An intelligent disease early warning method for inverted siphon based on multi-source data collection

Abstract

The present application relates to the technical field of inverted siphon disease detection, and particularly relates to an intelligent early warning method for inverted siphon diseases based on multi-source data acquisition, comprising: acquiring upstream and downstream water levels, flow, outlet turbidity, gate opening and reference turbidity data; determining a reference time period based on the flow and using the normal fluctuation range of the downstream water level to filter an effective water level difference sequence, generating a first marker when a one-way increasing trend is detected and determining a reference water level difference and a reference flow; in response to the first marker, dividing a flow interval based on the reference flow, correcting the outlet turbidity using the reference turbidity and judging the loss of sensitivity to generate a second marker; in response to the second marker, identifying the sand washing operation and correcting the water level difference to obtain a recovery offset; comparing with the historical recovery offset, generating an early warning signal when the degradation condition is met. Through multi-source data fusion and hierarchical identification, the present application realizes early warning of the evolution of sediments from live mud to dead mud, and overcomes the problem that traditional monitoring cannot distinguish the state of accumulated substances.

Description

Technical Field

[0001] This invention relates to the field of inverted siphon defect detection technology, and in particular to an intelligent early warning method for inverted siphon defects based on multi-source data acquisition. Background Technology

[0002] Inverted siphons are common hydraulic structures used in long-distance water conveyance and irrigation projects to cross rivers, valleys, or transportation routes. Their pipes are U-shaped and concave, and the inlet section typically includes a sedimentation basin to reduce the amount of sediment entering the pipe. In actual operation, these pipelines are equipped with upstream and downstream level gauges and flow meters to continuously monitor head loss and flow capacity changes during water conveyance. Some projects also install online turbidity meters at the outlet maintenance section to understand the outlet water quality or assess the effectiveness of sediment flushing. The real-time data collected by these instruments, along with the gate control system's opening records, are usually integrated into a SCADA system for centralized storage and monitoring. In daily operation and maintenance, technicians primarily determine whether significant sediment deposition has occurred in the inverted siphon by setting alarm limits or manually interpreting the magnitude and trend of the upstream and downstream water level difference, supplemented by the instantaneous turbidity readings provided by the turbidity meter. When the water level difference remains persistently high or the turbidity is abnormal, hydraulic sediment flushing is arranged in a timely manner, and the operation process is traced using gate opening records. This method of discrimination based on a single or a few indicator thresholds constitutes the basic framework for the current monitoring of sludge accumulation in inverted siphons.

[0003] For inverted siphon pipes, influenced by the periodic flow fluctuations caused by water supply scheduling, the sediment inside the pipe is not always in a single mode of static siltation or uniform scouring. Instead, it frequently experiences sediment settling at night under low flow rates and resuspension during the day under high flow rates. As the years of operation increase, the originally loosely structured floating silt layer, which can be easily carried away by hydraulics, gradually transforms into a dense, hardened bottom of dead silt under repeated pressure and drainage consolidation. Existing monitoring technologies rely solely on the absolute value or slow trend drift of the upstream and downstream water level difference, and cannot distinguish whether the blockage is caused by easily scourable floating silt or by hardened dead silt, let alone identify the stage of scouring sensitivity decay during the transformation of floating silt into dead silt. On-site personnel have no way of knowing whether the sediment layer at the bottom of the pipe is currently in a flushable, partially flushable, or completely inoperable state. This makes decision-making for low-pressure flushing operations lack targeted basis. Often, action is not taken during the window of opportunity when the floating mud can still be effectively removed, and only when the sediment layer has become deeply compacted and conventional hydraulic flushing is almost ineffective is a reactive measure taken, ultimately forcing a shutdown and costly mechanical dredging. Therefore, the industry urgently needs a method that can integrate multi-source conventional monitoring data to automatically identify and classify the stages of the evolution of the physical properties of sediments at the bottom of the pipe, so as to provide early warnings when flushing sensitivity is gradually lost and issue an alarm when irreversible compaction is truly established. Summary of the Invention

[0004] To address this issue, the present invention provides an intelligent early warning method for inverted siphon defects based on multi-source data acquisition, which overcomes the problem in existing technologies that cannot classify and identify the evolution of active to dead sediment at the bottom of the pipe in monitoring and early warning of inverted siphon defects.

[0005] To achieve the above objectives, this invention provides an intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition, comprising: Step S1: Obtain upstream and downstream water level data, flow rate data, outlet turbidity data, gate opening data, and reference turbidity data used to correct the outlet turbidity data of the inverted siphon. Step S2: Determine the reference time period based on the flow data; select a time interval within the reference time period where the downstream water level is in the normal fluctuation range; and determine the effective water level difference sequence based on the upstream and downstream water level difference within the selected time interval; perform trend detection on the effective water level difference sequence; when a unidirectional increasing trend exceeding the normal fluctuation range of the same period in history is detected, generate a first anomaly marker; and determine the water level difference and flow rate corresponding to triggering the unidirectional increasing trend as the reference water level difference and reference flow rate, respectively. Step S3: In response to the first anomaly marker, the subsequent flow process is divided into multiple flow intervals based on the reference flow rate, and the outlet turbidity data is corrected based on the reference turbidity data to obtain flushing turbidity change characteristics; when the flushing turbidity change characteristics meet the preset sensitivity loss condition, a second state marker is generated. Step S4: In response to the second state flag, identify the sand flushing operation event based on the gate opening data, and obtain upstream and downstream water level data and flow data before, during, and after the sand flushing operation; determine the water level change component caused by the independent downstream regulation based on the change of the downstream water level data, correct the upstream and downstream water level difference after the sand flushing operation based on the component, obtain the corrected water level difference, and determine the difference between the corrected water level difference and the reference water level difference as the recovery offset; Step S5: Compare the recovery offset corresponding to the current sand flushing operation with the recovery offset of the historical sand flushing operation recorded before the second state mark was generated. When the comparison result meets the preset degradation condition, generate an early warning signal.

[0006] Further, step S2 includes: Step S21: Based on the daily traffic variation range of each preset time period, determine the time period in which the traffic variation range is consistently lower than a preset threshold as the benchmark time period; Step S22: Divide the reference time period into multiple sub-time periods and obtain the upstream and downstream water level difference and the downstream water level change within each sub-time period. Step S23: For each sub-period, if the change in its downstream water level is within the normal fluctuation range of the downstream water level, then the difference between the upstream and downstream water levels of the sub-period is taken as an effective water level difference sample. Step S24: Based on the effective water level difference samples within the same reference time period, calculate the reference representative water level difference, and form the effective water level difference sequence from the reference representative water level differences over multiple consecutive days; Step S25: Based on the fluctuation characteristics of the historical effective water level difference sequence, determine the normal fluctuation boundary and detect the deviation of the current effective water level difference sequence from the normal fluctuation boundary. Step S26: When the deviation state is manifested as exceeding the normal fluctuation boundary at multiple consecutive time points and the deviation amount is monotonically increasing, it is determined that a unidirectional increasing trend has occurred, and the benchmark representative water level difference and benchmark representative flow rate on the day that the unidirectional increasing trend is triggered are respectively determined as the reference water level difference and reference flow rate.

