Industrial data acquisition method and system based on edge computing

Abstract

The application relates to the technical field of industrial Internet of Things edge computing, and discloses an industrial data acquisition method and system based on edge computing. The method comprises the following steps: dividing an edge domain according to target industrial scene configuration parameters, and dynamically configuring a data acquisition node for each edge domain; dynamically adjusting a communication link between the data acquisition node and an edge gateway according to real-time data stream characteristics reported by the data acquisition node; performing real-time analysis and load balancing processing on heterogeneous data received by the edge gateway; using a multilayer evaluation strategy deployed on the edge gateway to perform performance evaluation and anomaly detection on the processed data stream; and finally, dynamically optimizing configuration parameters and communication strategies of the data acquisition node according to the evaluation and detection results. Through dynamic adaptation of the communication link and multilayer collaborative evaluation, the reliability, real-time performance and system adaptability of industrial data acquisition are improved.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) edge computing technology, specifically to an industrial data acquisition method and system based on edge computing. Background Technology

[0002] In Industrial Internet of Things (IIoT) scenarios, to achieve state awareness and intelligent control of the entire production process, a large number of sensors and actuators need to be deployed on-site for data collection. Existing typical data acquisition solutions usually adopt a hierarchical, centralized architecture, where on-site data acquisition nodes are responsible for collecting equipment data and uploading it to an edge gateway or cloud processing center via pre-planned fixed communication paths. This fixed link configuration makes it difficult to effectively guarantee the reliability of communication paths and the real-time performance of data transmission in complex, dynamic, and frequently fluctuating industrial environments. Network congestion or single points of failure can easily lead to data loss or increased transmission delays, affecting the real-time and accurate judgment of the on-site status by upper-layer applications.

[0003] Current technologies for monitoring and managing data quality often focus on a single dimension. Common practices include monitoring network connectivity or bandwidth usage, lacking a multi-dimensional, cross-layered collaborative analysis and evaluation mechanism covering the entire data generation, transmission, and aggregation process. This single-point or single-layer monitoring model struggles to accurately identify complex problems caused by minor equipment malfunctions, network jitter, or data anomalies, failing to provide sufficient and accurate decision-making basis for the adaptive optimization of the data acquisition system. In industrial settings, heterogeneous devices with various communication protocols and data formats exist. When edge gateways aggregate this data, static load balancing strategies struggle to cope with dynamic fluctuations in data traffic caused by changes in production schedules, easily leading to uneven load on gateway processing units, creating performance bottlenecks, and impacting overall data processing efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide an industrial data acquisition method and system based on edge computing to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an industrial data acquisition method based on edge computing, the method comprising:

[0006] Obtain the configuration parameters of the target industrial scene, and divide the industrial scene into multiple edge domains based on the configuration parameters;

[0007] Based on the physical topology of each edge domain, at least one data acquisition node is dynamically configured for each edge domain;

[0008] Based on the characteristics of the real-time data stream reported by the data acquisition nodes, the communication link between the data acquisition nodes and the edge gateway is dynamically adjusted.

[0009] Based on the adjusted communication link, the heterogeneous data received by the edge gateway is parsed and load balanced in real time.

[0010] By deploying a multi-layered evaluation strategy on the edge gateway, performance evaluation and anomaly detection are performed on the data stream after load balancing.

[0011] Based on the results of performance evaluation and anomaly detection, the configuration parameters and communication strategies of the data acquisition nodes are dynamically optimized.

[0012] Preferably, the step of obtaining configuration parameters for the target industrial scene and dividing the industrial site into multiple edge domains based on the configuration parameters includes:

[0013] Collect equipment layout information, communication network topology, and real-time service requirements from the industrial site as configuration parameters;

[0014] Determine the physically adjacent device clusters based on the device layout information;

[0015] Based on the communication network topology, each device cluster and its associated communication network are divided into an independent edge domain;

[0016] Assign a unique domain identifier to each edge domain and establish an edge domain management list.

