Industrial Internet of Things data acquisition and elastic transmission method

Abstract

The invention relates to the field of industrial Internet of Things data processing, in particular to an industrial Internet of Things data acquisition and elastic transmission method. The method comprises the following steps: acquiring a process unit topology and monitoring target configuration, performing process unit code registration, probe layout scheme generation and acquisition session configuration structure registration processing, and generating an acquisition session configuration structure; time anchor point unified processing, change node alignment sequence generation and contact value grading threshold registration processing are executed, and a data importance grading table structure is generated; executing fingerprint detection session list generation, network state fingerprint matrix construction and network state category judgment processing, and generating a network state category structure; and carrying out elastic transmission strategy graph construction and strategy version record registration processing to generate an elastic transmission execution configuration structure. According to the invention, through multi-level data processing and intelligent state perception, accurate alignment of data acquisition and elastic self-adaption of transmission are realized, and the reliability and efficiency of an industrial Internet of Things system are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) data processing, and in particular to a method for industrial IoT data acquisition and flexible transmission. Background Technology

[0002] In the field of industrial IoT data processing, existing solutions based on process unit topology and monitoring target configuration typically generate probe deployment schemes by pre-setting acquisition rules, register the acquisition session configuration structure on the acquisition side in a fixed manner, and then schedule data transmission based on the general state judgment of the network side. However, these solutions suffer from limitations such as a lack of coordination between the acquisition session configuration structure and time anchor point processing, a lack of consistency in the generation of change node alignment sequences and the registration of connection value hierarchical thresholds, and the inability of fingerprint detection session list generation and network state fingerprint matrix construction to reliably support network state category determination. Existing methods often rely on local configurations or static strategies for single links. During data acquisition and transmission driven by the acquisition session configuration structure, when network state changes occur and the network state category structure needs to participate in scheduling, issues arise such as the difficulty in continuously connecting the construction of the elastic transmission strategy graph and the registration of strategy version records with the acquisition session configuration structure and the data importance hierarchy table structure. Furthermore, the generation of the elastic transmission execution configuration structure lacks traceable boundaries. This leads to inconsistencies between the monitoring link, data archiving link, and scheduling link in the production and processing scenario, affecting the continuous management of the process and monitoring target configuration. For the joint processing of the acquisition session configuration structure, data importance classification table structure, and network status category structure, existing technologies generally lack unified field standards and link constraints among session configuration registration, unified processing of time anchor points, processing of change node alignment sequence generation, processing of connection value classification threshold registration, processing of network status category determination, and processing of policy version record registration. It is difficult to form a continuous link in the application scenarios of industrial IoT data acquisition and elastic transmission methods, namely, the generation of acquisition session configuration structure, data importance classification table structure, network status category structure, and elastic transmission execution configuration structure. This makes it difficult to maintain consistency in the elastic transmission execution configuration structure in terms of policy scope, version consistency, and auditability. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method for industrial Internet of Things (IoT) data acquisition and flexible transmission, comprising: Acquire the process unit topology and monitoring target configuration, perform process unit coding registration, probe deployment scheme generation and acquisition session configuration structure registration processing, and generate acquisition session configuration structure; Based on the data collection session configuration structure, the system performs unified processing of time anchor points, generation of alignment sequences for changing nodes, and registration of hierarchical thresholds for connection values, generating a data importance hierarchy table structure. Based on the data collection session configuration structure and data importance hierarchy table structure, the fingerprint detection session list generation process, the network status fingerprint matrix construction process, and the network status category determination process are performed to generate the network status category structure. Based on the network state category structure, we construct the elastic transmission policy graph and register policy version records to generate the elastic transmission execution configuration structure.

[0004] Furthermore, the process of registering process unit codes also includes: The process unit coding registration process includes generating a stable process unit code for each process unit. The stable process unit code is generated by concatenating the original process unit identifier, the production line identifier, the process section identifier, and the equipment asset identifier and then standardizing it. The standardization process includes removing whitespace, unifying capitalization, merging symbols, supplementing prefixes, and resolving duplicate conflicts. The final code is then registered in the coding field of the process unit coding table. At the same time, an audit field containing a registration timestamp field, a registration source field, and a configuration version number field is written, and a code reverse lookup index is generated simultaneously.

[0005] Furthermore, the process of generating the probe deployment plan also includes: The probe deployment scheme includes probe type matching results, deployment location mapping results, sampling schedule, and channel mapping table. Based on the probe deployment scheme, a session primary key generation process is performed. This process involves generating a session primary key by combining a session time period identifier, process unit code, measurement point object identifier set summary, deployment version number field, and configuration version number field, and then obtaining the primary key string through primary key normalization. A degradable field set registration process is also performed, which includes registering a field retention list, field aggregation rules, field freeze flags, field reduction sampling rhythm field, and field compression candidate flag field. Finally, an importance weight field registration process is performed, which includes registering a measurement point object basic weight field, a monitoring topic weight field, and an alarm level weight field.

[0006] Furthermore, the process of unifying the execution time anchor points also includes: The unified time anchor processing includes reading the synchronization clock capability field and timestamp reference field from the time alignment probe, writing the collected records into the unified time anchor field, and generating a batch alignment index; performing missing test segment marking processing, which includes marking missing test segments and registering the missing test reason field based on the sampling time table, batch alignment index, sampling failure count field, and connection status field; constructing a multi-source time series data packet, which includes a session primary key, a sampling batch identifier field, a collection record set, a unified time anchor field set, a batch alignment index, and a missing test segment marking table; and performing measurement point object sequence extraction processing based on the multi-source time series data packet, which includes grouping by measurement point object identifier and sorting by the unified time anchor field to generate a measurement point object sequence, and writing it into the missing test mask field.

[0007] Furthermore, the process of registering and processing contact value hierarchical thresholds also includes: The connection value generation process includes normalizing and aggregating the linkage intensity to generate connection value fields and group connection value fields; performing hierarchical threshold registration processing, which includes mapping the connection value fields to discrete importance levels and registering the hierarchical threshold fields; and generating a data importance level table structure, which includes a session primary key, a threshold version number field, a hierarchical threshold field, a hierarchical mapping table, a measurement point object identifier, a connection value field, an importance level field, a hierarchical basis field, and a threshold caliber audit field.

[0008] Furthermore, the process of generating the fingerprint detection session list also includes: The fingerprint detection session list includes a session primary key, network path identifier, detection task identifier list, detection frequency field, detection window field, passive statistical collection item list, summary of bound measurement point object identifier set, importance classification field, threshold version number field, and session audit field reference. Based on the fingerprint detection session list, detection results and statistical results extraction processing is performed. The extraction processing includes reading detection result records from the detection result cache and reading statistical result records from the statistical collection cache, and performing time alignment. Network status indicator normalization processing is performed. The network status indicator normalization processing includes unit caliber normalization, outlier pruning, missing test marker filling, and baseline merging to generate round-trip latency indicators, latency jitter indicators, packet loss ratio indicators, and queue backlog indicators. Sliding window aggregation processing is performed. The sliding window aggregation processing includes windowed aggregation of the normalized network status indicators within the time range limited by the detection window field to generate a window indicator vector. Fingerprint vector concatenation processing is performed. The fingerprint vector concatenation processing includes concatenating the window indicator vectors into a fingerprint vector according to a fixed field serialization order.

[0009] Furthermore, the process of constructing the network state fingerprint matrix also includes: The network state fingerprint matrix includes a row index consisting of a session primary key, a network path identifier, and a window start and end time field, and column fields consisting of various indicator fields within the fingerprint vector. Based on the network state fingerprint matrix, a state mapping rule loading process is performed, which includes reading state mapping rules from the rule base. The state mapping rule includes a rule identifier, an entry condition field, an exit condition field, a condition threshold field, a rule priority field, and a rule version number field.

[0010] Furthermore, the network state category determination process also includes: The network status category determination process includes matching the fingerprint vector corresponding to each network path identifier and each window start and end time field with the entry condition field according to the rule priority field and generating a network status category field; performing hysteresis parameter registration processing, which includes registering the entry holding window field, exit holding window field, minimum dwell window field, and dual threshold caliber field; and generating a network status category structure, which includes the session primary key, network path identifier, window start and end time field, network status category field, trust level field, rule identifier reference field, rule version number field, hysteresis parameter registration table reference, and status audit field.

[0011] Furthermore, the process of constructing the elastic transport strategy graph also includes: The elastic transmission strategy graph includes a node table, an edge table, an action set table, an action binding table, a graph version candidate field, and a graph audit field. Based on the elastic transmission strategy graph, a version number field generation process is performed, which includes generating a version number field by combining the session primary key field, the graph version candidate field, and the registration timestamp field. An effective scope field extraction process is then performed, which includes extracting the session primary key range field, the network path identifier range field, the measurement point object identifier set summary range field, the importance classification range field, and the effective time window field.

[0012] Furthermore, the policy version record registration process also includes: The policy version record generation process includes generating a policy version record, which includes a version number field, an effective scope field, a graph summary field, a node summary field, an edge summary field, an action summary field, a rule version number field, a threshold version number field, and a difference summary field; performing a rollback point registration process, which includes registering a rollback point identifier field, a target version number field, a rollback reference field, a rollback verification field, and a rollback registration timestamp field; performing an audit field encapsulation process, which includes merging graph audit fields, edge audit fields, action binding audit fields, status audit fields, hysteresis adjustment audit fields, threshold caliber audit fields, threshold change audit fields, and session audit fields into an audit encapsulation field; and generating a policy version record.

[0013] The key innovations of this invention include: (1) Based on the acquisition session configuration structure, the process unit coding registration, probe deployment scheme generation and acquisition session configuration structure registration processing after acquiring the process unit topology and monitoring target configuration are integrated, so that the acquisition session configuration structure serves as a unified input carrier for subsequent unified processing of time anchor points, generation processing of change node alignment sequences, registration processing of connection value hierarchical thresholds and generation processing of fingerprint detection session list.

[0014] (2) Based on the acquisition session configuration structure, the link organization method of unified processing of time anchor points and generation of sequence of alignment of change nodes is adopted, and combined with the connection value hierarchical threshold registration processing, the multi-source acquisition data is transformed into the reusable data importance hierarchy table structure under the same time anchor caliber, so that the data importance hierarchy table structure can be directly called by the subsequent network status category determination processing.

[0015] (3) Based on the collection session configuration structure and the data importance classification table structure, the network status category structure is generated by using a continuous link of fingerprint detection session list generation processing, network status fingerprint matrix construction processing and network status category determination processing. Furthermore, the elastic transmission strategy graph construction and strategy version record registration processing are completed based on the network status category structure, so that the elastic transmission execution configuration structure forms a traceable execution configuration closed loop under the constraint of strategy version record.

[0016] The following are its main beneficial effects: (1) In view of the problem that the existing solutions for probe deployment scheme generation and acquisition session configuration structure registration are mostly separate configurations and difficult to form a consistent input caliber with subsequent links, the present invention uses the acquisition session configuration structure to organize the process unit coding registration, probe deployment scheme generation and acquisition session configuration structure registration, so that subsequent processing links can reference fields and connect links under the same session configuration caliber, thereby reducing the inconsistency of acquisition links, data archiving links and scheduling links caused by the dispersion of session configuration in production and processing scenarios.

[0017] (2) In view of the problem that the data importance is difficult to reuse stably due to the lack of coordination between the unified processing of time anchor points and the generation of the alignment sequence of change nodes in the existing scheme, and the lack of a unified standard in the registration of the threshold of the connection value classification, the present invention completes the unified processing of time anchor points and generates the alignment sequence of change nodes under the constraints of the acquisition session configuration structure, and then generates the data importance classification table structure by combining the registration of the threshold of the connection value classification. This enables the data importance to form a reusable classification result under the same time anchor standard, thereby reducing the risk of inconsistency in subsequent judgment links caused by the drift of the data importance standard in the scenario of network status change.

