A method and system for predicting scouring trend of full-pile wharf

Abstract

The application discloses a full-bench high-pile wharf scouring trend prediction method and system, and belongs to the field of port engineering monitoring. The method first constructs an embedded ultrasonic array monitoring system, arranges probes according to regions to realize accurate control, and synchronously collects acoustic data and environmental parameters; then, the data are subjected to noise reduction, inversion and fusion processing, combined with shallow water acoustic modeling and a scouring mechanism identification algorithm to distinguish normal sedimentation fluctuations from abnormal scouring events; finally, the current deposition state is predicted through an improved DWMA model, the scouring trend is predicted by means of an improved LSTM physical hybrid model, and risk visualization is realized based on a hierarchical early warning mechanism. The system can adapt to the dense pile foundation environment of the full-bench high-pile wharf, effectively solves the problems of large blind area and high error existing in traditional monitoring, significantly improves the monitoring accuracy and early warning timeliness, and provides strong support for wharf safety operation and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of port engineering monitoring, in particular to a full-bay high-pile wharf scouring trend prediction method and system. BACKGROUND

[0002] In recent years, full-bay high-pile wharfs are widely used in port engineering construction due to their stable structure and strong adaptability. The safety operation of the wharf is closely related to the scouring and deposition state of the seabed. Scouring of the seabed can lead to insufficient depth of the pile foundation, causing structural instability, while deposition can affect the navigation capacity of the wharf and the stress balance of the structure. Therefore, continuous monitoring and trend prediction of the scouring and deposition state of the seabed around the pile foundation are crucial to ensure the long-term safe operation of the wharf.

[0003] Currently, there are two major difficulties in the scouring and deposition monitoring and prediction of full-bay high-pile wharfs. On the one hand, there are significant limitations in equipment deployment. Traditional monitoring relies on an array of rigid sensors, but the dense pile foundation of full-bay high-pile wharfs and the narrow underwater space make it difficult to arrange sensors reasonably, resulting in large blind areas. At the same time, the strong turbulence in the underwater environment of the wharf can cause sensor drift, poor data stability, and an inability to accurately capture dynamic evolution details such as instantaneous scouring caused by local vortices, making it difficult to meet the seamless monitoring needs of the entire cross-section.

[0004] On the other hand, existing prediction models have obvious defects. Current prediction methods are mostly based on historical statistical data or simplified hydrodynamic equations, without fully considering the riverbed-flow coupling effect and integrating real-time monitoring data, resulting in large fluctuations in prediction errors, usually exceeding 20%. In addition, the model parameter update cycle is long, and it cannot quickly respond to short-term extreme hydrological events such as tidal mutations, making it difficult to accurately predict short-term scouring and deposition trends and provide timely and effective decision support for wharf operation and maintenance.

[0005] To solve the above problems, related fields have tried to optimize sensor deployment or improve prediction models, but none of them have formed a systematic solution for the structural characteristics of full-bay high-pile wharfs. Sensor deployment still cannot break through the spatial limitations caused by dense pile foundations, and the monitoring accuracy and coverage are insufficient; the prediction model has not effectively integrated real-time data and physical mechanisms, and the short-term prediction accuracy and dynamic response capability need to be improved, making it difficult to meet the actual needs of accurate monitoring, reliable prediction, and timely warning of scouring and deposition state for wharf safety operation. SUMMARY

[0006] Therefore, the present application provides a full-bay high-pile wharf scouring trend prediction method, characterized by the following steps:

[0007] Step (1), build an embedded ultrasonic array monitoring system, arrange ultrasonic probes according to the scouring risk distribution zoning of the underwater structure of the wharf, and optimize the installation parameters and array topology of the probes;

[0008] Step (2), collecting multi-dimensional monitoring data, including acoustic data and environmental parameters, the acoustic data being acquired by the ultrasonic probe;

[0009] Step (3), performing full-process processing on the monitoring data to improve data quality through time delay window dynamic control, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion and data fusion optimization;

[0010] Step (4), using an improved dynamic weighted moving average (DWMA) model, combining time decay weight and Kalman filter residual correction term, predicting the current sedimentation state based on the full-process processed monitoring data;

[0011] Step (5), constructing an improved LSTM physical hybrid model, fusing hydrodynamic terms and data-driven terms, setting adaptive weights and sliding time windows for incremental training, predicting risk indexes based on the current sedimentation state to realize graded early warning, and connecting to a digital twin platform for visual presentation.

