A method, system, electronic equipment, and software product for real-time reliability prediction and adaptive maintenance of a PID control system.

Abstract

This invention discloses a method, system, electronic device, and program product for real-time reliability prediction and adaptive maintenance of a PID control system, relating to the field of industrial digital twin operation and maintenance technology. The method includes: generating a probabilistic digital twin for each component based on current data acquired in real-time from the PID physical system; calibrating the degradation process of the probabilistic digital twin for each component based on historical data; predicting the reliability of the PID physical system at a future time using the calibrated probabilistic digital twin; and triggering a maintenance warning when the prediction result is less than an adaptive warning threshold. The adaptive warning threshold is adjusted based on the variance and gradient of the prediction result. This invention achieves an optimal balance between false alarms and missed alarms.

Description

Technical Field

[0001] This invention relates to the field of industrial digital twin operation and maintenance technology, and in particular to a method, system, electronic device, and program product for real-time reliability prediction and adaptive maintenance of a PID control system. Background Technology

[0002] The advancement of Industry 4.0 and the Industrial Internet has made predictive maintenance possible through data-driven approaches. While cloud computing platforms can process massive amounts of data, their latency may not meet the real-time requirements of control systems. Digital twin technology, as a key technology for realizing the integration of cyber-physical systems, has become a research hotspot both domestically and internationally.

[0003] Existing predictive and maintenance technologies for PID (Proportion-Integral-Differential) control systems, such as cloud-based digital twin solutions, model predictive control-based fault-tolerant solutions, and offline simulation solutions, all use fixed thresholds, which cannot cope with the uncertainty of the prediction itself, leading to false alarms and missed alarms; moreover, there are lag problems, which affect the maintenance efficiency of PID control systems.

[0004] Therefore, a new technology is urgently needed to overcome the aforementioned shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, electronic device, and program product for real-time reliability prediction and adaptive maintenance of PID control systems, in order to solve the technical problems of high latency, poor real-time performance, false alarms, and missed alarms in existing technologies. The various technical effects of the preferred solutions among the many technical solutions provided by this invention are detailed below.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] This invention provides a method for real-time reliability prediction and adaptive maintenance of a PID control system, the steps of which include:

[0008] Based on the current data of each component obtained in real time from the PID physical system, a probabilistic digital twin is generated for each component, wherein the current data includes key variable data, control command data and / or external intervention data of each component;

[0009] Based on the historical data of each component, the degradation process of the probabilistic digital twin corresponding to each component is parameter-calibrated, wherein the historical data includes the drift coefficient of each component under stress conditions and the diffusion coefficient during the degradation process;

[0010] The calibrated probabilistic digital twin is used to predict the reliability of the PID physical system at a future time. When the prediction result is less than the adaptive warning threshold, a maintenance warning is triggered. The adaptive warning threshold is adjusted based on the variance and gradient of the prediction result.

[0011] In one or more embodiments, the probabilistic digital twin includes a system state output vector equation and an observation output vector equation, wherein the state output vector equation is obtained based on vectorized key variables of each component, a control input vector, and process noise with a set distribution; and the observation output vector equation is obtained based on vectorized key variables of each component and observation noise with a set distribution.

[0012] In one or more embodiments, the method further includes determining the cumulative degradation of each component at the current time based on the drift coefficient of each component's key variables at the current time and the equivalent aging time, thereby obtaining a dynamic evolution probability digital twin of each component.

[0013] In one or more embodiments, the cumulative degradation includes standard Brownian motion as a variable.

[0014] In one or more embodiments, the step of parameter calibration of the degradation process of the probabilistic digital twin corresponding to each component based on historical data of each component includes:

[0015] At the initial time, multiple particles are initialized for the parameter to be estimated for each of the probabilistic digital twins. Each particle corresponds to the initial value and weight of the parameter to be estimated. Based on the given distribution and initial value of the parameter to be estimated, the probability of each particle at the initial time is calculated.

[0016] Predict the transition probability of each particle at the initial time based on its probability at the initial time, and obtain the parameter value of each particle at the next time based on the transition probability.

[0017] Based on the parameter value of each particle at the next time step, the given observation vector, and the given observation vector distribution, the observation probability of each particle at the next time step is obtained.

[0018] Update the weights for the next time step based on the parameter values ​​of each particle at the next time step and the weights at the initial time step.

