Dual-computer hot standby switching control method and system

Abstract

The invention relates to the technical field of automation control, and discloses a dual-computer hot standby switching control method and system, and the method comprises the following steps: collecting the operation state parameters of main control equipment, and forming the state vector of the system state; based on the state vector, establishing a state evolution model used for describing the time-varying characteristic of the operation state of the main control equipment; according to the state evolution model, estimating probability distribution of a state in a state space; calculating the state distribution entropy of the operation state of the main control equipment under the geometric information manifold; constructing a joint criterion function, and judging whether the main control equipment is in an unstable state or not according to the function; and when the criterion function satisfies a set switching condition, executing a switching control operation of the hot standby device, and completing control authority transfer of the master control and the standby control. Through the state evolution model and the entropy analysis module, real-time dynamic monitoring and stability evaluation of the operation state of the main control equipment are realized, and the effect of capturing the state change of the equipment in time and performing early warning is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation control technology, in particular to a dual-machine hot backup switching control method and system. BACKGROUND

[0002] In modern control systems, the main control device is one of the key components, and its stability directly affects the reliability and safety of the entire system. In order to ensure the stable operation of the main control device under various complex conditions, the traditional monitoring method usually relies on the static data monitoring of the device state, or evaluates the health state of the device through simple threshold judgment. These methods generally monitor the state based on the operating parameters of the device (such as temperature, voltage, load, etc.), and use periodic detection or fault triggering mechanism.

[0003] With the increase of system complexity, the traditional method gradually exposes the defect of being unable to cope with rapid changes or small changes, especially in a multi-dimensional and multi-level system environment, a single indicator or static monitoring is difficult to fully reflect the running state of the main control device. In order to solve these problems, in recent years, some researches have begun to introduce dynamic analysis methods based on state evolution model and entropy analysis, etc., through the prediction and stability analysis of the device running trend, to provide more accurate fault warning and health evaluation means. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a dual-machine hot backup switching control method and system, which solves the problem that the traditional technology usually relies on fixed threshold and simple index monitoring, and cannot capture the dynamic changes of the device running state, resulting in lack of sufficient early warning ability for rapid changes or potential faults of the system state.

[0005] To achieve the above purpose, the present application is realized by the following technical scheme: a dual-machine hot backup switching control method, comprising the following steps:

[0006] Collecting the operating state parameters of the main control device to form a state vector of the system state;

[0007] Based on the state vector, a state evolution model is established to describe the time-varying characteristics of the operating state of the main control device;

[0008] According to the state evolution model, the probability distribution of the state in the state space is estimated, and a geometric information manifold is further constructed to describe the state distribution structure;

[0009] The state distribution entropy of the main control device operating state under the geometric information manifold is calculated, and the change trend of the entropy with time is analyzed;

[0010] Construct a joint criterion function to comprehensively reflect the state deviation degree and operation stability of the main control device, and judge whether the main control device is in an unstable state based on the function;

[0011] When the criterion function meets the set switching conditions, the switching control operation of the hot standby device is executed to complete the transfer of control authority between the main control and the standby control.

[0012] Preferably, the state vector includes parameters of CPU utilization, memory usage, network fluctuation characteristic value and device temperature, which are normalized to form a unified dimensional input. The operating status of the main control device includes system performance status, service and communication status and physical and environmental status.

[0013] Preferably, the state evolution model is a stochastic differential equation model including a diffusion term, which can simultaneously describe the deterministic trend of system operation and the influence of external disturbances. The model can reflect the evolution trend of the system itself and the influence of environmental interference.

[0014] Preferably, a kernel density estimation method is used when estimating the state probability distribution, and the state samples are smoothed by a multi-dimensional Gaussian kernel function to obtain a continuous probability density distribution of the system in the state space.

[0015] Preferably, the geometric information manifold is a Riemann manifold structure established by constructing a Fisher information metric of a state probability density function, and is used to describe the information geometric characteristics of each point in the state space.

[0016] Preferably, the state distribution entropy is the integrated entropy on the Riemann information manifold, which is used to reflect the overall uncertainty of the system state distribution, and the time derivative of the entropy is used to characterize the changing trend of the system stability.

[0017] Preferably, the joint criterion function includes a distance term between the current state of the master device and the target steady state and a current system entropy term, and the two parts form a scalar function through a linear weighted combination, which is used to determine whether the hot standby switching conditions are met.

