A pre-treatment based flame resistant garment fabric batch traceability transfer method and system

Abstract

The application relates to a pretreatment-based flame-retardant garment fabric batch traceability transfer method and system, and the core scheme comprises the following steps: acquiring and structuring device operation state, operator identity and material attribute information through distributed sensor and control system protocol adaptation; fusing physical, operation constraint and material characteristic signal, constructing a multi-dimensional semantic feature vector, realizing node credibility modeling and anchor point screening, and giving the anchor point a unique identifier by using a time-space fingerprint; generating a causal confidence subgraph according to the anchor point set, supporting multi-flow parallel topology analysis and main trace flow screening; introducing a real-time health state criterion to dynamically maintain anchor point effectiveness, and supporting anchor point degradation and adaptive structure correction. The beneficial effects are that the production flow node verifiability is enhanced, the traceability accuracy and the topology structure flexibility are improved, and the high-reliability traceability demand in a complex production environment is adapted.

Description

Technical Field

[0001] This invention relates to the field of flame-retardant clothing fabric production and batch traceability modeling technology, and in particular to a batch traceability and transfer method and system for flame-retardant clothing fabric based on pretreatment. Background Technology

[0002] Batch traceability modeling technology in the flame-retardant clothing fabric production field has long focused on static topology modeling based on pre-set process flows. It typically employs centralized path planning, fixed process segment lists, or workstation chain compression to organize and archive batch flow data. In typical practice, enterprises mainly utilize MES systems or industrial IoT platforms to reduce production data through rule engines, constructing a three-level data relationship of "process / batch / equipment" to achieve full-process traceability of fabric batches. Under this approach, each flow and transfer of fabric batches must be compared one by one with the established flowchart or standard operating procedures. The attribution of process nodes is usually clarified through event sequence completion, chain segment repair, or manual rule judgment. When the actual production process undergoes deformations such as quality inspection re-inspection and retesting, order insertion, or multi-line parallelism, the system often corrects the original process by inserting placeholder markers, path splicing, and link merging to maintain topological integrity.

[0003] Currently available technologies, such as process log mining and global topology reconstruction, while improving data processing automation and topology robustness, are inadequate for handling the ambiguity of node ownership and the need for real-time correlation during batch parallel processing, especially when node ownership is highly dependent on global process settings or the completeness of event sequences. This leads to problems such as unclear node ownership in parallel paths, delayed link repair, process compatibility mismatches, and delayed clarification of contextual relationships between nodes, affecting the efficiency of management decisions and the accuracy of batch traceability. Especially in scenarios with strict process control and strong compliance, temporary node anomalies, production order insertions, equipment switching, and operator handovers can all cause traceability link breaks, node flow intersections, and even difficulties in historical correction, reducing the reliability and security of fabric batch traceability. Summary of the Invention

[0004] This application provides a batch traceability and transfer method and system for flame-retardant clothing fabrics based on pretreatment, which aims to solve one of the problems or issues of the prior art mentioned in the background.

[0005] This application provides a batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment, specifically including: S1: Obtain the original workstation log data during the flow of flame-retardant clothing fabric. The data includes equipment operating status, operator identification, and material physical property information to form an initial flow record set to be processed.

[0006] S2: Based on the initial flow record set, semantic strength analysis is performed using three types of dynamic indicators: physical mandatory signals, operational constraint signals, and material intrinsic signals, to generate a multi-dimensional semantic feature vector representing the credibility of nodes.

[0007] S3: Based on the intensity threshold determination results of at least two types of signals in the multidimensional semantic feature vector, select key process nodes with strong semantic constraints and assign them unique spatiotemporal fingerprints to construct a set of dynamic semantic anchor points as the topological skeleton.

[0008] S4: Taking each dynamic semantic anchor in the set of dynamic semantic anchors as the center, extract the predecessor and successor nodes that have a direct causal chain, calculate the weighted values ​​of time interval stability, device communication handshake success rate and operator identity consistency, and generate a local anchor subgraph labeled with causal confidence.

[0009] S5: For scenarios where multiple parallel paths in the local anchor subgraph converge at the same dynamic semantic anchor point, each converging path is identified as an independent causal flow and stored in an independent namespace.

[0010] S6: Based on the most recent dynamic semantic anchor event of the target batch, locate the starting position, expand backward along the independent causal flow to the starting dynamic semantic anchor, filter out the main tracing flow, and generate a candidate set of tracing paths containing auxiliary verification flows.

[0011] S7: Continuously monitor the failure signals of each dynamic semantic anchor in the set of dynamic semantic anchors. The failure signals include equipment downtime, operator off-duty status and number of times the same batch is triggered and missing. Generate a real-time status marker that represents the validity of the anchor.

[0012] S8: If the real-time status flag indicates that a specific dynamic semantic anchor point has failed, then a degradation process is performed to convert the specific dynamic semantic anchor point into a normal node, and its historical local anchor point subgraph is migrated to the knowledge base to be verified.

[0013] This application also provides a batch traceability and transfer system for flame-retardant clothing fabrics based on pretreatment, which uses a batch traceability and transfer method based on pretreatment to trace and transfer batches of flame-retardant clothing fabrics.

[0014] This application provides a batch traceability and transfer method and system for flame-retardant clothing fabrics based on pretreatment, which has the following beneficial effects: (1) To address the issue of ambiguous process node attribution caused by complex working conditions such as quality inspection re-inspection, process insertion, and multi-line parallelism in the production of flame-retardant clothing fabrics, this solution abandons the traditional traceability model that relies on preset flowcharts or centralized path planning, and proposes a two-layer modeling paradigm of "anchor point driven + relationship emergence," which significantly improves the stability and accuracy of the traceability structure under dynamic disturbance environments. By introducing a semantic strength analysis mechanism based on the fusion judgment of three types of dynamic signals: physical coercion, operational constraints, and material intrinsics, the system can automatically identify key process nodes with high semantic anchoring capabilities without prior rule templates, and assign them unique spatiotemporal fingerprints, thereby constructing a dynamic anchor point system based on real business constraints. This mechanism effectively overcomes the path breakage and attribution mismatch defects caused by traditional methods when facing non-standard processes, abnormal jumps, or data missing data, and realizes robust extraction and reliable labeling of key events in multi-source heterogeneous logs, greatly enhancing the adaptability of the traceability system in complex production scenarios.

[0015] (2) At the topology modeling level, this scheme innovatively adopts a local "anchor subgraph" instead of a global static graph. Each subgraph only covers the predecessor and successor causal chain nodes directly related to a specific anchor point. The "causal confidence" edge weights are generated by weighting the stability of time intervals, the success rate of equipment communication, and the consistency of operator identity, so that the topology relationship has the characteristics of dynamic adjustment and context sensitivity. When multiple parallel process flows converge at the same anchor point, the system does not force path merging, but retains the independent namespace of each "causal flow" and uses the cross-flow association matrix to record the process compatibility tags between them, so as to achieve logical isolation and semantic interconnection in the state of multi-path parallelism. On this basis, the tracing response executes a three-level strategy of "anchor backtracking - flow matching - context alignment", prioritizing the main tracing flow with the smallest parameter fluctuation, the best equipment status, and the most complete qualification coverage as the core tracing path. The remaining flows are embedded in the report as auxiliary verification information, which not only ensures the reliability of the results, but also improves the transparency of decision-making. This design avoids the computational redundancy and assumption bias brought about by traditional graph compression, event completion or digital twin modeling, and realizes a lightweight and highly interpretable real-time tracing response.

[0016] The aforementioned technical approaches collectively construct a novel traceability architecture that requires no predefined processes, does not rely on external prediction models, and does not introduce complex intermediate state constructions. This fundamentally solves the technical challenge of unclear node attribution under multi-line parallel and dynamic disturbance conditions. By internalizing the five-dimensional industrial constraints of "human-machine-material-method-environment" into a computable, propagable, and degenerate semantic anchor system, the system not only significantly improves the quality controllability and compliance auditability in the production process of flame-retardant clothing, but also possesses excellent scalability and engineering feasibility. It is suitable for the manufacturing field of special protective equipment with high safety levels and strong regulatory requirements, providing a new technical paradigm for building an autonomous, closed-loop feedback intelligent traceability system. Attached Figure Description

[0017] Figure 1 This is the main flowchart of a batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment.

[0018] Figure 2 This is a sub-flowchart of a batch traceability and transfer method for pre-treated flame-retardant clothing fabrics.

[0019] Figure 3 This is another sub-flowchart of a batch traceability and transfer method for pre-treated flame-retardant clothing fabrics. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0021] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0022] like Figure 1 As shown, this application provides a batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment, specifically including: S1: Obtain the original workstation log data during the flow of flame-retardant clothing fabric. The data includes equipment operating status, operator identification, and material physical property information to form an initial flow record set to be processed.

[0023] S2: Based on the initial flow record set, semantic strength analysis is performed using three types of dynamic indicators: physical mandatory signals, operational constraint signals, and material intrinsic signals, to generate a multi-dimensional semantic feature vector representing the credibility of nodes.

[0024] S3: Based on the intensity threshold determination results of at least two types of signals in the multidimensional semantic feature vector, select key process nodes with strong semantic constraints and assign them unique spatiotemporal fingerprints to construct a set of dynamic semantic anchor points as the topological skeleton.

[0025] S4: Taking each dynamic semantic anchor in the set of dynamic semantic anchors as the center, extract the predecessor and successor nodes that have a direct causal chain, calculate the weighted values ​​of time interval stability, device communication handshake success rate and operator identity consistency, and generate a local anchor subgraph labeled with causal confidence.

[0026] S5: For scenarios where multiple parallel paths in the local anchor subgraph converge at the same dynamic semantic anchor point, each converging path is identified as an independent causal flow and stored in an independent namespace.

[0027] S6: Based on the most recent dynamic semantic anchor event of the target batch, locate the starting position, expand backward along the independent causal flow to the starting dynamic semantic anchor, filter out the main tracing flow, and generate a candidate set of tracing paths containing auxiliary verification flows.

[0028] S7: Continuously monitor the failure signals of each dynamic semantic anchor in the set of dynamic semantic anchors. The failure signals include equipment downtime, operator off-duty status and number of times the same batch is triggered and missing. Generate a real-time status marker that represents the validity of the anchor.

[0029] S8: If the real-time status flag indicates that a specific dynamic semantic anchor point has failed, then a degradation process is performed to convert the specific dynamic semantic anchor point into a normal node, and its historical local anchor point subgraph is migrated to the knowledge base to be verified.

[0030] This application also provides a batch traceability and transfer system for flame-retardant clothing fabrics based on pretreatment, which uses a batch traceability and transfer method based on pretreatment to trace and transfer batches of flame-retardant clothing fabrics.

[0031] Step S1: Obtain the original workstation log data during the flow of flame-retardant clothing fabric. This data includes equipment operating status, operator identification, and material physical property information to form an initial flow record set to be processed. Specifically, this includes: S1.1: Acquire the raw sensor signals and control system event streams distributed across various process nodes during the flow of flame-retardant clothing fabric. Utilize an industrial protocol parsing engine to perform protocol adaptation and timestamp alignment on the multi-source heterogeneous data to generate a raw workstation log data stream with a unified time base.

[0032] S1.1 Acquires the raw sensor signals and control system event streams distributed across various process nodes during the flow of flame-retardant clothing fabric. Utilizes an industrial protocol parsing engine to perform protocol adaptation and timestamp alignment on the multi-source heterogeneous data to generate a raw workstation log data stream with a unified time base.

[0033] This sub-step, as the core preprocessing step of the data acquisition module, is responsible for solving the problem of data spatiotemporal misalignment caused by differences in equipment brands, inconsistent communication protocols, and network transmission delays at the flame-retardant clothing production site, and provides standardized input for the subsequent construction of high-precision dynamic semantic anchors.

[0034] The industrial protocol parsing engine deployed on the edge computing gateway monitors the network ports of each process node in real time, capturing the underlying data packets from heterogeneous devices such as heat setting machines, coating spraying lines, and flame retardant testers. These data packets cover ModbusTCP, OPC UA, Profinet, and proprietary TCP / IP protocol formats.

[0035] For the captured unstructured binary data stream, the protocol parsing engine performs byte-level unpacking operations based on the preset device location mapping table to extract process variable values ​​including temperature, pressure, speed, and valve opening, as well as discrete event status codes such as device start-up / stop, alarm triggering, and task completion.

[0036] The extracted process variables and event status codes are encapsulated into a standardized key-value pair structure, where the key name follows the naming convention of "device ID_measurement point type_parameter name" to ensure that the same physical quantities generated by devices from different manufacturers are comparable at the logical level.

