Intelligent operation and maintenance and redundancy control system for water conservancy gate based on digital twinning

Abstract

The application belongs to the technical field of control systems, and discloses a water conservancy gate intelligent operation and maintenance and redundancy control system based on digital twinning, which comprises a model calibration module, a self-calibration controller, a digital twinning model, an environmental disturbance classifier, a state decision module and a control execution unit. The system generates a theoretical energy response reference benchmark representing physical laws by using the digital twinning model calibrated online. The deviation characteristics of the actual energy response of the driving motor from the reference benchmark are used as the main decision basis, and the structural vibration signal is introduced to provide bypass evidence to distinguish physical faults and extreme environmental loads. The application establishes a closed-loop verification relationship among the instruction, the theoretical energy response and the actual energy response. The main decision in the energy domain and the bypass information in the vibration domain are associated and analyzed to provide a judgment basis based on physical laws for the control system when facing multi-source information conflicts.

Description

Technical Field

[0001] This invention relates to an intelligent operation and maintenance and redundancy control system for hydraulic gates based on digital twins, belonging to the field of control system technology. Background Technology

[0002] Currently, a common approach to automated control systems for large-scale water conservancy facilities is to construct a digital twin model and combine it with physical sensors such as position and torque sensors deployed on the equipment itself to form a cyber-physical fusion monitoring and control loop. As long as the feedback information from each sensor is consistent within a preset range and does not deviate significantly from the model's operating state, the system determines that its understanding of the physical entity's state is accurate and executes subsequent control commands accordingly. However, for facilities like water gates that need to operate reliably for a long time under complex conditions, the aforementioned control method presents a potential operational uncertainty when facing some slowly changing physical anomalies. For example, when a gate freezes due to low temperatures or is partially jammed by foreign objects, the drive motor increases its output to overcome the obstruction. The associated position sensor continues to provide expected displacement data as it rotates with the motor, and the digital twin model extrapolates based on this data. In this case, even if no independent faults occur in any of the information units of the entire control system and the feedback information is consistent, a system state judgment that does not match the physical reality will be formed, and inappropriate forces will continue to be applied to the equipment based on this judgment.

[0003] To address such issues, simply increasing the number of sensors or improving their measurement accuracy does not solve the fundamental problem. This is because when physical conditions cause multiple sensors to produce unexpectedly consistent feedback, increasing the number of information sources may actually reinforce erroneous judgments. Furthermore, even at the control method level, existing technologies largely focus on macroscopic predictions of future states, lacking an internal verification mechanism for the authenticity of current information. For example, Chinese invention patent CN118172036B discloses a method and system for intelligent operation and maintenance of water conservancy projects based on digital twins, which combines current water conservancy datasets with future weather data. The method inputs weather forecast data into a deep learning model to predict the state of water resources at a future point in time, and assesses flood or structural safety risks by comparing the current and future digital twin models. The essence of this method is to extrapolate future trends. The starting point of its entire logical chain, namely the first water resources dataset currently acquired, is assumed to be accurate. This leads to its inability to identify abnormal working conditions such as physical blockages, because under such conditions, the system acquires a set of initial data that appears normal but is completely inconsistent with physical facts. Any future predictions based on this will be wrong, and it will be unable to respond correctly to the physical risks that are currently occurring.

[0004] This phenomenon points to a deeper technical problem: the control system lacks a verification mechanism based on fundamental physical laws, independent of sensor feedback information itself, to determine the logical consistency between control commands, physical entity responses, and sensor feedback. Existing technologies have the following shortcomings: 1. The control system lacks a final method for determining the overall effectiveness of a multi-sensor information group. When all sensors are systematically misled by the same physical phenomenon, the system lacks the ability to detect this state. 2. Fault diagnosis logic relies on threshold analysis of single sensor signals, failing to identify complex abnormal states where multiple sensor readings are individually normal but physically contradictory. 3. Redundancy switching is usually triggered by explicit hardware or communication failures, lacking an effective identification and response mechanism for operational risks arising from system decisions based on data inconsistent with physical facts. Therefore, how to establish an inherent adjudication mechanism based on physical laws for the control system, enabling it to determine the overall effectiveness of the sensor information group and distinguish between the true state of the physical entity and the system's internal state determination results, thereby maintaining decision reliability under any operating condition, becomes the technical problem this invention aims to solve. Summary of the Invention

[0005] This invention provides an intelligent operation and maintenance and redundancy control system for hydraulic gates based on digital twins. Its main purpose is to solve the problem that existing control systems, due to their lack of inherent ability to determine the overall effectiveness of sensor information groups, may make incorrect decisions based on state judgments that do not conform to physical facts when faced with specific physical anomalies.

[0006] To achieve the above objectives, this invention provides an intelligent operation and maintenance and redundancy control system for hydraulic gates based on digital twins, the system comprising:

[0007] A model calibration module is configured to apply a sequence of diagnostic signals to the drive motor of the hydraulic gate under preset conditions, and simultaneously collect the dynamic response data of the drive motor and the physical state data of the hydraulic gate to calculate a set of parameters characterizing the current physical characteristics of the hydraulic gate.

[0008] A self-calibrating controller is configured to identify one or more standard operating profiles during the normal operation of a hydraulic gate, and to collect actual energy response time series data of the drive motor when the hydraulic gate performs the standard operating profile, and to re-identify and update the parameter set in the background based on the actual energy response time series data collected multiple times.

[0009] A digital twin model is configured to receive an updated set of parameters and, based on the control command and the set of parameters, calculate and generate a time-series envelope representing the theoretical energy response of the drive motor under the control command as a reference benchmark.

[0010] An environmental disturbance classifier is configured to acquire the structural vibration signal of a hydraulic gate and output a classification result characterizing the type of disturbance source based on the spectral characteristics of the structural vibration signal.

[0011] A state adjudication module is configured to call classification results to distinguish whether the deviation is caused by changes in environmental load or changes in the physical characteristics of the hydraulic gate when the real-time actual energy response time series of the drive motor deviates from the reference benchmark beyond a preset range, and generate a structured state adjudication result based on this distinction result.

