Automatic Control Redundancy System and Method for Non-Stop Operation
The redundancy system with multiple control nodes autonomously determines roles using health scores and CRDT-based state replication to address SPOF issues, ensuring uninterrupted and stable control in industrial systems.
Patent Information
- Application Number
- KR1020260029079
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2026-02-13
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2046-02-13
AI Technical Summary
Conventional control systems face issues with Single Point of Failure (SPOF) due to reliance on a single device for Master/Slave privilege management, leading to control gaps and potential data integrity problems during abrupt role transitions, especially in network partition situations.
A redundancy system with multiple control nodes that autonomously calculate and exchange health scores using a Distributed Consensus Protocol, employing a Health-Score Based Leader Election and CRDT-based state replication to determine roles and ensure uninterrupted control, even in failure scenarios.
The system maintains uninterrupted control by predictive failure detection and adaptive path selection, minimizing control gaps and data conflicts, ensuring stable operation in diverse industrial environments.
Smart Images

Figure 112026019167095-PAT00001_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to an automatic control redundancy system and method for control target devices installed in an industrial site, and more specifically, to a redundancy system and method that performs uninterrupted automatic control by having a plurality of control nodes calculate and exchange each node's health score in real time based on a Distributed Consensus Protocol, and autonomously determine the roles of an active controller and a standby controller according to the weighted sum of the said health scores.
[0002] Specifically, in the present invention, each control node incorporates a Self Health Assessment Module to periodically calculate a health score by synthesizing multiple indicators such as CPU load, memory availability, communication latency, sensor responsiveness, and internal diagnostic results, and exchanges this score with other control nodes via a multicast method. With all control nodes sharing the same health information, the node with the highest health score is elected as the active controller by a distributed consensus algorithm, and the remaining nodes are converted to standby controllers; in this process, role determination is achieved without the need for a single external arbiter or higher authority. Background Technology
[0003] Generally, centralized control methods such as SCADA (Supervisory Control and Data Acquisition), DCS (Distributed Control System), and BEMS (Building Energy Management System) are utilized for the equipment control of various social infrastructure facilities, including buildings, smart farms, water and sewage treatment plants, gas supply facilities, transportation infrastructure, and industrial plants. While these methods are efficient for operating large-scale facilities because a top-level server or central control system communicates with each lower-level device to perform integrated control and monitoring functions, they have a structural limitation in that the top-level server becomes a Single Point of Failure (SPOF).
[0004] To address these issues, a redundancy technology based on a Single Authority approach was conventionally proposed, in which a device controller (PLC / DDC) grants, revokes, and manages Master / Slave privileges for field controllers and remote central controllers. However, this method leaves a new single point of failure problem: since the device controller acts as the sole decision-maker for privilege switching and relies on its synchronization status verification and approval procedures to perform Master / Slave transitions, the decision to switch becomes impossible if the device controller fails or communication with it is severed.
[0005] Furthermore, conventional Master-Slave redundancy methods adopt a dichotomous role structure consisting of one Master and one Slave; consequently, a Slave must be immediately promoted when the Master fails. However, this process can lead to control gaps caused by data synchronization delays or incomplete state transitions. In particular, in environments with multiple field controllers, the decision of which Slave to promote to Master relies entirely on the judgment of a single authority; therefore, delays or errors in this decision pose a risk to the uninterrupted operation of the entire system.
