Water conservancy gate control system based on multi-source information fusion

Abstract

The invention relates to the technical field of water conservancy gate control, and discloses a water conservancy gate control system based on multi-source information fusion, and the system comprises the steps: driving a gate to complete the reciprocating motion of opening and closing, and carrying out the verification of an obtained power sequence based on physical symmetry, so as to establish a reference passing the internal quality verification; then, in actual operation, real-time power vibration is compared with the reference to diagnose the real-time operation state of the gate, by introducing a reference self-checking mechanism, the fragile mode that a traditional control system depends on initial data credible assumption is avoided, control logic is evolved into active process auditing from passive result confirmation, and the real-time operation state of the gate is diagnosed. According to the invention, the system can diagnose the health degree of the operation process of the system, thereby guaranteeing the purity of a diagnosis reference, carrying out the real-time early warning of potential operation risks, and achieving the improvement of the long-term operation reliability of hydraulic facilities.

Description

Technical Field

[0001] This invention relates to a hydraulic gate control system based on multi-source information fusion, belonging to the field of hydraulic gate control technology. Background Technology

[0002] In current water conservancy engineering applications, the reliable operation of large gates is fundamental to flood control scheduling and water resource management. The commonly used technical solution in this field involves using gate position sensors as feedback to form a closed-loop control system, with servo motors and other actuators driving the gate to achieve precise movement between preset open and closed positions. Simultaneously, to protect the drive motor, the system is typically equipped with conventional overload protection based on operating current monitoring. This technical combination, due to its straightforward structure and clear control objectives, has become the general technical foundation for achieving automated gate operation. However, the on-site operating conditions of water conservancy projects introduce a unique physical constraint to the above-mentioned general technical solution: the total load borne by the drive system consists of normal and dynamically changing hydraulic loads, and... The combination of occasional abnormal physical obstacles, such as siltation and floating debris jamming, presents an inherent limitation of existing control schemes. The sole criterion for judging whether an operation is complete is the geometric position of the gate. However, there is a lack of effective identification and analysis capabilities regarding the physical process of the power cost incurred by the drive system to achieve that position. A drive power within the normal range during high water levels in the flood season may be several times greater than the power required for severe mechanical jamming during low water levels in the dry season. This makes overload protection functions based on fixed or simple variable thresholds difficult to operate effectively in practical applications. If the threshold is set too high, it may miss damaging jamming at low water levels; if the threshold is set too low, false alarms will frequently occur at high water levels, making it difficult to perform its intended protective function.

[0003] The aforementioned limitations mean the system can only determine operation completion based on location information, but cannot detect any abnormal physical forces acting during the process. Each instance of overstress actuation caused by an obstacle encounter results in cumulative damage to the motor drive mechanism and even the gate structure itself, which cannot be monitored by the current system, creating potential operational risks. One direct technical approach is to increase the safety redundancy of the drive system and gate structure; however, this not only significantly increases engineering costs but also fails to fundamentally solve the problem of lack of abnormal event detection. In fact, stronger driving forces may exacerbate structural damage during abnormal events. The root cause of these physical limitations lies in the inherent defects of the control logic; even attempts to optimize the hardware structure using existing technologies have failed to overcome this bottleneck. For example, Chinese invention patent CN217997998U discloses a hydraulic gate control device. Although the device adds a slope waterproofing mechanism to its mechanical structure to improve the hardware protection effect, its core control logic is still at the closed-loop control level of the gate's geometric position. It only relies on the information fed back by the displacement sensor to determine whether the operation is completed. It completely lacks the ability to effectively identify and analyze the key physical process of the power cost paid by the drive system to achieve the position. It cannot distinguish between normal hydraulic load and abnormal physical obstacles, thus leaving potential operational risks to the physical structure itself.

[0004] Specifically, existing technologies have the following shortcomings: 1. Their control logic is based on normal assumptions about the driving process, and cannot identify abnormal stresses that exceed conventional physical laws experienced by the system when executing commands; 2. Their safety protection mechanisms only address the electrical parameters of the drive motor itself, rather than the overall operational health of the entire hydroelectric coupling system, lacking direct consideration of the safety of the hydraulic structure; 3. Their response to faults is passive, only interrupting the system after overload damage may have already occurred, lacking effective monitoring methods for initial faults or gradual system performance degradation. Therefore, how to enable the control system to utilize its existing information to accurately interpret the power cost borne by the drive gate in real time, and effectively separate the abnormal components caused by abnormal physical obstacles, thereby achieving online diagnosis and early warning of the gate's operating status, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a hydraulic gate control system based on multi-source information fusion. Its main purpose is to solve the problem that existing control methods cannot accurately interpret drive power, have difficulty distinguishing between normal hydraulic loads and abnormal physical obstacles, and thus cannot effectively diagnose and warn of potential operational risks of the gate.

[0006] To achieve the above objectives, the present invention provides a hydraulic gate control system based on multi-source information fusion, the system comprising: A data acquisition module is configured to simultaneously acquire a sequence of operating parameters characterizing the operating condition of the hydraulic gate and a sequence of actuator power costs characterizing its driving load during the operation of the hydraulic gate. A fingerprint construction module, connected to a data acquisition module, is configured to: instruct an actuator to drive a hydraulic gate to complete at least one reciprocating motion including an opening and closing process, and obtain the corresponding opening power cost sequence and closing power cost sequence from the data acquisition module; then, perform a symmetry check on the opening power cost sequence and closing power cost sequence, which is obtained by stripping the asymmetric power part caused by gravity or buoyancy effect in the two sequences to obtain two net resistance power sequences reflecting friction and additional resistance respectively, and perform a mirror symmetry comparison on the two net resistance power sequences; and only when the mirror symmetry comparison result meets a preset symmetry standard are the obtained opening and closing power cost sequences confirmed as valid reference data, and a reference health fingerprint is established based on the valid reference data; A real-time diagnostic module is configured to determine a baseline power cost from the baseline health fingerprint established by the fingerprint construction module based on the actual operating parameters acquired in real time, and diagnose the operating status of the hydraulic gate based on the deviation between the actual power cost acquired in real time and the baseline power cost.