[0012] A control execution unit is configured to select and execute the corresponding instruction set from the instruction set library based solely on the state adjudication result.

[0013] Preferably, the state adjudication module is further configured to: generate a state adjudication result that confirms the system's operation is self-consistent when the real-time actual energy response time sequence of the drive motor does not deviate from the reference benchmark beyond a preset range; and the control execution unit is configured to, upon receiving the state adjudication result that confirms the system's operation is self-consistent, verify that a sensor feedback used to characterize the physical state of the hydraulic gate is reliable, and execute closed-loop control based on the sensor feedback.

[0014] Preferably, the status adjudication module is further configured to: generate a status adjudication result indicating a physical anomaly when the deviation of the classification result is caused by a change in the physical characteristics of the hydraulic gate; and the control execution unit is configured to execute a set of instructions for triggering a safety protection mode upon receiving the status adjudication result indicating a physical anomaly, the set of instructions including suspending the current control task and outputting alarm information.

[0015] Preferably, the status adjudication module is further configured to: generate a status adjudication result indicating that a control channel abnormality has occurred when the value of the real-time actual energy response time sequence of the drive motor is much lower than the lower limit of the value defined by the reference benchmark and a control command has been issued; and the control execution unit is configured to execute a set of instructions for activating redundant control channels when it receives the status adjudication result indicating that a control channel abnormality has occurred.

[0016] Preferably, the environmental disturbance classifier is configured to output classification results according to the following rules: perform spectral analysis on the structural vibration signal to obtain the distribution of its energy in different frequency bands; if the energy is randomly distributed in a wide frequency band, the output classification result is environmental load change; if the energy exhibits concentrated peaks or harmonics in one or more specific frequency bands, the output classification result is change in the physical characteristics of the hydraulic gate.

[0017] Preferably, the model calibration module is further configured to: synchronously collect and analyze the structural vibration signals of the hydraulic gate under different health conditions under preset conditions, so as to establish a baseline vibration spectrum feature library that defines the frequency band distribution pattern of vibration energy under the health condition; and the environmental disturbance classifier is configured to compare the spectrum analysis results of the real-time acquired structural vibration signals with the baseline vibration spectrum feature library to identify peaks or harmonics in a specific frequency band.

[0018] Preferably, the system also includes a system health assessment module, which is configured to: continuously collect the residual between the real-time actual energy response time series of the drive motor and the reference benchmark without deviation exceeding a preset range, form a residual time series and store it in a historical database; and analyze the long-term evolution trend of the statistical characteristics of the residual time series in the historical database at a preset period to generate an assessment result characterizing the predictive health status of the hydraulic gate.

[0019] Preferably, the system health assessment module is configured to calculate and track at least the following statistical characteristics: the drift trend of the mean of the residual time series over time, which is used to characterize the existence of unidirectional deterioration factors; and the expansion trend of the variance of the residual time series over time, which is used to characterize the degree of decline in the operational stability of the hydraulic gate.

[0020] Preferably, the system health assessment module is further configured to: calculate a health index based on long-term evolution trend data of statistical characteristics using preset weighted fusion rules. The weighted fusion rule is set as follows: ,in, The first risk value is calculated based on the drift trend of the residual mean. The second risk value is calculated based on the expansion trend of the residual variance. and The preset weighting coefficients are used; and the health index is used as the basis. The historical evolution curve is extrapolated to predict the time point when it falls below the maintenance warning line, and predictive maintenance recommendations are generated.

[0021] Preferably, the self-calibration controller is configured to: learn and define a set of standard operating profiles uniquely identified by the start state, end state, and control command mode during the initial calibration phase of the model calibration module, and associate a set of initial energy response characteristics with each standard operating profile; during normal operation, when a match is detected between the current operation command and a standard operating profile in the standard operating profile library, a data acquisition is triggered to record the complete control command and actual energy response time series data pair; and after the number of data pairs aggregated for the same standard operating profile reaches a preset statistical significance threshold, the parameter set in the digital twin model is re-identified and updated in the background.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. By working collaboratively with the model calibration module and the state adjudication module, a closed-loop verification relationship is established among the control command, theoretical energy response, and actual energy response. During system operation, the state adjudication module does not judge whether the actual energy response exceeds a certain fixed threshold in isolation. Instead, it continuously compares the actual energy response with the theoretical energy response reference benchmark predicted in real time by the digital twin model based on the current control command. This comparison result is directly used to distinguish whether the deviation of the system's operating state is caused by the actual change in the physical characteristics of the gate or by the abnormal transmission of the control execution channel. This provides a physical law-based judgment basis for the system's decision to terminate the task or switch to a redundant channel, avoiding the control system from executing incorrect response actions due to the inability to identify the root cause of the fault.

[0024] 2. When the state adjudication module determines that the comparison result between the actual energy response and the theoretical reference benchmark exceeds the preset range, it does not immediately make a ruling. Instead, it performs correlation analysis between the deviation information of this energy domain and the disturbance source classification result based on the spectral characteristics of the structural vibration signal output by the environmental disturbance classifier. Since rigid physical obstacles and fluid environmental loads exhibit different physical characteristics in structural vibration, the system can use the bypass information of this vibration domain to perform secondary identification of the cause of deviation in the energy domain. This enables the control system to distinguish between extreme but normal environmental loads such as flood peaks and real physical jamming faults, maintaining the effectiveness of the adjudication logic in strong disturbance environments and avoiding unnecessary shutdown protection.

[0025] 3. This system utilizes a self-calibrating controller to continuously collect and aggregate control commands and actual energy response data pairs under these standardized operating profiles of the gate. Because the dynamic processes within the standard operating profiles are highly repeatable, the aggregated data can comprehensively reflect the gate's true physical characteristics evolving over time. Based on this data, the system re-identifies and updates the physical parameters in the digital twin model in the background, enabling the theoretical energy response, serving as the adjudication benchmark, to dynamically adapt to the gradual aging process of the gate's physical entity, thus maintaining the core adjudication mechanism throughout the equipment's entire lifecycle. To ensure accuracy, in addition to the real-time adjudication logic of the status adjudication module, this system also sets up a system health assessment module running in the background. This module does not process significant fault signals that trigger alarms, but continuously collects and records the tolerable small residuals between the actual energy response and the theoretical reference benchmark during normal self-consistent system operation, and forms a residual time series. By analyzing the evolution trend of long-term statistical characteristics such as mean drift and variance expansion of the residual series, the system can identify the gradual degradation tendency caused by factors such as component wear or lubrication failure before any macroscopic performance indicators deteriorate. Attached Figure Description

[0026] Figure 1 This invention presents the architecture and data flow diagram of a closed-loop control system based on digital twins.

