Non-optical identification and 3D reconstruction method of moving bed scour pits in hypergravity field
By using a multi-source non-optical signal recognition method, combined with various sensors and signal factors, the problem of unstable identification of the bed surface interface in the moving bed scour test in the hypergravity field was solved, and the three-dimensional reconstruction of the scour pit and the output of evolution index with high spatiotemporal resolution were realized.
Patent Information
- Application Number
- CN202611306701.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-26
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2046-08-26
AI Technical Summary
Existing technologies struggle to reliably identify the effective physical interface between the water body and the actual sediment bed in dynamic bed scour tests in hypergravity fields, resulting in numerous artifacts and temporal discontinuities in the scour elevation data, which fails to meet the requirements for high spatiotemporal resolution quantitative analysis.
A multi-source non-optical signal identification method is adopted, including acoustic echo, electrical sensing and pore water pressure sensor. The bed elevation is pre-screened and evaluated by combining signal quality factor, interface change intensity, spatial continuity and temporal continuity factor. The false echo caused by suspended particle layer and bubble is removed by a cross-modal identification mechanism that couples electrical concentration inversion and acoustic attenuation compensation.
It effectively reduces the probability of bed surface misjudgment in extremely non-transparent multiphase flow environments, achieves stable identification and three-dimensional reconstruction of scour pits, and outputs accurate scour pit evolution indicators.
Smart Images

Figure CN122817983B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of physical simulation and testing technology in geotechnical and marine engineering, and in particular to a non-optical method for identifying and reconstructing three-dimensional scour pits in a hypergravity field. Background Technology
[0002] Localized scouring caused by waves, currents, and their coupling effects can lead to reduced foundation depth and deterioration of bearing conditions in underwater structures. Using a geocentrifuge to reproduce the prototype's self-weight stress state in a hypergravity field and conducting wave-current coupled scouring centrifuge model tests is of technical research value for accurately revealing the sediment initiation law, transport mechanism, and bed evolution process. Obtaining continuous dynamic changes in bed morphology during scouring provides a data support foundation for assessing the physical boundary conditions and long-term service safety of structural engineering projects.
[0003] Existing wave-current scour tests primarily rely on optical methods such as transparent model box window imaging, PIV (particle image velocimetry) technology, or laser ranging to acquire data on the evolution of the bed surface morphology. However, in actual high-intensity hypergravity field moving bed scour tests, hydrodynamic forces cause a large amount of sediment to be suspended, forming a spatially unevenly distributed high-concentration turbid water layer, often accompanied by the generation of localized water bubbles. Optical signals experience severe scattering and absorption in the complex turbid water medium, making it difficult to penetrate the suspended water body and reach the actual dense bed surface at the bottom. Simultaneously, some conventional single-point acoustic ranging or contact measuring instruments are easily affected by the interfaces of suspended particles and turbulent disturbances when facing complex transitional media of water, sand, and air mixtures and strong hydrodynamic flow field disturbances. This leads to divergent echo signals, frequent multi-peak artifacts, or baseline drift in the measurement.
[0004] In summary, existing observation methods are insufficient to consistently and stably lock the effective physical interface between the water body and the actual sediment bed in extremely opaque test environments with turbid water, high suspended sediment, and strong disturbances. This results in a large number of artifacts and time discontinuities in the acquired scour elevation data, which cannot meet the needs for high spatiotemporal resolution quantitative analysis of the scour evolution process.
[0005] Therefore, there is an urgent need to study a monitoring and reconstruction method that can effectively overcome the obstruction of turbid water and strong hydrodynamic interference, and improve the stability and continuity of bed interface identification in complex multiphase flow media. Summary of the Invention
[0006] The purpose of this invention is to provide a non-optical method for identifying and reconstructing three-dimensional scour pits in a moving bed in a hypergravity field, in order to solve the aforementioned problems in the prior art.
[0007] Technical solution: A non-optical method for identifying and reconstructing three-dimensional scour pits in a moving bed in a hypergravity field, comprising: Multiple non-optical signals were acquired during the dynamic bed scouring model test in a hypergravity field, and multiple candidate bed surface elevations were extracted based on the multiple non-optical signals. Each of the candidate bed surface elevations is pre-screened and evaluated to obtain the confidence level of each candidate bed surface elevation; Based on the pre-screened candidate bed surface elevations and their corresponding confidence levels, state fusion is performed to obtain the estimated continuous bed surface elevation values for preset measuring points. Based on the estimated elevation values of the continuous bed surface at each of the preset measuring points, a three-dimensional bed surface is reconstructed, and the scour pit evolution index is output.
[0008] The candidate bed elevations are pre-screened and evaluated, and the evaluation process specifically includes: Obtain the near-bottom hydrodynamic intensity within the scour test section, and dynamically update the physical constraint factor for bed evolution based on the near-bottom hydrodynamic intensity; The confidence level is determined by weighting the following factors, including: Signal quality factor, used to characterize the stability of sensor response; Interface mutation intensity factor is used to characterize the degree of mutation of physical quantities at the candidate bed surface elevation after pre-screening; The spatial continuity factor is used to characterize the coordination between the current measuring point and adjacent high-confidence measuring points; The time continuity factor is used to characterize the continuity of the elevation change magnitude between the current moment and the previous moment; The physical constraint factor for bed evolution is determined based on preset or dynamically updated physical constraints for bed evolution.
[0009] Beneficial effects: Addressing the physical challenges of high-sediment-content turbid water and free bubbles causing numerous scattering artifacts in acoustic signals and making it difficult to pinpoint the true bed interface, a cross-modal identification mechanism coupling electrical concentration inversion and acoustic attenuation compensation was introduced. By calculating the expected echo intensity under the dense bed assumption and combining it with measured ratios for classification, false echoes caused by suspended particle layers and bubbles were eliminated, reducing the probability of bed misjudgment in extremely opaque multiphase flow environments. Attached Figure Description
[0010] Figure 1 This is a schematic diagram of the overall process of the non-optical identification and three-dimensional reconstruction method for the moving bed scour pit in the hypergravity field of the present invention.