[0019] Based on the weights of all particles, determine the number of effective particles. When the number of effective particles is less than a given number, return to the step corresponding to the initial time, and execute the above steps sequentially until the number of effective particles is not less than the given number.

[0020] In one or more embodiments, predicting the reliability of the PID physical system at a future time using the calibrated probabilistic digital twin includes:

[0021] The performance margin of each probabilistic digital twin is determined based on the actual steady-state error, the steady-state error under rated conditions, the maximum allowable steady-state error threshold, and the minimum allowable steady-state error threshold corresponding to the cumulative degradation amount determined at a future time as the degradation time. Given the observation data up to the current time, the probability that the performance margin of all probabilistic digital twins is always greater than zero from the current time to a future time is the reliability of the PID control system at a future time and at the current time.

[0022] In one or more embodiments, the remaining useful life of the PID control system is determined based on the reliability, wherein the remaining useful life is the time span between a future time when the reliability first falls below a given threshold and the current time.

[0023] As a joint invention, this invention also provides a real-time reliability prediction and adaptive maintenance system for a PID control system, comprising a PID physical subsystem, a probabilistic digital twin subsystem, and a centralized management and analysis subsystem connected in sequence. The PID physical subsystem is located at the physical layer, the probabilistic digital twin subsystem is located at the edge layer, and the centralized management and analysis subsystem is located at the cloud layer.

[0024] The PID physical subsystem is used to collect real-time data corresponding to each component of the PID physical subsystem during operation. The real-time data includes key variable data of the components, control command data and / or external intervention data.

[0025] The probabilistic digital twin subsystem runs in parallel with the PID physical subsystem and is used to receive the real-time data from the PID physical subsystem. Based on the real-time data, it implements the real-time reliability prediction and adaptive maintenance method for the PID control system described above. At the same time, it displays the prediction results and early warning strategies locally.

[0026] The centralized management and analysis subsystem is used to mine the historical data required by the probabilistic digital twin subsystem; at the same time, based on the prediction results of the probabilistic digital twin subsystem, it optimizes the parameters of the probabilistic digital twin and issues early warning strategies to the probabilistic digital twin subsystem.

[0027] As a joint invention, the present invention also provides an electronic device, comprising:

[0028] At least one processor;

[0029] At least one memory used to store processor-executable instructions;

[0030] The at least one processor implements the above-described method for real-time reliability prediction and adaptive maintenance of a PID control system by running the executable instructions.

[0031] As a joint invention, this invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the real-time reliability prediction and adaptive maintenance method for a PID control system described above.

[0032] Implementing one of the above-described technical solutions of the present invention has the following advantages or beneficial effects:

[0033] This invention abandons fixed thresholds and constructs a dynamically adjusted early warning threshold function. This function comprehensively considers the expected value of prediction reliability, the variance of uncertainty, and its rate of change. When prediction uncertainty is high, the threshold is automatically increased to reduce false alarms; when reliability declines rapidly, an early warning is triggered to buy time for response. This achieves an optimal balance between false alarms and missed alarms, greatly enhancing the practicality of early warning and the effectiveness of decision support. Attached Figure Description

[0034] The accompanying drawings used are briefly described below. It is obvious that the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings:

[0035] Figure 1 This is a flowchart of a real-time reliability prediction and adaptive maintenance method for a PID control system according to an embodiment of the present invention;

[0036] Figure 2 This is an architecture diagram of a real-time reliability prediction and adaptive maintenance system for a PID control system according to an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of a reliability prediction curve according to an embodiment of the present invention;

[0038] Figure 4 This is a block diagram of an electronic device structure according to an embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the present invention clearer, various exemplary embodiments described below will be referenced to the accompanying drawings, which form part of the exemplary embodiments, illustrating various exemplary embodiments that may be used to implement the present invention. Unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. It should be understood that they are merely examples of processes, methods, and apparatuses consistent with some aspects of the present invention disclosed as detailed in the appended claims, and other embodiments may be used, or structural and functional modifications may be made to the embodiments listed herein without departing from the scope and spirit of the present invention.

[0040] To illustrate the technical solution described in this invention, specific embodiments are described below, showing only the parts related to the embodiments of this invention.