[0007] Furthermore, the normal fluctuation boundary mentioned in step S25 is determined by upper and lower boundary values ​​based on the statistical distribution characteristics of the effective water level difference sequence in the same historical season.

[0008] Further, step S3 includes: Step S31: Divide the daily flow process into a baseline low flow range and at least one flushing flow range higher than the reference flow rate according to the reference flow rate. Step S32: Within the flushing flow range, based on the outlet turbidity data and the reference turbidity data, the corrected turbidity increment is obtained time-by-time, and a corrected turbidity increment sequence is formed. Step S33: Evaluate the amplitude characteristics of the corrected turbidity increment sequence and the synchronicity of the changes between the corrected turbidity increment sequence and the corresponding flow sequence. When both the amplitude characteristics and the synchronicity of the changes meet the preset sensitivity loss criteria, generate the second state flag.

[0009] Furthermore, the preset sensitivity loss criteria include: Within the scouring flow range, the amplitude index of the corrected turbidity increment sequence is lower than the lower limit of amplitude determined based on historical normal scouring events in the same range. After experiencing a predetermined number of daytime flushing periods, the amplitude index of the corrected turbidity increment sequence is lower than the preset low turbidity state threshold within the baseline low flow range.

[0010] Further, in step S33, the change synchronicity is to obtain a synchronicity quantification index of the corrected turbidity increment sequence and the flow sequence within a sliding time window. When the synchronicity quantification index is lower than a preset synchronicity threshold within multiple consecutive sliding time windows, it is determined that the change synchronicity meets the preset sensitivity loss standard.

[0011] Further, step S4 includes: Step S41: Based on the gate opening data and flow data, the gate action that meets the following conditions is identified as the sand flushing operation event: the flow starts from the initial level located in the vicinity of the reference flow, rises to a level exceeding the sand flushing flow trigger level within a preset short period of time, and remains above the trigger level for a duration greater than the preset minimum duration. Step S42: Obtain upstream and downstream water level data, downstream water level data, and flow rate data during the first time period before the sand flushing operation event and the second time period after the event. Step S43: Based on the changes in the downstream water level data before and after the flushing operation, determine the water level change component caused by the independent scheduling behavior of the downstream, and correct the upstream and downstream water level difference after the flushing operation based on the component to obtain the corrected water level difference. Step S44: After the sand flushing operation is completed and the flow rate recovers to the range near the reference flow rate, the difference between the corrected water level difference and the reference water level difference at this time is determined as the recovery offset.

[0012] Further, step S5 includes: Step S51: Extract the most recent historical sand flushing operation event from the history record before the second state marker was generated; Step S52: Obtain the historical reference water level difference, historical sludge flushing intensity, and historical recovery offset corresponding to the historical sludge flushing operation event; Step S53: When the difference between the reference water level difference of the current sand flushing operation and the historical reference water level difference is less than the first allowable deviation, and the difference between the current sand flushing intensity and the historical sand flushing intensity is less than the second allowable deviation, compare the recovery offset with the historical recovery offset. Step S54: When the recovery offset is less than the product of the historical recovery offset and the preset attenuation coefficient, and the difference between the historical recovery offset and the recovery offset is greater than the preset degradation tolerance value, it is determined that the preset degradation condition is met, and an early warning signal is generated.

[0013] Furthermore, the reference turbidity data is determined as follows: When upstream turbidity data is collected synchronously with the outlet turbidity data, the upstream turbidity data is used as the reference turbidity data; When upstream turbidity data is unavailable, the turbidity background value obtained by statistical analysis of outlet turbidity data from the same historical period without rainfall and with the same flow rate as the baseline low flow range will be used as the reference turbidity data.

[0014] Furthermore, the normal fluctuation range of downstream water level in step S23 is the range of variation determined by a preset statistical method based on the downstream water level data of the same historical period and when the flow rate is in the same order of magnitude as the benchmark period.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: by integrating multi-source monitoring data, the present invention screens the effective water level difference sequence under steady-state conditions and detects its unidirectional increasing trend to generate a first anomaly marker. After background turbidity correction and flow synchronization analysis, the loss of scour sensitivity is diagnosed to generate a second state marker. After excluding the influence of independent downstream scheduling, the sediment recovery offset is compared to generate an early warning. Thus, the present invention realizes the graded state identification and early warning of the evolution process of sediment from active mud to dead mud at the bottom of the pipe, overcoming the problem that the existing single-index monitoring cannot distinguish the changes in the physical state of sediment.

[0016] Furthermore, by selecting a baseline period with stable flow, the present invention uses the downstream water level variation to exclude boundary disturbance sub-periods to extract effective water level difference samples. Based on the unidirectional increasing trend of the historical statistical boundary detection sequence, it can detect the real flow resistance rise caused by sediment accumulation in a timely manner, reducing misjudgments caused by independent fluctuations in downstream water level.

[0017] Furthermore, by dividing the flow rate interval and correcting the outlet turbidity with a reference turbidity, the present invention extracts the amplitude characteristics of the scouring turbidity increment sequence and its synchronous change information with the flow rate, and generates a status marker based on a preset sensitivity loss standard, thereby accurately diagnosing early signs of scouring capacity decline during the transformation of floating mud into compaction.

[0018] Furthermore, this invention identifies sand flushing operation events based on gate action and flow response, and determines and deducts the water level change component caused by independent downstream scheduling based on the downstream water level change itself, correcting the upstream and downstream water level difference after sand flushing. The obtained recovery offset can reflect the true degree of removal of sediment at the bottom of the pipe, eliminating the interference of external scheduling on the effect evaluation.

[0019] Furthermore, this invention, under the premise that the reference water level difference and flushing intensity meet the comparable conditions, compares the current flushing recovery offset with the historical recovery offset, and issues an early warning when the recovery capacity declines, thus ensuring the accuracy of the determination of irreversible compaction of sediment at the bottom of the pipe and avoiding ineffective flushing and subsequent passive mechanical dredging. Attached Figure Description

[0020] Figure 1 This is a flowchart of the intelligent early warning method for inverted siphon defects based on multi-source data acquisition according to the present invention; Figure 2 This is a flowchart of step S2 of the intelligent early warning method for inverted siphon defects based on multi-source data acquisition of the present invention; Figure 3 This is a flowchart of step S3 of the intelligent early warning method for inverted siphon defects based on multi-source data acquisition of the present invention; Figure 4 This is a flowchart of step S4 of the intelligent early warning method for inverted siphon defects based on multi-source data acquisition according to the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; 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] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0024] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Please see Figure 1 The diagram shows a flowchart of an intelligent early warning method for inverted siphon defects based on multi-source data acquisition, according to an embodiment of the present invention. The method includes: Step S1: Obtain upstream and downstream water level data, flow rate data, outlet turbidity data, gate opening data, and reference turbidity data used to correct the outlet turbidity data of the inverted siphon. Specifically, the reference turbidity data is determined as follows: When upstream turbidity data is collected synchronously with the outlet turbidity data, the upstream turbidity data is used as the reference turbidity data; When upstream turbidity data is unavailable, the turbidity background value obtained by statistical analysis of outlet turbidity data from the same historical period without rainfall and with the same flow rate as the baseline low flow range will be used as the reference turbidity data.