[0018] Preferably, when the derivative value of the joint criterion function in the continuous time window exceeds a preset threshold, it is determined that the operating stability of the main control device has decreased, and the hot standby control switching operation is triggered.

[0019] Preferably, the hot standby switching control operation includes:

[0020] Synchronize the current status data of the master device to the backup device,

[0021] Disconnect the output channel of the master device.

[0022] Activate the control channel of the standby device to complete the control takeover.

[0023] And the control takeover result is consistent.

[0024] A dual hot standby switching control system comprises:

[0025] A state acquisition module is configured to acquire multi-dimensional running state parameters of the master control device and form a state vector.

[0026] A state modeling module is configured to construct a state evolution model for describing the running evolution process of the master control device.

[0027] A density estimation module is configured to estimate the probability density distribution of the state sample according to the state sample.

[0028] An information manifold construction module is configured to construct a Fisher information metric based on the density function to form a Riemann manifold.

[0029] An entropy analysis module is configured to calculate the state distribution entropy and analyze the time variation trend thereof.

[0030] A criterion decision module is configured to construct and evaluate a joint criterion function to judge the stability of the master control device.

[0031] A control execution module is configured to execute a hot standby control switching operation when the criterion meets a set condition.

[0032] The present application provides a dual hot standby switching control method and system.

[0033] 1. The present application realizes real-time dynamic monitoring and stability evaluation of the running state of the master control device through the state evolution model and the entropy analysis module, and achieves the effect of timely capturing the state change of the device and early warning.

[0034] 2. The present application realizes accurate modeling and trend prediction of the system state by smoothing the state sample through the kernel density estimation method and estimating the probability distribution of the state in the state space, and achieves the effect of obtaining higher precision system running situation prediction from the probability distribution.

[0035] 3. The present application realizes accurate detection of the stability change of the master control device by constructing a joint criterion function and combining derivative threshold judgment, and achieves the effect of timely triggering the hot standby switching operation to ensure smooth switching and control takeover between the master control device and the standby control device. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The present application is a dual hot standby switching control method.

[0037] Figure 2 The present application is a dual hot standby switching control system. DETAILED DESCRIPTION

[0038] The technical solutions of the present application will be described clearly and completely below with reference to the drawings of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0039] Please refer to the drawings of the present application Figure 1 The embodiment of the present application provides a dual-machine hot backup switching control method, comprising the following steps:

[0040] Collecting the running state parameters of the master control device to form a state vector of the system state;

[0041] Establishing a state evolution model based on the state vector for describing the time-varying characteristics of the running state of the master control device;

[0042] According to the state evolution model, estimating the probability distribution of the state in the state space, and further constructing a geometric information manifold for describing the distribution structure of the state;

[0043] Calculating the state distribution entropy of the running state of the master control device under the geometric information manifold, and analyzing the change trend of the entropy over time;

[0044] Constructing a joint criterion function for comprehensively reflecting the state deviation degree and running stability of the master control device, and judging whether the master control device is in an unstable state according to the function;

[0045] When the criterion function meets the set switching condition, performing the switching control operation of the hot standby device to complete the control authority transfer of the master control and the backup control.

[0046] The state vector includes CPU utilization, memory usage, network fluctuation eigenvalue and device temperature parameters, and forms a unified dimension input through normalization processing. The running state of the master control device includes system performance state, service and communication state and physical and environmental state.

[0047] Specifically, to realize the multi-dimensional and standardized representation of the running state of the master control device, the system takes the state vector as the core input to build a unified analysis model. The state vector is derived from the collection and normalization processing of the current running state parameters of the device, and specifically includes four main parameters: CPU utilization, memory usage, network fluctuation eigenvalue and device temperature. These parameters together constitute the state description of the system at time t:

[0048]

[0049] Wherein the meanings of each term are as follows:

[0050] a normalized value representing CPU utilization rate;

[0051] a normalized value representing memory usage rate;

[0052] a normalized value representing network fluctuation characteristic value (such as packet loss rate change, RTT jitter, etc.);

[0053] a normalized value representing device temperature (such as CPU or mainboard), in order to eliminate the unit difference between different physical quantities, so as to be able to be used as a unified input vector for mathematical modeling processing, the following linear normalization method is adopted to process the original parameters:

[0054]

[0055] In the formula, the meanings of the symbols are as follows:

[0056] x i (t): represents the observation value of the i-th original running parameter at time t;

[0057] represents the minimum boundary value of the i-th parameter in the historical collected data;

[0058] represents the maximum boundary value of the i-th parameter in the historical collected data;

[0059] a dimensionless characteristic value obtained after normalization, with a value range of [0, 1].

