Real-time state monitoring method and device for high-frequency time sequence data and medium

Abstract

The embodiment of the invention discloses a real-time state monitoring method and device for high-frequency time sequence data and a medium, and relates to the technical field of the Internet of Things, and the method comprises the steps: obtaining a state monitoring task of target production equipment, carrying out the initialization configuration, building a multi-topic subscription channel corresponding to the state monitoring task, and carrying out the multi-topic subscription through the multi-topic subscription channel, acquiring multi-channel high-frequency sensing data of a field side; multi-channel high-frequency sensing data is stored to a time sequence storage node preset on the edge side by utilizing communication among multiple processes, and the time sequence storage node comprises a super table structure storage library constructed based on TD engine; and performing multi-thread analysis on the multi-channel high-frequency sensing data in the time sequence storage node according to an AI model node preset on the edge side, and determining real-time state data of the target production equipment. Through collaborative design of a multi-topic subscription channel, edge side time sequence storage optimization and multi-thread AI analysis, the high-frequency data real-time processing capability and the analysis efficiency are remarkably improved.

Description

Technical Field

[0001] This specification relates to the field of Internet of Things technology, and in particular to a real-time status monitoring method, device and medium for high-frequency time-series data. Background Art

[0002] With the rapid development of industrial Internet of Things technology, the state monitoring of industrial equipment is gradually transforming from traditional regular maintenance to real-time and intelligent predictive maintenance. Real-time monitoring of equipment status and timely fault warning are the prerequisites for ensuring production efficiency, equipment and personnel safety, and reducing operation and maintenance costs. High-frequency time-series data (such as vibration, temperature, current, etc.) have become the key data source for achieving accurate status monitoring because they can capture the instantaneous dynamic characteristics during equipment operation. The acquisition and analysis of high-frequency time-series data have become the core requirements of status monitoring.

[0003] Traditional status monitoring systems mostly rely on low-frequency data acquisition (usually below 1000Hz), which cannot effectively capture the high-frequency dynamic characteristics of equipment, such as microsecond-level vibration signals, resulting in lagged anomaly detection. Moreover, the transmission of high-frequency data poses extremely high requirements on network bandwidth and real-time performance. Existing centralized cloud processing solutions often have difficulty achieving real-time response due to data transmission delays. In addition, in industrial scenarios, the status monitoring process needs to process multi-dimensional signals simultaneously. For example, the monitoring of the tool wear status of a numerically controlled machine tool needs to process multi-dimensional signals such as multi-directional vibration and cutting force simultaneously. Using single-threaded serial processing, it is impossible to efficiently process the parallel analysis requirements of multi-channel high-frequency data streams. In summary, in the traditional industrial equipment status monitoring process, there is a problem of insufficient real-time processing ability and poor analysis efficiency of high-frequency data, as it mostly relies on low-frequency data acquisition and ignores the high-frequency data characteristics. Summary of the Invention

[0004] One or more embodiments of this specification provide a real-time status monitoring method, device and medium for high-frequency time-series data to solve the following technical problems: In the traditional industrial equipment status monitoring process, there is a problem of insufficient real-time processing ability and poor analysis efficiency of high-frequency data, as it mostly relies on low-frequency data acquisition and ignores the high-frequency data characteristics.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of this specification provide a real-time status monitoring method for high-frequency timing data. The method includes: obtaining a status monitoring task of a target production device for initialization configuration, establishing a multi-topic subscription channel corresponding to the status monitoring task, and obtaining multi-channel high-frequency sensing data on the field side corresponding to the target production device through the multi-topic subscription channel; using inter-process communication to store the multi-channel high-frequency sensing data in a timing storage node preset on the edge side, where the timing storage node includes a supertable structure repository built based on TDengine; performing multi-threaded analysis on the multi-channel high-frequency sensing data in the timing storage node according to an AI model node preset on the edge side to determine the real-time status data of the target production device.

[0007] One or more embodiments of this specification provide a real-time status monitoring device for high-frequency timing data, including:

[0008] At least one processor; and,

[0009] A memory communicatively connected to the at least one processor; wherein,

[0010] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above method.

[0011] A non-volatile computer storage medium provided by one or more embodiments of this specification stores computer-executable instructions, and the computer-executable instructions are set to: execute the above method.

[0012] One or more of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects: Through the technical solution of the embodiments of this specification, through a multi-topic subscription channel, it supports ≥High-frequency data acquisition and real-time transmission at 10 kHz can effectively capture transient characteristics. Different from traditional condition monitoring systems that mostly rely on low-frequency data acquisition, it can effectively capture the instantaneous dynamic characteristics during equipment operation, solve the problem that traditional solutions cannot effectively capture high-frequency dynamic characteristics of equipment, and avoid lag in anomaly detection. The multi-channel high-frequency sensing data is stored in a time-series storage node preset on the edge side through inter-process communication, which changes the existing centralized cloud processing solution. In traditional centralized cloud processing, the transmission of high-frequency data poses extremely high requirements for network bandwidth and real-time performance, and real-time response is often difficult to achieve due to data transmission delay. In the embodiments of this specification, the data is stored on the edge side, reducing the data transmission volume and distance, decreasing the dependence on network bandwidth, and also reducing transmission delay, enabling real-time processing of high-frequency data and meeting the real-time requirements for real-time monitoring of equipment status. The time-series storage node uses a supertable structure repository built based on TDengine. As a database specifically designed for time-series data, TDengine can efficiently store and manage high-frequency time-series data, facilitating the storage and management of data at similar data collection points, reducing data redundancy, and improving data storage and query efficiency, making it more convenient and efficient to process a large amount of high-frequency time-series data and providing a good data storage foundation for subsequent data processing and analysis. In an industrial scenario, condition monitoring needs to process multi-dimensional signals simultaneously. Traditional single-threaded serial processing cannot efficiently handle the parallel analysis requirements of multi-channel high-frequency data streams. In the embodiments of this specification, through multi-threaded analysis, multi-channel high-frequency data can be processed in parallel, making full use of system resources, improving data processing efficiency, and quickly and accurately determining the real-time status data of target production equipment. Through a series of operations such as acquiring high-frequency data, processing and storing data on the edge side, and multi-threaded analysis, the equipment status can be monitored in real time and potential faults can be detected in a timely manner, realizing the transformation from traditional regular maintenance to real-time and intelligent predictive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0014] Figure 1 It is a flowchart showing a method for real-time status monitoring of high-frequency time-series data provided by an embodiment of this specification;