[0026] This invention provides an intelligent early warning system for inverted siphon defects based on multi-source data acquisition, comprising a data acquisition unit, a benchmark analysis unit, a scour sensitivity diagnosis unit, a sand flushing effect evaluation unit, and an early warning decision unit, for executing the early warning method described in any of the preceding embodiments.

[0027] The data acquisition unit is connected to the SCADA system, upstream and downstream water level gauges, main pipeline flow meter, outlet turbidity meter, and gate opening sensor of the inverted siphon line to acquire or periodically extract the required multi-source data in real time.

[0028] In this embodiment, step S1 is implemented by the data acquisition unit. The data acquisition unit collects water level and flow time series data from the water level gauge and flow meter via standard industrial communication protocols, acquires continuous outlet turbidity data from the online turbidity meter installed at the outlet maintenance section of the inverted siphon, and retrieves the gate opening action record from the gate control system. The data acquisition unit determines the reference turbidity data used to correct the outlet turbidity data using one of the following two methods: When an online turbidity meter is installed at the inlet of the inverted siphon or in the upstream channel it is connected to, and the turbidity meter and the outlet turbidity meter have the same time stamp synchronous acquisition capability, the data acquisition unit directly uses the real-time turbidity value of the upstream turbidity meter as the reference turbidity data at the corresponding time. When no upstream turbidity data is available for synchronous use, the data acquisition unit retrieves historical outlet turbidity values ​​from the historical database that meet the following filtering criteria: the time period is the same as the historical period, there are no effective rainfall records in the local area, and the pipeline flow rate is within a predetermined low flow range. The selected historical outlet turbidity dataset is then used to calculate the turbidity background value using a preset robust statistical method, and this background value is used as the reference turbidity data within the current analysis window. The predetermined low flow range is of the same or similar flow range as the baseline low flow range defined in subsequent steps based on the reference flow rate, ensuring that the background value represents the natural background of outlet turbidity under this low flow condition.

[0029] This invention introduces reference turbidity data to subtract background turbidity from the outlet turbidity, separating the turbidity increase caused by sediment scouring within the pipe from the turbidity fluctuations of the upstream inflow. During the operation of an inverted siphon, fluctuations in outlet turbidity may originate from two different types of disturbances: first, changes in the concentration of suspended sediment carried by the upstream channel flow itself; and second, sediment accumulated inside the pipe being stirred up and entering the water body at high flow velocities. If outlet turbidity is used directly for sludge diagnosis without differentiation, it is easy to misjudge active sludge scouring when the upstream inflow has increased sediment content, while it may incorrectly conclude that there is a tendency towards dead sediment when the upstream inflow is abnormally clear. This scheme, by subtracting the upstream background in real time when an upstream turbidity meter is available, and utilizing the statistical background of outlet turbidity under historical low-flow, clear weather conditions when an upstream turbidity meter is unavailable, ensures that subsequent assessments of scour turbidity variation characteristics focus entirely on the contribution of resuspended sediment within the pipe. This effectively eliminates the interference of background fluctuations on qualitative pattern recognition, guaranteeing higher reliability of conclusions on mud properties extracted from outlet turbidity passivation or synchronous changes in turbidity and flow rate.

[0030] Understandably, the statistical method for historical outlet turbidity background values ​​can be the median, truncated mean, or preset quantiles. Alternatively, a background modeling method based on probability density estimation can be used, as long as the obtained turbidity background value can stably represent the typical outlet turbidity level under that flow rate without pipe scouring. Similarly, upstream turbidity data can be first processed by moving average or time-delay alignment to eliminate instantaneous deviations caused by instrument noise and the time difference between upstream and downstream water delivery; this will not be elaborated further.

[0031] It is understandable that, based on the aforementioned online instrument data, this embodiment also incorporates in-pipe CCTV inspection as a supplementary detection and troubleshooting method. Specifically, when the inverted siphon pipe has low water levels or temporary drainage conditions, a pipe crawling robot carrying a high-definition camera can be used to enter the pipe and travel along the pipe section to collect real-time images of the pipe wall. This allows for direct observation and recording of apparent defects such as pipe wall cracks, corrosion, misaligned joints, deformation, and siltation at the bottom of the pipe. Based on the image records, the location of defects can be determined by distance, and the type and severity of defects can be classified and assessed. In-pipe CCTV inspection images are intuitive and accurate, but due to limitations in the water level inside the pipe, it is difficult to implement under high water levels or full-pipe flow conditions. It usually needs to be used in conjunction with other detection methods such as sonar detection, which will not be elaborated further here.

[0032] Step S2: Determine the reference time period based on the flow data; select a time interval within the reference time period where the downstream water level is in the normal fluctuation range; and determine the effective water level difference sequence based on the upstream and downstream water level difference within the selected time interval; perform trend detection on the effective water level difference sequence; when a unidirectional increasing trend exceeding the normal fluctuation range of the same period in history is detected, generate a first anomaly marker; and determine the water level difference and flow rate corresponding to triggering the unidirectional increasing trend as the reference water level difference and reference flow rate, respectively. Specifically, step S2 includes: Step S21: Based on the daily traffic variation range of each preset time period, determine the time period in which the traffic variation range is consistently lower than a preset threshold as the benchmark time period; Step S22: Divide the reference time period into multiple sub-time periods and obtain the upstream and downstream water level difference and the downstream water level change within each sub-time period. Step S23: For each sub-period, if the change in its downstream water level is within the normal fluctuation range of the downstream water level, then the difference between the upstream and downstream water levels of the sub-period is taken as an effective water level difference sample. Specifically, the normal fluctuation range of downstream water level in step S23 is the range of variation determined by a preset statistical method based on the downstream water level data of the same historical period and when the flow rate is in the same range as the benchmark period.

[0033] In this embodiment, step S2 is performed by the benchmark analysis unit. The benchmark analysis unit obtains the flow data and upstream and downstream water level data arranged in time series from the data acquisition unit. First, it divides each natural day into several time periods according to a preset time step and calculates the dispersion index of the flow series in each time period. The dispersion index can be a statistical quantity that characterizes the fluctuation range, such as the ratio of standard deviation, range to the mean of the time period, or interquartile range.