[0017] Preferably, the step of dynamically configuring at least one data acquisition node for each edge domain based on the physical topology of each edge domain includes:

[0018] Parse the physical topology of each edge domain in the edge domain management list;

[0019] Calculate the number of data acquisition nodes required for each edge domain based on the device distribution density in the physical topology;

[0020] Configure the device discovery protocol and data acquisition frequency for each data acquisition node, and establish a mapping table between data acquisition nodes and the edge domain management list;

[0021] The step of calculating the number of data acquisition nodes required for each edge domain based on the device distribution density in the physical topology includes: obtaining the physical location coordinates of all industrial devices in the edge domain, calculating the Euclidean distance between devices, calculating the device distribution density based on the Euclidean distance (the device distribution density is the number of devices per unit area), querying a preset density-node number mapping table to obtain the basic number of nodes based on the device distribution density, adjusting the basic number of nodes in conjunction with the historical data traffic peak of the edge domain, obtaining the final number of data acquisition nodes, and storing the final number of data acquisition nodes in the edge domain management list.

[0022] Preferably, the step of dynamically adjusting the communication link between the data acquisition node and the edge gateway based on the real-time data stream characteristics reported by the data acquisition node includes:

[0023] The system monitors the timestamp, data volume, and transmission rate of data packets reported by the data acquisition nodes. When the data volume exceeds a preset threshold, the link quality assessment process is initiated.

[0024] Based on the link quality assessment results, the optimal communication path is selected to establish a backup link between the data acquisition node and the edge gateway. After the backup link is successfully established, the real-time data stream is split and transmitted simultaneously on the main link and the backup link.

[0025] Preferably, the real-time parsing and load balancing of heterogeneous data received by the edge gateway based on the adjusted communication link includes:

[0026] The data packets received by the edge gateway are parsed according to the protocol, the edge domain identifier of the data source is identified, and the data packets are classified into different processing queues according to the edge domain identifier.

[0027] Monitor the load status of each processing queue. When the load of a processing queue reaches its processing capacity limit, dynamically schedule some data packets in the queue to idle processing queues, record the processing path of each data packet and update the data flow tracking table.

[0028] Preferably, the performance evaluation and anomaly detection of the data stream after load balancing through a multi-layer evaluation strategy deployed on the edge gateway includes:

[0029] In the first-level evaluation, the data throughput and processing latency of each processing queue are statistically analyzed;

[0030] In the second-level evaluation, the time-series characteristics of the data packets are analyzed to detect abnormal fluctuation patterns;

[0031] In the third-level evaluation, the characteristic differences between real-time data streams and historical normal data streams are compared;

[0032] The results of the three-tiered assessment are summarized to generate a comprehensive assessment report.

[0033] Preferably, the step of dynamically optimizing the configuration parameters and communication strategies of the data acquisition nodes based on the results of performance evaluation and anomaly detection includes:

[0034] Analyze the performance bottleneck areas identified in the comprehensive evaluation report;

[0035] For the edge domain corresponding to the performance bottleneck area, recalculate the optimal configuration parameters of the data acquisition nodes;

[0036] Update the device discovery protocol and data acquisition frequency of the data acquisition nodes based on the recalculated optimal configuration parameters;

[0037] Synchronously adjust the priority of the communication link between the data acquisition node and the edge gateway.

[0038] Preferably, the step of performing protocol parsing on the data packets received by the edge gateway to identify the edge domain identifier of the data source includes:

[0039] Extract the protocol header field of the data packet and parse the protocol type, which includes Modbus protocol or OPCUA protocol;

[0040] Match a predefined protocol template based on the protocol type, and obtain the storage location offset of the edge domain identifier from the protocol template;

[0041] Based on the storage location offset, read the binary value of the edge domain identifier from the payload of the data packet;

[0042] Convert the binary value to a string format and verify whether the string format conforms to the encoding rules of the edge domain identifier;

[0043] If the verification passes, the edge domain identifier is used for packet classification; otherwise, the packet is marked as invalid and discarded.

[0044] Preferably, the performance bottleneck areas identified in the comprehensive evaluation report include:

[0045] Performance metrics data are extracted from the comprehensive evaluation report, including data throughput, processing latency, and anomaly detection results.

[0046] Compare performance metrics data with preset thresholds to identify edge domain identifiers that are below the threshold;

[0047] Based on edge domain identifiers, locate the physical device clusters and communication paths corresponding to performance bottleneck areas;

[0048] Generate a detailed analysis report of the performance bottleneck area, including the cause of the bottleneck and optimization suggestions.

