A method and system for monitoring foundation settlement of a deep water breakwater

Abstract

The application provides a deep-water breakwater foundation settlement monitoring method and system, and relates to the technical field of ocean engineering. The method comprises the following steps: based on a breakwater design drawing, a plurality of hydrophone nodes are installed in at least two monitoring areas of an underwater foundation and connected to a monitoring host, and an underwater acoustic monitoring network is generated; the network nodes are controlled to synchronously collect underwater sound pressure fluctuation signals, and multi-channel sound pressure signal data is obtained; based on the data, two-by-two cross-correlation calculation is performed on different node channel signals, and a noise interference spectrum is obtained; time delay values and energy values of stable cross-correlation peaks in the spectrum are identified and measured, and cross-correlation function characteristic parameters are generated; finally, based on the parameters, a pre-trained settlement inversion model is used for calculation, and settlement distribution information of the breakwater foundation is generated, so that non-contact, wide-area, continuous and quantitative monitoring of the deep-water breakwater foundation settlement is realized.

Description

Technical Field

[0001] This invention relates to the field of marine engineering technology, and in particular to a method and system for monitoring the settlement of deep-water breakwater foundations. Background Technology

[0002] In deep-water breakwater engineering, foundation settlement is a key factor affecting the long-term safety and stability of the structure. Traditional settlement monitoring methods mainly rely on divers or underwater robots equipped with sensors for fixed-point measurements, or on local monitoring methods based on sonar, fiber optics, and other technologies. These methods have significant limitations in deep-water and complex hydrological environments: manual monitoring is costly, risky, and difficult to achieve long-term continuous observation; while existing instruments have limited monitoring range, complex deployment, are sensitive to environmental noise, and are unable to capture the subtle and distributed settlement evolution process of the foundation under continuous dynamic loads such as waves and water flow.

[0003] Especially in deep-water breakwaters, the underwater foundation structure is concealed, covers a large area, and is subject to complex loads. There is an urgent need for a foundation settlement monitoring technology that can achieve wide-area, continuous, and highly sensitive monitoring without interfering with the structure itself, so as to obtain settlement distribution information in real time and provide a reliable basis for engineering safety early warning and maintenance decisions. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, the first aspect of this invention proposes a method for monitoring the settlement of deep-water breakwater foundations, comprising:

[0005] S1: Based on the structural design drawing of the breakwater, multiple hydrophone nodes are fixedly installed in at least two monitoring areas of the underwater foundation structure of the breakwater, and all hydrophone nodes are connected to the monitoring host to generate an underwater acoustic monitoring network.

[0006] S2: Control the hydrophone nodes in the underwater acoustic monitoring network to synchronously record the sound pressure fluctuation signals from underwater within the set acquisition period, and obtain multi-channel sound pressure signal data;

[0007] S3: Based on multi-channel sound pressure signal data, pairwise cross-correlation calculations are performed on signals from different hydrophone node channels in the underwater acoustic monitoring network to obtain a noise interferogram characterizing the acoustic correlation between nodes;

[0008] S4: Identify and measure the time delay and energy values ​​corresponding to the stable cross-correlation peaks in the noise interferogram, and generate the characteristic parameters of the cross-correlation function; wherein, S4 includes:

[0009] S4.1: Based on the noise interferogram, perform peak detection on each cross-correlation function waveform to identify at least one major correlation peak in the waveform;

[0010] S4.2: Measure the lateral distance of the peak point of the main correlation peak relative to the zero time axis to obtain the time delay value;

[0011] S4.3: Calculate the signal energy integral of the waveform of the main correlation peak within a local time window near the time delay value to obtain the energy value;

[0012] S4.4: Based on the time delay value and energy value, combine and record the nodes to generate cross-correlation function characteristic parameters;

[0013] S5: Based on the characteristic parameters of the cross-correlation function, the settlement distribution information of the breakwater foundation is generated by using a pre-trained settlement inversion model.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0015] This invention effectively overcomes the challenges of monitoring foundation settlement in deep-water environments, achieving wide-area, continuous, and highly sensitive automated monitoring. Its effectiveness is specifically reflected in the synergistic effect of each step: First, by fixing multiple hydrophone nodes in at least two monitoring areas based on the breakwater structural design, an underwater acoustic monitoring network is generated, providing a distributed sensing foundation covering key areas for subsequent monitoring and ensuring spatial representativeness. Second, controlling the hydrophone nodes within the network to synchronously record sound pressure fluctuation signals yields multi-channel sound pressure signal data, ensuring temporal consistency and comparability of signal acquisition at different locations, providing a high-quality data source for subsequent analysis. Next, based on the multi-channel sound pressure signal data, pairwise cross-correlation calculations are performed on the signals from different node channels to obtain a noise interferogram. This step cleverly transforms environmental noise into usable acoustic interference signals, thus eliminating the need for active sound source emission and achieving passive monitoring sensitive to minute foundation deformations. Then, the time delay and energy values ​​of stable cross-correlation peaks in the noise interferogram are identified and measured to generate characteristic parameters of the cross-correlation function. This step quantitatively extracts characteristic quantities that reflect changes in the sound wave propagation path from the complex interferometric signal. The time delay value is related to the change in path difference, and the energy value is related to the change in scattering or attenuation, together forming an acoustic fingerprint characterizing changes in the foundation state. Finally, based on the characteristic parameters of the cross-correlation function, a pre-trained settlement inversion model is used to calculate and generate settlement distribution information. This step establishes a mapping between acoustic features and physical settlement through the model, realizing intelligent and quantitative inversion from acoustic signals to engineering settlement.

[0016] The entire process forms a closed loop, from networked data acquisition, noise interference signal extraction, key feature quantification to model-based information inversion. Each step is interconnected and works together to achieve non-contact, wide-area, continuous and high-precision dynamic monitoring of deep-water breakwater foundation settlement. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 The diagram shown is a flowchart illustrating a method for monitoring the foundation settlement of a deep-water breakwater according to an embodiment of the present invention.

[0019] Figure 2 The diagram shown is a structural schematic of a deep-water breakwater foundation settlement monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] The specific embodiments of the present invention will be described below.

[0022] Example 1

[0023] like Figure 1 As shown, the first aspect of this invention proposes a method for monitoring the settlement of deep-water breakwater foundations, comprising:

[0024] S1: Based on the structural design drawing of the breakwater, multiple hydrophone nodes are fixedly installed in at least two monitoring areas of the underwater foundation structure of the breakwater, and all hydrophone nodes are connected to the monitoring host to generate an underwater acoustic monitoring network.

[0025] S2: Control the hydrophone nodes in the underwater acoustic monitoring network to synchronously record the sound pressure fluctuation signals from underwater within the set acquisition period, and obtain multi-channel sound pressure signal data;

[0026] S3: Based on multi-channel sound pressure signal data, pairwise cross-correlation calculations are performed on signals from different hydrophone node channels in the underwater acoustic monitoring network to obtain a noise interferogram characterizing the acoustic correlation between nodes;

[0027] S4: Identify and measure the time delay and energy values ​​corresponding to the stable cross-correlation peaks in the noise interferogram, and generate the characteristic parameters of the cross-correlation function; wherein, S4 includes:

[0028] S4.1: Based on the noise interferogram, perform peak detection on each cross-correlation function waveform to identify at least one major correlation peak in the waveform;

[0029] S4.2: Measure the lateral distance of the peak point of the main correlation peak relative to the zero time axis to obtain the time delay value;

[0030] S4.3: Calculate the signal energy integral of the waveform of the main correlation peak within a local time window near the time delay value to obtain the energy value;

[0031] S4.4: Based on the time delay value and energy value, combine and record the nodes to generate cross-correlation function characteristic parameters;

[0032] S5: Based on the characteristic parameters of the cross-correlation function, the settlement distribution information of the breakwater foundation is generated by using a pre-trained settlement inversion model.