[0006] Furthermore, in conventional technology, synchronization integrity verification during a switchover is performed by a single device (PLC / DDC); therefore, there is a risk of incorrect switchover approval if there is an error in the device's verification logic or if the device is in an abnormal state. Consequently, a new redundancy technology is required that guarantees data integrity and control continuity while allowing multiple control nodes to autonomously and distributedly determine their roles without relying on a specific device. Prior art literature
[0007] KR 10-0279204 B1, Method for controlling controller redundancy of a field control device in an automatic control system and apparatus thereof KR 10-1566120 B1, Automatic control system for building facilities using a redundant scheduling method and method KR 10-1300743 B1, Automatic control system for building facilities using a fully redundant direct digital controller and control method thereof KR 10-2024-0141410 A, Redundant control system for governors for power plants US 9,405,644 B2, Redundant Automation System with Backup PLC US 10,002,091 B2, Master / Slave Management for Redundant Process Controller Modules US 10,073,748 B2, Failover System and Method of Deciding Master-Slave Relationship KR 10-2025-0122760 A, On-device based automatic control redundancy system 10-2902317 B1, Disaster Safety Type On-Device Automatic Control Redundancy System
[0008] ISA InTech, “Controller Redundancy Under the Hood”, 2022L. Lamport, “The Part-Time Parliament”, ACM TOCS, 1998D. Ongaro et al., "In Search of an Understandable Consensus Algorithm (Raft)", USENIX ATC, 2014 The problem to be solved
[0009] The present invention aims to solve the aforementioned conventional problems. The first objective of the present invention is to overcome the limitations of the conventional method of granting and revoking Master / Slave authority using a specific device (such as a PLC / DDC) as a single authority, and to provide an uninterrupted automatic control redundancy system in which the role determination function is maintained even in the event of a failure of any single device by having multiple control nodes autonomously calculate and exchange their respective health scores to determine the roles of active controllers and standby controllers through distributed consensus.
[0010] The second objective of the present invention is for each control node to calculate a health score in real time by synthesizing multidimensional indicators such as CPU load, memory availability, communication delay, sensor responsiveness, and internal diagnostic results, and to perform a graceful role transition based on this score, thereby resolving the control gap problem that occurred during conventional abrupt Master-Slave switching.
[0011] The third objective of the present invention is to provide a redundancy system that applies a Conflict-free Replicated Data Type (CRDT)-based state replication technique, enabling each control node to perform control independently even in a network partition situation, and then automatically merge the state without data conflicts upon network recovery.
[0012] The fourth objective of the present invention is to maximize control continuity by analyzing the downward trend of the health score before a failure occurs through a predictive failure detection function and initiating a role transition in advance, thereby maintaining a state where the transition is already completed when an actual failure occurs. means of solving the problem
[0013] To solve the above-mentioned problem, the automatic control redundancy system for uninterrupted operation according to the present invention comprises: three or more control nodes (10) that communicate with control target devices installed at an industrial site and control the operation of said control target devices, periodically calculate a health score by synthesizing a plurality of performance indicators through a self-health evaluation module, exchange the health score with other control nodes, and, by a distributed consensus protocol, elect the node with the highest health score as an active controller and switch the remaining nodes to standby controllers; a health evaluation module (20) embedded in each of the three or more control nodes (10) and calculates a health score by weighted summing a plurality of performance indicators including CPU load rate, memory availability rate, communication response delay, sensor data reception rate, and internal diagnostic score; and a distributed state replication unit (30) that replicates and merges health scores, operation data, setting values, alarm events, and log information among the three or more control nodes (10) based on CRDT (Conflict-free Replicated Data Type). It includes a multipath communication unit (40) that manages wired and wireless communication paths between the three or more control nodes (10) and distributes traffic according to path quality.
[0014] The control node (10) is implemented in at least one form among a field controller, a remote central controller, or an edge gateway, and each is equipped with complete control logic capable of independently performing automatic control and monitoring functions for the devices to be controlled. When performing the role of an active controller, the control node (10) directly transmits control commands to the devices to be controlled, and when performing the role of a standby controller, it maintains a hot standby state by reflecting operation data received from the active controller in real time.
[0015] The above health evaluation module (20) calculates a health score (H) ranging from 0 to 100 points by weighting and summing the following performance indicators as follows:
[0016] (Formula) H = w1×(100-CPU_load) + w2×MEM_avail + w3×f(COMM_delay) + w4×SENSOR_rate + w5×DIAG_score. Here, w1+w2+w3+w4+w5 = 1, and f is a function that normalizes the reciprocal of the communication delay to a range of 0 to 100. The default weights are set to w1=0.20, w2=0.15, w3=0.25, w4=0.20, and w5=0.20, and can be adjusted depending on the application field.