[0007] Preferably, the operating condition parameter sequence includes a gate position parameter sequence, a running speed parameter sequence, and a running direction parameter sequence, as well as an upstream water level parameter sequence and a downstream water level parameter sequence associated with the hydraulic gate; the fingerprint construction module is configured to establish a benchmark health fingerprint based on the correspondence between the operating condition parameter sequence and the effective benchmark data, wherein the benchmark health fingerprint is a multi-dimensional database that maps operating condition parameters to power costs.

[0008] Preferably, the system further includes an active detection module, which is configured to: under a preset diagnostic trigger condition, the instruction executor drives the hydraulic gate to complete a preset micro-diagnostic stroke at a diagnostic speed lower than the normal operating speed of the system; the real-time diagnostic module is also configured to acquire the diagnostic power cost sequence during the diagnostic stroke and compare the diagnostic power cost sequence with a benchmark friction curve stored in a benchmark health fingerprint to assess the long-term operational health trend of the hydraulic gate.

[0009] Preferably, the reference friction curve is acquired and stored in the reference health fingerprint during the initial establishment of the reference health fingerprint by executing the detection process of the active detection module.

[0010] Preferably, the real-time diagnostic module is specifically configured to: calculate the morphological similarity distance between the actual power cost sequence formed by the actual power cost and the benchmark power cost sequence determined from the benchmark health fingerprint by using a dynamic time warping algorithm, and determine the deviation; and when the morphological similarity distance exceeds the morphological similarity threshold determined according to the statistical distribution characteristics of the benchmark health fingerprint, diagnose the operating status as an abnormal state.

[0011] Preferably, the real-time diagnostic module is further configured to calculate a health aging factor characterizing the long-term frictional changes of the system based on the comparison results between the diagnostic power cost sequence and the benchmark friction curve. Its calculation rule is as follows ,in, The current friction characteristic value is calculated from the diagnostic power cost sequence. The reference friction characteristic value is calculated from the reference friction curve; the system also includes a maintenance early warning module, which is configured to detect health aging factors. If the value exceeds 1.5 for two or more consecutive diagnostic cycles, an early warning command will be output.

[0012] Preferably, the real-time diagnostic module is specifically configured to: diagnose the operating state as a resistive abnormal state when the actual power cost exceeds the reference power cost by a preset first threshold; and diagnose the operating state as an off-load abnormal state when the actual power cost is lower than the reference power cost by a preset second threshold.

[0013] Preferably, the system also includes a failure protection module, which is configured to execute a preset failure protection action after the real-time diagnostic module detects a resistive abnormality or an underload abnormality. The failure protection action includes stopping the actuator, instructing the actuator to run in reverse, or sending an alarm message to a higher-level monitoring system.

[0014] Preferably, the real-time diagnostic module is further configured to: when the real-time acquired actual operating parameters do not have a directly mapped reference power cost in the multidimensional database, call an interpolation calculation unit. The interpolation calculation unit is configured to select multiple reference operating points in the multidimensional database that are adjacent to the actual operating parameters in the multidimensional space, and calculate the interpolated reference power cost corresponding to the actual operating parameters based on the reference power cost corresponding to the multiple reference operating points through a preset interpolation algorithm, and use the interpolated reference power cost as the reference power cost.

[0015] Preferably, the symmetry verification operation is specifically configured as follows: a gravity and buoyancy effect model is established based on the gate structure parameters and real-time hydrological parameters, and the theoretical power components generated by gravity and buoyancy during the opening and closing processes are calculated using the model; then, the corresponding theoretical power components are subtracted from the opening power cost sequence to obtain the first net resistance power sequence, and the corresponding theoretical power components are subtracted from the closing power cost sequence to obtain the second net resistance power sequence; finally, a mirror symmetry comparison is performed between the first net resistance power sequence and the second net resistance power sequence.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By collecting the driving power of hydraulic gates under specific operating conditions, a power cost benchmark corresponding to hydraulic boundary conditions is established. Based on this, before adopting any benchmark data, the method forcibly performs a symmetry verification step, that is, acquiring the power sequences of the gate opening and closing strokes respectively. After removing the influence of the two deterministic physical quantities of gravity and buoyancy, the remaining power sequence mainly reflecting frictional resistance is mirrored. Only when the two show a preset symmetry is this data segment confirmed as a valid health fingerprint. This series of technical actions makes the system no longer dependent on the reliable assumptions about the initial state, but establishes a built-in quality inspection mechanism based on physical principles for diagnosing the validity of the benchmark itself, thereby solving the potential problem that the control system may fail to diagnose all subsequent diagnoses due to learning a faulty initial sample.

[0017] 2. While utilizing the verified health fingerprint to diagnose sudden resistance during gate operation in real time, this system also includes an active system health trend assessment mechanism. Under stable operating conditions where the gate is not in a task state, this mechanism drives the gate to complete a small reciprocating motion at a preset diagnostic speed lower than the normal operating speed. Under these specific physical conditions, the main part of the driving power directly reflects the static and dynamic friction characteristics of the system, while the influence of highly dynamic variables such as hydraulic load is effectively suppressed. By comparing the diagnostic power obtained this time with the baseline friction curve recorded under the initial health state, the control system can proactively and periodically quantify the long-term evolution trend of its own mechanical lubrication and wear state when not performing routine opening and closing tasks. This expands the dimension of health management from responding to immediate failures to predicting progressive systemic risks.