[0033] To monitor the settlement of deep-water breakwater foundations, it is first necessary to select at least two monitoring areas on the underwater foundation structure based on the breakwater's structural design drawings—the drawings that detail the breakwater's geometry, dimensions, materials, and underwater foundation construction. A monitoring area is a specifically designated, representative local area for systematic observation; for example, it can be selected in areas of concentrated load or sections with varying geological conditions. Within each selected monitoring area, multiple hydrophone nodes are fixedly installed. A hydrophone node is an underwater acoustic sensor that converts sensed sound pressure fluctuations into electrical signals, typically comprising a sensor unit, preliminary signal processing circuitry, and a waterproof enclosure. Installation means mechanically attaching it securely to the surface or a predetermined location within the foundation structure. Subsequently, all the distributed hydrophone nodes are physically and logically connected to a central monitoring host via wired or wireless underwater communication links, thereby constructing an underwater acoustic monitoring network covering the target area. This network is a system comprised of spatially distributed sensing nodes, data transmission channels, and a central processing unit. Its technological advantage lies in transforming a vast and concealed underwater foundation into a continuously perceptible acoustic detection field. Through simultaneous deployment in multiple areas, the monitoring system can acquire data from multiple spatial points, thus enabling the analysis of settlement distribution patterns rather than single-point settlement amounts. The derivation of this technological advantage lies in the fact that uneven settlement of the foundation leads to subtle changes in the relative positions of different monitoring areas. These changes in spatial geometry are indirectly captured by subsequently analyzing the relationships between acoustic signals received by different nodes in the network.

[0034] After establishing a stable underwater acoustic monitoring network, the monitoring host needs to send instructions to each hydrophone node in the network to control them to synchronously record data within a pre-set acquisition period, such as an observation window lasting several tens of minutes. Synchronization means that the data acquisition actions of all nodes start and end under a precise and unified time reference, which usually requires a high-precision underwater clock synchronization protocol or a unified timing signal trigger. During this period, each hydrophone node continuously records sound pressure fluctuation signals from the surrounding underwater environment. These signals mainly originate from background noise in the marine environment, such as wave breaking, turbulence, rainfall, and distant ship noise. The signals recorded by all nodes converge at the monitoring host, forming multi-channel sound pressure signal data. Here, "channel" corresponds to each independent hydrophone node; therefore, multi-channel data is essentially a set of time-series signals recorded at different spatial locations within the same time period. The core technical effect of this step is to obtain a raw observation dataset with strict time alignment and spatial diversity. Time synchronization ensures that signals from different nodes can be rigorously compared and analyzed on the time axis, which is an absolute prerequisite for subsequent calculation of the propagation delay difference between signals. The spatially distributed multiple channels capture the response of the sound wave field at different locations, providing a data foundation for inverting medium changes along the path by utilizing the differences in sound wave propagation paths.

[0035] After obtaining the multi-channel sound pressure signal data, the core signal processing stage begins. Based on this data, cross-correlation calculations are performed on the signals from all different hydrophone node channels in the underwater acoustic monitoring network. Cross-correlation is a mathematical operation used to measure the similarity between two signals at different time offsets, and the result is a function of time delay. Performing this calculation on all possible node pairs in the network yields a series of cross-correlation functions. Organizing these functions according to the node pair relationships forms a noise interferogram. The noise interferogram is a data representation that visually shows the correlation between environmental noise signals received by nodes at different locations and their distribution characteristics with time delay. The derivation of its technical effect is based on the principles of interference and propagation in acoustics: although the environmental noise received by a single node appears random and disordered, when multiple nodes receive sound waves from the same or similar noise sources, the different path lengths of the sound waves to different nodes will form stable peaks in the cross-correlation function. The time delays corresponding to these peaks contain the time difference information of the sound waves propagating between node pairs. When the foundation settles or deforms, it alters the acoustic properties of the underwater medium or the effective propagation paths between nodes, leading to systematic changes in the time delay and energy characteristics of these cross-correlation peaks. Therefore, noise interferograms are not meaningless noise, but rather transform chaotic environmental background noise into an analyzable "signal carrier" containing information about the state of the medium.

[0036] Subsequently, quantitative and stable characteristic parameters need to be extracted from the noise interferogram. First, it is necessary to identify and measure the cross-correlation peaks in the spectrum that remain stable under repeated analyses or analyses at different time segments. These stable peaks typically correspond to dominant acoustic paths or main noise sources with stable propagation paths and strong energy. For each identified major correlation peak, two key measurements are performed: first, the lateral distance of the peak point relative to the zero time axis is measured; this distance is the time delay value, which directly quantifies the time required for the sound wave to propagate between two specific hydrophone nodes; second, the signal energy integral of the waveform of the major correlation peak within a local time window near its time delay value is calculated to obtain the energy value, which reflects the strength of the sound wave energy transmitted along this specific propagation path. The time delay value and the energy value together constitute a basic description of the acoustic correlation state between a pair of nodes. Systematically combining and recording the extracted time delay values ​​and energy values ​​of all node pairs according to their corresponding node pair numbers generates the set of cross-correlation function characteristic parameters, as described in this method. The technical effect of this step is to achieve information condensation and characterization. It compresses the high-dimensional, highly detailed, and redundant original interferograms into a set of low-dimensional core feature indicators closely related to physical changes. The time delay value is directly related to the geometric length of the sound wave propagation path and the sound velocity in the medium, and is sensitive to foundation displacement; the energy value may be related to scattering loss or reflection coefficient along the path, and is sensitive to local changes in the foundation material or alterations in contact state. This step provides a clear, structured, and physically meaningful input vector for the next stage of data-driven inversion.

[0037] Finally, the generated cross-correlation function feature parameters are input into a pre-trained settlement inversion model for calculation. The settlement inversion model is a mathematical or computational model that establishes a mapping relationship between acoustic feature parameters (input) and foundation settlement distribution (output). This model is typically obtained through supervised training of machine learning algorithms based on a large amount of data under known working conditions (e.g., obtained through numerical simulation or physical model experiments). After receiving the feature parameters extracted during the current monitoring period, the model outputs the breakwater foundation settlement distribution information corresponding to the current state through complex nonlinear calculations. This information can be represented as numerical settlement values ​​for different areas or settlement contour maps. This step is crucial in transforming acoustic observations into directly usable engineering information, and its ultimate technical effectiveness depends on the groundwork laid by all the preceding steps.

[0038] The synergistic effect of the entire technology chain is reflected in the following aspects: network deployment provides the spatial framework and observational basis for inversion; synchronous acquisition ensures the quality of raw data and the feasibility of time delay analysis; cross-correlation calculation extracts interferometric patterns containing structural information from random noise; feature extraction quantifies the patterns into parameters that the model can process; and finally, the inversion model acts as a "decoder," translating acoustic parameters into intuitive settlement information. Thus, this method enables long-term, continuous, wide-area, and quantitative monitoring of deep-water breakwater foundation settlement using ubiquitous environmental noise without interfering with the structure or relying on active sound sources.

[0039] In some implementations, S1 includes:

[0040] S1.1: Based on the cross-sectional design of the breakwater, multiple deployment locations are selected on the underwater foundation surface on the seaward side, and hydrophone nodes are installed to generate the first hydrophone node group.

[0041] S1.2: Based on the cross-sectional design of the breakwater, multiple deployment locations are selected on the underwater foundation surface on the seaward side, and hydrophone nodes are installed to generate a second hydrophone node group.

[0042] S1.3: Based on the first hydrophone node group and the second hydrophone node group, all hydrophone nodes are connected to the same monitoring host through data transmission cables to generate an underwater acoustic monitoring network.

[0043] This embodiment specifies the steps for generating an underwater acoustic monitoring network, clarifying that the node deployment strategy is based on the cross-sectional design of the breakwater. The cross-sectional design drawing is an engineering drawing cut perpendicular to the length of the breakwater, clearly showing the structural composition of the breakwater from the seaward side to the leeward side, such as the underwater foundation, the breakwater body, and the cross-sectional shape of the top. First, based on this design drawing, multiple deployment locations are selected on the surface of the underwater foundation on the seaward side. The seaward side refers to the side of the breakwater that directly faces the open sea and bears the impact of dynamic loads such as waves and currents. Installing hydrophone nodes at these selected locations forms the first hydrophone node group. The technical effect of this deployment is specifically targeted, because the foundation on the seaward side directly bears external loads and is often the starting area for stress concentration and deformation. Deploying the node group here can preferentially capture changes in the acoustic response of the foundation under direct load, helping to detect potential instability signs or localized damage at an early stage. The reasoning behind this technical effect is that prioritizing the deployment of sensing resources in areas with the highest risk and most sensitive response can improve the early warning efficiency and reliability of the monitoring system.