[0018] (3) In view of the problems that the fingerprint detection session list generation and network status fingerprint matrix construction processes in the existing solutions are difficult to stably support the network status category determination process, and the elastic transmission strategy graph construction and strategy version record registration processes are difficult to form a continuous link with the previous hierarchical results, the present invention generates the network status category structure under the joint constraints of the collection session configuration structure and the data importance classification table structure, and completes the elastic transmission strategy graph construction and strategy version record registration processes on the basis of the network status category structure, thereby generating an elastic transmission execution configuration structure, so that the strategy version record forms a boundary constraint and traceability basis for the execution configuration, thereby maintaining the consistency of the collection, determination and execution configuration links under the network status change conditions in the production and processing scenarios. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an industrial Internet of Things (IoT) data acquisition and flexible transmission method provided in an embodiment of this application. Detailed Implementation

[0020] Example 1: Refer to Figure 1 This is a flowchart illustrating an industrial Internet of Things (IoT) data acquisition and flexible transmission method provided in an embodiment of the present invention. The process may include at least steps S100-S400: S100: Obtain the process unit topology and monitoring target configuration, perform process unit coding registration, probe deployment scheme generation and acquisition session configuration structure registration processing, and generate acquisition session configuration structure; S200: Based on the data acquisition session configuration structure, perform unified processing of time anchor points, generation of alignment sequences for changing nodes, and registration of hierarchical thresholds for connection values ​​to generate a data importance hierarchy table structure. S300, based on the acquisition session configuration structure and data importance hierarchy table structure, performs fingerprint detection session list generation processing, network status fingerprint matrix construction processing and network status category determination processing, and generates network status category structure; S400: Based on the network state category structure, construct the elastic transmission policy graph and register policy version records to generate the elastic transmission execution configuration structure.

[0021] Step S100 includes at least steps S110-S130: S110. Obtain the process unit topology and monitoring target configuration, perform process unit coding registration, measurement point object definition and measurement point object field template registration processing, and obtain the measurement point object list; Specifically, when an industrial IoT edge node accesses a configuration source, it first reads the process unit topology and monitoring target configuration. The process unit topology refers to a structured description of the set of process units in the production site and their connections. These connections include at least two of the following: material flow relationships, energy supply relationships, control interlock relationships, and data link relationships. Each process unit includes an upstream unit identifier, a downstream unit identifier, a connection type identifier, a connection direction identifier, and a connection validity period field. The monitoring target configuration refers to a set of monitoring targets set for production process supervision and operational analysis. It includes at least a target process unit identifier, a target monitoring topic identifier, a target observation window field, a sampling granularity field, an alarm constraint field, and a data retention constraint field. The configuration source can come from any of the following: workshop master station, field control station, or edge node local configuration storage area. The industrial IoT edge node reads the configuration source during the startup phase, process switching phase, and configuration version change trigger phase, and performs a check on the integrity field of the configuration source during the reading. The check items include the existence of required fields, field type conformity, uniqueness of the identifier field, and topology connectivity checks. If the check fails, a configuration exception record is written and a retry waiting window is entered. The retry waiting window is limited by the retry rhythm field in the monitoring target configuration and is triggered by the edge node timer.

[0022] In an engineering scenario embodiment, a production line in a discrete manufacturing workshop is divided into a material feeding unit, a processing unit, a quality inspection unit, and a packaging unit. The process unit topology records the material flow relationship between the material feeding unit and the processing unit, and also records the data link relationship between the processing unit and the quality inspection unit. The monitoring target configuration record periodically samples the spindle speed, feed rate, and equipment status fields of the processing unit, performs event-triggered sampling on the defect event field of the quality inspection unit, and registers the alarm-related field set in the alarm constraint field. The industrial IoT edge node pulls the configuration from the workshop master station and writes it to its local configuration storage area. During the configuration version change trigger phase, it writes the differences between the new and old versions into the configuration difference record. These differences include four categories: added process units, deleted measurement points, changes in field units, and changes in sampling granularity.

[0023] After obtaining the verified process unit topology, the industrial IoT edge node performs process unit code registration. Process unit code registration refers to generating a stable process unit code for each process unit and registering it in the process unit code table. The stable process unit code is generated by concatenating the original process unit identifier, the production line identifier, the process section identifier, and the equipment asset identifier, and then standardizing the process. The standardization process includes whitespace removal, case unification, symbol merging, prefix completion, and duplicate / conflict resolution. Conflict resolution uses a rule prioritizing topology level and then configuration update time to generate candidate codes. The final code is written to the coding field of the process unit code table, and simultaneously to an audit field, which includes a registration timestamp field, a registration source field, and a configuration version number field. During code registration, the industrial IoT edge node simultaneously generates a code reverse lookup index. This index is used to map the original device identifiers of subsequently connected devices back to the process unit codes. If a reverse lookup fails, a code missing record is written and the entry is entered into a manual review queue.

[0024] Furthermore, the industrial IoT edge node executes the definition of measurement point objects based on the process unit coding table and the monitoring target configuration. The measurement point object refers to the objectified encapsulation of the acquisition point, and it consists of a measurement point object identifier, its corresponding process unit code, a data field list, data field semantic tags, a sampling rhythm field, an acquisition channel constraint field, a cache constraint field, and an anomaly marker field. Specifically, the data field list includes at least two of the following: process quantity fields, status quantity fields, and event quantity fields; the data field semantic tags include at least one of the following: equipment operation, quality inspection, energy consumption metering, and safety interlocking; the sampling rhythm field includes a periodic sampling interval field, an event triggering condition field, and a sampling suppression window field; the acquisition channel constraint field includes an interface capability requirement field and a bandwidth budget field; the cache constraint field includes a cache upper limit field and an expired cleanup rule field; and the anomaly marker field includes a missing measurement marker field, an abnormal fragment marker field, and a recovery marker field. When defining a measurement point object, the Industrial Internet of Things (IIoT) edge node searches for the corresponding set of candidate data fields for each target monitoring theme identifier in the monitoring target configuration, and selects the minimum set of sampling rhythm fields according to the sampling granularity field. This minimum set refers to the set consisting of the periodic sampling interval field, the sampling suppression window field, and the buffer limit field, provided that the target observation time window field is covered. The event trigger condition field and the bandwidth budget field are written into the measurement point object as extended fields when the monitoring target configuration indicates their existence. If the monitoring target configuration provides both the periodic sampling interval field and the event trigger condition field for the same measurement point object identifier, the IIoT edge node defines the measurement point object as a hybrid sampling measurement point object and registers the priority field of the periodic sampling interval field and the event trigger condition field within the sampling rhythm field. The priority field indicates whether to pause periodic sampling or reduce the periodic sampling frequency within the time window when the event trigger occurs, and this pause or frequency reduction action is written into the sampling suppression window field. The IIoT edge node also writes a trigger debouncing window field into the hybrid sampling measurement point object. This trigger debouncing window field is used to merge and mark consecutively triggered events and avoid repeated triggering in subsequent sampling rhythm orchestration.

[0025] After the measurement point object is defined, the industrial IoT edge node performs a measurement point object field template registration process. The measurement point object field template refers to a templated description of the field name, field type, field value range, field unit, field precision, field default value, and field verification rules for each field within the measurement point object, and includes a field serialization order field and a field compression candidate flag field. The field template registration process includes unified field naming, unified unit caliber, enumerated value mapping, and verification rule loading. Unified field naming uses the field naming specification table in the monitoring target configuration; unified unit caliber converts units from different measurement sources to a unified unit and writes it into the field unit; enumerated value mapping maps the original code value of the state quantity field to a standard code value and writes it into the field value range; and verification rule loading binds at least one of the following types of field range verification, mutation verification, and drift verification to the field verification rules. When loading verification rules, the edge node of the Industrial Internet of Things writes the verification failure action field into the field template. The verification failure action field includes three actions: mark only, discard, and roll back to the default value. Among them, mark only is the core action set item, while discard and roll back to the default value are written as extended actions when the monitoring target configuration indicates strict control.

[0026] After registration, the industrial IoT edge node generates a list of measurement point objects. This list includes a list of measurement point object identifiers, a mapping table of their respective process unit codes, and a mapping table of measurement point object field templates. The list is written to the local session cache as an output field name and is also used as an input field name in S120 for probe type matching, deployment location mapping, and sampling rhythm orchestration. For seamless transitions between steps, the field compression candidate flag field in the measurement point object field template mapping table is referenced in the compression and reduction configuration process in S430, and the anomaly flag field is referenced in the missing measurement segment marking process in S210 and used as an anomaly segment filtering condition field in the change node extraction process in S220.

[0027] S120. Extract interface capabilities and spatial locations from the list of measurement points, perform probe type matching, deployment location mapping and sampling rhythm arrangement, and generate a probe deployment scheme. Specifically, after receiving the list of measurement points output by S110, the industrial IoT edge node extracts interface capabilities and spatial location from the list. The interface capabilities refer to the set of data access capabilities available from the data source side where the measurement point is located. These capabilities include fields for interface type, transmission medium, upper limit of acquisition frequency, upper limit of message size, synchronization clock capability, and number of acquisition channels. The interface type field includes at least one of wired, wireless, fieldbus, and control station interfaces; the transmission medium field includes at least one of twisted pair, fiber optic, cellular network, and private network link; and the synchronization clock capability field includes three states: timestamp reporting, clock synchronization, and no synchronization. The spatial location refers to the position description of the equipment or process unit to which the measurement point is attached in the workshop spatial coordinate system. This location includes fields for area identification, machine position identification, coordinate reference, and installation height, and may also include fields for obstruction level and environmental interference level. These fields are extended fields, written when the monitoring target configuration indicates a complex acquisition environment, and used for subsequent deployment location mapping constraints.

[0028] After extracting interface capabilities and spatial location, the industrial IoT edge node performs probe type matching. The probe type refers to the capability type of the probe responsible for data acquisition and edge preprocessing. The probe type includes at least three categories: acquisition probes, time alignment probes, and transmission probes. Acquisition probes are used to connect to data sources and generate raw acquisition records. Time alignment probes are used to write a unified time anchor field to the acquisition records locally. Transmission probes are used to encapsulate the acquisition records into data fragments to be transmitted and write them to the transmission queue. During probe type matching, the industrial IoT edge node selects the matching acquisition probe subtype based on the interface category field and the acquisition frequency upper limit field, and selects the time alignment method of the time alignment probe based on the synchronization clock capability field. When the synchronization clock capability field is "no time alignment," the time alignment probe uses a local clock stamp to write and registers the time drift estimation field. When the synchronization clock capability field is "clock alignment," the time alignment probe loads the time alignment plan and registers a time alignment failure retry window. The output of probe type matching includes probe type identifier and acquisition channel identifier. The acquisition channel identifier refers to the logical channel identifier assigned to the measurement point object inside the edge node. The logical channel identifier forms a one-to-one or one-to-many mapping with the measurement point object identifier and is recorded in the channel mapping table.

[0029] Furthermore, the industrial IoT edge nodes perform deployment location mapping. This deployment location mapping refers to the process of mapping measurement points to specific installation locations and access paths. The installation locations include process unit installation location identifiers, equipment installation location identifiers, and probe installation location identifiers. The access paths include uplink identifiers, edge access port identifiers, and gateway forwarding path identifiers. During deployment location mapping, the industrial IoT edge nodes select the nearest access point based on the area identifier field and machine location identifier field in the spatial location, and select the access port type based on the transmission medium field. When the transmission medium field indicates a private network link, the deployment location mapping writes the private network entry identifier field and registers the link priority field. When the transmission medium field indicates a cellular network, the deployment location mapping writes the cellular network access point identifier field and registers the signal quality sampling rhythm field. The signal quality sampling rhythm field is used for subsequent network path identifier registration. If the occlusion level field exists, an occlusion compensation configuration field is added to the deployment location mapping. The occlusion compensation configuration field includes an installation height adjustment field and an installation orientation adjustment field. If the environmental interference level field exists, an anti-interference acquisition window field is added to the deployment location mapping. The anti-interference acquisition window field is used to limit the sampling execution period and write the sampling rhythm arrangement input.

