An ultra-high availability virtual industrial intelligent control platform and method
By deploying distributed shared memory modules and real-time runtime suites in a virtual computing cluster, and isolating management and maintenance functions, and adopting multi-master synchronization and incremental data synchronization, the problem that core control functions of existing industrial control systems are easily affected under extreme conditions is solved, achieving high availability and fault tolerance.
Patent Information
- Application Number
- CN202611312906.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-11-17
AI Technical Summary
Existing industrial control systems are prone to core control functions being affected by node failures, network outages, or abnormalities in management and maintenance systems, resulting in insufficient availability and fault tolerance.
An ultra-high availability virtual industrial intelligent control platform was designed. By deploying distributed shared memory modules and real-time running suites in a virtual computing cluster, the data flow of the core control loop is ensured to be completed locally on the running node. The management and maintenance suite is isolated from the core control loop. Multi-master synchronization and incremental data synchronization methods are adopted, combined with task level adjustment and node health management, to ensure that the platform maintains high availability and fault tolerance under extreme conditions.
In the event of node failure, network interruption, or management and maintenance anomalies, the core control loop remains uninterrupted, ensuring high availability and fault tolerance of the platform, guaranteeing data consistency and security, and improving resource utilization efficiency.
Smart Images

Figure CN123113530A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial control technology, and in particular to an extremely high-availability virtual industrial intelligent control platform and method. Background Technology
[0002] In scenarios such as advanced control, smart power plants, and auxiliary network control, industrial control platforms need to access complex algorithms, AI models, and cross-system collaboration capabilities. They also need to be flexibly deployed on various hardware forms such as micro DPUs, industrial PCs, servers, virtual machines, and containers, depending on the project scale. At the same time, under extreme conditions such as node failures, they still need to ensure that core control loops such as communication acquisition, logic calculation, and output distribution are not interrupted.
[0003] In existing solutions, traditional distributed control systems mainly rely on dedicated controller hardware and closed engineering configuration environments to achieve continuous process control; some solutions utilize cloud-native high-availability architectures such as Kubernetes, Redis clusters, Redis Sentinels, and master-slave databases to achieve service continuity through master-slave switching, majority writes, or automatic service migration.
[0004] However, existing technologies have two related drawbacks. Firstly, current cloud-native high-availability architectures primarily rely on overall service failover, with fault migration or master-slave switching having windows of seconds or even sub-seconds. Furthermore, after node recovery, data freshness is not verified before providing services, easily leading to the reintroduction of outdated measurement data, task statuses, or alarm states into the control loop, failing to meet the deterministic and data security requirements of industrial control loops. Secondly, existing solutions do not distinguish between core control loops and non-core functions such as configuration downloads, historical queries, and report calculations. Under general scheduling strategies, non-real-time tasks may preempt resources needed for communication and logical calculation tasks. When the number of nodes decreases or resources are scarce, the core control loop lacks independent resource guarantees and tiered degradation mechanisms, posing a risk of overall availability degradation or even control interruption. Summary of the Invention
[0005] In view of the above-mentioned problems, the technical problem to be solved by the present invention is to provide an extremely high availability virtual industrial intelligent control platform and method to solve the technical problems of insufficient scalability and intelligent computing capabilities of existing industrial control systems, and the lack of design for the data path of the core industrial control loop in the existing high availability architecture, which leads to the core control function being easily affected when node failures, network interruptions or management and maintenance system abnormalities occur, and insufficient availability and fault tolerance.
[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides an extremely high-availability virtual industrial intelligent control platform, comprising: Virtual computing clusters, real-time runtime suites, and management and maintenance suites; A virtual computing cluster includes at least one running node, each running node is deployed with a real-time running suite, and multiple running nodes form a running group; The real-time runtime suite includes a distributed shared memory module, a communication task module, a logic computing module, and an auxiliary function module; The auxiliary function modules include a lightweight relational database module, a node health management module, and a recovery and synchronization module, which are used to store configuration data, detect running status, and synchronize real-time data and task status after the running node recovers from a failure, respectively. The communication task module is used to collect measurement point data from external cards or field devices and write it into the distributed shared memory module. It is also used to read the output measurement point data in the distributed shared memory module and send it to external cards or field devices. The logic calculation module is used to read the measurement point data in the distributed shared memory module to perform control logic calculations, and write the generated output measurement point data into the distributed shared memory module. Among them, the core control loop is formed between external cards or field devices, communication task modules, distributed shared memory modules and logic computing modules. Communication task modules and logic computing modules have priority to access the distributed shared memory modules in their respective running nodes. The management and maintenance suite provides functions for engineering configuration, screen configuration, historical data management and operation and maintenance, but does not directly participate in the core control loop.
