Control system redundancy abnormity monitoring and switching control method and system

Abstract

The invention discloses a control system redundancy abnormity monitoring and switching control method and system, and the method comprises the steps: constructing a comprehensive monitoring system, collecting PLC operation state, service operation index, network state and machine performance data through a multi-source channel, receiving a user instruction, marking a core service index, and configuring an independent monitoring parameter; inputting the collected data into an AI algorithm model for preprocessing, and training a dynamic monitoring model in combination with historical data to identify system abnormality; if the main system is abnormal, fault types are classified, the severity degree is calculated through an analytic hierarchy process, a strategy library is called to select a standby system to execute millisecond-level switching, and meanwhile switching information is recorded; and after switching, carrying out secondary judgment on the fault main system and executing corresponding recovery operation, if the fault main system is not recovered, analyzing fault data through AI learning to generate a feature library, and pre-judging potential faults and matching a recovery scheme in combination with an LSTM model. According to the invention, abnormity monitoring precision and switching real-time performance are improved, automatic fault recovery and self-learning are realized, and stable operation of the control system is guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of system control technology, specifically relating to a method and system for monitoring and switching redundancy anomalies in a control system. Background Technology

[0002] In key areas such as industrial automation, intelligent manufacturing, and energy management, the control system, as the core execution unit, directly determines production efficiency, operational safety, and equipment lifespan through its operational stability. As industrial systems become more complex, integrated, and require higher real-time performance, single control systems can no longer meet the demands for zero-interruption operation. Therefore, redundant control systems have become the mainstream solution to ensure continuous operation.

[0003] Existing technologies often employ fixed threshold monitoring, which fails to distinguish between "normal high load" and "fault anomaly" in industrial scenarios. This can easily lead to misjudgments triggering invalid switchovers or missed detections causing fault escalation. Furthermore, the lack of differentiated monitoring priorities for core services results in insufficient targeted anomaly responses. Inflexible switchover strategies can easily trigger secondary faults if the backup system harbors a similar underlying fault. Slow switchover processes cannot meet the millisecond-level real-time control requirements of industrial scenarios, potentially leading to product scrapping or safety incidents. After a primary / backup switchover, the faulty primary system requires manual troubleshooting and repair, which is inefficient and consumes backup system resources. The absence of intelligent learning mechanisms, failure to record fault data, and lack of strategy optimization make similar faults prone to recurrence, resulting in poor system adaptability to complex environments. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a method and system for monitoring and switching redundancy anomalies in a control system, so as to solve or partially solve the problems mentioned in the background art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for monitoring and switching control of redundancy anomalies in a control system, comprising the following steps: A comprehensive monitoring system is constructed. The comprehensive monitoring system first establishes a multi-source data acquisition channel, and collects PLC operating status data, service operation indicators, network status data and machine performance indicators in real time through the multi-source data acquisition channel. Receive service indicator specification instructions input by users through the human-computer interaction interface, mark the corresponding indicators as core service indicators according to the service indicator specification instructions, and configure independent monitoring and scheduling priority parameters and monitoring threshold parameters for the core service indicators; The data collected through the multi-source data acquisition channel is input into the AI ​​algorithm model. The AI ​​algorithm model performs standardized preprocessing on the input data, and then trains the model based on the preprocessed data and historical normal operation datasets and historical fault datasets to establish a dynamic monitoring model. The dynamic monitoring model analyzes data characteristics in real time to identify whether there is an abnormal state in the system. If the dynamic monitoring model detects an anomaly in the main system, it triggers a system switching control process. First, the abnormal data is classified to determine the fault type and severity level. Then, a preset response strategy library is invoked. Based on the fault type and severity level, the corresponding backup system selection rules are matched from the strategy library, and the target backup system is selected according to the rules.

[0006] As a preferred scheme for monitoring and switching control of redundancy anomalies in the control system, the PLC operating status data includes the level values ​​of PLC input and output signals, signal transmission frequency, instruction completion rate of program execution, and cycle time. The service performance metrics include CPU utilization, memory usage, service response time, and process survival rate. The network status data includes network bandwidth utilization, packet loss rate, network latency, and network jitter. The machine performance indicators include equipment operating temperature, power supply voltage, operating current, and fan speed.

[0007] As a preferred solution for monitoring and switching control of redundancy anomalies in control systems, the data recording module is activated in the system switching control process to collect and store the switching time, switching reason, main system status parameters before switching and target backup system status parameters after switching in real time, forming a switching information log. After the switchover is completed, an automatic recovery process is initiated for the faulty main system. First, the faulty main system is re-determined for the fault type, and then the corresponding recovery operation is executed based on the determination result.

[0008] As a preferred solution for monitoring and switching control of redundancy anomalies in control systems, if the faulty main system still fails to return to normal after the automatic recovery process is executed, the recorded switching information log and fault data are input into the learning module of the AI ​​algorithm model. The learning module analyzes and learns the fault cause, fault impact range and repair methods to generate a fault feature library. Based on the fault feature library and real-time monitoring data, the early warning module of the AI ​​algorithm model predicts potential faults. If a potential fault risk is predicted, an early warning message is generated. The preset repair strategy library is called to match the corresponding automatic repair scheme according to the fault type and execute the automatic repair operation.

[0009] As a preferred method for monitoring and switching redundancy in control systems, the dynamic monitoring model of the AI ​​algorithm is constructed using a neural network training method based on error backpropagation. The anomaly identification and judgment formula output by the dynamic monitoring model is as follows:

[0010]

[0011]

[0012] In the formula, For the model's linear output value, The number of dimensions in the input data. For the first Weighting coefficients of the input data. For the first The standardized values ​​of the input data. For model bias terms; This represents the final anomaly detection result value output by the model. This is an activation function used to convert linear output values. Map to the interval [0,1].

[0013] As a preferred method for monitoring and switching control of redundancy anomalies in control systems, the severity level classification of faults is determined by the analytic hierarchy process (AHP) to determine the weights of evaluation indicators. After calculating the indicator weights by constructing a judgment matrix, the severity value of the fault is calculated using a weighted summation formula. The formula is:

[0014] In the formula, The number of fault evaluation indicators, For the first The weighting coefficients of each evaluation indicator, For the first Standardized values ​​for each evaluation indicator; according to The value of divides the severity of the fault into 3 levels: when At that time, it was a minor fault. The fault was classified as moderate. This is a severe fault.

[0015] As a preferred scheme for redundancy anomaly monitoring and switching control methods in control systems, an automatic recovery process is initiated for the faulty main system. First, a secondary fault type determination is performed on the faulty main system. Based on the determination result, corresponding recovery operations are executed, specifically including: If the secondary fault type is determined to be a software configuration error, the configuration file verification algorithm is called to verify the integrity and correctness of the software configuration file of the faulty main system. After locating the erroneous configuration item, the erroneous configuration item is automatically corrected based on the preset standard configuration template. If the secondary fault type is determined to be a temporary hardware fault, an initialization command is sent to the hardware control module to perform initialization operations such as power-off restart and parameter reset on the faulty hardware device, thereby completing the recovery from the temporary hardware fault.