[0015] Figure 2 It is a schematic diagram of the system architecture of a system for real-time status monitoring of high-frequency time-series data provided by an embodiment of this specification;

[0016] Figure 3 It is a schematic diagram of a topic publishing and subscribing process provided by an embodiment of this specification;

[0017] Figure 4 It is a schematic diagram of a consumption process provided by an embodiment of this specification;

[0018] Figure 5 It is a schematic diagram of a timing storage process of a timing storage node provided by an embodiment of this specification;

[0019] Figure 6 It is a schematic diagram of a wear monitoring process provided by an embodiment of this specification;

[0020] Figure 7 It is a test schematic diagram of data transmission delay and packet loss rate indicators provided by an embodiment of this specification;

[0021] Figure 8 It is a test result schematic diagram of the operation time of each function on the edge side provided by an embodiment of this specification;

[0022] Figure 9 It is a structural schematic diagram of a real-time status monitoring device for high-frequency timing data provided by an embodiment of this specification. Specific embodiments

[0023] In order to enable those skilled in the art to better understand the technical solutions in this specification, the following will clearly and completely describe the technical solutions in the embodiments of this specification with reference to the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0024] An embodiment of this specification provides a real-time status monitoring method for high-frequency timing data. It should be noted that the execution entity in the embodiment of this specification can be a server, or any device with data processing capabilities, or a real-time status monitoring system. Figure 1 It is a schematic diagram of a real-time status monitoring method for high-frequency timing data provided by an embodiment of this specification, as Figure 1 shown, mainly including the following steps:

[0025] Step S101, obtain the status monitoring task of the target production equipment for initialization configuration, establish a multi-topic subscription channel corresponding to the status monitoring task, and obtain multi-channel high-frequency sensing data on the field side corresponding to the target production equipment through the multi-topic subscription channel.

[0026] In one embodiment of this specification, Figure 2 is a schematic diagram of the system architecture of a real-time state monitoring system for high-frequency time-series data provided by the embodiments of this specification. As Figure 2 shown, in order to achieve real-time and accurate data communication from the field layer to the edge side, the real-time communication between the field layer and the edge side is built based on Fast DDS. By deploying Fast DDS components on the field-side and edge-side systems, and using this component as the main component, field-edge interaction nodes and edge-field interaction nodes are formed, and a local area network is built using a router, so that the interaction nodes of the field layer and the interaction nodes of the edge side are in the same DDS data domain, and communication between any nodes independent of the device can be achieved within this data domain. Considering the continuous data stream characteristics of high-frequency time-series data, the interface method of topics is used to achieve communication between nodes.

[0027] The field layer consists of physical production equipment nodes, sensors and data acquisition devices, field controllers, and field-edge interaction nodes. The data acquisition module collects sensor measurement point data, such as vibration, temperature, current, etc., at a ≥ frequency of 10 kHz through industrial buses or DAQ devices. The core function of real-time state monitoring is deployed on the edge side. The overall edge-side nodes consist of an edge management node, a visualization interaction node, an AI node, a time-series storage node, and an edge-field interaction node. Among them, the edge management node is overall responsible for the deployment management of other edge nodes, such as dynamically allocating computing resources (CPU / memory), achieving load balancing between the AI node and the storage node, and maintaining the communication link between the edge side and the field layer to ensure data transmission reliability. The AI node deploys a data-driven state monitoring model to provide algorithm support for the real-time state monitoring task in the visualization interaction node. The state monitoring model here can be a lightweight machine learning model, such as pre-trained CNN, LSTM and other models, supporting online model updates. On the one hand, the edge-field interaction node receives dynamic high-frequency real-time data from the field layer and distributes the data on the edge side according to data governance requirements. On the other hand, it timely sends control commands, alarm notifications and other information on the edge side to the field layer control devices. When the visualization interaction node is running, it pulls high-frequency time-series data from the time-series storage node, performs real-time data processing and analysis, completes tasks such as signal monitoring and intelligent state monitoring near the field layer, and timely sends and stores the corresponding results. Based on the scheduling of the edge management layer, the cloud side interacts with the edge storage node and the AI node to achieve cloud offline big data processing storage and model training.

[0028] In as Figure 2In the architecture shown, real-time and reliable data connection and interaction ensure the fast and accurate response of the status monitoring task. To ensure the fast and accurate response of this task, the system requires real-time data connection, transmission, and control at the field layer, real-time communication between the field end and the edge side, and data connection and interaction between key functions such as data processing, storage, analysis, and visualization within the edge side. Taking the implementation process of the core status monitoring task as an example, data connection and interaction within the key nodes during the response process are achieved through tool components such as message middleware and rule engines.

[0029] The corresponding data connection and interaction within the field layer mainly consist of real-time data acquisition and its data connection and interaction with the field-edge interaction node, data connection and interaction between the field production equipment and the field control equipment, and data connection and interaction between the field control equipment and the field-edge interaction node. Among them, the data acquisition module performs high-frequency real-time data acquisition (acquisition frequency 9000 - 10000Hz) by connecting to sensors deployed on the field production equipment, and the field control equipment realizes the acquisition of operation status data of the field production equipment and the issuance of control commands. The data acquisition module and the field control equipment can transmit sensor data and equipment operation status data to the field-edge interaction node through process communication.