[0011] Figure 2 A flowchart for correcting multi-source non-optical signals that include acoustic echo signals.
[0012] Figure 3 This is a flowchart for extracting candidate bed surface elevations from multi-source non-optical signals.
[0013] Figure 4A flowchart for evaluating the confidence level of bed surface elevation based on physical constraints of bed surface evolution. Detailed Implementation
[0014] Example 1: Before describing the non-optical identification and three-dimensional reconstruction method of the moving bed scour pit in the hypergravity field, this example mainly describes the construction of the hardware environment of the experimental system, the configuration of multi-source non-optical sensors, and the pre-configuration process of offline calibration parameters.
[0015] Step 101: Construct the hardware system for the dynamic bed scouring test under the ultragravity environment of the geocentrifuge.
[0016] Specifically, this step requires establishing a physical platform for model testing suitable for hypergravity environments. This platform utilizes a geotechnical centrifuge, with a sealed model box installed inside the centrifuge basket. The model box contains a scour test section, used to fill the sediment material to be tested and to install a hydraulic structure model. Wave-generating components and flow-generating devices are installed around or inside the model box to generate wave and water flow coupling within the scour test section.
[0017] To accurately describe the bed evolution process, a three-dimensional coordinate system was pre-established for the model box, with the main direction of water flow set as the x-axis, the width of the model box set as the y-axis, and the vertical elevation set as the z-axis. Before the experiment started, a sediment bed was prepared in the test section using layered sand laying or static water pre-consolidation methods, and the initial bed elevation was recorded.
[0018] Step 102: Arrange a multi-source non-optical sensor array within the scouring test section.
[0019] Because the high-gravity moving bed scouring process inevitably involves turbid water with high sand content and air bubbles blocking the view, conventional optical observation methods often fail.
[0020] Therefore, this step involves deploying a multi-source non-optical sensing network within the scouring test section. These multi-source non-optical sensors include, but are not limited to, acoustic echo sensors, electrical sensing sensors, micropore water pressure sensors, and near-bottom pressure pulsation sensors.
[0021] Regarding the selection and placement of acoustic echo sensors, in order to balance the penetration capability of high-concentration suspended sediment with vertical recognition resolution, a miniature ultrasonic probe with an operating frequency of 1MHz to 2MHz is preferred.
[0022] For example, when the probe chip diameter is 3 to 5 millimeters, the vertical resolution corresponding to 1 MHz is approximately 0.75 millimeters. Simultaneously, the acoustic sensor's installation position must ensure that its minimum distance from the bed surface is greater than the far-field condition limit distance to avoid interference from near-field acoustic artifacts.
[0023] The electrical sensing sensors are arranged in a vertical array to capture abrupt changes in conductivity or impedance from the water body, suspended layer to the dense bed surface. Pore water pressure sensors are buried at different depths inside the bed surface, while near-bottom pressure pulsation sensors are arranged in the area above the near-bottom.
[0024] Step 103: Perform offline calibration of multi-source signal references and construction of electrical mapping functions under static water conditions.
[0025] In the still water or low-flow-rate clear water stage before the wave-generating and flow-generating devices are activated, it is necessary to collect the initial responses of various sensors as the reference zero point for subsequent dynamic calibration and drift correction.
[0026] Specifically, the initial bed surface echo flight time of the acoustic echo sensor is obtained, the initial conductivity or capacitance baseline of the electrical sensing sensor under clear water conditions is recorded, and the initial hydrostatic pressure distribution of each pore pressure sensor is recorded.
[0027] This step requires establishing a quantitative mapping function between electrical response characteristics and sediment concentration in the water body through offline mixing experiments. In a still water calibration container, turbid water samples with different known sediment concentrations are prepared, and the electrical response values corresponding to each concentration sample are measured. The mapping function f_cal between conductivity and sediment concentration is then fitted. This serves as the prior mathematical basis for subsequently inverting the vertical sediment concentration profile during dynamic turbid water flushing, providing input for correcting sound wave propagation attenuation.
[0028] Example 2: The overall process of the non-optical identification and three-dimensional reconstruction method of the moving bed scour pit in the hypergravity field is further explained.
[0029] like Figure 1 As shown, in one possible implementation, the following steps are included: Step 201: Acquire multi-source non-optical signals collected in the hypergravity field moving bed scour model test, and extract multiple candidate bed surface elevations based on the multi-source non-optical signals; During the operation of the geocentrifuge and the application of wave and water flow coupled loading, the acquisition system synchronously receives data streams fed back by the sensor array arranged in the scour test section, obtaining multi-source non-optical signals. The multi-source non-optical signals include acoustic echo sequences, baseline changes in conductivity or capacitance, and time history data of pore water pressure and near-bottom pressure pulsations.
[0030] In environments with turbid water and high-sediment suspended layers, the response of a single sensor often contains multiple reflection artifacts or is affected by changes in medium concentration. Therefore, possible bed locations are extracted based on the signal characteristics of each physical mechanism, forming multiple candidate bed elevations.
[0031] In this process, acoustic candidate elevations are extracted from locations of abrupt changes in acoustic reflection impedance, electrical candidate elevations are extracted from locations of abrupt changes in dielectric properties or conductivity gradients, and pore pressure candidate elevations are extracted by combining pore pressure phase hysteresis characteristics. Through parallel extraction processing, a set of elevation candidates based on different physical backgrounds is provided for each spatial location.
[0032] In some alternative implementations, before extracting multiple candidate bed elevations, the multi-source non-optical signals are pre-defined with spatiotemporal reference unification and parameter correction based on hypergravity conditions, which reduces mechanical vibration noise caused by centrifuge operation and acoustic wave refraction shift caused by medium concentration gradient.