[0060] The purpose of normalization is to unify all parameters to the same order of magnitude, so as to avoid the deviation or instability caused by the difference in numerical scale in the subsequent steps of probability modeling, density estimation, entropy calculation, etc. The selection of state parameters follows three core principles: first, the correlation with system running load; second, stable collection with high time resolution; third, sensitive response to running abnormities. In summary, the running state of the master device can be divided into the following dimensions:

[0061] composed of CPU utilization rate and memory usage rate, used to measure system load level and resource allocation efficiency, reflected by network fluctuation characteristic value, representing the communication stability and accessibility of the system in the process of executing tasks, represented by temperature, reflecting the working safety of the device under the current physical environment, indirectly representing potential hidden dangers such as hardware aging or heat dissipation problems. Through the above state representation method, the system can form a sequence of standardized state vectors that can be measured, compared and modeled at each time point, providing a raw basis for subsequent master device stability evaluation, information entropy evolution analysis and hot standby switching decision.

[0062] The state evolution model is a stochastic differential equation model containing a diffusion term, which can describe the deterministic trend of system operation and the influence of external disturbance at the same time, and the model can reflect the system evolution trend and environmental disturbance influence.

[0063] Specifically, the running state of the master control device is not only dominated by its own task execution rule, but also inevitably affected by external environmental disturbance, communication jitter, load fluctuation and other uncertain factors. In order to more realistically model and describe the running process of the master control system, the present application introduces a stochastic differential equation model (Stochastic Differential Equation, SDE) containing a diffusion term as the core expression form of system state evolution. The model takes into account the deterministic evolution trend of system state and the response of random disturbance, and can more truly reflect the dynamic complexity and disturbance uncertainty faced by the control system in actual operation. The model form is as follows:

[0064]

[0065] Wherein:

[0066] is the normalized state vector at time t, which represents the running state of the master control device in n state dimensions;

[0067] f(·): represents the deterministic evolution term of the system, which embodies the intrinsic dynamic trend of the device evolution over time under ideal conditions (without disturbance);

[0068] G(·): represents the diffusion intensity function, which is the sensitivity matrix of the state vector to the disturbance response;

[0069] W(t): is an n-dimensional Wiener process (standard Brownian motion), used to simulate the disturbance input;

[0070] is the incremental response of the state in a small time increment dt, which includes deterministic increment and random fluctuation increment.

[0071] The core advantage of this modeling method is to regard the state change as the sample path evolution in the probability process, and each path corresponds to the possible running track of the system. In the absence of disturbance, the system state is completely controlled by f, which embodies the internal driving characteristics of the device; while in the presence of significant disturbance, the GdW(t) part dominates the state evolution, which shows the passive response characteristics of the device to external changes. The deterministic term The deterministic term can be fitted according to historical state data, for example, using first-order linear regression, Bayesian dynamic modeling or neural network approximation, etc. The form is as follows:

[0072]

[0073] wherein:

[0074] is the state transition coefficient matrix

[0075] b(t): represents the bias drift vector of the system.

[0076] Diffusion term which can be constructed based on the variance estimation of state variables, and can also be simplified to a constant coefficient form, such as:

[0077]

[0078] where σ is a unified diffusion coefficient, I is the identity matrix, and represents the consistent response degree of all state components to the disturbance. This construction method is widely used in control theory and financial systems, and is simple, stable and easy to implement numerically. At the numerical solution level, the stochastic differential equation can be discretized using the Euler-Maruyama method, which is implemented as:

[0079]

[0080] where is a standard normal disturbance term sampled from a Gaussian process, used to simulate the disturbance impact on the system within a time interval Δt. The master system, under a certain initial state, not only determines its running state trajectory by its own structure (such as task load, process scheduling), but also is continuously affected by uncontrollable factors (such as network delay mutation, temperature rise, peripheral interference). The model output is the statistical distribution evolution trajectory of the state within a certain period of time in the future, rather than a unique trajectory, which makes the system have stronger flexibility and tolerance ability in judging the running stability, and also provides a theoretical basis for the calculation of criteria based on probability density and information entropy.