[0049] Preferably, the present invention also includes an industrial data acquisition system based on edge computing, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the industrial data acquisition method based on edge computing as described above.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] By analyzing the dynamic characteristics of data traffic, transmission latency, jitter, etc., reported by data acquisition nodes in real time, the system autonomously decides and switches the communication path for data upload or adjusts link parameters. This mechanism enables the data transmission process to proactively adapt to fluctuations in network conditions and changes in business load, avoid network congestion nodes and faulty links, improve the reliability and robustness of the data upload link, reduce the risk of data loss due to communication interruptions or delays, and ensure the real-time and continuous acquisition of critical industrial data.

[0052] A multi-layered evaluation strategy integrating network, application, and business layer metrics is deployed at the edge gateway to perform collaborative analysis and in-depth inspection of the aggregated data streams. Network layer metrics focus on transmission quality, application layer metrics examine the compliance of data content and format, and business layer metrics combine industrial logic to judge the rationality of data values. This three-dimensional evaluation approach can transcend the limitations of traditional single-layer monitoring, more accurately identifying complex faults and potential risks hidden in network jitter, minor equipment anomalies, or data generation sources. It achieves refined perception of data stream performance and quality, providing a more comprehensive and accurate foundation for subsequent optimization decisions based on anomaly detection and performance insights.

[0053] Based on the real-time and accurate system status and data quality insights output by the aforementioned multi-layered evaluation strategy, the system dynamically adjusts configuration parameters such as sampling frequency, data compression rate, and reporting cycle of downstream data acquisition nodes, and optimizes their communication strategies. This creates a closed-loop optimization system for the entire data acquisition system, from perception and evaluation to decision-making and execution. The system can self-adjust and optimize according to actual operating conditions and business needs, thereby maintaining efficient, stable, and adaptive operation in complex industrial environments, improving the intelligence level and long-term operational efficiency of the entire data acquisition infrastructure. Attached Figure Description

[0054] Figure 1 This is a schematic diagram illustrating the working principle of the edge computing-based industrial data acquisition method described in this invention.

[0055] Figure 2 A flowchart for dividing the edge region;

[0056] Figure 3 A flowchart for dynamically adjusting the communication link;

[0057] Figure 4 A comparison chart of the percentage of heavy load time before and after load balancing;

[0058] Figure 5 A comparison chart of performance indicators before and after optimization for an industrial data acquisition system. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Please see Figure 1 This invention provides an industrial data acquisition method and system based on edge computing. The method includes: acquiring configuration parameters of a target industrial scenario; dividing the industrial site into multiple edge domains based on the configuration parameters, including equipment layout, network topology, and real-time service requirements; dynamically configuring at least one data acquisition node for each edge domain based on its physical topology, with each data acquisition node responsible for acquiring data from industrial equipment; dynamically adjusting the communication link between the data acquisition node and the edge gateway based on the real-time data stream characteristics reported by the data acquisition node, including data volume, timestamp, and transmission rate; performing real-time parsing and load balancing on the heterogeneous data received by the edge gateway based on the adjusted communication link, where the heterogeneous data may originate from different protocols; and conducting performance evaluation and anomaly detection on the load-balanced data stream using a multi-layer evaluation strategy deployed on the edge gateway, involving throughput, latency, and anomaly pattern analysis. Finally, dynamically optimizing the configuration parameters and communication strategy of the data acquisition node based on the performance evaluation and anomaly detection results, including updating the acquisition frequency and link priority.

[0061] Example 1: See Figure 2In practical implementation, equipment layout information, communication network topology, and real-time service requirements from the industrial site are collected as configuration parameters. Equipment layout information is obtained through a high-precision positioning system deployed in the workshop or two-dimensional coordinates imported from computer-aided design documents. The communication network topology uses a simple network management protocol to poll network switches, generate a forwarding table, and construct a connection diagram. Real-time service requirements are issued in real time through the manufacturing execution system's application programming interface, which sends data acquisition tasks and priority instructions. Based on the equipment layout information, physically adjacent equipment clusters are determined. The division of equipment clusters is based on the premise that the Euclidean distance between any two industrial devices is less than a preset proximity threshold; for example, devices within a ten-meter radius are grouped into the same equipment cluster. Combined with the communication network topology, each equipment cluster and its directly connected communication network devices, such as switches and routers, are divided into an independent edge domain. This division process ensures that each edge domain has logically isolated network segments. A unique domain identifier is assigned to each edge domain, and an edge domain management list is established. The domain identifier is generated using a hybrid algorithm based on timestamps and geocoding. The edge domain management list is stored in a database table, containing the domain identifier, the equipment cluster member list, and network configuration information.