[0033] Step 202, pre-screening and evaluating each of the candidate bed elevations, including: calculating the cross-sensor physical consistency score of each of the candidate bed elevations, and obtaining the pre-screened candidate bed elevations and the confidence level of each of the pre-screened candidate bed elevations; After obtaining the candidate elevations output from each physical channel, their physical authenticity and data reliability need to be quantitatively assessed. Cross-sensor consistency verification is performed to eliminate isolated abnormal elevation points that only have a single sensor response and no other physical quantity verification.
[0034] Candidate elevations retained after pre-screening are entered into the evaluation model. The evaluation process comprehensively considers the local signal-to-noise ratio of the sensor echo signal, the abrupt gradient of the medium physical quantity at the candidate location, and the continuous and smooth performance of the candidate point in time and space.
[0035] Simultaneously, hydrodynamic boundary conditions are introduced to constrain the rationality of elevation changes. The aforementioned multi-dimensional evaluation indicators are calculated according to the set weight allocation logic to obtain the corresponding confidence levels, which are used to reflect the probability that the corresponding candidate bed surface elevation represents the true dense bed surface interface.
[0036] In some alternative implementations, when dealing with specific test soil materials, the confidence level calculation results can be optimized by adjusting the weight configuration in the evaluation system. For cohesive soil beds prone to surface softening and flocculent suspension, the weight ratios of the electrical gradient factor and the pore pressure cross-boundary response factor can be appropriately increased, while the influence of the acoustic echo factor can be reduced.
[0037] Step 203: Based on the elevation of each pre-screened candidate bed surface and the corresponding confidence level, perform state fusion to obtain the estimated value of the continuous bed surface elevation of the preset measuring points; This step transforms discrete, uncertain, transient candidate data into a smooth, continuous state that conforms to the laws of physical evolution over time. The preset measurement points correspond to spatial grid nodes divided in the model box coordinate system or normal mapping points of each sensor.
[0038] A recursive state-space model is established, using the elevations and rates of change of the preset measuring points as dynamic state variables. Within each data sampling period, the candidate bed elevations are used as observation inputs, and the confidence level is converted into the observation noise covariance weight in the state update equation. Candidate elevations with higher confidence levels are assigned higher confidence weights in the fusion update, guiding the state variables to converge towards that candidate value. Through continuous prediction and iterative updates, a continuous bed elevation estimate with uncertainty assessment is output.
[0039] Step 204: Based on the estimated elevation values of the continuous bed surface at each of the preset measuring points, perform three-dimensional bed surface reconstruction and output scour pit evolution index.
[0040] After obtaining the estimated elevation values of the continuous bed surface at all preset measuring points on the same time section, these values are mapped to a three-dimensional spatial coordinate system. A spatial interpolation algorithm is then used to fit the discrete point set into a continuous bed surface model.
[0041] Using the initial calibrated bed elevation as the zero-point reference, the scour depth difference at each coordinate position of the current reconstructed surface is calculated. Regions that meet the preset depth judgment conditions are extracted, and the evolution index of scour pits is statistically analyzed and output.
[0042] The scour pit evolution indicators specifically cover the maximum scour depth, local scour area, scour pit erosion volume, and the spatial evolution trajectory of the scour center.
[0043] In some alternative implementations, to avoid excessive smoothing of structure edges by spatial interpolation algorithms, a geometric high-gradient boundary preservation mechanism is incorporated during the 3D bed reconstruction process. By monitoring the local slope of the reconstructed mesh, isolation operations are performed at steep slope boundaries where slope changes abruptly, thus preserving the true morphological features of the scour pit sidewalls.
[0044] Example 3: Further explanation of the unification of the spatiotemporal reference of the signal and the parameter correction process based on hypergravity conditions.
[0045] In one possible implementation, the following steps are included: Step 301: Divide the sound wave propagation path into multiple thin layers, determine the sound velocity within each layer based on the sand concentration and temperature of each layer obtained by real-time inversion or pre-measurement, and correct the propagation direction and propagation time of the acoustic signal in the multi-source non-optical signal by combining the law of refraction, so as to obtain the corrected multi-source non-optical signal. Accordingly, the step of extracting multiple candidate bed surface elevations based on the multi-source non-optical signals specifically involves: extracting multiple candidate bed surface elevations based on the modified multi-source non-optical signals.
[0046] In a turbid water environment scoured by a moving bed in a hypergravity field, the concentration of suspended sediment is usually non-uniformly distributed vertically, exhibiting a vertical concentration gradient. This non-uniform distribution causes spatial variations in the acoustic properties of the water medium, making the propagation speed of sound no longer constant, and the propagation path also bends due to refraction.
[0047] Using the assumption of a single, uniform sound velocity propagating in a straight line will lead to systematic deviations in acoustic ranging in deep water or high-concentration areas. Therefore, it is necessary to introduce a layered medium propagation model for operational condition correction.
[0048] Specifically, the water depth is divided into a predetermined number of horizontal thin layers along the measurement direction of the acoustic sensor.
[0049] Local temperature data is acquired in real time by temperature sensors arranged in the scour test section, and the local volumetric sand concentration data is obtained by inversion of electrical response signals or by estimation from a pre-calibrated vertical concentration distribution model.
[0050] Local temperature data and local volumetric sand concentration data are acquired within each thin layer. The local volumetric sand concentration data can be initially calculated from the electrical response signals collected in real time by vertically distributed electrical sensing sensors using a calibration mapping function.
[0051] In practice, an electrical sensing sensor is used to obtain a preliminary vertical distribution of sand concentration, and then the acoustic signal is corrected for layered propagation paths based on this distribution. The local sound velocity within the corresponding thin layer is calculated based on a pre-stored empirical sound velocity model or the fluid acoustic state equation.