[0044] Following this, based on the cross-sectional design drawings, multiple locations were selected on the underwater foundation surface on the seaward side to install hydrophone nodes, creating a second set of hydrophone nodes. The seaward side refers to the relatively sheltered side of the breakwater away from the open sea, where the hydrodynamic environment is typically milder than that of the seaward side. The technical advantage of deploying the second set of nodes is that it establishes a comparative reference system. Deformation of the foundation on the seaward side is often related to deformation on the seaward side, and may be caused by load transfer, overall settlement, or differential settlement. By simultaneously monitoring both sides, overall response information in the cross-sectional direction of the breakwater can be obtained. Monitoring only one side may not be able to distinguish between local deformation and overall tilt, while deploying on both sides provides a data basis for identifying this difference.

[0045] After deploying the hydrophone node groups on both the seaward and leeward sides, underwater data transmission cables are used to connect all the hydrophone nodes in these two groups to a single monitoring host. This operation integrates the physical network, ultimately creating a complete underwater acoustic monitoring network spanning the entire cross-section of the breakwater and covering both the seaward and leeward sides. The deeper reasoning behind this technical advantage lies in the fact that this deployment, based on structural symmetry or correspondence, allows subsequent cross-correlation calculations to be performed not only between nodes on the same side but also across the breakwater structure, between nodes on the seaward and leeward sides. This cross-structure node pairing is uniquely sensitive to sensing the overall deformation patterns of the breakwater, such as tilting, torsion, or overall settlement. This is because when the breakwater tilts as a whole, the length of the sound wave propagation path across the breakwater and the conditions of the traversed medium undergo systematic changes, which are clearly reflected in the cross-correlation function characteristics of the nodes on both sides. Therefore, this deployment strategy greatly enhances the monitoring system's ability to perceive the overall deformation modes of the breakwater, so that the settlement distribution information obtained from the final inversion not only reflects the subsidence of local points, but also reveals the spatial deformation trend of the overall structure, significantly improving the engineering application value of the monitoring results and providing more comprehensive data support for assessing the overall stability and safety of the breakwater.

[0046] In some implementations, S1.1 includes:

[0047] S1.1a: Based on the length direction of the breakwater, determine multiple coordinates of the underwater foundation surface on the seaward side at uniform intervals, and generate node layout points;

[0048] S1.1b: Based on the node layout points, the hydrophone nodes are fixed to the surface of the underwater riprap foundation or pile foundation through the mounting base;

[0049] S1.1c: Adjust the orientation of the fixed hydrophone nodes so that the sensing surface faces the open water area, and generate the first hydrophone node group.

[0050] This embodiment further refines the specific installation process of the first hydrophone node group, ensuring the standardization of node layout, uniformity of spatial coverage, and optimization of signal reception quality. First, based on the length direction of the breakwater, i.e., the longitudinal extension direction with the breakwater axis as the reference, multiple installation coordinates on the underwater foundation surface on the seaward side are determined at uniform intervals, thus generating a series of node layout points. Uniform spacing means that an installation point is determined at fixed distance intervals (e.g., every twenty or fifty meters) along the breakwater line. The technical effect of this step is to achieve continuity and systematic monitoring along the longitudinal direction of the breakwater. Uniform point layout avoids monitoring blind spots, ensuring that when significant settlement occurs at any location along the breakwater line, a sufficiently close sensor can capture its impact. Simultaneously, the regular spatial distribution makes the collected data consistent and comparable in the longitudinal dimension, providing a structured data grid foundation for subsequent analysis of longitudinal settlement differences within the breakwater, identification of local settlement sections, or analysis of the propagation law of settlement along the breakwater line. If the point layout is random or unevenly spaced, it may lead to difficulties in data analysis and make it difficult to accurately determine the range and gradient of settlement.

[0051] After determining the specific coordinates of the node deployment points, the physical installation phase of the hydrophone nodes begins. The hydrophone nodes are securely fixed to the surface of the underwater riprap foundation or pile foundation using specially designed mounting bases. These mounting bases are mechanical components designed to withstand harsh underwater environments (such as corrosion, high pressure, and biofouling) and ensure long-term stability. They may be made of corrosion-resistant materials and achieve a rigid connection to the foundation structure through anchor bolts, adhesives, or gravity seats. Fixing the nodes to the surface of the riprap foundation (usually constructed of piled stones) or pile foundation (concrete or steel support structure) means that the sensor achieves direct physical coupling with the main foundation structure of the breakwater to be monitored. The technical effectiveness of this step is crucial; it is a prerequisite for ensuring that the monitoring data accurately reflects foundation deformation rather than the movement of the sensor itself. Only when the absolute position and spatial orientation of the sensor node remain highly stable during long-term monitoring can the temporal variation sequence of the recorded sound pressure signal be reliably interpreted as being due to changes in the external foundation condition. Loose installation will result in the inclusion of irrelevant sensor sway noise in the signal, severely interfering with or even masking the true foundation deformation signal.

[0052] After the physical fixing of the nodes is completed, the installation work is not finished. Fine-tuning of the orientation of the fixed hydrophone nodes is still required. The goal of this adjustment is to ensure that the sensing surface of the hydrophone, i.e., the direction in which its sound pressure sensitive element faces, is towards open water. Open water typically refers to the outside of the breakwater, the main direction of acoustic environmental noise (especially wave noise). The technical effect of this orientation adjustment is to maximize the signal-to-noise ratio and directional consistency of the received signal. Facing the sensing surface directly towards the main noise source allows the hydrophone to receive sound wave energy from that direction with the highest sensitivity, thus obtaining a stronger and higher-quality original acoustic signal. At the same time, this also gives the node a certain degree of directivity, helping to suppress noise from other non-target directions (such as multipath reflections from the seabed or interference from the back of structures), making the main component of the recorded signal, originating from open sea environmental noise and carrying information about the ground along the way, more prominent. These three sequential sub-steps—uniformly determining the locations along the breakwater line, achieving stable installation using dedicated bases, and fine-tuning the node orientation to optimize reception—constitute the complete process for the high-quality deployment of the first hydrophone node group. This process, encompassing spatial planning, mechanical fixation, and acoustic optimization, ensures the reliability and effectiveness of the key monitoring array on the coastal side, laying a solid foundation for obtaining high-quality, highly consistent raw data for the entire underwater acoustic monitoring network.

[0053] In some implementations, S2 includes:

[0054] S2.1: Based on a unified timing signal, control all hydrophone nodes in the underwater acoustic monitoring network to start synchronously and record the simulated sound pressure signal at a set sampling frequency to generate the original sound pressure time domain signal;

[0055] S2.2: Based on the original sound pressure time domain signal, a bandpass filter is used to filter out components outside the set frequency band, and the signal amplitude is normalized and adjusted to generate a pre-processed sound pressure signal;

[0056] S2.3: Based on the preprocessed sound pressure signal, encapsulate it according to channel and time sequence to generate multi-channel sound pressure signal data.

[0057] This embodiment details the preprocessing flow from synchronously acquiring raw signals to generating standardized multi-channel acoustic pressure signal data. This process is a crucial data preparation stage to ensure the accuracy and effectiveness of subsequent analysis. First, high-precision time synchronization of all hydrophone nodes needs to be achieved. This is controlled by a unified timing signal, which can be generated by the monitoring host and distributed through the network, or aligned by each node's built-in high-stability clock receiving external synchronization sources (such as GNSS signals relayed via surface buoys or dedicated underwater acoustic synchronization signals). Based on this unified timing signal, all hydrophone nodes begin digitizing the analog acoustic pressure signal at the same absolute time point using the exact same, pre-set sampling frequency, generating the raw acoustic pressure time-domain signal for each channel. The raw acoustic pressure time-domain signal is an unprocessed, time-ordered sequence of voltage values, representing the change in acoustic pressure at the sensor's location over time. The core technical effect of this step is establishing a unified and accurate time reference for the entire monitoring network. Microsecond or even nanosecond-level time synchronization errors may correspond to centimeter-level spatial positioning errors in acoustic monitoring. Therefore, strict time synchronization is the absolute foundation for all subsequent analyses based on signal propagation delay. It ensures that signals recorded at different spatial locations have a strict correspondence on the time axis, so that the calculated delay value can truly reflect the physical time difference of sound wave propagation, thereby accurately relating it to the geometric deformation of the foundation.