[0034] The continuous period when the dispersion index is consistently below a preset stability threshold is identified as the base period for the day. This stability threshold is set based on the statistical characteristics of the historical low-flow stable operation period of the line where the inverted siphon is located, so as to ensure that the base period corresponds to the nighttime or low water usage period when the flow inside the pipe is almost constant and the sediment is mainly deposited.

[0035] After identifying the baseline time period, the baseline analysis unit further divides the baseline time period into sub-time periods of equal length. The length of the sub-time period is determined according to the instrument sampling period, usually ranging from several minutes to tens of minutes.

[0036] For each sub-period, the representative value of the upstream and downstream water level difference and the downstream water level variation are calculated. The downstream water level variation can be expressed as the standard deviation of the downstream water level time series within that sub-period, the difference between the maximum and minimum values, or the offset relative to the mean of adjacent sub-periods. Subsequently, the benchmark analysis unit calls the normal fluctuation range of the downstream water level. This range is a reasonable range of variation determined by a preset robust statistical method, based on the downstream water level variation data of the same historical period when the pipeline flow is in the same low flow level range as the current benchmark period and there are no gate intervention records. For example, it may be a specific quantile range of historical samples. For each sub-period, if the downstream water level variation falls within the normal fluctuation range, the upstream and downstream water level difference of that sub-period is marked as a valid water level difference sample; if it exceeds the range, it is determined that there is an independent disturbance in the downstream boundary conditions within that sub-period, and the corresponding sample is removed. Through the above screening, a set of valid water level difference samples corresponding to the benchmark period is formed, thereby obtaining the benchmark representative water level difference used for trend detection.

[0037] This invention compares the downstream water level variation with the normal fluctuation range to eliminate sub-period samples affected by external factors such as independent downstream scheduling or river water level backflow from the baseline period. This ensures that the final effective water level difference reflects only the pipe's own resistance changes. In inverted siphon operation, the downstream water level is not solely determined by the pipe outlet flow rate; it often fluctuates independently due to the opening and closing of downstream diversion gates, tailrace regulation, or changes in the water level of the incoming river. This fluctuation is directly transmitted to the measured value of the upstream and downstream water level difference. If this is not identified and eliminated, the resulting water level difference sequence will contain a large number of spurious disturbances unrelated to the siltation at the bottom of the pipe, easily obscuring the weak trends caused by the slow accumulation of sediment. This solution uses the downstream water level variation as a criterion to filter out sub-periods with unstable downstream water levels that are significantly different from the normal calm state. This ensures that the data used to construct the baseline flow resistance sequence originates from a window with simple hydraulic conditions and silent boundary conditions, thereby significantly improving the signal-to-noise ratio and the accuracy of subsequent CUSUM unidirectional increasing trend detection and early warning triggering.

[0038] Understandably, in addition to using the quantile method based on the same historical flow rate, the establishment of the normal fluctuation range of downstream water level can also adopt the median absolute deviation method based on a sliding window or the adaptive interval construction method based on a machine learning anomaly detection model, as long as the determined interval can effectively distinguish between the normal slight fluctuations of the downstream boundary and the abnormal changes caused by human or natural disturbances.

[0039] Step S24: Based on the effective water level difference samples within the same reference time period, calculate the reference representative water level difference, and form the effective water level difference sequence from the reference representative water level differences over multiple consecutive days; Step S25: Based on the fluctuation characteristics of the historical effective water level difference sequence, determine the normal fluctuation boundary and detect the deviation of the current effective water level difference sequence from the normal fluctuation boundary. Specifically, the normal fluctuation boundary mentioned in step S25 is determined based on the statistical distribution characteristics of the effective water level difference sequence in the same historical season, which defines the upper and lower boundary values.

[0040] Step S26: When the deviation state is manifested as exceeding the normal fluctuation boundary at multiple consecutive time points and the deviation amount is monotonically increasing, it is determined that a unidirectional increasing trend has occurred, and the benchmark representative water level difference and benchmark representative flow rate on the day that the unidirectional increasing trend is triggered are respectively determined as the reference water level difference and reference flow rate.

[0041] In this embodiment, steps S24 to S26 are executed by the benchmark analysis unit. The benchmark analysis unit performs the following steps: for all valid water level difference samples retained after validity screening in the same benchmark time period, sorts them by value and takes the median as the benchmark representative water level difference for that day; and simultaneously calculates the arithmetic mean of the flow rate sequence within that benchmark time period as the benchmark representative flow rate for that day.

[0042] The benchmark analysis unit arranges the representative water level differences of multiple consecutive natural days in chronological order to form an effective water level difference sequence. Subsequently, the unit retrieves a historical set of effective water level difference sequences generated in the same season according to the same benchmark time period rules and effectiveness screening rules. By analyzing the statistical distribution of this historical set, the unit extracts the upper and lower boundary values ​​that can characterize the normal fluctuation range of the sequence, which are used as the upper and lower boundaries of normal fluctuation, respectively.

[0043] The benchmark analysis unit compares each benchmark representative water level difference in the current effective water level difference sequence with the upper boundary of the normal fluctuation. When multiple consecutive benchmark representative water level differences are observed to be above the upper boundary, and the values ​​of these benchmark representative water level differences show a strict increasing trend day by day, it is determined that the effective water level difference sequence has shown a unidirectional increasing trend. The benchmark representative water level difference on the day when this excess and the increasing sequence first appears is determined as the reference water level difference, and the benchmark representative flow rate on that day is determined as the reference flow rate.

[0044] This invention extracts the reference water level difference only under stable downstream water level conditions during a sub-period and adaptively determines the dynamic normal boundary based on the statistical distribution of the same historical period. This ensures that the long-term drift detection of flow resistance is no longer affected by independent downstream regulation or natural water level fluctuations over the years. During the discrimination process, the joint constraint of continuous exceedances with monotonically increasing deviations eliminates transient artifacts caused by isolated disturbances or instrument noise, ensuring that only irreversible, daily-intensifying accumulation of silt at the bottom of the pipe is flagged. The resulting reference water level difference and reference flow rate accurately reflect the system state at the onset of the basement uplift event, providing a stable and traceable reference coordinate for subsequent scour sensitivity diagnosis.

[0045] Understandably, the calculation of the benchmark water level difference can be replaced by the arithmetic mean or time-weighted average after outlier removal, in addition to using the median, as long as the representative value can robustly reflect the concentration level of the upstream and downstream water level differences within the benchmark period. The determination of the normal fluctuation boundary can be based on the quantile method of historical distribution, or by constructing the boundary by adding or subtracting a certain number of standard deviations from the historical series mean. The identification of a unidirectional increasing trend can be further verified using cumulative sum control chart algorithms or non-parametric trend test methods. When the cumulative offset or test statistic exceeds the preset confidence threshold, the trend is also considered valid; this will not be elaborated further.