[0037] Because the internal clocks of various distributed sensors and control systems drift and the network transmission path lengths vary, directly collected timestamps cannot reflect the true causal time sequence. Therefore, a time synchronization mechanism based on a global high-precision time server needs to be introduced.

[0038] The system arrival timestamp when the edge gateway receives data packets is obtained, and combined with the estimated protocol parsing time, the actual physical time of data generation is calculated in reverse, thus initially correcting the timing deviation caused by network jitter.

[0039] Linear interpolation is used to complete missing timestamp segments. For data breakpoints with consecutive sampling intervals exceeding a preset threshold, they are marked as invalid intervals and retransmission requests are triggered to ensure the continuity of the time series.

[0040] Establish a time window sliding mechanism based on a globally unified time base to map data streams from different protocol sources onto the same time axis, eliminating relative time offsets caused by protocol conversion delays.

[0041] Define a time alignment error tolerance threshold. When the timestamp deviation of multi-source data is less than the threshold, it is determined to be a related event that occurred at the same time and is merged into the same logical time slice. If the deviation exceeds the threshold, each independent timestamp is retained and marked with an asynchronous identifier.

[0042] The corrected unified timestamp is calculated using the following formula to quantify the elimination of the effects of clock drift:

[0043] Among them, t aligned For the corrected unified timestamp, t recv d represents the local system time at which the edge gateway receives data packets. net d is the estimated network transmission delay obtained by measuring network round-trip time (RTT). proc The fixed time estimate for protocol parsing and internal processing is given by δt, where δt is the cumulative drift of the local clock relative to the global high-precision time synchronization server.

[0044] Network transmission delay estimate d net The RTT data was obtained by periodically sending probe packets and recording the round-trip time. The least squares method was used to fit the RTT data from multiple recent measurements, and abnormal fluctuation values ​​were removed to obtain a stable delay benchmark.

[0045] The local clock cumulative drift δt is obtained by periodically synchronizing with the global high-precision time server via NTP or PTP protocol. The difference between the local clock and the standard time is calculated, and dynamic compensation is performed considering the temperature variation characteristics of the crystal oscillator frequency deviation.

[0046] The corrected unified timestamp is written into a standardized key-value pair structure to form an atomic data unit containing complete context information. Each unit uniquely corresponds to a specific device state or event at a physical moment.

[0047] The atomic data units are sorted in chronological order to generate a data stream queue with strict temporal logic, ensuring the temporal correctness of causal chain analysis in subsequent steps.

[0048] The system performs deduplication on duplicate data in the data stream queue, identifies and removes redundant records caused by network retransmission or device malfunction, and retains the earliest valid data unit.

[0049] Perform integrity checks on the deduplicated data stream, checking whether key fields (such as device ID, timestamp, parameter value) are empty or invalid, discarding data units that fail the check and recording error logs.

[0050] By adapting the above protocols and using clock synchronization processing, the scattered, heterogeneous, and asynchronous multi-source raw signals are transformed into raw workstation log data streams with a unified time base, standardized structure, and consistent timing. This achieves precise alignment of multi-source heterogeneous data in the spatiotemporal dimension, laying a solid data foundation for subsequent extraction of equipment operating status feature vectors and construction of dynamic semantic anchors.

[0051] S1.2: Based on the original workstation log data stream, extract the equipment operating status feature sequence including hot pressing temperature curve and coating spraying pressure value, and use sliding window filtering to remove high-frequency noise interference in order to construct an equipment operating status feature vector that characterizes the real-time operating condition of the production equipment.

[0052] The system receives raw workstation log data streams and extracts the sequences of hot press roller temperature, heating chamber air pressure, and spray nozzle outlet pressure. A sliding window filter (window width W, step size S) is used for smoothing: the mean or median within the window is calculated, and the standard deviation is calculated in real time; if the value exceeds a threshold, it is marked as an area to be reviewed. For the temperature curve, a second-order difference method is used to detect abrupt slope changes and remove false peaks; for the spray pressure, cross-validation using flow meters is used to correct drift errors. The filtered data is normalized to [0,1]. The data is cut into discrete segments corresponding to work batches according to the process cycle time. The mean, variance, peak value, valley value, waveform factor, and other statistical characteristics of each segment are calculated and concatenated to form a feature vector of equipment operating status, providing a high-quality data foundation for subsequent semantic anchor point determination.

[0053] S1.3: Using the operation trigger time in the device operation status feature vector as a synchronization benchmark, the biometric hash value and work badge code information uploaded by the access control system and the handheld terminal are read together, and the identity deduplication and permission verification logic is executed to generate a set of operator identity identifiers with unique constraints.

[0054] The system receives the device status feature vector from S1.2 and extracts the operation trigger time as a synchronization benchmark. A millisecond-level sliding window (dynamic width) is constructed centered on the trigger time. Access control logs and handheld terminal data within the window are retrieved to obtain the biometric hash value and work badge code of the matching workstation. Integrity is verified, and entries with abnormal formats are filtered. For redundant records of the same operator, the unique record closest to the trigger time and with the strongest signal is retained. The system compares the record with the production qualification database to verify special operation qualifications. If successful, an identity token containing the work number, feature hash, and qualification level is generated; otherwise, a null value is generated and an alarm is triggered. Identity tokens are aggregated by time to form a set of operator identity identifiers that correspond one-to-one with the device status vector, achieving precise solidification of the "human" element and providing personnel-dimensional constraints for constructing semantic anchors.

[0055] S1.4: Based on the work batch number corresponding to the operator identity set, call the fiber blending ratio, weight distribution and moisture content values ​​recorded by the online detection instrument, and process them through unit normalization and dimension conversion to form a material physical property information matrix describing the intrinsic characteristics of the fabric.

[0056] S1.5: Based on the equipment operating status feature vector, operator identity set, and material physical attribute information matrix, perform multi-dimensional data field mapping and structured encapsulation operations to aggregate discrete single-point monitoring data into a set of initial flow records to be processed containing complete context information.

[0057] Step S2: Based on the initial flow record set, semantic strength analysis is performed using three types of dynamic indicators: physical mandatory signals, operational constraint signals, and material intrinsic signals, to generate a multi-dimensional semantic feature vector representing the credibility of nodes. Specifically, this includes: S2.1: Analyze the equipment operation status data in the initial flow record set, extract the independent calibration actions triggered by the flame retardant performance retesting equipment and the unavoidable judgment results generated by the wash fastness tester, and generate a physical compulsion signal sequence that characterizes the degree of physical compulsion.

[0058] S2.2: Based on the batch information corresponding to the physical mandatory signal sequence, retrieve operator identification data, identify the process instructions that must be executed by operators with special protection qualifications for a specific batch, and the operation logs that the system automatically intercepts for unauthorized parameter modifications, and generate an operation constraint signal sequence that characterizes the degree of operation compliance.

[0059] Based on the batch identifier carried in the physical mandatory signal sequence, an association retrieval operation is performed in the initial flow record set to extract all operator identification data and corresponding process control instruction logs that are logically bound to that batch. This step aims to establish a spatiotemporal mapping relationship between material flow and personnel operations, providing basic data support for subsequent operational compliance determination.

[0060] The extracted operator identification data is parsed for qualification attributes. A pre-set special protection operation qualification database is called to compare whether the current operator has the legal certification required to perform specific flame retardant treatment procedures (such as high-temperature hot pressing and chemical coating impregnation). If the operator's qualification level is lower than the minimum safety threshold required by the procedure, the node is marked as a qualification violation; if the qualifications match, an initial confidence value for qualification compliance is generated.

[0061] The system synchronously reads the underlying security interception logs to identify parameter modification requests triggered during process execution and their processing results. It focuses on capturing unauthorized parameter adjustment behaviors that are automatically blocked by the system due to violations of process safety regulations, such as forcibly increasing hot-pressing temperatures during non-cooling phases or arbitrarily shutting down exhaust systems. The system counts the number of illegal commands intercepted within a unit time window to construct an operational risk frequency index.

[0062] The initial value of qualification compliance confidence is inversely mapped to the operational risk frequency index. For nodes with qualification violation records, an extremely low operational constraint strength score is directly assigned; for nodes with qualification compliance but high-frequency interception records, their strength score is reduced based on the product of the number of interceptions and the preset risk weight; for nodes with neither qualification violations nor unauthorized interception records, a high baseline strength score is maintained.

[0063] Weighted normalization is applied to standardize the scores obtained after the above processing to eliminate numerical bias caused by different units. The operational constraint signal strength component is calculated using the following formula:

[0064] Among them, S op To operate the constraint signal strength component, W qual C is the qualification weighting coefficient. qual W represents the confidence level of qualification compliance (value 0 or 1). log N represents the log weight coefficient. block N represents the number of unauthorized commands intercepted. total This represents the total number of instruction requests within that time period.

[0065] The calculated operational constraint signal intensity components are arranged in a time series to form a continuous operational constraint signal sequence. This sequence reflects the contribution of human intervention to process stability and compliance level during the transfer of flame-retardant clothing fabrics.

[0066] Through the dual verification mechanism based on qualification verification and log auditing, discrete operator identity data and system interception records are transformed into a sequence of operational constraint signals that characterize the degree of operational compliance. This effectively suppresses the problem of ambiguity in traceability nodes caused by human factors in multi-path parallel scenarios and significantly improves the accuracy of node attribution determination in topology modeling.

[0067] S2.3: Using the material physical property information associated with the operation constraint signal sequence, calculate the frequency and duration of the exclusive temperature control threshold alarm triggered by a certain batch of fabric due to the presence of a special aramid blend ratio, and generate a material intrinsic signal sequence characterizing the degree of material specificity.

[0068] Extract the material physical property information associated with the operational constraint signal sequence, locate the batch data record containing the special aramid blend ratio identifier, and use it as the initial input object for material intrinsic signal analysis.

[0069] Retrieve historical temperature control logs for the heat setting and coating curing processes associated with this batch of data records, and filter out the set of abnormal temperature control events where the temperature values ​​exceed the exclusive safety threshold range corresponding to this specific blending ratio.

[0070] The total number of occurrences of the abnormal temperature control event set within a preset time window is counted, the alarm trigger frequency per unit time is calculated, and a frequency domain characteristic index characterizing the sensitivity of the material to process parameters is generated.

[0071] The duration of each alarm in the abnormal temperature control event set is summed up to calculate the cumulative over-temperature duration within a single cycle, generating a time-domain characteristic index that characterizes the thermal stability risk of the material.

[0072] A quantitative calculation model for the intrinsic signal strength of materials is constructed, mapping alarm frequency and cumulative over-temperature duration to normalized intensity scores. The intrinsic signal strength value of materials is calculated using the following formula:

[0073] Where S is the intrinsic signal strength value of the material, e is the natural constant, w1 is the weighting coefficient of the alarm frequency, f is the alarm trigger frequency per unit time, w2 is the weighting coefficient of the cumulative timeout duration, and t is the cumulative over-temperature duration.

[0074] Based on the thermal decomposition characteristics of aramid fibers in flame-retardant clothing fabrics, initial values ​​for weighting coefficients w1 and w2 are set to ensure that high-frequency short-term alarms and low-frequency long-term alarms are comparable in signal strength.

[0075] The calculated intrinsic signal intensity values ​​of the material are linearly mapped and compressed to the range of 0 to 1, thus eliminating the problem of dimensional inconsistency caused by differences in the blending ratio of different batches of fabric.

[0076] The mapped intensity values ​​are encapsulated into structured data units containing batch ID, timestamp, and intensity value, forming a material intrinsic signal sequence that characterizes the degree of material specificity.

[0077] By using the weighted nonlinear mapping processing method based on the alarm frequency and duration of the dedicated temperature control threshold, the operational constraint correlation data of the previous step is transformed into a standardized material intrinsic signal strength index. This enables the quantitative characterization of the process deviation risk caused by the special physical properties of the material, and provides a highly discriminative material dimension input for the subsequent fusion of multi-dimensional semantic feature vectors.

[0078] S2.4: Based on the physical mandatory signal sequence, the operational constraint signal sequence, and the material intrinsic signal sequence, perform normalization processing and weighted fusion operation respectively to map the three types of heterogeneous signals into intensity scores with unified dimensions, and generate an original semantic intensity combination vector containing physical intensity components, operational intensity components, and material intensity components.

[0079] S2.5: Perform multi-dimensional spatial projection processing on the original semantic strength combination vector, and correct the weight of each component by combining the time decay factor to output the final multi-dimensional semantic feature vector that represents the credibility level of the node in the complex flow scenario.

[0080] Receive the original semantic strength combination vector output from step S2.4, which includes physical strength components, operational strength components, and material strength components. This vector represents the static semantic strength value of the current process node in three dimensions.

