Water conservancy pump station concrete structure leakage real-time monitoring and early warning system

Abstract

The invention relates to the technical field of structure health testing, in particular to a water conservancy pump station concrete structure leakage real-time monitoring and early warning system which comprises the steps that a distributed sensing array deployed on the surface of a concrete structure is used for collecting structure vibration signals excited by operation of a pump set; through static topology mode recognition, a continuous high-coherence abnormal path representing potential leakage is preliminarily judged; the method comprises the following steps: performing dynamic arbitration on time synchronism of response of each point on a path by using a static primary screening and dynamic confirmation mechanism and further using a transient vibration event generated in a system such as start and stop of a pump set as a verification signal, and outputting an early warning only after the path is confirmed to be a real physical channel. And unavoidable working condition switching events in structure operation are converted into a decisive basis for distinguishing real structure damage and benign state drift, so that the technical problem of false alarm caused by reference drift of a passive monitoring system is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to a real-time monitoring and early warning system for concrete structure leakage of water conservancy pumping station, belonging to the technical field of structural health testing. BACKGROUND

[0002] At present, for water conservancy pumping stations and other large-scale concrete structures that have been operating under high water head and high load conditions for a long time, the long-term integrity of the internal structure is the basis for ensuring the safety of the project; therefore, using sensors deployed on the surface of the structure to continuously monitor the vibration response signals of the structure and identifying abnormal changes caused by material degradation or damage is a common technical approach for evaluating the health status of the structure in the field; however, when this monitoring approach is applied to the actual operating scenario of a water conservancy pumping station, its effectiveness is subject to a constraint derived from the operating principle, that is, the structure vibration response on which the monitoring system relies is mainly excited by the strong, non-stationary, and wide-band background vibration generated by the pump unit itself; the energy of this background vibration is much higher than that of the structural response changes caused by the formation of early micro-leakage channels, making it difficult to separate the characteristic signals of the latter from the former, and the monitoring method that relies on identifying signal abnormalities from a relatively stable background loses its physical premise of application in this scenario.

[0003] Therefore, the industry has made related explorations, but most existing technical solutions focus on quality acceptance during the construction phase or rely on macroscopic manifestations after leakage occurs for judgment, and there are still logical limitations in dealing with early micro-crack evolution monitoring during the operation period of the structure that is masked by strong vibration; for example, a Chinese invention patent with the authorization announcement number CN120628495B discloses a leakage early warning system for water plant activated carbon pool construction, which tests by water injection after construction and comprehensively uses infrared thermal imaging technology to monitor the wetness representation value (i.e. low temperature area proportion) of the outer wall and the water level change rate for secondary determination, however, this method is essentially a lagging diagnosis based on the leakage result, its monitoring premise is that the leakage has developed to a sufficient extent to cause significant temperature drop or observable water level drop of the structure outer wall, this monitoring logic that relies on macroscopic physical manifestations cannot achieve early warning of the cause of the formation of the leakage path for structures such as water conservancy pumping stations that are under strong vibration conditions for a long time, the monitoring signal is easily disturbed by external factors such as environmental temperature and humidity, air flow, etc., resulting in false judgments, and it does not provide an internal mechanism for effectively extracting weak damage characteristic signals in the background noise of strong vibration.

[0004] The analysis shows that the prior art mainly has the following constraints: 1. There is a huge difference in energy between the damage information and the background vibration information in the structural response signal, and the characteristic signal of early damage is covered by high-energy background vibration; 2. As the main interference source, the pump set vibration and the structural response carrying damage information are both derived from the vibration of the structure itself, and cannot be effectively separated by conventional frequency band filtering; 3. The logical basis of the existing monitoring method is still to regard the pump set vibration as a negative interference factor that needs to be suppressed or overcome, and a new structure testing method that can use this vibration to test the structure has not been established. Therefore, how to establish a new structure testing method that no longer regards the strong vibration generated by the pump set as an obstacle to monitoring, but uses it as an information carrier for continuous detection of the structure, and reliably identifies the formation of an early leakage path representing the discontinuity of the internal medium, becomes a technical problem to be solved by the present application. SUMMARY

[0005] The present application provides a real-time monitoring and early warning system for concrete structure leakage of water conservancy pump station, which mainly aims to solve the problem that the existing monitoring method is difficult to reduce noise and has low signal-to-noise ratio under strong vibration background, and it is difficult to effectively identify early weak leakage signals.

[0006] To achieve the above purpose, the present application provides a real-time monitoring and early warning system for concrete structure leakage of water conservancy pump station, which comprises: a distributed sensing array deployed on the surface of the concrete structure; a full-time signal acquisition unit connected with the distributed sensing array and configured to synchronously acquire vibration signals output by each sensor excited by the pump set operated by the water conservancy pump station; a static topology pattern recognition subsystem configured to periodically calculate the coherence between all sensor pairs in the distributed sensing array based on the vibration signals to construct a real-time coherence matrix, and preliminarily identify a continuous high-coherence abnormal path by comparing the real-time coherence matrix with a health state reference matrix established in advance when the concrete structure is in a confirmed healthy state; A dynamic event decision arbitration subsystem comprises a pump set transient event monitoring unit configured to identify, from the vibration signals, a time instant of occurrence of a global transient vibration event caused by pump set start-stop or operating condition switching; and a path response synchronicity verification module configured to be activated only when the static topology pattern identification subsystem identifies a high-coherence abnormal path and the pump set transient event monitoring unit captures a transient vibration event, and to obtain a quantitative indicator of response synchronicity by calculating and comparing time stamps of responses of all sensor nodes on the path to the transient vibration event, and to confirm the authenticity of the high-coherence abnormal path when the synchronicity quantitative indicator satisfies a preset decision condition, and to generate and output a leakage warning signal only after the confirmation.