[0016] As a preferred method for monitoring and switching control of redundancy anomalies in a control system, the early warning module of the AI ​​algorithm model adopts a fault prediction model based on LSTM, and the output layer prediction formula is:

[0017] In the formula, for The probability value of fault prediction at time 1. This is the output layer weight matrix. For the hidden layers of the LSTM model Output status at any time, This is the output layer bias vector; is the activation function used to map the hidden layer output to a probability distribution of fault occurrence.

[0018] As a preferred method for monitoring and switching control of redundancy anomalies in control systems, the learning module of the AI ​​algorithm model expands the fault feature library through fault case matching and dynamic learning optimization. The specific process includes: A fault case library is pre-built, which stores historical fault feature vectors, historical repair schemes and repair effect evaluation data corresponding to historical faults. The dimensions of the historical fault feature vectors include CPU utilization, memory usage, network latency, abnormal PLC input and output signals and machine temperature parameters when the fault occurred. When it is determined that the faulty main system has not returned to normal, feature extraction is performed on the recorded switchover information logs and fault data to generate a new fault feature vector. ; The cosine similarity algorithm is used to calculate the new fault feature vector. Compared with the feature vectors of each historical fault in the fault case library The cosine similarity is calculated using the following formula:

[0019] In the formula, For the new fault feature vector Compared with historical fault feature vectors The similarity value ranges from 1 to 1. , The closer the value is to 1, the higher the degree of matching between the two fault characteristics; The dimension of the feature vector corresponds to the number of fault features; For the new fault feature vector The The component represents the number of new faults in the first position. The specific values ​​that each feature dimension can take; Historical fault feature vector The The component represents the historical fault in the first... The specific values ​​that each feature dimension can take.

[0020] Secondly, the present invention provides a control system redundancy anomaly monitoring and switching control system, employing the aforementioned control system redundancy anomaly monitoring and switching control method, comprising: A multi-source data acquisition unit is used to construct a comprehensive monitoring system. The comprehensive monitoring system first establishes a multi-source data acquisition channel, and collects PLC operating status data, service operation indicators, network status data and machine performance indicators in real time through the multi-source data acquisition channel. The core service indicator generation unit is used to receive the service indicator specification instruction input by the user through the human-computer interaction interface, mark the corresponding indicator as the core service indicator according to the service indicator specification instruction, and configure independent monitoring and scheduling priority parameters and monitoring threshold parameters for the core service indicator. The system anomaly identification unit is used to input the data collected through the multi-source data acquisition channel into the AI ​​algorithm model. The AI ​​algorithm model performs standardized preprocessing on the input data, and then trains the model based on the preprocessed data and historical normal operation datasets and historical fault datasets to establish a dynamic monitoring model. The dynamic monitoring model analyzes data characteristics in real time to identify whether the system has an abnormal state. The system switching control unit is used to trigger the system switching control process if the dynamic monitoring model detects an anomaly in the main system. First, the abnormal data is classified to determine the fault type and severity level. Then, a preset response strategy library is called. Based on the fault type and severity level, the corresponding backup system selection rule is matched from the strategy library, and the target backup system is selected according to the rule.

[0021] As a preferred solution for monitoring and switching redundancy in the control system, the PLC operating status data in the multi-source data acquisition unit includes the level values ​​of PLC input and output signals, signal transmission frequency, instruction completion rate of program execution, and cycle time. The service performance metrics include CPU utilization, memory usage, service response time, and process survival rate. The network status data includes network bandwidth utilization, packet loss rate, network latency, and network jitter. The machine performance indicators include equipment operating temperature, power supply voltage, operating current, and fan speed.

[0022] As a preferred solution for the control system redundancy anomaly monitoring and switching control system, the system switching control unit activates a data recording module to collect and store the switching time, switching reason, main system status parameters before switching and target backup system status parameters after switching in real time, forming a switching information log. After the switchover is completed, an automatic recovery process is initiated for the faulty main system. First, the faulty main system is re-determined for the fault type, and then the corresponding recovery operation is executed based on the determination result.

[0023] As a preferred solution for monitoring and switching control systems for redundancy in control systems, in the system switching control unit, if the faulty main system still fails to return to normal after the automatic recovery process is executed, the recorded switching information log and fault data are input into the learning module of the AI ​​algorithm model. The learning module analyzes and learns the fault cause, fault impact range and repair methods to generate a fault feature library. Based on the fault feature library and real-time monitoring data, the early warning module of the AI ​​algorithm model predicts potential faults. If a potential fault risk is predicted, an early warning message is generated. The preset repair strategy library is called to match the corresponding automatic repair scheme according to the fault type and execute the automatic repair operation.

[0024] As a preferred solution for monitoring and switching redundancy in a control system, the dynamic monitoring model of the AI ​​algorithm model in the system anomaly identification unit employs a neural network training method based on error backpropagation during its construction process. The anomaly identification judgment formula output by the dynamic monitoring model is as follows:

[0025]

[0026]

[0027] In the formula, For the model's linear output value, The number of dimensions in the input data. For the first Weighting coefficients of the input data. For the first The standardized values ​​of the input data. For model bias terms; This represents the final anomaly detection result value output by the model. This is an activation function used to convert linear output values. Map to the interval [0,1].

[0028] As a preferred scheme for a redundancy anomaly monitoring and switching control system, in the system switching control unit, the severity level of the fault is classified using the analytic hierarchy process (AHP) to determine the weights of the evaluation indicators. After calculating the indicator weights by constructing a judgment matrix, the severity value of the fault is calculated using a weighted summation formula. The formula is:

[0029] In the formula, The number of fault evaluation indicators, For the first The weighting coefficients of each evaluation indicator, For the first Standardized values ​​for each evaluation indicator; according to The value of divides the severity of the fault into 3 levels: when At that time, it was a minor fault. The fault was classified as moderate. This is a severe fault.

[0030] As a preferred embodiment of the redundancy anomaly monitoring and switching control system, the system switching control unit includes: If the secondary fault type is determined to be a software configuration error, the configuration file verification algorithm is called to verify the integrity and correctness of the software configuration file of the faulty main system. After locating the erroneous configuration item, the erroneous configuration item is automatically corrected based on the preset standard configuration template. If the secondary fault type is determined to be a temporary hardware fault, an initialization command is sent to the hardware control module to perform initialization operations such as power-off restart and parameter reset on the faulty hardware device, thereby completing the recovery from the temporary hardware fault.