[0052] Since sound waves are deflected when passing through the interfaces of medium layers with different sound velocities, the law of refraction is introduced to calculate the actual propagation angle of the sound wave within each thin layer. This propagation angle is recursively obtained from the ratio of the initial incident angle to the sound velocity in each layer. Based on the thickness of each layer, the local sound velocity, and the actual propagation angle, the flight time of the sound wave through each thin layer is calculated.
[0053] The relevant propagation time calculation logic is expressed by the following formula: dt _k =dh _k / (c _k *cos(θ _k )); Among them, dt _k Let dh be the propagation time of the sound wave in the kth thin layer. _k Let c be the thickness of the k-th thin layer. _k Let θ be the sound velocity within the k-th thin layer determined based on sand concentration and temperature. _k The propagation angle of the sound wave in the kth thin layer is derived by combining the law of refraction.
[0054] Based on this, the propagation times of all thin layers are summed to obtain the total one-way propagation time of the sound wave and the corresponding actual vertical propagation distance. Through the above layered calculation and accumulation, the time reference and spatial coordinate reference of the original acoustic echo signal are remapped to eliminate the refraction error and travel time error caused by the sound velocity gradient, resulting in a corrected multi-source non-optical signal. Subsequent candidate elevation extraction is based on this corrected multi-source non-optical signal after eliminating systematic errors.
[0055] In some optional implementations, a judgment mechanism is set to optimize the allocation of computational resources for cases where the vertical variation in sediment concentration is small. The overall vertical sediment concentration gradient within the scour test section is calculated, and it is determined whether the concentration change amplitude corresponding to this gradient exceeds a preset concentration change judgment threshold (e.g., 20% of the initial calibration value). If the concentration change amplitude does not exceed the judgment threshold, an equivalent uniform sound velocity linear propagation model is used instead of the aforementioned layered refraction model, and the propagation time is directly converted using the calibrated sound velocity in a single operation. This conditional execution branch reduces the computational overhead of real-time signal correction while ensuring spatial positioning accuracy.
[0056] Example 4: Further explanation of the specific extraction process for vertical sand concentration profile based on electrical response inversion and classification of physical properties of driving acoustic candidate peaks.
[0057] like Figure 2 As shown, in one possible implementation, the following steps are included: Step 401: Based on the pre-calibrated mapping relationship between electrical characteristics and sediment concentration, the electrical response signal is converted into a vertical sediment concentration profile; The corrected multi-source non-optical signal acquired at this time includes an electrical response signal vertically distributed along the water depth direction. Considering the baseline drift of the sensor caused by water temperature changes and centrifuge operation, the ratio of the electrical response signal at each height at the current moment to the initial calibration response is calculated to obtain the relative electrical response that eliminates the initial baseline drift.
[0058] Extract the mapping function between conductivity or capacitance and sediment concentration established offline in the early stage. Substitute the relative electrical response at each height into the mapping function for calculation to obtain the local volumetric sediment concentration data of the water body at each height interval, and convert the electrical response signal into a vertical sediment concentration profile.
[0059] The vertical sediment concentration profile reflects the spatial distribution characteristics of sediment concentration from the clear water layer, suspended sediment layer, moving particle layer to the compacted bed surface.
[0060] Step 402: Identify the locations of abrupt changes in the vertical gradient in the vertical sediment concentration profile, and extract the electrical candidate bed elevation as one of the candidate bed elevations; After obtaining the vertical sediment concentration profile, the ratio of the difference in sediment concentration between adjacent height intervals to the height difference is calculated to obtain the vertical sediment concentration gradient at each location.
[0061] Based on this, it is determined whether the vertical sediment concentration gradient at each location is greater than a preset gradient abrupt change threshold. When the vertical sediment concentration gradient in a certain height range is greater than this abrupt change threshold, it is identified as a transition point where the medium transitions from a water or suspended state to a compacted state. The spatial coordinate elevation corresponding to this abrupt change point is extracted to obtain the electrical candidate bed elevation, which is then used as one of the candidate bed elevations in the subsequent cross-sensor pre-screening.
[0062] Step 403: Input the vertical sand concentration profile into the acoustic attenuation model to calculate the desired echo intensity at the reflection interface; The corrected multi-source non-optical signal also includes acoustic echo signals acquired by an acoustic echo sensor. In sediment-laden water bodies, the energy attenuation of sound waves is determined by both the viscous absorption of pure water and the geometric scattering of suspended particles.
[0063] Extract the vertical sediment concentration profile generated in the aforementioned steps, and input the profile into an acoustic attenuation model that includes viscous absorption factor and particle scattering factor to calculate the acoustic attenuation coefficient at each depth of the water body.
[0064] Assuming the target reflection interface is a real, dense bed surface, and combining the initial emission intensity of the sound source, the geometric divergence attenuation law of sound waves with propagation distance, and the acoustic attenuation coefficients of each layer, an exponential attenuation accumulation calculation is performed to obtain the expected echo intensity at the reflection interface. This expected echo intensity characterizes the theoretically correct energy reference value that a sound wave should return when it collides with a dense bed surface with a large abrupt change in acoustic impedance after passing through a water body with a specific sediment concentration distribution.
[0065] Wherein, the expected echo intensity I at a distance r from the sound source _exp =I _0 *R _bed *(1 / r 2 )*exp(-2*Σ(α _k *dh _k )); In the formula, I _0 R represents the initial emitted pulse intensity of the sound source. _bed Let α be the acoustic impedance reflection coefficient assuming the reflecting interface is a dense bed surface, r be the propagation distance from the transmitting end to the reflecting interface, 1 / r² represent the geometric diffusion attenuation of the sound wave, and α be the acoustic impedance reflection coefficient. _k Let dh be the acoustic attenuation coefficient for the k-th depth interval. _k Let be the thickness of the k-th depth interval.