[0027] Figure 2 This is a logic diagram of the system operation mode and abnormal response decision of the present invention;

[0028] Figure 3 This is an interactive timing diagram of the fault diagnosis based on energy and vibration analysis of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. However, these descriptions are merely exemplary and not intended to limit the scope of protection of the present invention.

[0030] A digital twin-based intelligent operation and maintenance and redundancy control system for hydraulic gates begins its overall operation with a model calibration module. Through the collaborative processing of a self-calibrating controller, a digital twin model, an environmental disturbance classifier, and a state adjudication module, a structured state adjudication result is ultimately transmitted to the control execution unit to execute the corresponding control procedures. Simultaneously, a system health assessment module running in the background analyzes the long-term performance evolution trend of the system, collectively forming a complete control and operation closed loop from real-time fault adjudication to predictive health assessment. In the field of control system technology, when multiple physical sensors on which the system relies are systematically misled by the same external physical phenomenon, the control system lacks an inherent verification mechanism based on fundamental physical laws, independent of the sensor information itself, to determine the logical consistency of the entire information set. To address this issue, this system first establishes an initial benchmark representing physical laws for the entire adjudication system through a model calibration module. This model calibration module is configured to, under preset conditions, such as when the gate is confirmed to be in a healthy state after annual maintenance, transmit the data to the control unit. A diagnostic signal sequence is applied to the drive motor of the hydraulic gate. This sequence is a composite command set covering different speeds and load ranges, such as opening the gate to 10% of its travel at 5% of its rated speed, then opening it to 40% of its travel at 20% of its rated speed, and applying simulated step load commands during this process. Simultaneously with the application of this diagnostic signal sequence, the module synchronously collects the dynamic response data of the drive motor at a sampling frequency of no less than 1kHz, especially the time series data of its three-phase current and power, as well as the physical state data of the hydraulic gate, such as the position and speed information obtained through the encoder linked to the motor. Then, the model calibration module uses the collected data and employs a parameter estimation method based on the least squares method to calculate a set of physical parameters that can characterize the command energy response characteristics of the hydraulic gate in its current healthy state. This parameter set includes at least key physical parameters such as the load inertia equivalent to the motor shaft end, the dynamic and static friction coefficients, and the water resistance coefficients at different opening degrees. The completion of this calculation process enables the system to obtain an initial mathematical expression that can describe the physical characteristics of the gate.

[0031] However, the accuracy of any model built upon initial calibration will diminish with the long-term operation and wear of the physical entity, directly affecting the long-term reliability of the core decision-making mechanism. Therefore, the system is equipped with a self-calibrating controller to automatically address this model timeliness issue online. During its initial calibration phase, the self-calibrating controller analyzes historical operating data to learn and define a set of standard operating profiles uniquely identified by the initial state, ending state, and control command mode. For example, opening from the fully closed position at 15% of the rated speed to 30% of the flood control opening constitutes a standard operating profile, and each profile is associated with a set of initial energy response characteristics. During the routine operation of the hydraulic gate, the self-calibrating controller continuously monitors operating commands. When it identifies a match between the current command and a profile in the standard operating profile library, it triggers a chance-based data acquisition, recording the complete control command and actual energy response time series data pair for this standardized operation. When the number of data pairs aggregated for the same standard operating profile reaches a preset statistical significance threshold, such as 100 times, the self-calibrating controller calls least squares parameter estimation in the background. The process utilizes this data, rich in the latest physical characteristics, to re-identify and update the parameter set in the digital twin model. Through this opportunistic self-calibration using daily operation, the digital twin model, serving as the adjudication benchmark, can remain synchronized with physical reality. Correspondingly, a digital twin model receives this updated parameter set. Its core task is not to perform 3D visualization reproduction, but to act as a predictor of physical laws in the control loop. Based on any control command to be issued by the control execution unit and the updated parameter set within the model, it solves the rigid body dynamics equations of the system in real time and generates a time-series envelope characterizing the theoretical energy response of the drive motor under that command as a reference benchmark. This envelope defines the reasonable fluctuation range of motor current or power under the current command and gate health state. For example, for a command to start at 10% speed, the generated envelope may have a peak current of 50A at startup with an allowable fluctuation of ±5A, and a current of 20A during steady-speed operation with an allowable fluctuation of ±2A. This dynamically generated time-series envelope with clear upper and lower limits constitutes the benchmark for subsequent state adjudication.