[0081] A three-dimensional semantic feature space coordinate system is constructed, and the original semantic intensity combination vector is mapped to this space. The X-axis represents the physical mandatory signal intensity, the Y-axis represents the operational constraint signal intensity, and the Z-axis represents the material intrinsic signal intensity. The geometric positional relationship of each component in the multi-dimensional space is established.

[0082] A time decay factor function is introduced to correct the decrease in semantic credibility caused by the time difference between the data generation time and the current traceability time. This factor is constructed based on an exponential decay model to ensure that key processes that have occurred recently have a higher weight contribution.

[0083] The time decay factor is calculated using the following formula:

[0084] Where λ is the time decay coefficient, e is the base of the natural logarithm, α is the preset decay rate constant, and Δt is the difference between the current system time and the time of the node event.

[0085] Based on the typical cycle characteristics of flame-retardant clothing fabric production flow, the range of values ​​for the attenuation rate constant α is set so that the attenuation effect is weak within the normal flow interval, while the node weight is significantly reduced in abnormal stagnation or long-term lag scenarios.

[0086] Each component in the original semantic strength combination vector is multiplied by its corresponding time decay coefficient to generate a time-corrected dynamic semantic strength component, thus eliminating the interference of timing deviations caused by logistics waiting or process insertion on semantic judgment.

[0087] A multidimensional spatial projection transformation is performed to project the time-corrected three-dimensional dynamic semantic intensity components onto a unit hypersphere. The dimensional differences are eliminated through normalization, ensuring that the signal intensities of different dimensions are comparable in subsequent weighted fusion.

[0088] The magnitude of the projected vector is calculated as a preliminary measure of the overall semantic confidence of the node. This magnitude reflects the comprehensive tightness of the three types of constraints—physical, operational, and material—at the current moment.

[0089] By combining the preset importance weight matrix of each dimension, the projected components are weighted and summed. The weight matrix is ​​dynamically adjusted according to the sensitivity of different links in the flame retardant clothing production process to safety compliance, and the final multidimensional semantic feature vector score is calculated.

[0090] The calculated multidimensional semantic feature vector scores are encapsulated into a standardized data structure, and node IDs and timestamp labels are added to form the final multidimensional semantic feature vector that represents the credibility level of a node in a complex flow scenario.

[0091] By using the above-mentioned multi-dimensional spatial projection and time decay correction processing method, the original static intensity combination of the previous step is transformed into a dynamic semantic feature vector with time sensitivity and dimensional balance, realizing the fine quantification of node credibility and providing a highly discriminative judgment basis for the accurate selection of subsequent dynamic semantic anchor points.

[0092] like Figure 2 As shown, step S3 involves: based on the intensity threshold determination results of at least two types of signals in the multidimensional semantic feature vector, selecting key process nodes with strong semantic constraints and assigning them unique spatiotemporal fingerprints, thus constructing a dynamic semantic anchor point set as the topological skeleton. Specifically, this includes: S3.1: Perform pairwise combination logic operations on the physical mandatory signal component, the operation constraint signal component, and the material intrinsic signal component in the multidimensional semantic feature vector to generate three types of dual signal strength judgment matrices, including physical and operation combination, physical and material combination, and operation and material combination, so as to establish the preliminary candidate range of nodes to be selected as dynamic semantic anchors.

[0093] Receive the multidimensional semantic feature vector output from step S2. This vector contains physical mandatory signal components, operational constraint signal components, and material intrinsic signal components for each process node, serving as the basic data source for constructing the dual signal strength determination matrix.

[0094] The multidimensional semantic feature vector is deconstructed, and the three independent dimension signal intensity values ​​corresponding to each node are extracted and denoted as physical intensity value P, operational intensity value O and material intensity value M, respectively, to ensure that the values ​​of each component have been normalized and are in the same dimension space.

[0095] A combined physical and operational judgment logic is constructed, and the physical strength value P and the operational strength value O are mapped by Cartesian product to generate a first-class dual-signal joint feature space, which is used to characterize the node association strength that simultaneously possesses the dual attributes of mandatory equipment calibration and personnel qualification compliance.

[0096] A logic for determining the combination of physical and material strengths is constructed. The physical strength value P and the material strength value M are mapped by Cartesian product to generate a second type of dual-signal joint feature space, which is used to characterize the node association strength that simultaneously satisfies the mandatory detection conditions of the equipment and the triggering conditions of the special properties of the material.

[0097] A logic for determining the combination of operations and materials is constructed. The operation intensity value O and the material intensity value M are mapped by Cartesian product to generate a third type of dual-signal joint feature space, which is used to characterize the node association strength that simultaneously meets the special operation qualification requirements and the material intrinsic characteristics constraints.

[0098] For the three types of dual-signal joint feature spaces mentioned above, matrix encapsulation operations are performed respectively to reassemble discrete dual-dimensional signal pairs into structured data matrices, where the row index of the matrix corresponds to the process node ID, and the column index corresponds to the two signal dimensions involved in the combination and their interaction weight coefficients.

[0099] In the matrix generation process, a nonlinear coupling operator is introduced to perform weighted fusion of the two signal components in order to capture the complex semantic constraints that cannot be reflected by a single signal, and to ensure that the decision matrix can reflect the node credibility enhancement effect under the synergistic effect of multiple source signals.

[0100] Calculate the combined strength value of the two signals to quantify the semantic anchoring potential of a node under a specific combined dimension.

[0101] The generated three types of dual signal strength judgment matrices are subjected to integrity verification to ensure that all valid process nodes have corresponding combined strength records in the matrix, and empty rows caused by missing data collection are removed to ensure data consistency for subsequent threshold comparison.

[0102] The output includes three types of dual-signal strength judgment matrices: physical and operational combinations, physical and material combinations, and operational and material combinations. This establishes the initial candidate range for nodes to be selected as dynamic semantic anchors, providing multi-dimensional quantitative basis for the next step of precise screening based on dynamic thresholds.

[0103] Through the above-mentioned dual-signal combination logic operation and matrix encapsulation processing method, the multi-dimensional semantic feature vector of the previous step is transformed into a structured judgment matrix with cross-validation capability, realizing multi-angle quantitative evaluation of the semantic constraints of nodes, and significantly improving the robustness and accuracy of key process identification in complex flow scenarios.

[0104] S3.2: Based on the three types of dual signal strength judgment matrices, extract the real-time signal strength values ​​under each combination and compare them with the preset dynamic threshold range to screen out a list of qualified key process nodes that simultaneously meet the requirements of at least two types of signal strength thresholds, so as to eliminate ordinary flow nodes with weak semantic associations.

[0105] The three types of dual-signal strength determination matrices generated by step S3.1, including physical and operational combinations, physical and material combinations, and operational and material combinations, are received as the data input basis for this step.

[0106] Traverse each node index in the three types of dual-signal strength determination matrices, extract the real-time semantic strength value pair of the node under a specific signal combination, and form the original strength data tuple to be verified.

[0107] The system calls a preset dynamic threshold configuration library and retrieves the corresponding dual-signal collaborative judgment threshold range based on the production batch type and process complexity level of the current flame-retardant clothing fabric. The range includes two boundary parameters: a lower threshold and an upper threshold.

[0108] The extracted real-time semantic intensity values ​​are compared with the upper and lower limits of the dynamic threshold interval to determine whether each signal component independently falls within the valid confidence interval.

[0109] For the physical and operational combination decision matrix, perform a logical AND operation. Only when the physical mandatory signal strength score is greater than the lower limit of the physical threshold and the operational constraint signal strength score is greater than the lower limit of the operational threshold, mark the node as a candidate anchor point of the first type.

[0110] For the physical and material combination judgment matrix, a similar logical AND operation is performed. Only when the physical mandatory signal strength score is greater than the lower limit of the physical threshold and the material intrinsic signal strength score is greater than the lower limit of the material threshold, the node is marked as a second type of candidate anchor point.

[0111] For the operation and material combination decision matrix, perform a logical AND operation. Only when the operation constraint signal strength score is greater than the lower limit of the operation threshold and the material intrinsic signal strength score is greater than the lower limit of the material threshold, the node is marked as a third type of candidate anchor point.

[0112] Summarize the above three types of judgment results, perform logical OR aggregation on the marking status of the same node under different combinations, and if the node is marked in any combination, it is determined that it meets at least two types of signal strength threshold requirements.

[0113] Construct a list of qualified key process nodes, and write the node IDs that meet the judgment conditions, the corresponding dual signal combination types, and their real-time intensity values ​​into the list structure.

[0114] Remove nodes that do not meet the dual threshold requirements in any combination, classify them as ordinary flow nodes with weak semantic associations, and record the reasons for their exclusion for subsequent audit tracking.

[0115] Through the above-mentioned dual threshold comparison and multi-combination logic aggregation processing method, the multi-dimensional semantic feature vector generated in the previous step is transformed into a list of qualified key process nodes with high credibility screening results. This achieves the expected technical effect of accurately extracting core nodes with strong semantic constraints from all nodes, providing a precise set of target objects for subsequent spatiotemporal fingerprinting.

[0116] S3.3: For each node in the list of qualified key process nodes, collect its corresponding equipment operation identifier, high-precision timestamp data and operator biometric hash fragment, and serialize and splice the above multi-source heterogeneous information to generate an original spatiotemporal fingerprint code that represents the uniqueness of the node.

[0117] Receive the list of qualified key process nodes determined by step S3.2. Each node in the list has been determined to be a high-confidence flow node that simultaneously meets the requirements of at least two types of semantic signal strength thresholds, and serves as the core processing object for constructing dynamic semantic anchors.

[0118] For each qualified key process node in the list, the system calls the multi-source data acquisition interface in parallel to extract the equipment operation identifier, microsecond-level high-precision timestamp data, and operator biometric hash fragments that are physically bound to the node, forming a primitive heterogeneous information triplet to be fused.

[0119] The extracted equipment operation identifiers are standardized and cleaned to remove redundant prefixes and invalid suffixes, retaining only the unique equipment ID string that identifies the production equipment, ensuring the global uniqueness and resolvability of the identifier in the factory's IoT topology.

[0120] The acquired microsecond-level high-precision timestamp data is normalized and formatted, and converted into an integer time value based on the UTC standard. This eliminates time reference deviations caused by clock drift of distributed sensors or network latency, and establishes the absolute time coordinates of the nodes.

[0121] Integrity verification is performed on the operator's biometric hash fragment to verify whether it conforms to the preset SHA-256 digest length specification, and a fixed-length feature substring is extracted as the unique digital fingerprint of the person's identity, ensuring the traceability of identity while protecting privacy.

[0122] The processing logic for spatiotemporal fusion coding is constructed, and a serialization splicing template containing device, time, and personnel fields is defined. This template strictly specifies the arrangement order, delimiter type, and filling rules of the data in each field to ensure the structural consistency of the generated fingerprint.

[0123] The standardized device ID string is filled into the device field of the template, the normalized integer time value is filled into the time field, the extracted biometric substring is filled into the personnel field, and a specific invisible character is used as the field separator to perform a string-level linear concatenation operation.

[0124] The concatenated long string is converted into a byte stream encoding using the UTF-8 encoding standard to transform the character sequence into a binary byte array, eliminating encoding ambiguity caused by differences in character sets and providing a standardized input data stream for subsequent hash operations.

[0125] Through the serialization and splicing process of the above-mentioned spatiotemporal fusion coding, the scattered multidimensional heterogeneous information of equipment, time and personnel is aggregated into a compact and unique original spatiotemporal fingerprint code, realizing dual locking of key process nodes in both physical space and logical time dimensions, laying the data foundation for the subsequent generation of a tamper-proof final unique spatiotemporal fingerprint.

[0126] For example, taking the critical process node of "high-temperature hot pressing and setting of aramid blended fabric" in the flame-retardant clothing fabric production process as an example, this node is judged as a qualified critical process node in step S3.2 because it simultaneously meets the physical mandatory signal (the hot pressing temperature curve conforms to the exclusive temperature control threshold) and the operational constraint signal (executed by an operator with special protective equipment qualifications). The system first collects the equipment operation identifier corresponding to this node. The original data is "HP-Machine-007-V2.1", and after standardization and cleaning, the core ID "HP-007" is retained. Secondly, the system collects the high-precision timestamp of the completion time of this process. The original data is "2023-10-27T14:30:05.123456+08:00", which is converted into the UTC standard integer time value "1698388205123456" after time zone normalization and format unification processing. Next, the biometric hash fragment of the operator performing the procedure is collected. The original data is the SHA-256 digest "a1b2c3d4...z9y8x7w6". After integrity verification and truncation, the first 16 characteristic substrings "a1b2c3d4e5f6g7h8" are retained. Subsequently, a spatiotemporal fusion coding template is constructed, with the format set to "[DeviceID]||[Timestamp]||[BioHash]", where "||" is an invisible separator. "HP-007", "1698388205123456" and "a1b2c3d4e5f6g7h8" are sequentially filled into the template to generate the concatenated string "HP-007||1698388205123456||a1b2c3d4e5f6g7h8". Finally, the string is converted to UTF-8 encoding to obtain a 42-byte binary byte array, which is the original spatiotemporal fingerprint of the node. This fingerprint uniquely identifies "the hot press operation performed by the operator with biometric characteristics a1b2... at UTC time 1698388205123456 on the HP-007 hot press". Any minor alteration to the equipment, time, or personnel will result in a significant change in the fingerprint, thus achieving strong solidification and unique characterization of the node's identity, significantly improving the anti-interference capability and accuracy of the traceability topology skeleton.