[0081] Kernel density estimation method is used to estimate the state probability distribution, and multi-dimensional Gaussian kernel function is used to smooth the state samples to obtain the continuous probability density distribution of the system in the state space.

[0082] Specifically, to obtain the probability distribution characteristics of the current running state of the master control device in the state space, the system uses a kernel density estimation method to perform non-parametric modeling on the sampled state data. This method does not need to pre-set the distribution model form, can construct the continuous probability density function in the state space through the sample itself, is better applicable to the actual running scene with irregular and nonlinear structure of state evolution, and the core advantage of this modeling mode is to regard the state change as a sample path evolution in the probability process, and each path corresponds to a possible running track of the system. In the absence of disturbance, the system state is completely controlled by f, which reflects the internal driving characteristics of the device; and in the presence of significant disturbance, the GdW(t) part dominates the state evolution, which shows the passive response characteristics of the device to external changes. The deterministic term may be fitted according to historical state data, for example, by using a first-order linear regression, a Bayesian dynamic modeling or a neural network approximation, and the form is as follows:

[0083]

[0084] wherein: represents the estimated density value at the point ; N represents the number of samples (i.e., the number of state vectors in the sliding window); K h (·): the bandwidth of the kernel function h is a kernel function, which is commonly in the form of a Gaussian kernel function; is the kth historical state sample; h is the bandwidth coefficient of the kernel function, which controls the density smoothing degree. In the present application, the kernel function uses a multi-dimensional Gaussian kernel, and the form is as follows:

[0085]

[0086] wherein: represents the vector difference between the current state to be estimated and the sample point; is a bandwidth matrix, which is commonly in the form of a diagonal matrix H = h 2 I, wherein h is a uniform bandwidth factor; |H|: the determinant of the bandwidth matrix, which is used for normalization coefficient; n: the dimension of the state vector, through the above estimation method, a smooth and continuous probability density map can be formed in the entire state space. The density function describes the distribution form of the current state of the master control device in the state space, that is, the possibility of a certain state combination under the current running background.

[0087] The geometric information manifold is a Riemann manifold structure established by constructing the Fisher information metric of the probability density function, which is used to describe the information geometric characteristics of each point in the state space.

[0088] Specifically, the state vector is constructed with the probability density function On the basis of the estimation, the probability distribution in the state space is regarded as a manifold point embedded in the parameter space The manifold is constructed with the Fisher information as the inner product tensor, and the Riemann metric is constructed.

[0089] Let the family of probability density functions be denoted as:

[0090]

[0091] where θ is the parameter vector of the density function, and Θ is the parameter space. Each parameter point θ corresponds to a probability distribution function equivalent to a "geometric point" in the state space, and the Fisher information metric tensor g ij (θ) is defined in the probability space, and is expressed as follows:

[0092]

[0093] In the formula:

[0094] g ij (θ) is the i, j component of the Fisher information metric tensor; is the expectation of the probability distribution corresponding to the state variable .

[0095] is the logarithmic form of the probability density function;

[0096] θ i , θ j is the i, j distribution parameter.

[0097] The Fisher information metric describes the sensitivity of the distribution change in any two directions in the parameter space, and can be used to describe the "local distance" properties between probability distributions. The information distance is not the spatial geometric distance in the Euclidean sense, but describes the "distinguishability difference" of two probability distributions;

[0098] On this basis, the Riemann manifold constructed by the Fisher information metric is called the geometric information manifold, and has the following characteristics: each point represents a probability distribution; the geometric distance between the points represents the distribution difference degree, and the local curvature reflects the sensitive direction of the system state under perturbation. Since the state density function of the application is realized by kernel density estimation, the parameter space is the set of kernel function center positions and bandwidth parameters. For practical applications, the local Fisher information matrix can be estimated or approximated by numerical means, and then the local inner product structure on the information manifold is constructed.

[0099] The state distribution entropy is integral entropy on a Riemann information manifold, and is used for reflecting the overall uncertainty of the system state distribution, and the time derivative of the entropy is used for representing the change trend of the system stability.