[0030] To ensure the real-time, sequential, and integrity of the high-frequency sequential data transmission from the field layer to the edge side, communication is established between the field layer and the edge side with a distributed real-time message middleware, namely the Fast Data Distribution Service (Fast DDS). Nodes on both sides are based on the publish-subscribe method and topic mechanism, and real-time data connection is achieved through LAN TCP. During the edge-field transmission process, corresponding Quality of Service (QoS) policies are formulated to ensure the quality of the transmitted data. The edge-field interaction node is mainly composed of the edge-side middleware. After receiving the high-frequency sequential data, it will perform preprocessing operations such as data cleaning, and write the data into the sequential storage node in real-time in an efficient SQL write mode with multiple processes.

[0031] Through the service loading method, tasks such as signal monitoring and status monitoring are deployed. On the one hand, data is sequentially pulled from the sequential storage node through the consumer subscription mechanism to supply relevant services for data processing. On the other hand, the corresponding data-driven intelligent monitoring model in the AI node performs data analysis, and the corresponding analysis results are visualized and stored in real-time.

[0032] Obtain the status monitoring task of the target production equipment for initialization configuration, and establish a multi-topic subscription channel corresponding to the status monitoring task, specifically including: loading a preset AI analysis model according to the status monitoring task; parsing the input requirements of the AI analysis model to determine the sensor channels corresponding to the real-time data types required by the AI analysis model, and creating corresponding topics based on the sensor channels; creating consumers and subscribers for each topic through a preset communication middleware, and establishing a consumer group subscription relationship to determine the multi-topic subscription channel corresponding to the status monitoring task.

[0033] In one embodiment of this specification, based on the status monitoring task types of the target device, such as tool wear monitoring and bearing fault diagnosis, load a preset lightweight AI analysis model from the edge-side model repository, such as a CNN-LSTM fusion model and a random forest classifier. By parsing the model configuration file, extract the input data requirement parameters. Here, the input requirements include sensor type, sampling frequency, channel mapping, and data format, etc. The sensor type is used to determine the required physical quantities (vibration, temperature, current, etc.) and their range; the sampling frequency is used to match the high-frequency signal acquisition requirements ( ≥ 10000Hz); the channel mapping is used to establish the correspondence between the device physical interface and the logical channel; the data format is used to define the Protobuf message structure, including fields such as timestamp, device ID, and data value. It should be noted that the sensor channel mapping is the binding rule between the physical sensor and the logical communication channel, and its core is to establish an accurate correspondence between the physical sensors on the device and the data transmission channel (Topic) through multi-dimensional attribute association.

[0034] According to the parsing result, create corresponding topics based on the sensor channel mapping relationship. The topic naming rule is / sensors / {device ID}_{sensor type}, and define the Protobuf format data interface; dynamically associate the model with the data type. For example, the tool wear model needs to subscribe to the vibration (10000Hz) and three-axis cutting force (5000Hz) channels, and the bearing fault model needs to subscribe to the vibration (20000Hz) and temperature (1Hz) channels. Create independent consumers (DataReader) and subscribers (DataWriter) for each topic through the FastDDS middleware, and configure the QoS policy to include a reliability mode of RELIABLE, a historical data cache depth of ≥ 1000 items, and the transport protocol uses TCP / IP; establish a consumer group subscription relationship, bind the device physical address to the logical topic, and complete the initialization of the multi-channel communication link.

[0035] Through the above technical solution, the pre-set AI analysis model is loaded according to the status monitoring task, and its input requirements are parsed to determine the sensor channels corresponding to the real-time data types required by the AI analysis model, which can ensure that the acquired data is the data required by the AI analysis model, making the subsequent analysis more accurate and targeted, avoiding the waste of resources caused by acquiring irrelevant data, improving the efficiency and accuracy of data processing, and thus enhancing the accuracy of the status monitoring of the target production equipment; corresponding topics are created based on the sensor channels, and different types of sensor data correspond to different topics, facilitating subsequent data processing and analysis; through the pre-set communication middleware, consumers and subscribers are created for each topic, and a consumer group subscription relationship is established, thereby determining the multi-topic subscription channels. The use of the communication middleware ensures the stability and efficiency of data transmission, and the creation of consumers and subscribers and the establishment of the consumer group subscription relationship enable the data to be received and processed orderly, ensuring that the multi-channel high-frequency sensing data can be acquired in a timely and accurate manner, providing a reliable data basis for the subsequent real-time status monitoring of the target production equipment.

[0036] Through the multi-topic subscription channel, multi-channel high-frequency sensing data corresponding to the target production equipment is acquired, specifically including: determining the publisher and the corresponding on-site callback function pre-created in the on-site-edge interaction node on the on-site side, and the subscriber and the corresponding edge-side callback function pre-created in the edge-site interaction node on the edge side; receiving the real-time high-frequency sensing data corresponding to the target production equipment through the on-site callback function of the on-site-edge interaction node, performing data cleaning on the real-time high-frequency sensing data, and then publishing it to the target topic; subscribing to the data stream of the target topic through the subscriber and the corresponding edge-side callback function of the edge-site interaction node to acquire the multi-channel high-frequency sensing data corresponding to the target production equipment.

[0037] In one embodiment of the present specification, Figure 3 is a schematic flow chart of topic publishing and subscribing provided by the embodiment of the present specification. As Figure 3 shown, a data publisher and the corresponding callback function are pre-created in the on-site-edge interaction node on the on-site side, and the publisher QoS policy is configured; a subscriber and the corresponding callback function matching the publisher are pre-created in the edge-site interaction node on the edge side, and the subscriber QoS policy is configured. The on-site-edge interaction node creates a publisher and the corresponding callback function, and the edge-site interaction node creates a subscriber and the corresponding callback function. After the publisher and the subscriber define the same data interface type, a connection is established based on the "Force_X" topic.