[0017] The above distributed consensus protocol adopts a Health-Score Based Leader Election method that is a modified version of the Raft consensus algorithm, granting priority election rights to the node with the highest health score instead of the conventional random timeout-based election. Specifically, if the health score of the current active controller falls below a preset lower threshold (Threshold_low), the standby controller with the highest health score proposes a role transition, and if a majority (N / 2+1) or more of the control nodes agree to it, the role transition is confirmed.
[0018] The distributed state replication unit (30) applies a CRDT-based collision-free data merging technique so that, in a network partition situation, each control node performs control independently, and then automatically merges the state without a separate collision resolution procedure upon network recovery. The distributed state replication unit (30) applies an LWW-Register (Last-Writer-Wins Register) to driving data, a G-Set (Grow-Only Set) to alarm events and log information, and a PN-Counter to setting values to perform a merging strategy optimized for each data type.
[0019] The multipath communication unit (40) simultaneously operates at least two paths among a first wired path (Ethernet), a second wired path (RS-485 / RS-232), and a wireless path (Wi-Fi / 5G / LTE), and includes a path quality monitoring unit (41) that monitors the latency, packet loss rate, and bandwidth of each path in real time, and an adaptive path selection unit (42) that automatically selects an optimal path according to the results of the path quality monitoring unit (41). The adaptive path selection unit (42) maximizes communication reliability by distributing traffic in a weighted round-robin manner according to path quality. Effects of the invention
[0020] According to the present invention,
[0021] First, since multiple control nodes autonomously determine their roles through distributed consensus, unlike conventional methods where a specific device (such as a PLC or DDC) approves the transition as a single authority, the role determination function is maintained even in the event of a failure of any single device, fundamentally resolving the Single Point of Failure (SPOF) problem.
[0022] Second, by real-time calculation and exchange of health scores based on multidimensional performance indicators, it is possible to detect declining health trends before failure occurs and perform predictive transitions; thus, the control gap is minimized compared to the conventional reactive failover method.
[0023] Third, through CRDT-based distributed state replication, each control node performs control independently even in network partitioned situations, and since the state is automatically merged after network recovery without a separate conflict resolution procedure, the data collision problem that occurred in conventional P2P synchronization methods is fundamentally resolved.
[0024] Fourth, through adaptive traffic distribution and path quality-based automatic selection of multipath communication, it provides improved communication continuity compared to conventional wired / wireless simple fail-over methods.
[0025] Fifth, since the transition is completed before an actual failure occurs through predictive failure detection and gradual role transition, uninterrupted automatic control is possible without operator intervention. This allows for the application of the system to various social infrastructures, such as buildings, smart farms, water and sewage treatment plants, gas supply facilities, transportation infrastructure, and industrial plants, thereby minimizing social and economic damage that may occur in the event of failures or disasters. Brief explanation of the drawing
[0026] FIG. 1 is a block diagram showing the overall configuration of an automatic control redundancy system for uninterrupted operation according to the present invention. FIG. 2 is a block diagram showing the internal configuration of the health assessment module (20) and the process of calculating the health score according to the present invention. FIG. 3 is a flowchart illustrating a distributed consensus-based role determination process according to the present invention. Figure 4 is a sequence diagram illustrating the predictive fault detection and gradual role transition process according to the present invention. FIG. 5 is a block diagram showing the operation process of a CRDT-based distributed state replication unit (30) according to the present invention. FIG. 6 is a block diagram showing the internal configuration and adaptive path selection process of a multipath communication unit (40) according to the present invention. FIG. 7 is an overall flowchart of an automatic control redundancy method for uninterrupted operation according to the present invention. Specific details for implementing the invention
[0027] Hereinafter, embodiments of the present invention will be described in more detail with reference to the attached drawings. The system according to the present invention described below is not limited to the following embodiments, and its technical spirit extends to the scope in which any person with ordinary knowledge in the relevant technical field can modify and implement it without departing from the gist of the technology claimed in the claims.
[0028] 1. Overall System Configuration
[0029] Referring to FIG. 1, the automatic control redundancy system (1) for uninterrupted operation according to the present invention comprises a plurality of control nodes (10), a health assessment module (20), a distributed state replication unit (30), a multipath communication unit (40), a predictive fault detection unit (50), and a progressive role transfer unit (60).