[0066] Step 404: Calculate the ratio of the measured echo intensity to the expected echo intensity of the acoustic echo signal, and classify the physical properties of the echo peaks in the acoustic echo signal based on the ratio. After obtaining the expected echo intensity corresponding to the current propagation distance, a conventional bandpass filtering algorithm is used to remove high-frequency background noise from the acoustic echo signal, extract the measured echo peak, and obtain the measured echo intensity.
[0067] Calculate the ratio of the measured echo intensity to the expected echo intensity. Since the reflection impedance coefficients of a real dense bed surface and low-density suspended sediment masses differ by orders of magnitude, this ratio can directly map the physical mechanism properties of the echo peaks. Based on the numerical range within which this ratio falls, classify the physical properties of each echo peak in the acoustic echo signal.
[0068] In the first execution path, when the ratio is in the range of 0.5 to 2.0, it is determined that the measured echo intensity is consistent with the theoretical intensity expectation under the dense bed assumption, and the echo peak is classified as a high probability bed peak. In the second execution path, when the ratio is less than 0.5 and greater than or equal to 0.2, the measured echo energy is determined to be in the transition range between the bed surface and the suspension layer. The echo peak is classified as an uncertain peak and given a lower initial weight in the subsequent cross-sensor consistency evaluation. In the third execution path, when the ratio is less than 0.2, it is determined that the measured echo energy is weak and does not have the high reflection characteristics of a dense bed surface, and it is classified as a suspended layer peak caused by scattering of suspended layer particles. In the fourth execution path, when the ratio is greater than 2.0, it is determined that the measured echo energy has a gain that exceeds the theoretical expectation of the dense bed surface, and it is classified as an abnormal peak caused by free bubbles or interference from near-field structural components.
[0069] Through the above ratio discrimination mechanism, the acoustic artifact interference caused by high concentration of suspended sediment and free air bubbles can be removed at the level of acoustic impedance physical mechanism.
[0070] The above ratio range is set based on the magnitude of the difference in acoustic impedance between the dense bed surface and the suspended particle layer. Those skilled in the art can make adaptive adjustments according to the specific sediment type and sensor characteristics.
[0071] Step 405: The distance corresponding to the echo peak classified as a high-probability bed surface peak is converted into an acoustic candidate bed surface elevation, which is then used as one of the candidate bed surface elevations.
[0072] After completing the physical property classification, the data responses marked as suspended layer peaks and anomalous peaks are removed, while the uncertain peak echo data are retained. Their initial weights are then reduced in the subsequent confidence evaluation, and only the echo data classified as high-probability bed surface peaks are retained.
[0073] A time-series sliding window is used to perform temporal continuity state detection on the retained high-probability bed surface peaks, filtering out isolated noise peaks that flicker instantaneously and lack temporal correlation. The acoustic wave propagation distance corresponding to the high-probability bed surface peaks with temporal stability is converted into elevation values in the three-dimensional coordinate system of the model box by combining the pre-acquired sensor installation angle and coordinate offset, thus obtaining the acoustic candidate bed surface elevation.
[0074] The extracted result is output as one of the candidate bed surface elevations and incorporated into the multi-source candidate elevation set.
[0075] Example 5: A multiphysics-assisted candidate elevation extraction strategy is further described. As a supplement or alternative to acoustic and electrical feature extraction, this strategy provides redundant candidate data when the primary sensor is limited.
[0076] like Figure 3 As shown, in one possible implementation, the following steps are included: Step 501, the modified multi-source non-optical signal further includes at least one of pore water pressure signal, near-bottom pressure pulsation signal or flexible contact signal; The step of extracting multiple candidate bed surface elevations based on the corrected multi-source non-optical signals also includes at least one of the following operations: In the first execution path, the boundary location where the pore water pressure signal changes from a hysteretic response within the soil to a synchronous response in the water body is identified, and the candidate bed elevation of the pore pressure is extracted as one of the candidate bed elevations.
[0077] Specifically, the cross-boundary location refers to the spatial interface where the sensor's environment changes from a porous medium to a body of water. When a pore water pressure sensor embedded within the bed surface receives pressure changes transmitted by wave current loads, its response characteristics are related to the state of the surrounding medium.
[0078] Sensors located inside the bed surface are affected by the interaction between the sediment particle skeleton and pore fluid, resulting in amplitude attenuation and phase lag in their pore water pressure signals compared to the water pressure directly borne by the bed surface. As scouring continues, the thickness of the overlying sediment layer decreases.
[0079] When the bed elevation decreases to the depth of the sensor, the sensor is exposed to the water. At this point, the collected pore water pressure signal changes to a state of synchronous fluctuation with the overlying dynamic water pressure, with a decrease in amplitude attenuation and a reduction in phase lag approaching zero. By calculating the difference in pore pressure amplitude and phase lag between sensors at adjacent depths and comparing the changing trends of these parameters in the vertical spatial sequence, the depth coordinates at which the attenuation rate and phase lag experience a step decrease are determined. These depth coordinates are then extracted to obtain the candidate bed elevation for pore pressure.
[0080] In the second execution path, the location where the pressure pulsation energy in the near-bottom pressure pulsation signal changes abruptly is identified, and the pressure pulsation candidate bed surface elevation is extracted as one of the candidate bed surface elevations.
[0081] Specifically, when the bed morphology changes during scouring, the water flow through local scouring pits causes changes in the flow structure of the near-bottom boundary layer, leading to alterations in turbulent vortices and pressure pulsation characteristics. Near-bottom pressure pulsation signals collected by sensors placed in the near-bottom region are acquired, and the root mean square value of this signal within a preset time window and the integral value of pulsation energy in a specific high-frequency segment are calculated.
[0082] By comparing the pulsating energy integral values at different spatial measuring points, when a jump in the pulsating energy integral value corresponding to a certain spatial location is detected over time, or when the spatial gradient of pulsating energy between adjacent measuring points is greater than a preset energy gradient threshold, it is determined that a terrain change has occurred at that spatial location. The vertical coordinates corresponding to the spatial location where the energy change occurred are extracted to obtain the elevation of the pressure pulsation candidate bed surface.