[0032] When a control system faces a significant deviation in energy response, the root cause may stem from two different physical processes: one is physical impedance caused by rigid jamming, and the other is external environmental load caused by the enormous water pressure during a flood peak. While their energy manifestations may be similar, they exhibit different characteristics in structural vibration. Therefore, the system is equipped with an environmental disturbance classifier to provide independent corroboration based on bypass information. This classifier acquires the structural vibration signal of the hydraulic gate through one or more accelerometers and uses a spectrum analysis method based on Fast Fourier Transform (FFT) to obtain the energy of the vibration signal at 0.The energy distribution is measured within a frequency band from 1 Hz to 500 Hz. The classification rules are set as follows: if the energy exhibits a random distribution without significant peaks within a wide frequency band, such as between 50 Hz and 300 Hz, the output classification result is environmental load change; conversely, if the energy exhibits concentrated peaks exceeding three standard deviations from the baseline in one or more specific frequency bands, such as near the equipment's natural frequency of 25 Hz and its harmonic frequency of 50 Hz, the output classification result is change in the physical characteristics of the hydraulic gate. It should be noted that the vibration baseline used for comparison is synchronously collected and analyzed by the model calibration module under initial preset conditions. The system analyzes structural vibration signals under different healthy operating conditions, establishing a baseline vibration spectrum feature library that defines the frequency band distribution pattern of vibration energy under healthy conditions. This mechanism enables the system to identify the root cause of abnormal energy response, avoiding unnecessary downtime during critical moments such as flood control. The core adjudication logic of the system is executed by a state adjudication module, which continuously compares the real-time actual energy response time series of the drive motor with the theoretical energy response reference benchmark generated by the digital twin model. Its adjudication procedure is as follows: when the real-time actual energy response time series falls within the envelope defined by the reference benchmark, i.e., its deviation is small... At a preset 5% tolerance, the module generates a state decision result that confirms the system's self-consistent operation. In this state, the control execution unit is authorized to verify the reliable feedback from sensors characterizing the physical state of the hydraulic gate, such as the position encoder, and executes closed-loop control based on this sensor feedback. When the real-time actual energy response time series deviates beyond the preset range, for example, if the actual current continuously exceeds the theoretical envelope upper limit by 20% for more than 200ms, the state decision module does not immediately make a decision. Instead, it calls the classification result of the environmental disturbance classifier. If the classification result indicates that the deviation is due to changes in the physical characteristics of the hydraulic gate... If the vibration spectrum exhibits structurally abnormal characteristics, the module generates a status decision indicating a physical anomaly. Furthermore, if the real-time actual energy response time series value of the drive motor is significantly lower than the lower limit set by the reference benchmark—for example, if the motor current remains below 10% of the static current for 500ms after an start command has been issued—the module directly generates a status decision indicating a control channel anomaly. This structured decision, presented as a unique numerical code (e.g., 0 for self-consistency, 1 for physical anomaly, 2 for control channel anomaly), is transmitted to the control execution unit.

[0033] Furthermore, a control execution unit is configured to select and execute a corresponding instruction set from a preset instruction set library based on the state ruling result. For example, when receiving ruling result 1 indicating a physical anomaly, the control execution unit immediately selects and executes the instruction set from the library to trigger the safety protection mode. This instruction set includes suspending the current control task, executing a motor reversal unloading command, and outputting alarm information to the main control system. When receiving ruling result 2 indicating a control channel anomaly, the control execution unit executes the instruction set to activate redundant control channels, such as switching the control signal from the main PLC channel to the backup DCS channel and alerting the maintenance terminal to a main channel link failure. This deterministic response mechanism based on physical law ruling results avoids the control system from executing incorrect response actions when information sources conflict. Finally, to improve the system's maintenance capabilities from passive response to proactive prediction, the system also includes a system health assessment module that runs asynchronously in the background. When the state adjudication module determines that the system is operating consistently, this module continuously collects the normalized residuals between the actual energy response and the theoretical reference benchmark center value, forming a residual time series and storing it in a historical database. It then analyzes the long-term evolution trend of the statistical characteristics of the residual time series over the past 30 days in the historical database on a longer time scale, such as every 24 hours. The core analytical logic involves calculating and tracking at least two statistical characteristics: first, the drift trend of the residual time series mean over time, obtained through linear regression analysis of the daily mean, whose slope directly represents the existence of a unidirectional deterioration factor, such as continuously increasing friction; second, the expansion trend of the residual time series variance over time, also obtained through linear regression analysis of the daily variance, whose slope represents the degree of decline in the operational stability of the hydraulic gate. Based on these long-term evolution trend data of statistical characteristics, the module calculates a health index using preset weighted fusion rules. Its rules are set as follows ,in, For health index, The first risk value is calculated based on the slope of the drift trend of the residual mean. The second risk value is calculated based on the slope of the expansion trend of the residual variance, while and The system uses preset weighting coefficients, for example, set to 0.6 and 0.4 respectively, to determine the health index. The historical evolution curve is extrapolated to predict the time point when it may fall below the preset maintenance warning line, and predictive maintenance recommendations containing information on potentially deteriorated components are generated.

[0034] Example 1: The technical solution of the present invention operates as follows in a specific scenario of a hydraulic gate: Under a low water temperature condition during winter, the hydraulic gate receives a control command from the upper-level dispatching system for regulating the ecological flow of the downstream river, requiring the gate to be opened from the fully closed position to a target opening of 2%. At this time, due to the formation of a thin layer of ice between the gate sill and the gate leaf that is difficult to detect through conventional video monitoring, or a log with a density close to that of water being partially stuck in the gate slot on one side, the physical movement of the gate is subjected to an unexpected external obstacle; after receiving the control command, the control execution unit... When the drive circuit of the drive motor is connected, the motor begins to output torque according to the preset low-speed torque curve. However, due to the existence of the aforementioned physical obstacles, the actual physical displacement of the gate leaf does not occur, or its displacement is much smaller than the amount expected by the control command. In this case, an encoder mechanically linked to the drive motor, because it measures the rotation of the motor rather than the actual displacement of the gate body, feeds back physical state data that the gate is opening at the expected speed. After receiving this data, the conventional control system will determine that the system is operating normally and continue to issue control commands to apply forces that may cause structural damage to the physical entity.

[0035] In this embodiment, when the control execution unit issues the opening command, the digital twin model, based on the control command and the parameter set updated online by the internal self-calibration controller, calculates in real time and generates a time-series envelope characterizing the theoretical energy response of the drive motor under the command as a reference benchmark. This envelope defines the normal fluctuation of the motor current within a low value range when performing this 2% opening task under a healthy state without jamming. Simultaneously, the state adjudication module synchronously acquires the real-time actual energy response time series of the drive motor at high frequency and continuously compares it with the aforementioned reference benchmark. The system then detects that the actual motor current increases dramatically in a short period of time, continuously exceeding the upper limit of the theoretical energy response time series envelope by up to 300%, forming a significant deviation in the energy domain. At this time, the state adjudication module does not immediately determine this deviation as a physical fault, but instead calls the classification results of the environmental disturbance classifier. This classifier performs spectral analysis on the gate structure vibration signal obtained by the accelerometer. The analysis results show that the vibration energy does not exhibit the broadband random distribution characteristics unique to flood peaks or large-flow water impacts, but rather... At a specific high-frequency band related to structural friction, a concentrated energy peak with harmonic characteristics is observed. Upon receiving this classification result, the state adjudication module correlates the deviation in the energy domain with the characteristics of the vibration domain, ultimately generating a structured state adjudication result that identifies a physical anomaly, and transmits it to the control execution unit. Upon receiving the state adjudication result identifying a physical anomaly, the control execution unit immediately uses this result as the basis for decision-making, selecting and executing a set of instructions from the instruction set library to trigger the safety protection mode. This set of instructions includes stopping the current opening instruction, executing a small reverse torque instruction to release structural stress, and simultaneously sending an alarm containing physical jamming diagnostic information to the central control room and the mobile terminals of maintenance personnel. In this way, the system not only avoids physical damage to the gate or drive mechanism caused by the continuous application of improper forces, but also provides a direct basis for decision-making for subsequent troubleshooting work. The entire process transforms a judgment method that relies on the consistency of information from multiple sensors into a control method that uses fundamental physical laws as a benchmark and verifies the logical self-consistency between control commands and the energy response of the physical world.