[0031] As a preferred solution for monitoring and switching redundancy in a control system, the early warning module of the AI ​​algorithm model in the system anomaly identification unit adopts an LSTM-based fault prediction model, and the output layer prediction formula is:

[0032] In the formula, for The probability value of fault prediction at time 1. This is the output layer weight matrix. For the hidden layers of the LSTM model Output status at any time, This is the output layer bias vector; is the activation function used to map the hidden layer output to a probability distribution of fault occurrence.

[0033] As a preferred solution for monitoring and switching redundancy in a control system, the system anomaly identification unit includes: The fault case library construction subunit is used to pre-build the fault case library. The fault case library stores historical fault feature vectors, historical repair schemes and repair effect evaluation data corresponding to historical faults. The dimensions of the historical fault feature vectors include CPU utilization, memory usage, network latency, abnormal PLC input and output signals and machine temperature parameters when the fault occurred. The fault feature vector generation subunit is used to extract features from the recorded switchover information logs and fault data to generate a new fault feature vector when the faulty main system is determined to have not returned to normal. ; The similarity matching subunit is used to calculate new fault feature vectors using the cosine similarity algorithm. Compared with the feature vectors of each historical fault in the fault case library The cosine similarity is calculated using the following formula:

[0034] In the formula, For the new fault feature vector Compared with historical fault feature vectors The similarity value ranges from 1 to 1. , The closer the value is to 1, the higher the degree of matching between the two fault characteristics; The dimension of the feature vector corresponds to the number of fault features; For the new fault feature vector The The component represents the number of new faults in the first position. The specific values ​​that each feature dimension can take; Historical fault feature vector The The component represents the historical fault in the first... The specific values ​​that each feature dimension can take.

[0035] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the control system redundancy anomaly monitoring and switching control method of the first aspect or any possible implementation thereof.

[0036] Fourthly, the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform steps in the control system redundancy anomaly monitoring and switching control method of the first aspect or any possible implementation thereof.

[0037] The beneficial effects of the technical solution provided by this invention are as follows: First, this invention collects multi-dimensional data by constructing a comprehensive monitoring system and establishes a dynamic monitoring model by combining AI algorithms. It can automatically adjust the monitoring standards according to the real-time operating status of the system, effectively distinguishing between normal high load and abnormal fault states in industrial scenarios, and avoiding misjudgment and missed judgment caused by fixed threshold monitoring. At the same time, it supports users to specify core service indicators and assign them higher monitoring priorities, ensuring accurate capture of critical service anomalies and further improving monitoring reliability.

[0038] Secondly, it can intelligently select the optimal backup system based on the type and severity of the fault, avoiding the risk of secondary faults that may be caused by traditional fixed-sequence switching; and it achieves millisecond-level switching response through technologies such as multi-threaded parallel execution, which greatly shortens the switching time, meets the real-time fault tolerance requirements of industrial scenarios, and ensures business continuity and system stability.

[0039] Third, it can automatically attempt to repair faulty machines, reducing manual intervention and improving fault handling efficiency; at the same time, AI algorithms can learn fault data, analyze fault causes and repair methods, accumulate experience and predict potential faults, issue early warnings, reduce the probability of fault recurrence, and improve system adaptability and long-term operational stability.

[0040] Fourth, the switching information recorded during the switching process can provide data support for subsequent fault analysis and system optimization; the automatic recovery function enables the faulty main system to quickly return to the redundant backup role, avoids the backup system occupying resources for a long time, and ensures the overall redundancy of the system; the self-learning mechanism reduces the cost of repeated troubleshooting of similar faults, reduces the difficulty and workload of operation and maintenance, and indirectly improves industrial production efficiency. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in this invention or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram of the redundancy anomaly monitoring and switching control method for a control system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the control system architecture for redundancy anomaly monitoring and switching provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0044] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by those skilled in the art to which this invention pertains. The terms "comprising" or "including," or similar words used in the embodiments of this invention, mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects.

[0045] Existing redundant control systems mostly employ a fixed threshold monitoring mode, which sets static thresholds for PLC operating status, service indicators, network parameters, etc., and determines an anomaly when these thresholds are exceeded. However, in industrial scenarios, system loads fluctuate dynamically. For example, during peak production periods, CPU utilization may temporarily increase, or network bandwidth may be temporarily saturated. Static thresholds cannot distinguish between "normal high load" and "fault anomalies," leading to the following problems: Normal high loads may be misjudged as faults, frequently initiating master-slave switchover processes, wasting system resources and potentially affecting production continuity due to momentary data interruptions during the switchover process; the sensitivity to monitoring gradual faults is insufficient, failing to identify potential risks in advance, resulting in equipment downtime or data loss by the time the fault becomes apparent. Furthermore, existing systems lack differentiated monitoring priorities for "core services" and "non-core services." When multiple indicators fluctuate simultaneously, they cannot prioritize core services affecting critical production links, resulting in insufficient targeted anomaly responses.

[0046] Current redundancy switching methods mostly employ "fixed backup system" or "simple polling" strategies, failing to dynamically select the optimal backup system based on fault type and severity. If the preset backup system happens to have a latent fault originating from the same source as the primary system, switching to the backup system when the primary system fails can lead to a "secondary fault," expanding the scope of downtime. Traditional switching processes require "fault detection, manual confirmation, and manual triggering of switching" or "single-threaded step-by-step execution," with switching response times often in the hundreds of milliseconds or even seconds, which cannot meet the millisecond-level real-time control requirements of industrial scenarios and can easily cause product scrapping or safety accidents.

[0047] Existing redundant control systems, after completing the primary / backup switchover, largely rely on manual troubleshooting and repair of the faulty primary system, lacking automated recovery and intelligent learning mechanisms. The faulty primary system requires manual fault location, resulting in long repair cycles. This causes the backup system to continuously perform primary system functions, unable to return to its redundant backup role, leading to a decline in overall system redundancy. Furthermore, the system does not record fault data and cannot optimize monitoring and switching strategies using historical fault data. When faced with similar faults, it still needs to repeatedly execute inefficient troubleshooting and switching processes, resulting in poor system adaptability to complex industrial environments.

[0048] In summary, the technical deficiencies of current redundant control systems in terms of anomaly monitoring accuracy, switching control timeliness and reliability, and fault recovery and learning capabilities have become key bottlenecks restricting the development of industrial systems towards higher stability and intelligence. Therefore, there is an urgent need for a redundant anomaly monitoring and switching control scheme capable of dynamic and accurate monitoring, intelligent and rapid switching, automatic recovery, and self-learning to solve the aforementioned technical problems. The following are the specific details of the embodiments of the invention.