[0007] Secondly, the present invention provides an extremely high availability virtual industrial intelligent control method, comprising: Obtain the running status, task status, and resource status of the running nodes in the virtual industrial intelligent control platform; The availability status of the platform is determined based on the aforementioned operating status, task status, and resource status; The task level of the control task is determined based on the available status, and the operation mode of the corresponding control task is adjusted according to the task level. The system acquires real-time data and task status corresponding to the running node, and synchronizes the real-time data and task status with other running nodes in the running group to obtain synchronized data. The running group consists of multiple running nodes in the virtual industrial intelligent control platform. The synchronized data is subjected to status verification to obtain the verification result, and the service status of the running node is determined based on the verification result.
[0008] The technical problem to be solved by the present invention is to provide an extremely high availability virtual industrial intelligent control platform and method, which is used to solve the technical problems of existing virtual industrial intelligent control systems where control functions rely on cross-node communication or management and maintenance systems, core control functions are easily affected when nodes fail, networks are interrupted, or management and maintenance systems are abnormal, and availability and fault tolerance are insufficient.
[0009] The beneficial effects of this invention are: in extreme cases such as partial failure of running nodes, network interruption, or even abnormality of management and maintenance suite, since the data flow of the core control loop is strictly limited to the distributed shared memory module on the local running node, the communication task module and the logic calculation module do not need to rely on cross-node communication to complete reading and writing, and the relevant functions of management and maintenance suite are isolated outside the core control loop, as long as there is still one running node working normally, the core control loop of that node will not be interrupted due to abnormalities of other nodes, network, or management and maintenance suite, thereby enabling the platform as a whole to have high availability and fault tolerance.
[0010] Furthermore, the distributed shared memory modules of the multiple running nodes use a multi-master synchronization method to synchronize data. After the local measurement point data changes, the running node generates incremental data and sends it to other running nodes. After receiving the incremental data, the other running nodes compare the version number carried by the incremental data with the current version number of the corresponding local measurement point, and determine whether to update the local data based on the comparison result.
[0011] The beneficial effects of adopting the above-mentioned further scheme are: while maintaining the priority of each running node to read and write local shared memory instances and completing the core control loop without relying on cross-node communication, the measurement data between each running node can still be kept consistent through incremental synchronization and version number comparison, avoiding the overwriting of updated local data by old version data, and taking into account both the independence of the core control loop and the consistency of data between multiple nodes.
[0012] Furthermore, the distributed shared memory module is also equipped with a conflict determination mechanism. When the same measurement point data is written to by multiple running nodes at the same time in a short period of time, resulting in data conflicts with the same version number but different content, the write result that should be adopted is determined based on the write source, write priority, timestamp, version number and quality code.
[0013] The beneficial effects of adopting the above-mentioned further solution are: compared with using only version number or arrival time as the sole criterion, the method of multi-field hierarchical judgment based on write source, write priority, timestamp, version number and quality code avoids high-priority or high-quality data that should be retained being overwritten by poor-quality writes, thus ensuring the correctness and consistency of test point data in scenarios with concurrent writes from multiple running nodes.
[0014] Furthermore, the recovery and synchronization module in the auxiliary function module is used to restore the measurement point data, task version, alarm status and operation status missed during the failure period to the running node when it is reconnected to the system after the running node fails. After verification, the running node is put into service status.
[0015] The beneficial effect of adopting the above-mentioned further solution is that the fault recovery operation node will not be hastily brought online to participate in the read and write scheduling of the core control loop with expired or incomplete data. Instead, it will enter the service state only after going through the discovery, catch-up, and verification states in sequence and confirming that the data is complete and consistent. This ensures that the core control loop still has high data security and recovery reliability in scenarios such as node failure and restart.
[0016] Furthermore, the lightweight relational database module in the auxiliary function module is used to store the local measurement point configuration, communication task configuration, and computing task configuration of the running node, which can be read by the running node during startup, fault recovery, and operation; the node health management module is used to detect the process status, task status, network status, shared memory synchronization status, and resource status of the running node in a fixed periodic polling manner, and use the detection results as the basis for judgment.
[0017] The beneficial effects of adopting the above-mentioned further solution are that the running node can independently complete the configuration reading and status judgment without relying on the remote database or remote service where the management and maintenance suite is located, so as to know its own configuration content and running status, further reducing the dependence of the core control loop on external systems.
[0018] Furthermore, determining the platform's availability status based on the operating status, task status, and resource status includes: counting the number of currently operating nodes; if the number reaches one, the platform is determined to be in a core availability status; if the number of operating nodes operating normally exceeds half of the total number of nodes in its operating group, the platform is determined to be in a fully available status.
[0019] The beneficial effects of adopting the above-mentioned further solution are: the core availability state and the full availability state constitute two progressive judgment levels, enabling the platform to dynamically switch between ensuring only the core control capability and ensuring both control and management and maintenance functions based on the different number of normally operating nodes. Under extreme fault conditions, priority is given to ensuring the continuity of the core control loop, without having to forcibly maintain full functionality when the number of nodes is insufficient.