[0030] The plurality of control nodes (10) may include a first control node (10-1), a second control node (10-2), and a third control node (10-3), and each control node controls the operation of the control target devices (including sensors and actuators) installed at an industrial site by communicating with them via industrial protocols such as RS-485, Modbus, Ethernet / IP, and OPC-UA. The number of control nodes (10) is not limited to three, and configuring them with an odd number of two or more is advantageous for determining the majority of the distributed consensus.
[0031] Each of the above control nodes (10) can be implemented in at least one form of a PLC (Programmable Logic Controller), DDC (Direct Digital Controller), Industrial PC, or Edge Computing Gateway, and each is equipped with complete control logic and data collection logic capable of independently performing automatic control, monitoring, and HMI (Human Machine Interface) functions for the devices to be controlled.
[0032] 2. Health Assessment Module
[0033] Referring to FIG. 2, the health evaluation module (20) is embedded in each control node (10) to measure the real-time status of the node using multidimensional performance indicators and to calculate a health score (H) by synthesizing them. The health evaluation module (20) includes an indicator collection unit (21), a weighted summation unit (22), a trend analysis unit (23), and a score distribution unit (24).
[0034] The above indicator collection unit (21) collects the following performance indicators: (a) CPU load rate (CPU_load): current CPU usage rate measured in the range of 0 to 100%; (b) memory availability rate (MEM_avail): ratio of available memory; (c) communication response delay (COMM_delay): round-trip delay time for communication with other control nodes and controlled target devices; (d) sensor data reception rate (SENSOR_rate): rate of receiving data normally from connected sensors; (e) internal self-diagnosis score (DIAG_score): score obtained by combining self-diagnosis results such as watchdog timer, memory checksum, and I / O board diagnosis.
[0035] The above weighted summing unit (22) calculates the collected indicators into a health score (H) according to the following formula.
[0036] (Formula) H = w1×(100-CPU_load) + w2×MEM_avail + w3×f(COMM_delay) + w4×SENSOR_rate + w5×DIAG_score, where w1+w2+w3+w4+w5 = 1, and f is a function that normalizes the reciprocal of the communication delay to a range of 0 to 100. The default weights are set to w1=0.20, w2=0.15, w3=0.25, w4=0.20, w5=0.20, and can be adjusted according to the application field.
[0037] The above trend analysis unit (23) analyzes the moving average and slope of the health score calculated N times (e.g., 30 times) recently to detect a trend in which the health score continuously declines over a certain period. This trend information is transmitted to the predictive failure detection unit (50) and used as a trigger for a prior role transition.
[0038] The score distribution unit (24) periodically distributes the calculated health score to all other control nodes in a multicast manner.
[0039] 3. Determining Roles Based on Distributed Consensus
[0040] Referring to FIG. 3, the distributed consensus-based role determination process of the present invention is performed as follows. Each control node (10) receives the health scores of all control nodes and maintains a health table. If the health score of the current active controller falls below a lower threshold (Threshold_low) or if the health score of one of the standby controllers remains higher than the health score of the current active controller by an upper margin (Margin_high) for a set duration (e.g., 10 seconds), the standby controller proposes a role transition.
[0041] Other control nodes that receive the above role transition proposal verify the validity of the proposal by referring to their own health tables, and if they determine it is valid, they return an Agree message. When approval is received from a majority (N / 2+1) or more of the control nodes, the role transition is committed, the proposed standby controller is converted into a new active controller, and the existing active controller is converted into a standby controller.
[0042] In the event of a tie, the decision is made based on the priority of the Node ID, but the current active controller is given incumbency priority to prevent unnecessary role switching.
[0043] 4. Predictive failure detection and progressive role transition
[0044] Referring to FIG. 4, the predictive fault detection unit (50) receives information on the downward trend of the health score from the trend analysis unit (23) of the health evaluation module (20), and initiates a gradual role transition when the health score of the current active controller reaches a warning threshold (Threshold_warn). The warning threshold (Threshold_warn) is set to a value higher than the lower threshold (Threshold_low) to prepare for a role transition before an actual fault occurs.