[0041] Example 1: As Figure 1 As shown, this invention provides a method for real-time reliability prediction and adaptive maintenance of a PID control system, the steps of which include:

[0042] S100. Generate a probabilistic digital twin for each component based on the current data of each component obtained in real time from the PID physical system.

[0043] The aforementioned probabilistic digital twin can be defined as: a high-fidelity probabilistic model constructed in virtual space that can characterize the uncertainty of physical entities and be updated in real time. It not only reflects the system state but also quantifies the uncertainty of the state.

[0044] It is understood that the real-time data acquired above includes key variable data of each component in the PID physical system (such as position, velocity, temperature, etc.), control command data of the PID physical system, and / or external intervention data.

[0045] The aforementioned data can be acquired through corresponding sensors. After acquisition, it needs to be preprocessed and feature extracted to form vector data, which is used for probabilistic digital twin construction and reliability prediction.

[0046] To construct a digital twin in virtual space capable of accurately mapping and predicting the behavior of a PID physical system, its mathematical foundation must first be established. Due to the significant nonlinearity and non-Gaussianity of the control system degradation process, traditional Kalman filter algorithms based on the Gaussian assumption are no longer applicable. Therefore, in this embodiment, a probabilistic modeling method is chosen.

[0047] In one or more embodiments, the probabilistic digital twin includes a system state output vector equation (hereinafter Equation (1)) and an observation output vector equation (hereinafter Equation (2)), wherein,

[0048] The state output vector equation is obtained based on the vectorized key variables of each component, the control input vector, and the process noise with a set distribution; the observation output vector equation is obtained based on the vectorized key variables of each component and the observation noise with a set distribution.

[0049] In a specific embodiment, the state-space equation of the probabilistic digital twin is as follows:

[0050] (1);

[0051] (2);

[0052] in, The output vector represents the behavior state of the PID physical system at time t.

[0053] For the state vectors of the PID physical system components, these are key variables used to fully describe the internal dynamic behavior of the PID physical system;

[0054] A is the state matrix, which describes the internal dynamic characteristics of the PID physical system;

[0055] The control input vector describes the control commands and / or external interventions of the PID physical system;

[0056] B is the control input matrix, describing the control input influencing factors;

[0057] ~N(0, Q): Process noise (following a normal distribution with mean 0 and variance Q), characterizing Uncertainty caused by factors was not taken into account;

[0058] The observed output vector represents the measurable physical quantity of the PID physical system;

[0059] C is the observation matrix, which describes the relationship between the output vector and the observation output vector;

[0060] ∼N(0,R): Observation noise (following a normal distribution with mean 0 and variance Q), characterizing the sensor measurement error.

[0061] In specific embodiments, A and C are typically obtained through mechanistic modeling or system identification of the controlled object (such as a motor, valve, or other PID component). For example, differential equations can be established by analyzing physical laws (such as Newton's laws and circuit laws), and then discretized to obtain A and C. B reflects the influence of the control input u(t) on the state x(t), and can be obtained through the design parameters of the control system or experimental calibration.

[0062] The Q and R mentioned above are covariance matrices, which can be estimated through historical data statistics or noise analysis during system debugging. For example, data can be collected during steady-state system operation to calculate the statistical characteristics of noise.

[0063] In one or more embodiments, the method further includes determining the cumulative degradation of each component at the current time based on the drift coefficient of each component's key variables and the equivalent aging time, thereby obtaining a dynamic evolution probability digital twin of each component. The cumulative degradation includes standard Brownian motion as a variable.

[0064] Understandably, a high-fidelity digital twin must be able to simulate the aging process of a system during long-term operation. Therefore, based on the state-space model, a performance degradation model for components (which can be key components affecting the PID system) is introduced. This model aims to quantitatively describe how environmental stresses and workloads drive irreversible drift in component parameters, thus transforming the static skeleton into a dynamically evolving organism. The specific component degradation model (degradation equation) is shown below:

[0065] (3);

[0066] in, : The cumulative degradation amount of the i-th component at degradation time t;

[0067] Equivalent aging time, which is related to stress conditions, quantifies the accumulated aging effect from the start time to the present time. It is not an absolute physical time. For example, running under high load for 1 hour may be equivalent to the aging effect of running under rated load for 2 hours.