[0046] Step S3: In response to the first anomaly marker, the subsequent flow process is divided into multiple flow intervals based on the reference flow rate, and the outlet turbidity data is corrected based on the reference turbidity data to obtain flushing turbidity change characteristics; when the flushing turbidity change characteristics meet the preset sensitivity loss condition, a second state marker is generated. Specifically, step S3 includes: Step S31: Divide the daily flow process into a baseline low flow range and at least one flushing flow range higher than the reference flow rate according to the reference flow rate. Step S32: Within the flushing flow range, based on the outlet turbidity data and the reference turbidity data, the corrected turbidity increment is obtained time-by-time, and a corrected turbidity increment sequence is formed. Step S33: Evaluate the amplitude characteristics of the corrected turbidity increment sequence and the synchronicity of the changes between the corrected turbidity increment sequence and the corresponding flow sequence. When both the amplitude characteristics and the synchronicity of the changes meet the preset sensitivity loss criteria, generate the second state flag.

[0047] Specifically, the preset sensitivity loss criteria include: Within the scouring flow range, the amplitude index of the corrected turbidity increment sequence is lower than the lower limit of amplitude determined based on historical normal scouring events in the same range. After experiencing a predetermined number of daytime flushing periods, the amplitude index of the corrected turbidity increment sequence is lower than the preset low turbidity state threshold within the baseline low flow range.

[0048] Specifically, in step S33, the change synchronicity is to obtain a quantification index of the synchronicity between the corrected turbidity increment sequence and the flow sequence within a sliding time window. When the quantification index of synchronicity is lower than a preset synchronicity threshold within multiple consecutive sliding time windows, it is determined that the change synchronicity meets the preset sensitivity loss standard.

[0049] In this embodiment, steps S31 to S33 are executed by the scour sensitivity diagnosis unit. The scour sensitivity diagnosis unit is triggered to start after receiving the first anomaly flag, retrieves the reference flow from the benchmark analysis unit, and continuously accesses the real-time flow sequence of the day through the data acquisition unit.

[0050] The flushing sensitivity diagnostic unit uses the reference flow rate as the dividing anchor point. It defines the continuous period when the flow rate value is within the preset allowable deviation range of the reference flow rate and the flow fluctuation dispersion is lower than the preset stable limit as the benchmark low flow rate interval; and defines the continuous period when the flow rate value exceeds the preset multiple of the reference flow rate and the rise gradient meets the preset steep rise condition in adjacent sampling intervals as the flushing flow rate interval.

[0051] The flushing flow range retains at least one high-flow process with sufficient duration to meet the sample length requirements of subsequent sequence analysis.

[0052] Within each identified scouring flow range, the scouring sensitivity diagnostic unit aligns the outlet turbidity sequence with the reference turbidity sequence according to the acquisition time. For each alignment time, the outlet turbidity value is subtracted from the corresponding reference turbidity value. The difference is retained if it is positive and zero if it is negative, thus obtaining the corrected turbidity increment at that time.

[0053] The flushing sensitivity diagnostic unit arranges the corrected turbidity increments at all times within a complete flushing flow range in chronological order to form a corrected turbidity increment sequence.

[0054] For amplitude characteristics, the scour sensitivity diagnostic unit performs moving average smoothing on the corrected turbidity increment sequence, and then uses a preset high quantile as the amplitude characteristic index. The selection of this quantile is pre-calibrated based on the distribution law of corrected turbidity increments of scour events of the same magnitude under historical active sludge conditions, making the index robustly represent the degree of turbidity enhancement during scour. The unit compares this amplitude characteristic index with the lower limit of scour amplitude reference stored in the historical database. The lower limit of scour amplitude reference is a conservative lower bound determined by extracting the same amplitude characteristic index as mentioned above from multiple normal scour events corresponding to the same flow zone under historical active sludge conditions in the same pipeline, and then applying robust statistical principles. At the same time, after the daily scour process ends and the flow rate drops back to the baseline low flow range, the unit extracts a fixed-length time window in the middle of the baseline low flow range and calculates the time-series average of the corrected turbidity increment sequence within this window as the low-flow residual turbidity index. When the amplitude characteristic index of the flushing flow range is lower than the lower limit of the flushing amplitude reference, and in the subsequent consecutive daily flushing cycles, the daily low flow residual turbidity index is lower than the preset low turbidity normal maintenance limit, the unit determines that the amplitude characteristic meets the sensitivity loss condition.

[0055] For synchronicity of changes, the scour sensitivity diagnostic unit uses a sliding time window of fixed length, sliding synchronously along the flow sequence and the corrected turbidity increment sequence within the scour flow range. For each window, the Pearson linear correlation coefficient is calculated for the two sets of sequences as a quantitative indicator of turbidity synchronicity. When the Pearson linear correlation coefficients of multiple consecutive adjacent sliding windows are all below a preset lower threshold for synchronicity, the change in synchronicity is determined to meet the sensitivity loss condition.

[0056] The scour sensitivity diagnosis unit generates a second state flag only when both the amplitude characteristics and the synchronicity of change meet their respective sensitivity loss conditions, and outputs the flag, along with the current reference flow rate and reference water level difference, to the scour effect evaluation unit.

[0057] This invention, through the aforementioned process, transforms the decline in the scouring capacity of the bottom sediment layer into two independently measurable and mutually corroborating signals: one is that the absolute increase in resuspended sediment in the pipe during high flow periods is significantly lower than the historical active sediment level; the other is that the turbidity response to flow changes exhibits structural weakening. When the sediment at the bottom of the pipe is still loose floating mud, the shear force of the water flow is sufficient to massively uplift surface sediment, correcting the turbidity increase to amplify synchronously with the increase in flow. Furthermore, continuous diurnal scouring gradually reduces the amount of sediment at the bottom of the pipe, resulting in a daily decrease in residual turbidity at low flow rates. As the bottom sediment undergoes drainage consolidation and evolves into dead sediment, the bed surface's scouring resistance increases, and the same flow increase is no longer sufficient to erode a sufficient amount of sediment. The peak value of the turbidity increase continuously shrinks, thus breaking the synchronous linkage between turbidity and flow rate. This scheme does not rely on a fixed turbidity threshold, but instead uses the scouring behavior of the pipeline itself in the active mud stage to construct a dynamic reference standard. This not only adapts to changes in the upstream sandy background, but also captures early signals of loss of scouring sensitivity at the critical stage of the transition from floating mud to compaction, providing a reliable triggering condition for subsequent final verification through artificial sand flushing.

[0058] Understandably, the amplitude characteristic index of the corrected turbidity increment sequence can be replaced by the root mean square value or the cumulative integral value during the flushing period, in addition to using the smoothed high quantile, as long as the index can comprehensively characterize the intensity of turbidity release exceeding the background during the flushing period. Besides the Pearson linear correlation coefficient, the quantitative index of change synchronicity can also be the peak lag consistency measure of the cross-correlation function or the wavelet coherence coefficient, to measure the ability of turbidity to follow flow disturbances at different time scales.