[0058] After acquiring the raw sound pressure time-domain signals from each channel, preprocessing is required to improve signal quality. The first preprocessing step is to use a digital bandpass filter to filter out frequency components outside the designated frequency band. A bandpass filter is an electrical or digital processing module that allows signals within a specific frequency range (passband) to pass through, while attenuating or blocking signals at frequencies outside the passband. Its effectiveness is derived from the frequency domain separation characteristics of signal and noise: the effective components of marine environmental noise sensitive to structural deformation, as well as acoustic signals generated by structural vibration, deformation radiation, or modulation, often concentrate their energy within a specific frequency range. This range may be related to the structure's natural vibration frequency, the collision frequency between rocks, or the sensitive wavelength of sound wave interaction with the structure. Outside this range, there may be high-frequency electronic thermal noise, low-frequency tidal pressure fluctuations, or other irrelevant environmental interference. Bandpass filtering effectively preserves the signal within this "information-rich" frequency band while significantly suppressing irrelevant out-of-band noise, thereby improving the overall signal-to-noise ratio. This makes subsequent cross-correlation calculations more likely to identify stable, physically meaningful interference peaks, rather than being overwhelmed by broadband noise.

[0059] After filtering, the signal amplitude usually needs to be normalized. Normalization refers to scaling the amplitude of the filtered signal from each channel according to certain rules (e.g., adjusting its peak or root mean square value to the same reference level). The technical effect of this step is to eliminate or reduce signal strength differences between channels caused by non-geological factors. In reality, slight differences in the sensitivity of different hydrophone nodes, differences in hydrostatic pressure due to varying water depths at installation locations, or differences in noise intensity received due to different orientations will all cause inherent differences in the amplitude of the original signals from each channel. Without normalization, these amplitude differences will dominate the cross-correlation calculation, potentially masking weak but crucial time delay differences for monitoring. By normalizing, the signals from each channel are adjusted to similar amplitude levels, allowing the cross-correlation function to reflect the temporal similarity of the signal waveforms (i.e., time delay information) more accurately than the amplitude strength, thus highlighting the most critical characteristic for monitoring foundation displacement.

[0060] Finally, the preprocessed (i.e., filtered and normalized) sound pressure signals from each channel are encapsulated according to their corresponding node channel numbers and strict time order. Encapsulation means combining the data stream with metadata describing its attributes (such as channel ID, sampling rate, start timestamp, etc.) into a standard format data packet or file. The resulting multi-channel sound pressure signal data is a dataset with a clear structure, well-defined spatiotemporal information, and preliminary quality optimization. The technical advantage of this step is that it facilitates subsequent batch automated processing. It integrates scattered, independent time series into an ordered, complete data entity, making it easy to store, transmit, and input into the cross-correlation calculation module for efficient pairwise processing. From synchronous acquisition to establish a unified time reference, to bandpass filtering to improve the signal-to-noise ratio, to amplitude normalization to eliminate channel differences, and finally to encapsulation to form a regular dataset, these four sub-steps constitute an interlocking data preprocessing chain. Their synergistic effect ensures that the data input into the core analysis algorithm is high-quality, consistent, and reliable, providing a solid data quality guarantee for the entire monitoring method to accurately deduce foundation settlement information.

[0061] In some implementations, S2.2 includes:

[0062] S2.2a: Based on the dimensions and material properties of the underwater structure of the breakwater, calculate the range of acoustic wavelengths that are sensitive to small displacements of the structure, and determine the sensitive acoustic frequency bands;

[0063] S2.2b: Based on the sensitive acoustic frequency band, set the upper and lower passband frequencies of the bandpass filter;

[0064] S2.2c: Based on the upper and lower frequency limits, the original sound pressure time-domain signal is digitally filtered to generate a pre-processed sound pressure signal.

[0065] This embodiment provides a specific physical basis for setting the passband frequency of the bandpass filter, transforming the selection of this key technical parameter from empirical judgment to a deterministic process based on the analysis of the characteristics of the monitored object, significantly improving the scientific rigor and adaptability of the method. First, calculations and analyses are needed based on the specific dimensions and material properties of the underwater breakwater structure to determine the range of acoustic wavelengths sensitive to minute structural displacements, which are then converted into sensitive acoustic frequency bands. Here, the dimensions of the underwater breakwater structure may include characteristic dimensions such as the typical stone size of the riprap foundation, the average gap between stones, and the diameter and spacing of the pile foundations; material properties include the density and elastic modulus of the stone and concrete, as well as the propagation speed of sound waves within them. The physical principle of the calculation is that when the wavelength of the incident sound wave is comparable to these characteristic dimensions of the structure, significant scattering, diffraction, or resonance phenomena occur, and minute changes in the structure's geometry strongly modulate these acoustic interaction processes. Therefore, through theoretical analysis (such as scattering theory) or numerical simulation (such as finite element acoustic simulation), one or more frequency bands can be identified where the coupling effect between environmental noise and the underwater structure of the breakwater is strongest, meaning the acoustic signal is most sensitive to geometric deformation or material state changes in the structure. The technical benefit of this step is determining the optimal "monitoring channel" for the entire acoustic monitoring process. It avoids the problems of insufficient sensitivity or signal submersion in noise that may result from blindly selecting general frequency bands, achieving a match between the monitoring frequency band and the physical characteristics of the monitored target, thereby maximizing the proportion of effective information in the monitoring signal at the source.

[0066] After calculating the sensitive acoustic frequency band, the upper and lower limits of the digital bandpass filter are specifically set based on this band. For example, if calculations show that the frequency band sensitive to the deformation of a certain riprap foundation bed is between several kilohertz and ten thousand Hz, then the lower limit of the bandpass filter is set to several kilohertz, and the upper limit is set to ten thousand Hz, covering the sensitive frequency band. The derivation of its technical effect is direct and clear: precisely aligning the filter's passband with the structurally sensitive acoustic frequency band is equivalent to setting up an "information filter" in the frequency domain. This filter only allows those frequency components most likely to carry foundation deformation information to pass through, while resolutely blocking a large number of noise components outside this frequency band that are unrelated to deformation (such as extremely low-frequency water pressure fluctuations, extremely high-frequency electronic noise, etc.). This ensures that the energy of the pre-processed sound pressure signal output from the hydrophone node is mainly concentrated in the frequency range most valuable for the monitoring task.

[0067] Finally, based on the pre-defined upper and lower frequency limits, digital signal processing techniques are used to perform digital filtering on the raw sound pressure time-domain signal of each channel. This processing is completed in real-time or post-processing within the computing unit or embedded processor of the monitoring host, generating the final pre-processed sound pressure signal. The ultimate technical effect is that this physically-based, customized frequency selection significantly optimizes the quality of the raw data. It ensures that the acoustic signal used for subsequent cross-correlation calculations has undergone "spectrally purified" analysis, containing characteristic frequency components related to foundation deformation, while effectively suppressing background noise. This not only improves the significance of stable interference peaks in the cross-correlation function but also enhances the stability and reliability of characteristic parameters such as time delay and energy values ​​extracted from the noise interferogram. Therefore, this step directly links specific engineering structural properties with the core parameters of signal processing, making it a crucial step in improving the targeting, sensitivity, and accuracy of passive acoustic monitoring methods. This allows the method to adjust parameters according to the structural characteristics of different breakwaters (such as riprap foundations or pile foundations), giving it good engineering versatility and customizability, and ensuring that the monitoring system can maintain optimal monitoring performance when facing different types of underwater foundations.

[0068] In some implementations, following S4.4, the following are also included:

[0069] S4.5: Based on the cross-correlation function characteristic parameters generated in the current measurement cycle and the cross-correlation function characteristic parameters stored in the historical measurement cycles, the time delay value and energy value are calculated by subtraction to obtain the time delay offset and energy change.

[0070] In S5, the characteristic parameters of the cross-correlation function calculated by inputting the settlement inversion model are the time delay offset and the energy change.