[0007] Preferably, the static topology pattern identification subsystem is further configured to: call one or more health state reference matrices corresponding to one or more preset operating conditions of the pump set, the one or more health state reference matrices being established by running and collecting data under the confirmed health state of the concrete structure; and when performing the comparison, select a health state reference matrix corresponding to the actual operating condition of the pump set from the one or more health state reference matrices to perform the comparison operation.

[0008] Preferably, the system further comprises: a single-point response complexity calculation unit configured to calculate, based on the vibration signals collected by the full-time domain signal collection unit, an entropy measure index representing the complexity of the time series of vibration response signals of each independent sensor in the distributed sensing array in parallel; and a leakage precursor identification module configured to compare the entropy measure index of each independent sensor with a health state entropy reference, and when identifying that the entropy measure index of one or more adjacent sensors is continuously higher than the health state entropy reference, determine the structure region corresponding to the sensors as a leakage precursor risk area, and generate spatial position and time information representing the leakage precursor risk area.

[0009] Preferably, the system further comprises: a coherence spectrum analysis unit configured to calculate, after the dynamic event decision arbitration subsystem preliminarily identifies a high-coherence abnormal path, a coherence spectrum function of the vibration signals of the sensor pairs on the path; and a channel maturity qualitative indication unit configured to qualitatively evaluate the maturity state of the high-coherence abnormal path based on spectral pattern features of the coherence spectrum function, and output the maturity state obtained by the evaluation together with the leakage warning signal.

[0010] Preferably, the channel maturity qualitative indication unit is configured to perform the following operations: if the coherence spectral function presents one or more narrow-band peak patterns concentrated at specific frequencies, the maturity state of the high-coherence abnormal path is evaluated as the nascent stage; and if the coherence spectral function presents a flat high-value platform pattern in a continuous broad frequency range, the maturity state of the high-coherence abnormal path is evaluated as the mature stage.

[0011] Preferably, the path response synchronism verification module is configured to extract the response timestamp sequence of all sensor nodes on the high-coherence abnormal path to the transient vibration event , wherein N is the total number of sensor nodes on the path; calculate the standard deviation of the response timestamp sequence as a synchronism quantitative indicator; the preset determination condition is , wherein, is a time standard deviation threshold value determined according to historical health data statistics.

[0012] Preferably, the system further comprises a visual output module, and the visual output module is configured to determine the origin position and extension direction of the leakage according to the spatial topological position of the high-coherence abnormal path finally confirmed to be true in the distributed sensing array, and mark the origin position and extension direction in the three-dimensional structure model.

[0013] Preferably, the single-point response complexity calculation unit is configured to convert the time sequence of the vibration response signal of each independent sensor into a trajectory in a multi-dimensional phase space by a phase space reconstruction method; then, calculate the complexity of the sequence pattern of the phase space trajectory by using an permutation entropy algorithm, and take the calculation result as an entropy measurement indicator.

[0014] Preferably, the pump group transient event monitoring unit is configured to perform short-time energy analysis on the vibration signal, and when the change rate of the short-time energy of the vibration signal in a unit time exceeds an energy change rate threshold value determined according to historical operation data statistics, the occurrence of the transient vibration event is identified.

[0015] Preferably, the system further comprises a risk evolution evaluation module, and the risk evolution evaluation module is configured to receive the spatial position and time information representing the leakage precursor risk area generated by the leakage precursor identification module, and receive the spatial position and time information of the continuous high-coherence abnormal path identified by the static topological pattern identification subsystem; and perform spatio-temporal data correlation based on the spatial position and time information to establish a monitoring record chain from the appearance of the leakage precursor risk area to the formation of the continuous high-coherence abnormal path, and output the monitoring record chain to represent the evolution process of the leakage risk.

[0016] Compared with the prior art, the application has the beneficial effects that: 1. The distributed sensing array collects the structural vibration signals excited by the pump set, processes them through the coherence topology calculation core, and constructs a real-time coherence matrix representing the vibration synchronization of each part of the structure. The focus of this processing method is not the absolute size of the vibration signal energy or its frequency components, but the consistency of the response behavior between different points when the vibration is transmitted in the structure medium. When a physically continuous leakage channel is formed inside the structure, the water-filled channel becomes the dominant path for vibration transmission, and the vibration signal coherence between the two ends and along the way of the measuring points shows a stable change different from the dispersion propagation characteristics of the healthy state concrete. Therefore, by comparing the changes of the real-time coherence matrix and the reference state, the problem of identifying weak signals in a strong vibration background is transformed into a structural testing problem of identifying changes in the physical connectivity of the structure medium, so that the basis of monitoring is no longer subject to the vibration intensity fluctuations generated by the pump set.

[0017] 2. While calculating the coherence between sensor pairs to identify the formed leakage path, the system also performs complexity calculation on the vibration signal of each independent sensor in parallel to obtain an entropy metric index representing the medium state of the local area of the measuring point. When a region of micro-cracks is saturated with water but has not yet penetrated, the viscous motion of water will introduce random disturbances to the response of the region to environmental vibration, which is manifested as an increase in the entropy metric index. The system identifies the high-entropy metric index region as a pre-leakage risk area and cooperates with the continuous high-coherence path information identified through coherence analysis to establish a complete monitoring chain from micro-damage accumulation to macro-path formation. The parallel operation of this dual mechanism enables the system not only to locate the occurred structural damage but also to identify the incubation stage of the damage, providing a process-based decision basis for preventive maintenance.