[0045] The above-mentioned progressive role transition unit (60) performs the following three steps sequentially when the role transition is initiated: (Phase 1) Preparation phase - The node with the highest health score among the standby controllers is designated as the next active controller candidate, and the hot standby state is strengthened by fully synchronizing the latest operating data, set values, and control status with the candidate; (Phase 2) Dual execution phase - The current active controller and the candidate produce the same control command in parallel, but the actual output is performed only by the current active controller, and the candidate's control output is used only for verification to confirm control consistency. At this time, the candidate's output and the active controller's output are compared within a preset tolerance range, and must match consecutively for more than a preset number of times to proceed to Phase 3; (Phase 3) Transition confirmation phase - When the role transition is confirmed by distributed consensus, the candidate is promoted to the active controller to directly perform the control output, and the existing active controller is converted to a standby controller.
[0046] The above-mentioned gradual role transition unit (60) cancels the role transition and returns to normal operation when the health score of the current active controller recovers above a warning threshold during Phase 1 or Phase 2. This prevents unnecessary transitions and ensures operational stability.
[0047] 5. CRDT-based Distributed State Replication
[0048] Referring to FIG. 5, the distributed state replication unit (30) applies a CRDT-based collision-free data merging technique to ensure data consistency even in network partitioning situations. An optimized merging strategy is performed for each data type by applying LWW-Register to operation data, G-Set to alarm events and logs, and PN-Counter to set values.
[0049] When network partitioning occurs, each control node independently updates its local CRDT state, and upon network recovery, the distributed state replication unit (30) exchanges the CRDT states of each node and performs a type-specific merge operation so that all nodes converge to the same final state.
[0050] 6. Multipath Communication Unit
[0051] Referring to FIG. 6, the multipath communication unit (40) includes a path quality monitoring unit (41), an adaptive path selection unit (42), and a path-to-path data matching unit (43). The adaptive path selection unit (42) applies a weighted round-robin method that distributes data packets in proportion to the quality score of each communication path.
[0052] 7. Operational Examples
[0053] [First Embodiment - Health-based Autonomous Switching During Normal Operation] Three control nodes (10-1, 10-2, 10-3) are in operation, with 10-1 as the active controller and 10-2 and 10-3 as standby controllers. Due to an increase in CPU load and a decrease in memory availability of 10-1, the health score drops below the warning threshold. Since 10-2 has the highest health score, 10-2 proposes a role transition, and 10-3 approves it, thereby securing a majority. Phase 1 (synchronization enhancement), Phase 2 (dual execution verification - confirmation of output matching 5 consecutive times), and Phase 3 (transition confirmation) are sequentially performed by the gradual role transition unit (60), so that 10-2 becomes the new active controller and 10-1 switches to the standby controller.
[0054] [Second Embodiment - Independent Operation and Automatic Merging in Network Partitioning Situation] Communication between 10-1 and 10-2 / 10-3 is interrupted due to a network failure. In Partition A, where 10-1 is the active controller, 10-1 continues to perform control, while in Partition B, to which 10-2 / 10-3 belongs, 10-2 is elected as the active controller by distributed consensus and performs control. After network recovery, the distributed state replication unit (30) performs CRDT merging so that all nodes converge to the same final state.
[0055] [Third Embodiment - Predictive Pre-transition and Cancellation of Transition] A trend of continuous decline in the health score of 10-1 (active controller) is detected for 3 minutes. The current score (72 points) is higher than the lower threshold (50 points) but is close to the warning threshold (70 points), so the predictive fault detection unit (50) initiates a gradual role transition. In Phase 1, 10-2 is designated as a candidate to perform full synchronization, and in Phase 2, control consistency is verified through dual execution. If the health score of 10-1 recovers to 75 points, the transition is canceled in Phase 2 and returns to normal operation. If it does not recover and drops to 60 points, the transition is confirmed in Phase 3.