[0068] : Drift coefficient, which is the stress condition θ i (e.g., functions corresponding to temperature, vibration, load);

[0069] : Diffusion coefficient, characterizing the intensity of random fluctuations in the degradation process;

[0070] : Standard Brownian motion distribution, simulating random uncertainty in the degradation process.

[0071] In a specific embodiment, the diffusion coefficient Estimate using the following methods:

[0072] Historical data fitting: The degradation trajectory of the component is obtained by fitting it through maximum likelihood estimation or Bayesian inference using data recorded in accelerated aging tests or historical operation and maintenance data.

[0073] Online calibration: The particle filter algorithm (Equations 5 to 9) in step S200 below will be updated in real time. This allows it to adapt to the uncertainties of the actual degradation process.

[0074] In this embodiment, B(t) is the standard Brownian motion (Wiener process), which is mathematically defined as follows:

[0075] B(t) ~ N(0,t), which represents a normal distribution with a mean of 0 and a variance of t. In simulations, this can be achieved through discretization. For example, B(t+Δt) = B(t)+ξ· , where ξ~N(0,1) is standard Gaussian noise.

[0076] It is understandable that the state-space equations (Equations 1 and 2) described above describe the short-term dynamic behavior of the PID physical system, such as the real-time response of the PID controller. The component degradation model (Equation 3) described above describes the long-term performance degradation of components, such as wear and aging, and its cumulative degradation amount. This will affect the parameters in the state-space equations (such as matrices A, B, and C). The two are coupled. That is, the cumulative degradation... These parameters will be embedded into the state-space model as time-varying parameters. For example, valve wear may cause changes in the elements of the state matrix A, thereby affecting the system dynamics.

[0077] S200. Based on the historical data of each component, perform parameter calibration on the degradation process of the probabilistic digital twin corresponding to each component. The historical data includes the drift coefficient that each component can obtain under stress conditions and the diffusion coefficient that each component can obtain during the degradation (aging) process.

[0078] It is understandable that the aforementioned historical data can be obtained from the cloud, including but not limited to drift coefficients and diffusion coefficients characterizing the aging process of components obtained through public channels, or research data that can be obtained directly through other means.

[0079] Based on the above embodiments, after establishing a model including a degradation mechanism, it is necessary to ensure that the digital twin and the physical entity remain synchronized at all times. Due to the inaccuracy of the initial model parameters and the time-varying nature of the operating environment, this embodiment provides a mechanism for continuously calibrating model parameters using real-time observation data through a Bayesian filtering framework. The core formula of the Bayesian update process is:

[0080] (4);

[0081] in,

[0082] Θ={μ1, σ1, μ2, σ2,…}: The set of parameters to be estimated for the degradation model corresponding to each component, that is, the set consisting of drift coefficients and diffusion coefficients;

[0083] ={y1, y2, …,yt}: The sequence of observation data from the starting time 1 to the current time t, obtained from the observation output vector equation;

[0084] The prior distribution of the drift coefficient represents the parameter knowledge based on historical data;

[0085] Likelihood function: represents the probability of observing data at the current time t given a drift coefficient;

[0086] The posterior distribution of the drift coefficient, combined with updated knowledge based on new evidence.

[0087] The above , , This can be obtained through the corresponding historical data.

[0088] Understandably, given the high nonlinearity and non-Gaussian nature of the model, analytically solving the Bayesian filter described above is extremely difficult. This embodiment employs the Monte Carlo method of the particle filter algorithm for numerical implementation. This algorithm approximates the posterior probability distribution of the parameters using a swarm of weighted particles.

[0089] In one or more embodiments, parameter calibration is performed on the degradation process of the probabilistic digital twin corresponding to each component based on historical data of each component, including:

[0090] At the initial time, multiple particles (N) are initialized for the parameters to be estimated in each probabilistic digital twin. Each particle corresponds to an initial value and weight of the parameter to be estimated. Based on the given distribution of the parameter to be estimated (obtained from historical data) and the initial values, the probability of each particle at the initial time is calculated. The mathematical estimation expression is as follows:

[0091] (5);

[0092] P is a given distribution.