[0020] Furthermore, for the running nodes within the running group, the method further includes: dividing the running nodes into computing groups, and assigning the measurement point data, communication tasks, and logical computing tasks to the corresponding computing groups; setting a shared measurement point identifier for the measurement point data that needs to be used across computing groups; publishing the measurement point data with the shared measurement point identifier to a shared topic; having other computing groups subscribe to the shared topic, and writing the subscribed measurement point data to the mirror measurement point of the corresponding computing group.
[0021] The beneficial effects of adopting the above-mentioned further scheme are: after dividing the running nodes into computing groups, high-frequency control logic and strongly coupled interlocking operations are restricted to be completed within the same computing group. Data transfer across computing groups is completed through marking, publishing and subscribing, which avoids the impact of cross-group high-frequency dependencies on the core control loop, while reducing unnecessary cross-group synchronization pressure.
[0022] The beneficial effects of adopting the above-mentioned further solutions are: by adjusting the operation mode of control tasks according to task level, limited computing and network resources can be prioritized to serve tasks that are of high importance to the core control loop, and the normal operation of the core control loop can still be guaranteed even in resource-constrained fault scenarios.
[0023] Furthermore, the process of acquiring the real-time data and task status corresponding to the running node, and synchronizing the real-time data and task status with other running nodes in the same running group to obtain synchronized data, includes: when a running node rejoins the running group after a failure and going offline, it sends a rejoin request to other running nodes in the running group; upon receiving the request, a running node in the running group that is operating normally sends its locally stored measurement point snapshots, task version, alarm status, and operation status to the running node to obtain the synchronized data.
[0024] The beneficial effect of adopting the above-mentioned further solution is that the fault-recovered operating node can supplement the missing measurement point data, task version, alarm status and operation status during the fault by requesting data from the normally operating nodes in the operating group, thus avoiding the inability of the fault-recovered operating node to accurately participate in the subsequent control process due to the lack of data generated during the fault.
[0025] Furthermore, the step of performing status verification on the synchronized data to obtain verification results, and determining the service status of the running node based on the verification results, includes: verifying the version number, timestamp, quality code, data integrity, and configuration consistency of the synchronized data; if the verification result is that all verifications pass, then the running node is determined to enter the service status; if the verification result shows that there are failed items, then the running node is determined to not yet meet the conditions for entering the service status, and the data is re-acquired and synchronized and verified again.
[0026] The beneficial effects of adopting the above-mentioned further solutions are: before the fault-recovered running nodes enter the service state and participate in the read and write scheduling of the core control loop, they need to undergo multiple checks such as version number, timestamp, quality code, data integrity and configuration consistency. This avoids running nodes carrying expired or inconsistent data from hastily entering the service state and improves the overall operational security of the platform. Attached Figure Description
[0027] Figure 1 The overall architecture diagram provided for embodiments of the present invention; Figure 2 This is a data flow diagram of the core control loop provided in an embodiment of the present invention; Figure 3 This is a diagram of a multi-master distributed shared memory synchronization structure provided in an embodiment of the present invention; Figure 4 This is a node fault and recovery state machine diagram provided in an embodiment of the present invention; Figure 5 This is a synchronization diagram of calculation groups and cross-group shared measurement points provided in an embodiment of the present invention; Figure 6 This is a flowchart illustrating the task hierarchy degradation process provided in an embodiment of the present invention. Detailed Implementation
[0028] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.
[0029] Example 1 Combination Figure 1 As shown, the ultra-high availability virtual industrial intelligent control platform provided by the present invention consists of three parts: a virtual computing cluster, a real-time running suite, and a management and maintenance suite. The virtual computing cluster contains at least one running node, and each running node is equipped with a real-time running suite, which is responsible for the real-time data processing and control calculation of that node.
[0030] Within each running node, the real-time runtime suite is further divided into a distributed shared memory module, a communication task module, a logic calculation module, and an auxiliary function module.
[0031] The auxiliary function module does not directly participate in control calculations, but provides supporting functions for the nodes. It includes a lightweight relational database module, a node health management module, and a recovery and synchronization module, which are respectively responsible for saving configuration data, detecting the node's own running status, and resynchronizing real-time data and task status back to the node after a failure and recovery.
[0032] Data flow within a node revolves around a distributed shared memory module, as shown in the reference. Figure 2As shown, the measurement point data generated by field devices or external cards is first collected by the communication task module and written into the distributed shared memory module of this node. In addition, the logic calculation module does not directly take data from the communication task module, but reads these measurement point data from the shared memory, completes the control logic operation, and then writes the output result of the operation back to the same shared memory, instead of sending it directly to the communication task module.
[0033] Finally, the communication task module reads the output from the shared memory and sends it to the field device or external card. In this way, the field device, the communication task module, the shared memory, and the logic calculation module are connected in series to form a complete core control loop.
[0034] The key to the stable operation of this loop lies in the fact that the communication task module and the logic calculation module always access the shared memory within this node, without needing to read or write across nodes. Because of this, as long as there is another running node working normally, its own core control loop will not be interrupted due to problems with other nodes or the network.