[0083] In the third execution path, the location where the contact state changes abruptly in the flexible contact signal is identified, and the strain candidate bed surface elevation is extracted as one of the candidate bed surface elevations.
[0084] Specifically, flexible sensing elements are attached to the surface of a structure or inside a shallow bed. Before scouring occurs, the surface of the sensing element is covered with silt and sand, and is subjected to contact compression from the surrounding soil, placing the flexible contact signal in an initial compressed state.
[0085] When sediment particles are eroded and detached from the bed surface, the surface of the sensing element changes from a soil-covered state to a water-contact state. The contact compressive stress applied to it decreases, leading to a drop in the strain response level. The time derivative of the flexible contact signal is calculated. When the absolute value of the time derivative exceeds a preset response abrupt change threshold, it is determined that the sediment cover layer at that location has been removed. The installation elevation corresponding to this contact state abrupt change point is extracted to obtain the candidate bed surface elevation for strain.
[0086] In some optional implementations, the auxiliary candidate bed elevations extracted by the three execution paths described above can be invoked independently or in parallel combinations. When the acoustic echo signal experiences frame drops due to high-concentration bubble obstruction, or when the electrical response signal becomes saturated due to fluid salinity fluctuations, the system activates the aforementioned pore pressure, pressure pulsation, or strain identification paths based on pre-configured signal validity discrimination logic to maintain data flow in the candidate bed elevation extraction stage.
[0087] Example 6: Further explanation is given of the specific process of calculating the cross-sensor physical consistency score and pre-screening each candidate bed surface elevation based on the cross-sensor physical consistency score.
[0088] In one possible implementation, the following steps are included: Step 601: Count the number of heterogeneous sensor physical responses supported within a preset range of the candidate bed surface elevation, and determine the cross-sensor physical consistency score. In a turbid water environment scourned by a moving bed in a hypergravity field, a single sensor is easily affected by non-Gaussian noise, which can produce false signals.
[0089] To prevent false elevation points lacking multi-physical verification from entering the subsequent state fusion stage and causing estimation divergence, a physical consistency verification mechanism is constructed before fusion. Candidate bed elevations at the same spatial measurement point are acquired, and a preset range is set for each candidate bed elevation. This preset range is typically set to twice the vertical resolution of the acoustic sensors. All heterogeneous sensors deployed in the measurement point area are traversed, and it is determined whether each heterogeneous sensor has a physical response within the preset range that supports the candidate elevation.
[0090] The specific judgment logic is set as follows: For acoustic sensors, check whether there is echo data classified as high-probability bed surface peaks within the preset range; For electrical sensing sensors, check whether there are conductivity abrupt changes within the preset range that are higher than a predetermined abrupt change gradient; For pore pressure sensors, check whether the pore pressure sensor closest to the candidate bed elevation in terms of spatial distance exhibits cross-boundary characteristics between water and soil responses.
[0091] If a heterogeneous sensor meets the above criteria, it is determined that it provides physical response support, and its corresponding indicator function is set to 1; otherwise, its corresponding indicator function is set to 0.
[0092] Based on the indication function states of each heterogeneous sensor, a cross-sensor physical consistency score is obtained through numerical calculation, as follows: κ _i_j =(1 / N_ sensor )*Σ_s(δ _s ); In the formula, κ _i_j For the cross-sensor physical consistency score corresponding to the j-th candidate bed elevation at the i-th preset measuring point, N _sensor Σ_s represents the total number of heterogeneous sensors participating in the evaluation within the measurement point area, Σ_s represents the summation operator that accumulates the output values of each sensor, and δ_s represents the total number of heterogeneous sensors participating in the evaluation within the measurement point area. _s Provide a physical response indication function for the s-th heterogeneous sensor.
[0093] In some optional implementations, if a certain type of sensor has a detection blind zone in a specific area, such as a blank area for the pore pressure sensor on the back surface of a structure, then when calculating the cross-sensor physical consistency score, the sensor is removed from the evaluation system, and the total number of heterogeneous sensors participating in the evaluation is reduced simultaneously to ensure the spatial adaptability of the scoring logic.
[0094] Step 602: If the cross-sensor physical consistency score is lower than a preset threshold, the weight of the candidate bed surface elevation is reduced or the candidate bed surface elevation is removed to obtain the pre-screened candidate bed surface elevation.
[0095] After obtaining the cross-sensor physical consistency score, it is compared with a pre-set grading threshold, and corresponding preprocessing actions are performed based on the comparison result. The first preset threshold is set to 0.5, and the second preset threshold is set to 0.25. In the first execution path, when the cross-sensor physical consistency score is greater than or equal to the first preset threshold, it is determined that the candidate bed elevation is supported by the physical characteristics of more than half of the heterogeneous sensors, and the system retains the candidate bed elevation without performing weight attenuation.
[0096] In the second execution path, when the cross-sensor physical consistency score is greater than or equal to the second preset threshold and less than the first preset threshold, the physical support for the candidate bed elevation is determined to be weak. The system retains the candidate bed elevation but multiplies its initial confidence parameter by a decay factor. The decay factor is set as the ratio of the cross-sensor physical consistency score to the first preset threshold.
[0097] In the third execution path, when the cross-sensor physical consistency score is less than the second preset threshold, the candidate bed elevation is determined to be supported only by a single sensor or without any physical response support, belonging to isolated noise generated by free bubbles or local disturbances. The system rejects and blocks this candidate bed elevation, terminating its transmission to subsequent fusion stages. After the above judgment and blocking processing, the system outputs pre-screened candidate bed elevations after filtering outliers.
[0098] Example 7: This example further explains the construction process of the confidence evaluation framework and the dynamic physical constraints of the hydrodynamic bed evolution. Since there are parallel schemes for calculating near-bottom hydrodynamic intensity, this example will describe these two alternative acquisition paths respectively.