[0018] 3. After the structural integrity self-checking and early warning unit initially identifies a highly coherent abnormal path, it does not immediately generate an early warning. Instead, it initiates a verification mechanism based on pump group transient events. This mechanism monitors global transient vibration events generated in the structure when the pump group starts or stops or switches operating conditions, and analyzes the synchronicity of the responses of each sensor along the initially identified highly coherent path as the transient event propagates. A real physical leakage channel has a fixed and rapid characteristic in transmitting transient vibrations, and the responses of each point along the path are highly synchronized in time. False high coherence caused by benign baseline drift due to factors such as temperature changes or concrete creep does not have this synchronous response capability to dynamic events. By introducing this dynamic verification step, this invention utilizes the inherent events that inevitably occur during structural operation to finally confirm the preliminary diagnostic results based on static pattern comparison, thereby distinguishing between real structural damage and benign state drift, and ensuring the reliability of the early warning results. Attached Figure Description

[0019] Fig. 1 This is a flowchart illustrating the dual decision-making mechanism and data processing of the monitoring and early warning system of the present invention. Fig. 2 This is a schematic diagram illustrating the relationship between the maturity of the leakage channel and the coherence spectrum characteristics of the present invention. Fig. 3 This is a multi-level functional module architecture diagram of the monitoring and early warning system of the present invention. Detailed Implementation

[0020] 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 with reference to the accompanying drawings. However, those skilled in the art will understand that, in the present invention, modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and such modifications and equivalent substitutions all fall within the protection scope of the claims of the present invention.

[0021] This invention discloses a real-time monitoring and early warning system for leakage in concrete structures of hydraulic pumping stations. Its overall architecture mainly consists of a distributed sensing array deployed on the surface of the concrete structure, a connected all-time-domain signal acquisition unit, a static topology pattern recognition subsystem, and a dynamic event decision arbitration subsystem. The all-time-domain signal acquisition unit is responsible for digitizing the structural vibration signals acquired by the distributed sensing array and transmitting them to the subsequent processing unit. The static topology pattern recognition subsystem continuously analyzes the vibration signals and uses pattern comparison based on spatial coherence topology to initially identify potential leakage paths. The dynamic event decision arbitration subsystem utilizes transient events during pump operation to make a final decision arbitration based on dynamic response synchronization of the preliminary static identification results, thereby outputting a doubly verified early warning signal. In a specific implementation scenario, the distributed sensing array consists of multiple piezoelectric accelerometers. These sensors are deployed on the surface of the concrete structure of the hydraulic pumping station to be monitored in a preset geometric layout to ensure… Capable of capturing broadband vibrations excited by pump operation, the selected sensors must have an effective frequency response range of no less than 0.5Hz to 5kHz and a sampling accuracy of no less than 24 bits. The full-time domain signal acquisition unit is connected to each sensor in the array and configured to synchronously acquire data with a unified sampling clock. The sampling frequency is set to, for example, 4096Hz to ensure the fidelity of high-frequency components in the signal. Its output is a multi-channel, time-based raw vibration signal data stream. The static topology pattern recognition subsystem needs to address the technical problem that the structural response changes caused by early minor leaks have energy far lower than the high-energy background vibrations generated by pump operation, making it difficult to effectively identify them using traditional signal amplitude or spectrum analysis methods. To address this technical problem, this subsystem is configured to execute a procedure that transforms vibration signal analysis into a structural medium connectivity test. The specific steps are as follows: First, the system acquires a continuous vibration data stream from the full-time domain signal acquisition unit and divides it into segments of length T (e.g., ...). The system generates continuous data frames (seconds); secondly, within the coherent topology calculation core of this subsystem, it calculates the data from any two different sensors within each data frame. and Output signal time series and Perform cross-correlation coefficient calculations to construct a Real-time coherence matrix ,in For the total number of sensors, the nth in the matrix Line number Column elements That is, a sensor and vibration signal coherence quantization value between them; a physically continuous water leakage channel will become the dominant path for vibration propagation, causing the vibration signals between the sensors at both ends and along the channel to exhibit higher synchronization than the general concrete medium propagation, which is reflected in that the corresponding values are significantly higher than those of other sensors; the system compares this real-time coherence matrix with a reference coherence matrix representing the structure's health state , and the comparison operation is to perform a topological path search algorithm, i.e., to find in whether there is a continuous path composed of multiple sensor nodes, with the coherence value of each segment of the connection on the path consistently higher than a preset judgment threshold ; the determination of the judgment threshold can be completed by an initial state calibration procedure, i.e., during the initial deployment of the system and confirmation of the structure's health, the system collects vibration data for at least 24 hours, calculates the reference coherence matrix under the healthy state, and takes the average value and standard deviation of all non-diagonal elements in the matrix , then the judgment threshold can be set as , where is a coefficient set according to the required monitoring sensitivity, for example ; when such a continuous high-coherence abnormal path is identified, the subsystem completes a preliminary identification, and outputs the sensor node sequence information contained in the path to the dynamic event decision arbitration subsystem; through this procedure, the system converts the problem of identifying weak signals in strong background vibration into a pattern recognition problem of changes in the physical connectivity topology of the structure's internal medium.

[0022] The dynamic event decision arbitration subsystem needs to solve the technical problem that the overall state of the structure may change slowly due to seasonal changes in temperature or concrete creep and other non-damage factors, which may cause the entire coherence topological map to drift, and in turn may cause the static topological pattern recognition subsystem to misjudge the benign state drift as a leakage path. To address this technical problem, the subsystem adopts a path response synchronization verification procedure based on pump transient events to confirm the preliminary results of static identification; the pump transient event monitoring unit in the subsystem identifies the occurrence of global transient vibration events such as pump start-stop or working condition switching by performing short-time energy analysis on the full-time vibration signal, for example, when the change rate of the calculated short-time energy of the vibration signal in unit time exceeds an energy change rate threshold determined according to historical operation data statistics, a transient vibration event and its occurrence time are identified The path response synchronization verification module within the static topology pattern recognition subsystem is activated only when the static topology pattern recognition subsystem identifies a highly coherent abnormal path and the pump group transient event monitoring unit captures a transient vibration event. Once activated, this module accurately extracts the response timestamp sequence of all sensor nodes on the highly coherent path to the transient event. ,in , This represents the total number of sensor nodes along the path; subsequently, the response timestamp sequence is calculated. Standard deviation As a quantitative indicator characterizing response synchronicity, a real physical leakage channel exhibits a fixed and rapid characteristic in the transmission of transient vibrations, with highly synchronized responses at each point along the path. In contrast, spurious high coherence generated by benign baseline drift lacks this ability to synchronously respond to dynamic events. Therefore, the system uses this synchronicity quantification indicator. Does it meet a preset judgment condition, for example? ,in A time standard deviation threshold, such as 5ms, determined based on historical health data statistics, is used to finally confirm the authenticity of the highly coherent abnormal path. Only after its authenticity is confirmed will the system finally generate and output a leakage warning signal. This transforms the operating condition switching events generated during the structural operation into a basis for distinguishing between real structural damage and benign state drift.