[0127] S3.4: Perform irreversible hash encryption transformation and integrity check bit addition processing on the original spatiotemporal fingerprint code to generate a final unique spatiotemporal fingerprint with anti-tampering characteristics, and bind the final unique spatiotemporal fingerprint as the primary key index to the corresponding qualified key process node to complete the strong identity solidification of the node.

[0128] Receive the original spatiotemporal fingerprint code generated in step S3.3. This fingerprint code is composed of the device operation identifier, high-precision timestamp data and operator biometric hash fragment serialized and concatenated, and serves as the initial input data stream for anti-tampering processing.

[0129] The original spatiotemporal fingerprint code is subjected to character set standardization and cleaning operations to remove invisible control characters and non-standard encoded bytes, ensuring that the input data conforms to the UTF-8 encoding standard and eliminating the risk of format ambiguity caused by differences in heterogeneous system data sources.

[0130] The SHA-256 secure hash algorithm is used to perform one-way hashing on the standardized original spatiotemporal fingerprint code, mapping the variable-length input data into a fixed-length 256-bit binary digest sequence, thereby realizing digital fingerprint compression of the data content.

[0131] A dynamic salt value generation mechanism is introduced, which extracts the current microsecond-level random number from the system entropy pool and combines it with the server hardware fingerprint as a salt value factor. The salt value factor is then appended to the head of the hash digest sequence to enhance the resistance to rainbow table attacks.

[0132] A second hash operation is performed on the data block after the salt value is added to generate an intermediate check code with higher collision resistance, ensuring that even if there are slight differences in the original spatiotemporal fingerprint code, the generated check code will produce a significant change like an avalanche effect.

[0133] The CRC-32 algorithm is used to calculate the integrity check bits of the intermediate check code and generate a 32-bit check polynomial remainder, which is used to detect single-bit or multi-bit errors that may occur during data transmission or storage.

[0134] The intermediate checksum generated by the secondary hash is encapsulated in a structured manner with the CRC-32 checksum, and the data is reassembled according to the byte alignment format of 'hash header + salt value identifier + main digest + checksum tail' to form a final unique spatiotemporal fingerprint with self-verification capability.

[0135] Establish a global anchor point index database, and build a B+ tree index structure with the final unique spatiotemporal fingerprint as the primary key. Mount the process attribute data, semantic strength vector and context association pointer of qualified key process nodes under this primary key.

[0136] Perform consistency checks on index-bound transactions, verify the uniqueness constraint of the primary key, and ensure that the same spatiotemporal fingerprint corresponds to only one physical process node instance, preventing topology node identity conflicts caused by concurrent writes.

[0137] By using irreversible hash encryption transformation and integrity check bit addition processing, the original spatiotemporal fingerprint code from the previous step is transformed into a final unique spatiotemporal fingerprint that is tamper-proof, verifiable, and globally unique. This fingerprint is then used as the primary key index to solidify the strong identity of the node, achieving the expected technical effect of ensuring the identity of the core nodes in the traceability topology skeleton and protecting data integrity.

[0138] S3.5: Based on all qualified key process nodes that have been bound to a final unique spatiotemporal fingerprint, structured aggregation and encapsulation are performed according to the time sequence and process logic relationship to construct a dynamic semantic anchor set with elastic expansion capability, which serves as the core topological skeleton for subsequent local anchor subgraph generation.

[0139] like Figure 3 As shown, step S4 involves: taking each dynamic semantic anchor in the set of dynamic semantic anchors as the center, extracting predecessor and successor nodes with direct causal chains, calculating weighted values ​​for time interval stability, device communication handshake success rate, and operator identity consistency, and generating a local anchor subgraph labeled with causal confidence. Specifically, this includes: S4.1: Based on each dynamic semantic anchor in the set of dynamic semantic anchors, retrieve the predecessor process node and successor process node with direct physical connection or logical dependency relationship in the initial flow record set, and generate a candidate causal chain node set containing complete context information to establish the boundary range of the local topology.

[0140] S4.2: For adjacent node pairs in the candidate causal chain node set, extract timestamp sequence data and perform sliding window variance analysis to calculate the time interval stability coefficient between adjacent processes, and generate a quantitative evaluation index of the first dimension characterizing the temporal continuity, so as to eliminate false causal associations caused by abnormal waiting or order insertion.

[0141] The raw timestamp sequence data of adjacent process node pairs in the candidate causal chain node set is obtained. The data contains precise records of the completion time of the predecessor node and the start time of the successor node. The extracted timestamp sequence is preprocessed to remove outliers caused by sensor clock drift, ensuring the continuity and consistency of the time series data.

[0142] A sliding window mechanism is used to segment the timestamp sequence after cleaning. The window length is set to a multiple of the preset standard process cycle, and the step size is a single sampling interval, so as to cover the time fluctuation characteristics of different batches of fabric during the circulation process.

[0143] Calculate the time interval difference sequence between adjacent nodes within each sliding window. This difference sequence reflects the actual time distribution of material residence or transfer between two processes.

[0144] Based on the time interval difference sequence, its statistical variance is calculated to quantify the dispersion of time fluctuations. The smaller the variance, the tighter and more stable the process connections; the larger the variance, the more abnormal waiting, order insertion, or path jump interference exists.

[0145] A time decay factor is introduced to weight and correct the variance contribution of historical windows, giving higher weight to recent time windows to reflect the immediate impact of the current production status on traceability reliability. The time interval stability coefficient between adjacent processes is calculated.

[0146] When the actual time interval variance is significantly lower than the benchmark threshold, the exponential term approaches negative infinity and the stability coefficient S approaches 1, indicating extremely high temporal continuity. When the actual variance is much higher than the benchmark threshold, the exponential term approaches positive infinity and the stability coefficient S approaches 0, indicating extremely low temporal continuity.

[0147] The calculated time interval stability coefficient is mapped to a dimensionless value between 0 and 1, serving as the first dimension of quantitative evaluation index characterizing temporal continuity.

[0148] The evaluation metric is attached to the corresponding connection edge of the candidate causal chain node set for subsequent fusion calculation of comprehensive causal confidence.

[0149] By using the above-mentioned sliding window variance analysis and nonlinear mapping processing, the original timestamp sequence obtained in the previous step is transformed into a time interval stability coefficient that characterizes the tightness of process connection. This effectively eliminates false causal relationships caused by abnormal waiting, process insertion, or path skipping, and improves the temporal logic accuracy of local topology construction.

[0150] For example, in the causal chain analysis of flame-retardant clothing fabric flowing from the "heat pressing and setting" node to the "flame-retardant coating spraying" node, the sliding window length is set to 30 minutes, the step size to 5 minutes, the sensitivity adjustment factor α to 0.5, and the standard process time interval baseline variance threshold σ0 to 4.0 (minutes squared). Twelve sets of time interval data for a batch of fabric between these two processes are collected, and after preprocessing, a time interval difference sequence is obtained. The weighted variance σ2 within the current sliding window is calculated to be 2.5. Substituting into the formula to calculate the exponential part: 0.5 multiplied by (2.5 minus 4.0) equals -0.75. e raised to the power of -0.75 is approximately 0.472. The denominator is 1 plus 0.472, which equals 1.472. Finally, the time interval stability coefficient S is calculated to be approximately 0.68. If another parallel path experiences a delay due to equipment failure, its time interval variance σ² surges to 25.0. Substituting this into the formula, the exponential part is calculated as 0.5 multiplied by (25.0 minus 4.0), which equals 10.5. Since e raised to the power of 10.5 is extremely large, the stability coefficient S approaches 0.00003. Based on this, the system determines that the former is a highly reliable causal relationship, while the latter is a low-reliability or spurious relationship. Therefore, when constructing the local anchor subgraph, the former is given a high weight, effectively isolating the interference from abnormal flow paths.

[0151] S4.3: Based on the device communication logs in the candidate causal chain node set, parse the handshake protocol interaction records and count the ratio of successful response times to total request times to calculate the device communication handshake success rate and generate a second-dimensional quantitative evaluation index characterizing the reliability of the hardware link, so as to eliminate the link break noise caused by network fluctuations or device failures.

[0152] The device communication logs of adjacent node pairs in the candidate causal chain node set are parsed to extract the original interaction record sequence containing the request initiation time, response reception time, and status code. Based on the message structure definition of industrial Ethernet or fieldbus protocols, the SYN (synchronization), ACK (acknowledgment), and FIN (end) control bit identifiers in the handshake protocol are identified. Non-standard format or redundant data fields not specified by the protocol are filtered out to generate a standardized handshake event metadata list.

[0153] For the standardized handshake event metadata list, grouping and aggregating according to device unique identifiers and time windows, and counting the total number of communication requests initiated by each predecessor node to the successor node within a preset observation period. Traversing all response records within this period, filtering out interaction records with a success status code and a response latency below a preset timeout threshold, and accumulating the number of valid handshake successes, forming a raw count pair reflecting the quality of a single link interaction.

[0154] A ratio calculation model is constructed by using the number of successful handshakes and the total number of communication requests to quantitatively characterize the stability of the hardware link at the physical connection level. The success rate of device communication handshakes directly maps to the physical health of the data transmission channel between the underlying IoT gateway, PLC controller, and sensors, eliminating occasional packet loss noise caused by network congestion, electromagnetic interference, or loose interfaces.

[0155] The calculated device communication handshake success rate is normalized and mapped to a second-dimensional quantitative evaluation index within the range of 0 to 1. When the handshake success rate is lower than a preset minimum confidence threshold, the causal chain is considered to have a risk of breaking, and it is marked as a low-confidence connection in the edge attributes of the local anchor subgraph. When the handshake success rate is higher than a preset high-confidence threshold, it is marked as a high-confidence connection, thus providing a noise-resistant hardware link reliability basis for subsequent multi-dimensional weight fusion.

[0156] By using the above-mentioned quantitative processing method based on protocol parsing and statistical ratios, the candidate causal chain node set generated in the previous step is transformed into a second-dimensional quantitative evaluation index characterizing the reliability of the hardware link, thereby achieving the expected technical effect of eliminating link break noise caused by network fluctuations or equipment failures.

[0157] For example, in the flow from the hot pressing and setting process to the coating spraying process of flame-retardant clothing fabric, the preceding node is the HP-2000 hot press PLC controller, and the succeeding node is the CS-500 automatic spraying robot gateway. The system is set to observe the entire flow time of the current batch of fabric between these two processes, i.e., 15 minutes. During this period, the hot press PLC sends a status heartbeat and material placement signal to the spraying robot gateway every 2 seconds, initiating a total of 450 communication requests. Due to electromagnetic interference generated by the starting of large motors in the workshop, 12 requests did not receive a response, 8 responses timed out (exceeding 500ms), and 5 responses had incorrect status codes. The system analyzes the communication logs and identifies 425 successful handshakes. Substituting into the formula, the equipment communication handshake success rate P is calculated. handshake = 425 / 450 ≈ 0.944. The minimum confidence threshold is set to 0.85, and the high confidence threshold is set to 0.95. Since 0.944 falls between the two, the system normalizes this indicator to 0.944 and marks this causal chain as a medium-confidence connection. If, at this time, in another parallel path, the number of successful connections is only 300 due to poor network cable contact, its handshake success rate is 0.667, which is lower than the minimum confidence threshold. The system automatically marks this path as low-confidence and significantly reduces its weight in the subsequent weighted fusion step S4.5, thereby effectively isolating false tracing paths caused by hardware failures.

[0158] S4.4: Using the operator biometric hash fragments bound in the candidate causal chain node set, execute identity consistency comparison logic to verify the continuity and identity of the qualifications of the operator subject of the predecessor node and the successor node, calculate the operator identity consistency score, and generate a third-dimensional quantitative evaluation index that represents the compliance of human operation, so as to ensure that the transfer process complies with the special protection qualification constraints.