[0062] In practical implementation, the physical topology of each edge domain in the edge domain management list is parsed. The physical topology is represented using a graph data structure, where nodes represent industrial equipment and edges represent physical links. Based on the equipment distribution density in the physical topology, the required number of data acquisition nodes for each edge domain is calculated. Equipment distribution density characterizes the concentration of equipment per unit area. The planar coordinates of all industrial equipment within the edge domain are obtained, and the area of ​​the minimum convex polygon covering all equipment is calculated. Based on the number of equipment and the area, the equipment distribution density is calculated using the following formula:

[0063] ;

[0064] in: Indicates the density of equipment distribution. This represents the total number of industrial devices within the edge domain. This represents the area of ​​the smallest convex polygon formed by the device location coordinates. A preset density-node number mapping table is queried. This table is a two-dimensional array storing the correspondence between density intervals and recommended node numbers, obtaining the basic node number based on the device distribution density. Historical data traffic peaks in the edge domain are combined. Historical data traffic peaks are extracted from system monitoring logs, showing the maximum data transmission rate over a past period. The ratio of the peak value to the single-node processing capacity is used as a correction factor to adjust the basic node number, resulting in the final number of data acquisition nodes. The final number of data acquisition nodes is written to the corresponding field in the edge domain management list. A device discovery protocol and data acquisition frequency are configured for each data acquisition node. The device discovery protocol is selected based on mainstream industrial communication standards within the edge domain, such as Modbus TCP or OPCUA. The data acquisition frequency is set according to the sampling interval specified in the real-time business requirements. A mapping table between data acquisition nodes and the edge domain management list is established. This mapping table is a relational database table containing the data acquisition node identifier, IP address, edge domain identifier, protocol type, and acquisition frequency parameters.

[0065] In some embodiments, the process of collecting device layout information integrates RFID reader data to update device location changes in real time. In some embodiments, the collection of communication network topology employs active probing technology, sending Internet Control Message Protocol (ICP-MS) requests to network devices to construct a topology map. Optionally, the partitioning of device clusters can use a density clustering algorithm to automatically identify high-density areas. In a specific implementation, the density clustering algorithm processes the physical location coordinates in the device layout information, calculates the Euclidean distance between industrial devices, and automatically partitions device clusters based on device distribution density. The algorithm sets a core distance parameter and a minimum number of neighbors, traversing from any device point. If the number of devices in the neighborhood of a point reaches the minimum number of neighbors, the point is marked as a core point and a new cluster is created. Subsequently, all density-reachable device points are recursively grouped into the same cluster, thereby identifying high-density areas that form physically adjacent device clusters. Optionally, the generation of edge domain identifiers can be achieved by incorporating hash functions to ensure global uniqueness. It is understood that the calculation of device distribution density relies on geometric spatial analysis.

[0066] Example 2: See Figure 3In practice, the monitoring system tracks the timestamp, data volume, and transmission rate of data packets reported by the data acquisition nodes. The timestamp is written to a specific field in the packet header by the real-time clock circuit inside the data acquisition node when the data packet is assembled. The data volume is obtained by parsing the data packet length field to determine the number of bytes. The transmission rate is calculated by the network interface driver of the edge gateway, which measures the number of bits received per second. When the detected data volume exceeds a preset threshold, a link quality assessment process is initiated. The preset threshold is set to 90% of the maximum allowed number of bytes per data packet, based on the transmission capacity of the data acquisition node model. The link quality assessment process includes sending Internet Control Message Protocol echo request messages to the data acquisition nodes to detect connectivity.

[0067] In practice, based on the link quality assessment results, the optimal communication path is selected to establish a backup link between the data acquisition node and the edge gateway. The link quality assessment results include the round-trip time and data packet delivery success rate. The selection of the optimal communication path is based on the link quality score, which is calculated using the following formula:

[0068] ;

[0069] in: Indicates the link quality score. Indicates the data packet delivery success rate. This indicates the measured round-trip time delay. This indicates the maximum allowable latency threshold for the system. (Select) The path with the highest value is selected as the optimal communication path. The process of establishing a backup link involves issuing flow table entries through the northbound interface of the software-defined network controller, creating a new forwarding path on the switch between the data acquisition node and the edge gateway. After the backup link is successfully established, the real-time data stream is split between the primary and backup links for simultaneous transmission. The splitting operation employs a load balancing algorithm based on packet hash values, inputting the five-tuple information of the data packets into a hash function and selecting the transmission path based on the result.