[0093] The transition probability of each particle at the initial time (e.g., t-1) is predicted, and the parameter value of each particle at the next time (t) is obtained based on the transition probability. The mathematical estimation process is as follows:

[0094] (6);

[0095] in, This can be obtained using random walk or parametric evolution models, for example... = +η, where η is Gaussian noise. Alternatively, in step S200, each particle can evolve independently according to the transfer model, thereby approximating the posterior distribution of the parameters. This method is applicable to nonlinear non-Gaussian systems.

[0096] Based on the parameter value of each particle at the next time step, the given observation vector, and the given observation vector distribution (obtained from historical data), its observation probability at the next time step is obtained; the weights at the next time step are updated based on the parameter value of each particle at the next time step and the weights at the initial time step. The mathematical estimation process is as follows:

[0097] (7);

[0098] (8);

[0099] in, Given the parameter values ​​of the i-th particle at the next time step (t), and the distribution of the observation vector given the observation vector. Let be the weight of the i-th particle in the next time step (t).

[0100] Based on the weights of all particles, determine the number of effective particles. If the number of effective particles is less than a given number, return to the step corresponding to the initial time. Repeat the above steps until the number of effective particles is not less than the given number. Wherein, the number of effective particles... The expression is as follows:

[0101] (9).

[0102] The given number can be set to N / 2 or N / 3.

[0103] Based on the above embodiments, the Bayesian update process updates the probability distribution of model parameters through an iterative cycle of "initialization → prediction → update → resampling". This process ensures that the digital twin can continuously correct its understanding of the physical system's state as real-time data is continuously input.

[0104] S300 uses a calibrated probabilistic digital twin to predict the reliability of the PID physical system at a future time. When the prediction result is less than the adaptive early warning threshold, a maintenance early warning is triggered.

[0105] In one or more embodiments, using a calibrated probabilistic digital twin to predict the reliability of a PID physical system at a future time includes:

[0106] The performance margin function is determined based on the actual steady-state error corresponding to the cumulative degradation amount at a future time, the steady-state error under rated operating conditions, the maximum allowable steady-state error threshold, and the minimum allowable steady-state error threshold for each probabilistic digital twin. If the performance margin obtained through the performance margin function is still greater than zero, then the corresponding probabilistic digital twin possesses reliability.

[0107] In this embodiment, reliability is essentially the probability that system performance will remain within a safe threshold. A performance margin function can be defined as follows:

[0108] (10);

[0109] in, Let be the actual steady-state error at degradation time t. When using the above formula (3) for prediction, it can be obtained from the above formula (3). get .

[0110] This refers to the steady-state error under rated operating conditions. For the maximum allowed, The minimum steady-state error threshold, , , This information can be obtained from the instruction manual of the corresponding component or through relevant experiments.

[0111] It should be noted that the probability of the event "at some point in the future, the performance margin obtained through the performance margin function will still be greater than zero" can be estimated by statistically inferring the future states of the probabilistic digital twin set (i.e., the particle swarm). As shown below:

[0112] (11);

[0113] in, {}: Indicator function; value is 1 if the condition is met, otherwise 0. Let t be the current time, and t be the future time. Let be the distribution function to be estimated.

[0114] Furthermore, It can be estimated using the Monte Carlo method; its expression is as follows:

[0115] (12);

[0116] Among them, M (i) (t+Δt) represents the predicted future performance margin for the i-th particle. The number of particles.

[0117] In one or more embodiments, observation data up to the current time... Under the given conditions, the probability that the performance margin M(t+τ) of all probabilistic digital twins is always greater than zero from the current time t to a future time t+τ is the reliability of the PID control system within time τ. The prediction result includes the reliability of the PID control system within time τ, and the reliability is used to measure the reliability mentioned above.

[0118] Furthermore, it also includes determining the remaining useful life of the PID control system based on reliability, where the remaining useful life is the time span between a future time when the reliability first falls below a given threshold and the current time.

[0119] It should be further explained that, in addition to predicting reliability at a future point in time, operational decisions are more concerned with how much longer the system can continue to function normally. This leads to the concept of Remaining Useful Life (RUL), whose complete probability distribution provides richer decision-making information than a single estimate.

[0120] Remaining useful life (RUL) probability distribution:

[0121] (13);

[0122] The time span is the remaining useful life.

[0123] In one or more embodiments, the adaptive warning threshold is adjusted based on the variance and gradient of the prediction results. Specifically, the adaptive warning threshold based on the reliability prediction results is:

[0124] (14);

[0125] Basic reliability threshold; can be set.