[0018] 3. When performing the comparison step of real-time diagnosis, instead of comparing the instantaneous values ​​of the two sets of power data at the same location point, the dynamic time warping algorithm is preferred to calculate the overall morphological similarity between the actual operating power sequence and the power sequence in the benchmark health fingerprint database. The introduction of this algorithm changes the core of the comparison logic from pursuing absolute numerical equivalence to judging the macroscopic conformity of the inherent trend and rhythm of the two curves. Therefore, for non-faulty small speed changes caused by power grid fluctuations or controller response differences in real engineering, the system can automatically absorb its local scaling on the time axis during the comparison process, avoiding a large number of false alarms caused by such time-domain noise. This transforms the power curve-based diagnosis method from a theoretical model into a stable solution with high reliability and practicality in real complex engineering environments.

[0019] 4. By combining a benchmark self-verification mechanism based on physical symmetry with an active detection mechanism based on diagnostic perturbation, a logically closed-loop health management system with self-updating and dynamic adaptability is constructed. The former ensures that the starting point for each system learning is reliable, guaranteeing the purity of the diagnostic benchmark from the source. The latter provides the system with the ability to continuously monitor its own state as it slowly drifts from this reliable starting point. When the long-term trend monitoring shows a significant but still safe change in the system's frictional characteristics, a fingerprint update process including self-verification can be triggered again after planned maintenance confirms the restoration of the physical state. The synergistic effect of this series of mechanisms enables the entire control system to dynamically adapt to the natural performance degradation throughout its entire life cycle, and its operational reliability no longer depends solely on the ideal state at the time of initial installation. Attached Figure Description

[0020] Figure 1 This is a functional architecture diagram of the closed-loop health management system of the present invention; Figure 2 This is a schematic diagram of the health fingerprint verification principle based on power symmetry of the present invention; Figure 3 This is a timing diagram of the abnormal diagnosis and response interaction of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] The hydraulic gate control system based on multi-source information fusion claimed in this invention has an overall architecture that mainly includes a data acquisition module, a fingerprint construction module, a real-time diagnosis module, an active detection module, and a failure protection module. The data acquisition module is responsible for simultaneously acquiring the sequence of operating parameters characterizing the current operating condition of the gate and the sequence of actuator power costs characterizing the load state of the drive system during the operation of the hydraulic gate. The fingerprint construction module, based on the output of the data acquisition module, establishes a benchmark health fingerprint that reflects the operating characteristics of the gate system under a preset health state through a set of procedures based on physical symmetry verification. The core task of the real-time diagnosis module is to compare the real-time acquired power costs with the benchmark power costs retrieved from the benchmark health fingerprint based on the real-time operating conditions during the daily operation of the gate, thereby performing an immediate diagnosis of the health status of the operation process. The active detection module, as a supplement to the real-time diagnosis, provides the ability to actively quantify and evaluate the long-term performance evolution trend of the system during system idle periods. The failure protection module, as the execution end of the diagnostic results, executes preset protective actions after the system is diagnosed as being in an abnormal state. Together, they constitute a complete closed-loop control system from benchmark establishment, state monitoring, trend prediction to failure protection.

[0023] In a specific application scenario, such as the cluster scheduling and control of the flood discharge gates of a large reservoir, to address the technical problem that traditional control methods cannot accurately distinguish between normal hydraulic loads and abnormal physical obstacles such as silt and floating objects, resulting in a lack of perception of potential operational risks of the gates, the system claimed in this invention is configured to operate using the following procedure: First, in the fingerprint construction phase, to ensure the validity of the baseline data upon which all subsequent diagnoses are based, the fingerprint construction module is configured to execute a verification procedure based on physical symmetry. This procedure instructs the executor to drive the hydraulic gate to complete at least one complete opening-closing reciprocating motion. During this process, the data acquisition module acquires the opening power cost sequence of the opening process and the closing power cost sequence of the closing process, respectively. Then, the fingerprint construction module calls a mechanism based on gate structural parameters, such as the weight and volume of the gate leaf, and real-time hydrological parameters, such as upstream water level, etc. Downstream water level, a gravity and buoyancy effect model is established. Using this model, the theoretical power components generated by gravity and buoyancy during the opening and closing processes are calculated. Then, the module subtracts the corresponding theoretical power components from the original opening power cost sequence to obtain the first net resistance power sequence, which mainly reflects friction and additional resistance. The second net resistance power sequence is obtained from the closing power cost sequence in the same way. Finally, the module performs a mirror symmetry comparison on these two net resistance power sequences. Only when the overlap of the two curve shapes after mirror flipping is higher than a preset symmetry standard, such as the root mean square value of the difference between the corresponding points of the two curves being less than 5% of the total power average, is it determined that the learning process has not been interfered with by a unidirectional fixed obstacle. The initially obtained complete opening and closing power cost sequences are adopted as valid benchmark data and stored in a multidimensional database that maps operating parameters with power costs to form a benchmark health fingerprint.