[0012] Further, in step (1), the ultrasonic probe is arranged at the front and rear capping tables of each bent pile and at the junction of adjacent bent piles, and the array topology is optimized through finite element simulation; the optimized array topology specifically includes optimizing the probe working frequency, layout spacing and pointing angle through water tank simulation and numerical analysis; the embedded ultrasonic array works in a multi-unit cooperative scanning mode.

[0013] Further, in step (2), the acoustic data includes echo time delay, amplitude decay rate and frequency shift, and the environmental parameters include hydrological data and meteorological correlation parameters; the hydrological data includes flow rate, flow direction, tidal level and sediment concentration, and the meteorological correlation parameters include wind speed and wave height.

[0014] Further, in step (3), the time delay window width calculation formula is where D(t) is the real-time water depth, c is the sound speed, is the standard deviation of environmental noise, and k is the noise constant; the echo path discrimination separates reflected waves using a cross-correlation algorithm formula: where is the time delay, x(t) is the original ultrasonic pulse signal transmitted, and y(t+τ) is the mixed echo signal received; the acoustic impedance inversion is achieved by calculating the reflection coefficient, and the calculation formula of the reflection coefficient is where , is the medium density, , is the acoustic impedance of the adjacent medium.

[0015] Further, the monitoring data is transmitted to the shore-based data processing module for waveform analysis, noise suppression and echo peak identification, and then combined with the historical monitoring sequence to construct the erosion evolution trend curve.

[0016] Further, in the step (4), the dynamic weighted moving average (DWMA) model calculation formula is: , wherein, is the local erosion depth at time t, is the model calibration coefficient, is the number of ultrasonic probes participating in the calculation, is the time decay weight, is the echo delay variation of the i-th probe, is the spatial adaptation coefficient, is the sound speed of the i-th probe after temperature and salinity compensation correction, is the initial erosion depth, is a random error term.

[0017] Further, in the step (4), the time decay weight calculation formula is: , wherein , t is the time, is the time of the i-th event or data point; by optimizing the beta coefficient to adapt to the spatial distribution of multiple probes, and combining the local erosion rate difference of multiple probe collaborative inversion, the sensor drift error is corrected by Kalman filter.

[0018] Further, in the step (5), the LSTM physical hybrid model calculation formula is: , wherein, is the hydrodynamic term, u is the flow velocity vector, H is the erosion depth, is the hydrodynamic term weight coefficient, is the feature mapping function of the LSTM model, is the sediment concentration, is the pile foundation vibration spectrum, is the dynamic noise suppression term weight, is the dynamic noise suppression term.

[0019] Further, in the step (5), the LSTM physical hybrid model adopts an incremental training mechanism, and based on the classification results of normal deposition fluctuations and abnormal erosion events, a hierarchical early warning threshold is drawn, and the seabed morphology change is presented in the form of three-dimensional visualization through the digital twin platform.

[0020] The application further provides a full-bay high-pile wharf scouring trend prediction system for realizing the full-bay high-pile wharf scouring trend prediction method.

[0021] The embedded ultrasonic array monitoring unit is used to construct an embedded ultrasonic array monitoring system, arrange ultrasonic probes according to wharf underwater structure scouring risk distribution zoning, and optimize probe installation parameters and array topology.

[0022] The multi-dimensional data acquisition unit is used to acquire multi-dimensional monitoring data, wherein the multi-dimensional monitoring data comprises acoustic data and environmental parameters, and the acoustic data is acquired by the ultrasonic probe.