[0099] like Figure 4 As shown, in one possible implementation, the following steps are included: Step 701: Obtain the near-bottom hydrodynamic intensity within the scour test section.
[0100] In the first execution path, the near-bottom pressure pulsation signal in the scour test section is collected in real time; the root mean square value of the near-bottom pressure pulsation signal within a preset short time window is calculated, and the near-bottom hydrodynamic intensity is determined based on the ratio of the root mean square value to the comprehensive empirical coefficient. The near-bottom hydrodynamic intensity is used to evaluate the elevation of the pre-screened candidate bed surface.
[0101] Specifically, near-bottom pressure pulsation signals can reflect the turbulent disturbance energy near the bed surface and have a positive physical correlation with the fluid shear force acting on the surface of sediment particles. Pressure sensors are placed above the area of the bed surface to be measured to continuously receive high-frequency pressure response time-series data.
[0102] A preset short time window is selected, typically with a length of 5 to 20 main wave cycles. The root mean square (RMS) value of the pressure pulsation signal within this time window is calculated, representing the statistical average level of fluid pulsation energy over a specific time scale.
[0103] Based on this, a dimensionless comprehensive empirical coefficient is introduced, which is derived from previous theoretical calibration. By dividing the calculated root mean square value by this comprehensive empirical coefficient, the pressure pulsation energy can be mapped into an estimated value of the bed shear stress characterizing the near-bottom hydrodynamic intensity.
[0104] The relevant calculation logic is expressed by the following formula: τ _b (t)=P _rms / C2; Where, τ _b (t) represents the near-bottom hydrodynamic intensity (near-bottom shear stress) derived at the current moment, P _rms C1 represents the root mean square value of the near-bottom pressure pulsation signal within a preset short time window, and C2 is a comprehensive empirical coefficient. The comprehensive empirical coefficient C2 can be determined through a comparative calibration test under known bed shear stress conditions.
[0105] Specifically, under controlled conditions of clear water or known sediment concentration, the root mean square value of near-bottom pressure fluctuations and the bed shear stress value obtained using standard methods (such as logarithmic velocity profiling or direct shear force measurement) are simultaneously measured. The comprehensive empirical coefficient C2 is then obtained through regression analysis. Those skilled in the art can perform adaptive calibration based on specific sediment types and flow field conditions.
[0106] In the second execution path, the near-bottom pressure pulsation signal in the scouring test section is collected in real time; combined with a preset velocity scale, the near-bottom pressure pulsation signal is processed using the turbulent kinetic energy method to obtain the near-bottom hydrodynamic intensity, which is used to evaluate the elevation of the pre-screened candidate bed surface.
[0107] Specifically, this approach is based on the turbulent kinetic energy theory in fluid mechanics. After obtaining the root mean square value of the near-bottom pressure fluctuation signal, it is necessary to additionally obtain a preset velocity scale parameter that reflects the characteristics of the local flow field. The root mean square value of the pressure fluctuation, the fluid density, and the preset velocity scale are input into the turbulent kinetic energy estimation model for product calculation to obtain the estimated value of the near-bottom hydrodynamic intensity.
[0108] In one specific implementation, the formula for calculating turbulent kinetic energy is: τ _b (t)=C1*ρ*kt; Where, τ _b (t) represents the near-bottom hydrodynamic intensity, C1 is a dimensionless empirical coefficient related to the turbulent boundary layer structure, ρ is the fluid density, and kt is the turbulent kinetic energy estimated based on the root mean square value of near-bottom pressure fluctuations and a preset velocity scale. The empirical coefficient C1 can be determined through comparative calibration experiments, and those skilled in the art can select a specific calibration method according to the actual flow field conditions.
[0109] In one specific implementation, the preset velocity scale can be taken as the near-bottom mainstream average velocity or the wave trajectory velocity amplitude. The basic calculation idea of the turbulent kinetic energy method is to combine the root mean square value of the near-bottom pressure fluctuation with the fluid density and the preset velocity scale to obtain an estimated value of the near-bottom turbulent kinetic energy, and then convert it into near-bottom shear stress based on the turbulent boundary layer theory. The turbulent kinetic energy method is a well-known near-bottom hydrodynamic estimation method in the art, and those skilled in the art can choose the specific velocity scale calibration method according to the actual flow field conditions.
[0110] The elevation of each pre-screened candidate bed surface is evaluated in conjunction with the physical constraints of the bed surface evolution to obtain the corresponding confidence level, specifically including the following steps.
[0111] Step 702: Dynamically update the physical constraint factor of bed evolution based on the near-bottom hydrodynamic intensity.
[0112] The near-bottom hydrodynamic intensity is compared with the pre-calibrated critical starting hydrodynamic intensity for sediment. After obtaining the near-bottom hydrodynamic intensity at the current moment, it is compared with the pre-calibrated critical starting hydrodynamic intensity for sediment. Based on the comparison results, the dynamic maximum allowable scour rate and dynamic stable slope angle in the physical constraint factors of the bed surface evolution are dynamically updated. Wherein, when the near-bottom hydrodynamic intensity exceeds the critical starting hydrodynamic intensity of sediment, the dynamic maximum allowable scour rate is increased, and the natural angle of repose of sediment under still water conditions is reduced to decrease the dynamic stable slope angle. The dynamic maximum permissible scour rate and the dynamic stable slope angle are used to evaluate the pre-screened candidate bed elevation or to project the physical constraints onto the estimated continuous bed elevation.
[0113] Specifically, traditional bed surface recognition algorithms often use fixed rates or slope thresholds to eliminate outliers. During periods of strong scouring, these algorithms tend to misinterpret genuine elevation drops as noise and suppress them, while during periods of weak scouring, they easily misinterpret minute sensor fluctuations as genuine terrain changes. To address this issue, this embodiment introduces a dynamic constraint adaptive update logic based on hydrodynamic mechanisms.