[0035] In addition, the platform is equipped with a management and maintenance suite, which consists of an engineer station, an operator station, a historical data cluster, and a standard relational database. It is connected to the virtual computing cluster through a redundant industrial network and is responsible for engineering configuration, screen configuration, historical data archiving, and daily maintenance. This part of the function is specifically placed outside the core control loop and does not participate in the real-time data processing link. In this way, even if the management and maintenance suite malfunctions or the network is temporarily interrupted, it will not affect the control loops that are running inside each operating node.
[0036] It is through this design, in which data within nodes is strictly transferred through shared memory, control loops do not rely on cross-node communication, and management and maintenance functions are isolated from real-time control functions, that the platform can maintain the continuous operation of core control functions even in extreme situations such as partial node failures, network interruptions, or even abnormalities on the management and maintenance side, thus possessing high availability and fault tolerance.
[0037] In this embodiment, multiple running nodes in the virtual computing cluster form a running group. In this embodiment, the distributed shared memory module is not a centrally deployed single instance, but rather each running node maintains its own shared memory instance, such as... Figure 3 As shown, each of running node 1, running node 2, and running node 3 has its own local shared memory instance. Local communication tasks and logical calculation tasks prioritize reading and writing to this shared memory instance within their own node, and can complete the read and write operations without waiting for confirmation from other nodes.
[0038] The shared memory instances across multiple nodes operate using a multi-master synchronization method. Specifically, each node maintains a complete copy of its local measurement point data. When a local measurement point changes, it packages the measurement point's identifier, value, quality code, version number, and timestamp into an incremental data entry and sends it to other nodes through the measurement point incremental synchronization channel. Upon receiving the incremental data, the receiving node first retrieves the version number currently stored for that measurement point in its local shared memory instance and compares it with the version number carried in the incremental data. If the version number of the incremental data is greater than the local version number, the measurement point's value, quality code, timestamp, and version number fields are completely overwritten with the corresponding content in the incremental data. If the version number of the incremental data is not greater than the local version number, the incremental data entry is discarded without any update.
[0039] By comparing version numbers one by one before deciding whether to overwrite a field, the visibility of measurement data between nodes is maintained, while avoiding the overwriting of updated local data by older versions.
[0040] Based on this, when the same measurement point data is written by multiple nodes simultaneously in a short period of time, after each node receives the write request from each other through incremental synchronization, data conflicts may occur where the version number is the same but the content is different. At this time, it is no longer possible to determine the choice based solely on version number comparison. In this embodiment, the distributed shared memory module adopts a field conflict determination mechanism to solve this problem: first, the source of the write is determined, and the write generated by the local node takes priority over the remote write synchronized from other nodes. If the write sources have the same priority, the write priority field is compared, with writes generated by manual operations taking precedence over writes generated by automatic tasks. If the write priorities are still the same, the timestamp (newer one), version number (incrementing one), and quality code (higher quality one) are compared in turn to determine which write result should be adopted and updated to the effective measurement point data.
[0041] The quality code is used to characterize the reliability of the measurement point data. It is divided into multiple levels according to a preset quality grade. The higher the grade, the higher the reliability of the corresponding measurement point data.
[0042] Compared to simply using the last arriving write as the standard, this multi-field hierarchical judgment method avoids high-priority or high-quality data that should be retained being overwritten by later but lower-quality writes, thus ensuring the correctness and consistency of measurement point data in multi-node concurrent write scenarios.
[0043] Furthermore, this embodiment also involves the recovery process after a running node failure. When the failed running node reconnects to the system, it does not directly participate in the control loop in its pre-failure state, but rather goes through several stages in sequence: startup, discovery, catch-up, verification, pre-service, and service. Figure 4As shown, the node first starts its own process, then finds its running group in the network and determines the synchronization object; Once the catch-up phase begins, a complete snapshot of the measurement points is obtained from any surviving node in the operation group, and then the incremental log of that node is synchronized, thereby gradually restoring the measurement point data, task version, alarm status, and operation status missed during the fault to the local machine. During the catch-up phase, incremental logs are continuously synchronized. When the version number of the measurement data of the fault recovery running node is consistent with the version number of the measurement data of the synchronized surviving running node, the catch-up phase is considered complete and the verification phase begins. If the surviving running node generates new incremental logs during the catch-up process, the fault recovery running node continues to synchronize that part of the incremental logs until the version numbers of the two are consistent.
[0044] After synchronization is complete, the verification phase begins, comparing five fields in sequence: version number, timestamp, quality code, data integrity, and configuration consistency. The version currently confirmed by the running group refers to the version number with the largest value among the surviving running nodes for the same test point data. During the verification phase, the fault recovery running node compares its local version number updated during the catch-up phase with the version number held by the surviving running nodes one by one. If it reaches or exceeds the largest version number among the surviving running nodes, the verification passes. However, only when all verifications pass can it enter the pre-service state. At this time, although the node has been connected to the running group, it does not undertake core control tasks. Only after all statuses are confirmed to be stable will the system finally enter the service state and officially participate in read / write operations and task scheduling. If the verification fails, the system will revert to the catch-up phase to continue synchronization, rather than directly entering the service state with unverified outdated data.