[0036] To objectively verify the technical advantages of the present invention in distinguishing between anomalies caused by changes in the physical characteristics of hydraulic gates and anomalies caused by changes in extreme environmental loads, and to highlight the non-obviousness of the present invention's technical solution compared to conventional ideas in the prior art, the following comparative examples are provided.

[0037] Comparative Example 1: This comparative example aims to reproduce an optimized scheme that is closest in function to what a person skilled in the art can construct based on existing technology. The system architecture of this scheme is basically the same as that of the aforementioned embodiments. It also uses the energy response deviation of the drive motor as the main criterion and introduces structural vibration signals for auxiliary judgment. The essential difference is that the function of its environmental disturbance classifier is replaced by a vibration intensity analysis module. The technical logic of this module is: to collect structural vibration signals through an accelerometer and calculate the total energy value or root mean square value of the signal within a preset time window. When the total energy value exceeds a preset fixed threshold used to distinguish normal operation from significant vibration, it is determined that there is a risk of physical anomaly. This scheme represents a conventional technical approach, which is that the simultaneous occurrence of large energy deviation and large vibration intensity indicates a physical fault.

[0038] To test the effectiveness of the solution under real and complex working conditions, the same 1:10 scale test platform as in Embodiment 2 of this invention was used, and two typical working conditions that exhibit severe energy response and vibration intensity were simulated. Working condition A (physical jamming): During the gate opening process, a mechanical stop driven by a servo motor applies an instantaneous rigid physical obstruction to the gate leaf; Working condition B (flood peak passage): A high-velocity, high-turbulence water flow impact is formed in front of the gate through a high-power water pump and a guide plate, and the peak load generated is consistent with that of working condition A. The test process and results: In the comparative system, the alarm threshold of the vibration intensity analysis module was calibrated by collecting vibration data from 100 standard operating profiles under healthy operating conditions and taking three times the standard deviation of the maximum total energy value to ensure that it has sufficient anti-interference capability under normal disturbances. Subsequently, the system executed the start command under operating conditions A and B respectively. The status adjudication module continuously monitored the energy response deviation and called the results of the vibration intensity analysis module when the deviation exceeded the range. The key response data and the final adjudication results are recorded in Table 1 below.

[0039] Table 1: Response data and decision results of Comparative Example 1 under simulated operating conditions.

[0040]

[0041] The experimental results show that under condition A (physical jamming), the rigid obstruction simultaneously generated a huge energy deviation and severe structural vibration, and the comparative system made a correct fault judgment. However, under condition B (flood peak passage), although the huge water flow impact is a normal external environmental load in nature, it also caused extremely high energy response deviation and total vibration energy far exceeding the threshold. Since the system can only judge the presence or intensity of vibration, but cannot analyze its internal physical causes, it confused the two conditions. At the critical moment of flood control, the normal load was judged as a physical fault and unnecessary shutdown protection was executed, which constituted a serious misjudgment. This result proves that simply superimposing energy information and vibration intensity information cannot solve the inherent judgment dilemma of existing technology when facing multi-source information conflict.

[0042] Example 2: To objectively verify the effectiveness of the technical solution of this invention in distinguishing between anomalies caused by changes in the physical characteristics of a hydraulic gate and anomalies caused by changes in extreme environmental loads, a test platform containing a scale model was built. This platform replicated the hydraulic gate and its drive mechanism at a 1:10 scale. The drive motor was an AC asynchronous motor, equipped with a Hall effect current sensor for collecting the three-phase current of the motor (sampling frequency set to 1kHz) and a microelectromechanical system (MEMS) accelerometer for collecting vibration signals from the gate structure (sampling frequency set to 5kHz). Data acquisition and control algorithm execution were performed using a processor capable of performing Fast Fourier Transform (FFT) operations. The experiment included two control groups and one test group. Control group 1 employed a judgment mechanism that relied solely on energy response deviation; that is, an alarm was triggered when the real-time actual energy response time series exceeded the envelope of the theoretical energy response time series, without introducing structural vibration signals for analysis. Group 2, based on control group 1, added a simple threshold judgment on the total energy of the structural vibration signal, i.e., when the energy response deviates and the total vibration energy exceeds a preset value, an alarm is triggered. The test group adopted the complete technical solution of this invention, i.e., when the energy response deviates, an environmental disturbance classifier is called to classify the spectral characteristics of the structural vibration signal, and a final decision is made based on the classification results. The test process simulated two typical challenging working conditions: Condition A, simulating physical jamming, in which a mechanical stop driven by a servo motor applies an instantaneous, rigid physical obstruction to the gate leaf during the gate's opening command; Condition B, simulating a flood peak, in which a high-velocity, high-turbulence water flow impact is formed in front of the gate through a high-power water pump and a guide plate, and the equivalent load generated is numerically consistent with the peak load generated by the physical jamming in Condition A. Under these two working conditions, the response behavior of the three test groups was recorded respectively. To clearly show this differentiation process, the key data are recorded in Table 2.

[0043] Table 2: A comparison of response data and adjudication results for different test groups under simulated operating conditions.