[0126] The standard deviation of reliability prediction, which characterizes the prediction uncertainty, can be obtained by formula (3) above.

[0127] : Reliability change rate (gradient of prediction results)

[0128] (15);

[0129] β: Adaptive adjustment coefficient, obtained by linearly fitting formula (14) based on the corresponding prediction base data.

[0130] When the predicted reliability R(t+Δt) < An alert is triggered at any time.

[0131] It is understandable that the adaptive early warning mechanism corresponding to the aforementioned adaptive early warning threshold transforms the prediction results into effective operation and maintenance actions. Its threshold can be dynamically adjusted according to the uncertainty (variance) and changing trend (gradient) of the prediction reliability, thereby achieving the best balance between false alarms and false alarms.

[0132] It should be noted that this implementation can introduce a multi-level early warning strategy, providing differentiated alarms based on the risk level.

[0133] like Figure 3 As shown, the reliability prediction curve illustrates the trend of system reliability over time, including the predicted mean, adaptive warning threshold, and failure threshold. The adaptive warning mechanism can dynamically adjust the adaptive warning threshold based on prediction uncertainty, effectively balancing false alarms and missed alarms.

[0134] In summary, to improve the accuracy of early warnings, this embodiment provides an intelligent adaptive early warning mechanism. This mechanism abandons fixed thresholds and constructs a dynamically adjusted early warning threshold function. This function comprehensively considers the expected value of prediction reliability, the variance of uncertainty, and its rate of change. When prediction uncertainty is high, the threshold is automatically increased to reduce false alarms; when reliability declines rapidly, an early warning is triggered in advance to gain time for response. This achieves an optimal balance between false alarms and missed alarms, greatly enhancing the practicality of early warnings and the effectiveness of decision support.

[0135] To address the nonlinear and non-Gaussian characteristics of control system degradation processes, this embodiment provides an online parameter identification algorithm based on particle filtering. This algorithm uses particle swarm optimization to characterize the posterior probability distribution of model parameters, employs sequential Bayesian updates, and fuses observational data in real time to dynamically calibrate degradation model parameters (such as drift coefficients) of key components in the digital twin. This method not only accurately tracks the time-varying characteristics of parameters but also provides complete parameter uncertainty quantification, significantly improving the model's adaptability and prediction accuracy under complex operating conditions.

[0136] Example 2: Figure 3As shown, this embodiment provides a real-time reliability prediction and adaptive maintenance system for a PID control system, including a PID physical subsystem, a probabilistic digital twin subsystem, and a centralized management and analysis subsystem connected in sequence. The PID physical subsystem is located at the physical layer (field devices), the probabilistic digital twin subsystem is located at the edge layer, and the centralized management and analysis subsystem is located at the cloud layer. Specifically,

[0137] The PID physical subsystem is used to collect real-time data corresponding to each component of the PID physical subsystem during operation. The real-time data includes key variable data of the components, control command data and / or external intervention data.

[0138] The probabilistic digital twin subsystem runs in parallel with the PID physical subsystem. It receives real-time data from the PID physical subsystem and implements the real-time reliability prediction and adaptive maintenance method for the PID control system described in Example 1 based on the real-time data. At the same time, the prediction results and early warning strategies are displayed locally.

[0139] The centralized management and analysis subsystem is used to mine the historical data required by the probabilistic digital twin subsystem; at the same time, based on the prediction results of the probabilistic digital twin subsystem, it optimizes the parameters of the probabilistic digital twin and issues early warning strategies to the probabilistic digital twin subsystem.

[0140] In one or more embodiments, the PID physical subsystem includes a PID controller, an actuator, a controlled object, and a sensor connected in sequence, with the sensor connected to the probabilistic digital twin subsystem.

[0141] Furthermore, the centralized management and analysis subsystem, namely the remote cloud platform, includes a data lake and historical database, an AI model training and optimization platform, and a global reliability monitoring center, all connected in sequence. The data lake and historical database are used to mine historical data required by the probabilistic digital twin subsystem, while the AI ​​model training and optimization platform is used to optimize the parameters of the probabilistic digital twin based on its prediction results and issue early warning strategies to the probabilistic digital twin subsystem.