[0024] Furthermore, after the system enters the daily real-time diagnostic phase, to address the technical problem in real water conservancy engineering environments where fluctuations in grid voltage or slight differences in controller response cause the actual operating speed curve of the gate to not coincide with the baseline curve on the time axis, thus making the instantaneous power value comparison method based on the same location point prone to false alarms, the real-time diagnostic module of this invention is configured to perform the comparison operation using a dynamic time warping algorithm. Specifically, when the gate performs a lowering task, the data acquisition module acquires the actual power cost sequence formed in real time. Simultaneously, the real-time diagnostic module, based on the synchronously acquired actual operating condition parameters, such as the gate position changing from 10 meters to 2 meters, the average speed of 0.1 m / s, and the upstream water level of 50 meters, determines the baseline power cost sequence matching this operating condition from the multidimensional database of baseline health fingerprints through interpolation calculation. Subsequently, dynamic time warping is performed. The algorithm takes these two time series as input and finds the optimal matching path between the data points of the two series. This path allows for local fast-forwarding or slow-motion on the time axis, thereby calculating a minimum warping distance value that reflects the overall similarity of the two curves. The system's diagnostic logic is based on this distance value. For example, if the calculated minimum warping distance value is 0.85, and the morphological similarity threshold determined based on historical statistical data is 1.5, then the system determines that the power curve morphology of this operation is consistent with the health benchmark, and the operating state is healthy. Conversely, if the actual power curve exhibits a drastic fluctuation that does not exist in the benchmark curve due to foreign object obstruction, the dynamic time warping algorithm cannot compensate for this morphological distortion by stretching the time axis, thus calculating a minimum warping distance value much greater than 1.5. The system then diagnoses the operating state as a resistive abnormal state.

[0025] Furthermore, to compensate for the limitations of passive real-time diagnostics in early detection of gradual, slowly increasing systemic friction, such as compaction of mud and sand within the guide rails or lubrication system failure, this system also integrates an active detection module. This module is configured to automatically execute a micro-perturbation health diagnostic process under a preset diagnostic trigger condition, such as 2:00 AM daily during the non-flood season. In this process, the module instructs the servo motor to drive the gate to complete a small stroke, such as moving it upwards by 20 centimeters and then returning, at a constant diagnostic speed lower than the normal operating speed, for example, 10% of the normal speed. Due to the extremely low and constant speed, the influence of changes in inertial force and hydraulic load can be ignored. Therefore, the baseline value of the diagnostic power cost sequence collected during this period can directly reflect the current static and dynamic friction characteristics of the system. The system will compare the friction characteristic curve obtained from this detection with the baseline friction curve obtained and stored through the same detection process when the baseline health fingerprint was first established. To achieve quantitative evaluation, the real-time diagnostic module is also configured to calculate a health aging factor characterizing the long-term changes in system friction. Its calculation rule is as follows ,in, Frictional characteristic values, such as average power, are calculated from the current diagnostic power cost sequence. The reference friction characteristic value is calculated from the reference friction curve, when the healthy aging factor... For example, if the value exceeds 1.5 for three consecutive diagnostic cycles, the system's built-in maintenance warning module will output a warning command suggesting maintenance and lubrication. This method of using the drive system to perform diagnostic actions under specific operating conditions provides the system with the ability to perform predictive maintenance on its long-term health trend without adding hardware.

[0026] Example 1: In a scenario where a large water conservancy project is conducting pre-discharge scheduling during the flood season, the control system commands Gate 3 to be lowered from 50% to 20% opening, at which point the upstream water level is 35.2 meters. During the lowering process, a floating object weds into the guide rail and gate leaf on the side of the gate. Because the hydraulic load is not large at this time, although the additional resistance generated by the floating object has caused abnormal stress on the drive mechanism, the instantaneous value of the overall actuator power cost is still lower than the fixed overload protection threshold set to cope with high water level conditions. At this time, the real-time diagnostic module of the control system continuously compares the actual power cost sequence obtained by the data acquisition module with the benchmark power cost sequence determined from the benchmark health fingerprint based on the actual operating parameters such as the current gate position, speed, and 35.2-meter water level. It should be noted that the benchmark health fingerprint is adopted by the system as a highly reliable comparison benchmark because it has passed the physical symmetry-based verification procedure executed by the fingerprint construction module from the beginning of its establishment. The procedure compares the high symmetry of the net resistance power sequence after removing gravity and buoyancy effects during the reciprocating motion of the gate, proactively filtering out any contamination of the baseline data that may be caused by minor deformations or unidirectional obstacles in the initial state. This ensures that the baseline power cost sequence used for comparison can truly reflect the physical process of the gate operating without faults under this condition. Therefore, although the current total power value does not trigger overload protection, the dynamic time warping algorithm used by the real-time diagnostic module, when calculating the overall morphological similarity between the actual power cost sequence and the baseline power cost sequence, identifies a significant difference in the shape of the two curves that cannot be bridged by local scaling of the time axis. The minimum warping distance value calculated by the algorithm exceeds the preset morphological similarity threshold. This judgment is not based on a single numerical value, but on the fact that the energy dissipation pattern of the entire physical process has deviated from its healthy state profile. This avoids the technical contradiction between insufficient sensitivity of the protection threshold at high water levels and easy underreporting at low water levels in traditional methods.