[0023] The monitoring data processing unit is used to perform full-process processing on the monitoring data, and improve data quality through time delay window dynamic control, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion and data fusion optimization.

[0024] The siltation state prediction unit is used to predict the current time siltation state based on the full-process processed monitoring data by using an improved dynamic weighted moving average (DWMA) model in combination with a time attenuation weight and a Kalman filter residual correction term.

[0025] The scouring trend prediction and early warning unit is used to construct an improved LSTM physical hybrid model, fuse a hydrodynamic term and a data driven term, set adaptive weights and a sliding time window for incremental training, realize graded early warning based on the risk index predicted based on the current time siltation state, and access a digital twin platform to realize visual presentation.

[0026] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, the specific embodiments of the application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the specific embodiments of the application or the technical scheme in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0028] Figure 1 The flowchart of the full-bay high-pile wharf scouring trend prediction method.

[0029] Figure 2 This is a schematic diagram of the scour trend prediction system architecture for a full-span high-pile wharf. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0031] As attached Figure 1 and attached Figure 2 As shown, this invention discloses a method for predicting the scour trend of a full-span high-pile wharf, characterized by comprising the following steps:

[0032] Step (1): Construct an embedded ultrasonic array monitoring system, deploy ultrasonic probes according to the risk distribution of underwater structure scour at the wharf, and optimize probe installation parameters and array topology;

[0033] The ultrasonic probes are deployed on the front and rear supports of each rack, as well as at the junction of adjacent racks. The array topology is optimized through finite element simulation. The optimization of the array topology specifically includes optimizing the probe operating frequency, deployment spacing, and pointing angle through water tank simulation and numerical analysis. The embedded ultrasonic array operates in a multi-unit collaborative scanning mode.

[0034] Specifically, the zoning layout scheme is arranged as follows:

[0035] For the front abutment (facing the flow): 53 main probes are set up according to the frame unit, and 26 auxiliary probes are added at the junction of adjacent frames to focus on covering the gaps between pile groups (which are prone to forming eddy current zones). The probe spacing is 1.2m to form a "main-auxiliary" cross monitoring network.

[0036] For the rear abutment (backflow surface): due to the gradual change in sedimentation, one main probe is installed for every two frame units, for a total of 27 probes. No auxiliary probes are added at the junctions. The probe spacing is 2.4m.

[0037] For the corner of the dock (area where the water flow direction changes abruptly): 8 additional wide-angle probes (pointing angle 150°) are installed to cover the blind spots of traditional probes.

[0038] The optimization array topology is simulated by finite element simulation, specifically including: adopting ANSYS Fluent and ANSYS APDL joint simulation, first simulating the underwater flow field distribution (flow velocity, vortex intensity) of the wharf to determine 12 high-risk scour points; then establishing a three-dimensional model of the ultrasonic array, setting the sound velocity of water to 1500 m / s and the sound velocity of riverbed sediment to 1800 m / s, simulating the sound wave propagation path and reflected signal intensity;

[0039] Based on the optimization results, the probe spacing is adjusted from 1.5 m to 1.2 m (front pile cap), and the monitoring blind area coverage rate is reduced from the initial 6.8% to 2.9%; by increasing the wide-angle probe, the blind area at the corner of the wharf is completely eliminated, and finally the monitoring blind area coverage rate of the whole wharf is controlled at 2.1%, meeting the design requirement of ≤5%.

[0040] Step (2), collecting multi-dimensional monitoring data, including acoustic data and environmental parameters, the acoustic data being obtained by the ultrasonic probe.

[0041] Further, in step (2), the acoustic data includes echo time delay, amplitude attenuation rate and frequency offset, and the environmental parameters include hydrological data and meteorological correlation parameters; the hydrological data includes flow velocity, flow direction, tidal level and sediment concentration, and the meteorological correlation parameters include wind speed and wave height.

[0042] In order to realize effective collection of multi-dimensional monitoring data, each monitoring unit is configured with one data acquisition terminal (integrated with the array control module), and the acquisition frequency can be dynamically adjusted: 1 time / minute under normal working conditions, 1 time / 10 seconds in typhoon season (wind speed ≥10 m / s), to ensure the timeliness of data in extreme environment.