[0071] After generating the cross-correlation function characteristic parameters, this method does not directly use them for the final calculation. Instead, it introduces a crucial differential processing step. Specifically, this step involves comparing the cross-correlation function characteristic parameters generated in the current measurement period with those stored in the system for historical measurement periods. Historical measurement periods are typically selected at the initial stage of breakwater construction, after loading stabilization, or at a point confirmed as a baseline state. The stored characteristic parameters represent the initial acoustic "fingerprint" of the system under conditions of no new settlement or known stability. For each pair of hydrophone nodes, two differential calculations are performed: the first is to subtract the time delay value of the node pair measured in the current period from the time delay value recorded for the same node pair in the historical period; the difference is the time delay offset. The second is to subtract the energy value of the historical period from the energy value of the current period to obtain the energy change. The time delay offset is a scalar that can be positive or negative. Its physical meaning is how much the time difference of sound wave propagation between two specific hydrophones has changed compared to the baseline state; a positive value indicates an increase in propagation time, and a negative value indicates a decrease. The change in energy is also a signed numerical value, reflecting the enhancement or weakening of the acoustic signal energy along this propagation path relative to the reference state.

[0072] The derivation of the technical effectiveness of this step is based on a deep analysis of the composition of the monitoring signal. The raw time delay and energy values ​​extracted directly from a single monitoring session are a composite quantity that mixes multiple types of information. It contains at least three main components: the first component is the physical change in the sound wave propagation path caused by new settlement or deformation of the foundation, which is the core signal we want to extract; the second component is the inherent static deviation of the entire monitoring system that does not change over time, such as the fixed geometric deviation caused by the imperfectly precise installation position of the hydrophone nodes, the slight inconsistencies in the response characteristics of each hydrophone and its preamplifier circuit, and the local differences in sound velocity caused by the inhomogeneity of the initial state of the underwater foundation material; the third component is the random fluctuation of the environmental noise field itself on a short time scale. If these absolute characteristic parameters, which mix the target signal, system static deviation, and random noise, are directly input into the settlement inversion model, the model will have to painstakingly separate the settlement signal of interest from the complex input. This not only greatly increases the difficulty of the model learning task and requires a larger amount of training data, but also easily leads to model overfitting or unstable inversion results, because static bias and random noise will interfere with the model's capture of the stable mapping relationship between settlement and acoustic features.

[0073] By introducing differential calculations to generate time delay offsets and energy changes, a "baseline cancellation" or "high-pass filtering" operation in signal processing is essentially performed. The core principle is that the inherent static bias of the system remains almost constant between two measurements (current and historical). Therefore, when performing differential calculations, these static bias components are effectively canceled out. Similarly, the influence of components in environmental noise that have long-term statistical stationarity is significantly weakened after differential calculation. Ground settlement is a slow but continuously accumulating process; the changes in acoustic characteristics it causes gradually manifest over time and are "sedimented" into the current measurement values. Therefore, the time delay offsets and energy changes retained after differential calculation primarily reflect the changes in acoustic characteristics introduced by changes in ground condition from the historical reference time to the current time. This makes subsequent input data "purer" and "more focused" on dynamic changes.

[0074] Therefore, in the calculation stage of the settlement inversion model, the characteristic parameters of the cross-correlation function input to the model are no longer the original time delay and energy values, but rather the time delay offset and energy change after differential processing. This change brings several significant technical benefits. First, it greatly improves the accuracy and reliability of the inversion model. Because interference noise in the input data is suppressed in advance, the model can learn the physical relationship between acoustic feature changes and foundation settlement displacement more directly and clearly, resulting in a more accurate model and less fluctuation in prediction results. Second, it significantly enhances the sensitivity of the monitoring system to minute settlements. Because differential processing amplifies the "change" part, even weak acoustic feature changes caused by micron- or millimeter-level settlement can be effectively extracted from the background, which is beneficial for early warning. Furthermore, this method improves adaptability to different engineering sites. Regardless of the initial installation conditions and geological background of different breakwater projects, as long as their own stable baseline state data is obtained, subsequent monitoring is transformed into the measurement and inversion of "changes," which makes the core algorithm highly versatile. In summary, from historical data retrieval and differential calculation to using dynamic changes as model input, this series of operations constitutes an ingenious signal preprocessing and feature enhancement process. It transforms an absolute quantity measurement system that may be troubled by various initial conditions into a highly sensitive monitoring system focused on capturing and quantifying "changes," thereby laying a solid data foundation for achieving high-precision and high-reliability dynamic assessment of foundation settlement.

[0075] In some implementations, the settlement inversion model is pre-trained through the following steps:

[0076] Based on the design drawings, material parameters, and sound velocity profile of the breakwater, a numerical model including structural geometry and acoustic propagation characteristics is established.

[0077] In the numerical model, a virtual receiving point is set up that corresponds to the location of the underwater acoustic monitoring network, and a simulated marine environmental noise source is loaded to generate a simulated sound field.

[0078] Based on the simulated sound field, the acoustic cross-correlation function between each virtual receiving point is calculated under different preset settlement conditions, and a simulated noise interferogram set is generated.

[0079] Based on the simulated noise interferogram set, simulated characteristic parameters under different working conditions are extracted, and a database of mapping relationship between settlement displacement field and changes in simulated characteristic parameters is constructed.

[0080] Based on the mapping relationship database, the selected machine learning model is trained under supervision to obtain the settlement inversion model.

[0081] The settlement inversion model is the core intelligent component of the entire monitoring method, and its pre-training process is a systematic engineering project integrating numerical simulation, physical modeling, and machine learning. The first step in training is to construct a numerical model that can accurately reflect the actual monitoring scenario. This requires the comprehensive use of three key types of information: the breakwater design drawings, material parameters, and the sound velocity profile of the water area. The design drawings provide detailed three-dimensional geometric dimensions and spatial relationships of the breakwater's underwater foundation structure (such as riprap foundation and pile foundation), the breakwater body, and even the superstructure; this forms the spatial skeleton of the model. Material parameters include the density, elastic modulus, Poisson's ratio, and more critical acoustic parameters (such as P-wave velocity, S-wave velocity, and attenuation coefficient) of the building materials such as concrete, stone, and steel. These parameters determine the physical laws governing the interaction between sound waves and the structure (such as reflection, refraction, scattering, and transmission). The sound velocity profile of the water area describes the curve of the speed of sound waves propagating in the seawater medium as a function of water depth; this is a key environmental factor affecting the sound wave propagation path and propagation time. By integrating the above information into specialized computational acoustics software or simulation platforms based on finite element and boundary element methods, a high-fidelity numerical model encompassing structural geometry and acoustic propagation characteristics can be established. The technical advantage of this model lies in creating a fully controllable "digital twin" breakwater with arbitrarily configurable operating conditions. This makes it possible to simulate various settlement scenarios and their acoustic responses in a computer, overcoming the fundamental obstacle of conducting destructive experiments on real structures to obtain training data.

[0082] In the established numerical model, virtual receiving points need to be precisely set up to perfectly match the actual deployment scheme of the future underwater acoustic monitoring network. This means that if the actual engineering plan is to install hydrophone nodes at ten specific coordinates, then ten virtual observation points with ideal point receiver characteristics must be set up at the same spatial coordinate location in the numerical model. These virtual receiving points will be used to "record" the sound pressure signals in the simulated environment. Subsequently, sound sources capable of simulating real marine environmental noise are loaded into the aquatic environment simulated by the model. These noise sources are not single point sources, but distributed sources or random sound fields configured according to the statistical characteristics of marine environmental noise (such as spatial distribution following a certain random model, frequency spectrum conforming to classic marine noise spectra such as the Knudsen spectrum, and intensity conforming to typical sea states), thereby generating a simulated sound field that approximates reality. Running this numerical model containing noise sources, each virtual receiving point will output a series of time-domain sound pressure signals. This process simulates the process of real hydrophone nodes collecting multi-channel sound pressure signal data, and the simulated data naturally contains the acoustic characteristics under a specific structural state (at this time, the original design state without deformation).

[0083] Next, this numerical simulation platform was used to conduct proactive and systematic numerical experiments. A series of different preset settlement conditions were artificially set in the model. These conditions needed to cover various possible settlement modes, such as: uniform settlement in localized areas (e.g., at the toe of the breakwater), uneven settlement along the breakwater line (forming settlement troughs), differential settlement between the seaward and leeward sides (leading to breakwater tilting), and complex settlement involving multiple modes. For each preset settlement condition, the geometric coordinates of the corresponding part of the breakwater foundation in the numerical model were adjusted or different material properties were assigned (simulating soil softening), and then the acoustic simulation calculations were rerun. For each condition, the acoustic cross-correlation function between all pairs of virtual receiving points was simulated, just like processing real data, thereby generating simulated noise interferograms corresponding one-to-one with that settlement condition. Repeating this process for all preset conditions yielded a large and diverse set of simulated noise interferograms. This atlas is identical in data format to the field measured atlas, but its key advantage is that each atlas corresponds to a known and precisely quantified settlement displacement field (i.e., preset working conditions).