[0030] After completing probe type matching and deployment location mapping, the industrial IoT edge node performs sampling rhythm orchestration. Sampling rhythm orchestration refers to the unified scheduling and conflict resolution of the sampling rhythm fields of the measurement point objects, generating a sampling schedule for concurrent acquisition of multiple measurement points. Sampling rhythm orchestration includes at least the merging of periodic sampling interval fields, loading of event trigger condition fields, and overlaying of sampling suppression window fields. Specifically, merging of periodic sampling interval fields merges similar periods into the same sampling batch and generates a batch identifier field; loading of event trigger condition fields writes the trigger event source identifier and trigger threshold field into the event trigger rule table and binds them to the batch identifier field; and overlaying of sampling suppression window fields takes the intersection of the anti-interference acquisition window field and the sampling suppression window field to generate a valid sampling window field. Sampling rhythm orchestration also records channel congestion constraint fields, which include a maximum number of packets per batch and a maximum packet size per batch, serving as constraint parameters for subsequent session primary key generation and execution queue configuration.

[0031] After completing the above processing, the industrial IoT edge node generates a probe deployment scheme. The probe deployment scheme includes probe type matching results, deployment location mapping results, sampling schedule, and channel mapping table, and includes a deployment version number field and a deployment design field. The industrial IoT edge node writes the probe deployment scheme as an output field name into its local session cache, and uses it as an input field name in S130 for session primary key generation, degradable field set registration, and importance weight field registration. Simultaneously, the sampling schedule is referenced in the sampling execution process in S210, and the channel mapping table is referenced in the network path identifier registration process in S310.

[0032] S130. Generate session primary key, register degradable field set and importance weight field for probe deployment scheme, and generate collection session configuration structure; Specifically, after receiving the probe deployment plan output by the S120, the industrial IoT edge node first performs session primary key generation. The session primary key refers to a unique identifier field for a single data acquisition and transmission session. The session primary key is generated by combining the session time period identifier, process unit code, measurement point object identifier set summary, deployment version number field, and configuration version number field, and is obtained as a primary key string through primary key normalization. The primary key normalization process includes field order fixing, separator merging, length truncation, and conflict resolution. Conflict resolution is achieved by appending a sequence number field and writing it to the primary key conflict record table. After the session primary key is generated, the industrial IoT edge node writes the session primary key to the session primary key table and registers the session start and end time field, session status field, and session retrieval flag field in the session primary key table. The session status field includes four states: initialization, running, suspended, and terminated. State switching is triggered by the batch identifier field in the sampling rhythm orchestration and recorded in the session audit field. The session audit field and the audit field encapsulated in subsequent policy version records use the same field name and are constrained by the same audit caliber template. When registering in the session primary key table, the industrial IoT edge node synchronously writes the session renewal condition field. The session renewal condition field includes three types of triggering conditions: configuration version number change, deployment version number change, and sampling time schedule change. When the triggering condition occurs, the session status field is updated to suspended and enters the renewal processing flow, which is driven by the edge node task scheduler.

[0033] Furthermore, the industrial IoT edge node performs a degradable field set registration. The degradable field set refers to the set of data fields whose expression precision or upload frequency can be adjusted when network status changes or resource constraints occur. The degradable field set includes a field retention list, field aggregation rules, field freeze flags, field reduction sampling rhythm fields, and field compression candidate flag fields. Specifically, the field retention list refers to the set of field names that still need to be retained in the degradation scenario; the field aggregation rules refer to the aggregation method and aggregation window fields used for the degraded fields; the field freeze flag refers to the flag field that suspends updates for the degraded fields; the field reduction sampling rhythm field refers to the sampling rhythm substitution value used in the degradation scenario; and the field compression candidate flag fields are associated with the corresponding fields in the measurement point object field template mapping table output by S110. When registering the set of degradeable fields, the Industrial Internet of Things (IIoT) edge node selects the minimum set of the field retention list based on the data retention constraint fields and alarm constraint fields in the monitoring target configuration. This minimum set includes at least an alarm association field, a time anchor field, and a measurement point object identifier field. For fields not within the minimum set, the IIoT edge node loads field aggregation rules according to field type and registers the field reduction rhythm field. For process quantity fields, it loads interval aggregation rules; for state quantity fields, it loads stable segment aggregation rules; and for event quantity fields, it loads event compression rules. If the monitoring target configuration indicates the existence of high-frequency redundant fields, the IIoT edge node writes these fields into the field freeze flag candidate set and registers the freeze trigger condition field within the field freeze flag. The freeze trigger condition field includes three categories: network status category trigger conditions, queue level trigger conditions, and session status trigger conditions. The network status category trigger conditions are loaded by the compression and reduction configuration processing of S430 after the network status category structure is output by S330. The queue level trigger conditions are loaded during the execution queue configuration processing of S430. The session status trigger conditions are triggered by the session status field in the session primary key table. To avoid conflicts between freezing actions and alarm-related fields, industrial IoT edge nodes execute alarm-related field exclusion rules when generating the field freeze flag candidate set. The exclusion rules indicate that alarm-related fields must not be written into the field freeze flag candidate set, and write the field exclusion audit field when the exclusion rule is hit.

[0034] After completing the registration of the degradable field set, the Industrial Internet of Things (IIoT) edge node performs the registration of the importance weight field. The importance weight field refers to the weight field used to characterize the relative importance of the measurement point object before the data importance classification table structure is generated. The importance weight field includes the measurement point object basic weight field, the monitoring topic weight field, and the alarm level weight field, and also includes a weight source field and a weight update time field. During importance weight field registration, the IIoT edge node generates an initial weight combination based on the target monitoring topic identifier and alarm constraint field in the monitoring target configuration. The minimum set of core parameters for the initial weight combination includes the measurement point object basic weight field and the alarm level weight field. The monitoring topic weight field, as an extended field, is written when multiple monitoring topics overlap and participates in the prior caliber of classification threshold registration in the subsequent connection value generation stage. The IIoT edge node simultaneously generates a weight verification rule field when registering weights. The weight verification rule field includes two types: weight range verification and weight drift verification. When weight drift verification is triggered, it is written to the weight anomaly record table and enters the weight recalculation waiting window. The weight recalculation waiting window is triggered by the configuration version number field, the batch identifier field, and the session renewal condition field. To support automated operation, the industrial IoT edge node registers the recalculation task identifier field when writing to the weight anomaly record table. The recalculation task identifier field is consumed by the task scheduler and the weight recalculation is performed during the sampling interval. After the recalculation is completed, the weight update time field is updated and written to the weight change audit field.

[0035] After completing the generation of the session primary key, registration of the set of degradable fields, and registration of the importance weight field, the industrial IoT edge node generates a data acquisition session configuration structure. This data acquisition session configuration structure includes a session primary key, a reference to the list of measurement points, a reference to the probe deployment scheme, a sampling schedule, a channel mapping table, a set of degradable fields, an importance weight field, a configuration version number field, a deployment version number field, and a session audit field. The session audit field includes a registration timestamp field, a registration source field, a most recent change timestamp field, and a change reason field. The industrial IoT edge node writes the acquisition session configuration structure as an output field name into the session configuration storage area, and uses the acquisition session configuration structure as an input field name in S210 for sampling execution, time anchor unification, and missing segment marking processing. Simultaneously, the acquisition session configuration structure is used as an input field name in S310 for network path identification registration, active probe task orchestration, and passive statistical acquisition item binding processing. Furthermore, the field reduction acquisition rhythm field, field freeze marker candidate set, and freeze trigger condition field in the degradable field set are referenced in the compression reduction acquisition configuration processing and retransmission window configuration processing in S430. The importance weight field is referenced in the hierarchical threshold registration processing in S230, and the session audit field is referenced and written into the policy version record in the audit field encapsulation processing in S420. In summary, the technical effects of this step are as follows: This step binds the list of measurement points and probe deployment schemes to the session primary key, and solidifies the set of degradable fields and importance weight fields within the acquisition session configuration structure, enabling subsequent elastic transmission actions to have executable inputs at the field level; at the same time, it forms a traceable configuration evolution link through the configuration version number field, deployment version number field, and session audit field, reducing session identification conflicts caused by configuration changes and suppressing the drift of degradation rules.

[0036] Step S200 includes at least steps S210-S230: S210. Based on the acquisition session configuration structure, perform sampling execution, time anchor point unification, and missing segment marking processing to obtain multi-source time-series data packets; Specifically, this step is triggered by the industrial IoT edge node when the session status field is in runtime. The industrial IoT edge node reads the session primary key, measurement point object list reference, probe deployment scheme reference, sampling schedule, channel mapping table, degradable field set, importance weight field, and configuration version number field from the acquisition session configuration structure output in the previous step. The acquisition session configuration structure is stored in the edge node's session configuration storage area according to the session primary key index. The edge node's session scheduler generates sampling batch trigger events according to the sampling schedule and writes the batch identifier field into the sampling execution context. In addition to the acquisition session configuration structure, the input sources for sampling execution also include the underlying driver status corresponding to the acquisition channel identifier indicated by the channel mapping table. The underlying driver status is reported by the acquisition probe and includes the current connection status field, the most recent sampling timestamp field, the sampling failure count field, and the reconnection wait window field. When each sampling batch trigger event occurs, the edge node first performs a consistency check on the session status field of the corresponding session primary key in the session primary key table. If the session status field is in the pending or terminated state, the sampling of that batch is paused and a sampling suspension record is written. The sampling suspension record includes the session primary key, batch identifier field, suspension reason field, and registration timestamp field.

[0037] During the sampling execution phase, the edge node iterates through the list of measurement point object identifiers in the measurement point object list and aligns them with the acquisition channel identifiers. Sampling throttling is then performed according to the upper limit of the acquisition frequency field and the upper limit of the message size field in the acquisition channel constraint field. The specific sampling action is executed by the acquisition probe. The acquisition probe selects the corresponding access adapter according to the interface category field. In wired interface scenarios, the access adapter calls the wired port sampling driver; in wireless interface scenarios, it calls the wireless access sampling driver; in fieldbus interface scenarios, it calls the bus message subscription driver; and in control station interface scenarios, it calls the control station data subscription driver. The acquisition probe encapsulates the sampled raw data fragments into acquisition records and writes them to the edge node's acquisition buffer. These acquisition records include the measurement point object identifier, acquisition channel identifier, original timestamp field, original field value set, and sampling batch identifier field. The field names, field types, and field unit calibers of the original field value set are derived from the field template mapping table of the measurement point object. During encapsulation, the acquisition probe executes at least one of the following verification rules within the field template registration process: field range verification, mutation verification, and drift verification, and writes the verification conclusion to the field verification status field. When the field verification status field indicates "mark only," the acquisition probe retains the original field value and writes the field anomaly flag to the anomaly flag field. When the field verification status field indicates "discard," the acquisition probe writes the field default value and registers the field discard flag. The edge node simultaneously writes the sampling failure count field to the sampling execution context and enters the sampling failure handling branch when the count exceeds a threshold. The sampling failure handling branch includes timing the reconnection wait window field, resetting the connection status field, and writing the sampling failure audit field. The sampling failure audit field uses the same field name as the session audit field and is written to the audit extension area of ​​the session configuration storage area.

[0038] After sampling is completed, the edge nodes perform unified time anchor processing on the collected records. The time anchor refers to a unified time reference field used for cross-channel alignment. This unified time anchor processing is performed by the time alignment probe, which reads the synchronization clock capability field and the timestamp reference field. When the synchronization clock capability field indicates clock synchronization, the time alignment probe reads the synchronization plan and writes the synchronization deviation record into the time deviation record field after successful synchronization. When the synchronization clock capability field indicates timestamp reporting, the time alignment probe performs unit conversion and time zone standardization on the original timestamp field and writes it into the unified time anchor field. When the synchronization clock capability field indicates no synchronization, the time alignment probe uses the edge node's local clock to write into the unified time anchor field and registers the time drift estimation field. The time drift estimation field is calculated by the time alignment probe based on the difference between the most recent sampled timestamp field and the edge node's local clock, and written into the drift sequence cache according to the sampling batch identifier field. The drift sequence cache has the same sliding window convergence caliber as the subsequent network status indicator normalization stage, facilitating the reuse of the same window configuration across steps. After the time alignment probe writes a unified time anchor field to each acquisition record, it generates a batch alignment index according to the sampling batch identifier field. The batch alignment index includes a summary of the measurement point object identifier set, an alignment start and end time field, and an alignment integrity marker field. The alignment integrity marker field indicates whether there are missing segments in the batch sampling.