[0043] The data transmission link adopts a three-level transmission architecture of "edge node-land-based base station-cloud platform", specifically, the edge node acquisition terminal aggregates the probe data in the unit through RS485 bus, and completes preliminary noise reduction based on mean filtering;

[0044] The land-based base station is deployed with several 4G / 5G base stations, which receive edge node data through LoRa wireless communication and store them to the local server synchronously;

[0045] The cloud platform uses Aliyun server, and the land-based base station pushes complete data to the cloud once every preset time for remote access and data analysis by the cloud platform.

[0046] Step (3), processing the monitoring data in the whole process, and improving the data quality through time delay window dynamic control, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion and data fusion optimization;

[0047] Further, in step (3), the time delay window width calculation formula is wherein D(t) is the real-time water depth, c is the sound speed, is the standard deviation of the ambient noise, k is the noise constant; the standard deviation of the ambient noise Obtained by 10 consecutive no-signal measurements (shielding the probe emitting end).

[0048] The 1. wavelet threshold denoising adopts a Symlet5 base function to perform 5-layer wavelet decomposition on the echo signal, and the coefficients of each layer after decomposition are processed using a heuristic threshold (Heuristic SURE), and the threshold calculation is wherein N is the signal length; is the standard deviation of the noise);

[0049] The echo path discrimination separates the reflected waves using a cross-correlation algorithm formula wherein is the time delay, x(t) is the original ultrasonic pulse signal emitted by the probe (such as the sound wave emitted by the probe), and y(t+τ) is the received mixed echo signal (containing direct waves, pile foundation reflected waves, and riverbed reflected waves); the cross-correlation algorithm is used to calculate the cross-correlation coefficient of the original signal and the reference signal, i.e., the mixed echo signal, a cross-correlation coefficient range is set, direct waves, pile foundation reflected waves, and riverbed reflected waves are distinguished, and accurate separation of the three types of echoes is achieved.

[0050] The acoustic impedance inversion is achieved by calculating the reflection coefficient, and the calculation formula of the reflection coefficient is wherein , is the medium density, , is the acoustic impedance of the adjacent medium; specifically, the sound pressure distribution under different deposition thicknesses is simulated, the riverbed reflected wave sound pressure amplitude is extracted, and the deposition thickness is inverted in combination with the reflection coefficient formula.

[0051] Further, the monitoring data is classified into patterns using shallow water acoustic modeling and scour mechanism identification algorithms, and normal deposition fluctuations and abnormal scour events are distinguished, the monitoring data is first transmitted to a shore-based data processing module for waveform analysis, noise suppression, and echo peak identification, and then combined with historical monitoring sequences to construct a scour evolution trend curve.

[0052] The shallow water acoustic modeling specifically includes building a three-dimensional layered model for the shallow water environment, inputting real-time parameters to correct the sound speed and attenuation coefficient, setting the pile foundation (hard boundary), the pile cap (gradual boundary), and the riverbed (layered boundary) and correcting the multi-path error; the finite difference time domain (FDTD) method is used to solve the acoustic wave equation, and the simulation results show that under normal deposition conditions, only a single riverbed interface echo peak appears, and under abnormal scour events, there are echo missing sections or multi-interface chaotic echoes, which provide characteristic basis for subsequent mechanism identification.

[0053] The flushing mechanism recognition algorithm specifically includes distinguishing events in "feature extraction + secondary classification": extracting three types of features of interface stability, sound field distribution, and dynamic change, the secondary classification includes primary judgment by 1000+ sample random forest (accuracy > 95%), and secondary false anomaly elimination according to historical threshold, and output of an event report containing the basis.

[0054] In step (4), an improved dynamic weighted moving average (DWMA) model is used to predict the current sedimentation state based on the monitoring data processed in the whole process by combining a time decay weight and a Kalman filter residual correction term.