[0070] In some embodiments, timestamp synchronization employs a precision time protocol to keep the clock deviation between the data acquisition node and the edge gateway within microseconds. In some embodiments, transmission rate statistics employ a sliding time window algorithm, calculating the average of the most recent ten time windows as the current rate. Optionally, the link quality assessment process can integrate service-aware functionality, setting different assessment weights for different types of data flows. Optionally, the establishment of backup links can employ a multipath transmission control protocol, implementing link aggregation at the transport layer. It is understood that the packet delivery success rate needs to be calculated by sending probe packets and counting the number of response packets.

[0071] Example 3: In specific implementation, the data packets received by the edge gateway undergo protocol parsing to identify the edge domain identifier of the data source. The protocol parsing process begins with extracting the protocol header fields of the data packets. The protocol header fields are extracted, and the protocol type is parsed. Protocol types include Modbus or OPCUA protocols. Extraction of the protocol header fields is achieved by parsing the payload portion of the Ethernet frame. For example, for the Modbus TCP protocol, the transaction identifier and protocol identifier fields are parsed to confirm the protocol type. A predefined protocol template is matched according to the protocol type. The protocol templates are stored in the edge gateway's local database, and each template defines the data structure specification for a specific protocol in JSON format. The storage location offset of the edge domain identifier is obtained from the protocol template. The storage location offset is an integer value representing the starting byte position of the edge domain identifier in the data packet payload. Based on the storage location offset, the binary value of the edge domain identifier is read from the data packet payload. The read operation uses pointer arithmetic to directly access the specified offset address in memory, extracting a fixed-length byte sequence. The binary value is converted to a string format. The conversion process decodes each byte into its corresponding character according to ASCII encoding rules. Verify that the string format conforms to the encoding rules of the edge domain identifier. The encoding rules require that the string must consist of uppercase letters, numbers, and hyphens, and be at least sixteen characters long. If the verification passes, the edge domain identifier is used for packet classification; otherwise, the packet is marked as invalid and discarded. Invalid packets are moved to the isolation buffer, and an alarm event is generated.

[0072] In implementation, data packets are classified into different processing queues based on edge domain identifiers. Each processing queue corresponds to an edge domain identifier, and the queues are implemented as a circular buffer structure. The load status of each processing queue is monitored, and the load status is evaluated by sampling the queue length and the average processing time of data packets in real time. When the load of a processing queue reaches its processing capacity limit, some data packets in the queue are dynamically scheduled to idle processing queues. The processing capacity limit is determined by the maximum queue depth configured by the system and the number of processor cores. The dynamic scheduling process is triggered based on the load rate indicator, which is calculated using the following formula:

[0073] ;

[0074] in: Indicates load rate, This indicates the number of data packets currently in the processing queue. This indicates the maximum capacity of the processing queue. When... When the value exceeds the threshold of 0.75, the system extracts a certain number of data packets from the head of the queue and reroutes them to idle queues with a load rate below 0.5. The processing path of each data packet is recorded and updated in the data flow trace table. The data flow trace table uses a hash table data structure and stores path information, including the source queue identifier, the destination queue identifier, and the scheduling timestamp, with the data packet sequence number as the key.

[0075] In some embodiments, protocol template matching employs a multi-pattern matching algorithm, simultaneously matching multiple protocol features to improve recognition speed. In some embodiments, the load status monitoring of the processing queue integrates a real-time stream processing engine to continuously calculate load metrics. Optionally, the verification process for edge domain identifiers can include a digital signature verification step to ensure identifier integrity. Optionally, the dynamic scheduling algorithm can introduce a packet size weighting factor, prioritizing the scheduling of smaller packets to quickly reduce queue load. It is understood that protocol parsing requires handling the conversion between network byte order and host byte order.