[0027] Based on the calculation results, the real-time diagnostic module diagnoses the gate's operating status as a resistive anomaly and immediately transmits the diagnostic result to the failure protection module. The failure protection module then executes the preset failure protection action, instructing the actuator to stop running and simultaneously sending an alarm message to the upper-level monitoring system, indicating that gate No. 3 has an operational anomaly at a specific location. The operators conduct on-site investigation based on this information and clear the floating obstruction. During this process, the system does not equate power anomaly with motor overload, but rather transforms the physical quantity of drive power cost from an electrical parameter used to assess the motor's own safety into a diagnostic criterion for auditing the health of the entire hydraulic structure's operation. This allows an initial fault that is easily overlooked under traditional technical frameworks and could potentially cause cumulative structural damage to be identified and intervened in a timely manner. It should be noted that in the scenario of this embodiment, after receiving the resistive anomaly diagnostic result, the failure protection module's preset failure protection action is further configured as a serialized procedure containing logical judgments. The module first executes the instruction to stop the actuator, causing the gate to move. The kinetic energy is instantly unloaded, and the actuator power cost is continuously monitored under this static state. If the power cost remains above the no-load reference value after stopping, the system determines that the resistive anomaly originates from a persistent hard obstacle that is jamming the door leaf and the door slot. In view of this, to avoid the possibility that reverse operation may aggravate structural stress, the failure protection module will no longer execute subsequent actions and will only maintain the alarm state. In another optional implementation, if the system detects that the power cost quickly drops back to the no-load reference after stopping operation, it determines that the resistive anomaly may originate from a non-fixed obstacle wedged in at a specific angle. At this time, the failure protection module is configured to execute the second step instruction, that is, to instruct the actuator to reverse a small stroke, such as 20 centimeters, at a low speed. The purpose of this is to try to make the non-fixed obstacle fall off under the impact of the reverse water flow or its own gravity, so that the system has the opportunity to automatically restore normal operation. This protection action selection procedure, which includes diagnostic logic, transforms the system's fault response from a simple passive disconnection to a dynamic process with preliminary self-repair attempts.

[0028] Example 2: To objectively verify the diagnostic capability of the control system claimed in this invention for typical operational anomalies under different hydraulic conditions, a test platform capable of simulating a hydraulic environment was built. This platform mainly consists of a scaled-down steel gate model, a servo motor-driven gate hoist, and a water tank with adjustable upstream and downstream water levels. The servo motor's drive controller integrates a power monitoring unit, whose data acquisition module can record the actuator power cost sequence at a sampling frequency of 100Hz and a resolution of 0.1W. Simultaneously, the system is equipped with a laser displacement sensor with a measurement range of 0 to 2 meters and an accuracy of 1 millimeter to acquire gate position parameters, and an ultrasonic level gauge with a measurement accuracy of 5 millimeters to acquire water level parameters. To simulate typical fault scenarios, a [missing information - likely a device or mechanism] was added to the gate guide rail of the test platform. An electromagnetic brake, which can be program-controlled to generate different damping forces, is used to simulate different degrees of resistive anomalies. An electronically controlled clutch is integrated into the transmission chain to simulate off-load anomalies. This experiment set up an experimental group using the technical solution of this invention and a control group using traditional set-value overload protection technology, and conducted comparative tests on three typical working conditions. The overload protection threshold of the control group was set to 120% of the rated power of the motor, i.e., 5.0kW, according to conventional engineering practice. The experimental group first performed a complete opening-closing reciprocating motion under clean conditions without any artificial fault setting, under three working conditions: high water level of 65 meters, medium water level of 50 meters, and low water level of 35 meters. Through a verification procedure based on physical symmetry, its baseline health fingerprint was established and solidified. The specific working condition settings and diagnostic results of the experiment are shown in Table 1.

[0029] Table 1: Comparison of diagnostic results for the two control systems under different operating conditions.

[0030] In operating condition 1, the system simulated normal flood discharge operation at high water levels. Due to the large hydraulic load, the actual peak power measured by the test group reached 5.2kW, exceeding the control group's set threshold of 5.0kW. Therefore, the control group system reported an overload false alarm. However, the test group's real-time diagnostic module, based on the high water level of 65.0 meters, determined from its baseline health fingerprint that the corresponding baseline peak power should be 5.1kW. The deviation between the two was within the normal range. Furthermore, its calculated minimum regularization distance was 0.45, less than the set morphological similarity threshold of 1.5. Therefore, the system was judged to be operating healthily. In operating condition 2, the system simulated moderate-intensity siltation or blockage at low water levels. The actual peak power was 4.8kW, which did not reach the control group's protection threshold. Therefore, the control group failed to detect this risk. However, the test group system, based on the low water level of 35.0 meters, determined its baseline peak power... The power value should be 3.2kW, but there is a significant deviation between the two values. Furthermore, due to jamming, the power curve shape is distorted, and the calculated minimum normalization distance increases to 3.86. Therefore, the system diagnoses this state as a resistive anomaly. In operating condition number 3, the system simulates a drive shaft breakage while running at medium water level, causing the actual power to drop to 1.5kW. The control group showed no response, while the test group system found that this value was far lower than the benchmark value of 4.1kW determined based on the operating condition. At the same time, the minimum normalization distance also exceeded the threshold, so it was judged as an underload anomaly. The test data shows that the test group using the technical solution of this invention can effectively identify various abnormal operating states under high dynamic hydraulic load backgrounds that cannot be covered by traditional fixed-value protection methods by associating real-time operating condition parameters with a benchmark health fingerprint that has been verified by intrinsic quality and using comparison logic based on morphological similarity. The diagnostic results are not affected by water level changes.