[0055] Further, in step (4), the calculation formula of the dynamic weighted moving average (DWMA) model is: , wherein is the local flushing depth at time t, is a model calibration coefficient, is the number of ultrasonic probes participating in the calculation, is a time decay weight, is the echo delay change of the i-th probe, is a spatial adaptation coefficient, is the sound speed of the i-th probe after temperature and salinity compensation correction, is the initial flushing depth (reference depth), is a random error term.

[0056] Further, in step (4), the calculation formula of the time decay weight is: , wherein t is the time, is the time of the i-th event or data point, and is the historical time; the space adaptation coefficient is adapted by optimizing the space adaptation coefficient, and the local flushing rate difference is combined with the multi-probe collaborative inversion to correct the sensor drift error by means of Kalman filter.

[0057] In step (5), an improved LSTM physical hybrid model is constructed to fuse the hydrodynamic term and the data-driven term, set adaptive weight and sliding time window for incremental training, predict the risk index based on the current sedimentation state to realize graded warning, and access the digital twin platform for visual presentation.

[0058] Further, in step (5), the calculation formula of the LSTM physical hybrid model is: , wherein is the hydrodynamic term, u is the flow velocity vector (real-time monitoring input), and H is the flushing depth, is the hydrodynamic term weight coefficient, is the feature mapping function of the LSTM model, for the sediment concentration, for the pile foundation vibration spectrum, for the dynamic noise suppression term weight, for the dynamic noise suppression term, the noise suppression term is dynamically adjusted based on Kalman filter residual, and the interference of environmental noise on prediction is suppressed.

[0059] The LSTM network, the input layer is the sediment concentration, the pile foundation vibration spectrum (set frequency range, extract a certain number of characteristic values), the past sediment thickness sequence for a certain length of time, the hidden layer is set to a certain number of layers (each layer is set to a certain number of neurons), and the output layer is the predicted value of the scour rate, and the weight is adjusted through an adaptive algorithm

[0060] Further, in the step (5), the LSTM physical hybrid model adopts an incremental training mechanism, and a hierarchical warning threshold is determined based on the classification results of normal deposition fluctuations and abnormal scour events, and the seabed morphology changes are presented in a three-dimensional visual form through a digital twin platform.

[0061] The risk index is calculated in real time, and the formula is The calculation is as follows: wherein According to the safety standard of the wharf pile foundation (the scour depth exceeding the value is easy to lead to the instability of the pile foundation), k is a trend sensitive coefficient, The scour rate is set to a threshold range corresponding to different risk levels, respectively corresponding to a safe state, an attention state and a warning state; when the risk index reaches the warning threshold, the corresponding warning is triggered.

[0062] The digital twin platform accesses the monitoring data and the prediction results in real time, and the scour risk areas are marked with different colors in the three-dimensional interface, corresponding to different risk levels, and the real-time data and historical trend curves can be viewed by clicking any monitoring point, the real-time data including the deposition thickness, the flow rate, the risk index and the like; when the warning is triggered, the platform automatically sends a message and a push notification to the mobile phone APP of the operation and maintenance personnel, including the warning position, the risk level, the recommended measures such as riprap protection and flow rate monitoring encryption, and the control response time is within a set range.

[0063] Embodiment two

[0064] The application further provides a full-pile high-pile wharf scour trend prediction system for realizing the full-pile high-pile wharf scour trend prediction method.

[0065] The embedded ultrasonic array monitoring unit is used to construct an embedded ultrasonic array monitoring system, arrange ultrasonic probes according to the scour risk distribution zoning of the underwater structure of the wharf, and optimize the probe installation parameters and array topology.

[0066] The multi-dimensional data acquisition unit is configured to acquire multi-dimensional monitoring data, wherein the multi-dimensional monitoring data comprises acoustic data and environmental parameters, and the acoustic data is acquired by the ultrasonic probe.

[0067] The monitoring data processing unit is configured to perform full-process processing on the monitoring data, and improve data quality through time delay window dynamic control, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion and data fusion optimization.