[0100] Specifically, in the dynamic evaluation index system of the running state of the master control device, in order to comprehensively reflect the overall uncertainty degree of the system state distribution, the definition of the state distribution entropy is introduced. The entropy value is based on a Riemann information manifold, is constructed on a state probability density function, and describes the information dispersion and chaos degree of the current state in the state space. The definition of the state distribution entropy is as follows:

[0101] Suppose that the probability density distribution of the current state of the system at any time t is Wherein is a normalized state vector in the state space. The information entropy is defined as follows:

[0102]

[0103] Wherein:

[0104] The state distribution entropy of the system at time t is represented by The Riemann information manifold embedded in the state space is represented by dV g The volume element defined by the Fisher information metric g on the information manifold is represented by The state probability density function corresponding to time t (obtained by KDE estimation) is represented by The logarithmic entropy kernel of the density function is represented by. The integral entropy is essentially a generalization of the Shannon entropy on the Riemann manifold, and reflects the uniformity of the current state distribution under the information geometric structure. If the state distribution is highly concentrated (i.e., the state is highly predictable), the entropy value is low. Conversely, if the state is widely dispersed or has a multi-peak structure (i.e., the state is highly uncertain), the entropy value will be significantly increased. In order to dynamically monitor the change trend of the system running state stability, the derivative of the state distribution entropy with respect to time is further calculated, that is:

[0105]

[0106] If It is indicated that the state density tends to be concentrated, and the system evolution process is gradually stable.

[0107] If It is indicated that the state density tends to be divergent, the system fluctuation is enhanced, and there may be potential abnormalities.

[0108] If the derivative value fluctuates or changes dramatically in a certain interval, it indicates that the system is in a boundary state, critical state or load transition interval. The entropy derivative can be used as a trend criterion for monitoring indicators to assist in determining whether the master-backup role switching is needed to avoid possible system operation failures in advance. In actual calculations, the entropy value and its derivative can be estimated by numerical integration of the density function within a sliding window, using Monte Carlo sampling or grid approximation. To enhance responsiveness, the system can set a dynamic threshold for the entropy derivative:

[0109]

[0110] where δ crit is an empirically set change threshold used to define the tolerance range of system uncertainty evolution.

[0111] The derivative term represents the rate of change of the uncertainty of the system state distribution structure in a physical sense, and has the following interpretation:

[0112] The joint criterion function includes the distance between the current state of the master device and the target steady state, and the current system entropy value. The two parts are combined by linear weighting to form a scalar function, which is used to determine whether the hot standby switching condition is met.

[0113] Specifically, to achieve smooth transition and high reliability operation of the system at critical moments, a joint criterion function is designed to determine whether to perform hot standby switching operation. The function considers two core indicators: the distance between the current state of the master device and the target state: this item measures the difference between the current running state and the expected stable working state of the system, which can usually be modeled by Euclidean distance, Manhattan distance or Mahalanobis distance, reflecting the degree of deviation from normal operation;

[0114] The current system entropy value: this indicator is used to quantify the uncertainty or chaos within the system. The higher the entropy value, the more unstable the system state, and the more likely it is to appear abnormal or fail. These two indicators are combined into a unified scalar criterion function by linear weighting, i.e.:

[0115] J = a · D(x, x * ) + b · H(x);

[0116] Where:

[0117] D(x, x * ) represents the current state x and the target steady state x *distance between the current state and the steady state; H(x) is the entropy value corresponding to the current system state, and a and β are weight coefficients adjusted according to the sensitivity of the system to steady state deviation and entropy value; when the value of the joint criterion function J exceeds the preset threshold, the system considers that the current state is no longer suitable for continuing to undertake the control task by the master device, thereby triggering the hot standby switching mechanism to hand over the control right to the backup device to ensure the continuous and stable operation of the system.

[0118] When the derivative value of the joint criterion function within a continuous time window exceeds the preset threshold, it is determined that the running stability of the master device decreases, and a hot standby control switching operation is triggered.

[0119] Specifically, relying only on the static value of the joint criterion function may not be able to timely reflect the mutation or potential instability trend of the system state. Therefore, the derivative value of the joint criterion function within a continuous time window is introduced as a sensitive index of dynamic change, which is used to capture the sudden fluctuations of the system running state. When the absolute value of the first derivative of the criterion function continuously exceeds the preset threshold within a set sliding time window (for example, several seconds to tens of seconds), it indicates that the system state is rapidly deteriorating or mutating. At this time, it can be reasonably determined that the running stability of the master device has shown a significant downward trend, and it is not an occasional abnormality caused by short-term disturbance or noise.