[0044]

[0045] Referring to Table 2, Control Group 1, relying solely on energy response deviation for judgment, identified the massive load as a physical anomaly in both Condition A and Condition B, triggering a shutdown alarm. The alarm in Condition B was a false alarm caused by extreme environmental load. Control Group 2 introduced the judgment of total vibration energy, but because the total vibration energy in both conditions was significantly higher than normal, it failed to effectively distinguish between the two conditions, also resulting in false alarms. In Condition A, the experimental group's environmental disturbance classifier, through spectrum analysis, identified a concentrated peak in the structural vibration signal within a specific frequency band of 125Hz, classifying it as a change in the physical characteristics of the hydraulic gate, and subsequently, the state adjudication module... The correct decision was made regarding the occurrence of a physical anomaly, triggering a shutdown protection mechanism. Under condition B, the energy response deviation of the test group was also captured by the state adjudication module. However, after performing spectral analysis on the structural vibration signal, the environmental disturbance classifier found that its energy was randomly distributed in a wide frequency band from 40Hz to 250Hz without a specific peak, which conformed to the extreme environmental load pattern defined in the baseline vibration spectrum feature library. Therefore, the output classification result was environmental load change. Based on this classification result, the state adjudication module attributed the current energy deviation to normal external load, generating a state adjudication result that determined the system operation to be self-consistent. The control system continued to execute the start-up task, avoiding unnecessary shutdown.

[0046] Example 3: This example combines Figures 1 to 3 This section describes the intelligent operation and maintenance and redundant control system for water conservancy gates based on digital twins, such as... Figure 1As shown, this illustrates the collaborative working relationship between the various modules of the system. The entire closed-loop control and operation process begins with an externally input control command. This command is sent to the digital twin model, which, based on the command and the updated parameter set, calculates and generates a reference benchmark for the theoretical energy response and transmits it to the state adjudication module. Simultaneously, the command drives the physical entity of the hydraulic gate to generate an actual physical response. After executing the command, the physical entity outputs two types of key data: the actual energy response characterizing its power consumption and the structural vibration signal characterizing its operating state. The actual energy response is fed back to the self-calibration controller for updating the model parameters in the background based on actual data under a standard operating profile, and is also sent to the state adjudication module. The structural vibration signal is input to the environmental disturbance classifier. The classifier analyzes the spectral characteristics of the structural vibration signal and outputs the disturbance source classification result, which is then provided as bypass corroborating information to the state adjudication module. As the decision-making core of the system, the state adjudication module compares the deviation between the theoretical and actual energy responses and combines the bypass information provided by the environmental disturbance classifier to make a comprehensive adjudication, generating a structured state adjudication result. This result is uniquely transmitted to the control execution unit, which selects and executes the corresponding execution action from the instruction set library based on the adjudication result, such as normal operation, safety protection, or redundancy switching. In addition, during normal system operation, the state adjudication module continuously outputs the residual sequence under normal operation to the system health assessment module, which analyzes the long-term trend of the residuals to generate predictive maintenance suggestions, which are ultimately used to formulate and execute maintenance plans.

[0047] like Figure 2 As shown, the entire system's operation is triggered by maintenance personnel or the superior system. It can initiate a one-time initial model calibration task or instruct the system to enter the regular execution gate normal operation mode. In the normal operation mode, the system executes three core background tasks in parallel, including online self-calibration of the model using actual operating data, evaluation of system health based on small residual sequences, and providing evidence for real-time decision-making by classifying environmental disturbances. The outputs of these tasks collectively serve a core online process, namely, adjudicating the system's operating status. This adjudication process, based on the comprehensive analysis results of real-time data, will trigger the activation of redundant control channels when an anomaly in the control channel is identified, and will trigger the execution of the safety protection mode when a physical anomaly is identified.

[0048] like Figure 3As shown, the process begins with the control execution unit issuing a flood control activation command to the drive motor. Due to the impact of the massive water flow, the drive motor reports that its actual current is too high, for example, reaching 150% of the baseline. After detecting this deviation in energy response, the state adjudication module does not immediately determine it as a fault. Instead, it requests disturbance source analysis from the environmental disturbance classifier. The vibration sensor then collects the vibration signal during the flood season and transmits the data to the environmental disturbance classifier. After spectrum analysis, the classifier identifies that the vibration energy is randomly distributed in a wide frequency band of 40-250Hz and generates a classification result of environmental load change. After receiving the classification result, the state adjudication module confirms that the current state is a normal load through correlation analysis and sends back a self-consistent adjudication result of system operation to the control execution unit. Finally, after confirming the reliability of the sensor feedback, the control execution unit continues to execute the flood control task and records the operating data under this extreme environment, thus completing an intelligent adjudication that avoids false alarms caused by extreme environmental load.

[0049] Example 4: After the initial deployment or major maintenance and renovation of a digital twin-based intelligent operation and redundancy control system for hydraulic gates, one engineering problem is how to calibrate a general digital twin model into a dedicated model that matches the real-time characteristics of a specific physical gate, and establish a reproducible initial benchmark and parameter thresholds for subsequent online self-calibration and status adjudication. This example discloses a procedure for debugging and calibrating the system. The execution of this procedure begins with an offline initial parameter identification phase. In this phase, the initial state must first be defined, ensuring that the hydraulic gate is in a confirmed healthy operating condition and decoupled from external loads, specifically in a waterless environment. The process involves configuring the system with a processor capable of at least 1kHz data acquisition frequency and floating-point arithmetic. Subsequently, the model calibration module is activated, applying a preset diagnostic signal sequence to the drive motor. This sequence is a sweeping sinusoidal signal that scans the frequency range from 0.1Hz to 10Hz to fully excite the gate system's response characteristics under different dynamic conditions. Within 10 minutes of applying this sequence, the system synchronously acquires the actual energy response time series data of the drive motor and the physical state data of the position sensor. Based on this set of input and output data, the least squares method in system identification is used to calculate the parameter set characterizing the initial characteristics of the physical entity.