[0068] The deposition state prediction unit is configured to use an improved dynamic weighted moving average (DWMA) model, combine a time attenuation weight and a Kalman filter residual correction term, and predict the deposition state at the current time based on the monitoring data processed by the full-process processing.

[0069] The scouring trend prediction and early warning unit is configured to build an improved LSTM physical hybrid model, fuse a hydrodynamic term and a data-driven term, set an adaptive weight and a sliding time window for incremental training, realize graded early warning based on a risk index predicted by the deposition state at the current time, and access a digital twin platform to realize visual presentation.

[0070] It can be understood that the system and unit provided by the embodiment can also be used to implement each step in the method provided by other embodiments of the application.

[0071] The application also provides a computer device. The computer device is in the form of a general-purpose computing device. The components of the computer device can include but are not limited to one or more processors or processing units, system memory, and a bus connecting different system components.

[0072] The computer device typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device, including volatile and non-volatile media, removable and non-removable media.

[0073] The system memory can include a computer system readable medium in the form of volatile memory, and the memory can include at least one program product having a set (for example, at least one) of program modules configured to perform the functions of the embodiments of the application.

[0074] The processing unit performs various functional applications and data processing by running the programs stored in the system memory, for example, implements the method provided by other embodiments of the application.

[0075] The application also provides a storage medium containing computer executable instructions, and a computer program is stored on the storage medium, which is executed by a processor to implement the method provided by other embodiments of the application.

[0076] Note that the above merely describes preferred embodiments of the application and the principles of the technology applied. Those skilled in the art will understand that the application is not limited to the specific embodiments described herein, and that various obvious changes, modifications and substitutions can be made to the application without departing from the scope of the application. Therefore, although the application has been described in detail by the above embodiments, the application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the application, and the scope of the application is determined by the appended claims.

Claims

1. A method for predicting scouring trend of a full-pile wharf, characterized in that, Includes the following steps: Step (1): Construct an embedded ultrasonic array monitoring system, deploy ultrasonic probes according to the risk distribution of underwater structure scour at the wharf, and optimize probe installation parameters and array topology; Step (2): Collect multi-dimensional monitoring data, including acoustic data and environmental parameters, wherein the acoustic data is acquired by the ultrasonic probe; Step (3) involves processing the monitoring data throughout the entire process, improving data quality through dynamic control of time delay window, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion, and data fusion optimization. Step (4): An improved dynamic weighted moving average (DWMA) model is adopted, which combines time decay weights and Kalman filter residual correction terms to predict the current siltation status based on the monitoring data after the entire process is processed. Step (5): Construct an improved LSTM physical hybrid model, integrate hydrodynamic terms and data-driven terms, set adaptive weights and sliding time windows for incremental training, predict the risk index based on the current siltation state to achieve graded early warning, and connect to the digital twin platform for visualization. 2.The method according to claim 1, wherein, In step (1), the ultrasonic probes are placed on the front and rear supports of each rack and at the junction of adjacent racks. The array topology is optimized by finite element simulation. The optimization of the array topology specifically includes optimizing the probe working frequency, placement spacing and pointing angle by water tank simulation and numerical analysis. The embedded ultrasonic array works in a multi-unit collaborative scanning mode. 3.The method according to claim 1, wherein, In step (2), the acoustic data includes echo delay, amplitude attenuation rate and frequency offset, and the environmental parameters include hydrological data and meteorological parameters; the hydrological data includes flow velocity, flow direction, tide level and sediment content, and the meteorological parameters include wind speed and wave height.

4. The method for predicting the scour trend of a full-span high-pile wharf according to claim 1, characterized in that, In step (3), the formula for calculating the time delay window width is: Where D(t) is the real-time water depth and c is the speed of sound. Let k be the standard deviation of the environmental noise, and k be the noise constant; the cross-correlation algorithm formula for the echo path discrimination and separation of reflected waves is as follows: ,in Let x(t) be the original transmitted ultrasonic pulse signal and y(t+τ) be the received mixed echo signal, with a time delay of τ. The acoustic impedance inversion is achieved by calculating the reflection coefficient, and the formula for calculating the reflection coefficient is as follows: ,in , For the density of the medium, , The acoustic impedance of the adjacent medium.