[0159] Extract the operator biometric hash fragments and corresponding special protection qualification metadata bound to the predecessor and successor nodes in the candidate causal chain node set to construct the original comparison dataset for identity consistency verification. This dataset contains the operator's unique identifier, qualification certificate validity period, authorized work scope, and biometric fingerprint digest, serving as the basic input for subsequent logical verification.

[0160] A precise matching operation is performed on the operator biometric hash fragments of the predecessor and successor nodes to determine whether the two nodes are operated by the same physical entity. If the hash values ​​are completely identical, they are marked as operating by the same entity; if they are inconsistent, a cross-entity qualification continuity verification process is initiated to distinguish between normal job rotation and unauthorized proxy operation.

[0161] In scenarios where different operating entities are identified or where further verification of qualification continuity is required, the qualification cancellation or transfer records of the predecessor node operator are retrieved and aligned with the qualification activation or acceptance records of the successor node operator within a time window. The time interval between the two events is calculated. If the interval is less than a preset handover threshold and the qualification level satisfies the inheritance relationship, the qualification transfer is considered to be continuous.

[0162] Verify whether the special protective equipment (SPE) qualifications held by operators at both the predecessor and successor nodes cover the safety requirements of the current process. By querying the qualification database, confirm that operators possess the legal permits to perform high-risk processes such as flame-retardant fabric hot pressing and coating spraying. If any operator's qualification is missing or expired, set the compliance score for that dimension to zero and mark it as a high-risk failure point.

[0163] Based on the identity determination results, qualification continuity status, and qualification coverage, a weighted scoring model is used to calculate the operator identity consistency score. This model comprehensively considers the deterministic weight of biometric matching, the temporal tightness weight of qualification transfer, and the matching degree weight of qualification level to generate quantitative compliance evaluation indicators.

[0164] The calculated operator identity consistency score is mapped to a range of 0 to 1, generating a third-dimensional quantitative evaluation index that characterizes the compliance of human operations. This index directly reflects the legality and consistency of personnel operations during the workflow, and is used to eliminate false causal relationships caused by illegal substitution or unqualified personnel.

[0165] Through the above-mentioned biometric matching, qualification continuity verification and coverage assessment, the candidate causal chain node set of the previous step is transformed into a quantitative evaluation index containing the operator identity consistency score, which realizes the accurate measurement of the compliance of human operation in the process of flame retardant clothing fabric circulation and ensures the credibility of personnel elements in the traceability topology.

[0166] For example, in the flow traceability from the padding process to the drying process of a flame-retardant aramid fabric production line, the system extracts the biometric hash value of operator A at the padding node. A and the biometric hash value of drying node operator B. B After comparison, the hash... A With Hash BInconsistency triggers cross-entity verification. The system retrieves that operator A completed the qualification transfer at time T1, and operator B received the qualification at time T2. The time interval Δt = T2 - T1 is 5 minutes, which is less than the preset transfer threshold of 10 minutes. Therefore, the qualification continuity coefficient C is calculated to be 0.95. Further verification reveals that both operators A and B hold "Special Flame Retardant Treatment Operation Certificates" and the certificates are valid, covering the current impregnation and drying processes. Therefore, the qualification coverage coefficient E is 1. Substituting into the formula, let w1 = 0.6, w2 = 0.4, and I = 0 due to biometric mismatch. Then, the operator identity consistency score S = 0.6 * 0 + 0.4 * 0.95 * 1 = 0.38. If in another scenario, the previous and subsequent nodes belong to the same operator C, Hash C If a match is found, I=1. At this point, regardless of the continuity of qualifications, S = 0.6 * 1 + 0.4 * C * E, with a maximum of 1.0. This score of 0.38 is marked as medium confidence, suggesting that the strength of the causal relationship needs to be comprehensively judged in conjunction with other dimensions such as the success rate of device communication handshakes. This effectively identifies potential risks of unauthorized job rotation and significantly improves the accuracy of personnel associations in the traceability graph.

[0167] S4.5: Based on the time interval stability coefficient, device communication handshake success rate and operator identity consistency score, generate a comprehensive causal confidence value, and attach this value as an attribute label to the connection edge of the candidate causal chain node set. Output a local anchor subgraph labeled with causal confidence to complete the construction of a topology unit with error correction capability.

[0168] The time interval stability coefficient generated by the preceding step S4.2, the device communication handshake success rate generated by S4.3, and the operator identity consistency score generated by S4.4 are received as the basic input data for constructing the causal confidence of the local anchor subgraph.

[0169] The quantitative evaluation indicators of the above three dimensions are subjected to Z-Score standardization to eliminate the numerical distribution bias caused by the difference in physical dimensions. Each indicator is mapped to a standard normal distribution space with a mean of 0 and a standard deviation of 1, ensuring that signals of different properties are comparable in subsequent fusion operations.

[0170] Based on the coupling characteristics of the five elements of 'human, machine, material, method, and environment' in the production process of flame-retardant clothing fabrics, a dynamic weight allocation model is constructed. This model automatically adjusts the contribution weight of each dimension according to the semantic strength level of the current process node. For key processes with high safety levels, a higher weight coefficient is assigned to the operator's identity consistency to strengthen compliance constraints.

[0171] The weighted geometric mean is used to calculate the overall causal confidence score, which better reflects the weakest link effect than the arithmetic mean. That is, a low confidence score in any dimension will significantly lower the overall score, thereby effectively identifying and suppressing weak links with potential risks.

[0172] The calculated comprehensive causal confidence score is used as an attribute label and attached to the directed connection edges of adjacent node pairs in the candidate causal chain node set, forming a topological edge structure with confidence weights.

[0173] Traverse all anchors in the dynamic semantic anchor set, repeatedly perform the confidence calculation and edge label mounting operations described above, and integrate all local causal chains into a subgraph structure with the anchors as the center and connected radially.

[0174] The generated local anchor subgraph is subjected to connectivity verification, and isolated nodes or broken edges with confidence scores below a preset threshold (such as 0.6) are removed to ensure the tightness and effectiveness of the internal logical relationships of the subgraph.

[0175] By using multi-dimensional weighted fusion and geometric average processing, the scattered time, communication and personnel indicators from the previous step are transformed into comprehensive causal confidence data with unified dimensions. This enables quantitative representation of causal relationships and noise filtering in local topological structures, and constructs traceability topological units with adaptive error correction capabilities.

[0176] Step S5: For scenarios where multiple parallel paths in the local anchor subgraph converge at the same dynamic semantic anchor point, each converging path is identified as an independent causal flow and stored in an independent namespace. Specifically, this includes: S5.1: Perform flow direction feature extraction processing on multiple parallel path data converging to the same dynamic semantic anchor point in the local anchor subgraph. Based on the differences in the timestamp sequence of path nodes and the uniqueness of the device communication handshake identifier, generate an independent causal flow identifier set that represents the independent evolution characteristics of each path.

[0177] Obtain local anchor subgraph data labeled with causal confidence scores. This subgraph contains multiple predecessor path nodes converging to the same dynamic semantic anchor point and their associated temporal and communication metadata. For the parallel path branching structure existing in the local anchor subgraph, extract the original timestamp sequence of all nodes on each path and construct a multi-dimensional temporal feature matrix to characterize the evolution trajectory of each path in the physical time dimension.

[0178] The time difference between adjacent nodes is calculated on the multidimensional time feature matrix to generate a time interval sequence within each path. The fluctuation of the time interval is evaluated using the sliding window standard deviation to quantify the temporal stability of the path flow and eliminate abnormal timing jitter caused by equipment failure or human intervention.

[0179] Based on the time series stability assessment results, the timestamp sequences of each path are normalized to eliminate the dimensional influence caused by the inherent time differences in different process steps, and a standardized time feature vector is generated to ensure that subsequent difference comparisons are conducted under the same benchmark.

[0180] Calculate the Euclidean distance between the standardized time feature vectors of any two parallel paths to construct a measure of temporal difference between paths. This measure reflects the degree of deviation between the two paths in terms of flow rhythm and serves as the first criterion for distinguishing independent causal flows.

[0181] The device communication logs of each node in the local anchor subgraph are extracted synchronously, and the handshake protocol identifier, session ID and data packet sequence number are parsed to construct the device communication fingerprint sequence to characterize the uniqueness of the interaction at the hardware link level.

[0182] The device communication fingerprint sequence is subjected to hash mapping processing to generate a fixed-length communication handshake identifier. The uniqueness ratio of consecutive handshake identifiers on each path is counted, the entropy value of the path communication link is calculated, and the randomness and unpredictability of communication behavior are quantified.

[0183] By combining the temporal difference metric between the paths with the path communication link entropy, a two-dimensional feature fusion model is constructed. A comprehensive flow direction feature index is generated by weighted summation. This index comprehensively reflects the independent evolution characteristics of the paths in the temporal and spatial communication dimensions.

[0184] A clustering threshold is set for the flow direction feature index. Unsupervised clustering analysis is performed on all precursor paths that converge to the same dynamic semantic anchor point. Paths with flow direction feature indices that differ from the threshold are divided into different clusters, with each cluster representing an independent causal evolution direction.

[0185] Each identified independent cluster is assigned a unique causal flow identifier, which is composed of an anchor spatiotemporal fingerprint fragment, a cluster number, and a generation timestamp, ensuring the global uniqueness and traceability of the identifier.

[0186] All path node data that are determined to be independent causal flows are bound to their corresponding causal flow identifiers to form structured data units, thus completing the semantic mapping from the original topological nodes to the logical causal flows.

[0187] By using the above-mentioned flow feature extraction and clustering identifier generation processing methods, the fuzzy parallel local anchor point subgraphs in the previous step are transformed into a set of independent causal flow identifiers with clear independent evolution characteristics. This achieves the initial decoupling and clarification of node affiliation relationships in multi-path parallel scenarios, laying the data foundation for subsequent namespace isolated storage.

[0188] For example, in a parallel process of impregnation and setting on a flame-retardant clothing fabric production line, a local anchor point subgraph shows three precursor paths converging to the dynamic semantic anchor point "high-temperature setting". The system extracts the timestamp sequences of five nodes on each of these three paths: the standard deviation of the time interval for path A is 0.5 seconds, for path B it is 0.6 seconds, and for path C it is 2.1 seconds. After normalization, the temporal difference metric for paths A and B is calculated to be 0.12, for paths A and C it is 1.85, and for paths B and C it is 1.78. Simultaneously, the equipment communication logs are analyzed, and the entropy values ​​of the communication handshake identifiers for paths A and B are 0.85 and 0.82, respectively, while the entropy value for path C is 0.45 (indicating retransmission). The temporal difference threshold is set to 1.0, and the communication entropy difference threshold is set to 0.2. According to the two-dimensional feature fusion model, paths A and B have relatively small differences in their comprehensive flow characteristic indices and are clustered into one class; path C differs significantly from the former two and is clustered into a separate class. The system assigns the causal flow identifier "CF-20231027-001" to the first type and "CF-20231027-002" to the second type. The final set of independent causal flow identifiers contains two elements, corresponding to the normal parallel production lines A / B and the abnormal re-inspection line C, respectively, realizing the logical separation of physical paths.

[0189] S5.2: Based on the set of independent causal stream identifiers, perform namespace isolation allocation operation to construct an independent namespace container with unique addressing capability for each independent causal stream, and generate isolated causal stream storage units that carry data in a single direction and do not interfere with each other.

[0190] Receive the set of independent causal flow identifiers generated in step S5.1. This set contains unique flow direction feature IDs of multiple parallel paths converging to the same dynamic semantic anchor point, and serves as the input data source for namespace-isolated allocation.

[0191] Traverse each independent causal stream identifier in the set of independent causal stream identifiers, extract its corresponding spatiotemporal fingerprint fragment and device communication handshake sequence number, and construct the original key-value pair data structure for hash mapping calculation.

[0192] A consistent hashing algorithm is used to hash the original key-value pairs, mapping discrete path identifiers to a predefined ring hash space. This ensures that the data migration range is minimized when adding or removing causal streams, and generates an initial hash slot allocation index.

[0193] Based on the initial hash slot allocation index, a logically isolated storage container is instantiated for each independent causal flow. The container maintains an independent node attribute table, edge relationship graph and metadata index area, forming a physically continuous but logically mutually exclusive data storage unit.

[0194] A namespace descriptor is written to the header of the isolated causal stream storage unit. This descriptor contains the stream ID, creation timestamp, associated dynamic semantic anchor fingerprint, and access control list, establishing the unique addressing capability and security boundary of the storage unit.

[0195] The node time-series data, device status vectors, and operator identity hashes of each parallel path extracted in S5.1 are written into the corresponding isolated causal flow storage units according to their respective independent causal flow identifiers, thereby realizing the split storage of multi-source heterogeneous data.

[0196] Integrity checks are performed on the isolated causal stream storage units after writing, the Merkle root hash value of the stored content is calculated, and the check value is bound to the namespace descriptor to prevent data tampering or crosstalk during subsequent processing.