[0038] Implement the reception, data cleaning, and real-time publishing of high-frequency time-series data at the field layer within the callback function of the field-edge interaction node. Receive the sensing data of the target production equipment in real time through the field-side callback function. After performing anti-aliasing filtering and dimension normalization on the original data, publish it to the corresponding topic. The callback function of the edge-interaction node implements preprocessing tasks such as data subscription, data segmentation, and conversion for high-frequency time-series data, as well as writing to the inter-process communication queue. That is, the edge-side subscriber pulls the topic data stream and performs data segmentation and format conversion. Write the processed data to the time-series database in batches through the inter-process communication queue. It should be noted that during the data transmission process between the field layer and the edge side, define the policy parameters of Depth and History to configure QoS to ensure communication quality. The function implementation of other nodes is the same as above. Thus, multiple interaction nodes are developed and registered at the field end and the edge side. In addition, use the startup management tool to uniformly manage the operation of multiple interaction node processes, ensure the normal startup, stop, and monitoring of each node, promptly discover and solve problems that occur during the operation of the nodes, and ensure the continuity and accuracy of the status monitoring of the target production equipment

[0039] Through the above technical solution, in the field-side callback function of the field-edge interaction node, clean the real-time high-frequency sensing data. During the data transmission process between the field layer and the edge side, configure QoS (Quality of Service) by defining the policy parameters of Depth and History to ensure the reliability, integrity, and timeliness of data transmission, reduce problems such as data loss and delay, and ensure that the multi-channel high-frequency sensing data of the target production equipment is accurately transmitted from the field side to the edge side. Develop and register multiple interaction nodes at the field end and the edge side. Each node can independently complete the data processing and transmission tasks. The distributed architecture can support the status monitoring of a large number of target production equipment. By increasing the number of interaction nodes, it can easily handle the access of more devices. The callback function of the edge-field interaction node implements preprocessing tasks such as data subscription, data segmentation, and conversion for high-frequency time-series data. Data segmentation splits the continuous data stream into appropriate blocks for subsequent processing. Format conversion makes the data meet the storage requirements of the time-series database, improving the efficiency and quality of data storage. Write the processed data to the time-series database in batches through the inter-process communication queue, reducing the number of database writes, lowering the database load, and improving the performance of data storage.

[0040] Through the subscribers of the edge-field interaction node and the corresponding edge-side callback functions, subscribe to the data stream of the target topic to obtain multi-channel high-frequency sensing data corresponding to the target production device, specifically including: triggering a data pulling operation through the edge-side callback function of the subscriber, and batch pulling data blocks from the time-series storage node; automatically submitting the database consumption position in the edge-side callback function to persist the data consumption progress to the distributed coordination service; determining whether the pulled data block is empty, and if the data block is not empty, performing deserialization processing and supertable format conversion on the data block to obtain multi-channel high-frequency sensing data corresponding to the target production device.

[0041] Figure 4 A consumption process schematic diagram provided by an embodiment of this specification, as Figure 4 shown, after creating a consumer group and subscribing to the target topic, the load balancing strategy of the consumer group can be configured as the polling mode. Trigger a data pulling operation through the edge-side callback function of the subscriber, and batch pull data blocks from the time-series storage node. Here, each data block can be set to contain 1000 - 5000 time-series data points. Automatically submit the database consumption position in the edge-side callback function to ensure that the data consumption progress is persisted to the distributed coordination service. Determine whether the pulled data block is empty. If the data block is empty, pull again. The number of empty pulls can be accumulated, and when the number exceeds the threshold, trigger a health check operation. If the data block is not empty, perform multi-level data parsing, such as Protobuf deserialization and TDengine supertable format conversion. Push the parsed data to the multi-threaded processing queue, and the AI model node concurrently performs feature extraction and status analysis. Monitor the data subscription status. When receiving a termination instruction, close the corresponding process, call the unsubscribe interface of the subscriber, release the topic binding resources, close the consumer group process, and empty the cache queue, and record the final consumption position to the log file.

[0042] In an embodiment of this specification, create consumers and corresponding topics for the required service functions. Among them, a consumer group is set in the creation configuration of the consumer to facilitate sharing the consumption progress of the same consumer group. And according to the task requirements of online monitoring and signal monitoring, set the initial position of the consumer subscription to "latest", that is, start subscribing from the latest data, and set the automatic submission of the consumption position to facilitate positioning the consumption position of the data in the database. By creating and subscribing to the corresponding topics, consumers can pull data from the database according to the data query conditions defined by the topics. Thus, during the operation of the service function, consumers can continuously pull data from the database and automatically feedback the consumption position to the database for the data consumption of the service function. When the service function ends, the consumer needs to cancel the topic subscription and close the corresponding process to release the cache.

[0043] Through the above technical solutions, the edge-side callback function of the subscriber pulls data blocks in batches from the time-series storage node. The batch operation reduces the number of interactions with the database, reduces the communication overhead, and greatly improves the efficiency of data pulling. It is especially suitable for the processing scenario of high-frequency time-series data and can quickly obtain a large amount of data for subsequent analysis. In the edge-side callback function, the database consumption position is automatically submitted, and the data consumption progress is persisted to the distributed coordination service, ensuring that even in the event of a failure or restart, the position of data consumption can be accurately recorded, avoiding duplicate processing or omission of data, and ensuring the integrity and continuity of data processing. A consumer group is created and a load balancing strategy in polling mode is configured. The polling mode can evenly distribute the data processing tasks among the consumers, avoiding the situation where some consumers are overloaded while others are idle. It is judged whether the pulled data block is empty. If it is empty, it is pulled again, which can timely detect possible problems in the data pulling process, such as data storage node failure or data loss, etc., ensuring that valid data can be continuously and stably obtained. The data subscription status is monitored. When a termination instruction is received, the system can close the corresponding process according to a predetermined process, call the unsubscribe interface, release the topic binding resources, close the consumer group process and empty the cache queue, and at the same time record the final consumption position to the log file, which can avoid resource leakage and data loss and ensure the stability and reliability of the system in normal shutdown or abnormal situations.