[0049] See Figure 1 This invention provides a method for monitoring and switching control of redundancy anomalies in a control system, comprising the following steps: S1. Construct a comprehensive monitoring system. This system first establishes multi-source data acquisition channels to collect PLC operating status data, service operation indicators, network status data, and machine performance indicators in real time. The multi-source data acquisition channels adopt a distributed data acquisition architecture, deploying data acquisition terminals at nodes such as PLC devices, servers, network switches, and hardware sensors to achieve parallel acquisition of multi-dimensional data. This architecture is based on industrial Ethernet and edge computing technology, allowing acquisition terminals to process raw data locally, reducing data transmission latency and ensuring data real-time performance.

[0050] S2. Receive service indicator specification instructions input by the user through the human-computer interaction interface. Mark the corresponding indicators as core service indicators according to the instructions, and configure independent monitoring scheduling priority parameters and monitoring threshold parameters for the core service indicators. The human-computer interaction interface provides a visual operation entry point, allowing users to select core service indicators based on business importance. The system has a built-in indicator management module that allocates independent scheduling threads and cache resources to the marked core service indicators, ensuring that core indicator data is processed first during monitoring and improving response speed. Simultaneously, the independent monitoring threshold parameters support differentiated configuration, avoiding resource waste and insufficient monitoring targeting through a threshold layering mechanism.

[0051] S3. The data collected through the multi-source data acquisition channels is input into the AI ​​algorithm model. The AI ​​algorithm model performs standardized preprocessing on the input data, and then trains the model based on the preprocessed data and historical normal operation datasets and historical fault datasets to establish a dynamic monitoring model. The dynamic monitoring model analyzes data characteristics in real time to identify whether there are abnormal states in the system. The data standardization preprocessing adopts the Min-Max standardization method, which maps raw data of different magnitudes and units to the [0,1] interval, eliminating the interference of data magnitude differences on model training and improving the model convergence speed and analysis accuracy. The AI ​​algorithm model is built based on a neural network with backpropagation of error. It learns the characteristic patterns of stable system operation through historical normal data and learns the typical characteristics of various faults through historical fault data. During the training process, the network weights and bias terms are dynamically adjusted to form a dynamic monitoring model that adapts to the system's operating rules.

[0052] S4. If the dynamic monitoring model detects an anomaly in the main system, it triggers a system switching control process. First, the abnormal data is classified to determine the fault type and severity level. Then, a preset response strategy library is invoked. Based on the fault type and severity level, the corresponding backup system selection rules are matched from the strategy library, and a target backup system is selected according to these rules. The abnormal data classification is based on a feature matching algorithm, extracting key features from the abnormal data and comparing them with a preset fault type feature library to achieve rapid fault type determination. The fault severity level classification uses the analytic hierarchy process (AHP). A judgment matrix is ​​constructed to quantify the weights of each evaluation indicator, and a weighted summation formula is used to calculate the severity value, achieving objective level classification. The preset response strategy library stores the backup system selection logic corresponding to different fault types and levels. The matching process is executed quickly through a rule engine to ensure the rationality and efficiency of the target backup system selection.

[0053] In this embodiment, the PLC operating status data includes the level values ​​of PLC input and output signals, signal transmission frequency, instruction completion rate of program execution, and cycle time; the service operation indicators include CPU utilization, memory usage, service response time, and process survival rate; the network status data includes network bandwidth utilization, data packet loss rate, network latency, and network jitter value; and the machine performance indicators include equipment operating temperature, power supply voltage, operating current, and fan speed.

[0054] Specifically, PLC operating status data directly reflects the operation of the control core. Input / output signal levels and transmission frequencies are used to determine if signal transmission is normal, while instruction completion rate and cycle time are used to evaluate PLC program execution efficiency. Service operation indicators are core monitoring criteria at the software level; CPU utilization and memory usage reflect server resource load, while service response time and process survival rate reflect service availability. Network status data affects data transmission reliability; bandwidth utilization, packet loss rate, latency, and jitter directly relate to the communication quality between the primary and backup systems and the timeliness of control command transmission. Machine performance indicators focus on the physical state of hardware devices; temperature, voltage, current, and fan speed are used to predict hardware failures. The selection of various indicators covers the software, hardware, and communication aspects of the control system, ensuring comprehensive anomaly monitoring.

[0055] In this embodiment, during the system switching control process, the data recording module is activated to collect and store the switching time, switching reason, main system status parameters before switching, and target backup system status parameters after switching in real time, forming a switching information log. After the switching is completed, an automatic recovery process is initiated for the faulty main system. First, the faulty main system is subjected to a secondary fault type determination, and the corresponding recovery operation is performed based on the determination result.

[0056] Specifically, the data recording module employs structured log storage technology, classifying and storing switch-related parameters according to timestamps, fault information, and system status, supporting subsequent rapid retrieval and analysis. Industrial-grade solid-state drives can be used as the storage medium to ensure no log data loss. A secondary fault determination before the automatic recovery process begins, based on the initial determination and supplementary data from the faulty main system after the switchover, further improves the accuracy of fault type identification, preventing recovery operation failure due to a single misjudgment. The recovery operation uses a categorized handling logic, calling corresponding repair algorithms and instructions for different fault types, ensuring the targeted and efficient nature of the recovery operation and reducing manual intervention costs.

[0057] In this embodiment, if the main system still fails to return to normal after the automatic recovery process is executed, the recorded switching information log and fault data are input into the learning module of the AI ​​algorithm model. The learning module analyzes and learns the cause of the fault, the scope of the fault's impact, and the repair method to generate a fault feature library.

[0058] Specifically, the AI ​​algorithm model's learning module integrates case-based reasoning and reinforcement learning techniques to extract features from switch information logs and fault data, forming structured fault samples. By analyzing the causal relationships within these samples—such as service interruptions due to configuration errors or performance degradation caused by hardware aging—the core causes and patterns of fault occurrence are uncovered. Simultaneously, corresponding repair attempts and their results are correlated, forming comprehensive fault handling knowledge. The fault feature library is stored in a vector database, transforming fault features into high-dimensional vectors to support rapid similarity matching, providing data support for the rapid handling of similar faults. The learning module continuously accumulates fault samples, constantly optimizing the fault feature extraction algorithm and knowledge association logic to enhance the system's ability to recognize complex faults.

[0059] Specifically, based on the fault feature library and real-time monitoring data, the early warning module of the AI ​​algorithm model makes a prediction of potential faults. If a potential fault risk is predicted, an early warning message is generated. The preset repair strategy library is called to match the corresponding automatic repair scheme according to the fault type and execute the automatic repair operation.

