the differential data sequence based on the abnormal timestamp, performs hash calculation and generates a consistency verification certificate. The system relies on the calibration time window to quantify the distribution characteristics of environmental background noise, and adaptively calibrates the judgment benchmark based on mathematical statistical parameters to suppress the interference of the underlying physical environment fluctuation on the information entropy audit mechanism, and alleviate the technical contradiction between data anti-tampering accuracy and computing power consumption under dynamic network conditions.
[0047] Example 4: In an industrial field deployment scenario where a plastic business data redundancy disaster recovery system is connected to a new heterogeneous IoT architecture, the heterogeneous configuration of physical devices leads to an unknown state in the protocol header distribution of business data packets. The system initiates a pre-calibration procedure in an offline isolation environment to establish static attribute feature codes and initial security benchmarks. The heterogeneous data diversion module collects the baseline packet stream of the extrusion equipment in a clean physical local area network with the external wide area network disconnected, extracts fixed-length protocol header fields from the baseline packet stream, and counts the frequency of occurrence of various protocol header fields in 100,000 consecutive packet samples. Protocol header fields with a frequency lower than the baseline normal lower limit are extracted as static attribute feature codes pointing to production formula data, and the static attribute feature codes are solidified into the mapping rule table in the system memory.
[0048] The high-frequency timing synchronization module, based on the solidified static attribute feature code, diverts the baseline message stream and injects a differential data sequence free from external interference into the information entropy audit module. Within a controlled calibration period without tampering, the information entropy audit module continuously reads the differential values entering the sliding data window and calculates the reference sliding entropy characteristic value. According to the formula The initial safety baseline for calibrating the system for online operation, among which, Refers to the reference sliding entropy eigenvalue. Refers to the range of independent differential values within the calibration period. This refers to the distribution probability of independent differential numerical intervals in a controlled physical environment. The information entropy audit module locks the initial security benchmark and sets it as the calculation benchmark point for deriving dynamic security thresholds when the system switches to wide area network disaster recovery status. The system eliminates the interference of unknown initial hardware noise or latent intervention signals in the startup period of the anti-tampering audit mechanism from unknown industrial site initial hardware noise or latent intervention signals. In the early stage of cross-platform deployment, it establishes a disaster recovery synchronization starting point with a definite physical source and data extraction basis.
[0049] Example 5: In a system deployment scenario adapting to heterogeneous network bandwidth and sampling rate of extrusion equipment, the length of the fixed sliding data window is limited by the physical constraints of sudden change response delay and background noise suppression. The information entropy audit module acquires 10,000 baseline differential data samples transmitted by the high-frequency timing synchronization module, and truncates the baseline differential data samples according to the incremental byte length gradient to construct candidate test windows. The standard deviation and information entropy volatility of the numerical distribution within each candidate test window are calculated respectively, and a discrete mapping relationship is established with the window length as the input feature and the information entropy volatility as the output parameter. When the standard deviation decays to the preset noise floor range and the first derivative of the information entropy volatility jumps to the predetermined zero threshold interval, the information entropy audit module extracts the corresponding byte length value and writes it into the memory register, and determines the corresponding byte length value as the fixed byte length of the sliding data window. ,in, The sample set size used to calculate the probability distribution is characterized. In the actual logic solution module, the physical definition of information entropy volatility is the ratio of the absolute value of the backward difference of the entropy values measured in two adjacent sliding test windows to the entropy value of the preceding test window. Since discrete sample features do not have the analytical differentiability of continuous functions, their so-called first derivative jump feature is specifically equivalent to discrete difference quotient operation. When the results of five consecutive difference quotient calculations of the information entropy volatility sequence itself all fall within the predetermined minimum constant zero threshold interval of 10 to the power of negative fourth, it is determined that the statistical variation of the system approaches the steady-state convergence point.