[0031] Example 3: This example combines Figures 1 to 3 A description of a hydraulic gate control system based on multi-source information fusion, such as... Figure 1 As shown, this architecture is based on the hydraulic gates and actuators at the physical device layer. Their operating status is synchronously acquired by the data acquisition module through sensor data. This module splits the acquired raw power and operating condition sequence into two paths. One path is sent to the fingerprint construction module, which is responsible for performing symmetry verification and establishing a baseline health fingerprint as a diagnostic benchmark. The other path sends the real-time power and operating condition sequence to the real-time diagnostic module, which compares the real-time sequence with the benchmark retrieved from the baseline health fingerprint to diagnose the operating status. The diagnostic results are sent to the failure protection module, which receives the diagnostic results and executes preset protection actions, issuing control commands to the physical device layer. At the same time, the system also includes an active detection module, which performs micro-motion detection on the physical device according to the control commands to assess the long-term health trend. Based on the assessment results, it can issue fingerprint update or calibration commands to the fingerprint construction module, thus forming a logically closed-loop health management system.

[0032] like Figure 2 As shown, the diagram illustrates the relationship between the opening power sequence, the mirrored closing power sequence, and the calculated net resistance power sequence during a complete reciprocating motion. The horizontal axis represents the gate position (%), and the vertical axis represents the power cost (kW). The solid line in the diagram represents the opening power sequence, the long dashed line represents the mirrored closing power sequence, and the dotted line represents the net resistance power sequence. After removing the asymmetric components caused by gravity, buoyancy, etc., from the original opening and closing power sequences, the two net resistance power sequences should be highly overlapping in shape, as shown in the figure. Only when this mirror symmetry meets the preset standard will the system accept the collected data as a valid health benchmark.

[0033] like Figure 3 As shown in the diagram, the participants include the hydraulic gate, data acquisition module, real-time diagnostic module, baseline health fingerprint, failure protection module, and upper-level monitoring system. The process begins during the normal operation of the hydraulic gate. The data acquisition module acquires the real-time operating parameter sequence and the actual power cost sequence and transmits them to the real-time diagnostic module. This module then queries the baseline power cost from the baseline health fingerprint and obtains the returned data. Next, it compares and calculates the morphological similarity distance using the dynamic time warping method. If the similarity distance is normal, monitoring continues. If the similarity distance exceeds the threshold, it is diagnosed as an abnormal state, and the abnormal diagnosis result is sent to the failure protection module, which executes the failure protection action and issues instructions to the hydraulic gate. At the same time, it sends alarm information to the upper-level monitoring system until the abnormality is handled.

[0034] Example 4: To determine key parameters relied upon by the core diagnostic logic of the control system of this invention after initial deployment or major maintenance, such as the symmetry standard used to determine the validity of the benchmark data and the morphological similarity threshold used for real-time diagnosis, and to avoid uncertainties caused by relying on experience-based settings, this example discloses a systematic offline calibration and parameter self-tuning procedure. The initial state of this procedure is a hydraulic gate system that has been physically inspected and confirmed to be in mechanical good condition and with clean flow channels, and a control and data acquisition hardware environment with functional specifications. After the procedure is started, the system first performs preliminary acquisition of the benchmark health fingerprint. The data acquisition and verification process involves the controller instructing the actuator to continuously execute 10 complete opening and closing reciprocating motions at a standard speed covering the main working range. The data acquisition module simultaneously records the operating parameter sequence and actuator power cost sequence for all 20 strokes. For each reciprocating motion, the fingerprint construction module calculates the mirror symmetry comparison result between the two net resistance power sequences for the opening and closing strokes, following the method described in the specific implementation. This result is quantified using the root mean square error (RMSE) of the difference between corresponding points on the two curves, resulting in 10 RMSE values. The system then calculates the statistical mean of these 10 error values. with standard deviation And the symmetry standard is determined as Any travel data whose symmetry error exceeds the standard during the initial acquisition will be considered to be subject to random interference and discarded.

[0035] Based on the power cost sequence set confirmed as valid under multiple healthy states through the above-mentioned process, the system then performs morphological similarity threshold calibration. The real-time diagnostic module is configured to call the dynamic time warping algorithm to perform pairwise similarity calculations on all power cost sequences with the same running direction within the set. For example, if there are 8 valid power sequences for the opening stroke, the system will perform 28 dynamic time warping calculations to obtain a statistical sample consisting of 28 minimum warping distance values. The system then calculates the statistical distribution characteristics of this sample and sets the morphological similarity threshold to the value corresponding to the 99th percentile of the sample. In this way, the threshold becomes a boundary that can reflect the inherent variation range of the power curve of the specific gate system under healthy states. Simultaneously, in constructing... When using a multidimensional database to store baseline health fingerprints, the system is configured to index operating parameters using a kd-tree structure for subsequent nearest neighbor lookups. Furthermore, when the actual operating parameters acquired in real-time do not have a directly mapped baseline power cost in the database, the interpolation unit invoked by the system is configured to use a trilinear interpolation algorithm. This involves selecting the eight nearest baseline operating points based on the relative position of the actual operating point in the three-dimensional space defined by the gate position, operating speed, and upstream water level. A weighted average is then calculated based on the baseline power costs corresponding to these eight points to obtain the interpolated baseline power cost corresponding to the actual operating condition. After this series of calibration procedures, the core judgment criteria of the entire control system are given a traceable source, and the system enters a standby state with parameter status determined and diagnostic logic closed-loop.

[0036] Example 5: In a control gate project containing multiple parallel-operating hydraulic gates, to address the problem that each gate has specific frictional characteristics even under the same operating conditions due to manufacturing tolerances, installation differences, and uneven wear during long-term operation, the control system of this invention is configured to independently execute the offline calibration and parameter self-tuning procedures described in Example 3 for each gate during initial deployment. Specifically, the system will establish an independent set of baseline health fingerprints, symmetry standards, and morphological similarity thresholds for each gate, from gate 1 to gate N, each uniquely corresponding to its physical entity. These parameter sets are bound and stored with the gate's unique equipment identification code. During subsequent real-time diagnostics, when the system needs to assess the status of any gate, such as gate 2, it will first read the gate's equipment identification code and, based on this code, retrieve the parameter set specific to gate 2 from memory for comparison calculation.