[0060] Specifically, the early warning module employs an LSTM-based fault prediction model. LSTM models possess time-series data processing capabilities, enabling them to capture trend changes in monitored data. The model performs time-series correlation analysis between real-time monitoring data and historical data in a fault feature library. The output layer's softmax activation function outputs the probability distribution of fault occurrence. When the probability of a "fault occurrence" category reaches a preset threshold, an early warning message is generated. This message, containing the predicted fault type, potential impact range, and suggested handling measures, is pushed to the maintenance terminal and human-machine interface via the industrial control network. A preset repair strategy library stores standardized repair solutions for various faults. The matching process is based on fault type and severity. Repair operations are remotely commanded to the faulty equipment, proactively preventing fault occurrence or mitigating its impact.

[0061] In this embodiment, the dynamic monitoring model construction process of the AI ​​algorithm model adopts a neural network training method based on error backpropagation, and the anomaly identification judgment formula output by the dynamic monitoring model is:

[0062]

[0063]

[0064] In the formula, For the model's linear output value, The number of dimensions in the input data. For the first Weighting coefficients of the input data. For the first The standardized values ​​of the input data. For model bias terms; This represents the final anomaly detection result value output by the model. This is an activation function used to convert linear output values. Mapped to the [0,1] interval. The neural network training method using error backpropagation iteratively calculates the output through forward propagation and corrects the parameters through backpropagation, allowing the model to gradually learn data features. In forward propagation, the linear output value S is obtained by weighted summation of the input data in each dimension with their corresponding weights, plus a bias term, reflecting the linear correlation between the input data and the model features; the sigmoid activation function maps the linear output S to the [0,1] interval, giving the output value Y probabilistic meaning, facilitating the explicit identification of abnormal states through thresholding. Weight coefficients The magnitude reflects the degree of influence of the corresponding input data on anomaly detection, and is dynamically adjusted based on the prediction error during training. The value of b allows the model to be continuously optimized, improving the accuracy of anomaly detection.

[0065] In one possible embodiment, the severity level classification of faults uses the analytic hierarchy process (AHP) to determine the weights of the evaluation indicators. After calculating the indicator weights by constructing a judgment matrix, the severity value of the fault is calculated using a weighted summation formula. The formula is:

[0066] In the formula, The number of fault evaluation indicators, For the first The weighting coefficients of each evaluation indicator, For the first Standardized values ​​for each evaluation indicator; based on The value of divides the severity of the fault into 3 levels: when At that time, it was a minor fault. The fault was classified as moderate. This is a severe fault.

[0067] Specifically, the Analytic Hierarchy Process (AHP) decomposes the assessment of fault severity into a hierarchical structure consisting of a target layer (severity level), a criterion layer (evaluation indicators), and a solution layer (indicator values). It quantifies the relative importance of different evaluation indicators by constructing a judgment matrix, and then ensures the rationality of weight allocation through consistency checks, ultimately obtaining normalized weight coefficients. Standardized values ​​for evaluation indicators The original indicator values ​​are mapped to the [0,10] interval to ensure a consistent value range for different indicators and avoid the weighted summation result being affected by differences in indicator magnitude. The weighted summation formula uses weight coefficients... Highlighting the impact of key indicators, the calculations obtained L The value objectively reflects the severity of the fault, providing a quantitative basis for the selection of switching strategies and repair solutions.

[0068] In one possible embodiment, an automatic recovery process is initiated for the faulty main system. First, a secondary fault type determination is performed on the faulty main system. Based on the determination result, corresponding recovery operations are executed. Specifically, if the secondary fault type is determined to be a software configuration error, a configuration file verification algorithm is invoked to verify the integrity and correctness of the software configuration file of the faulty main system. After locating the erroneous configuration item, the erroneous configuration item is automatically corrected based on a preset standard configuration template. If the secondary fault type is determined to be a temporary hardware fault, an initialization command is sent to the hardware control module to perform initialization operations such as power-off restart and parameter reset on the faulty hardware device, thereby completing the temporary hardware fault recovery.

[0069] Specifically, software configuration error recovery is based on a configuration file verification algorithm. This algorithm combines hash verification and syntax parsing. First, it verifies the integrity of the configuration file by comparing hash values, and then uses a syntax parser to check the format and logical rationality of configuration parameters to locate erroneous configuration items. A standard configuration template stores the optimal configuration parameters for this type of system. During the correction process, erroneous configuration items are directly replaced to ensure the configuration file is restored to a normal state. Recovery from temporary hardware failures is based on the instruction interaction capabilities of the hardware control module. A power-off restart operation cuts off and restores power supply to the power module, clearing temporary hardware caches and abnormal states. A parameter reset operation restores the hardware's operating parameters to factory default values ​​or optimal operating parameters, resolving temporary failures caused by parameter drift and avoiding unnecessary hardware replacement.

[0070] In one possible embodiment, the early warning module of the AI ​​algorithm model adopts an LSTM-based fault prediction model, and the output layer prediction formula is:

[0071] In the formula, for The probability value of fault prediction at time 1. This is the output layer weight matrix. For the hidden layers of the LSTM model Output status at any time, This is the output layer bias vector; is the activation function used to map the hidden layer output to a probability distribution of fault occurrence.

[0072] Specifically, the LSTM model, through its gating mechanism of input gate, forget gate, and output gate, can effectively capture long-term dependencies in time-series data, making it suitable for handling the trend characteristics of system operation data over time. Hidden layer output state. It integrates current input data with historical state information to reflect the dynamic changes in the system's operating state. Output layer weight matrix. With bias vector The hidden layer states are mapped to the output space. The softmax activation function transforms the mapping result into a probability distribution for each fault type (and normal state), making... It possesses a clear probabilistic meaning. Through analysis... The probability values ​​of each fault type can predict possible system faults in the future, providing a basis for taking preventive measures in advance.

[0073] In one possible embodiment, the learning module of the AI ​​algorithm model expands the fault feature library through fault case matching and dynamic learning optimization. The specific process includes: pre-constructing a fault case library, which stores historical fault feature vectors, historical repair schemes, and repair effect evaluation data corresponding to historical faults. The dimensions of the historical fault feature vectors include CPU usage, memory usage, network latency, PLC input / output signal anomaly types, and machine temperature parameters at the time of the fault. When it is determined that the main fault system has not returned to normal, features are extracted from the recorded switching information logs and fault data to generate a new fault feature vector A. The cosine similarity algorithm is used to calculate the similarity between the new fault feature vector A and each historical fault feature vector B in the fault case library. The cosine similarity calculation formula is:

[0074] In the formula, For the new fault feature vector Compared with historical fault feature vectors The similarity value ranges from 1 to 1. , The closer the value is to 1, the higher the degree of matching between the two fault characteristics; The dimension of the feature vector corresponds to the number of fault features; For the new fault feature vector The The component represents the number of new faults in the first position. The specific values ​​that each feature dimension can take; Historical fault feature vector The The component represents the historical fault in the first... The specific values ​​that each feature dimension can take.