[0050] When the system faces a situation where link congestion causes discrete loss of differential data sequences transmitted by the high-frequency timing synchronization module, the service state reconstruction module intercepts the state distortion caused by missing data based on the consistency verification certificate. The beacon alignment unit extracts the logical timestamp and sequence index number carried in the global timing beacon, compares the sequence index number increments of adjacent messages in the receiving queue, and when the difference between adjacent sequence index numbers is greater than a preset tolerance constant, the service state reconstruction module blocks the mapping action of subsequent differential data sequences and sends a retransmission instruction carrying the missing timestamp to the source end. The system receives the retransmission message and compares the hash using the consistency verification certificate. The beacon alignment unit retransmits the message according to the monotonically increasing order of the logical timestamp. The message is inserted into the breakpoint address space of the receiving queue. The service status reconstruction module reads the receiving queue to update the service reconstruction status of the disaster recovery end. To support the above-mentioned breakpoint address space insertion and out-of-order reordering, the physical memory residence depth of the receiving queue is dynamically expanded and set to a ring storage pool that can concurrently accommodate 20,000 differential data frames, based on the product of the maximum round-trip response time of the current wide area transmission link and its sampling frequency. The beacon alignment unit relies on the built-in binary tree addressing matching algorithm to scan the timestamp increment trajectory of the messages residing in the receiving queue in parallel, accurately locate the logical out-of-order gaps caused by network jitter, and accurately embed the retransmission messages that have passed the retransmission verification into the corresponding clock gaps, thereby seamlessly connecting the service flow.
[0051] To address the scenario where attackers inject implicit negative event records by mixing interfering events—which conform to the communication protocol syntax but exhibit a slight negative logical shift—into a regular differential data sequence, the information entropy audit module extracts the first-order difference mean and second-order difference variance of the difference values within the sliding data window, constructing a feature value containing the sliding entropy. The three-dimensional feature vectors of the first-order difference mean and second-order difference variance are generated. These three feature parameters belong to different physical dimensions and exhibit numerical scaling differences spanning multiple orders of magnitude. Before the information entropy audit module performs spatial distance calculation, it triggers a data decoupling and standardization procedure. It reads the historical expected parameters and historical standard deviation parameters of the three feature baselines stored in non-volatile memory. Using deviation standardization logic, it extracts the three feature values within the current sliding data window, subtracts the corresponding historical expected parameters from each, and divides the resulting difference by the corresponding historical standard deviation parameter. The output is a dimensionless standard three-dimensional feature vector, free from physical unit interference. The Euclidean distance between the three-dimensional feature vector and the pre-stored normal baseline vector is then calculated. ,in, The scale representing the spatial deviation of the current sequence relative to normal business logic, when the sliding entropy eigenvalue... Below the safety threshold and Euclidean distance When the distance exceeds the predetermined spatial boundary, the information entropy audit module outputs an implicit negative event record injection judgment signal, triggering the security consensus verification module to extract the corresponding data segment, perform hash calculation, and generate a consistency verification certificate. This blocks the flow of logically tampered data hidden under the fluctuation characteristics of the normal entropy value into the disaster recovery node. Here, the predetermined spatial boundary is represented by a three-dimensional decision hypersphere with its center anchored at the endpoint of the non-poisoned baseline vector. Its physical radius is set as the result of multiplying the historical highest peak value of the measured Euclidean distance during the controlled and interference-free operation phase of the system by a tolerance coefficient of 1.5. To ensure that the distance measurement in Euclidean space can reflect the true physical weight of the variation in each dimension in a balanced manner, based on principal component dimensionality reduction measurement analysis, the Euclidean distance coefficients of the three coordinate axes mapped by the sliding entropy eigenvalue, the first-order difference mean, and the second-order difference variance in this three-dimensional feature vector are weighted to 0.60, 0.25, and 0.15, respectively, to ensure that the indicators of the dominant probability topological distortion play a decisive role in guiding distance deviation.
[0052] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.