[0050] After obtaining the initial parameter set, the procedure proceeds to the calibration stage of key thresholds. The digital twin model uses the parameter set calculated in the previous step to perform back-calculation of the theoretical energy response under the diagnostic signal sequence, and generates a time series of the theoretical energy response. The system then calculates the residual time series between the actual energy response time series and the theoretical energy response time series during the offline calibration process. By performing statistical analysis on the residual time series, its standard deviation is calculated. The state adjudication module is used to determine whether the system operation deviates from a preset range in terms of self-consistency. Its upper and lower limits are set to ±3 of the center value of the theoretical energy response time series envelope. This statistically based calibration method provides a judgment criterion coupled with specific equipment characteristics for subsequent real-time adjudication. The next step in this procedure is the learning and definition of online operating profiles. After completing offline calibration, the system enters regular operation and a 100-hour opportunistic learning mode. In this mode, the self-calibration controller records all control command sequences issued by the operator and their corresponding start and end states, and performs cluster analysis on these operation records, specifically using the K-means clustering algorithm to group operations with similar start states, end states, and control command patterns into the same category. After the learning cycle ends, the system selects the top 5 categories with the most cluster centroids, defines their characteristic command patterns and state boundaries as a set of standard operating profiles, and stores them in the database. Simultaneously, the threshold for the number of data pairs used to trigger background parameter re-identification is set to 100 times. This value is based on the fact that, in simulation testing, aggregating 100 sets of data can achieve a 95% confidence level in the parameter identification results. Through this procedure, the system... The actual operation history established a reference profile library for online self-calibration; in addition, the procedure also includes the establishment of a baseline vibration spectrum feature library; during the same period of offline initial parameter identification, the environmental disturbance classifier synchronously collects the structural vibration signals of the gate under healthy operating conditions and when responding to the diagnostic signal sequence; the system segments the collected vibration signals and performs a fast Fourier transform on each segment to obtain a series of vibration spectra representing the healthy state; the baseline vibration spectrum feature library is constructed as the envelope of the statistical mean and standard deviation of these spectra, which defines the vibration characteristics of the gate under fault-free and external environmental disturbance conditions; subsequently, in real-time adjudication, when the environmental disturbance classifier analyzes the real-time vibration signal, if the peak energy of its spectrum in a specific frequency band exceeds three times the standard deviation of the mean energy of the baseline spectrum, it is judged as a structural abnormal mode. This series of calibration and learning processes ensures that all core models, reference libraries and judgment thresholds of the entire control system have been personalized based on the specific physical entity characteristics before being put into formal operation.

[0051] Example 5: To calibrate the internal parameters of the system health assessment module, a backtracking analysis procedure based on historical data was adopted. This procedure first acquires complete operational data for a hydraulic gate that has experienced progressive wear failure, covering the six months prior to the failure, including control commands, actual energy response time series, and the final failure record. The system health assessment module uses this data, with a 30-day sliding window, to calculate the mean and variance of the residual time series daily, and performs linear regression analysis on the temporal evolution trends of these two statistical characteristics. Furthermore, the first risk value is determined. The normalized slope was determined as the trend of the residual mean drift, while the second risk value The normalized slope, weighted coefficient, is determined as the trend of residual variance expansion. and The design involves analyzing multiple historical failure cases and employing the analytic hierarchy process (AHP) to determine the contribution weights of unidirectional degradation factors (represented by mean drift) and operational instability (represented by variance expansion) to ultimately lead to physical failures. These weights are then used as the basis for determining the final physical failure outcome. and The set value is used to determine the health index. The calculations provide a traceable basis.

[0052] When faced with boundary conditions where the core control component malfunctions, the technical solution of this invention employs the following preset response method: During periodic self-checks, if the system detects that the self-calibration controller has failed to update the parameter set in the digital twin model for more than three calibration cycles, the system determines that the self-calibration function is malfunctioning. At this time, the control system is configured to automatically enter a preset degraded operation mode. In this mode, the system first locks the parameter set used by the digital twin model to the state of the last successful update, without making any further adjustments. Simultaneously, the state adjudication module uses this to determine whether the system operation deviates from a preset range based on the initial calibration within ±3... Relaxed to ±6 To a certain extent, this is compatible with prediction biases that may be caused by model aging, avoiding false alarms. In addition, the system will immediately send a level-two maintenance alarm to the operation and maintenance management platform regarding the abnormal function of the self-calibration controller, prompting the operation and maintenance personnel to intervene. This mechanism enables the system to maintain basic fault protection capabilities with higher fault tolerance even when its internal key software modules fail.

[0053] Example 6: This example discloses a self-testing procedure for verifying the health status of sensors within a control system, addressing the potential risk of decreased reliability of the entire decision logic due to sensor failure. The procedure is set to automatically execute every 24 hours when the system is in standby mode. Its core function is to verify the effectiveness of the accelerometer, a key input to the environmental disturbance classifier. After the self-testing procedure is initiated, the system first confirms that the hydraulic gate is completely stationary. Subsequently, the model calibration module applies a diagnostic electrical pulse signal to the drive motor, with a predefined energy amplitude insufficient to cause macroscopic displacement of the gate. This pulse signal is a square wave pulse with a width of 100ms and an amplitude of 5% of the rated voltage. This electrical pulse generates a weak electromagnetic and mechanical impact in the motor windings and transmission chain, which then propagates along the gate structure in the form of elastic waves, forming a predictable weak structural vibration response. During this period, the environmental disturbance classifier synchronously acquires the output signal of the accelerometer and performs spectral analysis on it.

[0054] The system compares the vibration spectrum acquired during this self-test with the standard vibration response templates pre-stored in the baseline vibration spectrum feature library, which correspond to the diagnostic electrical pulse signal. If the similarity between the acquired spectrum features and the standard template is higher than a preset 90% threshold, the system determines that the accelerometer is working normally. Conversely, if the sensor output signal is flat and unresponsive, or its spectrum features are significantly inconsistent with the standard template, the system determines that the sensor has failed or its channel has been interrupted, and immediately marks the functional status of the environmental disturbance classifier as unavailable. In this state, the control system will enter a more conservative fault handling mode. That is, in subsequent operation, once the state adjudication module detects a deviation in energy response, it will no longer call the classification results of the environmental disturbance classifier, but will default to attributing the deviation to a physical anomaly and directly execute the safety protection process of shutdown alarm, while sending a specific maintenance alarm about the accelerometer failure to the operation and maintenance platform.