[0075] Specifically, the fault case library uses a structured storage method, transforming the multi-dimensional features of historical faults into feature vectors of a unified dimension, while simultaneously associating corresponding repair solutions and effect data to form complete fault handling knowledge units. New fault feature vectors AThe generation process uses feature extraction algorithms to filter key features from switchover information logs and fault data, ensuring that the vectors accurately represent the core attributes of the new fault. The cosine similarity algorithm calculates the cosine of the angle between two vectors to quantify their directional consistency, thereby determining the degree of similarity between the new fault and historical faults. For example, when... When the value is ≥0.85, it indicates that the two fault characteristics are highly matched, and historical repair solutions can be directly reused to improve fault handling efficiency; when When the similarity score is less than 0.85, it is identified as a novel fault. Its feature vector, processing procedure, and repair plan are added to the case library. At the same time, the feature dimension weights are updated through reinforcement learning algorithm to optimize the similarity matching accuracy, thereby realizing the dynamic expansion of the fault case library and the continuous improvement of the system's learning ability.

[0076] The application scenarios of this invention are as follows: Scenario 1: Industrial Automation Production Line Control System: In continuous industrial automation production lines such as automotive welding and electronic component placement, the control system needs to drive robotic arms, conveyors, and testing equipment to work collaboratively in real time. Any interruption can lead to production line shutdown and product scrapping. This invention can collect PLC operating status, service operation indicators, and machine performance indicators in real time through multi-source data acquisition channels. An AI dynamic monitoring model accurately distinguishes between "normal high load" and "fault / abnormal" conditions. When a main system anomaly is identified, a backup system is quickly matched based on the severity of the fault, achieving millisecond-level switching and preventing production line downtime. Simultaneously, the faulty main system is automatically restored, and similar fault data is accumulated through an AI learning module to optimize subsequent monitoring accuracy, ensuring continuous and stable operation of the production line.

[0077] Scenario 2: Smart Grid Power Dispatch and Control System: In a smart grid, the power dispatch and control system is responsible for real-time allocation of power resources and monitoring of substation equipment status. It must meet millisecond-level response and high reliability requirements; otherwise, regional power outages may occur. This invention can collect network status data (such as network latency and packet loss rate between the dispatch center and substations), service operation indicators (such as memory usage and process survival rate of power load calculation services), and PLC operation status (such as substation switch control signals). Core service indicators such as power load dispatch and switch status control are marked as high priority, and stricter monitoring thresholds are configured. When the main dispatch system triggers a fault due to network jitter or service anomalies, the severity of the fault is quickly calculated using the analytic hierarchy process (AHP). A strategy library is invoked to select a backup dispatch system with sufficient resources for millisecond-level switching, ensuring uninterrupted power dispatch. For the faulty main system (if determined to be a software configuration error), it is automatically corrected based on a standard configuration template, and after recovery, it returns to its redundant role. Simultaneously, the LSTM early warning module predicts potential network latency risks based on historical fault data and pushes early warning information in advance, ensuring the safety of power grid dispatch.

[0078] Scenario 3: High-end CNC Machine Tool Control System: High-end CNC machine tools are used for precision parts machining, requiring extremely high real-time performance and stability from the control system. Even minor control delays or faults can lead to substandard part accuracy. This invention can collect machine performance indicators (such as machine tool spindle temperature and feed axis current), PLC operating status (such as spindle speed control signals and program instruction completion rate), and service operation indicators (such as the response time of CNC machining code parsing service) in real time to build a dynamic monitoring model. When the main control system experiences an abnormal increase in spindle temperature (determined as a minor fault) or an interruption in code parsing service (determined as a major fault), it quickly matches the corresponding backup system: a backup system of the same precision is selected for minor faults, and a backup system with higher redundancy is selected for major faults. Millisecond-level switching is achieved through multi-threaded parallel execution to avoid machining interruptions. During the recovery phase of the faulty main system, if it is a temporary hardware fault, an initialization command is sent to reset parameters. After recovery, the system is rejoined in the redundancy system. The AI ​​learning module synchronously updates the fault feature library, improving the monitoring and prediction capabilities of machine tool sensor faults and ensuring precision machining accuracy.

[0079] Scenario 4: Urban Rail Transit Signal Control System: Urban rail transit signal control systems need to control train intervals and station stopping accuracy in real time to ensure safe and efficient train operation. System failures may lead to train delays or operational interruptions. This invention can collect network status data (such as the communication bandwidth utilization rate between the signal control center and the train, and network jitter value), PLC operating status (such as turnout switching control signals and train positioning signals), and service operation indicators (such as the CPU utilization rate of train dispatching services), marking core services such as turnout control and train positioning as high priority. When the main signal system experiences an abnormal turnout switching signal (determined as a severe fault), it quickly calls the strategy library to select a backup signal system without the same source fault, performs a millisecond-level switch, and ensures normal turnout switching; if the secondary determination of the faulty main system is a temporary hardware fault, it performs a power-off restart initialization operation, and after recovery, it serves as a backup system; the AI ​​learning module analyzes the fault data, generates a turnout sensor fault feature library, and uses the LSTM early warning module to predict the risk of similar faults, notifying maintenance personnel in advance to replace aging sensors and reduce the probability of operational interruptions.

[0080] It should be noted that the method of this embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this embodiment, and these multiple devices will interact with each other to complete the control system redundancy anomaly monitoring and switching control method.

[0081] It should be noted that the above description describes some embodiments of the present invention. In some cases, the described actions or steps can be performed in a different order than that shown in the above embodiments and the desired result can still be achieved. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] See Figure 2 Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this invention also provides a control system redundancy anomaly monitoring and switching control system, employing a control system redundancy anomaly monitoring and switching control method of the above embodiments, including: The multi-source data acquisition unit 100 is used to construct a comprehensive monitoring system. The comprehensive monitoring system first establishes a multi-source data acquisition channel, and collects PLC operating status data, service operation indicators, network status data and machine performance indicators in real time through the multi-source data acquisition channel. The core service indicator generation unit 200 is used to receive the service indicator specification instruction input by the user through the human-computer interaction interface, mark the corresponding indicator as the core service indicator according to the service indicator specification instruction, and configure independent monitoring and scheduling priority parameters and monitoring threshold parameters for the core service indicator. The system anomaly identification unit 300 is used to input the data collected through the multi-source data acquisition channel into the AI ​​algorithm model. The AI ​​algorithm model performs standardized preprocessing on the input data, and then trains the model based on the preprocessed data and historical normal operation datasets and historical fault datasets to establish a dynamic monitoring model. The dynamic monitoring model analyzes data characteristics in real time to identify whether the system has an abnormal state. The system switching control unit 400 is used to trigger the system switching control process if the dynamic monitoring model identifies an anomaly in the main system. First, the abnormal data is classified to determine the fault type and severity level. Then, a preset response strategy library is called. According to the fault type and severity level, the corresponding backup system selection rule is matched from the strategy library, and the target backup system is selected according to the rule.