[0056] 8. Application Fields
[0057] This invention is applicable to various social infrastructures such as buildings, smart farms, water and sewage treatment plants, gas supply facilities, transportation infrastructure, and industrial plants, and ensures uninterrupted operation and stable control by having multiple control nodes autonomously determine their roles through distributed consensus even in the event of failure or disaster. Explanation of the symbols
[0058] 1 : Automatic control redundancy system 10: Control Node 10-1: First control node 10-2: Second Control Node 10-3: Third Control Node 20: Health Assessment Module 21 : Indicator Collection Unit 22 : Weighted summation part 23 : Trend Analysis Department 24 : Score Distribution Section 30: Distributed State Replica 40: Multipath communication unit 41: Path Quality Monitoring Unit 42: Adaptive Path Selection 43: Data alignment between paths 50: Predictive fault detection unit 60: Gradual role transition
Claims
Claim 1 In an automatic control redundancy system for uninterrupted operation of control target devices installed at an industrial site, three or more control nodes (10) are equipped with complete control logic capable of communicating with the control target devices and independently controlling the operation of each control target device, and periodically calculate a Health Score by weighted summing a plurality of performance indicators including CPU load rate, memory availability rate, communication response delay, sensor data reception rate, and internal self-diagnosis score, and exchange the Health Score with other control nodes in a multicast manner; A distributed state replication unit (30) that replicates driving data, setting values, alarm events, and log information between the three or more control nodes (10) based on CRDT (Conflict-free Replicated Data Type), applies a logical timestamp-based LWW-Register (Last-Writer-Wins Register) to driving data which is a real-time measurement value, applies a G-Set (Grow-Only Set) to alarm events and log information to merge them through a union operation, and applies a PN-Counter to operator setting values to determine the final value by synthesizing the increase / decrease history, thereby automatically merging the state independently updated by each control node when a network partition occurs without a conflict resolution procedure during network recovery; An automatic control redundancy system for uninterrupted operation, comprising: three or more control nodes (10), wherein the node with the highest health score is elected as the Active Controller by a distributed consensus protocol without the intervention of a single external arbitrator or higher authority, and the remaining nodes are converted into Standby Controllers, wherein the election is confirmed by the approval of a majority (N / 2+1) or more of the control nodes. Claim 2 An automatic control redundancy system for uninterrupted operation, characterized in that, in claim 1, each control node (10) incorporates a health evaluation module (20), and the health evaluation module (20) includes: an indicator collection unit (21) that collects CPU load rate, memory availability rate, communication response delay, sensor data reception rate, and internal self-diagnosis score; a weighted summation unit (22) that calculates a health score in the range of 0 to 100 points by weighted summing the indicators using the following formula; and a score distribution unit (24) that distributes the calculated health score to all other control nodes in a multicast manner (formula) H = w1×(100-CPU_load) + w2×MEM_avail + w3×f(COMM_delay) + w4×SENSOR_rate + w5×DIAG_score (where w1+w2+w3+w4+w5=1, f is a function that normalizes the reciprocal of the communication delay to 0~100). Claim 3 An automatic control redundancy system for uninterrupted operation according to claim 1, wherein the distributed consensus protocol is characterized in that when either of the following conditions is satisfied: (a) a first condition in which the health score of the current active controller falls below a preset lower threshold (Threshold_low), or (b) a second condition in which the health score of one of the standby controllers is maintained at a state higher than the health score of the current active controller by a preset upper margin (Margin_high) for a preset duration, the standby controller proposes a role transition, and if a majority of control nodes verify the validity of the proposal by referring to their own health tables and agree, the role transition is committed. Claim 4 delete Claim 5 In an automatic control redundancy system for uninterrupted operation of control target devices installed at an industrial site, the system comprises: a plurality of control nodes (10) equipped with complete control logic capable of independently controlling the operation of each of the control target devices, and periodically calculating and exchanging a health score that integrates multiple performance indicators; a trend analysis unit (23) embedded in each of the control nodes (10) that detects a downward trend in health score by analyzing the moving average and slope of the health score calculated N times recently; and a predictive failure detection unit (50) that initiates a gradual role transition before an actual failure occurs when it is determined by the trend analysis unit (23) that the current active