[0039] During the unified processing of time anchor points, edge nodes perform missing segment marking in parallel. Missing segment marking refers to marking time segments that did not produce acquisition records according to the sampling schedule or produced acquisition records but whose field default values ​​exceeded a threshold, and registering the missing reason field. Missing segment marking processing relies at least on the sampling schedule, batch alignment index, sampling failure count field, and connection status field. Edge nodes first expand the list of measurement point object identifiers to be collected in the current batch according to the sampling schedule, and then retrieve the actual collection record set that arrived in the acquisition buffer by the session primary key and sampling batch identifier fields. If a collection record corresponding to a certain measurement point object identifier is missing, a missing entry is written to the missing segment marking table. The missing entry includes the session primary key, measurement point object identifier, missing start and end time fields, missing reason field, and registration timestamp field. The missing test reason field is determined by the edge node according to priority rules. These rules sequentially check the connection status field, reconnection wait window field, sampling failure count field, field discard flag, and the freeze trigger condition field of the field freeze flag candidate set. If the freeze trigger condition field points to a session status trigger condition, the missing test reason field is registered as a session suspension. If it points to a queue level trigger condition, it is registered as a queue level constraint. If the connection status field indicates a disconnection and the reconnection wait window field has not expired, the missing test reason field is registered as a connection disconnection wait window. The edge node performs deduplication and merging on the missing test segment marker table. Deduplication and merging merge adjacent entries based on the measurement point object identifier and the missing test start and end time fields and writes them to the merge marker field. The merge marker field is used by subsequent change node extraction stages to filter missing test segment boundaries.

[0040] After sampling, time anchor unification, and missing segment marking are completed, the edge node constructs a multi-source time-series data packet. This multi-source time-series data packet consists of a session primary key, a sampling batch identifier field, a collection record set, a unified time anchor field set, a batch alignment index, and a missing segment marking table. Each collection record in the collection record set includes a field verification status field and an anomaly marker field, and retains the correspondence between the measurement point object identifier and the collection channel identifier. The edge node writes the multi-source time-series data packet into the time-series data packet buffer and registers the data packet sequence number field. The data packet sequence number field is generated incrementally according to the session primary key and written into the data packet audit field. The multi-source time-series data packet is generated as an output field name at the end of this step and serves as an input field name for S220 in subsequent steps for extracting the measurement point object sequence from the multi-source time-series data packet, performing change node extraction, and change node alignment processing. Simultaneously, the missing segment marking table within the multi-source time-series data packet is referenced as an anomaly segment filtering condition field in subsequent steps and serves as one of the sources of audit association fields when generating the policy version record.

[0041] S220. Extract the sequence of measurement point objects from the multi-source time series data packets, perform change node extraction, change node type labeling and change node alignment processing, and generate a change node aligned sequence. Specifically, this step is triggered by the industrial IoT edge node in batches after the multi-source time-series data packets are generated. The triggering condition is driven by the sampling batch identifier field. The edge node reads the multi-source time-series data packets from the time-series data packet buffer, parses the session primary key and locates the measurement point object list reference, and then obtains the measurement point object field template mapping table and data field semantic tags. The measurement point object sequence refers to the time-series segment sequence aggregated by the measurement point object identifier. The edge node first groups the collection record set in the multi-source time-series data packet by the measurement point object identifier, and sorts the collection records with the same measurement point object identifier by a unified time anchor field to generate the measurement point object sequence. In the sorting stage, the edge node limits the sequence boundary according to the alignment start and end time field in the batch alignment index, and writes the segments covered by the missing measurement start and end time field into the missing measurement mask field according to the missing measurement segment marking table. The missing measurement mask field serves as the input constraint for the subsequent change node extraction stage. Each record in the sequence of measurement points retains the original set of field values ​​and the field verification status field. Edge nodes write records that are discarded according to the field verification status field into the abnormal fragment mark field and register the default fragment mark in the missing test mask field. The default fragment mark is consistent with the merge mark field of the missing test fragment mark table to avoid the same abnormality being registered repeatedly in multiple places.

[0042] After the sequence of measurement points is constructed, the edge nodes undergo change node extraction processing. These change nodes refer to key time points or key time period boundaries in the measurement point sequence that characterize state transitions, trend reversals, or abrupt changes. This step limits change nodes to three annotation categories: abrupt change, step change, and trend reversal. An abrupt change refers to a field value jumping beyond the field range verification threshold within adjacent sampling intervals. A step change refers to a field value completing an amplitude transition within a short window and maintaining a new level within a subsequent stable window. A trend reversal refers to a field value changing from rising to falling or from falling to rising in the trend direction formed by several consecutive sampling points. The change node extraction process is performed by the change node extraction unit. This unit reads the field precision, field value range, and verification rule loading caliber from the field template mapping table of the measurement point object. For process quantity fields, it prioritizes trend reversal detection; for state quantity fields, it prioritizes step detection; and for event quantity fields, it prioritizes abrupt change detection. When the measurement point object is a mixed-sample measurement point object, the change node extraction unit simultaneously reads the event trigger condition field and writes the occurrence time of the trigger event into the candidate change node set. The candidate change node set is then merged and deduplicated from change nodes in the same period sampling sequence. During processing, the change node extraction unit executes a skip strategy for segments covered by the missing test mask field. The skip strategy means that change nodes are not generated within the missing test segment, but boundary change nodes are allowed to be generated at the boundary of the missing test segment. Boundary change nodes are marked as missing test boundaries and written into the reference index of the missing test reason field.

[0043] Furthermore, the edge nodes perform change node type labeling processing on the candidate change node set. The change node type labeling refers to writing a change node type field, a change node credibility field, and a change node source field for each change node. The change node type field takes one of three categories: mutation, step, or trend reversal. The change node credibility field is calculated by the change node extraction unit based on the field verification status field, the missing measurement mask field, and the sampling suppression window field, and written into the credibility level. The credibility level is registered as a discrete level. The change node source field writes one of the following sources: periodic sampling, event triggering, or boundary completion. When generating the credibility level, the edge nodes use a minimum set parameter caliber, which includes the field verification status field, the missing measurement mask field, and the alignment integrity marker field. The sampling suppression window field and the trigger debouncing window field are used as extended fields in the calculation and written into the extended source marker when mixed sampling measurement point objects exist. For change nodes with the lowest credibility level, edge nodes are not directly discarded. Instead, they are written into the low credibility tag field and their weight in the alignment is reduced according to the rules in the subsequent alignment stage. The weight is written into the alignment weight field and is compatible with the lag step size in the subsequent linkage strength calculation stage.

[0044] After completing the extraction and labeling of change nodes, edge nodes undergo change node alignment processing. Change node alignment refers to projecting change nodes from different measurement point object sequences onto a unified time anchor axis and establishing alignment relationships across measurement point objects, outputting an aligned change node sequence. The alignment process is performed by the change node alignment unit, which reads the sampling time schedule and batch alignment index from the acquisition session configuration structure. First, it establishes an alignment time window field on the unified time anchor axis, constrained by both the sampling interval field and the sampling suppression window field, and generates an alignment candidate window for each change node. Subsequently, the change node alignment unit matches change nodes across different measurement point objects according to the alignment candidate windows. The matching rules include three types: same-window matching, adjacent-window matching, and lag matching. Same-window matching means the change node falls within the same alignment time window field; adjacent-window matching means the change node falls within adjacent alignment time window fields and the alignment weight field meets a threshold; and lag matching means the change node is offset across windows, but the offset falls within the lag step size candidate range. The candidate range of lag step size is determined by the drift verification caliber of the previous measurement point object field template registration process and the alignment time window field of this step. The edge node registers the candidate range of lag step size as the alignment lag configuration field and writes it into the alignment audit field, so that the same caliber can be referenced in the subsequent hierarchical threshold registration stage.

[0045] In the engineering scenario implementation, edge nodes construct a sequence of measurement points for the spindle speed, feed rate, and equipment status fields of the machining process unit, and a sequence of measurement points for the defect event field of the quality inspection process unit. Within a certain sampling batch, the equipment status field exhibits a step change node and switches from running to stopped; the spindle speed field exhibits a sudden change node at an adjacent time anchor point and quickly returns to zero; the feed rate field exhibits a trend reversal node and changes from decreasing to stabilizing; and the defect event field writes candidate change nodes at the event trigger time. The change node alignment unit uses the unified time anchor point field as the axis to determine that the step change node of the equipment status field is a window match with the sudden change node of the spindle speed field, determines that the candidate change node of the defect event field is an adjacent window match with the step change node of the equipment status field, and writes it into the alignment weight field. The trend reversal node of the feed rate field is determined to be a lag match with the sudden change node of the spindle speed field because it falls within the lag step candidate range. The alignment result is written to the alignment group identifier field in the alignment sequence of the variable node. The alignment group identifier field is used to describe a summary of the set of measurement point object identifiers contained in the same alignment relationship.

[0046] After completing the change node alignment process, the edge node generates a change node alignment sequence. This sequence includes at least the session primary key, sampling batch identifier field, alignment group identifier field, measurement point object identifier, change node type field, unified time anchor point field, alignment time window field, alignment weight field, confidence level, and missing measurement boundary marker. It also retains an alignment audit field to record the matching rule type and alignment lag configuration field used. The edge node writes the change node alignment sequence as an output field name to the alignment sequence cache and uses it as an input field name for S230 for linkage strength calculation, connection value generation, and hierarchical threshold registration. Simultaneously, the alignment group identifier field output in this step can be used as the grouping criterion for measurement point object grouping in the subsequent network path identifier registration stage, participating in the passive statistical collection item binding process.

[0047] S230. Perform linkage strength calculation, connection value generation and hierarchical threshold registration on the alignment sequence of the changing nodes, and generate a data importance hierarchical table structure. Specifically, this step is triggered in batches by the industrial IoT edge node after the change node alignment sequence is generated. The edge node reads the change node alignment sequence from the alignment sequence cache and performs linkage strength calculation by combining the importance weight field and session audit field in the acquisition session configuration structure. The linkage strength refers to the degree of consistency of changes between different measurement point objects on the same time anchor axis. This step limits the core calculation caliber of linkage strength to three elements: the number of synchronous changes, the lag step size, and the direction consistency, and writes them into the linkage strength record. The number of synchronous changes refers to the number of times change nodes appear simultaneously under the same alignment time window field or adjacent window matching rule. The lag step size refers to the discrete step size value of the time offset of change nodes under the lag matching rule. The direction consistency refers to the discrete determination of whether the trend direction of trend reversal change nodes is consistent within the alignment group. When calculating the number of synchronous changes, edge nodes use the alignment group identifier field as the aggregation key to count the number of same-window and adjacent-window matches of different measurement point object identifier pairs within each alignment group identifier field. Change nodes with the lowest confidence level are weighted lower using the alignment weight field. When calculating the lag step, edge nodes read the alignment lag configuration field from the alignment audit field, map the difference in the unified time anchor field in the alignment sequence of change nodes to the offset step, and write it into the lag step field. When calculating directional consistency, edge nodes only participate in the calculation for trend reversal change nodes. They generate a direction marker field based on the change node type field and the direction of change of the preceding and following field values, and then count the proportion of directional consistency within the alignment group and write it into the direction consistency field. In addition to the above three core fields, the linkage strength record also includes a missing measurement impact marker field. This missing measurement impact marker field is jointly determined by the missing measurement boundary marker and the merge marker field of the missing measurement segment marker table, and is used to effectively correct the number of synchronous changes when a missing measurement boundary exists.