[0059] Step S4: In response to the second state flag, identify the sand flushing operation event based on the gate opening data, and obtain upstream and downstream water level data and flow data before, during, and after the sand flushing operation; determine the water level change component caused by the independent downstream regulation based on the change of the downstream water level data, correct the upstream and downstream water level difference after the sand flushing operation based on the component, obtain the corrected water level difference, and determine the difference between the corrected water level difference and the reference water level difference as the recovery offset; Specifically, step S4 includes: Step S41: Based on the gate opening data and flow data, the gate action that meets the following conditions is identified as the sand flushing operation event: the flow starts from the initial level located in the vicinity of the reference flow, rises to a level exceeding the sand flushing flow trigger level within a preset short period of time, and remains above the trigger level for a duration greater than the preset minimum duration. Step S42: Obtain upstream and downstream water level data, downstream water level data, and flow rate data during the first time period before the sand flushing operation event and the second time period after the event. Step S43: Based on the changes in the downstream water level data before and after the flushing operation, determine the water level change component caused by the independent scheduling behavior of the downstream, and correct the upstream and downstream water level difference after the flushing operation based on the component to obtain the corrected water level difference. Step S44: After the sand flushing operation is completed and the flow rate recovers to the range near the reference flow rate, the difference between the corrected water level difference and the reference water level difference at this time is determined as the recovery offset.

[0060] In this embodiment, step S4 is performed by the flushing effect evaluation unit. The flushing effect evaluation unit is triggered upon receiving the second status flag output by the flushing sensitivity diagnosis unit and retrieves historical gate opening data and real-time and historical flow data from the data acquisition unit. The flushing effect evaluation unit searches for gate opening signals and corresponding flow responses using a time-series scan method, identifying gate actions that simultaneously meet the following conditions as valid flushing operation events: the initial flow value is within the vicinity of the reference flow; the flow rises to exceed a preset flushing flow trigger level within a preset short period; and the duration of the flow exceeding the trigger level is greater than a preset minimum maintenance duration. The vicinity of the reference flow is composed of preset tolerance widths above and below the reference flow value. The flushing flow trigger level is pre-calibrated based on the hydraulic characteristics of the inverted siphon and historical flushing operation records to ensure that the identified operation event corresponds to a high-intensity flushing operation actively implemented by maintenance personnel, rather than routine flow fluctuations.

[0061] After identifying the sand flushing operation event, the sand flushing effect assessment unit obtains the upstream and downstream water level sequences, downstream water level sequence, and flow rate sequence for the first time period before sand flushing and the second time period after sand flushing, respectively, based on the start and end times of the event. The start point of the first time period is earlier than the start time of the sand flushing operation, and the time period is long enough to ensure that the data within it is not affected by the pre-disturbance of the sand flushing operation. The start point of the second time period is later than the end time of the sand flushing operation, and the time period is long enough to allow the pipeline water flow to return to a quasi-steady state.

[0062] After acquiring the aforementioned sequence, the sediment flushing effect assessment unit performs decoupling and compensation for the downstream independent scheduling component. The unit analyzes the self-range of the downstream water level sequence before and after the sediment flushing operation. When a long-range unidirectional displacement of the downstream water level, different from the flow rate trend, is detected during or immediately after the sediment flushing operation, it is determined that the displacement originates from downstream independent scheduling. The unit then calls a pre-established steady-state relationship model between downstream water level and flow rate based on historical operational data. Using the measured flow rate during the sediment flushing operation as input, it obtains the expected downstream water level under conditions without downstream independent scheduling. The difference between the measured downstream water level and this expected value is determined as the water level change component caused by downstream independent scheduling. Subsequently, the unit subtracts this water level change component from the measured upstream and downstream water level difference after the sediment flushing operation to obtain the corrected water level difference. This correction operation can be performed hourly within the second time period after sediment flushing. The unit ultimately takes the stable center value of the corrected water level difference within this time period as the representative value for subsequent calculations.

[0063] The sediment flushing effect assessment unit continuously monitors the flow rate decline process. When it is confirmed that the flow rate has recovered to the vicinity of the reference flow rate and the upstream and downstream water level difference has stabilized after correction, the unit reads the representative value of the corrected water level difference at that moment and subtracts it from the reference water level difference transmitted from the benchmark analysis unit. The resulting difference is determined as the recovery offset of this sediment flushing operation.

[0064] This invention identifies and deducts the interference component of downstream independent scheduling behavior on the measured water level difference, ensuring that the recovered offset only represents the true decrease in flow resistance caused by the flushing away of sediment within the pipe. In actual operation of inverted siphons, the downstream end is often connected to a sluice gate or controlled by the tailrace channel water level. During sediment flushing operations, downstream may experience simultaneous artificial regulation or natural water level fluctuations. If these are not distinguished, the measured decrease in upstream and downstream water level difference may be influenced by non-sediment flushing factors, leading to an overestimation of the flushing effect. This scheme, based on the detection of abnormal trends in the downstream water level sequence and combined with a historical flow-downstream water level relationship model, quantitatively separates the downstream water level offset caused by independent scheduling. This ensures that the net value calculation of flow resistance changes before and after sediment flushing is not contaminated by changes in downstream boundary conditions, establishing a reliable data foundation for subsequent comparisons between the current recovered offset and historical recovered offsets under the same benchmark.

[0065] It is understandable that the determination of the water level change component caused by the independent downstream scheduling behavior can be achieved by using the residual analysis method based on the steady-state relationship model, or by correlating the rate of change of downstream water level during the flushing operation period with the time scale of the downstream gate opening and closing records, or by directly fitting the natural drift baseline after separating the external disturbance segment by inputting the downstream water level sequence into the change point detection algorithm. This will not be elaborated further.

[0066] Furthermore, during sand flushing operations, the pipeline is typically under high water level and high flow velocity conditions, making CCTV inspection difficult. This embodiment also employs sonar inspection to supplement the underwater detection of the pipe. Specifically, before and after the sand flushing operation, sonar probes are placed in the water inside the pipe. Utilizing the propagation and reflection characteristics of sound waves in water, acoustic scanning imaging is performed on the cross-sectional morphology of the underwater pipe section, the thickness of sediment at the bottom of the pipe, and major defects in the pipe wall. By comparing the changes in sediment thickness at the bottom of the pipe before and after the sand flushing operation, the aforementioned assessment results of the recovered offset are supplemented and verified. This also compensates for the inability of CCTV inspection to operate under high water level and full-flow conditions. Sonar inspection is unaffected by water turbidity and can operate in all weather conditions. However, its output is acoustic images, and its ability to identify defects such as minute cracks and corrosion in the pipe wall is limited. These defects need to be combined with other inspection results and interpreted by professionals, which will not be elaborated further here.

[0067] Step S5: Compare the recovery offset corresponding to the current sand flushing operation with the recovery offset of the historical sand flushing operation recorded before the second state mark was generated. When the comparison result meets the preset degradation condition, generate an early warning signal.