[0023] In some implementations, to further enhance the monitoring scope, the system can also run a single-point response complexity calculation unit in parallel. The technical problem addressed by this unit is that before a macroscopic leakage path is finally completed, the gradual and diffuse accumulation of micro-damage within the structure cannot form the continuous, highly coherent path relied upon by the static topology pattern recognition subsystem. To solve this problem, this unit calculates an entropy metric that characterizes the complexity of the vibration signal time series collected by each independent sensor in the distributed sensing array, for example, using the permutation entropy algorithm. When a microcrack in a local area is saturated with water but has not yet been connected, the viscous movement of water will introduce nonlinear random disturbances into the response of that area to environmental vibrations, which manifests as an increase in the entropy metric. The system compares the entropy metric of each independent sensor with a healthy entropy benchmark. When the entropy metric of one or more neighboring sensors is identified as consistently higher than the benchmark, the structural area corresponding to that sensor is determined as a leakage precursor risk zone, and information characterizing the spatial location and time of the risk zone is generated. This function complements the main path monitoring function, establishing a monitoring chain from micro-damage accumulation to macroscopic path formation.

[0024] In addition, in order to qualitatively evaluate the severity of the confirmed leakage path, the system can further include a coherence spectrum analysis unit and a channel maturity qualitative indication unit, which solves the technical problem that the operation and maintenance manager needs to distinguish between a newly formed leakage and a mature leakage to determine the priority of disposal, and the technical principle is that a newly formed and immature leakage path has physical characteristics that make it like a band-pass filter, which can only effectively conduct vibration energy of a specific narrow frequency band, while a mature and well-developed path is like a wideband waveguide that can efficiently conduct vibrations in a wide frequency range, so when the dynamic event decision arbitration subsystem confirms a high-coherence abnormal path, the coherence spectrum analysis unit is activated to calculate the coherence spectrum function of the vibration signals of the sensor pairs on the path , and then the channel maturity qualitative indication unit evaluates based on the spectral pattern characteristics of the coherence spectrum function: if the function presents one or more narrow-band peak patterns concentrated at a specific frequency, the maturity state of the path is evaluated as the newborn period, and if the function presents a flat high-value platform pattern in a continuous wide frequency range, it is evaluated as the mature period, and the evaluation result is output together with the warning signal.

[0025] In some embodiments, the pump group transient event monitoring unit is configured to process the vibration signal time series output by one or more reference sensors, calculate the short-time energy of the signal through a sliding time window (for example, window length 1 second, sliding step 0.1 second), and first-order difference the short-time energy sequence to obtain its unit time change rate. When the absolute value of the change rate continuously exceeds an energy change rate threshold determined based on historical operation data statistics for a plurality of consecutive sampling points, the system identifies the occurrence of a global transient vibration event. The visualization output module is configured to, after receiving the high-coherence abnormal path information finally confirmed to be true, read the preset spatial topological position coordinates of each sensor node included in the path in the distributed sensing array, and project the sequence of these coordinate points into a preloaded pump station three-dimensional structure model. By highlighting the path rendering on the model, the origin position and extension direction of the leakage are indicated. Further, the system can also include a risk evolution evaluation module configured to continuously receive and store the spatial position and time information of the leakage precursor risk area generated by the leakage precursor identification module, and the spatial position and time information of the continuous high-coherence abnormal path identified by the static topological pattern identification subsystem, and periodically perform spatio-temporal data correlation analysis on the two types of information, i.e., check whether the spatial position of the newly generated continuous high-coherence abnormal path overlaps or is adjacent to any of the leakage precursor risk areas that appeared within a preset time window (for example, within the past 30 days). When it is confirmed that there is such a spatio-temporal correlation, the system establishes and outputs a monitoring record chain containing the precursor event and the path formation event to represent the evolution process of the leakage risk at this location.

[0026] Embodiment 1: In the continuous operation scenario of a large water diversion project pump station, the pump station concrete structure is long-term in a strong vibration environment generated by high water head load and joint operation of multiple pump groups. Especially in periods of significant seasonal temperature changes, the structure will produce a slow and uneven stress field redistribution due to thermal expansion and contraction. Such working conditions make it difficult for conventional monitoring methods to distinguish between the slow drift of vibration transmission characteristics caused by this benign, non-damaging factor and the damage signal caused by the newly generated micro-leakage channel in the structure. The system of the present application is continuously deployed and operated in this scenario. The vibration signals collected by the distributed sensing array are sent to the static topological pattern identification subsystem for periodic processing. Within a period of time, the subsystem identifies a continuous high-coherence abnormal path located in the side wall area of the pump house connected by multiple sensor nodes by comparing the real-time coherence matrix with the healthy state benchmark matrix. The appearance of this path is consistent with the characteristics of a potential leakage channel from the static topological pattern. However, the system does not immediately output a warning at this time, but maintains attention to the path. At the same time, the dynamic event decision arbitration subsystem is in standby state.