[0048] After the linkage intensity record is generated, the edge nodes perform the connection value generation process. The connection value refers to the normalized convergence result of the linkage intensity, which is used to characterize the degree of correlation between the measurement point objects and to be registered in the data importance classification table structure. The connection value generation is performed by the connection value generation unit. The connection value generation unit reads the number of synchronous changes, lag step size, and directional consistency from the linkage intensity record, and reads the importance weight field from the acquisition session configuration structure as the prior weight input. The connection value generation unit first normalizes the number of synchronous changes according to the length of the sampling batch identifier field and writes it into the normalized synchronization field, then normalizes the lag step size according to the width of the aligned time window field and writes it into the normalized lag field, and normalizes the directional consistency according to the number of valid nodes participating in the trend turning point change type within the aligned group and writes it into the normalized direction field. Subsequently, the connection value generation unit merges the normalized synchronization field, normalized lag field, and normalized direction field with the measurement point object basic weight field and alarm level weight field according to the configured weight combination, writes the fusion rule into the connection value fusion rule field and registers it in the extension area of ​​the session audit field, so as to facilitate the encapsulation of the same audit caliber in the subsequent strategy version record generation stage. In this step, the minimum set parameter caliber for weight combination is fixed as the basic weight field of the measurement point object and the alarm level weight field. The monitoring topic weight field, as an extended field, only participates in the fusion and is written into the extended weight tag field when there are multiple monitoring topic overlay tags in the acquisition session configuration structure. The connection value generation unit outputs a connection value field for each measurement point object identifier and outputs a group connection value field for each alignment group identifier field. The group connection value field is used to guide the acquisition granularity of passive statistical acquisition item binding in the subsequent fingerprint detection session list generation stage.

[0049] Furthermore, the edge nodes perform hierarchical threshold registration processing on the contact value field and generate a data importance hierarchy table structure. The hierarchical threshold registration refers to mapping the contact value field to discrete importance levels and registering hierarchical threshold fields. These hierarchical threshold fields include at least a high-level threshold field, a medium-level threshold field, and a low-level threshold field, and also include a threshold caliber audit field and a threshold version number field. The hierarchical threshold registration is performed by the hierarchical threshold registration unit. This unit reads the contact value fusion rule field and the weight source field, first generating a threshold candidate set based on the distribution of contact value fields in this batch according to the measurement point object identifier, and then combining the configuration version number field and the deployment version number field in the session audit field to solidify the threshold caliber. When the configuration version number field or the deployment version number field changes, the threshold version number field is updated and written to the threshold change audit field. The threshold change audit field maintains the same field name as the audit field encapsulation field in subsequent policy version records. The hierarchical threshold registration unit uses the missing measurement impact marker field as a boundary constraint when generating the threshold candidate set. This marker field indicates the identifier of the measurement point object significantly affected by missing measurements. The hierarchical threshold registration unit writes its connection value field into the candidate threshold exclusion cache and records the exclusion reason field. The value of the exclusion reason field is consistent with that of the missing measurement reason field to avoid threshold drift caused by missing measurements. After the threshold is solidified, the hierarchical threshold registration unit compares the connection value field of each measurement point object identifier with the hierarchical threshold field, generates an importance hierarchical field, and writes it into the hierarchical mapping table. The importance hierarchical field is a discrete level and retains the hierarchical basis field, which references the version identifier of the connection value fusion rule field.

[0050] In the engineering scenario implementation, the edge node forms window matching and adjacent window matching with the spindle speed field and equipment status field of the machining process unit in multiple sampling batches. The lag step size field is concentrated in a small step size range, the direction consistency field shows consistency, and the connection value field is registered as high level. The defect event field of the quality inspection process unit forms adjacent window matching with the equipment status field of the machining process unit. However, in some batches, the candidate change node is merged due to the influence of the event-triggered debouncing window, and the connection value field is registered as medium level. For some auxiliary process quantity fields, the missing measurement segment mark table shows the connection disconnection waiting window, which causes the missing measurement impact mark field to appear. The hierarchical threshold registration unit includes them in the candidate threshold exclusion cache and records the exclusion reason field in the threshold caliber audit field, so that the threshold version number field and the configuration version number field of the session audit field form a traceable correspondence.

[0051] After completing the linkage strength calculation, connection value generation, and hierarchical threshold registration, the edge node generates a data importance hierarchy table structure. This data importance hierarchy table structure includes a session primary key, threshold version number field, hierarchical threshold field, hierarchical mapping table, measurement point object identifier, connection value field, importance hierarchy field, hierarchical basis field, and threshold caliber audit field. It also retains a missing measurement impact marker field for caliber control in subsequent passive statistical collection item binding processing. The edge node writes the data importance hierarchy table structure as an output field name into the hierarchy table storage area and uses it as an input field name for S310 for network path identifier registration, active probe task orchestration, and passive statistical collection item binding processing. Simultaneously, the importance hierarchy field in the data importance hierarchy table structure participates in action set binding processing during the subsequent elastic transmission strategy graph construction phase and forms a sequential connection with the importance weight field in the collection session configuration structure, facilitating the encapsulation of the same audit link during the strategy version record generation phase. In summary, the technical effects of this step are as follows: This step transforms the alignment sequence of changing nodes into linkage strength records and generates a connection value field. Then, it forms a data importance classification table structure through hierarchical threshold registration, linking the importance classification source with the sampling batch identifier field, alignment audit field, and threshold version number field. Compared to classification methods that rely solely on fixed field tables or manual rules, this step introduces a missing test impact marker field and a threshold caliber audit field. The classification threshold maintains a consistent versioned registration caliber when the configuration version number field changes, thereby reducing the disturbance of missing tests and configuration changes to the classification mapping table.

[0052] Step S300 includes at least steps S310-S330: S310. Based on the collection session configuration structure and data importance classification table structure, network path identifier registration, active detection task orchestration and passive statistical collection item binding processing are performed to obtain a fingerprint detection session list. Specifically, this step is initiated by the industrial IoT edge node after the sampling batch trigger event corresponding to the session primary key arrives, and reads the acquisition session configuration structure and data importance classification table structure output from the previous steps as input. The acquisition session configuration structure provides a channel mapping table, a sampling time table, a session audit field, and a set of degradable fields. The data importance classification table structure provides the correspondence between the measurement point object identifier and the importance classification field, the connection value field, and the threshold version number field. The edge node first generates a network path identifier at the channel mapping table dimension. The network path identifier refers to the transmission path description field between the acquisition channel identifier on the end side and the aggregation side. Its structure includes at least the access port identifier, uplink identifier, gateway forwarding path identifier, link medium identifier, bearer network identifier, and path validity period field. Among them, the access port identifier comes from the deployment location mapping result, the uplink identifier comes from the edge node network configuration, the gateway forwarding path identifier comes from the gateway routing table snapshot, the link medium identifier comes from the transmission medium field, and the bearer network identifier comes from the network access policy record. Edge nodes perform field normalization on the above fields. Field normalization includes fixing the identifier prefix, mapping the caliber dictionary, filling in default values ​​and resolving conflicts, and write the normalized network path identifier into the path identifier table. The path identifier table is synchronously written to the path audit field. The path audit field references the configuration version number field and the deployment version number field in the session audit field, which are used for audit association in the subsequent policy version record generation stage.

[0053] After completing the network path identifier registration, the edge node performs active probe task orchestration. An active probe task refers to a task set in which a probe generates probe messages and sends them to the sending port indicated by the network path identifier, then collects acknowledgments at the receiving end and generates probe result records. An active probe task includes at least a probe task identifier, a probe frequency field, a probe window field, a probe message type field, a probe timeout field, and a failure retry window field. The probe is the probe task execution module within the edge node; its input is the path identifier table and the sampling time table, and its output is the probe result record. During orchestration, the edge node expands the sampling time table into a probe window field and registers the probe frequency field hierarchically according to the importance level field: for the measure point object identifier set summary with a high importance level field, a denser probe window field is orchestrated and a shorter failure retry window field is registered; for the measure point object identifier set summary with a low importance level field, a sparser probe window field is orchestrated and a longer failure retry window field is registered. The minimum set of probe message type fields includes timestamp probe messages and link reachability probe messages. Both types of messages are encapsulated by the probe using a unified field template. The field template includes a path identifier reference field, a session primary key field, a probe task identifier field, and a sending time anchor point field. If the acquisition channel constraint field has a strict constraint on the message size upper limit field, the probe uses a compressed field template and writes the compression mark into the audit extension field of the probe result record.

[0054] Furthermore, edge nodes perform passive statistical collection item binding processing within the same orchestrated link. Passive statistical collection items refer to a set of collection items that periodically or event-wise sample the operational statistical fields of network interfaces, switching forwarding units, cellular access units, or gateway forwarding units. Their structure includes at least a statistical item identifier, a collection source identifier, a sampling rhythm field, a list of statistical fields, and a statistical caliber field. The minimum set of the statistical field list includes round-trip delay statistics, delay jitter statistics, packet loss ratio statistics, and queue backlog statistics. The statistical caliber field includes at least a window aggregation caliber and anomaly fragment marking caliber. The retransmission count statistics, link switching count statistics, and signal strength statistics are extended fields, bound when the bearer network identifier is a cellular network or multiple uplink identifiers exist. Edge nodes bind passive statistical collection items to collection channel identifiers according to network path identifiers, and write the importance classification field in the data importance classification table structure into the binding record. When the importance classification field is at a high level, the binding record registers a shorter sampling rhythm field, and expands the statistical field list to include the link switching count statistical field, which facilitates the identification of path jitter in the subsequent hysteresis parameter registration stage.

[0055] After completing the network path identifier registration, active probe task orchestration, and passive statistical collection item binding processes, the edge node generates a fingerprint probe session list. The fingerprint probe session list is structured list data oriented towards probe task execution and statistical collection execution. It includes at least a session primary key, network path identifier, probe task identifier list, probe frequency field, probe window field, passive statistical collection item list, summary of the bound measurement point object identifier set, importance classification field, threshold version number field, and session audit field reference. The edge node writes the fingerprint probe session list to the probe session list storage area as the output field name for this step, and uses this fingerprint probe session list as the input field name for subsequent step S320, for extracting probe and statistical results from the fingerprint probe session list, and for executing network status indicator normalization and sliding window aggregation calls. Simultaneously, the network path identifier in the fingerprint probe session list is referenced as a path constraint field for action set binding in the subsequent elastic transmission strategy graph construction stage, and the session audit field reference in the fingerprint probe session list participates in audit field encapsulation during the strategy version record generation stage.

[0056] S320. Extract the detection results and statistical results from the fingerprint detection session list, perform network status index normalization, sliding window aggregation and fingerprint vector splicing to generate a network status fingerprint matrix. Specifically, this step is triggered by the fingerprint aggregation unit of the industrial IoT edge node according to the detection window field. The fingerprint aggregation unit reads the fingerprint detection session list and locates the detection task identifier list, then reads the detection result record from the detection result cache and the statistical result record corresponding to the passive statistical collection item from the statistical collection cache. The detection result record includes at least a path identifier reference field, a detection task identifier field, a sending time anchor field, a receiving time anchor field, a receipt status field, and a timeout flag field; the statistical result record includes at least a statistical item identifier, a collection source identifier, a sampling time anchor field, a list of statistical fields, and an abnormal fragment flag field. During the extraction phase, the fingerprint aggregation unit performs time alignment on the detection result record and statistical result record under the same path identifier reference field. Time alignment uses a unified time anchor field as a reference, and records that cannot be aligned are written to an alignment failure record table, which includes a session primary key, a path identifier reference field, a failure reason field, and a registration timestamp field.