[0083] In this embodiment, the PLC operating status data in the multi-source data acquisition unit 100 includes the level value of PLC input and output signals, signal transmission frequency, instruction completion rate of program execution, and cycle time. The service performance metrics include CPU utilization, memory usage, service response time, and process survival rate. The network status data includes network bandwidth utilization, packet loss rate, network latency, and network jitter. The machine performance indicators include equipment operating temperature, power supply voltage, operating current, and fan speed.

[0084] In this embodiment, the system switching control unit 400 activates the data recording module to collect and store the switching time, switching reason, main system status parameters before switching and target backup system status parameters after switching in real time, forming a switching information log; After the switchover is completed, an automatic recovery process is initiated for the faulty main system. First, the faulty main system is re-determined for the fault type, and then the corresponding recovery operation is executed based on the determination result.

[0085] In this embodiment, if the faulty main system still fails to return to normal after the automatic recovery process is executed, the recorded switching information log and fault data are input into the learning module of the AI ​​algorithm model in the system switching control unit 400. The learning module analyzes and learns the fault cause, fault impact range and repair method to generate a fault feature library. Based on the fault feature library and real-time monitoring data, the early warning module of the AI ​​algorithm model predicts potential faults. If a potential fault risk is predicted, an early warning message is generated. The preset repair strategy library is called to match the corresponding automatic repair scheme according to the fault type and execute the automatic repair operation.

[0086] In this embodiment, the dynamic monitoring model construction process of the AI ​​algorithm model in the system anomaly identification unit 300 adopts a neural network training method based on error backpropagation, and the anomaly identification judgment formula output by the dynamic monitoring model is:

[0087]

[0088]

[0089] In the formula, For the model's linear output value, The number of dimensions in the input data. For the first Weighting coefficients of the input data. For the first The standardized values ​​of the input data. For model bias terms; This represents the final anomaly detection result value output by the model. This is an activation function used to convert linear output values. Map to the interval [0,1].

[0090] In one possible embodiment, in the system switching control unit 400, the severity level classification of faults is determined by the analytic hierarchy process (AHP) to determine the weights of the evaluation indicators. After calculating the indicator weights by constructing a judgment matrix, the severity value of the fault is calculated using a weighted summation formula. The formula is:

[0091] In the formula, The number of fault evaluation indicators, For the first The weighting coefficients of each evaluation indicator, For the first Standardized values ​​for each evaluation indicator; according to The value of divides the severity of the fault into 3 levels: when At that time, it was a minor fault. The fault was classified as moderate. This is a severe fault.

[0092] In one possible embodiment, the system switching control unit 400 includes: If the secondary fault type is determined to be a software configuration error, the configuration file verification algorithm is called to verify the integrity and correctness of the software configuration file of the faulty main system. After locating the erroneous configuration item, the erroneous configuration item is automatically corrected based on the preset standard configuration template. If the secondary fault type is determined to be a temporary hardware fault, an initialization command is sent to the hardware control module to perform initialization operations such as power-off restart and parameter reset on the faulty hardware device, thereby completing the recovery from the temporary hardware fault.

[0093] In one possible embodiment, in the system anomaly identification unit 300, the early warning module of the AI ​​algorithm model adopts an LSTM-based fault prediction model, and the output layer prediction formula is:

[0094] In the formula, for The probability value of fault prediction at time 1. This is the output layer weight matrix. For the hidden layers of the LSTM model Output status at any time, This is the output layer bias vector; is the activation function used to map the hidden layer output to a probability distribution of fault occurrence.

[0095] In one possible embodiment, the system anomaly identification unit 300 includes: The fault case library construction subunit 301 is used to pre-build the fault case library. The fault case library stores historical fault feature vectors, historical repair schemes and repair effect evaluation data corresponding to historical faults. The dimensions of the historical fault feature vectors include CPU utilization, memory usage, network latency, abnormal PLC input and output signals and machine temperature parameters when the fault occurred. The fault feature vector generation subunit 302 is used to extract features from the recorded switchover information logs and fault data to generate a new fault feature vector when it is determined that the faulty main system has not returned to normal. ; Similarity matching subunit 303 is used to calculate new fault feature vectors using a cosine similarity algorithm. Compared with the feature vectors of each historical fault in the fault case library The cosine similarity is calculated using the following formula:

[0096] In the formula, For the new fault feature vector Compared with historical fault feature vectors The similarity value ranges from 1 to 1. , The closer the value is to 1, the higher the degree of matching between the two fault characteristics; The dimension of the feature vector corresponds to the number of fault features; For the new fault feature vector The The component represents the number of new faults in the first position. The specific values ​​that each feature dimension can take; Historical fault feature vector The The component represents the historical fault in the first... The specific values ​​that each feature dimension can take.

[0097] The system described in the above embodiments is used to implement a corresponding control system redundancy anomaly monitoring and switching control method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0098] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a control system redundancy anomaly monitoring and switching control method as described in any of the above embodiments.

[0099] Figure 3This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 510, a memory 520, an input / output interface 530, a communication interface 540, and a bus 550. The processor 510, memory 520, input / output interface 530, and communication interface 540 are interconnected internally via the bus 550.

[0100] The processor 510 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0101] The memory 520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 520 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 520 and is called and executed by the processor 510.

[0102] Input / output interface 530 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0103] The communication interface 540 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0104] Bus 550 includes a pathway for transmitting information between various components of the device, such as processor 510, memory 520, input / output interface 530, and communication interface 540.

[0105] It should be noted that although the above-described device only shows the processor 510, memory 520, input / output interface 530, communication interface 540, and bus 550, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0106] The electronic devices described above are used to implement a corresponding redundancy anomaly monitoring and switching control method for a control system in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0107] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions, which are used to cause the computer to execute a control system redundancy anomaly monitoring and switching control method as described in any of the above embodiments.

[0108] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0109] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute a control system redundancy anomaly monitoring and switching control method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0110] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the embodiments of the invention as described above, which are not provided in detail for the sake of brevity.

[0111] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of the invention, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of the invention, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of the invention will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of the invention, it will be apparent to those skilled in the art that the embodiments of the invention may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0112] Although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAMDRAM) may be used with the embodiments discussed.

[0113] The embodiments of this invention are intended to cover all such substitutions, modifications, and variations falling within the scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this invention should be included within the scope of protection of this invention.