[0197] Through the hash mapping and logical isolation mechanism described above, the multiple parallel paths identified in the previous step are transformed into isolated causal stream storage units with unique addressing capabilities and no interference between them. This enables the structural decoupling and independent management of multi-path data in complex flow scenarios, providing a clean data foundation for subsequent process compatibility analysis.

[0198] S5.3: Perform process rule matching analysis on the parallel path node attribute information stored in the isolated causal flow storage unit, extract key constraint parameters including consistency of quality inspection standards, sharing of raw material batches and coverage of operation qualifications, and generate process compatibility label vectors describing the logical compatibility relationship between different causal flows.

[0199] The system reads attribute data of parallel path nodes from the isolated causal flow storage unit. This data includes quality inspection report metadata, raw material batch code sequences, and operator qualification certificate hash values. The quality inspection report metadata is structured and parsed to extract the flame retardant performance test standard version number, wash fastness grade judgment threshold, and breaking strength test method identifier, constructing a quality inspection standard feature vector. The quality inspection standard feature vectors in different independent causal flows are compared bit-by-bit to calculate the consistency between standard version and judgment threshold, generating a quality inspection consistency score characterizing the homogeneity of quality control. The system retrieves raw material warehousing records associated with the starting nodes of each independent causal flow, extracting fiber blending ratio codes, yarn supplier IDs, and dye batch numbers to form a raw material traceability fingerprint set. The number of overlapping elements in the raw material traceability fingerprints between any two independent causal flows is calculated using the set intersection, and combined with the raw material input weight ratio, a raw material batch sharing coefficient is calculated to quantify the homogeneity of material sources. The system parses the operator biometric hash fragments bound to all process nodes in each independent causal flow and maps them to the enterprise human resources database to obtain the corresponding special protective operation qualification category and validity status. The system statistically analyzes the set of high-risk process qualification types covered in each independent causal flow, calculates the union coverage rate and intersection matching degree of the qualification sets between any two flows, generates an operational qualification coverage index, and assesses the complementary or redundant state of personnel compliance under multi-path parallelism. The quality inspection consistency score, raw material batch sharing coefficient, and operational qualification coverage index are normalized to eliminate the impact of dimensional differences on subsequent weight allocation. Based on the flame-retardant clothing production process specifications, initial allocation ratios for quality inspection standard weights, raw material homogeneity weights, and qualification coverage weights are set, and a weighted summation is used to calculate the comprehensive process compatibility score. Based on the high and low ranges of the comprehensive process compatibility score, discrete process compatibility labels are generated, including four status identifiers: 'fully compatible', 'partially compatible', 'potentially conflicting', and 'mutually exclusive'. The above three types of original index values, comprehensive scores, and final labels are encapsulated into a fixed-length array to generate a process compatibility label vector describing the logical compatibility relationship between two specific independent causal flows. By extracting and weighted fusion of multidimensional constraint parameters, the parallel path data stored in isolation in the previous step is transformed into process compatibility label vectors that characterize the logical relationship between flows. This enables the quantitative self-clarification of node affiliation in complex parallel scenarios and provides standardized input for the subsequent construction of cross-flow correlation matrices.

[0200] For example, in a scenario where a batch of aramid flame-retardant fabric is diverted to two parallel drying paths after the padding process, the system extracts attribute data for path A and path B from the isolated causal flow storage unit. Path A executes a 'high-temperature rapid drying' process, while path B executes a 'low-temperature slow drying' process. From a quality inspection perspective, analysis reveals that both paths comply with the GB / T 5455-2014 flame retardant testing standard, and both have a wash fastness threshold of level 4. The quality inspection standard feature vectors match perfectly, resulting in a quality inspection consistency score Sq of 1.0. From a raw material perspective, path A uses fabric raw material batch number 'AR-2023-05-A', while path B uses 'AR-2023-05-B'. Verification shows that both originate from the same supplier and the same batch of aramid raw yarn, differing only in dyeing batch. The raw material traceability fingerprint overlap is 85%, and combined with the input weight ratio, the raw material batch sharing coefficient Sr is calculated to be 0.85. In terms of qualifications, the operator of path A holds a 'High-Temperature Equipment Operation Certificate' and a 'Hazardous Chemicals Handling Certificate', while the operator of path B holds a 'Low-Temperature Equipment Operation Certificate' and a 'Hazardous Chemicals Handling Certificate'. The union of their qualifications covers all the high-risk qualifications required for this process, with an intersection matching degree of 60%. After normalization, the operation qualification coverage index So is 0.9. We set weight coefficients w1=0.4, w2=0.3, and w3=0.3, and substituted them into the formula to calculate the comprehensive process compatibility score C = 0.4×1.0 + 0.3×0.85 + 0.3×0.9 = 0.4 + 0.255 + 0.27 = 0.925. Since the score of 0.925 is higher than the preset high compatibility threshold of 0.85, the system generates a 'Fully Compatible' label. The final output process compatibility label vector is [1.0, 0.85, 0.9, 0.925, 'Fully Compatible']. This vector accurately reflects that although the two parallel paths have different process parameters, they are highly homogeneous and complementary in terms of quality standards, raw material sources and personnel qualifications. This significantly improves the accuracy of identifying the parallel nodes in the subsequent traceability map and avoids erroneous segmentation caused by process differences.

[0201] S5.4: Based on the process compatibility tag vector, perform a multidimensional relation matrix construction operation to calculate the compatibility value and conflict probability value between any two independent causal flows, and generate a cross-flow correlation matrix that records the topological correlation state between all flows.

[0202] Receive the process compatibility label vector generated in step S5.3. This vector contains quantitative indicators in three dimensions: quality inspection standard consistency coefficient, raw material batch sharing index, and operation qualification coverage overlap rate. These indicators serve as the basic input data for constructing the cross-flow correlation matrix.

[0203] Standardization mapping is performed on each dimension index in the process compatibility label vector to uniformly map compatibility parameters of different dimensions to the [0,1] interval, eliminating numerical deviations caused by differences in the physical meaning of the indexes, and generating a standardized inter-flow compatibility feature matrix.

[0204] Based on the standardized inter-flow compatibility feature matrix, a weighted adjacency matrix model is constructed, where the rows and columns of the matrix correspond to the identifiers of independent causal flows that converge at the same dynamic semantic anchor point, and the values ​​of the matrix elements represent the strength of the topological association between any two causal flows.

[0205] The compatibility score between any two independent causal flows i and j is calculated using a weighted summation method.

[0206] Based on the calculated compatibility value, the conflict probability between any two independent causal flows is further derived. The conflict probability is negatively correlated with the compatibility. The conflict probability is calculated using the following formula:

[0207] in, Let i be the probability value of the conflict between causal flows i and j. This is the compatibility value between causal flow i and causal flow j. The closer this value is to 1, the more significant the differences are between the two paths in terms of process logic, material source, or personnel qualifications, and careful differentiation is required during tracing. The closer this value is to 0, the more homogeneous the two paths are, and they can be regarded as equivalent tracing paths.

[0208] The calculated compatibility values ​​and conflict probability values ​​are filled into the corresponding positions of the weighted adjacency matrix to form a symmetric matrix structure. The diagonal elements of the matrix are set to 1, indicating that the causal flow is completely compatible with itself.

[0209] The generated weighted adjacency matrix is ​​sparsified, and a conflict probability threshold is set. When the conflict probability value of any two causal flows exceeds the threshold, the corresponding compatibility value in the matrix is ​​set to zero to cut off the low-confidence inter-flow association and reduce noise interference in subsequent tracing calculations.

[0210] Through the above weighted adjacency matrix construction operation, the discrete process compatibility label generated in the previous step is transformed into a structured cross-flow correlation matrix. This matrix fully records the topological correlation status and logical compatibility between each independent causal flow in a multi-path parallel scenario, realizing a quantitative description and conflict warning of complex intertwined flow relationships, and providing accurate data support for the selection of main and auxiliary flows in subsequent source tracing paths.

[0211] S5.5: The cross-flow association matrix is ​​structured and encapsulated, and the matrix data is bidirectionally indexed and bound to the corresponding independent namespace container to generate a multi-path parallel tracing graph substructure that supports fast backtracking query and dynamic update, so as to complete the self-clarification of node ownership in complex flow scenarios.

[0212] Step S6: Based on the most recent dynamic semantic anchor event of the target batch, locate the starting position, expand backward along the independent causal flow to the starting dynamic semantic anchor, and filter out the main tracing flow to generate a candidate set of tracing paths containing auxiliary verification flows. Specifically, this includes: S6.1: Obtain the real-time flow identifier and dynamic semantic anchor set of the target batch, perform location processing on the most recently triggered dynamic semantic anchor, and generate the initial anchor coordinates representing the starting point of the trace and the associated independent causal flow namespace list.

[0213] The S6.1 sub-step aims to address how to accurately locate the starting point of a target batch from a massive number of dynamic semantic anchors in complex parallel workflow scenarios, and to clarify its independent causal flow namespace, providing precise coordinate references and context boundaries for subsequent reverse path development. This step follows the multi-path parallel tracing graph sub-structure generated in S5, utilizing the uniqueness and temporal correlation of spatiotemporal fingerprints to achieve instantaneous locking of the current state of the target batch.

[0214] The system receives a real-time flow identifier for the target batch from the upper-layer business system. This identifier includes a unique batch code, the current process node ID, and the latest event trigger timestamp, serving as the initial retrieval key for traceability queries. Based on this unique batch code, the system performs a full scan or hash lookup in the dynamic semantic anchor set index to extract all dynamic semantic anchor records that have historically interacted with this batch, forming a candidate anchor time series table.

[0215] For each dynamic semantic anchor in the candidate anchor time series list, its bound final unique spatiotemporal fingerprint is parsed, and the embedded high-precision timestamp data is extracted. Using the maximum time value as the filter, the timestamps of all anchors in the list are compared, and the dynamic semantic anchor corresponding to the timestamp with the largest value is identified. This anchor is determined to be the most recently triggered key process node of the target batch, i.e., the starting physical location of the current traceability.

[0216] Obtain the local anchor subgraph structure information constructed in step S5 for the initial dynamic semantic anchor, and read the set of independent causal flow identifiers directly associated with it. Based on the set of independent causal flow identifiers, query the pre-established namespace mapping table, retrieve the independent namespace container address and metadata description corresponding to each independent causal flow identifier, and generate a list of independent causal flow namespaces containing namespace IDs, storage paths, and flow attribute characteristics.

[0217] The initial dynamic physical location coordinates are structurally encapsulated with a list of independent causal flow namespaces to construct an initial anchor point coordinate object. This object not only contains a spatiotemporal fingerprint reference to the starting node, but also carries address pointers to all possible predecessor path branches, ensuring that subsequent reverse traversal can cover all potential parallel causal links and avoid missing any possible tracing path.

[0218] Through the above spatiotemporal fingerprint matching and namespace association processing, the discrete batch flow state is transformed into initial anchor point coordinates with clear topological boundaries and an associated list of independent causal flow namespaces. This achieves a technological leap from global fuzzy search to local precise positioning, laying a solid data foundation for the efficient and unambiguous reverse causal chain traversal in step S6.2.

[0219] S6.2: Based on the initial anchor point coordinates and the independent causal flow namespace list, perform a reverse causal chain traversal operation, and retrieve the predecessor node in reverse along the causal confidence edge in each independent causal flow to construct a full backtracking path tree structure covering the current node to the starting dynamic semantic anchor point.

[0220] Based on the initial anchor coordinates and the associated list of independent causal flow namespaces, the reverse causal chain traversal engine is initiated, an empty full backtracking path tree data structure is initialized, and the current traversal level is set as the root node level. The node index table within each namespace container in the independent causal flow namespace list is read, and a set of candidate node IDs with direct predecessor relationships to the current dynamic semantic anchor is extracted as the first-level queue of nodes to be processed. For each candidate node in the queue of nodes to be processed, its corresponding causal confidence edge attribute in the local anchor subgraph is retrieved, and the time interval stability coefficient, device communication handshake success rate, and operator identity consistency score of that edge are obtained. The comprehensive confidence weight of each incident edge is calculated using weighted fusion, and weakly related edges below a preset truncation threshold are pruned, retaining only high-confidence causal connections to eliminate false path branches caused by noise interference. For the retained valid predecessor nodes, their bound spatiotemporal fingerprint information is parsed, and the device ID, operation timestamp, and operator biometric hash fragment are extracted to construct a node state snapshot containing complete context information. The system determines whether the current node is the starting dynamic semantic anchor or a source process node without more predecessor nodes. If so, the node is marked as a leaf node and the recursive search of that branch is terminated. If not, the node is added to the full backtracking path tree structure, and its predecessor node set is used as the next level of pending node queue, continuing the reverse retrieval logic. During the recursive traversal, the path length and cumulative causal confidence from the current dynamic semantic anchor to each ancestor node are recorded in real time, forming path metadata with depth information and accumulated weight values. When the reverse traversal of all independent causal flows is completed, the subtree structures generated by each flow are integrated. By merging shared nodes, overlapping common process nodes in different causal flows are identified, cross-flow node mapping relationships are established, and finally, a full backtracking path tree structure covering the current node to the starting dynamic semantic anchor is generated. By using the above-mentioned reverse causal chain traversal and multi-flow subtree integration processing method, discrete independent causal flow data is transformed into a full backtracking path tree with hierarchical structure and weight attributes, realizing the automated reconstruction and structured expression of the complete tracing chain in complex parallel flow scenarios.