[0027] During the subsequent operation, the pump station performed a planned pump set switching operation for adjusting flow, which resulted in a global, short-duration transient vibration event in the structure. This event was captured by the pump set transient event monitoring unit in the dynamic event decision arbitration subsystem, which immediately activated the path response synchronization verification module. The module then extracted the response time stamp sequence of all sensor nodes on the previously identified abnormal path with high coherence, which was identified by the static topology pattern recognition subsystem , and calculated the standard deviation of the time sequence , which showed that the value of was greater than the preset time standard deviation threshold , indicating that the transient vibration had no high time synchronization in the conduction between the points on the path, and the response mode was consistent with the propagation characteristics of the vibration in the healthy and diffuse concrete medium. Based on this dynamic arbitration result, the system determined that the abnormal path with high coherence was not a real physical leakage channel, but a non-damage feature caused by the slow drift of the structural stress field due to temperature changes. Therefore, the system finally suppressed the output of this early warning. The operation mode of the system is to combine the static topology pattern recognition subsystem and the dynamic event decision arbitration subsystem. The former is used to preliminarily identify potential abnormal paths, and the latter uses pump set switching transient events in the operation process to finally confirm the preliminary identification results. This method of using structural operation-generated vibrations as detection and verification signal sources provides a reliable criterion for distinguishing between real structural damage and benign state drift without relying on external excitation.

[0028] Example 2: To objectively verify the effectiveness of the system of the present application in distinguishing real structural damage from benign state drift under strong vibration background, and the synergistic effect of the dual mechanisms of static topology pattern recognition and dynamic event decision arbitration, the following verification test is carried out; the test is carried out on a 2mx2mx1m C30 concrete test block, a 4x4 distributed sensing array composed of 16 piezoelectric acceleration sensors is deployed on the surface of the test block to simulate the layout of field sensors, all sensors are connected to a synchronous data acquisition system, the sampling frequency is set to 4096Hz to obtain the structural vibration signals, the test platform includes a wide-band mechanical exciter, which continuously outputs random vibration signals from 10Hz to 500Hz to the test block to simulate the background vibration generated by the operation of the pump set in the water conservancy pumping station, the platform also integrates an electric heating system, which can generate a slow and non-uniform temperature field change in the test block through process control to simulate the benign state drift of the structure caused by seasonal temperature changes, at the same time, a pipe with a controllable small opening is embedded in the test block and connected to a precision water injection pump to artificially create a physically continuous water leakage channel at a specified location and time.

[0029] Two treatment groups are set in this test: the test group uses the complete system of the present application, i.e. both the static topology pattern recognition subsystem and the dynamic event decision arbitration subsystem are enabled; the control group only enables the static topology pattern recognition subsystem and does not have the function of dynamic event decision arbitration to simulate the prior art method that relies only on static pattern comparison, the test is divided into two stages; the first stage is the benign state drift test, under the condition that the mechanical exciter is continuously running, the electric heating system is started to slowly heat the test block to simulate the structural state change caused by temperature stress, in this process, the static topology pattern recognition subsystem of the control group identified a continuous high coherence abnormal path connecting sensors A5 to C7 at the 12th minute of its operation and immediately outputted a warning signal, at the same time, the static topology pattern recognition subsystem of the test group also identified the same path, but its dynamic event decision arbitration subsystem did not immediately confirm the warning, at the 15th minute of operation, a global transient vibration event simulating the start-stop of the pump set was applied to the test block by a transient impact hammer, the dynamic event decision arbitration subsystem of the test group captured this event and calculated the standard deviation of the response time stamps of each sensor on the high coherence abnormal path which is greater than the preset time standard deviation threshold , the system determines that the path is not a real physical channel and suppresses the output of the early warning signal; the second stage is a real leakage path test, stop the electric heating system and wait for the test block temperature to recover stable, then start the precision water injection pump to inject water into the embedded pipeline, forming a physically continuous leakage channel inside the test block. In this process, the static topology pattern recognition subsystem of both the control group and the test group identified a continuous high coherence abnormal path connecting sensors B2 to D4 corresponding to the location of the leakage channel at the 8th minute after the start of water injection. The control group immediately outputs the early warning signal. When running to the 10th minute, a same global transient vibration event is applied again. After capturing this event, the dynamic event decision arbitration subsystem of the test group calculates the standard deviation of the response time stamp of each sensor on the path to be 2.8 ms, which is less than the preset time standard deviation threshold , the system confirms that the path is a real physical leakage channel, and generates and outputs the leakage early warning signal.

[0030] Table 1: Test data comparison table.

[0031]

[0032] Referring to Table 1, the test results show that the control group using only the static topology pattern recognition produces a false alarm in the first stage, while the test group using the system of the present application successfully suppresses the false alarm caused by benign drift in the first stage by introducing the dynamic event decision arbitration subsystem as a second re-verification, and uses the physical feature of transient event response synchronicity to confirm the true leakage path in the second stage. The test data confirms that the double decision mechanism of the present application has the ability to distinguish between real damage and non-damage state changes when monitoring structural leakage in a strong vibration background; it needs to be further explained that in the second stage of the real leakage path test, within the period after the start of the precision water injection pump and before the static topology pattern recognition subsystem first identifies the continuously high-coherence abnormal path at the 8th minute, the single-point response complexity calculation unit of the system has monitored the entropy metric indicators of sensors B2, B3 and C3 located near the preset leakage area, which show a sustained, more than preset threshold rise compared to their healthy state entropy benchmark. The system has thus generated and determined this area as a leakage precursor risk area at the 5th minute, and marked it in the form of a highlighted risk cloud in the visual output module; and after the dynamic event decision arbitration subsystem finally confirms that the path connecting sensors B2 to D4 is a real physical leakage channel at the 10th minute, the coherence spectrum analysis unit of the system is automatically activated, and the coherence spectrum function between each sensor pair on this path shows a flat high-value platform shape in a continuous frequency range of more than 200 Hz, and the channel maturity qualitative indication unit assesses the maturity state of this high-coherence abnormal path as mature, and outputs this assessment result together with the final leakage warning signal.

[0033] To further verify the necessity and technical effects of the dynamic event decision arbitration mechanism of the present application, the following comparative examples are set to compare and illustrate.