[0044] Step S102: Use inter-process communication to store multi-channel high-frequency sensing data in a time-series storage node preset on the edge side.

[0045] Among them, the time-series storage node includes a supertable structure repository built based on TDengine;

[0046] Using inter-process communication to store the multi-channel high-frequency sensing data in a time-series storage node preset on the edge side specifically includes: creating multiple processes and corresponding process queues in advance, where the processes include an edge middleware receiving process, a database writing process, and a monitoring process; starting the services of the multiple processes and process queues, and establishing a connection relationship between the edge middleware and the time-series database of the time-series storage node; through the edge middleware receiving process, encapsulate the multi-channel high-frequency sensing data in time series, generate high-frequency time-series data and write it into the process queue; the database writing process reads the high-frequency time-series data from the process queue and fills the data fields according to the predefined supertable structure; through the monitoring process, monitor the cache data volume in real time to obtain the current cache data volume, and judge whether the current cache data volume reaches a preset single-batch writing threshold. If it is satisfied, generate a batch SQL insert statement to store it in the time-series database.

[0047] In one embodiment of this specification, the timing storage nodes on the edge side are built relying on the TDengine OSS time-series database, and the real-time storage of high-frequency time-series data is realized by writing multi-process SQL statements within the edge-field interaction nodes. After deploying TDEngine and the corresponding Python connector on the edge side, relevant data modeling and efficient database writing are carried out. For data modeling, database creation and the creation of supertables and tables are performed. The database is created under the conditions of comprehensively considering factors such as the data collection frequency at the field end, the data file retention time, and the later data subscription and consumption. The timestamp accuracy of the database is defined at the nanosecond level. To facilitate the data storage management of similar data collection points, the supertable creation function of TDEngine is used. Considering that the data types and storage requirements of high-frequency time-series data are similar, a supertable is created and then sub-tables of high-frequency time-series data for each channel are created using this supertable as a template.

[0048] After completing the modeling, the efficient storage of data is realized by means of inter-process communication. Figure 5 It is a schematic diagram of the timing storage process of a timing storage node provided by an embodiment of this specification, as Figure 5 shown. On the edge side, a database writing process and a monitoring process are created through a multi-process method. An edge middleware receiving process, a database writing process, and a monitoring process are created, and the inter-process communication queue is initialized. The above processes and queue services are started. The receiving process, the database writing process, and the monitoring process of the edge-side subscriber run synchronously, and the inter-process communication is carried out using the message queue method to establish a connection between the edge middleware and the time-series database. During the data storage process, the edge-field interaction node first links to the time-series database. The callback function of the edge-side subscriber puts the received and preprocessed data into the message queue in a timely manner. The database writing process reads the data from the message queue. To improve the writing efficiency of the database, the data writing process adopts the batch writing method. After the amount of data cached in the process reaches the set batch writing threshold, the corresponding program generates SQL statements and writes them into the database. The specific implementation process is as follows: The multi-channel high-frequency sensing data is encapsulated in time series by the edge middleware receiving process and written into the inter-process queue. The database writing process reads the time-series data from the queue and fills the data fields according to the predefined supertable structure; it is judged whether the current cached data volume reaches the single-batch writing threshold. If it is satisfied, a batch SQL insertion statement is generated and submitted to the time-series database; the external termination instruction is monitored. If a stop writing signal is received, the database connection is disconnected and the queue cache is cleared; the database writing process, the edge middleware receiving process, and the monitoring process are closed in sequence to release system resources.

[0049] Through the above technical solutions, by creating an edge middleware receiving process, a database writing process, and a monitoring process, parallel operations of data reception, processing, and storage are achieved. The computing resources of the multi-core processor can be fully utilized, avoiding potential performance bottlenecks in single-process processing, greatly improving the efficiency of data processing and storage, and ensuring that high-frequency time-series data can be processed and stored in a timely manner. The message queue is used for inter-process communication, ensuring the orderly transfer of data between different processes. The edge middleware receiving process puts the encapsulated data into the queue, and the database writing process reads the data from the queue. This decoupled method enables each process to run independently, improving the stability and scalability of the system. Even if a certain process experiences a temporary failure, it will not affect the normal operation of other processes, and the data can wait in the queue for processing. Relying on the TDengine OSS time-series database, which is specifically optimized for time-series data, can efficiently handle the storage and query of high-frequency time-series data. Its timestamp accuracy is defined at the nanosecond level, meeting the requirements of high-precision data storage and ensuring the accuracy and integrity of the data. Using the supertable creation function of TDengine, sub-tables of high-frequency time-series data for each channel are created based on the supertable, facilitating the data storage management of similar data collection points, reducing data redundancy, and improving data storage efficiency. The supertable structure makes data query and analysis more convenient, and quick filtering and aggregation operations can be performed according to the attributes of the supertable. The database writing process adopts the batch writing method. When the amount of cached data reaches the set batch writing threshold, batch SQL insertion statements are generated and submitted to the database, reducing the number of interactions with the database, reducing the load on the database, and improving the writing performance. Compared with writing one by one, batch writing can significantly shorten the data storage time, especially suitable for high-frequency data scenarios. The monitoring process monitors the amount of cached data in real time to ensure that data can be written to the database in a timely manner. When the amount of cached data reaches the threshold, a batch writing operation is triggered in a timely manner to avoid data accumulation in the cache causing memory overflow. The monitoring process can also monitor external termination instructions and, when receiving a stop writing signal, can safely disconnect the database connection, empty the queue cache, and close relevant processes to release resources, ensuring data security and system stability.

[0050] Step S103: According to the AI model nodes pre-set on the edge side, perform multi-threaded analysis on the multi-channel high-frequency sensing data in the time-series storage node to determine the real-time status data of the target production equipment.