[0045] This recovery process avoids hastily bringing faulty nodes online with expired or incomplete data, thus ensuring that the core control loop still has high data security and recovery reliability in scenarios such as node failure and restart.
[0046] In addition, the lightweight relational database module and the node health management module included in the auxiliary function module in this embodiment also play specific supporting roles. The lightweight relational database module stores the local measurement point configuration, communication task configuration, and computing task configuration of the running node. These configurations are generated and downloaded by the management and maintenance suite after the engineer completes the configuration. The management and maintenance suite first performs integrity and permission verification on the configuration package, and then sends it to the target running node. After receiving it, the running node writes it into its local lightweight relational database module, which can be read at any time during startup, fault recovery, and operation, without relying on the remote database where the management and maintenance suite is located. After writing, the node health management module loads the configuration and performs semantic verification. Only after the verification is passed will the corresponding communication task module and logical computing module run according to the new configuration version. Among them, the node health management module detects the process status, task status, network status, shared memory synchronization status, and resource status of the node in a fixed periodic polling manner, and uses the detection results as the basis for judgment for the recovery synchronization module to determine whether the node has the conditions to resume providing services.
[0047] The existence of these two modules allows the running nodes to perform configuration reading and status judgment independently of the management and maintenance suite, further reducing the dependence of the core control loop on external systems.
[0048] In this embodiment, the real-time runtime suite can also include a message publish-subscribe module. This module, implemented using a distributed shared memory module, is used for event notification within the node and between nodes. Specifically, when the value, alarm status, or task status of a measurement point in the distributed shared memory module changes, the message publish-subscribe module publishes this change as an event. The communication task module, logic calculation module, or other modules subscribed to this event can then promptly learn of the measurement point change without needing to continuously poll the shared memory to detect data updates. By employing this mechanism, the latency of data change detection between modules is reduced to some extent, while also minimizing unnecessary polling overhead.
[0049] Furthermore, the isolation between the real-time runtime suite and the management and maintenance suite in this embodiment is not limited to functional division; it can also be further achieved through resource-level isolation: the communication components, logical computing components, and shared memory components corresponding to the communication task module, logical computing module, and distributed shared memory module can be bound to dedicated CPU cores or set to real-time priority scheduling; while the historical database, object storage, reporting services, and large-scale query operations on which the management and maintenance suite depends are subject to corresponding resource limits, thereby preventing them from preempting the computing resources required by the real-time runtime suite when resources are scarce. This resource-level isolation method, combined with the aforementioned functional-level isolation, ensures that the real-time runtime suite can still obtain stable computing resources even when the management and maintenance suite's load increases, further improving the determinism of the core control loop operation.
[0050] The running nodes in this embodiment can adopt different hardware forms during specific deployment, including but not limited to ARM microprocessor units, industrial control computers, physical servers, virtual machines or container nodes, while the industrial network can adopt network forms that can meet the real-time requirements of the field, such as redundant Ethernet, 10 Gigabit Ethernet, TSN enhanced network.
[0051] Based on the above structure, this platform can adapt to deployment scenarios of different scales: in small-scale scenarios, only two low-power processing units can be used as running nodes, with an additional maintenance station to deploy the management and maintenance suite; in medium-scale scenarios, three to six industrial control computers or servers can be used to form a computing group, with separate operator stations, engineer stations, and historical stations; in large-scale scenarios, server clusters, virtualization platforms, or container platforms can be used, divided into multiple computing groups according to process areas, and the historical database, object storage, and reporting services can be deployed independently. Furthermore, the above implementation methods with different hardware forms and deployment scales do not change the basic structure of the aforementioned core control loop, priority read / write of shared memory, and isolation of the management and maintenance suite.
[0052] Example 2 This embodiment provides an ultra-high availability virtual industrial intelligent control method, which is applied to the ultra-high availability virtual industrial intelligent control platform described in Embodiment 1 above.
[0053] In this embodiment, multiple running nodes in the virtual computing cluster form a running group. First, the node health management module on each running node continuously detects the process status, network status, and shared memory synchronization status of the node, and writes the detection results as running status, task status, and resource status into the distributed shared memory module of the node for other nodes and upper-layer judgment logic to read, thereby obtaining the running status, task status, and resource status of each running node in the platform.
[0054] The platform's availability is then determined based on the running status, task status, and resource status of each running node. First, the number of currently active running nodes is counted: as long as at least one running node is active, the platform is considered to be in a core available state. At this time, the platform can still maintain communication acquisition, logic calculation, output distribution, and critical alarm capabilities. Multiple running nodes form a running group. If the number of normally operating nodes exceeds half of the total number of nodes in the running group, the platform is further considered to be in a fully available state. At this time, the platform can still maintain complete functions such as configuration download, screen monitoring, historical trends, reports, and system maintenance.