[0076] Example 4: In the specific implementation, during the first-layer evaluation, the data throughput and processing latency of each processing queue are statistically analyzed. Data throughput is obtained by recording the total number of data packets successfully processed by each processing queue within a fixed time window using the edge gateway's built-in counter and dividing by the time window length. Processing latency is obtained by timestamping each data packet and calculating the average time difference from entering the processing queue to leaving the processing queue. The statistical results are updated to the performance monitoring database in real time. In the specific implementation, during the second-layer evaluation, the time-series characteristics of the data packets are analyzed to detect abnormal fluctuation patterns. The time-series characteristics include the data packet arrival rate sequence and the processing time sequence. The data packet arrival rate sequence is generated by sampling the number of data packets arriving at the processing queue within each millisecond interval. The processing time sequence is generated by recording the processing time of each data packet. Abnormal fluctuation pattern detection uses a sliding window-based Z-score algorithm, calculating the Z-score of the sequence value within the current window and comparing it with a threshold to identify anomalies. In practical implementation, during the third-layer evaluation, the feature difference between real-time data streams and historical normal data streams is compared. Historical normal data streams retrieve feature vectors of data streams without anomaly markers within a specified historical time period from the edge gateway's archive storage. These feature vectors include average throughput, latency variance, and packet size distribution. The feature difference is calculated by determining the Euclidean distance between the real-time and historical feature vectors, using the following formula:

[0077] ;

[0078] in: Indicates the degree of feature difference. The dimension of the feature vector. This represents the k-th feature value of the real-time data stream. This represents the k-th feature value of the historical normal data stream. In practice, the results of the three-layer evaluation are summarized to generate a comprehensive evaluation report. The comprehensive evaluation report is generated in a structured document format, including the identifier of each processing queue, the throughput and latency values ​​of the first-layer evaluation, the number of outliers in the second-layer evaluation, the feature difference value of the third-layer evaluation, and the timestamp. The report is stored in the log system of the edge gateway for subsequent analysis.

[0079] In some embodiments, referring to Table 1, the statistical integration hardware accelerator supports direct reading of network interface card register values ​​to improve sampling frequency. In some embodiments, abnormal fluctuation pattern detection employs wavelet transform to decompose the multi-resolution features of the time series to identify transient anomalies. Optionally, the retrieval of historical normal data streams can be based on a tag system, loading only historical data matching the current real-time data stream service type. Optionally, the generation of the comprehensive evaluation report can employ a template engine to automatically populate the evaluation results into a predefined format document. It is understood that the Z-score algorithm relies on the calculation of the sequence mean and standard deviation.

[0080] Table 1: Comprehensive Evaluation Report Form

[0081]

[0082] See Figure 4 This chart visually demonstrates the optimization effect of load balancing on each heavily loaded queue. Specifically, the chart compares the percentage of heavy load time before and after load balancing, using the processing queue as the dimension: for example, queue Q-001 had a heavy load percentage of 20% before load balancing, which decreased to 18% after processing; queue Q-005 had a heavy load percentage of 40% before load balancing, which was optimized to 25% after processing. The core support for this type of data is the edge gateway's real-time parsing and load scheduling mechanism for heterogeneous data: the edge gateway first classifies data packets into corresponding processing queues according to edge domain identifiers, and then dynamically schedules data packets from overloaded queues to idle queues by monitoring the queue load status. The quantitative results in this chart directly reflect the mitigation effect of the load balancing strategy on the heavy load state of the queues, providing a data basis for subsequent performance optimization based on multi-layer evaluation strategies.

[0083] Example 5: In specific implementation, the performance bottleneck areas identified in the comprehensive evaluation report are parsed. The parsing process reads the comprehensive evaluation report document from the edge gateway's log system. The comprehensive evaluation report document stores multi-level evaluation results in JSON format. Performance index data is extracted from the comprehensive evaluation report. The performance index data includes data throughput, processing latency, and anomaly detection results. The data throughput value is read from the "Data Throughput" field in the report, the processing latency value is read from the "Processing Latency" field, and the anomaly detection results are read from the "Number of Anomalies" and "Feature Difference" fields. The performance index data is compared with preset thresholds to identify edge domain identifiers below the thresholds. The preset thresholds are set according to the system service level agreement, for example, the data throughput threshold is set to 1000 packets / second, the processing latency threshold is set to 20 milliseconds, and the feature difference threshold is set to 0.5. The edge domain identifiers are obtained by mapping from the "Processing Queue Identifier" field in the report, and the mapping relationship is based on the correspondence table between processing queues and edge domains. Based on the edge domain identifiers, the physical device clusters and communication paths corresponding to the performance bottleneck areas are located. The physical device clusters are obtained by querying the edge domain management list, and the communication paths are determined by analyzing the switch ports and links associated with the edge domain identifiers in the network topology diagram. Generate a detailed analysis report on the performance bottleneck area, including the cause of the bottleneck and optimization suggestions. The cause of the bottleneck is inferred by correlating performance indicators with network monitoring data. For example, low throughput may be due to network congestion. Optimization suggestions include adjusting the data collection frequency or increasing link bandwidth. The detailed analysis report is stored in XML format and sent to the management platform.