controller's health score has reached a warning threshold (Threshold_warn) higher than a lower threshold (Threshold_low). When a role transition is initiated by the predictive fault detection unit (50) above, the system comprises: (Phase 1) a preparation step in which the node with the highest health score among the standby controllers is designated as a candidate and the operation data and control state are fully synchronized; (Phase 2) a dual execution step in which the current active controller and the candidate produce the same control command in parallel, but the actual output is performed only by the current active controller, and the control output of the candidate and the control output of the current active controller are compared and verified within a preset tolerance range, and if a discrepancy occurs, the control logic of the candidate is resynchronized or the role transition is canceled; and (Phase 3) a transition confirmation step in which the role transition is confirmed by distributed consensus and the candidate is promoted to an active controller; characterized by including a gradual role transition unit (60). Claim 6 delete Claim 7 In claim 5, the gradual role transition unit (60) is characterized by canceling the role transition and returning to normal operation when the health score of the current active controller recovers above a warning threshold (Threshold_warn) during the preparation phase (Phase 1) or dual execution phase (Phase 2). Claim 8 An automatic control redundancy system for uninterrupted operation, characterized in that, in claim 1 or 5, the system further comprises a path quality monitoring unit (41) that measures the delay time, packet loss rate, and bandwidth availability in real time for each of the multiple communication paths between the control nodes (10) and calculates a quality score for each path, and a multipath communication unit (40) that includes an adaptive path selection unit (42) of a weighted round-robin method that distributes and transmits data packets to multiple paths in proportion to the quality score for each path. Claim 9 A method for automatic control redundancy for uninterrupted operation of control target devices installed at an industrial site, comprising: (a) a health score calculation and sharing step in which each of three or more control nodes (10) periodically calculates a health score by weighted summing multiple performance indicators including CPU load rate, memory availability rate, communication response delay, sensor data reception rate, and internal self-diagnosis score, and maintains a health table by exchanging the health score with other control nodes; (b) a distributed consensus-based role determination step in which, without the intervention of a single external arbitrator or higher authority, the control node with the highest health score is elected as the active controller by a distributed consensus protocol, and the election is confirmed by the approval of a majority (N / 2+1) or more of the control nodes; (c) a redundancy operation step in which the active controller controls the control target devices, and the standby controller maintains a hot standby state by receiving real-time replication of operation data, set values, alarm events, and log information based on CRDT (Conflict-free Replicated Data Type); (d) a predictive progressive transition step for uninterrupted operation characterized by sequentially performing candidate designation and synchronization enhancement (Phase 1) when a warning threshold is reached by analyzing the downward trend of the health score of the active controller, before an actual failure occurs, and double execution verification (Phase 2) in which the control output of the candidate and the control output of the current active controller are compared and verified within a preset tolerance range, and proceed to the next step only if it is confirmed that they match continuously for more than a preset number of times, and cancel the role transition when a mismatch occurs, and transition confirmation by distributed consensus (Phase 3). Claim 10 In claim 9, when network partitioning occurs in step (c) above, control nodes within each partition independently perform the distributed consensus of step (b) above to elect an active controller for each partition and perform independent operation, and when the network is restored, LWW-Register (Last-Writer-Wins Register) for operation data, G-Set (Grow-Only Set) for alarm events and log information, and PN-Counter for operator setting values are applied respectively, and a distributed state replication unit (30) is used to automatically converge the CRDT (Conflict-free Replicated Data Type) state independently updated in each partition by type-specific merge operation. Claim 11 delete Claim 12 A computer-readable recording medium having a program recorded thereon for executing the method of any one of paragraphs 9 to 10 on a computer.
Citation Information
Patent Citations
Intelligent monitoring method and system for heating and ventilation of thousand-level dust-free workshop based on digital twinning
CN120065894A
Dual System and Dual Change Method
KR101865872B1
On-Device Redundant Automatic Control System
KR1020250144961A
AI-Based Redundant Integrated Environmental Control System for Smart Farming
KR1020250157291A
Multi-path routing method and node apparatus in tactical network
KR102167028B1