[0057] After extraction and time alignment, the fingerprint aggregation unit performs network status indicator normalization processing. Network status indicators refer to the set of network operation indicator fields calculated from detection result records and statistical result records. The minimum set includes round-trip latency, latency jitter, packet loss ratio, and queue backlog indicators. Link reachability ratio, retransmission ratio, link switching frequency, and signal strength variation amplitude indicators are used as an extended set, and their generation is determined by whether the corresponding fields are present in the bearer network identifier and statistical field list. Network status indicator normalization processing includes unit standardization, outlier pruning, missing test marker filling, and baseline merging. Unit standardization converts the units of statistical fields from different data acquisition sources to a unified standard and writes the conversion rules into the normalization audit field. Outlier pruning writes records indicating failure in the receipt status field or timeout in the timeout marker field into the outlier fragment marker field and registers the outlier placeholder value marker in the normalization output. Missing test marker filling writes the missing test mask field into the window that matches the alignment failure record table and registers the missing test reason field. Baseline merging is maintained by the fingerprint aggregation unit in the path identifier reference field dimension of the baseline cache. The baseline cache is isolated by the session primary key and contains the indicator distribution summary field of the most recent window. When the session audit field references the configuration version number field or the deployment version number field changes, the baseline cache switches to the new version and writes to the baseline switch audit field, so that the subsequent state mapping rule loading stage can refer to the consistent version caliber.

[0058] Furthermore, the fingerprint aggregation unit performs sliding window aggregation processing. Sliding window aggregation refers to the processing link that performs windowed aggregation of normalized network status indicators within the time range defined by the probe window field, generating a window indicator vector. The window indicator vector includes at least the window start and end time field, the indicator mean field, the indicator fluctuation range field, and the anomaly ratio field, and retains the anomaly segment marking caliber for subsequent hysteresis parameter registration. The window length field and step field of sliding window aggregation come from the probe window field of the fingerprint probe session list. Within the same window, the fingerprint aggregation unit uses a consistent aggregation caliber for probe result records and statistical result records: for round-trip delay indicators and delay jitter indicators, the mean is calculated after extracting steady-state segments; for packet loss ratio indicators, the ratio is calculated after counting aggregation; and for queue backlog indicators, peak aggregation is used and the peak position field is registered. If the statistical field list includes a link switching count statistics field, the fingerprint aggregation unit counts link switching events within the window and generates a link switching frequency indicator, while simultaneously writing this indicator into the window indicator vector and registering the event source marking field.

[0059] After completing the sliding window aggregation, the fingerprint aggregation unit performs fingerprint vector concatenation. Fingerprint vector concatenation refers to the process of concatenating window indicator vectors under the same session primary key and the same network path identifier into a fingerprint vector according to a fixed field serialization order, and then writing the fingerprint vector into the fingerprint matrix row. The fixed field serialization order comes from the fixed field order rules in the normalized audit fields and maintains consistency with the field serialization order field registered in the previous measurement point object field template. The minimum set of fingerprint vectors includes the window start and end time field, round-trip latency indicator, latency jitter indicator, packet loss ratio indicator, queue backlog indicator, and anomaly ratio field; the link switching frequency indicator and signal strength change amplitude indicator are concatenated as extended fields when the carrying network identifier meets the conditions. During concatenation, the fingerprint aggregation unit writes the path identifier reference field and threshold version number field into the fingerprint vector header and writes the session primary key into the fingerprint matrix index field, facilitating subsequent status category determination by tracing back to the data importance hierarchy table structure according to the session primary key.

[0060] In the engineering scenario implementation, the same production line edge node has dual uplink identifiers. The primary link is the forwarding path from the workshop wired industrial network to the factory aggregation gateway, and the backup link is the forwarding path from the cellular network access to the operation-side gateway. The fingerprint detection session list binds a high-level measurement point object identifier set summary of the processing technology unit under the network path identifier corresponding to the primary link, and arranges denser detection window fields; under the network path identifier corresponding to the backup link, it binds a medium-level measurement point object identifier set summary of the quality inspection technology unit, and arranges sparser detection window fields. The fingerprint aggregation unit simultaneously reads the queue backlog statistics field of the switching and forwarding unit and the receipt status field of the detection probe within the primary link window, and simultaneously reads the signal strength statistics field and link switching count statistics field of the cellular access unit within the backup link window. After network status indicators are normalized and converged by sliding window, the main link window indicator vector shows a combination of increased queue backlog and increased packet loss ratio, while the backup link window indicator vector shows a combination of increased link switching frequency and fluctuating signal strength change. The fingerprint vector splicing process writes the two types of combinations into different fingerprint matrix rows and retains the corresponding network path identifier in the matrix index field.

[0061] After completing network status indicator normalization, sliding window aggregation, and fingerprint vector concatenation, the fingerprint aggregation unit generates a network status fingerprint matrix. The network status fingerprint matrix is ​​two-dimensional structured data. The row index consists of the session primary key, network path identifier, and window start and end time fields. The column fields consist of the indicator fields within the fingerprint vector and include references to the normalization audit field and baseline switching audit field. Edge nodes write the network status fingerprint matrix into the fingerprint matrix storage area as the output field name for this step and use it as the input field name for subsequent step S330, for loading state mapping rules, determining state categories, and registering hysteresis parameters. Simultaneously, the matrix index field of the network status fingerprint matrix can be used in the subsequent elastic transmission strategy graph construction phase to locate the network status categories and input criteria required by the policy nodes.

[0062] S330. Load state mapping rules, determine state categories, and register hysteresis parameters on the network state fingerprint matrix to generate a network state category structure. Specifically, this step is triggered by the state determination unit of the industrial IoT edge node when a new window row is added to the network state fingerprint matrix. The state determination unit reads the row index field of the network state fingerprint matrix and groups it according to the network path identifier, and then loads the state mapping rules. The state mapping rule refers to the set of rules that map the combination of indicator fields in the fingerprint vector to discrete network state categories. Its structure includes at least a rule identifier, an entry condition field, an exit condition field, a condition threshold field, a rule priority field, and a rule version number field. The entry condition field and the exit condition field both reference the indicator field names in the fingerprint vector. The condition threshold field is provided by the threshold table and written into the threshold caliber audit field. The rule priority field is used to select the effective rule when multiple rules are hit at the same time. The state mapping rule loading is read from the rule base by the state determination unit. The rule base is stored locally on the edge node and indexed by the rule version number field. The state determination unit writes the read rule version number field into the rule loading audit field and associates it with the baseline switching audit field of the network state fingerprint matrix. When the rule version number field changes in accordance with the configuration version number field of the collection session configuration structure, the state determination unit performs rule hot switching and writes the hot switching record. The hot switching record includes the session primary key, network path identifier, old rule version number field, new rule version number field, and registration timestamp field.

[0063] After loading the state mapping rules, the state determination unit performs state category determination. State category determination refers to the process of matching the fingerprint vector corresponding to each network path identifier and each window start and end time field in order of rule priority field into the condition field and generating the network state category field. In this embodiment, the value of the network state category field includes five categories: stable, congested, intermittent packet loss, path switching, and signal fluctuation. If the rule library has an extended category, it is appended to the extended category flag field and the rule identifier reference field is retained. When matching, the state determination unit first reads the anomaly ratio field and the missing test mask field of the fingerprint vector. For window rows where the anomaly ratio field exceeds the threshold or the missing test mask field is matched, the state determination unit writes an untrusted flag field and transfers the window row to the conservative determination branch. The conservative determination branch still performs rule matching, but when outputting the network state category field, it simultaneously writes the trust level field and lowers the trust level. The status determination unit performs consistency checks on the determination results of the same path within a continuous window. The consistency check determines whether status rollback is allowed based on the exit condition field, and writes the check result into the status audit field. The status audit field and the audit field encapsulation field of the subsequent policy version record use the same field name to facilitate cross-step audit traceability.

[0064] Furthermore, the state determination unit performs hysteresis parameter registration before outputting the network state category field. Hysteresis parameters refer to the set of suppression parameters that control the switching of the network state category field between windows. Their structure includes at least an entry hold window field, an exit hold window field, a minimum dwell window field, and a dual threshold caliber field. The entry hold window field limits the number of windows that must be continuously satisfied after the entry condition field is hit; the exit hold window field limits the number of windows that must be continuously satisfied after the exit condition field is hit; the minimum dwell window field limits the shortest window span for maintaining the state category field; and the dual threshold caliber field uses different condition threshold fields for the entry and exit condition fields. The minimum set of hysteresis parameter registration includes the entry hold window field and the minimum dwell window field. The exit hold window field and the dual threshold caliber field are registered as extended fields when they exist in the rule base declaration. The status determination unit writes the hysteresis parameters into the hysteresis parameter registration table and creates an index by the session primary key, network path identifier and rule version number fields. When the link switching frequency index exists in the fingerprint vector and exceeds the threshold, the status determination unit raises the minimum dwell window field to a stricter standard and writes it into the hysteresis adjustment audit field to suppress the frequent oscillation of path switching categories between adjacent windows.

[0065] In the engineering scenario implementation, if the network path identifier corresponding to the primary link shows an increase in queue backlog and a corresponding increase in packet loss ratio within a continuous window, the status determination unit matches the congestion rule according to the entry condition field and writes it as congestion in the network status category field. If the network path identifier corresponding to the backup link shows an increase in link switching frequency and a corresponding fluctuation in signal strength variation within a continuous window, the status determination unit matches the path switching rule according to the entry condition field and writes it as path switching in the network status category field. Because the abnormal percentage field of the backup link window row is relatively high, the status determination unit registers the credibility level field as a lower level and increases the entry holding window field caliber in the hysteresis parameter registration table. The network status category field is only stably registered after the entry condition field has been continuously matched in multiple windows. The rule loading audit field and hysteresis adjustment audit field in the above determination process are both written into the status audit field, which facilitates the encapsulation of audit links during the policy version record generation stage.

[0066] After completing the loading of state mapping rules, state category determination, and hysteresis parameter registration, the edge node generates a network state category structure. This structure includes at least the session primary key, network path identifier, window start and end time field, network state category field, trust level field, rule identifier reference field, rule version number field, hysteresis parameter registration table reference, and state audit field. It also retains the threshold version number field to establish a correspondence with the data importance classification table structure. The edge node writes the network state category structure into the state category storage area as the output field name for this step and uses it as the input field name for subsequent step S410 for policy node definition, state switching edge definition, and action set binding processing. Simultaneously, the state audit field is referenced in the audit field encapsulation process of subsequent step S420, and the rule version number field and threshold version number field form a traceable version alignment relationship in the policy version record. In summary, the technical effects of this step are as follows: This step transforms the network state fingerprint matrix into a network state category structure with a rule version number field and a hysteresis parameter. State switching no longer relies on real-time judgment of single-point thresholds, but instead introduces a combination of entry condition fields, exit condition fields, and minimum dwell window fields. At the same time, active detection results and passive statistical results are unified and included in the audit field within the same window. The congestion, packet loss, handover, and fluctuation states at the network path identification level have input fields that can be directly called when constructing the elastic transmission strategy graph in the future.

[0067] Step S400 includes at least steps S410-S430: S410. Based on the network state category structure, perform policy node definition, state switching edge definition, and action set binding to obtain the elastic transmission policy graph. Specifically, this step is triggered by the policy orchestration unit of the industrial IoT edge node after receiving the network state category structure output from the previous step. The network state category structure, along with the acquisition session configuration structure and the data importance classification table structure, are used as inputs for joint orchestration. The policy orchestration unit is a policy generation module deployed on the control plane of the edge node. Its inputs include a session primary key, network path identifier, network state category field, trust level field, rule version number field, hysteresis parameter registration table reference, importance classification field, degradable field set, threshold version number field, and session audit field reference. The policy orchestration unit first executes the policy node definition. The policy node refers to a discrete state unit in the policy graph. A policy node consists of a node identifier field, a session primary key field, a network path identifier field, a network state category field, an importance classification field, a trust threshold field, an entry condition reference field, and an exit condition reference field. The node identifier field is generated by concatenating the session primary key field with the network path identifier field, network state category field, and importance classification field, and is registered as the node primary key. The entry condition reference field and the exit condition reference field reference the entry hold window field, the exit hold window field, the minimum dwell window field, and the dual threshold caliber field in the network state category structure, respectively. The credibility threshold field references the judgment caliber of the credibility level field and writes it into the node constraint field. When the credibility level field is at a low level, the node constraint field registers a conservative constraint flag. The conservative constraint flag triggers a conservative action selection branch in the subsequent action set binding stage.