[0051] Based on the AI model nodes pre - set on the edge side, perform multi - thread analysis on the multi - channel high - frequency sensing data in this time - series storage node to determine the real - time status data of the target production device. Specifically, in this AI model node, load the corresponding AI analysis model according to this status monitoring task; parse the AI analysis model to create multiple analysis sub - threads corresponding to the AI analysis model. Among them, the analysis sub - threads include a database pulling thread, a feature processing thread, and a feature recognition thread; through these multiple analysis sub - threads, perform multi - thread analysis on the multi - channel high - frequency sensing data to determine the real - time status data of the target production device. Through these multiple analysis sub - threads, perform multi - thread analysis on the multi - channel high - frequency sensing data to determine the real - time status data of the target production device. Specifically, through the database pulling thread, pull multi - channel high - frequency sensing data in this time - series storage node; through the feature processing thread, extract and combine features of the multi - channel high - frequency sensing data to send to the feature recognition thread, and based on the feature recognition thread, generate the real - time status recognition result of the target production device.

[0052] In an embodiment of this specification, in the edge - side AI model node, according to the status monitoring tasks of the target device, such as tool wear monitoring and bearing fault diagnosis, call lightweight AI models matching the tasks, such as CNN - LSTM fusion models and random forest classifiers, from the pre - set model library, parse the model configuration file to determine the input data requirements, and based on the model requirements, create three types of independently running sub - threads. The database pulling thread is used to batch - pull multi - channel high - frequency sensing data from the time - series storage node. The feature processing thread performs joint time - frequency domain feature extraction on the original data, including time - domain features, frequency - domain features, and cross - channel features. The time - domain features can be kurtosis coefficient, waveform factor, and root - mean - square value; the frequency - domain features can be band - energy entropy based on wavelet packet transform; the cross - channel features can be the cross - correlation peak and lag time of multi - sensor signals. The feature recognition thread loads the AI model to perform real - time inference on the feature vector and outputs the status determination result, such as wear rate percentage and fault mode code.

[0053] Achieve efficient analysis through the decoupling and asynchronous communication mechanism between multi - threads. The database pulling thread extracts data blocks from the time - series storage node according to the time window through the SQL interface, and after Protobuf deserialization, stores them in the shared memory queue. The data pulling period ≤ is 50ms to ensure real - time performance. The feature processing thread reads data blocks from the shared queue, performs data segmentation operations to divide data subsets according to device ID and sensor type, and performs parallel computing to synchronously extract features from multi - channel data to generate a standardized feature vector; combines the time - domain, frequency - domain, and cross - channel features into a multi - dimensional feature matrix according to the preset weights and pushes it to the feature recognition thread through the ZeroMQ message queue.

[0054] The feature recognition thread receives the feature matrix, inputs the feature matrix into the AI model, and outputs the original determination result, such as the abnormal probability of the vibration signal spectrum; performs weighted voting on the output results of multiple models (e.g., the weight of the CNN-LSTM model accounts for 70%, and the random forest model accounts for 30%) to generate a comprehensive health index; when the health index exceeds the dynamic threshold (e.g., wear rate > 90%) or the continuous decline rate is abnormal, triggers an alarm instruction and sends it to the on-site control device. The status recognition result is written into the time-series storage node through the edge-field interaction node, and at the same time is pushed to the visualization interaction node to be displayed in real time in the form of a waveform diagram and a heat map. The edge management node monitors the thread resource occupancy rate in real time. When the load of the feature processing thread > 80%, automatically expands new thread instances, preferentially processes high-frequency vibration signal analysis tasks, and ensures the processing priority of key data.

[0055] After pulling multi-channel high-frequency sensing data from the time-series storage node through the database pulling thread, the method further includes: determining whether the target production equipment is in an operating state; if so, caching the multi-channel high-frequency sensing data in the operating state, and determining the cached data volume; when the cached data volume is not less than the preset quantity threshold, sending it to the feature processing thread.

[0056] In an embodiment of the present specification, after the database pulling thread pulls multi-channel high-frequency sensing data from the time-series storage node (based on the TDengine super table), parses the device operation flag bit in the data block (such as the PLC status register value). If the flag bit is "running" (RUN = 1), it is determined that the device is in an operating state. If the flag bit is missing, perform secondary verification. If the current sensor data is continuously ≥ the no-load current threshold (such as 5A), it is determined to be in an operating state, or, if the root mean square value (RMS) of the vibration signal > the shutdown baseline value (such as 0.1m / s 2 ), it is determined to be in an operating state. The determination result is written into the distributed state cache through the edge management node and pushed to the visualization interaction node to display the device operation status indicator in real time. If the device is in an operating state, store the pulled high-frequency sensing data in the circular buffer, divide independent buffer areas according to the device ID and sensor type to avoid data aliasing. Attach timestamp, data volume, and device ID meta-information to each data block to generate a cache index table. The preset quantity threshold here is dynamically adjusted according to the device type, for example, the threshold > 22000 pieces. When the cached data volume ≥When the threshold is reached, extract a complete data block from the circular buffer, encapsulate it into a Protobuf format message, push the data to the feature processing thread through the message queue, reset the buffer pointer, and release the memory of the processed data. Through the device status perception and intelligent caching mechanism, it solves the defects of resource waste, timing disorder, and insufficient real-time performance in traditional high-frequency data processing, and meets the core requirements of high efficiency, reliability, and resource controllability in industrial equipment monitoring scenarios.

[0057] Figure 6 This is a schematic diagram of the process of wear monitoring provided by the embodiments of this specification. As Figure 6 shown, taking the tool wear state monitoring in the numerical control machining process as an example, the implementation of the state monitoring task based on high-frequency real-time data is specifically described. In the field layer, the numerical control machine tool is used as the on-site production equipment. Cutting force and vibration sensors are deployed on this equipment, and three-axis cutting force and three-axis vibration high-frequency signals are collected in real time through the data acquisition equipment during the machining process. The data acquisition frequency is 10,000 Hz. In order to perform real-time state monitoring of tool wear during milling, multiple sub-threads are created to achieve wear state identification and ensure real-time performance. After deploying the data-driven wear monitoring algorithm, through multiple functional modules such as database pulling, feature processing, and model reasoning, the current tool wear state can be identified in real time. At the same time, through the visualization interaction node on the edge side, the triggering of the state monitoring task, the visualization of the corresponding results, and the distribution can be realized.