[0034] Comparative Example 1: To verify the limitations of using only static topology pattern recognition for leakage monitoring, this comparative example uses a monitoring system that does not include the dynamic event decision arbitration subsystem of this invention. That is, after identifying a continuous high-coherence abnormal path through static comparison, the system directly outputs an early warning signal. Apart from this, the test platform, sensor array, data acquisition parameters, background vibration simulation method, benign state drift simulation method, and actual leakage path manufacturing method of this comparative example are completely consistent with those in Example 2. The test is also divided into two stages; the first stage is the benign state drift test. While maintaining continuous operation of the mechanical vibrator, the electric heating system is activated to slowly heat the test block to simulate non-damaging structural stress field changes caused by seasonal temperature variations. At the 12-minute mark of this stage, the static topology pattern recognition subsystem of the monitoring system identifies a path in the real-time coherence matrix connecting sensors A5 to C7, where the coherence values ​​between nodes are consistently higher than a preset threshold. The first stage involves a continuous, highly coherent abnormal path. Due to the lack of a dynamic verification mechanism, the system immediately determines that leakage has occurred based on this static identification result and generates and outputs an early warning signal. In a real engineering scenario, this early warning signal will trigger an unnecessary emergency investigation procedure, which is expected to take 4 man-hours and may lead to preventive load reduction of the pump unit. The second stage is a real leakage path test. After the electric heating system is stopped and the temperature of the test block recovers and stabilizes, the precision water injection pump is started while the mechanical vibrator continues to run, forming a physical leakage channel inside the test block that corresponds to the preset position (connecting sensors B2 to D4). When this stage reaches the 8th minute, the static topology pattern recognition subsystem of the monitoring system also identifies a continuous, highly coherent abnormal path that completely corresponds to the position of the physical channel and connects sensors B2 to D4, and then generates and outputs an early warning signal. The test data is recorded in the table below.

[0035] Table 2: Data Recording Table for Comparative Example 1.

[0036]

[0037] The experimental results show that although the monitoring method relying solely on static topology pattern recognition can identify the actual physical leakage channels in the second stage, it misjudges non-damaging structural state drift caused by temperature changes as leakage in the first stage, thus generating false alarms. This result confirms that without a mechanism for dynamic arbitration and confirmation of the preliminary identification results, the system cannot distinguish between real structural damage and benign state changes, making it difficult to meet the requirements of early warning reliability for practical engineering applications.

[0038] Example 3: This example combines Figs. 1 to 3 This document describes a real-time monitoring and early warning system for leakage in the concrete structure of a water conservancy pumping station.Fig. 1 As shown, the distributed sensing array deployed on the surface of the concrete structure of the water conservancy pumping station perceives the vibration and synchronously collects the multi-channel vibration signals by the full-time domain signal acquisition unit. The vibration signal data stream is sent to the single-point response complexity calculation unit to calculate the entropy metric index and identify the leakage precursor, thereby defining the leakage precursor risk area. On the other hand, the vibration signal data stream is sent to the static topology pattern recognition subsystem, which preliminarily identifies the continuously high-coherence abnormal path by comparing with the healthy state benchmark library. The preliminary identification result is sent to the dynamic event decision arbitration subsystem as a signal to be confirmed. The arbitration subsystem receives the pump set transient event generated by the pump set start-stop or working condition switching as a verification signal, uses the transient event to verify the authenticity of the path to suppress false positives, and finally outputs the confirmed path leakage warning signal. The warning signal is then sent to the channel maturity qualitative indication unit to evaluate the path development stage based on the coherence spectrum, and is synchronously sent to the visualization output module, which marks the leakage location, direction and risk in the three-dimensional model combined with the information of the leakage precursor risk area.

[0039] As shown in Fig. 2 , the horizontal axis of the graph is frequency, unit Hz, and the vertical axis is the coherence spectrum function . The narrowband characteristic of a nascent leakage channel is indicated by a solid line, and its coherence spectrum function shows concentrated peaks at one or more specific narrowband frequencies. The wideband platform characteristic of a mature leakage channel is indicated by a dashed line, and its coherence spectrum function shows a flat high-value platform in a continuous wide frequency range. As shown in Fig. 3 , the architecture diagram clearly reveals the composition and logical relationship of the system from bottom to top. The bottom layer is the concrete structure of the water conservancy pumping station as the monitoring object and the pump set operating excitation vibration signal source as the excitation source. The physical sensing layer is the distributed sensing array piezoelectric acceleration sensor. The data acquisition layer is the full-time domain signal acquisition unit. The core processing layer is composed of the static topology pattern recognition subsystem and the dynamic event decision arbitration subsystem in parallel. The static subsystem is responsible for calculating the real-time coherence matrix and comparing with the healthy state benchmark to identify the high-coherence abnormal path. The dynamic subsystem is responsible for monitoring the pump set transient event and calculating the path response synchronicity 、 to evaluate the maturity of the channel. All processing results are finally collected to the top-level warning decision core and output layer, which includes the risk evolution evaluation module and the visualization output module, and generates the system output including the leakage warning signal, path location and direction, channel maturity status and risk evolution process.