[0055] After determining the platform's availability status, each control task is classified into different levels based on its importance to the core control loop. When the platform is in a core available state but has not reached a fully available state, control tasks with lower importance will have their execution frequency reduced or be suspended. This allows limited computing and network resources to be allocated preferentially to control tasks with higher importance to the core control loop, ensuring that the core control loop can still be prioritized for protection even under resource-constrained conditions.
[0056] When a running node goes offline due to a fault and then rejoins the running group, it obtains the corresponding measurement point snapshot, task version, alarm status, and operation status from other running nodes in the running group that are still alive. It then synchronizes this data to the newly recovered running node, allowing the running node to fill in the data missed during the fault and obtain synchronized data.
[0057] For the synchronized data, each item—version number, timestamp, quality code, data integrity, and configuration consistency—is verified: the version number must meet the version currently confirmed by the running group; the timestamp must be within the allowable deviation range of the running group; the quality code must meet the pre-set quality level; data integrity is verified and compared to confirm that no data has been lost or corrupted; and configuration consistency is confirmed by comparing the local configuration with the current configuration of the running group. If all the above verifications pass, the running node is determined to enter the service state to provide services. If any verification fails, the running node is determined not to meet the conditions for entering the service state, and data is retrieved from a surviving running node and synchronized and verified again.
[0058] This embodiment continuously acquires the status of running nodes and determines the platform's availability accordingly. This allows the platform to dynamically adjust its functional scope based on the scale of currently surviving nodes, prioritizing core control capabilities under extreme failure conditions. By adjusting the operation mode of control tasks according to task levels, limited computing resources can be prioritized for control tasks of higher importance. Through data synchronization and item-by-item status verification of recovered running nodes, the embodiment avoids, to some extent, allowing recovered running nodes to directly participate in the control process with outdated or incomplete data, thereby improving the overall operational security and resource utilization efficiency of the platform.
[0059] Furthermore, in this embodiment, the running status, task status, and resource status represent different judgment dimensions: the running status indicates whether the running node is currently operating normally; the task status indicates the current execution status of each control task on the running node; and the resource status indicates the current resource occupancy of the running node. The aforementioned determination of the core availability status and the full availability status is based on the number of normally operating nodes obtained from the statistics of the running status of each running node. Among them, the core availability status and the full availability status constitute two progressive judgment levels. The former only requires that the core control link is not interrupted, while the latter requires that most nodes in the running group are operating normally. The distinction between the two allows the platform to dynamically switch between "ensuring control only" and "ensuring control and ensuring management and maintenance functions" based on the different number of normally operating nodes, without having to forcibly maintain full functionality when the number of nodes is insufficient. Thus, in fault scenarios with limited resources, the continuity of the core control loop can still be prioritized.
[0060] Furthermore, this embodiment also relates to the organization of running nodes at the compute group level. During the engineering configuration phase of the management and maintenance suite, the running nodes are divided into multiple compute groups according to the on-site process zoning or control coupling relationship, as shown in the following example. Figure 5 The calculation groups A and B are divided into calculation groups. The measurement point data, communication tasks and logic calculation tasks are also assigned to the corresponding calculation groups. Since the running nodes in the same calculation group maintain copies of the measurement point data through local shared memory and support local priority reading and writing, the high-frequency control logic and strongly coupled interlocking operations in the same calculation group do not need to cross the group. The local shared memory reading and writing of each node in the group can complete the closure.
[0061] The computing task configuration records which running node the logical computing task of each measurement point data in the computing group belongs to. Other nodes in the same group only retain a local copy of the measurement point data and participate in multi-master synchronization, but do not execute the task. If the node responsible for execution fails, the task will be taken over by other nodes in the group.
[0062] If data from a certain measurement point needs to be used by other computing groups, a shared measurement point identifier is set for that measurement point, marking it as a shared measurement point. When this measurement point data changes, its value, quality code, and version number are published to a cross-group shared topic. Other computing groups then subscribe to this topic, receive the data, and write it to the corresponding mirror measurement point within their own computing group. During the write operation, the received version number directly overwrites the original content of the mirror measurement point. The mirror measurement point is only read by the current computing group; nodes within the current computing group do not write to it. Therefore, there are no conflicts from simultaneous writes by multiple parties, and conflict resolution is unnecessary. For example, shared measurement point data generated by computing group A, after being transmitted to computing group B via the shared topic, only resides on the mirror measurement point of computing group B. Low-frequency logic operations, monitoring displays, or historical records within computing group B can use it, but it cannot enter the shared memory used by computing group B for high-frequency control closed-loop operations. The operation of computing group B's own core control loop will not be affected by it.