[0084] Based on this simulated noise interferometric spectrum, simulated feature parameters identical to those in the real data processing can be extracted, such as the time delay and energy values ​​of each node pair. Then, by calculating the changes in the simulated feature parameters relative to the non-settlement baseline state under each settlement condition (i.e., time delay offset and energy change), and correlating them with the known settlement displacement field (usually represented as a numerical matrix of settlement values ​​for each region or node) under that condition, a massive mapping database can be gradually constructed. This database essentially stores a complex, high-dimensional, nonlinear mapping relationship: "what acoustic feature change pattern (input) corresponds to what foundation spatial settlement pattern (output)." Finally, a machine learning model with strong nonlinear fitting capabilities, such as a deep neural network, random forest, or gradient boosting decision tree, is selected. Using the constructed mapping database as the training and validation set, the selected machine learning model is subjected to supervised training. During training, the goal of model learning is to predict the corresponding foundation settlement displacement field as accurately as possible when given a set of acoustic feature parameter changes (i.e., time delay offset and energy change). After sufficient iterative training, hyperparameter tuning, and model validation, the resulting computational model, capable of predicting settlement spatial distribution from acoustic features, is the settlement inversion model to be deployed and applied.

[0085] The technical advantage of the entire pre-training method lies in its ability to generate massive, diverse, and absolutely accurately labeled training samples economically, efficiently, and non-destructively through high-fidelity numerical simulation technology, solving the pain point of the extreme difficulty in obtaining "settlement-acoustic" paired data in practical engineering. This ensures that the trained settlement inversion model not only has a solid physical foundation but also possesses strong generalization ability and engineering applicability. It is the theoretical cornerstone and core computational engine that enables this method to accurately invert complex settlement distributions from random noise.

[0086] In some implementations, following S5, the following are also included:

[0087] S6: Acquire displacement monitoring data from the displacement monitoring system deployed on the top of the breakwater during the same time period;

[0088] S7: Spatially match the settlement distribution information with the displacement monitoring data at the horizontal projection position, and compare the consistency between the vertical settlement trend and the embankment top displacement trend to generate data comparison results;

[0089] S8: Based on the data comparison results, the output values ​​of the settlement inversion model are proportionally calibrated or the local inversion path parameters are optimized to generate a calibrated settlement monitoring report.

[0090] After obtaining the settlement distribution information of the foundation through the settlement inversion model, this method introduces an important cross-validation and data fusion step, aiming to improve the reliability and engineering acceptance of the final results by comparing with mature independent monitoring technologies. This step begins with the simultaneous acquisition of data from another monitoring system. Specifically, it is necessary to acquire displacement monitoring data recorded by a displacement monitoring system deployed on the top of the breakwater during the same time period as the underwater acoustic data acquisition. The breakwater top displacement monitoring system typically refers to a real-time dynamic measurement system based on the Global Navigation Satellite System (GNSS) or a system that uses a high-precision total station for automated periodic measurement. These systems, by deploying fixed observation piers or prisms on the breakwater top, can directly and continuously measure the absolute coordinate changes of each monitoring point in three-dimensional space, and through data processing, can accurately calculate the settlement in the vertical direction. The emphasis on acquiring data "within the same time period" is to ensure that the two independent systems are observing the deformation response of the breakwater structure under the same environmental loads (such as the same tide level and wave conditions) and the same time process, thus ensuring the comparability of the two sets of data in the time dimension and avoiding misjudgment of trends due to asynchronous monitoring time.

[0091] After obtaining these two sets of monitoring data from different sources and based on different principles, in-depth data fusion analysis is required. The first step is spatial matching, which involves associating the settlement distribution information obtained from underwater acoustic inversion with the breakwater displacement monitoring data in terms of horizontal projection positions. Each inversion result in the settlement distribution information (which can be regarded as a virtual monitoring point) is associated with a horizontal coordinate (e.g., east and north coordinates), while each GNSS monitoring point on the breakwater top also has its precise horizontal coordinates. Since the underwater monitoring points are located on the foundation and the breakwater top monitoring points are located on the top of the structure, their elevations are different. Therefore, it is necessary to project their horizontal coordinates onto the same reference plane (such as mean sea level or design elevation plane) for position comparison. Spatial analysis algorithms are used, such as finding the nearest breakwater top monitoring point on the horizontal plane for each underwater inversion point, or establishing a correlation by normal projection according to the breakwater axis, thereby finding the most likely corresponding measured reference point on the breakwater top for the underwater results. After completing spatial matching, the second step is comparative analysis. The key focus is to analyze and compare the vertical settlement trend revealed by underwater acoustic inversion with the displacement trend of the dike crest reflected by direct measurements of the dike crest, and whether there is consistency between the two. "Trend" here is a comprehensive concept, which includes spatial distribution trends (e.g., whether both show that the settlement in the middle section of the dike is greater than that at both ends), relative magnitude trends (e.g., whether both show that the settlement at point A is greater than that at point B), and temporal evolution trends (e.g., whether the changes in settlement rate shown by both are synchronized over multiple consecutive periods).

[0092] Based on the above data comparison results, the system can perform targeted calibration or optimization of the settlement inversion model output, and finally generate a calibrated settlement monitoring report. The derivation of the technical effectiveness of this step stems from the understanding of the characteristics of different monitoring technologies. Levee top displacement monitoring (such as GNSS) is a direct geometric measurement technology that measures the absolute spatial position change of the top of the structure. It is a mature technology with high accuracy, and the results are intuitive and often used as a reference benchmark for verifying other methods. However, it cannot directly measure underwater foundation settlement, and the sampling points are limited to the levee top, resulting in limited spatial resolution. Underwater acoustic inversion methods, on the other hand, are an indirect physical inversion technology that can provide information on the large-scale, spatially continuous settlement distribution within the foundation. However, the absolute accuracy of the inversion results is affected by various factors such as model errors and uncertainties in the sound velocity field. By combining the two, the high-precision, high-reliability "point" absolute measurements from the levee top can be used to calibrate, verify, or correct the "area" relative settlement field obtained from underwater acoustic inversion. If the comparison results show a high degree of consistency between the two trends, it confirms that the acoustic inversion model is reliable in capturing settlement spatial patterns and dynamic changes. At this point, the absolute settlement value of the levee crest can be used to perform overall scaling factor calibration or zero-point shift on the underwater inversion results, thereby giving the acoustic inversion results more accurate absolute numerical meaning, upgrading it from an excellent "relative change monitor" to a reliable "absolute settlement measurer." If local inconsistencies are found, they can provide clues for model optimization or engineering anomaly diagnosis. Therefore, this fusion and calibration process essentially constructs a multi-layered, mutually verifying monitoring system. It significantly enhances the overall robustness of the monitoring system, the reliability of the results, and the engineering practical value of the outcomes. This ensures that the output settlement monitoring report is no longer a product of a single technology, but a comprehensive conclusion verified and optimized through multi-source data, providing a more solid and reliable decision-making basis for engineering safety management.

[0093] In some implementations, S7 includes:

[0094] S7.1: Calculate the settlement of each monitoring point on the top of the embankment based on the three-dimensional coordinate data provided by the displacement monitoring system;

[0095] S7.2: Based on the horizontal coordinates of the underwater monitoring points in the settlement distribution information, project and associate them with the nearest monitoring point on the top of the dike;

[0096] S7.3: Based on the data after projection correlation, analyze the numerical relationship and trend of underwater settlement gradient and levee crest settlement at the same vertical location;

[0097] S8 includes:

[0098] S8.1: Based on the analysis results of S7.3, if the trend of change is consistent, the absolute value of the settlement obtained by underwater acoustic inversion is linearly calibrated using the settlement of the dike crest.

[0099] S8.2: Based on the analysis results of S7.3, if there is a deviation in the local trend of change, a correction coefficient for the acoustic propagation path of the corresponding monitoring area will be added to the settlement inversion model to generate a calibrated settlement monitoring report.