[0058] The specific implementation process is as follows: During the initialization process, the feature processing thread, wear monitoring, and database data subscription sub-thread are declared first. Then, database consumers and corresponding topics are created for the sensor signals required for model inference, and the consumers subscribe to the corresponding topics. When the function is running, the data-driven algorithm to be deployed is selected in advance. After the wear monitoring function is enabled, the corresponding data subscription sub-thread starts running, continuously pulling data from the database and determining whether the data corresponds to the cutting state during the operation of the machine tool. The data during the cutting process is cached. After the amount of cached data reaches the set threshold Ns (such as Ns > 22000), these cached data are sent to the feature processing thread. In the feature processing thread, the received original multi-channel sensor data will go through data segmentation, feature extraction, and normalization processes in sequence, and be transformed into the data input structure required for wear monitoring inference. Then, these model inputs will be sent to the wear monitoring thread for wear state determination. In the wear monitoring thread, the deployed data-driven algorithm will determine the state of the current model input, and the inferred wear state will be sent to the visualization interface. The visualization interface will update and display the current wear state and make monitoring records, and when the monitoring result storage is triggered, the state monitoring records will be saved to the database. In addition, when the tool wear monitoring algorithm recognizes that the tool wear state is about to fail, the alarm mechanism will be triggered, and relevant warning information will be transmitted to the on-site layer control system for reminder.

[0059] Real-time performance is one of the important performance indicators for evaluating the on-line tool wear state monitoring. The running times of links such as the data transmission delay from the on-site side to the edge side, the database writing time, the database subscription time, the feature processing, and the model inference time mainly determine the real-time response ability of the on-line tool wear monitoring. Therefore, the response time performance of the above-mentioned links is tested. For the data transmission delay and packet loss rate indicators between the on-site side and the edge side, performance tests are respectively carried out with data sending amounts of 400,000, 5 million, 10 million, 25 million, and 50 million. Figure 7 This is a test schematic diagram of the data transmission delay and packet loss rate indicators provided by the embodiments of this specification, and the performance curve as shown in Figure 7 is obtained. In addition, based on the amount of sensor data required for a single on-line wear monitoring (about 25,000), the running times of each function on the edge side are tested. Figure 8 This is a test result schematic diagram of the running times of each function on the edge side provided by the embodiments of this specification. The writing rate of the time-series database on the edge side is 100%, and the data writing functions of each channel are implemented in a multi-process parallel manner, as shown in Figure 8As shown, the single-channel database write time is 11.19325 ms. Similarly, the database subscription function for each channel is implemented in a multi-threaded parallel manner, and the database subscription time is 5.12325 ms. Inside the feature processing thread, the original data is first segmented, and then the segmented data is subjected to feature extraction and normalization processing in the time domain, frequency domain, and time-frequency domain. The feature processing thread and the data-driven model inference times are 78.15453 ms and 24.93661 ms respectively. Therefore, the overall time for a single wear monitoring function on the edge side is approximately 119.4 ms.

[0060] Through the technical solution of the embodiments of this specification, through multiple topic subscription channels, it supports ≥ High-frequency data acquisition and real-time transmission at 10 kHz, effectively capturing transient features. Different from traditional condition monitoring systems that mostly rely on low-frequency data acquisition, it can effectively capture the instantaneous dynamic features during equipment operation, solve the problem that traditional solutions cannot effectively capture the high-frequency dynamic features of equipment, and avoid the lag in anomaly detection; using inter-process communication to store multi-channel high-frequency sensing data in a time-series storage node preset on the edge side, changing the existing centralized cloud processing solution. In traditional centralized cloud processing, the transmission of high-frequency data poses extremely high requirements on network bandwidth and real-time performance, and it is often difficult to achieve real-time response due to data transmission delays. The embodiments of this specification store the data on the edge side, reducing the data transmission volume and distance, reducing the dependence on network bandwidth, and at the same time reducing transmission delays, enabling real-time processing of high-frequency data and meeting the real-time requirements for real-time monitoring of equipment status; the time-series storage node adopts a supertable structure repository based on TDengine. As a database specifically designed for time-series data, TDengine can efficiently store and manage high-frequency time-series data, facilitating the storage and management of data at similar data acquisition points, reducing data redundancy, and improving data storage and query efficiency, making it more convenient and efficient to process a large amount of high-frequency time-series data and providing a good data storage foundation for subsequent data processing and analysis; in an industrial scenario, condition monitoring needs to process multi-dimensional signals simultaneously. Traditionally, single-threaded serial processing cannot efficiently handle the parallel analysis requirements of multi-channel high-frequency data streams. The embodiments of this specification can process multi-channel high-frequency data in parallel through multi-threaded analysis, making full use of system resources, improving data processing efficiency, and being able to quickly and accurately determine the real-time status data of the target production equipment; through a series of operations such as obtaining high-frequency data, processing and storing data on the edge side, and multi-threaded analysis, it can monitor the equipment status in real time and detect potential faults in a timely manner, realizing the transformation from traditional regular maintenance to real-time and intelligent predictive maintenance.

[0061] The embodiments of this specification also provide a real-time status monitoring device for high-frequency time-series data, such as Figure 9As shown, the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above method.

[0062] An embodiment of this specification also provides a non-volatile computer storage medium storing computer-executable instructions configured to execute the above method.

[0063] The various embodiments in this specification are described in a progressive manner. For the same or similar parts among the various embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0064] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the specific order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0065] The device and medium provided by the embodiments of this specification correspond one-to-one with the method. Therefore, the device and medium also have beneficial technical effects similar to those of their corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be elaborated here.