Claims
1. A data redundancy disaster recovery system for plastics business, characterized in that, The system includes a heterogeneous data offloading module, a high-frequency time-series synchronization module, a security consensus verification module, an information entropy auditing module, and a business state reconstruction module. The heterogeneous data diversion module is used to process and classify the acquired business data messages, distribute the differential data sequence in the business data message to the high-frequency time synchronization module, or divert the production formula data in the business data message to the security consensus verification module. The information entropy audit module is used to acquire the differential data sequence transmitted by the high-frequency timing synchronization module and maintain a sliding data window with a fixed byte length in memory. By statistically analyzing the probability characteristics of the numerical distribution within the sliding data window, the sliding entropy characteristic value, which characterizes the distribution characteristics of the differential data sequence, is determined. The information entropy audit module is also used to send a verification command to the security consensus verification module when the sliding entropy characteristic value exceeds a preset security threshold. The security consensus verification module is used to respond to verification commands by performing hash operations on the differential data sequence to generate a consistency verification certificate, and then transmit the consistency verification certificate and the global time-series beacon to the business state reconstruction module. The business status reconstruction module is used to filter the legality of differential data sequence processing based on consistency verification credentials, and to reorganize the filtered data in conjunction with global time-series beacons to generate the business reconstruction status of the disaster recovery end.
2. The plastic business data redundancy disaster recovery system according to claim 1, characterized in that, When the information entropy audit module processes the audit logic of differential data sequences, it includes the following sub-steps: Step S11, collect the differential values entering the sliding data window and determine the statistical frequency of each value interval; Step S12, calculate the corresponding distribution probability based on the statistical frequency and use the distribution probability to determine the sliding entropy feature value. Step S13: Monitor the offset gradient of the sliding entropy feature value in the time domain, and update the safety threshold when the offset gradient reaches the preset offset step size.
3. The plastic business data redundancy disaster recovery system according to claim 1, characterized in that, The information entropy audit module also includes a dynamic baseline calibration unit; the dynamic baseline calibration unit is used to obtain historical entropy samples of the business status reconstruction module within a preset operating cycle, and dynamically correct the safety threshold processing by calculating the variance envelope of the historical entropy samples.
4. The plastics business data redundancy disaster recovery system according to claim 1, characterized in that, The security consensus verification module includes a fingerprint calculation unit and a credential encapsulation unit. The fingerprint calculation unit is used to perform one-way hash operations on the differential data segments within the sliding data window after receiving the verification command, and extract a globally consistent fingerprint to characterize the data integrity.
5. A plastics business data redundancy disaster recovery system according to claim 4, characterized in that, The credential encapsulation unit is used to associate and encapsulate the global consistency fingerprint with the current business recipe version number to generate a consistency verification credential.
6. A plastics business data redundancy disaster recovery system according to claim 1, characterized in that, The high-frequency timing synchronization module adopts a unidirectional push logic architecture based on the UDP protocol to carry the timing condition differential sequence generated by the plastic production equipment.
7. A plastics business data redundancy disaster recovery system according to claim 1, characterized in that, The business status reconstruction module includes a beacon alignment unit. The beacon alignment unit is used to parse global time-series beacons and, based on the logical timestamps carried by the global time-series beacons, to perform time-series rearrangement of differential data sequences that are out of order due to network transmission jitter.
8. A plastics business data redundancy disaster recovery system according to claim 7, characterized in that, The business state reconstruction module is also used to directly map the differential data sequence to the memory image when the sliding entropy feature value is lower than the safety threshold, so as to complete the business state generation in the no-verification mode and maintain the response speed of the data link within 1ms.
9. A plastics business data redundancy disaster recovery system according to claim 1, characterized in that, The heterogeneous data diversion module is also used to identify the header file identifier of the business data message, and when the header file identifier points to static production parameters, it directs the business data message to the security consensus verification module for processing and encrypted storage.