5. The method for predicting the scour trend of a full-span high-pile wharf according to claim 4, characterized in that, The monitoring data is classified into patterns using shallow water acoustic modeling and scour mechanism identification algorithms to distinguish between normal sedimentary fluctuations and abnormal scour events. The monitoring data is first transmitted to the shore-based data processing module for waveform analysis, noise suppression and echo peak identification, and then combined with historical monitoring sequences to construct a scour evolution trend curve.

6. The method for predicting the scour trend of a full-span high-pile wharf according to claim 1, characterized in that, In step (4), the calculation formula for the Dynamic Weighted Moving Average (DWMA) model is as follows: ,in, Let be the local scour depth at time t. For model calibration coefficients, This represents the number of ultrasound probes used in the calculation. For time decay weight, Let be the change in echo delay of the i-th probe. For spatial adaptation coefficient, Let be the velocity of sound from the i-th probe after temperature and salinity compensation correction. This is the initial scour depth. This is the random error term.

7. The method for predicting the scour trend of a full-span high-pile wharf according to claim 6, characterized in that, In step (4), the formula for calculating the time decay weight is: ,in t represents the time at that time. Let be the time of the i-th event or data point; the spatial distribution of multiple probes is optimized and adapted by the β coefficient, and the local scour rate difference is combined with the multi-probe collaborative inversion, and the sensor drift error is corrected by Kalman filtering.

8. The method for predicting the scour trend of a full-span high-pile wharf according to claim 1, characterized in that, In step (5), the calculation formula for the LSTM physical hybrid model is as follows: ,in, Here, u is the hydrodynamic term, H is the velocity vector, and H is the scour depth. This is the weighting coefficient for the hydrodynamic term. This is the feature mapping function of the LSTM model. For sand content, This is the vibration spectrum of the pile foundation. For the weight of the dynamic noise suppression term, This is a dynamic noise suppression term.

9. The method for predicting the scour trend of a full-span high-pile wharf according to claim 8, characterized in that, In step (5), the LSTM physical hybrid model adopts an incremental training mechanism, defines graded early warning thresholds based on the classification results of normal sedimentary fluctuations and abnormal scour events, and presents seabed morphological changes in a three-dimensional visualization form through a digital twin platform.

10. A scour trend prediction system for a full-span high-pile wharf, used to implement the scour trend prediction method for a full-span high-pile wharf as described in any one of claims 1-9, characterized in that, include: Embedded ultrasonic array monitoring unit, multi-dimensional data acquisition unit, monitoring data processing unit, siltation state prediction unit, and scour trend prediction and early warning unit; The embedded ultrasonic array monitoring unit is used to construct an embedded ultrasonic array monitoring system, deploy ultrasonic probes according to the risk distribution of underwater structure scour at the wharf, and optimize probe installation parameters and array topology. The multi-dimensional data acquisition unit is used to acquire multi-dimensional monitoring data, which includes acoustic data and environmental parameters. The acoustic data is acquired by the ultrasonic probe. The monitoring data processing unit is used to process the monitoring data throughout the entire process, and improve the data quality through dynamic control of time delay window, wavelet threshold denoising, echo path discrimination, acoustic impedance inversion and data fusion optimization. The siltation state prediction unit is used to predict the siltation state at the current moment based on the monitoring data after the entire process, by adopting an improved dynamic weighted moving average (DWMA) model, combined with time decay weights and Kalman filter residual correction terms. The scour trend prediction and early warning unit is used to construct an improved LSTM physical hybrid model, which integrates hydrodynamic terms and data-driven terms, sets adaptive weights and sliding time windows for incremental training, predicts risk index based on the current siltation state to achieve graded early warning, and connects to a digital twin platform for visualization.

Images