[0063] In this approach, after running nodes are divided into computing groups, the control loops within a group can be closed using local shared memory. Data transfer between groups only involves marking, publishing, and subscribing. High-frequency control loops are confined to a single computing group, and the latency of cross-group communication will not hinder real-time control. Both the control loops within a computing group and the cross-group data transfer rely on the normal operation of nodes within the running group. If a running node fails, the running group must first address the resulting task gap, rather than immediately attempting to catch up and synchronize. Once the node health management module detects a failure in a running node, the running group marks it as offline. For non-core tasks, the execution frequency is reduced or execution is paused according to their importance. Each control task is specifically divided into three levels based on its importance to the core control loop: core tasks, reduced-frequency tasks, and pauseable tasks. Figure 6 As shown: Core tasks include safety interlocks, critical control loops, critical communication links, critical alarms, and operational outputs. These tasks continue to execute normally and prioritize CPU and network resources as long as any node is alive. Reduced-frequency tasks include non-critical monitoring points, shared measurement points across computing groups, trend push notifications, and general alarms. These tasks are rate-limited or merged when the node health management module assesses resource scarcity or a decrease in the number of nodes, thus reducing their execution frequency. Paused tasks include reports, batch engineering synchronization, non-real-time analysis, object storage organization, and historical compression. These tasks do not participate in core control loops and are paused during extreme failures or resource scarcity, resuming only after resources are restored. Furthermore, the fault status, task takeover status, and alarm status of failed nodes are written to a distributed shared memory module, allowing the operator station to continue reading real-time data and issuing operational commands through any normally operating node.
[0064] The operation group first stabilizes the execution of core tasks in this way. When a certain operation node goes offline due to a failure and rejoins the operation group, it will first send a rejoin request to other operation nodes in the operation group. After receiving the request, any operation node in the operation group that is operating normally will send the measurement point snapshot, task version, alarm status and operation status stored in its local shared memory instance to the failure-recovered operation node. The measurement point snapshot is a complete copy of the measurement point data currently stored in the local shared memory instance of the sending node. The task version records the version number of the control logic and interlocking strategy that the node should run. The alarm status and operation status record the alarm information that the sending node has not yet processed and the operation records issued by the operator, respectively.
[0065] After the fault recovery node receives the above data, since its local data has failed to be continuously updated during the fault and has become completely invalid, it no longer compares the version numbers one by one. Instead, it writes the received measurement point snapshots, task versions, alarm statuses, and operation statuses into the local shared memory instance, replacing all the original old data, thus obtaining the synchronized data. At this time, the running node has not yet entered the state of providing services to the outside world, and the written data does not participate in the read and write scheduling of the core control loop.
[0066] After the synchronized data is written to the local shared memory instance, it must be verified item by item before it can be accepted: Verification involves comparing the version number to see if it has reached the version currently confirmed by the running group; comparing the timestamp to see if it falls within the allowable deviation range of the running group; comparing the quality code to see if it meets the pre-set quality level; verifying data integrity by calculating and confirming that the data has not been lost or damaged during transmission; and finally verifying configuration consistency by comparing the local measurement point configuration, communication task configuration, and calculation task configuration with the currently effective configuration of the running group to confirm consistency. Only when all five verifications pass can the running node enter the service state to provide services externally, allowing it to participate in read / write and task scheduling; if any one of them fails, it means the data is incomplete or inconsistent, requiring a new request to be sent to the running nodes in the running group to retrieve the data again and verify it once more.
[0067] Once the running node enters service status, the core communication tasks, logical calculation tasks, and critical alarm tasks that were previously executed by other normally operating nodes are returned to the running node for execution. Non-core tasks that were previously reduced in execution frequency or suspended due to resource constraints are also restored to their original execution frequency and restarted. The running group is thus restored to the task allocation state before the running node failed.
[0068] In some embodiments, the ultra-high availability virtual industrial intelligent control platform of the present invention can be implemented in a combination of hardware and software. As an example, the logic computing module can be a processor in the form of a hardware decoding processor, which is programmed to execute the ultra-high availability virtual industrial intelligent control method of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.
[0069] The invention describes how modules such as the distributed shared memory module, communication task module, logic calculation module, and auxiliary function module involved in the embodiments can be implemented in software or hardware. The names of these modules do not necessarily limit the functionality of the module itself.
[0070] An electronic device according to an embodiment of the present invention may be an industrial control computer, physical server, or virtual machine constituting a running node, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned highly available virtual industrial intelligent control methods. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the highly available virtual industrial intelligent control method shown in any embodiment of the present invention by calling the computer program. The processor may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof, and may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of the present invention. The memory can be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, magnetic disk storage media, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory is used to store application code (computer program) that executes the present invention, and is controlled by a processor to execute the application code stored in the memory to implement the content shown in the foregoing method embodiments. The electronic device can also be a terminal device, which can be any device capable of installing applications, including at least one of smartphones, tablets, laptops, and desktop computers.
[0071] This invention provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements any of the aforementioned highly available virtual industrial intelligent control methods. Optionally, the computer-readable storage medium may be a read-only memory (ROM), random access memory (RAM), compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, or optical data storage device, etc. It may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or any combination thereof, having an electrical connection with one or more wires, a portable computer disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof. The computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device. The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.