Claims

1. A method for monitoring and switching redundancy anomalies in a control system, characterized in that, The method comprises the following steps: An integrated monitoring system is constructed, which first establishes a multi-source data collection channel, through which PLC operation state data, service operation indicators, network state data and machine performance indicators are collected in real time; A service indicator designation instruction input by a user through a man-machine interface is received, and corresponding indicators are marked as core service indicators according to the service indicator designation instruction, and independent monitoring and scheduling priority parameters and monitoring threshold parameters are configured for the core service indicators; The data collected through the multi-source data collection channel are input into an AI algorithm model, the AI algorithm model performs standardization preprocessing on the input data, and then performs model training based on the preprocessed data and historical normal operation data sets and historical fault data sets, establishes a dynamic monitoring model, analyzes data features in real time through the dynamic monitoring model, and identifies whether the system has an abnormal state; If the dynamic monitoring model identifies that the main system has an abnormality, a system switching control process is triggered, the abnormal data are first classified to determine the fault type and severity level, a preset coping strategy library is then called, corresponding backup system selection rules are matched from the strategy library according to the fault type and severity level, and a target backup system is selected according to the rules.

2. The method of claim 1, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, The PLC operation state data include the level value, signal transmission frequency and instruction completion rate of program execution of PLC input and output signals, and cycle time consumption; The service operation indicators include CPU usage, memory occupancy, service response time and process survival rate; The network state data include network bandwidth utilization, data packet loss rate, network delay and network jitter value; The machine performance indicators include equipment operating temperature, power supply voltage, working current and fan speed.

3. The method of claim 1, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, In the system switching control process, a data recording module is started, and switching time, switching reason, main system state parameters before switching and target backup system state parameters after switching are collected and stored in real time to form a switching information log; After switching is completed, an automatic recovery process is started for the fault main system, the fault type of the fault main system is determined again, and corresponding recovery operations are performed according to the determination result.

4. The method of claim 3, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, If the fault main system is still not restored to normal after the automatic recovery process is performed, the recorded switching information log and fault data are input into a learning module of the AI algorithm model, the learning module is used to analyze and learn the fault cause, fault influence range and repair method, and a fault feature library is generated; Based on the fault feature library and real-time monitoring data, a potential fault is predicted through a warning module of the AI algorithm model, and warning information is generated if a potential fault risk is predicted; A preset repair strategy library is called, a corresponding automatic repair scheme is matched according to the fault type, and an automatic repair operation is performed.

5. The method of claim 4, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, The dynamic monitoring model of the AI algorithm model adopts a neural network training method based on error back propagation, and the abnormality identification determination formula output by the dynamic monitoring model is: ; ; ; wherein, is a linear output value of the model, is a number of dimensions of input data, is a weight coefficient of the i-th dimensional input data, is a normalized value of the i-th dimensional input data, is a bias term of the model; is an abnormality determination result value of the final output of the model; is an activation function for mapping the linear output value to the interval [0, 1].

6. The method of claim 1, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, The severity level of a fault is determined using the analytic hierarchy process (AHP) to identify the weights of the evaluation indicators. After constructing a judgment matrix to calculate the indicator weights, a weighted summation formula is used to calculate the severity value of the fault. The formula is: ; In the formula, is the number of fault evaluation indexes, is the weight coefficient of the evaluation index, is the normalized value of the evaluation index; According to the values of the fault severity is divided into 3 levels: when it is a light fault, when it is a moderate fault, and when it is a severe fault.

7. The method of claim 3, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, An automatic recovery process is started for the fault main system, the fault type of the fault main system is determined again, and corresponding recovery operations are performed according to the determination result, which specifically includes: If the secondary determination fault type is a software configuration error, a configuration file verification algorithm is called to verify the integrity and correctness of the software configuration file of the fault main system, locate the error configuration item, and automatically correct the error configuration item based on a preset standard configuration template; If the secondary determination fault type is a hardware temporary fault, an initialization instruction is sent to the hardware control module to perform power-off restart and parameter reset initialization operations on the fault hardware device to complete the hardware temporary fault recovery.

8. The method of claim 4, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, The early warning module of the AI algorithm model adopts a fault prediction model based on LSTM, and the output layer prediction formula is: ; wherein, is the failure prediction probability value at time instant, is the output layer weight matrix, is the output state of the LSTM model hidden layer at time instant, is the output state of the LSTM model hidden layer at time instant, is the output layer bias vector; is an activation function used to map the hidden layer output to a probability distribution of failure occurrence.

9. The method of claim 4, wherein the control system redundancy anomaly monitoring and switchover control method is characterized by, The learning module of the AI algorithm model realizes fault feature library expansion through fault case matching and dynamic learning optimization, and the specific process includes: A fault case library is constructed in advance, which stores historical fault feature vectors, historical repair schemes and repair effect evaluation data corresponding to historical faults, wherein the dimensions of the historical fault feature vectors include CPU usage, memory occupancy, network delay, PLC input and output signal abnormal type and machine temperature parameters when the fault occurs; When it is determined that the main system in failure is not restored to normal, the recorded switching information log and the failure data are feature-extracted to generate a new failure feature vector ; The cosine similarity algorithm is used to calculate a new fault feature vector The cosine similarity algorithm is used to calculate a new fault feature vector The cosine similarity algorithm is used to calculate a new fault feature vector ; In the formula, For the new fault feature vector Compared with historical fault feature vectors The similarity value ranges from 1 to 1. , The closer the value is to 1, the higher the degree of matching between the two fault characteristics; The dimension of the feature vector corresponds to the number of fault features; For the new fault feature vector The The component represents the number of new faults in the first position. The specific values ​​that each feature dimension can take; Historical fault feature vector The The component represents the historical fault in the first... The specific values ​​that each feature dimension can take.

10. A control system redundancy exception monitoring and switchover control system employing the control system redundancy exception monitoring and switchover control method of any one of claims 1 to 9, characterized by It comprises: A multi-source data acquisition unit is used to construct a comprehensive monitoring system, which first establishes a multi-source data acquisition channel to collect PLC running state data, service running indicators, network state data and machine performance indicators in real time through the multi-source data acquisition channel; A core service indicator generation unit is configured to receive service indicator designation instructions input by a user through a human-computer interaction interface, mark corresponding indicators as core service indicators according to the service indicator designation instructions, and configure independent monitoring scheduling priority parameters and monitoring threshold parameters for the core service indicators; A system anomaly recognition unit is configured to input data collected through the multi-source data acquisition channel into an AI algorithm model, perform standardization preprocessing on the input data, and then perform model training based on the preprocessed data and historical normal running data set and historical fault data set to establish a dynamic monitoring model, analyze data features in real time through the dynamic monitoring model, and identify whether the system has an abnormal state; A system switching control unit is configured to trigger a system switching control process if the dynamic monitoring model identifies that the main system has an abnormality, classify the abnormal data, determine the fault type and severity level, call a preset response strategy library, match corresponding backup system selection rules from the strategy library according to the fault type and severity level, and select a target backup system according to the rules.

Images