[0100] This embodiment provides a more detailed and operational description of the data comparison and calibration process. First, in S7.1, the raw data from the levee crest displacement monitoring is processed. The displacement monitoring system (such as a GNSS receiver) directly outputs the three-dimensional coordinates of each monitoring point in a specific coordinate system (such as an engineering-independent coordinate system), including east-facing coordinates, north-facing coordinates, and elevation. To obtain the settlement, a stable benchmark point or benchmark network needs to be selected or established. Then, the change in the vertical direction (usually the elevation direction) of each monitoring point relative to its initial value or the previous period's value is calculated. This change is the levee crest settlement at that point. This is a transformation process from raw observations to a target quantity with clear engineering significance, providing a unified metric for subsequent comparison with underwater data.

[0101] Next, in S7.2, the core spatial projection correlation operation is performed. Settlement distribution information is derived from underwater acoustic inversion, and the results are typically given as grid points or specific location points, each containing a horizontal position and an estimated settlement value. Since there is a spatial height difference between the underwater foundation and the breakwater top structure, direct coordinate comparison is meaningless; therefore, projection correlation is necessary. Specifically, the horizontal coordinates of each underwater monitoring point in the settlement distribution information are projected vertically upwards onto the plane of the breakwater top or the centerline of the structure. Then, on this projection plane, the actual monitoring point on the breakwater top that is closest to the projected point is found, and these are established as a correlation pair. The physical assumption of this correlation method is that, in the vertical projection relationship, the settlement of the foundation will be reflected roughly synchronously in the displacement of the breakwater top directly above it through the structural transfer. Therefore, the inverted settlement at this underwater point should theoretically have the strongest correlation with the measured settlement at the corresponding breakwater top, which establishes a reasonable spatial correspondence framework for subsequent quantitative comparison.

[0102] Then, in S7.3, based on the established set of projection correlation pairs, in-depth numerical relationship and trend analysis is conducted. The analysis is multifaceted: First, it analyzes numerical relationships, calculating the ratio or difference between the underwater settlement gradient (i.e., the inverted settlement) and the breakwater crest settlement for each pair of correlation data, observing whether this relationship is essentially constant throughout the breakwater or exhibits spatial variation; second, it analyzes spatial distribution trends, plotting underwater and crest settlement curves along the breakwater line, observing whether the undulations, peaks, and troughs of the two curves match; third, it analyzes temporal trends, analyzing whether the rate of change (acceleration) of underwater settlement and the rate of change of crest settlement remain consistent at the same correlation location for data from multiple consecutive monitoring periods. This consistency analysis is crucial for determining whether the acoustic inversion model reliably captures the actual deformation process.

[0103] Based on the detailed analysis results in S7.3, S8 presents targeted calibration strategies. S8.1 addresses the case of global systematic bias. If the analysis results show that at all or most associated locations, the underwater settlement gradient and the crest settlement change trend curves are highly similar in shape (i.e., rising, falling, or remaining stable simultaneously), but there is a generally stable proportional relationship or fixed offset in numerical magnitude, this strongly suggests a systematic scale factor error or zero-point drift in the mapping from acoustic feature changes to absolute settlement in the settlement inversion model. This error may stem from the difference between the training data and the actual sound velocity on site, or from the model's own estimation bias of the absolute magnitude. In this case, a linear calibration method can be used for overall correction. For example, the average ratio of crest settlement to underwater settlement gradient in all associated pairs can be calculated as a global proportional correction coefficient, or the average difference between the two can be calculated as a global offset. This coefficient or offset is then applied to all underwater acoustic inversion results to perform overall absolute value calibration. The technical effect of this operation is direct and effective. It cleverly "transfers" the high absolute accuracy of the GNSS system on the levee to the underwater acoustic inversion results, enabling the latter to maintain its high spatial resolution advantage while possessing reliable absolute quantitative accuracy.

[0104] S8.2 is used to handle local anomalies. If the analysis reveals that the trend is consistent across most areas, but a significant and persistent deviation exists between the underwater settlement trend and the crest settlement trend in a specific local area (e.g., a section of the levee or near a structure), this is often a signal requiring vigilance and in-depth analysis. Such local deviations can be caused by various complex factors: for example, the underwater foundation may have special geological conditions (such as weak interlayers) causing its deformation to be not completely synchronized with the deformation of the superstructure; or the acoustic propagation path in this area may be abnormally disturbed by newly formed obstacles (such as construction residues or localized siltation) or localized structural damage (such as cracks or cavities), causing the acoustic response to deviate from the model's predictions based on the healthy structure assumption. In this case, simple global linear calibration cannot solve the problem and may even distort the true information of the area. A more reasonable approach is to perform targeted optimization of the hydrophone nodes and their acoustic propagation paths involved in the local anomaly area within the framework of the settlement inversion model. For example, a path-specific correction coefficient or different weighting factors can be introduced into the model's inversion algorithm for the acoustic characteristic parameters (time delay offset, energy change) from node pairs in these anomalous regions; or, environmental auxiliary variables reflecting the specific characteristics of the region can be added to the model's input layer. Through this local parameter optimization, the model can adapt to the complex and non-uniform conditions actually occurring on-site, thereby generating more reasonable inversion results for that region. The final calibrated settlement monitoring report will include a calibrated and optimized high-precision settlement distribution map, data consistency analysis charts, special explanations of local anomalous areas, and potential risk warnings. This report integrates multiple pieces of information, including direct measurement and indirect inversion, overall trends and local details, providing the project operator with a comprehensive, reliable, and directly applicable high-value technical result for safety assessment and maintenance decisions.

[0105] Example 2

[0106] like Figure 2 As shown, in a second aspect, the present invention provides a deep-water breakwater foundation settlement monitoring system. The system employs a deep-water breakwater foundation settlement monitoring method provided in any of the above embodiments. The system includes:

[0107] The underwater sensor network deployment module is used to perform step S1: Based on the structural design drawing of the breakwater, multiple hydrophone nodes are fixedly installed in at least two monitoring areas of the underwater foundation structure of the breakwater, and all hydrophone nodes are connected to the monitoring host to generate an underwater acoustic monitoring network.

[0108] The synchronous acoustic data acquisition module is used to execute step S2: control the hydrophone nodes in the underwater acoustic monitoring network to synchronously record the sound pressure fluctuation signals from underwater within the set acquisition period, and obtain multi-channel sound pressure signal data;

[0109] The cross-correlation interference calculation module is used to perform step S3: based on multi-channel sound pressure signal data, perform pairwise cross-correlation calculations on the signals from different hydrophone node channels in the underwater acoustic monitoring network to obtain a noise interference spectrum characterizing the acoustic correlation between each node;

[0110] The acoustic feature extraction module is used to perform step S4: identifying and measuring the time delay and energy values ​​corresponding to the stable cross-correlation peaks in the noise interferogram, and generating cross-correlation function feature parameters; wherein, S4 includes:

[0111] S4.1: Based on the noise interferogram, perform peak detection on each cross-correlation function waveform to identify at least one major correlation peak in the waveform;

[0112] S4.2: Measure the lateral distance of the peak point of the main correlation peak relative to the zero time axis to obtain the time delay value;

[0113] S4.3: Calculate the signal energy integral of the waveform of the main correlation peak within a local time window near the time delay value to obtain the energy value;

[0114] S4.4: Based on the time delay value and energy value, combine and record the nodes to generate cross-correlation function characteristic parameters;

[0115] The settlement information inversion module is used to execute step S5: based on the characteristic parameters of the cross-correlation function, the pre-trained settlement inversion model is used to calculate and generate the settlement distribution information of the breakwater foundation.