[0040] Embodiment 4: This embodiment aims to describe the standardization calibration procedure of each core decision threshold in the system of the present application, to ensure that the monitoring sensitivity and reliability performance of the system meet the preset technical requirements when it is deployed in a specific engineering application; when the system of the present application is first deployed in a water conservancy pumping station with a confirmed structural health, in order to adapt the internal algorithm and decision logic of the system to the vibration transmission characteristics of the specific concrete structure and the operation conditions of the pump set, it is necessary to perform a systematic calibration procedure in the initial state, the purpose of this procedure is to determine the key decision threshold in the static topology pattern recognition subsystem and the dynamic event decision arbitration subsystem using the vibration response data of the structure in the healthy state, the initial state of this procedure is defined as: all hardware of the system, including the distributed sensing array and the full-time signal acquisition unit, have been installed and are operating normally, and the pumping station can operate stably in at least one or more typical working conditions according to its regular operation plan; this calibration procedure first enters a baseline state learning phase for not less than 72 hours, during which the system does not generate any early warning, its task is to continuously and synchronously acquire and store the full-quantity vibration signal raw data stream covering different typical operation conditions of the pump set, including single-pump low-load operation, multi-pump high-load operation, and at least two complete pump set start-stop transient processes. After data acquisition is completed, the system enters the offline calibration calculation phase, first, for the continuous high-coherence abnormal path recognition threshold in the static topology pattern recognition subsystem , the system segments the collected data according to different stable working conditions, and for each working condition data segment, it calculates its health state baseline matrix according to the method , in order to determine the coefficient in the formula , the system performs a sensitivity and false alarm rate trade-off analysis, i.e. taking the value of as a variable, increasing from 2.0 to 6.0 with a step of 0.1, repeatedly testing the health data itself in the baseline state learning phase for path search, and counting the number of high-coherence paths misidentified by the system due to background vibration random fluctuations under different values. Through this test, a curve representing the relationship between value and system false alarm rate can be obtained, according to which a minimum value that can control the false alarm rate in the healthy state to below is selected. In this embodiment, the value of is finally determined to be 4.5 through this procedure.

[0041] Secondly, for the time standard deviation threshold Calibration is performed, and the system extracts two or more global transient vibration event data segments of pump group start-stop from the data of the reference state learning stage. For each transient event, the system randomly selects a large number of sensor pairs that are not adjacent in space and do not form a direct propagation path in physics in the sensor array, calculates the standard deviation of the response time stamp sequence of these pseudo paths for the transient event, and thus obtains a statistical distribution of the vibration dispersion propagation time difference in the healthy structure, The value is set on the basis of the need to significantly distinguish from such dispersion propagation characteristics, and therefore, the 5% quantile point of the statistical distribution is set as the value of In this embodiment, the value of is 4.8 ms; finally, the specific logic of the path search algorithm in the static topology pattern recognition subsystem is clarified, which is configured to connect all sensor pairs in the real-time coherence matrix whose values are greater than the threshold value as an edge of the topology graph, and uses a breadth-first search algorithm to search for whether there is a simple path with a length of not less than 3 nodes in the graph from each sensor node with at least one valid connection. Once any path meeting this condition is found, it is determined that a continuous high-coherence abnormal path is preliminarily identified.

[0042] Embodiment 5: This embodiment aims to describe the adaptive management and updating procedure of the multi-task condition health state reference matrix in the system, to ensure the stability of the system when the pump group running combination mode changes and the structure state evolves slowly for a long time; in a complex water conservancy pumping station with multiple pump groups and variable running combination modes, the static topology pattern recognition subsystem of the system is further configured to, when performing the initial state calibration procedure, establish the corresponding health state reference matrix for each typical running condition, and extract a power spectrum feature vector that can represent the running condition from the vibration signals collected by one or more reference sensors, and store the feature vector and the corresponding health state reference matrix in association, thereby constructing a reference matrix library containing multiple entries and indexed by the working condition feature vectors. After the system enters the formal monitoring mode, the static topology pattern recognition subsystem will first calculate the power spectrum feature vector of the current vibration signal in real time before performing the comparison operation, and match it with each working condition feature vector stored in the reference matrix library to determine the actual running condition of the current pump group. Subsequently, the system automatically calls the health state reference matrix corresponding to the working condition to perform the subsequent real-time coherence matrix comparison operation.

[0043] To cope with the problem of health state reference matrix drift caused by long-term creep of structural materials or annual cycle of ambient temperature, etc., the system is built-in with a set of reference adaptive updating mechanism, which is linked with the dynamic event decision arbitration subsystem. When a continuous high-coherence abnormal path is preliminarily identified by the static topology pattern recognition subsystem, but the path is subsequently determined as a non-true physical channel by the dynamic event decision arbitration subsystem according to the result of the transient event response synchronicity check and is suppressed from the warning, the system will record the event which is statically identified but dynamically denied. When the cumulative number of such events for a same spatial path under a specific operating condition exceeds a preset frequency threshold within a preset statistical period, the system determines that the health state reference matrix corresponding to the operating condition has drifted significantly. At this time, the system will automatically trigger a reference updating process for the specific condition, i.e., using the vibration signal data collected under the health state and confirmed by the dynamic event decision arbitration subsystem within the statistical period to recalculate and overwrite the original drifted health state reference matrix.

[0044] Embodiment 6: This embodiment aims to describe the standardization setting procedure of the key algorithm parameters in the single-point response complexity calculation unit and the channel maturity qualitative indication unit in the system of the present application, so as to ensure that the system has reproducible monitoring performance when deployed on different structures. In a specific engineering application, in order to enable the permutation entropy algorithm used by the single-point response complexity calculation unit to effectively represent the vibration response characteristics of a specific concrete structure, the core parameters, i.e., the embedding dimension and the delay time , need to be adapted. This calibration procedure is performed after the system completes the initial deployment and collects the health state reference vibration data covering typical operating conditions. First, to determine the embedding dimension , the system analyzes the vibration signal time series under the health state using the pseudo-neighbor point method, finds the minimum embedding dimension corresponding to the first time when the proportion of pseudo-neighbor points drops to a preset low value (such as 1%) by calculating the change of the distance between adjacent points in the phase space under different embedding dimensions, and takes the value as the embedding dimension . Second, to determine the delay time , the system calculates the information correlation degree between the original time series and its delayed sequence using the mutual information method, takes the delay time corresponding to the first minimum point of the mutual information function as the delay time . Through this procedure, the two key parameters of the permutation entropy algorithm are determined by the vibration data of the structure itself.