[0120] Based on this judgment logic, the system immediately triggers a hot standby control switching operation to take over the current task by the backup control device, thereby avoiding the master device causing system failure or service interruption when the stability continues to deteriorate, and ensuring the continuity and reliability of the system operation. This dynamic criterion supplements the static threshold determination mechanism, so that the hot standby switching strategy not only has steady-state response capability, but also has real-time response capability to rapid changes in system state, significantly improving the sensitivity and robustness of fault response.

[0121] The hot standby switching control operation includes:

[0122] synchronizing the current state data of the master device to the backup device,

[0123] disconnecting the output channel of the master device,

[0124] activating the control channel of the backup device to complete control takeover,

[0125] and performing consistency verification on the control takeover result.

[0126] Specifically, once the hot standby switching criterion is met, the system will automatically perform the hot standby switching control operation to ensure smooth transfer of control right and continuous operation of system function. The operation process usually includes the following key steps:

[0127] First, the key operating states of the current master device (such as internal control variables, system parameters, actuator states, etc.) are synchronized in real time to the backup device in a high-precision manner. This process ensures that the backup device can continue running from the latest state of the master when it takes over the control task, avoiding control impact or system instability caused by state jumps. After synchronization is complete, the system immediately disconnects the control output channel of the master device to prevent it from continuing to send control instructions to the controlled object, avoiding competition or conflict with the backup device. This step is usually implemented through logical disconnection or hardware relay switching. Subsequently, the system activates the control output channel of the backup device, allowing it to formally take over the master task. The backup device quickly recovers the control link based on the synchronized state data and sends new control signals to the controlled system, maintaining system operation continuity. To ensure that the system runs as expected after switching, consistency verification of the takeover process is required. This may include output behavior comparison, controlled object response monitoring, and key variable trend analysis. Once an abnormal takeover is detected, the system can further take measures such as warning, rollback, or manual intervention;

[0128] This hot standby switching mechanism has the characteristics of fast response, state alignment, conflict isolation, and safety verification, and is the core means to ensure the high availability and fault tolerance of critical control systems. The entire process needs to be completed within milliseconds to seconds to minimize the disturbance to system operation.

[0129] Please refer to the attached Figure 2 A dual-machine hot standby switching control system, comprising:

[0130] A state acquisition module for acquiring multi-dimensional operating state parameters of the master device and forming a state vector;

[0131] A state modeling module for constructing a state evolution model describing the evolution process of the master device;

[0132] A density estimation module for estimating the probability density distribution of the state samples;

[0133] An information manifold construction module for constructing a Fisher information metric based on the density function to form a Riemann manifold;

[0134] An entropy analysis module for calculating the state distribution entropy and analyzing its time variation trend;

[0135] A criterion decision module for constructing and evaluating a joint criterion function to determine the stability of the master;

[0136] A control execution module for executing a hot standby control switching operation when the criterion meets the set conditions.

[0137] Specifically, the state acquisition module is configured to acquire multi-dimensional running state parameters of the master control device in real time, such as temperature, voltage, current, CPU load, control deviation, etc. The state vector containing key dynamic characteristics is constructed through sensor acquisition or software monitoring, providing a data basis for subsequent modeling and analysis.

[0138] The state modeling module is responsible for constructing a state evolution model capable of describing the running evolution law of the master control device based on historical and real-time state vector data. The model uses autoregressive model, Markov chain, state space model, etc. to predict future state trends and judge the stability of running behavior.

[0139] The density estimation module is configured to estimate the probability density distribution of the collected state samples in the state space. Non-parametric or semi-parametric methods such as kernel density estimation (KDE), Gaussian mixture model (GMM), or variational inference can be used to reveal the uncertainty structure of the system state;

[0140] The information manifold construction module uses Fisher information metric to construct Riemann information manifold based on the probability distribution function obtained by density estimation, which describes the difference between various states in the state space from a geometric perspective. This information geometric perspective helps to more essentially describe the system dynamic evolution path and its stability;