[0068] After defining the policy nodes, the policy orchestration unit executes the state switching edge definition. The state switching edge refers to the directed edge connecting two policy nodes in the policy graph. It consists of an edge identifier field, a source node identifier field, a target node identifier field, a trigger condition field, an edge effective window field, an edge priority field, and an edge audit field. The trigger condition field is generated driven by a network state category field change event. The policy orchestration unit reads the window start and end time field, the state audit field, and the hysteresis parameter registration table reference from the network state category structure. Based on the minimum dwell window field, it generates an effective window field and writes the entry hold window field and exit hold window field into the continuous satisfaction criteria of the trigger condition field. For multiple network state category field switches occurring in the same network path identifier field, the policy orchestration unit registers the edge priority field using a combination of the rule version number field and the threshold version number field. The edge priority field is used as the adjudication criterion in multi-sided competition scenarios. When the path switching category and the congestion category alternate in adjacent windows, the policy orchestration unit registers a suppression flag in the trigger condition field and references the hysteresis adjustment audit field to prevent edges from frequently flipping within short windows. The audit field is associated with the session audit field and includes references to the rule version number field, threshold version number field, and baseline switch audit field, which serve as audit material for the subsequent policy version record generation stage.

[0069] Furthermore, the strategy orchestration unit performs action set binding processing. The action set refers to the set of transmission control actions executed by the data plane during the effective period of a certain strategy node. The action set consists of an action set identifier field, an action type field, an action parameter field, an action constraint field, and an action audit field. The action type field includes at least three categories: queue actions, compression / reduction actions, and retransmission actions. The action parameter field corresponds to queue configuration parameters, compression / reduction configuration parameters, and retransmission window configuration parameters, respectively. The action constraint field references the set of degradable fields and the importance classification field. The minimum parameter set for queue actions includes a priority queue identifier field, a queue quota field, a queue level trigger condition field, and an enqueue shaping parameter field. The minimum parameter set for compression / reduction actions includes a field selection mask field, a compression method identifier field, a reduction rhythm field, and an abnormal fragment retention flag field. The minimum parameter set for retransmission actions includes a retransmission window length field, a retransmission trigger condition field, a deduplication index field, and a retransmission batch identifier field. When binding, the strategy orchestration unit uses the importance classification field in the data importance classification table structure as the primary factor for action selection, the network status category field as the primary factor for action intensity, and limits the set of degradable fields to the field selection range for compression and reduction actions. When the network status category field is stable, the queue action registers the regular queue quota field and closes the congestion trigger branch of the retransmission trigger condition field. When the network status category field is congested or intermittent packet loss, the queue action registers a stricter queue level trigger condition field, the compression and reduction action only covers the set of degradable fields in the field selection mask field, and the retransmission action registers the trigger condition field for delayed retransmission and writes the generation caliber of the retransmission batch identifier field. When the network status category field is path switching or signal fluctuation, the strategy orchestration unit registers the path constraint flag in the action constraint field and expands the deduplication index field to cover the network path identifier field, so as to maintain a consistent caliber for subsequent retransmission deduplication in dual-path coexistence scenarios. The action audit field is merged with the side audit field to generate the action binding audit field. The action binding audit field records the node identifier field, action set identifier field, parameter caliber summary field and registration timestamp field.

[0070] In an engineering scenario embodiment, a production line edge node has two sets of network path identifiers: wired and cellular. The data acquisition session configuration structure registers the summary of the measurement point object identifier set for the processing unit as high-level and the summary of the measurement point object identifier set for the quality inspection unit as medium-level. Image summary and auxiliary process quantity fields are registered in the degradable field set. The network state category structure determines congestion on the wired path and path switching on the cellular path. Based on this, the policy orchestration unit generates two sets of policy nodes and defines cross-node state switching edges: congested nodes on the wired path are bound to queue actions similar to compression and downsampling actions, and the retransmission action is registered as a delayed retransmission branch; path switching nodes on the cellular path are bound to queue actions similar to retransmission actions, and the network path identifier field is overwritten in the deduplication index field. The strategy orchestration unit writes the aforementioned nodes, edges, and action sets into the graph structure storage area and generates an elastic transmission strategy graph. The elastic transmission strategy graph includes at least a node table, an edge table, an action set table, an action binding table, a graph version candidate field, and a graph audit field. The graph audit field references the session audit field and records the action binding audit field. The elastic transmission strategy graph is generated as the output field name for this step and serves as the input field name for strategy version record generation in subsequent steps (S420). It also serves as the input field name for the execution configuration orchestration of queue configuration processing, compression and downsampling configuration processing, and retransmission window configuration processing in S430.

[0071] S420. Extract the version number field and the effective scope field from the elastic transport strategy diagram, perform strategy version record generation, rollback point registration and audit field encapsulation, and generate strategy version record; Specifically, this step is triggered by the version management unit of the industrial IoT edge node after the elastic transmission strategy graph is generated. The version management unit reads the elastic transmission strategy graph from the graph structure storage area as input, and simultaneously reads the session audit field reference, rule version number field, threshold version number field, and baseline switch audit field reference as version caliber material. The version number field refers to the unique identifier field of the strategy version. When generating the version number field, the version management unit uses the session primary key field, graph version candidate field, and registration timestamp field to generate the version primary key, and writes it into the version number field after the version primary key is generated. When the graph version candidate field under the same session primary key field is consistent with the previous version, the version management unit still generates a new version number field and registers a mark that the difference is empty in the difference summary field. The difference summary field is used for subsequent rollback adjudication. The effective scope field refers to the object boundary description field to which the strategy version applies. The effective scope field includes at least the session primary key range field, network path identifier range field, measurement point object identifier set summary range field, importance classification range field, and effective time window field. The version management unit extracts the network path identifier field set from the node table of the elastic transmission strategy diagram, the importance classification field set from the action binding table, and the summary range of the measurement point object identifier set involved in the field selection mask field from the action set table. It also merges the effective window fields around the window start and end time field to generate the effective time window field. When the configuration version number field or deployment version number field of the acquisition session configuration structure changes, the version management unit registers the version switch effective point field in the effective scope field. The version switch effective point field maintains the same reference caliber as the baseline switch audit field.

[0072] After extracting and registering the version number and effective scope fields, the version management unit generates policy version records. These policy version records are structured records for policy issuance, tracing, and rollback, and include at least the version number, effective scope, graph summary, node summary, edge summary, action summary, rule version number, threshold version number, and difference summary fields. The graph summary field is generated by calculating the summary fingerprint from the node table, edge table, action set table, and action binding table of the elastic transmission policy graph. The summary fingerprint is written to the graph summary field and simultaneously to the verification field. The node summary field registers the summary of the node identifier field set, the edge summary field registers the summary of the edge identifier field set, and the action summary field registers the summary of the action set identifier field set and includes a summary field of key action parameter definitions. During the generation of policy version records, the version management unit performs rollback point registration as a parallel process. The rollback point refers to a rollback anchor point that can be used to restore to an existing policy version. The rollback point consists of a rollback point identifier field, a target version number field, a rollback reference field, a rollback verification field, and a rollback registration timestamp field. The version management unit retrieves the policy version record snapshot corresponding to the target version number field in the local version repository and writes the snapshot address into the rollback reference field. At the same time, it writes the graph summary field and the verification field into the rollback verification field. When the target version number field snapshot does not exist in the version repository, the version management unit registers a rollback unavailable flag and writes the missing reason field into the rollback registration record. The missing reason field and the audit field encapsulation link have the same field name, which is convenient for explaining the abnormal handling of the version repository during subsequent review.

[0073] Furthermore, the version management unit performs audit field encapsulation. Audit field encapsulation refers to the process of writing audit fields generated by the policy generation link, data classification link, and network judgment link into the policy version record in a unified format. The version management unit reads graph audit fields, edge audit fields, and action binding audit fields from the elastic transmission policy graph; reads status audit fields and hysteresis adjustment audit fields from the network status category structure; reads threshold caliber audit fields and threshold change audit fields from the data importance classification table structure; and reads session audit fields, configuration version number fields, and deployment version number fields from the acquisition session configuration structure. These fields are then merged and written into the audit encapsulation fields of the policy version record. The audit encapsulation fields at least include a source identifier field, a registration timestamp field, a version caliber field, a difference summary field, an anomaly record field, and a processing caliber field. When there is a conservative judgment branch caused by an alignment failure record table or a missing test fragment marker table, the version management unit writes the untrusted marker field and the conservative constraint marker into the anomaly record field, and writes the corresponding processing caliber field into the processing caliber field. After the audit field encapsulation is completed, the version management unit generates a policy version record and writes it to the version record storage area. The policy version record is used as the output field name for this step and as the input field name for S430 in subsequent steps. It is linked with the elastic transmission policy diagram for the execution configuration solidification of queue configuration processing, compression and reduction configuration processing, and retransmission window configuration processing. At the same time, the version number field in the policy version record is written to the device-side execution status record as the version binding field of the execution configuration during the data delivery stage, which is convenient for subsequent traceability.

[0074] S430: Based on the elastic transmission strategy diagram and strategy version record, execute queue configuration processing, compression and downsampling configuration processing, and retransmission window configuration processing to generate the elastic transmission execution configuration structure. Specifically, this step is triggered by the execution configuration generation unit of the industrial IoT edge node after the policy version record is generated. The execution configuration generation unit reads the policy version record from the version record storage area and reads the elastic transmission policy graph from the graph structure storage area as input. Subsequently, it completes the solidification and distribution preparation of the three types of execution configurations on the control plane. The elastic transmission execution configuration structure refers to the unified encapsulation structure of queue configuration, compression and reduction configuration, and retransmission window configuration required for data plane operation. It includes at least the session primary key field, version number field, effective scope field, network path identifier field, node identifier field, action set identifier field, queue configuration segment, compression and reduction configuration segment, retransmission window configuration segment, and execution audit field. The execution configuration generation unit first performs queue configuration processing, which generates queue configuration segments for entries in the action set table whose action type field is queue action. The queue configuration section includes at least the following fields: priority queue identifier, queue quota, queue level trigger condition, enqueue shaping parameter, dequeue scheduling parameter, and drop policy flag. The priority queue identifier field is generated by mapping fields according to their importance. The queue quota field registers a regular quota when the network state category is stable, and a compressed quota with a congestion threshold field when the network state category is congested or experiencing intermittent packet loss. The enqueue shaping parameter field is generated jointly by the sampling rhythm orchestration caliber and the queue level trigger condition field. It includes at least the burst window field, smooth window field, and shaping start / stop condition field. The dequeue scheduling parameter field registers the inter-queue polling caliber, weight caliber, and dwell window caliber, and writes the minimum dwell window field into the dwell window caliber to avoid scheduling jitter caused by frequent queue switching during state transitions. The discard policy flag field is not a fixed discard action, but a placeholder flag after the set of degradeable fields is compressed and reduced in size. When the configuration generation unit registers the discard policy flag field, it associates the field selection mask field with the abnormal fragment retention flag field to prevent abnormal fragments from being covered by compression and reduction.