[0084] In practical implementation, for the edge domain corresponding to the performance bottleneck area, the optimal configuration parameters of the data acquisition nodes are recalculated. This recalculation process is based on performance metric data and the device resource model. The device resource model defines the processing capacity and network bandwidth constraints of the data acquisition nodes, and the calculation of the optimal configuration parameters introduces a formula:

[0085] ;

[0086] in: This indicates the adjusted data acquisition frequency. This represents the actual data throughput in the current performance metrics data. This represents the baseline data throughput within the preset threshold. This indicates the initial data acquisition frequency. Based on the recalculated optimal configuration parameters, the device discovery protocol and data acquisition frequency of the data acquisition nodes are updated. This update operation is achieved by sending configuration commands to the data acquisition nodes, with the commands encapsulating the new parameters using the NETCONF protocol. The communication link priority between the data acquisition nodes and the edge gateway is synchronously adjusted. This communication link priority is achieved by modifying the Quality of Service (QoS) policy in the router, for example, elevating the traffic priority of the bottleneck edge domain to a higher level.

[0087] In some embodiments, performance metric data extraction employs a streaming processing engine to parse the comprehensive evaluation report stream in real time. In some embodiments, the location of the physical device cluster is integrated with a geographic information system to visualize the device locations. Optionally, bottleneck cause analysis can apply root cause analysis algorithms to automatically identify underlying fault points. Optionally, communication link priority adjustment can employ a centralized configuration strategy using a software-defined network controller. It is understood that the device resource model needs to be pre-calibrated to reflect actual hardware performance.

[0088] See Figure 5 In the overall optimization effect analysis of the industrial data acquisition system, the comprehensive differences before and after optimization were quantitatively presented through four core performance indicators: data throughput, processing latency, number of outliers, and system stability. Specifically, in terms of data throughput (relative value), the optimized indicator was significantly higher than the unoptimized one, reflecting the improvement in the system's data carrying capacity; in terms of processing latency (relative value), the optimized indicator was much lower than the unoptimized one, reflecting the improvement in data processing timeliness; in terms of the number of outliers (relative value), the optimized indicator was significantly reduced, indicating the enhanced system's anomaly detection and data quality control capabilities; and in terms of system stability (relative value), the optimized indicator was higher than the unoptimized one, confirming the improvement in system operational reliability. The changes in each indicator intuitively demonstrate the technical effectiveness of the edge computing-based industrial data acquisition method in resource scheduling, link optimization, and load balancing.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0090] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An industrial data acquisition method based on edge computing, characterized in that, The method includes: Obtain the configuration parameters of the target industrial scene, and divide the industrial scene into multiple edge domains based on the configuration parameters, including: Collect equipment layout information, communication network topology, and real-time service requirements from the industrial site as configuration parameters; Assign a unique domain identifier to each edge domain and establish an edge domain management list; Based on the physical topology of each edge domain, at least one data acquisition node is dynamically configured for each edge domain; Based on the characteristics of the real-time data stream reported by the data acquisition nodes, the communication link between the data acquisition nodes and the edge gateway is dynamically adjusted. Based on the adjusted communication link, real-time parsing and load balancing are performed on the heterogeneous data received by the edge gateway, including: The data packets received by the edge gateway are parsed according to the protocol, the edge domain identifier of the data source is identified, and the data packets are classified into different processing queues according to the edge domain identifier. Monitor the load status of each processing queue. When the load of a processing queue reaches its processing capacity limit, dynamically schedule some data packets of the queue to an idle processing queue, record the processing path of each data packet and update the data flow tracking table. The step of parsing the data packets received by the edge gateway to identify the edge domain identifier of the data source includes: Extract the protocol header field of the data packet and parse the protocol type, which includes Modbus protocol or OPCUA protocol; Match a predefined protocol template based on the protocol type, and obtain the storage location offset of the edge domain identifier from the protocol template; Based on the storage location offset, read the binary value of the edge domain identifier from the payload of the data packet; Convert the binary value to a string format and verify whether the string format conforms to the encoding rules of the edge domain identifier; If the verification passes, the edge domain identifier is used for packet classification; otherwise, the packet is marked as invalid and discarded. By deploying a multi-layered evaluation strategy on the edge gateway, performance evaluation and anomaly detection are performed on the data stream after load balancing. Based on the results of performance evaluation and anomaly detection, the configuration parameters and communication strategies of the data acquisition nodes are dynamically optimized.