[0141] The entropy analysis module is configured to calculate the information entropy of the current state distribution, reflecting the uncertainty level of the system and analyzing its evolution trend over time. An abnormal increase in entropy value usually indicates an increase in volatility of the system state or a potential risk of loss of control, providing a basis for fault warning;

[0142] The criterion decision module is responsible for constructing and evaluating a joint criterion function based on "state deviation + entropy increase trend", and comprehensively judging whether the master control device is in a stable working state. This module analyzes the value and derivative of the joint function, and when the judgment result meets the set switching condition, it issues a switching instruction;

[0143] The control execution module, upon receiving the criterion decision result, immediately performs the hot standby switching control operation, including state synchronization, master control offline, standby control online, and consistency verification steps, to realize smooth transfer of control rights and ensure continuous and safe operation of the system.

[0144] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A dual-machine hot standby switching control method, characterized in that: The following steps are involved: Collect the operating status parameters of the main control device to form the state vector of the system status; Establishing a state evolution model based on the state vector for describing the time-varying characteristics of the operating state of the main control device; According to the state evolution model, the probability distribution of the state in the state space is estimated, and a geometric information manifold for describing the state distribution structure is further constructed; Calculating the state distribution entropy of the operating state of the main control device under the geometric information manifold, and analyzing the change trend of the entropy over time; Construct a joint criterion function to comprehensively reflect the state deviation degree and operation stability of the main control device, and judge whether the main control device is in an unstable state based on the function; When the criterion function meets the set switching conditions, the switching control operation of the hot standby device is executed to complete the transfer of control authority between the main control and the standby control.

2. The dual-machine hot standby switching control method according to claim 1, characterized in that: The state vector includes parameters of CPU utilization, memory usage, network fluctuation characteristic value and device temperature, which are normalized to form a unified dimensional input. The operating status of the main control device includes system performance status, service and communication status and physical and environmental status.

3. The dual-machine hot standby switching control method according to claim 1, characterized in that: The state evolution model is a stochastic differential equation model containing a diffusion term, which can simultaneously describe the deterministic trend of system operation and the influence of external disturbances. The model can reflect the system ontology evolution trend and the influence of environmental interference.

4. The dual-machine hot standby switching control method according to claim 1, characterized in that: The kernel density estimation method is used to estimate the state probability distribution. The state samples are smoothed by the multi-dimensional Gaussian kernel function to obtain the continuous probability density distribution of the system in the state space.

5. The dual-machine hot standby switching control method according to claim 1, characterized in that: The geometric information manifold is a Riemann manifold structure established by constructing the Fisher information metric of the state probability density function, and is used to describe the information geometric characteristics of each point in the state space.

6. The dual-machine hot standby switching control method according to claim 1, characterized in that: The state distribution entropy is the integrated entropy on the Riemann information manifold, which is used to reflect the overall uncertainty of the system state distribution. The time derivative of the entropy is used to characterize the changing trend of the system stability.

7. The dual-machine hot standby switching control method according to claim 1, characterized in that: The joint criterion function includes a distance term between the current state of the master device and the target steady state and a current system entropy term. The two parts form a scalar function through a linear weighted combination, which is used to determine whether the hot standby switching conditions are met.

8. The dual-machine hot standby switching control method according to claim 1, characterized in that: When the derivative value of the joint criterion function in the continuous time window exceeds a preset threshold, it is determined that the operating stability of the main control device has decreased, and the hot standby control switching operation is triggered.

9. The dual-machine hot standby switching control method according to claim 1, characterized in that: The hot standby switching control operation includes: Synchronize the current status data of the master device to the backup device, Disconnect the output channel of the master device. Activate the control channel of the standby device to complete the control takeover. And perform consistency check on the control takeover results.

10. A dual-machine hot standby switching control system, characterized in that: A dual-machine hot standby switching control method as claimed in claim 1, comprising: The state acquisition module is used to obtain the multi-dimensional operating state parameters of the main control device and form a state vector; The state modeling module is used to build a state evolution model for describing the operation evolution process of the main control device; Density estimation module, used to estimate the probability density distribution of state samples; Information manifold construction module, used to construct Fisher information metric based on density function to form Riemann manifold; Entropy analysis module, used to calculate the state distribution entropy and analyze its time change trend; Criteria decision module, used to construct and evaluate the joint criterion function to judge the stability of the main control; The control execution module is used to execute the hot standby control switching operation when the judgment criterion meets the set conditions.

Images