[0068] Specifically, step S5 includes: Step S51: Extract the most recent historical sand flushing operation event from the history record before the second state marker was generated; Step S52: Obtain the historical reference water level difference, historical sludge flushing intensity, and historical recovery offset corresponding to the historical sludge flushing operation event; Step S53: When the difference between the reference water level difference of the current sand flushing operation and the historical reference water level difference is less than the first allowable deviation, and the difference between the current sand flushing intensity and the historical sand flushing intensity is less than the second allowable deviation, compare the recovery offset with the historical recovery offset. Step S54: When the recovery offset is less than the product of the historical recovery offset and the preset attenuation coefficient, and the difference between the historical recovery offset and the recovery offset is greater than the preset degradation tolerance value, it is determined that the preset degradation condition is met, and an early warning signal is generated.

[0069] In this embodiment, steps S51 to S54 are executed by the early warning decision unit. After the sand flushing effect evaluation unit outputs the recovery offset, reference water level difference, and sand flushing intensity of this sand flushing operation, the early warning decision unit retrieves the most recent valid historical sand flushing operation event record from the historical event record database, which is the one that occurred before the time point when the second state marker was generated.

[0070] Specifically, the historical sediment flushing operation event record includes at least three parameters: historical reference water level difference, historical sediment flushing intensity, and historical recovery offset. The historical reference water level difference is the reference water level difference transmitted by the benchmark analysis unit before the historical sediment flushing operation is triggered. The historical sediment flushing intensity is the scouring energy characterization value calculated and recorded by the sediment flushing effect evaluation unit based on the flow data during the operation. The historical recovery offset is the difference between the corrected water level difference and the corresponding reference water level difference calculated and recorded by the sediment flushing effect evaluation unit after the operation.

[0071] After retrieving the aforementioned historical parameters from the historical event record database, the early warning decision unit proceeds to the comparability verification step. The early warning decision unit first calculates the absolute difference between the reference water level difference corresponding to the current sand flushing operation and the historical reference water level difference. If this absolute difference is less than a pre-set first allowable deviation, the early warning decision unit determines that the current sand flushing operation is the same as or similar to the pipeline base resistance state before the start of a historical sand flushing operation, and is therefore comparable. Otherwise, the early warning decision unit terminates the comparison and marks the reason as a discrepancy in the reference water level difference.

[0072] Furthermore, the early warning decision unit calculates the absolute difference between the current sand flushing intensity and the historical sand flushing intensity. When the absolute difference is less than the pre-set second allowable deviation, the early warning decision unit determines that the scouring energy applied by the two sand flushing operations is of the same or similar magnitude and is comparable; otherwise, the comparison is terminated and the reason is marked.

[0073] Specifically, the first allowable deviation is determined based on the interannual normal fluctuation range of the effective water level difference sequence in the same historical period, and the second allowable deviation is determined based on the statistical dispersion of the flushing intensity of the same type of flushing operation under the condition of active mud in previous years. Both are pre-set in the configuration parameters of the early warning decision unit.

[0074] When both of the above comparability conditions are met simultaneously, the early warning decision unit performs a degradation comparison of the flushing recovery offset. The early warning decision unit multiplies the historical recovery offset by a preset attenuation coefficient to obtain an attenuation reference limit; simultaneously, it calculates the difference between the historical recovery offset and the current recovery offset. The early warning decision unit compares the current recovery offset with the attenuation reference limit and the difference with a preset degradation tolerance value. When the current recovery offset is less than the attenuation reference limit, and the difference between the historical recovery offset and the current recovery offset is greater than the degradation tolerance value, the early warning decision unit determines that the flushing recovery force at the bottom of the pipe has experienced a generational decline, and both of the above conditions satisfy the preset degradation condition. When the preset degradation condition is met, the early warning decision unit generates an early warning signal, indicating that an irreversible hardened bed has formed at the bottom of the inverted siphon, requiring shutdown for mechanical dredging. If either condition is not met, no early warning signal is generated, and the comparison result between the current recovery offset and the historical recovery offset is stored in the event log database for subsequent tracking.

[0075] This invention ensures the fairness of the comparison benchmark by setting dual comparability thresholds of reference water level difference and flushing intensity. Then, through a combined judgment of attenuation coefficient and degradation tolerance, it captures the critical state where the flushing recovery capacity has truly undergone a qualitative change and cannot be reversed by hydraulic dredging. In the active mud state, the net reduction in flow resistance obtained with the same flushing intensity may fluctuate slightly due to variations in mud thickness, but it remains within the normal range and will not fall below the attenuation limit set according to the statistical law of recovery capacity in the active mud stage. When the bottom of the pipe gradually hardens into dead mud, the flushing operation can only flush up the surface thin mud; the deep solidified mud bed is hardly eroded, and the recovery offset shrinks sharply, with its reduction relative to the historical active mud level exceeding the allowable range of random fluctuations. This judgment logic ensures that the warning signal strictly points to a fundamental change in the rheological properties of the bottom mud, thereby avoiding misjudgments caused by normal changes in mud volume or differences in operational details.

[0076] It is understandable that, in addition to using the difference between the peak flushing flow rate and the initial flow rate, the intensity of the flushing can also be characterized by the cumulative integral value of the flow rate exceeding the trigger level during the flushing duration, or by the equivalent flushing work calculated by multiplying the gate opening increment by the high flow duration, as long as the same calculation method is used in the two comparisons.

[0077] Furthermore, this embodiment also incorporates ground-based ground-penetrating radar (GPR) detection to assess the condition of the soil outside the inverted siphon. Specifically, survey lines are laid out on the ground above the inverted siphon along the pipeline axis. The GPR emits high-frequency electromagnetic waves into the soil outside the pipe and receives reflected waves from different media interfaces (such as concrete pipe wall and soil, soil and cavities). By analyzing the travel time, amplitude, and phase characteristics of the reflected waves, the system detects whether there are abnormalities such as cavities, voids, or looseness in the soil outside the pipe. This allows for the timely detection of hidden defects caused by pipe leakage, which could lead to soil loss around the pipe and threaten the safety of the pipeline and its superstructure.

[0078] It should be noted that conventional detection methods such as in-pipe CCTV detection, sonar detection, and ground-based ground-penetrating radar detection can be used to conduct on-site verification and defect location for suspicious pipe sections indicated by the early warning signal generated by this invention. This cross-verifies the early warning results with the detection results, thereby improving the accuracy of defect identification and location. These methods can also be used as periodic inspection methods, complementing and reinforcing the intelligent early warning method described in this invention. It is understood that the specific implementation methods, equipment selection, parameter settings, and data processing methods of the above-mentioned detection methods can all be implemented with reference to conventional techniques in the field. Their combination with this invention does not change the technical solution defined in the foregoing steps of this invention. Those skilled in the art can select and combine the above-mentioned detection methods according to actual working conditions.