2. The industrial data acquisition method based on edge computing according to claim 1, characterized in that, The process of obtaining configuration parameters for the target industrial scenario and dividing the industrial site into multiple edge domains based on these parameters includes: Determine the physically adjacent device clusters based on the device layout information; Based on the communication network topology, each device cluster and its associated communication network are divided into an independent edge domain.

3. The industrial data acquisition method based on edge computing according to claim 2, characterized in that, The physical topology based on each edge domain, dynamically configuring at least one data acquisition node for each edge domain, includes: Parse the physical topology of each edge domain in the edge domain management list; Calculate the number of data acquisition nodes required for each edge domain based on the device distribution density in the physical topology; Configure the device discovery protocol and data acquisition frequency for each data acquisition node, and establish a mapping table between data acquisition nodes and the edge domain management list; The step of calculating the number of data acquisition nodes required for each edge domain based on the device distribution density in the physical topology includes: obtaining the physical location coordinates of all industrial devices in the edge domain, calculating the Euclidean distance between devices, calculating the device distribution density based on the Euclidean distance (the device distribution density is the number of devices per unit area), querying a preset density-node number mapping table to obtain the basic number of nodes based on the device distribution density, adjusting the basic number of nodes in conjunction with the historical data traffic peak of the edge domain, obtaining the final number of data acquisition nodes, and storing the final number of data acquisition nodes in the edge domain management list.

4. The industrial data acquisition method based on edge computing according to claim 3, characterized in that, The step of dynamically adjusting the communication link between the data acquisition node and the edge gateway based on the real-time data stream characteristics reported by the data acquisition node includes: The system monitors the timestamp, data volume, and transmission rate of data packets reported by the data acquisition nodes. When the data volume exceeds a preset threshold, the link quality assessment process is initiated. Based on the link quality assessment results, the optimal communication path is selected to establish a backup link between the data acquisition node and the edge gateway. After the backup link is successfully established, the real-time data stream is split and transmitted simultaneously on the main link and the backup link.

5. The industrial data acquisition method based on edge computing according to claim 4, characterized in that, The multi-layered evaluation strategy deployed on the edge gateway performs performance evaluation and anomaly detection on the data stream after load balancing, including: In the first-level evaluation, the data throughput and processing latency of each processing queue are statistically analyzed; In the second-level evaluation, the time-series characteristics of the data packets are analyzed to detect abnormal fluctuation patterns; In the third-level evaluation, the characteristic differences between real-time data streams and historical normal data streams are compared; The results of the three-tiered assessment are summarized to generate a comprehensive assessment report.

6. The industrial data acquisition method based on edge computing according to claim 5, characterized in that, The process of dynamically optimizing the configuration parameters and communication strategies of data acquisition nodes based on performance evaluation and anomaly detection results includes: Analyze the performance bottleneck areas identified in the comprehensive evaluation report; For the edge domain corresponding to the performance bottleneck area, recalculate the optimal configuration parameters of the data acquisition nodes; Update the device discovery protocol and data acquisition frequency of the data acquisition nodes based on the recalculated optimal configuration parameters; Synchronously adjust the priority of the communication link between the data acquisition node and the edge gateway.

7. The industrial data acquisition method based on edge computing according to claim 6, characterized in that, The performance bottleneck areas identified in the comprehensive evaluation report include: Performance metrics data are extracted from the comprehensive evaluation report, including data throughput, processing latency, and anomaly detection results. Compare performance metrics data with preset thresholds to identify edge domain identifiers that are below the threshold; Based on edge domain identifiers, locate the physical device clusters and communication paths corresponding to performance bottleneck areas; Generate a detailed analysis report of the performance bottleneck area, including the cause of the bottleneck and optimization suggestions.

8. An industrial data acquisition system based on edge computing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the industrial data acquisition method based on edge computing as described in any one of claims 1 to 7.

Images