[0116] This system corresponds to the method provided in Example 1, and will not be described in detail here.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the settlement of deep-water breakwater foundations, characterized in that, include: S1: Based on the structural design drawing of the breakwater, multiple hydrophone nodes are fixedly installed in at least two monitoring areas of the underwater foundation structure of the breakwater, and all hydrophone nodes are connected to the monitoring host to generate an underwater acoustic monitoring network. S2: Control the hydrophone nodes in the underwater acoustic monitoring network to synchronously record the sound pressure fluctuation signals from underwater within the set acquisition period, and obtain multi-channel sound pressure signal data; S3: Based on multi-channel sound pressure signal data, pairwise cross-correlation calculations are performed on signals from different hydrophone node channels in the underwater acoustic monitoring network to obtain a noise interferogram characterizing the acoustic correlation between nodes; S4: Identify and measure the time delay and energy values ​​corresponding to the stable cross-correlation peaks in the noise interferogram, and generate the characteristic parameters of the cross-correlation function; wherein, S4 includes: S4.1: Based on the noise interferogram, perform peak detection on each cross-correlation function waveform to identify at least one major correlation peak in the waveform; S4.2: Measure the lateral distance of the peak point of the main correlation peak relative to the zero time axis to obtain the time delay value; S4.3: Calculate the signal energy integral of the waveform of the main correlation peak within a local time window near the time delay value to obtain the energy value; S4.4: Based on the time delay value and energy value, combine and record the nodes to generate cross-correlation function characteristic parameters; S5: Based on the characteristic parameters of the cross-correlation function, the settlement distribution information of the breakwater foundation is generated by using a pre-trained settlement inversion model.

2. The method for monitoring the settlement of deep-water breakwater foundations according to claim 1, characterized in that, S1 includes: S1.1: Based on the cross-sectional design of the breakwater, multiple deployment locations are selected on the underwater foundation surface on the seaward side, and hydrophone nodes are installed to generate the first hydrophone node group. S1.2: Based on the cross-sectional design of the breakwater, multiple deployment locations are selected on the underwater foundation surface on the seaward side, and hydrophone nodes are installed to generate a second hydrophone node group. S1.3: Based on the first hydrophone node group and the second hydrophone node group, all hydrophone nodes are connected to the same monitoring host through data transmission cables to generate an underwater acoustic monitoring network.

3. The method for monitoring the settlement of deep-water breakwater foundation according to claim 2, characterized in that, S1.1 includes: S1.1a: Based on the length direction of the breakwater, determine multiple coordinates of the underwater foundation surface on the seaward side at uniform intervals, and generate node layout points; S1.1b: Based on the node layout points, the hydrophone nodes are fixed to the surface of the underwater riprap foundation or pile foundation through the mounting base; S1.1c: Adjust the orientation of the fixed hydrophone nodes so that the sensing surface faces the open water area, and generate the first hydrophone node group.

4. The method for monitoring the settlement of deep-water breakwater foundation according to claim 1, characterized in that, S2 include: S2.1: Based on a unified timing signal, control all hydrophone nodes in the underwater acoustic monitoring network to start synchronously and record the simulated sound pressure signal at a set sampling frequency to generate the original sound pressure time domain signal; S2.2: Based on the original sound pressure time domain signal, a bandpass filter is used to filter out components outside the set frequency band, and the signal amplitude is normalized and adjusted to generate a pre-processed sound pressure signal; S2.3: Based on the preprocessed sound pressure signal, encapsulate it according to channel and time sequence to generate multi-channel sound pressure signal data.

5. The method for monitoring the settlement of deep-water breakwater foundations according to claim 4, characterized in that, S2.2 includes: S2.2a: Based on the dimensions and material properties of the underwater structure of the breakwater, calculate the range of acoustic wavelengths that are sensitive to small displacements of the structure, and determine the sensitive acoustic frequency bands; S2.2b: Based on the sensitive acoustic frequency band, set the upper and lower passband frequencies of the bandpass filter; S2.2c: Based on the upper and lower frequency limits, the original sound pressure time-domain signal is digitally filtered to generate a pre-processed sound pressure signal.

6. The method for monitoring the settlement of deep-water breakwater foundations according to claim 1, characterized in that, Following S4.4, it also includes: S4.5: Based on the cross-correlation function characteristic parameters generated in the current measurement cycle and the cross-correlation function characteristic parameters stored in the historical measurement cycles, the time delay value and energy value are calculated by subtraction to obtain the time delay offset and energy change. In S5, the characteristic parameters of the cross-correlation function calculated by inputting the settlement inversion model are the time delay offset and the energy change.

7. The method for monitoring the settlement of deep-water breakwater foundation according to claim 1, characterized in that, The settlement inversion model is pre-trained through the following steps: Based on the design drawings, material parameters, and sound velocity profile of the breakwater, a numerical model including structural geometry and acoustic propagation characteristics is established. In the numerical model, a virtual receiving point is set up that corresponds to the location of the underwater acoustic monitoring network, and a simulated marine environmental noise source is loaded to generate a simulated sound field. Based on the simulated sound field, the acoustic cross-correlation function between each virtual receiving point is calculated under different preset settlement conditions, and a simulated noise interferogram set is generated. Based on the simulated noise interferogram set, simulated characteristic parameters under different working conditions are extracted, and a database of mapping relationship between settlement displacement field and changes in simulated characteristic parameters is constructed. Based on the mapping relationship database, the selected machine learning model is trained under supervision to obtain the settlement inversion model.

8. The method for monitoring the settlement of deep-water breakwater foundations according to claim 1, characterized in that, Following S5, it also includes: S6: Acquire displacement monitoring data from the displacement monitoring system deployed on the top of the breakwater during the same time period; S7: Spatially match the settlement distribution information with the displacement monitoring data at the horizontal projection position, and compare the consistency between the vertical settlement trend and the embankment top displacement trend to generate data comparison results; S8: Based on the data comparison results, the output values ​​of the settlement inversion model are proportionally calibrated or the local inversion path parameters are optimized to generate a calibrated settlement monitoring report.

9. A method for monitoring the settlement of deep-water breakwater foundations according to claim 8, characterized in that, S7 includes: S7.1: Calculate the settlement of each monitoring point on the top of the embankment based on the three-dimensional coordinate data provided by the displacement monitoring system; S7.2: Based on the horizontal coordinates of the underwater monitoring points in the settlement distribution information, project and associate them with the nearest monitoring point on the top of the dike; S7.3: Based on the data after projection correlation, analyze the numerical relationship and trend of underwater settlement gradient and levee crest settlement at the same vertical location; S8 includes: S8.1: Based on the analysis results of S7.3, if the trend of change is consistent, the absolute value of the settlement obtained by underwater acoustic inversion is linearly calibrated using the settlement of the dike crest. S8.2: Based on the analysis results of S7.3, if there is a deviation in the local trend of change, a correction coefficient for the acoustic propagation path of the corresponding monitoring area will be added to the settlement inversion model to generate a calibrated settlement monitoring report.

10. A deep-water breakwater foundation settlement monitoring system, characterized in that, The system employs a deep-water breakwater foundation settlement monitoring method as described in any one of claims 1 to 9, and the system comprises: The underwater sensor network deployment module is used to perform step S1: Based on the structural design drawing of the breakwater, multiple hydrophone nodes are fixedly installed in at least two monitoring areas of the underwater foundation structure of the breakwater, and all hydrophone nodes are connected to the monitoring host to generate an underwater acoustic monitoring network. The synchronous acoustic data acquisition module is used to execute step S2: control the hydrophone nodes in the underwater acoustic monitoring network to synchronously record the sound pressure fluctuation signals from underwater within the set acquisition period, and obtain multi-channel sound pressure signal data; The cross-correlation interference calculation module is used to perform step S3: based on multi-channel sound pressure signal data, perform pairwise cross-correlation calculations on the signals from different hydrophone node channels in the underwater acoustic monitoring network to obtain a noise interference spectrum characterizing the acoustic correlation between each node; The acoustic feature extraction module is used to perform step S4: identifying and measuring the time delay and energy values ​​corresponding to the stable cross-correlation peaks in the noise interferogram, and generating cross-correlation function feature parameters; wherein, S4 includes: S4.1: Based on the noise interferogram, perform peak detection on each cross-correlation function waveform to identify at least one major correlation peak in the waveform; S4.2: Measure the lateral distance of the peak point of the main correlation peak relative to the zero time axis to obtain the time delay value; S4.3: Calculate the signal energy integral of the waveform of the main correlation peak within a local time window near the time delay value to obtain the energy value; S4.4: Based on the time delay value and energy value, combine and record the nodes to generate cross-correlation function characteristic parameters; The settlement information inversion module is used to execute step S5: based on the characteristic parameters of the cross-correlation function, the pre-trained settlement inversion model is used to calculate and generate the settlement distribution information of the breakwater foundation.

Images