[0079] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for intelligent early warning of inverted siphon pipe defects based on multi-source data acquisition, characterized in that, include: Step S1: Obtain upstream and downstream water level data, flow rate data, outlet turbidity data, gate opening data, and reference turbidity data used to correct the outlet turbidity data of the inverted siphon. Step S2: Determine the reference time period based on the flow data; select a time interval within the reference time period where the downstream water level is in the normal fluctuation range; and determine the effective water level difference sequence based on the upstream and downstream water level difference within the selected time interval; perform trend detection on the effective water level difference sequence; when a unidirectional increasing trend exceeding the normal fluctuation range of the same period in history is detected, generate a first anomaly marker; and determine the water level difference and flow rate corresponding to triggering the unidirectional increasing trend as the reference water level difference and reference flow rate, respectively. Step S3: In response to the first anomaly marker, the subsequent flow process is divided into multiple flow intervals based on the reference flow rate, and the outlet turbidity data is corrected based on the reference turbidity data to obtain the flushing turbidity change characteristics. When the turbidity change characteristic of the scouring meets the preset sensitivity loss condition, a second state marker is generated; Step S4: In response to the second state flag, identify the sand flushing operation event based on the gate opening data, and obtain upstream and downstream water level data and flow data before, during, and after the sand flushing operation; determine the water level change component caused by the independent downstream regulation based on the change of the downstream water level data, correct the upstream and downstream water level difference after the sand flushing operation based on the component, obtain the corrected water level difference, and determine the difference between the corrected water level difference and the reference water level difference as the recovery offset; Step S5: Compare the recovery offset corresponding to the current sand flushing operation with the recovery offset of the historical sand flushing operation recorded before the second state mark was generated. When the comparison result meets the preset degradation condition, generate an early warning signal.

2. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 1, characterized in that, Step S2 includes: Step S21: Based on the daily traffic change rate for each preset time period, determine the time period in which the traffic change rate is consistently lower than a preset threshold as the baseline time period; Step S22: Divide the reference time period into multiple sub-time periods and obtain the upstream and downstream water level difference and the downstream water level change within each sub-time period. Step S23: For each sub-period, if the change in its downstream water level is within the normal fluctuation range of the downstream water level, then the difference between the upstream and downstream water levels of the sub-period is taken as an effective water level difference sample. Step S24: Based on the effective water level difference samples within the same reference time period, calculate the reference representative water level difference, and form the effective water level difference sequence from the reference representative water level differences over multiple consecutive days; Step S25: Based on the fluctuation characteristics of the historical effective water level difference sequence, determine the normal fluctuation boundary and detect the deviation of the current effective water level difference sequence from the normal fluctuation boundary. Step S26: When the deviation state is manifested as exceeding the normal fluctuation boundary at multiple consecutive time points and the deviation amount is monotonically increasing, it is determined that a unidirectional increasing trend has occurred, and the benchmark representative water level difference and benchmark representative flow rate on the day that the unidirectional increasing trend is triggered are respectively determined as the reference water level difference and reference flow rate.

3. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 2, characterized in that, The normal fluctuation boundary mentioned in step S25 is determined based on the statistical distribution characteristics of the effective water level difference sequence in the same historical season, which defines the upper and lower boundary values.

4. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 3, characterized in that, Step S3 includes: Step S31: Divide the daily flow process into a baseline low flow range and at least one flushing flow range higher than the reference flow rate according to the reference flow rate. Step S32: Within the flushing flow range, based on the outlet turbidity data and the reference turbidity data, the corrected turbidity increment is obtained time-by-time, and a corrected turbidity increment sequence is formed. Step S33: Evaluate the amplitude characteristics of the corrected turbidity increment sequence and the synchronicity of the changes between the corrected turbidity increment sequence and the corresponding flow sequence. When both the amplitude characteristics and the synchronicity of the changes meet the preset sensitivity loss criteria, generate the second state flag.

5. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 4, characterized in that, The preset sensitivity loss criteria include: Within the scouring flow range, the amplitude index of the corrected turbidity increment sequence is lower than the lower limit of amplitude determined based on historical normal scouring events in the same range. After experiencing a predetermined number of daytime flushing periods, the amplitude index of the corrected turbidity increment sequence is lower than the preset low turbidity state threshold within the baseline low flow range.

6. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 5, characterized in that, In step S33, the change synchronicity is to obtain a synchronicity quantification index of the corrected turbidity increment sequence and the flow sequence within a sliding time window. When the synchronicity quantification index is lower than a preset synchronicity threshold within multiple consecutive sliding time windows, it is determined that the change synchronicity meets the preset sensitivity loss standard.

7. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 6, characterized in that, Step S4 includes: Step S41: Based on the gate opening data and flow data, the gate action that meets the following conditions is identified as the sand flushing operation event: the flow starts from the initial level located in the vicinity of the reference flow, rises to a level exceeding the sand flushing flow trigger level within a preset short period of time, and remains above the trigger level for a duration greater than the preset minimum duration. Step S42: Obtain upstream and downstream water level data, downstream water level data, and flow rate data during the first time period before the sand flushing operation event and the second time period after the event. Step S43: Based on the changes in the downstream water level data before and after the flushing operation, determine the water level change component caused by the independent scheduling behavior of the downstream, and correct the upstream and downstream water level difference after the flushing operation based on the component to obtain the corrected water level difference. Step S44: After the sand flushing operation is completed and the flow rate recovers to the range near the reference flow rate, the difference between the corrected water level difference and the reference water level difference at this time is determined as the recovery offset.

8. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 7, characterized in that, Step S5 includes: Step S51: Extract the most recent historical sand flushing operation event from the history record before the second state marker was generated; Step S52: Obtain the historical reference water level difference, historical sludge flushing intensity, and historical recovery offset corresponding to the historical sludge flushing operation event; Step S53: When the difference between the reference water level difference of the current sand flushing operation and the historical reference water level difference is less than the first allowable deviation, and the difference between the current sand flushing intensity and the historical sand flushing intensity is less than the second allowable deviation, compare the recovery offset with the historical recovery offset. Step S54: When the recovery offset is less than the product of the historical recovery offset and the preset attenuation coefficient, and the difference between the historical recovery offset and the recovery offset is greater than the preset degradation tolerance value, it is determined that the preset degradation condition is met, and an early warning signal is generated.

9. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 8, characterized in that, The reference turbidity data is determined as follows: When upstream turbidity data is collected synchronously with the outlet turbidity data, the upstream turbidity data is used as the reference turbidity data; When upstream turbidity data is unavailable, the turbidity background value obtained by statistical analysis of outlet turbidity data from the same historical period without rainfall and with the same flow rate as the baseline low flow range will be used as the reference turbidity data.

10. The intelligent early warning method for inverted siphon pipe defects based on multi-source data acquisition according to claim 9, characterized in that, The normal fluctuation range of downstream water level in step S23 is the range of variation determined by a preset statistical method based on the downstream water level data of the same historical period and when the flow rate is in the same order of magnitude as the reference period.

Images