[0066] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, a system, or a computer program product. Therefore, this specification can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Moreover, this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This specification is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 one or more of the blocks for implementing the specified functions.

[0068] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 one or more of the blocks.

[0069] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 one or more of the blocks.

[0070] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0071] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0072] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0073] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0074] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. A real-time status monitoring method for high-frequency time-series data, characterized in that, The method includes: Obtain the status monitoring task of the target production equipment for initialization configuration, establish a multi-topic subscription channel corresponding to the status monitoring task, and obtain multi-channel high-frequency sensing data on the field side corresponding to the target production equipment through the multi-topic subscription channel; Use inter-process communication to store the multi-channel high-frequency sensing data in a time-series storage node preset on the edge side, where the time-series storage node includes a supertable structure repository built based on TDengine; According to the AI model node preset on the edge side, perform multi-threaded analysis on the multi-channel high-frequency sensing data in the time-series storage node to determine the real-time status data of the target production equipment.

2. The real-time status monitoring method for high-frequency timing data according to claim 1, characterized in that Obtain the status monitoring task of the target production equipment for initialization configuration, and establish a multi-topic subscription channel corresponding to the status monitoring task, which specifically includes: Load the preset AI analysis model according to the status monitoring task; Analyze the input requirements of the AI analysis model to determine the sensor channels corresponding to the real-time data types required by the AI analysis model, and create corresponding topics based on the sensor channels; Create consumers and subscribers for each topic through the preset communication middleware, and establish a consumer group subscription relationship to determine the multi-topic subscription channel corresponding to the status monitoring task.

3. A real-time status monitoring method for high-frequency time-series data according to claim 1, characterized in that Obtain the multi-channel high-frequency sensing data corresponding to the target production equipment through the multi-topic subscription channel, which specifically includes: Determine the publisher and the corresponding field-side callback function created in the field-edge interaction node on the field side in advance, and the subscriber and the corresponding edge-side callback function created in the edge-field interaction node on the edge side in advance; Receive the real-time high-frequency sensing data corresponding to the target production equipment through the field-side callback function of the field-edge interaction node, perform data cleaning on the real-time high-frequency sensing data, and then publish it to the target topic; Subscribe to the data stream of the target topic through the subscriber and the corresponding edge-side callback function of the edge-field interaction node to obtain the multi-channel high-frequency sensing data corresponding to the target production equipment.

4. A real-time status monitoring method for high-frequency time-series data according to claim 1, characterized in that, Use inter-process communication to store the multi-channel high-frequency sensing data in a time-series storage node preset on the edge side, which specifically includes: Create multiple processes and corresponding process queues in advance, where the processes include an edge middleware receiving process, a database writing process, and a monitoring process; Start the multiple processes and process queue services, and establish a connection relationship between the edge middleware and the time-series database of the time-series storage node; Through the edge middleware receiving process, encapsulate the multi-channel high-frequency sensing data in time series, generate high-frequency time-series data, and then write it into the process queue; The database writing process reads the high-frequency time-series data from the process queue and fills the data fields according to the predefined supertable structure; Through the monitoring process, monitor the cache data volume in real time to obtain the current cache data volume, and judge whether the current cache data volume reaches the preset single-batch writing threshold. If it is satisfied, generate a batch SQL insert statement to store it in the time-series database.

5. The real-time status monitoring method for high-frequency time-series data according to claim 3, wherein Subscribe to the data stream of the target topic through the subscribers of the edge-site interaction node and the corresponding edge-side callback functions to obtain the multi-channel high-frequency sensing data corresponding to the target production device, specifically including: Trigger a data pulling operation through the edge-side callback function of the subscriber to batch pull data blocks from the time-series storage node; Automatically submit the database consumption position in the edge-side callback function to persist the data consumption progress to the distributed coordination service; Determine whether the pulled data block is empty. If the data block is not empty, perform deserialization processing and supertable format conversion on the data block to obtain the multi-channel high-frequency sensing data corresponding to the target production device.

6. The real-time status monitoring method for high-frequency time-series data according to claim 1, characterized in that According to the AI model node pre-set on the edge side, perform multi-threaded analysis on the multi-channel high-frequency sensing data in the time-series storage node to determine the real-time status data of the target production device, specifically including: In the AI model node, load the corresponding AI analysis model according to the status monitoring task; Parse the AI analysis model to create multiple analysis sub-threads corresponding to the AI analysis model, where the analysis sub-threads include a database pulling thread, a feature processing thread, and a feature recognition thread; Through the multiple analysis sub-threads, perform multi-threaded analysis on the multi-channel high-frequency sensing data to determine the real-time status data of the target production device.

7. A real-time status monitoring method for high-frequency time-series data according to claim 6, characterized in that, Through the multiple analysis sub-threads, perform multi-threaded analysis on the multi-channel high-frequency sensing data to determine the real-time status data of the target production device, specifically including: Through the database pulling thread, pull multi-channel high-frequency sensing data in the time-series storage node; Through the feature processing thread, extract and combine features of the multi-channel high-frequency sensing data to send to the feature recognition thread, and based on the feature recognition thread, generate the real-time status recognition result of the target production device.

8. A real-time status monitoring method for high-frequency time-series data according to claim 7, characterized in that After pulling the multi-channel high-frequency sensing data in the time-series storage node through the database pulling thread, the method further includes: Determine whether the target production device is in a running state; If so, cache the multi-channel high-frequency sensing data in the running state and determine the cache data volume; When the cache data volume is not less than the preset quantity threshold, send it to the feature processing thread.

9. A real-time status monitoring device for high-frequency time-series data, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method according to any one of claims 1-8.

10. A non-volatile computer storage medium stores computer-executable instructions, characterized in that, The computer-executable instructions are set to: execute the method according to any one of claims 1-8.

Images