[0221] S6.3: Perform process type clustering on the nodes in the full backtracking path tree structure, extract the operation parameter sequence of nodes of the same type, and use statistical distribution to calculate and generate the operation parameter distribution interval feature vector characterizing the stability of each independent causal flow process.

[0222] The node data in the full backtracking path tree structure is clustered by process type to extract the operation parameter sequences of nodes with the same type. Based on the full backtracking path tree structure built by S6.2, all process nodes in each independent causal flow are traversed, and the process type identifier in the node attributes is read. Hash mapping is used to group nodes with the same process type identifier into the same logical cluster, forming a node set containing specific process categories such as hot pressing, flame retardant coating spraying, and washability testing. For each process type cluster, the nodes are arranged in reverse order of timestamp to ensure the consistency of the operation parameter sequence with the tracing direction, generating a standardized isomorphic process node list.

[0223] Key operational parameter sequences are extracted from a list of isomorphic process nodes to construct a multidimensional process state vector. For each type of process, key quality indicators are defined according to the flame-retardant clothing fabric production process standards, including physical quantities such as hot-pressing temperature, pressure value, residence time, coating thickness, and drying rate. Real-time monitoring values ​​of the above physical quantities are parsed from node logs, and null values ​​and abnormal jump points are removed to form a continuous operational parameter time series. For missing data points, linear interpolation is used to complete the sequence based on adjacent valid nodes to ensure the integrity of the parameter sequence. The completed parameter sequence is encapsulated into a multidimensional vector, where each dimension corresponds to a specific process control parameter, generating an original process feature matrix characterizing the actual execution state of the process in a specific causal flow.

[0224] The original process characteristic matrix is ​​processed using statistical distribution calculations to quantify the process stability of each independent causal flow. For the parameter sequence of isomorphic process nodes in each independent causal flow, statistical moments such as mean, standard deviation, skewness, and kurtosis are calculated. The dispersion index of the operating parameters is calculated using the following formula to characterize the volatility of process execution:

[0225] Where D is the dispersion index, N is the number of isomorphic process nodes in the causal flow, and x i Let be the measured values ​​of the operation parameters for the i-th node. This is the arithmetic mean of the parameter sequence. This formula quantifies the degree to which the parameter deviates from the central tendency by calculating the sample variance; a smaller D value indicates more stable process control of the causal flow at that stage.

[0226] Further calculations were performed to determine the morphological characteristics of the parameter distribution and to assess the standardization of process execution. The skewness coefficient was calculated using the following formula to determine whether there was a systematic deviation in the parameter distribution:

[0227] Where S is the skewness coefficient and σ is the standard deviation. An S value close to 0 indicates a symmetrical distribution, a positive value indicates right skewness, and a negative value indicates left skewness, used to identify the existence of unidirectional process drift. The kurtosis coefficient is also calculated to assess the thickness of the tails in the parameter distribution, in order to identify the impact of occasional extreme outliers on process stability.

[0228] By integrating the dispersion index, skewness coefficient, and kurtosis coefficient, a feature vector representing the distribution interval of operational parameters characterizing the process stability of each independent causal flow is generated. The calculated statistics are then Z-score standardized to eliminate differences in the dimensions of different parameters. The standardized statistics are then concatenated according to process type to form a high-dimensional feature vector. This vector not only contains information on the central tendency of the parameters but also fully preserves the distribution pattern and fluctuation range information, accurately characterizing the process consistency level of each independent causal flow at a specific process. Through the above processing method, the full backtracking path tree structure of the previous step is transformed into a quantifiable feature vector representing the distribution interval of operational parameters, achieving a refined characterization of the process stability of each parallel path and providing accurate data support for the subsequent selection of the main backtracking flow.

[0229] S6.4: Based on the feature vector of the distribution range of the operation parameters, perform multi-flow parameter volatility comparison and qualification coverage assessment, and use a weighted sorting strategy to select the main path with the smallest parameter volatility and the most complete operator qualification coverage, so as to establish a unique main trace flow and mark the remaining paths as auxiliary verification flows.

[0230] Receive the feature vector of the distribution range of operation parameters generated in step S6.3. This vector contains the statistical distribution data of each independent causal flow in the full backtracking path tree at nodes of the same process type, and serves as the input basis for multi-flow parameter volatility comparison and qualification coverage assessment.

[0231] For each independent causal flow in the feature vector of the operating parameter distribution interval, the standard deviation sequence and range sequence of its key process parameters are extracted to construct an original set of fluctuation indicators characterizing process stability. For key processes in flame-retardant clothing fabric production, such as hot pressing and coating impregnation, the dispersion of physical quantities such as temperature, pressure, and speed is calculated to form a multidimensional fluctuation feature matrix.

[0232] An operator qualification coverage assessment mechanism is introduced, which retrieves the operator biometric hash fragments corresponding to all nodes in each independent causal flow and matches and verifies them with the special protection qualification database. The proportion of operator nodes with valid special operation qualifications in each causal flow is statistically analyzed, as well as whether high-risk procedures (such as chemical additive preparation) are performed by personnel with the highest level of qualification, generating a qualification compliance score vector.

[0233] A comprehensive evaluation function for screening main causal flows is constructed, which weights and integrates process stability indicators and qualification compliance indicators. The comprehensive traceability confidence score of the i-th independent causal flow is calculated using the following formula. i :

[0234] Where, σ i The weighted standard deviation of the key process parameter for the i-th causal flow reflects the level of parameter fluctuation; Q i Let α be the qualification compliance score for the i-th causal flow, with a value ranging from 0 to 1; α and β are the fluctuation suppression weight and qualification enhancement weight, respectively, and satisfy α+β=1.

[0235] The calculated confidence score for the comprehensive tracing of each independent causal flow. i Perform a descending sorting operation to identify the causal flow with the highest confidence as the primary tracing flow candidate. At the same time, set a confidence threshold interval. If the difference between the second-highest confidence flow and the highest confidence flow is less than a preset tolerance, then the second-highest confidence flow is marked as a strong auxiliary verification flow; otherwise, it is marked as a weak auxiliary verification flow or discarded.

[0236] The selected main tracing flow undergoes integrity verification to check whether it covers all key nodes of the entire lifecycle from the initial dynamic semantic anchor point to the most recent anchor event of the current target batch. If any nodes are missing, the context information of the missing nodes is extracted from the causal flow marked as a strong auxiliary verification flow, and the structural gaps in the main tracing flow are filled using spatiotemporal fingerprint alignment technology to ensure the logical continuity of the main tracing chain.

[0237] The final established main traceability flow data is structured and encapsulated, including the main path node sequence, the mean and standard deviation of key process parameters, and the entire qualification and compliance certification chain. The remaining tagged auxiliary verification flows are encapsulated into supporting data packages, recording their differences from the main traceability flow and compatibility tags. Through this processing, the operational parameter distribution characteristics of the previous step are transformed into a candidate set of traceability paths with clear primary and secondary relationships. This enables the uniqueness establishment and credibility stratification of traceability paths in complex parallel scenarios, significantly improving the accuracy and legal validity of traceability results.

[0238] S6.5: Integrate the main tracing stream and the marked auxiliary verification stream, perform data structure encapsulation processing, embed the main tracing stream as the core tracing chain and the auxiliary verification stream as credibility verification information into a unified data format to generate a source path candidate set containing complete context information.

[0239] Step S7: Continuously monitor the failure signals of each dynamic semantic anchor point in the dynamic semantic anchor point set. The failure signals include equipment downtime, operator absence status, and the number of times the same batch of triggers is missing, and generate a real-time status marker characterizing the effectiveness of the anchor point. Specifically, this includes: S7.1: Obtain the device operation log stream and operator biometric hash fragment corresponding to each anchor point in the dynamic semantic anchor point set, use the time series sliding window to truncate the continuous unresponsive period of the device, and calculate the off-duty duration based on the operator positioning beacon signal to generate the anchor point basic failure factor set containing device downtime data and operator off-duty status data.

[0240] S7.2: Based on the anchor point basic failure factor set, extract the process trigger record sequence of the same batch of fabric within a preset time period, use frequency statistics to count the cumulative number of times a specific dynamic semantic anchor point is not triggered, and combine the equipment downtime data and operator off-duty status data generated in the previous steps to construct a multi-dimensional original failure feature vector characterizing the degree of anchor point abnormality.

[0241] Based on the anchor point-based failure factor set, the sequence of process trigger records for the same batch of fabric within a preset time period is extracted. Using the batch number as the primary key index, all historical process logs associated with the batch are retrieved from the distributed time-series database. Valid trigger events with timestamps within the current monitoring window are filtered out, forming a raw trigger event queue arranged in ascending chronological order.

[0242] The original trigger event queue is deduplicated and its integrity is verified. The device ID of each trigger event is compared with the device whitelist defined by the dynamic semantic anchor point, and redundant records caused by sensor false alarms are removed. The event payload is checked to see if it contains a complete snapshot of process parameters. Records missing key parameters are marked as invalid triggers and removed from the queue, generating a cleaned standard trigger event sequence.

[0243] The system counts the cumulative number of times a specific dynamic semantic anchor point is not triggered within a preset time period. It then obtains the standard process cycle model corresponding to the dynamic semantic anchor point and calculates the theoretically expected number of triggers within the current monitoring window. The expected number is subtracted from the actual number of valid triggers detected to obtain the absolute missing frequency. If the actual number of triggers is zero and the expected number is greater than zero, a complete missed trigger is directly identified, and the maximum missing frequency value is recorded.

[0244] Combining the equipment downtime data and operator absence status data generated in the previous steps, a multi-dimensional original failure feature vector is constructed. The numerical value of the continuous unresponsive equipment time is mapped to the first dimension component, the numerical value of the operator absence duration is mapped to the second dimension component, and the cumulative number of missed triggers in the same batch is mapped to the third dimension component. The above three components are concatenated in a fixed order to form a three-dimensional column vector structure.

[0245] Through the above data processing flow, discrete equipment status, personnel behavior, and process trigger frequency are transformed into structured multidimensional original failure feature vectors, realizing the quantitative representation of the health status of dynamic semantic anchors, and providing a unified data foundation for subsequent normalized weighted fusion and failure determination.

[0246] S7.3: Perform normalization preprocessing on the multidimensional original failure feature vector to eliminate the difference in dimensions, and use a weighted fusion strategy to map the weights of equipment downtime, operator off-duty status and number of times the same batch is triggered to a unified evaluation space, and calculate the anchor point sub-item failure confidence sequence representing the risk level of a single dimension.

[0247] S7.4: Based on the anchor point sub-item failure confidence sequence, apply multi-threshold comparison logic to determine whether each sub-item index exceeds the preset failure threshold, mark the sub-item failure confidence exceeding the threshold as a high-risk failure identifier, and aggregate all high-risk failure identifiers through logical OR operation to generate a preliminary validity judgment result indicating the overall health status of the anchor point.

[0248] The system receives normalized confidence sequences from S7.3 output across three dimensions: equipment downtime, operator absence status, and number of missed triggers in the same batch. A dynamic failure threshold judgment matrix is ​​constructed, and corresponding thresholds are retrieved based on the strong constraint level of the process semantics. Each component is compared with the threshold, generating a Boolean judgment vector; exceeding the threshold marks a high-risk failure. Logical OR aggregation is used: if any component is high-risk, the entire system is judged as failed, generating a preliminary validity judgment flag; if none exceed the threshold, the system is judged as healthy and valid. This achieves rapid binary classification, providing a basis for anchor point degradation, avoiding misjudgments, and ensuring immediate response to severe failures.

[0249] S7.5: Based on the preliminary validity judgment result, the anchor point health status is mapped to a standardized Boolean value or an enumeration type label using a status coding rule, and the final real-time status label representing the anchor point validity is output. This label is directly used as the trigger condition for determining whether to perform anchor point degradation processing and historical subgraph migration operations in subsequent steps.