[0075] After completing the queue configuration processing, the execution configuration generation unit performs compression and reduction configuration processing. This processing generates compression and reduction configuration segments for entries in the action set table whose action type field is "compression and reduction action". Each compression and reduction configuration segment includes at least the following fields: field selection mask field, field template reference field, compression method identifier field, compression dictionary identifier field, reduction rhythm field, reduction trigger condition field, abnormal fragment retention flag field, and compression audit field. The field selection mask field is derived from the elastic transmission strategy graph action binding table and is limited to the set of degradable fields. The field template reference field references the template version number field of the measurement point object field template mapping table. The compression method identifier field registers the lossless or lossy compression caliber. The compression dictionary identifier field, when structured fields exist, registers the dictionary snapshot identifier and writes the snapshot address to the dictionary reference field. The sampling reduction rhythm field is generated jointly by the sampling time table and the network status category field. The sampling reduction trigger condition field references the queue level trigger condition field and the packet loss ratio threshold, and writes the trigger hit record into the compression audit field. When the network status category field is path switching or signal fluctuation, the compression and sampling reduction configuration section adds a path stability window constraint field to the sampling reduction trigger condition field. The path stability window constraint field references the entry hold window field and the minimum dwell window field to avoid frequent changes in the sampling reduction caliber due to short window jitter. The abnormal segment retention mark field is an essential field of the minimum set in this step. The execution configuration generation unit associates the missing segment mark table merge mark field with the missing boundary mark in the change node alignment sequence and writes it into the retention rule field. The retention rule field stipulates that abnormal segments are still output according to the original field template during the compression and sampling reduction action, to avoid abnormal segments having their field caliber changed by compression and sampling reduction.

[0076] Furthermore, the configuration generation unit performs retransmission window configuration processing, which generates retransmission window configuration segments for entries in the action set table whose action type field is a retransmission action. Each retransmission window configuration segment includes at least the following fields: retransmission window length, retransmission trigger condition, retransmission batch identifier, retransmission queue identifier, deduplication index, retransmission priority, retransmission storage quota, and retransmission audit. The retransmission window length field is generated by merging the effective time window field with the effective window field on the same side, and is combined with the queue level trigger condition field to register a delayed retransmission window. The retransmission trigger condition field references the network state category field to switch from a congested or intermittent packet loss state to a stable state. The trigger condition field also references the exit hold window field to avoid triggering a large number of retransmissions immediately after the rollback, preventing the queue from accumulating again. The retransmission batch identifier field is generated based on the session primary key field and the version number field, and is incremented each time a retransmission action is triggered. The retransmission queue identifier field is linked to the priority queue identifier field, registering a higher retransmission priority field when the importance level field is high. The deduplication index field serves as the index for deduplication in retransmissions. Its minimum set includes the measurement point object identifier, the unified time anchor point field, and the version number field. When the network path identifier field exists in a dual-path scenario, the deduplication index field is expanded to include the network path identifier field to prevent the same data segment from being repeatedly corrupted during the dual-path retransmission phase. The retransmission storage quota field registers the retransmission buffer quota and expiration cleanup criteria on the edge node side. The expiration cleanup criteria reference the retransmission window length field and writes the cleanup record to the retransmission audit field. The retransmission audit field is associated with the execution audit field during the encapsulation phase.

[0077] After generating the three types of configuration segments, the execution configuration generation unit performs a consistency check on the node identifier field and action set identifier field with the three types of configuration segments, and writes the check result into the execution audit field. When the check result indicates that the template reference field and the template version number field are inconsistent, the execution configuration generation unit registers a configuration blocking flag and writes the blocking reason field into the execution audit field. At the same time, it triggers the version management unit to register the rollback point link, and the rollback target version number field is taken from the version number field of the previous policy version record. After completing the consistency check, the execution configuration generation unit generates an elastic transmission execution configuration structure and writes the elastic transmission execution configuration structure into the execution configuration storage area. At the same time, it writes the structure into the delivery queue to the data plane delivery interface. The delivery queue entries include a session primary key field, a version number field, a network path identifier field, a delivery timestamp field, and a confirmation receipt field. When the confirmation receipt field does not return within the specified confirmation window, the execution configuration generation unit registers a delivery failure flag in the execution audit field and writes it into the retry window field. The retry window field and the failure retry window field use the same caliber to facilitate sharing the same retry scheduler with the active probe task orchestration link.

[0078] In an engineering scenario implementation, an edge node is in a congested network state on the wired bearer path within the factory area. The corresponding strategy node in the elastic transmission strategy diagram binds queue actions to the same compression and reduction actions. The execution configuration generation unit generates a queue configuration segment and registers the congestion threshold field in the queue level trigger condition field. Simultaneously, in the compression and reduction configuration segment, the field selection mask field is limited to a set of degradeable fields, and an abnormal fragment retention flag field is registered. When a subsequent window determines that the network state category has fallen back to stable and the exit retention window field has been hit, the strategy node switches sides and triggers a retransmission action. The execution configuration generation unit registers the retransmission trigger condition field in the retransmission window configuration segment and generates a retransmission batch identifier field. The retransmission queue identifier field is registered as a high-priority queue identifier field, and the deduplication index field covers the measurement point object identifier, the unified time anchor field, and the version number field. After distribution is completed, the data plane binds the execution configuration according to the version number field. During operation, the distribution receipt field, trigger hit record, and cleanup record are all written to the execution audit field and established with a traceable association with the audit encapsulation field of the strategy version record. In summary, the technical effects of this step are as follows: This step transforms the nodes, edges, and action sets in the elastic transmission strategy diagram into deployable queue configuration segments, compression and reduction configuration segments, and retransmission window configuration segments, and solidifies them into an elastic transmission execution configuration structure under the constraints of the version number field and the effective scope field. Compared with the approach of only making temporary adjustments to the data plane, the execution configuration structure introduces the linkage between the template version number field, the deduplication index field, the abnormal fragment retention mark field, and the execution audit field. The configuration switching and rollback links have traceable version anchors and consistent audit records.

Claims

1. A method for industrial Internet of Things (IoT) data acquisition and flexible transmission, characterized in that, include: Acquire the process unit topology and monitoring target configuration, perform process unit coding registration, probe deployment scheme generation and acquisition session configuration structure registration processing, and generate acquisition session configuration structure; Based on the data collection session configuration structure, the system performs unified processing of time anchor points, generation of alignment sequences for changing nodes, and registration of hierarchical thresholds for connection values, generating a data importance hierarchy table structure. Based on the data collection session configuration structure and data importance hierarchy table structure, the fingerprint detection session list generation process, the network status fingerprint matrix construction process, and the network status category determination process are performed to generate the network status category structure. Based on the network state category structure, we construct the elastic transmission policy graph and register policy version records to generate the elastic transmission execution configuration structure.

2. The method according to claim 1, characterized in that, The process of registering process unit codes also includes: The process unit coding registration process includes generating a stable process unit code for each process unit. The stable process unit code is generated by concatenating the original process unit identifier, the production line identifier, the process section identifier, and the equipment asset identifier and then standardizing it. The standardization process includes removing whitespace, unifying capitalization, merging symbols, supplementing prefixes, and resolving duplicate conflicts. The final code is then registered in the coding field of the process unit coding table. At the same time, an audit field containing a registration timestamp field, a registration source field, and a configuration version number field is written, and a code reverse lookup index is generated simultaneously.

3. The method according to claim 1, characterized in that, The process of generating a probe deployment plan also includes: The probe deployment scheme includes probe type matching results, deployment location mapping results, sampling schedule, and channel mapping table. Based on the probe deployment scheme, a session primary key generation process is performed. This process involves generating a session primary key by combining a session time period identifier, process unit code, measurement point object identifier set summary, deployment version number field, and configuration version number field, and then obtaining the primary key string through primary key normalization. A degradable field set registration process is also performed, which includes registering a field retention list, field aggregation rules, field freeze flags, field reduction sampling rhythm field, and field compression candidate flag field. Finally, an importance weight field registration process is performed, which includes registering a measurement point object basic weight field, a monitoring topic weight field, and an alarm level weight field.

4. The method according to claim 1, characterized in that, The process of uniformly processing time anchors also includes: The unified time anchor processing includes reading the synchronization clock capability field and timestamp reference field from the time alignment probe, writing the collected records into the unified time anchor field, and generating a batch alignment index; performing missing test segment marking processing, which includes marking missing test segments and registering the missing test reason field based on the sampling time table, batch alignment index, sampling failure count field, and connection status field; constructing a multi-source time series data packet, which includes a session primary key, a sampling batch identifier field, a collection record set, a unified time anchor field set, a batch alignment index, and a missing test segment marking table; and performing measurement point object sequence extraction processing based on the multi-source time series data packet, which includes grouping by measurement point object identifier and sorting by the unified time anchor field to generate a measurement point object sequence, and writing it into the missing test mask field.

5. The method according to claim 1, characterized in that, The process of registering and processing contact value hierarchical thresholds also includes: The connection value generation process includes normalizing and aggregating the linkage intensity to generate connection value fields and group connection value fields; performing hierarchical threshold registration processing, which includes mapping the connection value fields to discrete importance levels and registering the hierarchical threshold fields; and generating a data importance level table structure, which includes a session primary key, a threshold version number field, a hierarchical threshold field, a hierarchical mapping table, a measurement point object identifier, a connection value field, an importance level field, a hierarchical basis field, and a threshold caliber audit field.

6. The method according to claim 1, characterized in that, The process of generating a fingerprint detection session list also includes: The fingerprint detection session list includes a session primary key, network path identifier, detection task identifier list, detection frequency field, detection window field, passive statistical collection item list, summary of bound measurement point object identifier set, importance classification field, threshold version number field, and session audit field reference. Based on the fingerprint detection session list, detection results and statistical results extraction processing is performed. The extraction processing includes reading detection result records from the detection result cache and reading statistical result records from the statistical collection cache, and performing time alignment. Network status indicator normalization processing is performed. The network status indicator normalization processing includes unit caliber normalization, outlier pruning, missing test marker filling, and baseline merging to generate round-trip latency indicators, latency jitter indicators, packet loss ratio indicators, and queue backlog indicators. Sliding window aggregation processing is performed. The sliding window aggregation processing includes windowed aggregation of the normalized network status indicators within the time range limited by the detection window field to generate a window indicator vector. Fingerprint vector concatenation processing is performed. The fingerprint vector concatenation processing includes concatenating the window indicator vectors into a fingerprint vector according to a fixed field serialization order.

7. The method according to claim 1, characterized in that, The process of constructing the network state fingerprint matrix also includes: The network state fingerprint matrix includes a row index consisting of a session primary key, a network path identifier, and a window start and end time field, and column fields consisting of various indicator fields within the fingerprint vector. Based on the network state fingerprint matrix, a state mapping rule loading process is performed, which includes reading state mapping rules from the rule base. The state mapping rule includes a rule identifier, an entry condition field, an exit condition field, a condition threshold field, a rule priority field, and a rule version number field.

8. The method according to claim 1, characterized in that, The network state category determination process also includes: The network status category determination process includes matching the fingerprint vector corresponding to each network path identifier and each window start and end time field with the entry condition field according to the rule priority field and generating a network status category field; performing hysteresis parameter registration processing, which includes registering the entry holding window field, exit holding window field, minimum dwell window field, and dual threshold caliber field; and generating a network status category structure, which includes the session primary key, network path identifier, window start and end time field, network status category field, trust level field, rule identifier reference field, rule version number field, hysteresis parameter registration table reference, and status audit field.

9. The method according to claim 1, characterized in that, The process of constructing a resilient transport strategy graph also includes: The elastic transmission strategy graph includes a node table, an edge table, an action set table, an action binding table, a graph version candidate field, and a graph audit field. Based on the elastic transmission strategy graph, a version number field generation process is performed, which includes generating a version number field by combining the session primary key field, the graph version candidate field, and the registration timestamp field. An effective scope field extraction process is then performed, which includes extracting the session primary key range field, the network path identifier range field, the measurement point object identifier set summary range field, the importance classification range field, and the effective time window field.

10. The method according to claim 1, characterized in that, The policy version record registration process also includes: The policy version record generation process includes generating a policy version record, which includes a version number field, an effective scope field, a graph summary field, a node summary field, an edge summary field, an action summary field, a rule version number field, a threshold version number field, and a difference summary field; performing a rollback point registration process, which includes registering a rollback point identifier field, a target version number field, a rollback reference field, a rollback verification field, and a rollback registration timestamp field; performing an audit field encapsulation process, which includes merging graph audit fields, edge audit fields, action binding audit fields, status audit fields, hysteresis adjustment audit fields, threshold caliber audit fields, threshold change audit fields, and session audit fields into an audit encapsulation field; and generating a policy version record.

Images