[0142] The probabilistic digital twin subsystem includes a data acquisition interface for connecting sensors, a data lake, and a historical database; a lightweight probabilistic digital twin; a Bayesian filtering algorithm library; and a real-time prediction engine. The Bayesian filtering algorithm library is used to calibrate the parameters of the lightweight probabilistic digital twin, using the same method as in Example 1. The real-time prediction engine uses the calibrated probabilistic digital twin to predict and warn of the reliability of the PID physical system at a future time, and displays the prediction and warning strategies locally, also using the same method as in Example 1.

[0143] The implementation methods not illustrated in this embodiment are consistent with those in Embodiment 1, and will not be repeated here.

[0144] This embodiment provides an edge-cloud collaborative architecture for reliability monitoring of PID control systems. This architecture deploys a lightweight probabilistic digital twin and a real-time prediction engine at the edge, enabling millisecond-level processing of sensor data and online reliability calculations, ensuring low-latency response for critical controls. The cloud platform is responsible for deep mining of massive historical data and global model optimization, and achieves knowledge sharing among multiple systems through federated learning. An adaptive communication strategy is adopted between the edge and the cloud to dynamically balance real-time performance and computational accuracy, reducing reliability early warning latency from seconds in traditional cloud platform solutions to milliseconds, providing a highly reliable and scalable monitoring infrastructure for industrial sites.

[0145] This system utilizes a calibrated probabilistic digital twin to perform short-term (hourly), medium-term (daily), and long-term (monthly) reliability predictions in parallel. Short-term predictions serve real-time monitoring and emergency response, medium-term predictions guide maintenance planning and resource allocation, and long-term predictions support life assessment and strategic planning. The prediction results at each scale are integrated through a unified interface, taking into account equipment criticality and cost factors to generate optimal maintenance strategy recommendations. It also features self-monitoring capabilities for predictive performance, forming a complete predictive maintenance closed-loop management system.

[0146] Example 3, as Figure 4 As shown, this embodiment also provides an electronic device, including:

[0147] One or more processors;

[0148] One or more memories are used to store one or more computer programs, and one or more processors are used to execute the one or more computer programs stored in the memories, so that the one or more processors perform the features / steps of a real-time reliability prediction and adaptive maintenance method for a PID control system as described in Embodiment 1.

[0149] Furthermore, at the hardware level, the device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other necessary hardware. One or more embodiments of this application can be implemented in software, for example, by the processor reading the corresponding computer program from the non-volatile memory into memory and then running it.

[0150] Of course, in addition to the software implementation, one or more embodiments of this application do not exclude other implementation methods, such as logic devices or a combination of hardware and software. That is to say, the execution subject of the above processing flow is not limited to each logic module, but can also be hardware or logic devices.

[0151] In Example 4, those skilled in the art will understand that all or part of the features / steps of the above method embodiments can be implemented by methods, data processing systems, or computer programs. These features can be implemented without hardware, entirely in software, or a combination of hardware and software. The aforementioned computer program can be stored in one or more computer-readable storage media. When the computer program is executed (e.g., by a processor), it performs the steps of the above-described embodiment of a real-time reliability prediction and adaptive maintenance method for a PID control system.

[0152] The aforementioned storage media capable of storing program code include: static disks, solid-state drives, random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), optical storage devices, magnetic storage devices, flash memory, magnetic disks or optical disks, and / or combinations of the above devices, that is, they can be implemented by any type of volatile or non-volatile storage devices or combinations thereof.

[0153] Example 5:

[0154] This embodiment provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements a real-time reliability prediction and adaptive maintenance method for a PID control system as described in Embodiment 1.