[0045] To enable the channel maturity qualitative indication unit to quantitatively distinguish the severity of the leakage path, the system needs to establish a corresponding relationship model between the coherence spectrum pattern and the physical maturity, the construction of which can be completed through offline calibration experiments. In the concrete block test platform, a series of leakage channels representing different maturity states are manufactured in the pre-embedded pipeline by a precision water injection pump at different, increasing constant flow rates, and the coherence spectrum function of the sensor vibration signal at both ends of the leakage channel under each flow rate is recorded ; then, the system calculates a quantitative indicator representing the bandwidth of each recorded coherence spectrum function, i.e. the effective coherence bandwidth , which is defined as the total width of all frequency intervals with a coherence value greater than 50% of its peak value; by correlating the leakage state under different flow rates with the calculated , a lookup table or a function relationship can be established to determine a bandwidth threshold for distinguishing the newborn period from the mature period ; when the system enters the formal monitoring mode, when the channel maturity qualitative indication unit calculates the effective coherence bandwidth of a newly discovered high-coherence abnormal path, by comparing it with the bandwidth threshold , the maturity state of the path can be objectively evaluated.

[0046] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0047] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations, characterized in that, The system includes: A distributed sensing array deployed on the surface of a concrete structure; A full-time domain signal acquisition unit is connected to a distributed sensing array and is configured to synchronously acquire vibration signals output by various sensors excited by the pump sets operated by the water pump station. A static topology pattern recognition subsystem is configured to periodically calculate the coherence between all sensor pairs in a distributed sensing array based on vibration signals to construct a real-time coherence matrix, and to preliminarily identify a continuous high-coherence anomaly path by comparing the real-time coherence matrix with a pre-established health-state benchmark matrix in which the concrete structure is in a confirmed healthy state. A dynamic event decision arbitration subsystem includes a pump group transient event monitoring unit and a path response synchronization verification module. The pump group transient event monitoring unit is configured to identify the occurrence time of global transient vibration events caused by pump group start-up, shutdown, or operating condition switching from vibration signals. The path response synchronization verification module is configured to be activated only when the static topology pattern recognition subsystem identifies a highly coherent abnormal path and the pump group transient event monitoring unit captures the transient vibration event. It calculates and compares the response timestamps of all sensor nodes on the path to the transient vibration event to obtain a quantitative index characterizing the response synchronization. When the synchronization quantitative index meets a preset judgment condition, the authenticity of the highly coherent abnormal path is confirmed, and a leakage warning signal is generated and output only after confirmation.

2. The real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The static topology pattern recognition subsystem is further configured to: invoke one or more health-state reference matrices, which are established by collecting data on one or more preset operating conditions of the pump group, respectively, while the concrete structure is in a confirmed healthy state; and when performing comparison, select the corresponding health-state reference matrix from the one or more health-state reference matrices according to the actual operating conditions of the current pump group to perform the comparison operation.

3. The real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The system also includes: a single-point response complexity calculation unit, configured to calculate an entropy metric representing the time series complexity of the vibration response signal of each independent sensor in the distributed sensing array based on the vibration signal acquired by the full-time domain signal acquisition unit; and a leakage precursor identification module, configured to compare the entropy metric of each independent sensor with a healthy state entropy benchmark. When the entropy metric of one or more neighboring sensors is identified as consistently higher than the healthy state entropy benchmark, the structural region corresponding to the sensor is identified as a leakage precursor risk zone, and spatial location and temporal information representing the leakage precursor risk zone are generated.

4. The real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The system also includes: a coherence spectrum analysis unit, configured to calculate the coherence spectrum function of the vibration signal of the sensor pair on the path after the dynamic event decision arbitration subsystem initially identifies the high coherence anomaly path; and a channel maturity qualitative indicator unit, configured to qualitatively assess the maturity status of the high coherence anomaly path based on the spectral morphology characteristics of the coherence spectrum function, and output the assessed maturity status along with the leakage warning signal.

5. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 4, characterized in that, The channel maturity qualitative indicator unit is configured to perform the following operations: if the coherence spectral function exhibits one or more narrowband peaks concentrated at a specific frequency, the maturity status of the high coherence anomaly path is assessed as nascent; and if the coherence spectral function exhibits a flat high-value plateau over a continuous wide frequency range, the maturity status of the high coherence anomaly path is assessed as mature.

6. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The path response synchronization verification module is configured to extract the response timestamp sequence of all sensor nodes on the highly coherent abnormal path to transient vibration events. ,in N is the total number of sensor nodes on the path; Calculate the response timestamp sequence Standard deviation As a quantitative indicator of synchronicity; The preset judgment condition is ,in, This is a time standard deviation threshold determined based on historical health data statistics.

7. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The system also includes a visualization output module, which is configured to determine the origin and extension direction of the leakage based on the spatial topological location of the finally confirmed high coherence anomaly path in the distributed sensing array, and mark the origin and extension direction in the three-dimensional structural model.

8. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 3, characterized in that, The single-point response complexity calculation unit is configured to: convert the time series of vibration response signals of each independent sensor into trajectories in multi-dimensional phase space through a phase space reconstruction method; then, calculate the complexity of the sequence pattern of the phase space trajectory using the permutation entropy algorithm, and use the calculation result as an entropy metric.

9. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 1, characterized in that, The pump unit transient event monitoring unit is configured to perform short-time energy analysis on vibration signals, and identify the occurrence of transient vibration events when the rate of change of short-time energy of the vibration signal exceeds a threshold of energy change rate determined based on historical operating data.

10. A real-time monitoring and early warning system for leakage in concrete structures of water conservancy pumping stations according to claim 3, characterized in that, The system also includes a risk evolution assessment module, which is configured to receive spatial location and temporal information representing the risk zone of leakage precursors generated by the leakage precursor identification module, and to receive spatial location and temporal information of continuous high coherence anomaly paths identified by the static topology pattern recognition subsystem. Based on spatial location and temporal information, spatiotemporal data correlation is performed to establish a monitoring and recording chain from the emergence of the early risk zone of leakage to the formation of a continuous high coherence anomaly path, and the monitoring and recording chain is output to characterize the evolution process of leakage risk.

Images