[0250] Receive the anchor point health status assessment results output from S7.4. Construct a state machine coding mapping table, defining four enumerated states: effective active, sub-health warning, critical failure, and complete failure, and setting state transition rules. If all indicators do not trigger the high-risk flag, map to the effective active state and assign the highest topology weight. If only non-critical indicators issue warnings while critical signals are normal, map to the sub-health warning and reduce the confidence contribution weight. If multiple high-risk or critical signals trigger the failure threshold, force mapping to complete failure and generate a failure event packet containing a timestamp, cause code, and batch ID. Use a finite state automaton to verify the state transition path to prevent frequent jitter. Encapsulate the final state as a real-time state marker containing a status code, confidence decay coefficient, and timestamp, and broadcast it as a standard message to provide deterministic triggering conditions for anchor point degradation and subgraph migration.

[0251] Step S8: If the real-time status flag indicates that a specific dynamic semantic anchor point has failed, then a degradation process is performed to convert the specific dynamic semantic anchor point into a normal node, and its historical local anchor point subgraph is migrated to the knowledge base to be verified. Specifically, this includes: S8.1: Based on the equipment downtime, operator off-duty status and number of missing triggers in the same batch in the real-time status flag, perform failure judgment processing on specific dynamic semantic anchors in the dynamic semantic anchor set to generate anchor failure judgment results containing failure confirmation identifiers.

[0252] Receives dynamic semantic anchor health status markers from S7 output (including equipment downtime, operator absence status, and number of missing triggers in the same batch). Parses downtime: if it exceeds the equipment tolerance threshold, a hardware failure flag is generated. Parses absence status: if it exceeds the interruption time limit or involves unauthorized replacement, a personnel failure flag is generated. Counts missing triggers: if the number is lower than the historical baseline trigger rate, a logical failure flag is generated. Constructs a multi-dimensional judgment matrix; failure of any flag triggers a global failure, generating an anchor failure judgment result object containing a failure confirmation flag, the dominant dimension, and a timestamp, providing a decision-making basis for subsequent topology corrections.

[0253] S8.2: Based on the anchor point failure determination result, perform semantic constraint stripping processing on the unique spatiotemporal fingerprint of the specific dynamic semantic anchor point to generate an ordinary node identity identifier that has lost strong semantic constraints.

[0254] S8.3: Using the ordinary node identity identifier, the causal confidence of the historical local anchor subgraph constructed with the specific dynamic semantic anchor as the center is recalculated to generate skeletonized subgraph data that removes the weighted influence of the original anchor.

[0255] The system receives ordinary nodes that have lost strong semantic constraints and historical local subgraphs centered on failed anchor points (including predecessor and successor nodes and original causal confidence scores). It parses the topology, decomposing the "predecessor-failed anchor-successor" triples into candidate "predecessor-successor" direct connections to be reconstructed. For each pair, it calculates the time interval stability coefficient, device communication handshake success rate, and operator identity consistency score, weighting and fusing these three factors to generate a new comprehensive causal confidence score, replacing the original confidence score that passed through the failed anchor point. It then traverses all affected connection edges, removing the influence of anchor point mediation, generating skeletonized subgraph data without failed anchor points, thus achieving smooth correction of the topology structure.

[0256] S8.4: Based on the skeletonized subgraph data, perform decoupling and reorganization operations on the cross-flow association matrix to generate a flattened flow relationship record that retains only the process compatibility label and removes the independent namespace isolation attribute.

[0257] The system receives skeletonized subgraph data, decouples the cross-flow association matrix, and extracts the process compatibility tags and inter-flow topology states corresponding to the failed dynamic semantic anchors. Based on the tag mapping parsing logic compatibility relationships, it transforms the radial associations centered on the anchors into point-to-point planar topology. It unbinds the namespace from the failed anchors through hash mapping and merges parallel paths into the global flow index. It removes independent namespace identifiers, retains process type, timestamp, and handshake identifier, and directly attaches compatibility tags to adjacent node edges, generating a flattened flow relationship record and eliminating the risk of path breakage caused by anchor failures.

[0258] S8.5: Write the flattened flow relationship record into the temporary storage area of ​​the knowledge base to be verified, and trigger the edge reasoning module to reconstruct the association by combining the newly generated dynamic semantic anchors to complete the adaptive correction of the tracing topology and output the updated tracing map.

[0259] The knowledge base to be verified is an isolated storage area used to temporarily store subgraph data that has been downgraded and migrated due to anchor failure. Its architecture employs a hybrid storage of key-value pairs and graph database: the spatiotemporal fingerprint of the failed anchor is used as the primary key, associated with historical local anchor subgraphs (including predecessor and successor nodes, causal confidence edges, process compatibility tags, etc.); additionally, the failure cause code, migration timestamp, and verification tag are included. The knowledge base supports incremental writing and expired archiving, and provides a query interface for the edge reasoning module, allowing newly generated dynamic semantic anchors to be re-associated with the temporarily stored data in the base through spatiotemporal fingerprint alignment or process compatibility matching. This design ensures the cleanliness of the main traceability topology while preserving historical context for recovery or auditing, serving as a temporary buffer in the system's adaptive correction loop.

[0260] A batch traceability and transfer system for flame-retardant clothing fabrics based on pretreatment is provided, which uses a batch traceability and transfer method based on pretreatment to trace and transfer batches of flame-retardant clothing fabrics.

[0261] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0262] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0263] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment, specifically including: S1: Obtain the original workstation log data during the flow of flame-retardant clothing fabric. The data includes equipment operating status, operator identification, and material physical property information to form an initial flow record set to be processed. S2: Based on the initial flow record set, semantic intensity analysis is performed using three types of dynamic indicators: physical mandatory signals, operational constraint signals, and material intrinsic signals, to generate a multi-dimensional semantic feature vector characterizing the credibility of nodes; S3: Based on the intensity threshold determination results of at least two types of signals in the multidimensional semantic feature vector, select key process nodes with strong semantic constraints and assign them unique spatiotemporal fingerprints to construct a set of dynamic semantic anchor points as the topological skeleton. S4: Taking each dynamic semantic anchor point in the set of dynamic semantic anchor points as the center, extract the predecessor and successor nodes that have a direct causal chain, calculate the weighted values ​​of time interval stability, device communication handshake success rate and operator identity consistency, and generate a local anchor point subgraph labeled with causal confidence. S5: For scenarios where multiple parallel paths in the local anchor subgraph converge at the same dynamic semantic anchor point, each converging path is identified as an independent causal flow and stored in an independent namespace; S6: Based on the most recent dynamic semantic anchor event of the target batch, locate the starting position, expand backward along the independent causal flow to the starting dynamic semantic anchor, filter out the main tracing flow, and generate a candidate set of tracing paths containing auxiliary verification flows.

2. The batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment according to claim 1, characterized in that, Following S6, the following is also included: S7: Continuously monitor the failure signals of each dynamic semantic anchor point in the set of dynamic semantic anchor points. The failure signals include equipment downtime, operator off-duty status and number of times the same batch is triggered and missing. Generate a real-time status marker that represents the effectiveness of the anchor point. S8: If the real-time status flag indicates that a specific dynamic semantic anchor point has failed, then a degradation process is performed to convert the specific dynamic semantic anchor point into a normal node, and its historical local anchor point subgraph is migrated to the knowledge base to be verified.

3. The batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment according to claim 1, characterized in that, S3 specifically includes: The physical mandatory signal component, the operational constraint signal component, and the material intrinsic signal component in the multidimensional semantic feature vector are combined in pairs to generate three types of dual signal strength judgment matrices, so as to establish the preliminary candidate range of nodes to be selected as dynamic semantic anchors. Based on the three types of dual signal strength determination matrices, the real-time signal strength values ​​under each combination are extracted and compared with the preset dynamic threshold range to screen out a list of qualified key process nodes that simultaneously meet the requirements of at least two types of signal strength thresholds, so as to eliminate ordinary flow nodes with weak semantic associations. For each node in the list of qualified key process nodes, its corresponding equipment operation identifier, high-precision timestamp data, and operator biometric hash fragment are collected. The above multi-source heterogeneous information is serialized and spliced ​​to generate the original spatiotemporal fingerprint code. The original spatiotemporal fingerprint code is subjected to irreversible hash encryption transformation and integrity check bit addition processing to generate a final unique spatiotemporal fingerprint with anti-tampering characteristics. The final unique spatiotemporal fingerprint is then used as a primary key index and bound to the corresponding qualified key process node to complete the strong identity solidification of the node. Based on all qualified key process nodes that have been bound with a final unique spatiotemporal fingerprint, they are structured and aggregated according to the time sequence and process logic relationship to build a dynamic semantic anchor set with elastic expansion capability, which serves as the core topological skeleton for subsequent local anchor subgraph generation.

4. The batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment according to claim 3, characterized in that, The three types of dual-signal strength determination matrices include physical and operational combinations, physical and material combinations, and operational and material combinations.

5. The batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment according to claim 3, characterized in that, The original spatiotemporal fingerprint code represents the uniqueness of the node.

6. The batch traceability and transfer method for flame-retardant clothing fabric based on pretreatment according to claim 1, characterized in that, S4 specifically includes: Based on each dynamic semantic anchor in the set of dynamic semantic anchors, the predecessor process node and successor process node with direct physical connection or logical dependency relationship in the initial flow record set are retrieved to generate a candidate causal chain node set containing complete context information, so as to establish the boundary range of the local topology. For adjacent node pairs in the candidate causal chain node set, timestamp sequence data is extracted and sliding window variance analysis is performed to calculate the time interval stability coefficient between adjacent processes and generate a quantitative evaluation index of the first dimension characterizing the temporal continuity in order to exclude false causal associations caused by abnormal waiting or order insertion. Based on the device communication logs in the candidate causal chain node set, the handshake protocol interaction records are parsed and the ratio of successful response times to total request times is calculated. The device communication handshake success rate is calculated, and a second-dimensional quantitative evaluation index characterizing the reliability of the hardware link is generated to eliminate link break noise caused by network fluctuations or device failures. Using the operator biometric hash fragments bound to the candidate causal chain nodes, the identity consistency comparison logic is executed to verify the continuity and identity of the qualifications of the operator subjects of the predecessor node and the successor node, calculate the operator identity consistency score, and generate a third-dimensional quantitative evaluation index that represents the compliance of human operation, so as to ensure that the transfer process complies with the special protection qualification constraints. Based on the time interval stability coefficient, device communication handshake success rate, and operator identity consistency score, a comprehensive causal confidence value is generated and attached as an attribute label to the connection edge of the candidate causal chain node set. A local anchor subgraph is then output to complete the construction of a topology unit with error correction capability.

7. The batch traceability and transfer method for flame-retardant clothing fabric based on pretreatment according to claim 6, characterized in that, The local anchor point sub-icons are labeled with causal confidence levels.

8. The batch traceability and transfer method for flame-retardant clothing fabrics based on pretreatment according to claim 1, characterized in that, S5 specifically includes: The flow direction feature extraction process is performed on multiple parallel path data converging to the same dynamic semantic anchor point in the local anchor subgraph. Based on the differences in the timestamp sequence of path nodes and the uniqueness of the device communication handshake identifier, an independent set of causal flow identifiers is generated. Based on the set of independent causal stream identifiers, a namespace isolation allocation operation is performed to construct an independent namespace container with unique addressing capability for each independent causal stream, generating an isolated causal stream storage unit that carries data in a single direction and does not interfere with each other. The parallel path node attribute information stored in the isolated causal flow storage unit is subjected to process rule matching analysis to extract key constraint parameters including consistency of quality inspection standards, sharing of raw material batches and coverage of operation qualifications, and process compatibility label vectors describing the logical compatibility relationship between different causal flows are generated. Based on the process compatibility tag vector, perform a multidimensional relation matrix construction operation to calculate the compatibility value and conflict probability value between any two independent causal flows, and generate a cross-flow correlation matrix that records the topological correlation state between all flows. The cross-flow association matrix is ​​structured and encapsulated, and the matrix data is bidirectionally indexed and bound to the corresponding independent namespace container to generate a multi-path parallel tracing graph substructure that supports fast backtracking query and dynamic update, so as to complete the self-clarification of node affiliation in complex flow scenarios.

9. A batch traceability and transfer method for flame-retardant clothing fabric based on pretreatment according to claim 8, characterized in that, The set of independent causal flow identifiers represents the independent evolution characteristics of each path.

10. A batch traceability and transfer system for flame-retardant clothing fabrics based on pretreatment, characterized in that, The flame-retardant clothing fabric batch traceability and transfer method based on pretreatment, as described in claim 1, is used to trace and transfer batches of flame-retardant clothing fabric.

Images