[0155] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the processes of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0156] The above description is merely a preferred embodiment of the present invention. Those skilled in the art will understand that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the present invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A PID control system real-time reliability prediction and adaptive maintenance method, characterized in that, The steps include: According to the current data of each component obtained from the PID physical system in real time, a probabilistic digital twin is generated for each component, and the current data includes key variable data, control instruction data and / or external intervention data of each component; According to the historical data of each component, the degradation process of the probabilistic digital twin corresponding to each component is parameter calibrated, and the historical data includes the drift coefficient of each component under stress conditions and the diffusion coefficient in the degradation process; The calibrated probabilistic digital twin is used to predict the reliability of the PID physical system at a future time, and when the prediction result is less than the adaptive warning threshold, a maintenance warning is triggered; wherein the adaptive warning threshold is adjusted by the variance and gradient of the prediction result; The parameter calibration of the degradation process of the probabilistic digital twin corresponding to each component according to the historical data of each component includes: At the initial time, a plurality of particles are initialized for the to-be-estimated parameters of each probabilistic digital twin, each particle corresponds to an initial value and a weight of the to-be-estimated parameter, and the probability of each particle at the initial time is calculated according to the given distribution and initial value of the to-be-estimated parameter; The transition probability of each particle at the next time is predicted according to the probability of each particle at the initial time, and the parameter value of each particle at the next time is obtained according to the transition probability; The observation probability of each particle at the next time is obtained according to the parameter value of each particle at the next time, the given observation vector and the given observation vector distribution; The weight at the next time is updated according to the parameter value of each particle at the next time and the weight at the initial time; The number of effective particles is determined according to the weights corresponding to all particles, and when the number of effective particles is less than a given number, the step corresponding to the initial time is returned, and the above steps are sequentially executed until the number of effective particles is not less than the given number; The calibrated probabilistic digital twin is used to predict the reliability of the PID physical system at a future time, including: The performance margin of each probabilistic digital twin is determined according to the actual steady-state error corresponding to the cumulative degradation amount determined as the degradation time of each probabilistic digital twin at a future time, the steady-state error under the rated operating condition, the maximum steady-state error threshold allowed, and the minimum steady-state error threshold allowed; Under the condition of given observation data up to the current time, the probability that the performance margin of all probabilistic digital twins from the current time to a future time is always greater than zero is the reliability of the PID control system from the current time to a future time.

2. The PID control system real-time reliability prediction and adaptive maintenance method according to claim 1, characterized in that, The probabilistic digital twin includes a system state output vector equation and an observation output vector equation, wherein The state output vector equation is obtained according to the vectorized key variables, control input vector and process noise with a given distribution of each component; The observation output vector equation is obtained according to the vectorized key variables and observation noise with a given distribution of each component.

3. The method of claim 2, wherein the PID control system real-time reliability prediction and adaptive maintenance method is characterized by, The cumulative degradation amount of each component at the current time is determined according to the drift coefficient and the equivalent aging time of a key variable of each component at the current time, and a dynamic evolution probability digital twin of each component is obtained.

4. The method of claim 3, wherein the PID control system real-time reliability prediction and adaptive maintenance method is characterized by, The cumulative degradation amount includes a standard Brownian motion as a variable.

5. The method of claim 1, wherein the PID control system real-time reliability prediction and adaptive maintenance method is characterized by, The remaining useful life of the PID control system is determined according to the reliability, and the remaining useful life is a time span between a future time when the reliability first falls below a given threshold and the current time.

6. A PID control system real-time reliability prediction and adaptive maintenance system, characterized in that, The PID physical subsystem is arranged at a physical layer, the probability digital twin subsystem is arranged at an edge layer, and the centralized management and analysis subsystem is arranged at a cloud layer. The PID physical subsystem is configured to collect real-time data corresponding to each component of the PID physical subsystem when the PID physical subsystem is running, and the real-time data includes key variable data, control instruction data, and / or external intervention data of the component. The probability digital twin subsystem is configured to receive the real-time data of the PID physical subsystem, implement the PID control system real-time reliability prediction and adaptive maintenance method according to any one of claims 1 to 5 according to the real-time data, and display a prediction result and an early warning strategy locally. The centralized management and analysis subsystem is configured to mine historical data required by the probability digital twin subsystem, optimize parameters of the probability digital twin according to the prediction result of the probability digital twin subsystem, and issue an early warning strategy to the probability digital twin subsystem.

7. An electronic device, comprising: The PID control system real-time reliability prediction and adaptive maintenance method comprises: at least one processor; at least one memory for storing processor-executable instructions; wherein the at least one processor implements the PID control system real-time reliability prediction and adaptive maintenance method according to any one of claims 1 to 5 by executing the executable instructions.

8. A computer program product, characterised in that, The PID control system real-time reliability prediction and adaptive maintenance method comprises a computer program or instructions, which are executed by a processor to implement the PID control system real-time reliability prediction and adaptive maintenance method according to any one of claims 1 to 5.

Images