[0072] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.
[0073] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.
[0074] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.
[0075] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be embodied in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.
[0076] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A highly available virtual industrial intelligent control platform, characterized in that, include: Virtual computing clusters, real-time runtime suites, and management and maintenance suites; A virtual computing cluster includes at least one running node, each running node is deployed with a real-time running suite, and multiple running nodes form a running group; The real-time runtime suite includes a distributed shared memory module, a communication task module, a logic computing module, and an auxiliary function module; The auxiliary function modules include a lightweight relational database module, a node health management module, and a recovery and synchronization module, which are used to store configuration data, detect running status, and synchronize real-time data and task status after the running node recovers from a failure, respectively. The communication task module is used to collect measurement point data from external cards or field devices and write it into the distributed shared memory module. It is also used to read the output measurement point data in the distributed shared memory module and send it to external cards or field devices. The logic calculation module is used to read the measurement point data in the distributed shared memory module to perform control logic calculations, and write the generated output measurement point data into the distributed shared memory module. Among them, the core control loop is formed between external cards or field devices, communication task modules, distributed shared memory modules and logic computing modules. Communication task modules and logic computing modules have priority to access the distributed shared memory modules in their respective running nodes. The management and maintenance suite provides functions for engineering configuration, screen configuration, historical data management and operation and maintenance, but does not directly participate in the core control loop.
2. The platform according to claim 1, characterized in that, The distributed shared memory module includes: Multiple shared memory instances are deployed on corresponding running nodes, and the distributed shared memory module stores measurement point data using measurement points as the basic data unit. The shared memory instances in the multiple running nodes operate in a multi-master synchronous mode. Each running node maintains a local copy of the measurement point data and performs reading and writing of the measurement point data based on the local copy of the measurement point data.
3. The platform according to claim 2, characterized in that, For multiple write operations on the same measurement point data, the distributed shared memory module determines conflicts based on the write source, write priority, timestamp, version number, and quality code, and updates the corresponding measurement point data.
4. The platform according to claim 1, characterized in that, The recovery and synchronization module is used to enable the running node that has recovered from the fault to enter the startup state, discovery state, catch-up state, verification state, pre-service state and service state in sequence. In the catch-up state, it synchronizes the measurement point data, task version, alarm state and operation state. After passing the verification state, it enters the service state.
5. The platform according to claim 1, characterized in that, The lightweight relational database module in the auxiliary function module is used to store the measurement point configuration, communication task configuration, and computing task configuration corresponding to the running node; the node health management module in the auxiliary function module is used to detect the process status, task status, network status, shared memory synchronization status, and resource status of the running node.
6. A highly available virtual industrial intelligent control method, characterized in that, include: Obtain the running status, task status, and resource status of the running nodes in the virtual industrial intelligent control platform; The availability status of the platform is determined based on the aforementioned operating status, task status, and resource status; The task level of the control task is determined based on the available status, and the operation mode of the corresponding control task is adjusted according to the task level. The system acquires real-time data and task status corresponding to the running node, and synchronizes the real-time data and task status with other running nodes in the running group to obtain synchronized data. The running group consists of multiple running nodes in the virtual industrial intelligent control platform. The synchronized data is subjected to status verification to obtain the verification result, and the service status of the running node is determined based on the verification result.
7. The method according to claim 6, characterized in that, Determining the platform's availability based on the running status, task status, and resource status includes: Obtain the running status, task status, and resource status of each running node; When at least one running node is alive, the available state is determined to be the core available state; When the number of normally operating nodes is greater than half the total number of nodes in the operating group, the available state is determined to be a fully available state.
8. The method according to claim 6, characterized in that, For the running nodes within the running group, it also includes: The running nodes are divided into computing groups, and the measurement point data, communication tasks and logical computing tasks are respectively assigned to the corresponding computing groups; Set a shared measurement point identifier for the measurement point data that needs to be used across computing groups; Publish the measurement point data with the shared measurement point identifier to the shared topic; Other computing groups subscribe to the shared topic and write the obtained measurement point data to the mirror measurement points of the corresponding computing group.
9. The method according to claim 6, characterized in that, The process of acquiring real-time data and task status corresponding to the running node, and synchronizing the real-time data and task status with other running nodes within the same running group to obtain synchronized data, includes: Obtain the measurement point snapshot, task version, alarm status, and operation status corresponding to the fault recovery operation node from the surviving operation nodes in the operation group, and synchronize the measurement point snapshot, task version, alarm status, and operation status to the fault recovery operation node to obtain the synchronized data.
10. The method according to claim 6, characterized in that, The step of performing a status check on the synchronized data to obtain the check result includes: The version number, timestamp, quality code, data integrity, and configuration consistency in the synchronized data are verified to obtain the verification results. The service status of the running node is determined based on the verification result.