[0037] Furthermore, to adapt to the performance drift of hydraulic structures throughout their entire life cycle caused by mechanical wear or changes in lubrication conditions, the system is also equipped with a dynamic update procedure for a baseline health fingerprint based on active detection results. For example, in a specific implementation, the system periodically acquires health aging factors that reflect long-term frictional changes in the system through an active detection module. When a certain gate When the value continuously exceeds the preset maintenance threshold of 1.5 for two or more consecutive diagnostic cycles, the system will output a planned maintenance instruction to the operation and maintenance unit. After the instruction is responded to and the relevant physical maintenance work, such as lubricating the guide rail, is completed, the authorized operation and maintenance personnel can issue a fingerprint update instruction to the system. After receiving the instruction, the system will again execute the complete calibration process including symmetry verification in Embodiment 3 on the gate to generate a brand-new benchmark health fingerprint that can reflect its current health status. Once the new fingerprint is confirmed, it will replace the original old fingerprint as the benchmark for subsequent real-time diagnosis. This procedure enables the system's diagnostic logic to dynamically adapt to the slow performance evolution of its monitored object in order to maintain diagnostic accuracy during long-term operation.

[0038] Example 6: For the health aging factor used in the control system of this invention for early warning of long-term performance degradation The maintenance threshold is calibrated, and the response mechanism of the fingerprint construction module when encountering a non-ideal initial state is solidified. This embodiment discloses a pre-emptive parameter optimization and anomaly handling procedure. This procedure is executed on a test platform equipped with an adjustable electromagnetic brake to simulate progressive friction increase. First, with the gate system in a confirmed clean and well-lubricated initial healthy state, the diagnostic peristalsis process of the active detection module is executed to measure and store its reference friction characteristic values. .

[0039] Subsequently, by controlling the electromagnetic brake, an additional, constant braking force is applied to the gate guide rail. The magnitude of this braking force, after calibration, is equivalent to a 50% increase in the gate's mechanical friction. Under this state, the diagnostic creep process is executed again, and the corresponding current friction characteristic value is measured. ,in accordance with The calculation rules are used to calculate the results under this working condition. The numerical value, 1.5, is set as the preset maintenance threshold for the system to output planned maintenance instructions. This process anchors the basis for maintenance decisions from an empirical estimate to a reproducible parameter related to the specific degree of physical wear. Simultaneously, to address potential, subtle, persistent physical anomalies during initial system deployment, such as minor permanent deformation of the guide rail, the fingerprint construction module of this invention is configured to, during symmetry verification, if the calculated root mean square error value exceeds [a certain threshold] in the initial acquisition of 10 consecutive reciprocating motions... The travel data consists of more than 5 trips, or the standard deviation of all error values. If the system exceeds an upper limit threshold representing the degree of data dispersion, it will terminate the process of establishing the baseline health fingerprint and send an alarm for an abnormal baseline learning environment to a higher-level monitoring system. This mechanism can prevent the system from mistakenly learning a state that is already diseased from the beginning as a healthy baseline.

[0040] To further verify the key role of the physical symmetry-based verification procedure in ensuring the purity of reference data in the system claimed in this invention, the following comparative example 1 is provided.

[0041] Comparative Example 1: The only difference between this comparative example and the test group in Example 2 is that its control system, when establishing the baseline health fingerprint, does not execute the verification procedure based on physical symmetry of this invention, but instead adopts a conventional technical path known in the art, that is, assuming that the system is in good condition during initial deployment and directly using the power cost sequence collected for the first time as the baseline; the test process is as follows: when the system is initially deployed and baseline data is collected, in order to simulate a one-way initial fault that is not easily detected on site, an obstacle that only generates significant frictional resistance during the gate closing (lowering) stroke is artificially set in the guide rail on one side of the gate; under this state, the command system executes A complete opening-closing reciprocating motion was performed to collect baseline data. Under the condition of a high water level of 65.0 meters, the peak power during the opening process was measured to be 5.1 kW, consistent with the healthy state. However, due to the influence of a unidirectional obstruction during the closing process, its peak power increased to 5.8 kW. Because the control system of this comparative example lacked symmetry verification capability, it confirmed the collected contaminated asymmetric data as a valid baseline and established an erroneous baseline health fingerprint, which stored the erroneous record of a baseline peak power of 5.8 kW under the high water level of 65.0 meters. Afterwards, the obstruction was removed, restoring the gate to a fully healthy physical state. For direct comparison with condition number 1, the exact same high water level normal operation test as condition number 1 was performed again. The diagnostic results of this comparative example were compared with the diagnostic results of condition number 1 in Embodiment 2 of this invention, as detailed in Table 2.

[0042] Table 2: Comparison of diagnostic results between Comparative Example 1 and the present invention in the initial baseline contaminated scenario.

[0043] The results of Comparative Example 1 show that without the reference self-verification mechanism proposed in this invention, the control system cannot identify the contamination in the initial reference data. It incorrectly learns an excessively high power value (5.8kW) contaminated by a unidirectional fault as a healthy reference. Therefore, when the gate operates in a subsequent truly healthy state, its normal power cost (5.2kW) is significantly lower than this erroneous reference and is misjudged by the system as an underload anomaly. This experimental result confirms that without a prior verification step based on physical symmetry, conventional technical solutions cannot guarantee the purity of the diagnostic reference. When faced with engineering realities where the initial state is not ideal, there is an inherent risk of outputting a seriously erroneous diagnosis.