[0055] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart operation and maintenance and redundancy control system for hydraulic gates based on digital twins, characterized in that, The system includes: A model calibration module is configured to apply a sequence of diagnostic signals to the drive motor of the hydraulic gate under preset conditions, and simultaneously collect the dynamic response data of the drive motor and the physical state data of the hydraulic gate to calculate a set of parameters characterizing the current physical characteristics of the hydraulic gate. A self-calibrating controller is configured to identify one or more standard operating profiles during the normal operation of a hydraulic gate, and to collect actual energy response time series data of the drive motor when the hydraulic gate performs the standard operating profile, and to re-identify and update the parameter set in the background based on the actual energy response time series data collected multiple times. A digital twin model is configured to receive an updated set of parameters and, based on the control command and the set of parameters, calculate and generate a time-series envelope representing the theoretical energy response of the drive motor under the control command as a reference benchmark. An environmental disturbance classifier is configured to acquire the structural vibration signal of a hydraulic gate and output a classification result characterizing the type of disturbance source based on the spectral characteristics of the structural vibration signal. A state adjudication module is configured to call classification results to distinguish whether the deviation is caused by changes in environmental load or changes in the physical characteristics of the hydraulic gate when the real-time actual energy response time series of the drive motor deviates from the reference benchmark beyond a preset range, and generate a structured state adjudication result based on this distinction result. A control execution unit is configured to select and execute the corresponding instruction set from the instruction set library based solely on the state adjudication result.

2. The intelligent operation and redundancy control system for hydraulic gates based on digital twins according to claim 1, characterized in that, The state adjudication module is further configured to generate a state adjudication result that confirms the system's self-consistency when the real-time actual energy response time series of the drive motor does not deviate from the reference benchmark beyond a preset range. Furthermore, the control execution unit is configured to, upon receiving the self-consistent state decision result of the system, confirm the reliability of a sensor feedback used to characterize the physical state of the hydraulic gate, and execute closed-loop control based on the sensor feedback.

3. The intelligent operation and redundancy control system for hydraulic gates based on digital twins according to claim 1, characterized in that, The status adjudication module is further configured to generate a status adjudication result indicating a physical anomaly when the deviation of the classification result is caused by a change in the physical characteristics of the hydraulic gate; and the control execution unit is configured to execute a set of instructions for triggering a safety protection mode upon receiving the status adjudication result indicating a physical anomaly, the set of instructions including suspending the current control task and outputting alarm information.

4. The intelligent operation and redundancy control system for hydraulic gates based on digital twins according to claim 1, characterized in that, The status adjudication module is further configured to generate a status adjudication result indicating that a control channel abnormality has occurred when the value of the real-time actual energy response time sequence of the drive motor is much lower than the lower limit of the value defined by the reference benchmark and a control command has been issued. Furthermore, the control execution unit is configured to execute a set of instructions to activate redundant control channels upon receiving a status decision result indicating that a control channel abnormality has occurred.

5. The intelligent operation and redundancy control system for hydraulic gates based on digital twins according to claim 1, characterized in that, The environmental disturbance classifier is configured to output classification results according to the following rules: perform spectral analysis on the structural vibration signal to obtain the distribution of its energy in different frequency bands; if the energy is randomly distributed in a wide frequency band, the output classification result is environmental load change; if the energy shows concentrated peaks or harmonics in one or more specific frequency bands, the output classification result is change in the physical characteristics of the hydraulic gate.

6. The intelligent operation and maintenance and redundancy control system for hydraulic gates based on digital twins according to claim 5, characterized in that, The model calibration module is further configured to: under preset conditions, synchronously collect and analyze the structural vibration signals of the hydraulic gate under different health conditions, so as to establish a baseline vibration spectrum feature library that defines the vibration energy frequency band distribution pattern under the health condition; Furthermore, the environmental disturbance classifier is configured to compare the spectral analysis results of the real-time acquired structural vibration signals with a baseline vibration spectral feature library.

7. The intelligent operation and redundancy control system for hydraulic gates based on digital twins according to claim 1, characterized in that, The system also includes a system health assessment module, which is configured to: continuously collect the residual between the real-time actual energy response time series of the drive motor and the reference benchmark without deviation from the preset range, form a residual time series and store it in the historical database; and analyze the long-term evolution trend of the statistical characteristics of the residual time series in the historical database at a preset period to generate an assessment result characterizing the predictive health status of the hydraulic gate.

8. The intelligent operation and maintenance and redundancy control system for hydraulic gates based on digital twins according to claim 7, characterized in that, The system health assessment module is configured to calculate and track at least the following statistical characteristics: the drift trend of the mean of the residual time series over time, which is used to characterize the existence of unidirectional deterioration factors; And the expansion trend of the variance of the residual time series over time, which is used to characterize the degree of decline in the operational stability of hydraulic gates.

9. A smart operation and redundancy control system for hydraulic gates based on digital twins as described in claim 8, characterized in that, The system health assessment module is further configured to calculate a health index based on long-term evolution trend data of statistical characteristics using preset weighted fusion rules. The weighted fusion rule is set as follows: ,in, The first risk value is calculated based on the drift trend of the residual mean. The second risk value is calculated based on the expansion trend of the residual variance. and The preset weighting coefficients are used; and the health index is used as the basis. The historical evolution curve is extrapolated to predict the time point when it falls below the maintenance warning line, and predictive maintenance recommendations are generated.

10. A smart operation and redundancy control system for hydraulic gates based on digital twins as described in claim 1, characterized in that, The self-calibration controller is configured to: learn and define a set of standard operating profiles uniquely identified by the start state, end state, and control command mode during the initial calibration phase of the model calibration module, and associate a set of initial energy response characteristics with each standard operating profile; during normal operation, when a match is detected between the current operation command and a standard operating profile in the standard operating profile library, a data acquisition is triggered to record the complete control command and actual energy response time series data pairs; and after the number of data pairs aggregated for the same standard operating profile reaches a preset statistical significance threshold, the parameter set in the digital twin model is re-identified and updated in the background.

Images