[0044] 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.

[0045] 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 water conservancy gate control system based on multi-source information fusion, characterized in that, The system comprises: a data acquisition module configured to synchronously acquire a working condition parameter sequence representing a working condition of the water gate and an actuator power cost sequence representing a driving load of the water gate during a movement of the water gate; a fingerprint construction module connected with the data acquisition module, the fingerprint construction module being configured to: instruct an actuator to drive the water gate to complete at least one reciprocating movement including an opening process and a closing process, and acquire a corresponding opening power cost sequence and a closing power cost sequence from the data acquisition module; then perform symmetry checking on the opening power cost sequence and the closing power cost sequence, the checking being performed by stripping off asymmetric power parts in the opening power cost sequence and the closing power cost sequence due to gravity or buoyancy effects to obtain two net resistance power sequences respectively reflecting friction and additional resistance, and performing mirror symmetry comparison on the two net resistance power sequences; and only when a result of the mirror symmetry comparison meets a preset symmetry standard, the acquired opening and closing power cost sequences are confirmed as valid reference data, and a reference health fingerprint is established based on the valid reference data; a real-time diagnosis module configured to determine a reference power cost from the reference health fingerprint established by the fingerprint construction module according to actual working condition parameters acquired in real time, and diagnose a running state of the water gate based on a deviation between an actual power cost acquired in real time and the reference power cost.

2. The water conservancy gate control system based on multi-source information fusion according to claim 1, characterized in that, The working condition parameter sequence includes a gate position parameter sequence, a running speed parameter sequence and a running direction parameter sequence, and an upstream water level parameter sequence and a downstream water level parameter sequence associated with the water gate; the fingerprint construction module is configured to establish the reference health fingerprint based on a corresponding relationship between the working condition parameter sequence and the valid reference data, wherein the reference health fingerprint is a multidimensional database mapping the working condition parameters and the power cost.

3. The water gate control system based on multi-source information fusion according to claim 1, characterized in that, The system further comprises an active detection module configured to: under a preset diagnosis triggering condition, instruct the actuator to drive the water gate to complete a preset small diagnosis stroke at a diagnosis speed lower than a normal running speed of the system; the real-time diagnosis module is further configured to acquire a diagnosis power cost sequence during the diagnosis stroke, and compare the diagnosis power cost sequence with a reference friction curve stored in the reference health fingerprint.

4. The water gate control system based on multi-source information fusion according to claim 3, characterized in that, The reference friction curve is synchronously acquired and stored in the reference health fingerprint by performing a detection process of the active detection module during a first establishment of the reference health fingerprint.

5. The water gate control system based on multi-source information fusion according to claim 1, characterized in that, The real-time diagnosis module is specifically configured to: determine the deviation by calculating a shape similarity distance between an actual power cost sequence formed by the actual power cost and a reference power cost sequence determined from the reference health fingerprint by using a dynamic time warping algorithm; and diagnose the running state as an abnormal state when the shape similarity distance exceeds a shape similarity threshold value determined according to statistical distribution characteristics of the reference health fingerprint. ​ 6. The water gate control system based on multi-source information fusion according to claim 3, characterized in that, The real-time diagnosis module is further configured to calculate a health aging factor representing long-term change of system friction based on a comparison result of the diagnosis power penalty sequence and the reference friction curve , and the calculation rule is , wherein is a current friction characteristic value calculated from the diagnosis power penalty sequence, is a reference friction characteristic value calculated from the reference friction curve; and the system further comprises a maintenance warning module configured to output a warning instruction when the health aging factor continuously exceeds 1.5 for two or more consecutive diagnosis cycles.

7. The water gate control system based on multi-source information fusion according to claim 1, characterized in that, The real-time diagnostic module is specifically configured to: diagnose the operating state as a resistive abnormal state when the actual power cost exceeds the reference power cost by a preset first threshold; and diagnose the operating state as an off-load abnormal state when the actual power cost is lower than the reference power cost by a preset second threshold.

8. The water gate control system based on multi-source information fusion according to claim 7, characterized in that, The system also includes a failure protection module, which is configured to execute a preset failure protection action after the real-time diagnostic module detects a resistive abnormality or an underload abnormality. The failure protection action includes stopping the actuator, instructing the actuator to run in reverse, or sending an alarm message to a higher-level monitoring system.

9. The water gate control system based on multi-source information fusion according to claim 2, characterized in that, The real-time diagnostic module is also configured to: when the actual operating parameters acquired in real time do not have a directly mapped reference power cost in the multidimensional database, call an interpolation calculation unit. The interpolation calculation unit is configured to select multiple reference operating points in the multidimensional database that are adjacent to the actual operating parameters in the multidimensional space, and calculate the interpolated reference power cost corresponding to the actual operating parameters based on the reference power cost corresponding to the multiple reference operating points through a preset interpolation algorithm, and use the interpolated reference power cost as the reference power cost.

10. The water gate control system based on multi-source information fusion according to claim 1, characterized in that, The symmetry verification operation is specifically configured as follows: a gravity and buoyancy effect model is established based on the gate structure parameters and real-time hydrological parameters, and the theoretical power components generated by gravity and buoyancy during the opening and closing processes are calculated using this model; then, the corresponding theoretical power components are subtracted from the opening power cost sequence to obtain the first net resistance power sequence, and the corresponding theoretical power components are subtracted from the closing power cost sequence to obtain the second net resistance power sequence; finally, a mirror symmetry comparison is performed between the first net resistance power sequence and the second net resistance power sequence.

Images