[0114] Obtain the near-bottom hydrodynamic strength (i.e., near-bottom shear stress τ) _b (t) and the critical starting hydrodynamic intensity of the target sediment (i.e., critical shear stress τ) calibrated in advance by a flume test. _cr When τ _b (t) is greater than τ _cr At this time, the sediment particles are in an active, mobile state. At this time, calculate τ. _b (t) and τ _cr The ratio is subtracted by 1, and the resulting dimensionless excess dynamic parameter is exponentially calculated. Finally, it is multiplied by the reference scour rate to obtain the dynamic maximum permissible scour rate after relaxation at the current moment.
[0115] Specifically, v _max (t)=v _ref *(τ _b (t) / τ _cr -1) nv ; In the formula, v _max (t) represents the maximum permissible scouring rate after dynamic updates, v _ref τ is the preset reference flushing rate. _b (t) represents the near-bottom hydrodynamic intensity, τ _cr The critical starting hydrodynamic intensity for sediment is given by nv, which is a dimensionless velocity index. For sandy beds, the value of nv is usually set between 1.5 and 2.0.
[0116] Similarly, strong hydrodynamic forces not only accelerate the scouring rate but also weaken the anti-sliding stability of the scouring pit slope. Therefore, when τ _b When (t) is large, the natural angle of repose under still water conditions is numerically reduced using a formula that includes a hydrodynamic reduction factor, resulting in a dynamic stable slope angle that is smaller than the still water angle of repose.
[0117] In one example, the dynamic stability slope angle φ _max (t)=φ _0 *max(0,1-β*(τ _b (t) / τ _cr -1)); Where, φ _max (t) represents the dynamically updated stable slope angle, φ _0θ is the natural angle of repose of sediment under still water conditions, β is the hydrodynamic reduction factor, and max(0, ·) is the non-negative operation.
[0118] Step 703: Evaluate the elevation of each of the pre-screened candidate bed surfaces in conjunction with the physical constraints of bed surface evolution to obtain the corresponding confidence level. The confidence level is determined by weighting the signal quality factor, interface mutation intensity factor, spatial continuity factor, temporal continuity factor, and physical constraint factor of bed surface evolution.
[0119] The confidence level is determined by a weighted average of the signal quality factor, the interface mutation intensity factor, the spatial continuity factor, the temporal continuity factor, and the physical constraint factor of the bed surface evolution. The signal quality factor is determined based on the signal-to-noise ratio of the sensor response signal and is used to characterize the stability of the sensor response. The interface mutation intensity factor is determined based on the gradient value of the physical quantity at the elevation of the pre-screening candidate bed surface, and is used to characterize the degree of mutation of the physical quantity at the elevation of the pre-screening candidate bed surface. The spatial continuity factor is determined based on the elevation deviation between the current measuring point and the adjacent measuring points whose confidence level is higher than a preset confidence threshold, and is used to characterize the coordination between the current measuring point and the adjacent measuring points. The time continuity factor is determined based on the elevation difference between the current time and the previous time, and is used to characterize the continuity of the elevation change range between the current time and the previous time. The physical constraint factor for bed evolution is determined based on preset or dynamically updated physical constraints for bed evolution.
[0120] After completing the consistency pre-screening and dynamically updating the physical constraint boundaries, the multi-dimensional confidence evaluation stage begins. For each retained candidate bed elevation, its evaluation index is calculated in five independent dimensions.
[0121] All five evaluation factors are normalized to a dimensionless interval of 0 to 1, where 1 represents performance in that dimension. The specific calculation of each factor can be achieved using conventional signal quality evaluation methods, gradient normalization methods, and spatiotemporal difference mapping methods in this field. Those skilled in the art can select the specific normalization function form according to the sensor type and actual data characteristics.
[0122] First, calculate the signal quality factor, which characterizes the signal-to-noise ratio or baseline stability of the sensor's physical response itself.
[0123] Second, calculate the interface abrupt change intensity factor, which characterizes the magnitude of the numerical jump gradient of the physical parameters under investigation (such as acoustic energy, electrical conductivity, etc.) at the candidate elevation location.
[0124] Third, calculate the spatial continuity factor, which measures the numerical deviation between the current candidate elevation and the confirmed high-confidence bed elevation in adjacent areas. The smaller the deviation, the higher the spatial continuity factor.
[0125] Fourth, calculate the time continuity factor, which measures the rate of change between the current candidate elevation and the historical elevation of the measuring point at the previous moment.
[0126] Fifth, calculate the physical constraint factor of the bed surface. Compare the local slope change introduced by the candidate elevation with the dynamic stable slope angle calculated in step 702; compare the instantaneous rate change introduced by the candidate elevation with the dynamic maximum allowable scour rate calculated in step 702. If the comparison result does not exceed the limit, assign a high score to the constraint factor; if it exceeds the limit, penalize the constraint factor using a Gaussian decay function or a step cutoff function.
[0127] After obtaining the evaluation factors for the five independent dimensions mentioned above, a weighted summation method is used to calculate the overall confidence score.
[0128] The relevant confidence level calculation logic is expressed by the following formula: C _i_j (t)=a1*Q _i_j +a2*G _i_j +a3*S _i_j +a4*T _i_j +a5*Φ _i_j ; Among them, C _i_j (t) represents the final confidence score of the j-th pre-screened candidate bed elevation at the i-th measuring point at time t. _i_j G _i_j S _i_j T _i_j Φ _i_j These represent the signal quality factor, interface mutation intensity factor, spatial continuity factor, temporal continuity factor, and bed surface physical constraint factor calculated above, respectively. a1 to a5 represent the pre-configured weighting coefficients corresponding to each factor.
[0129] In some optional implementations, for typical sandy bed scouring environments, the pre-configured weighting coefficients can be assigned values within the following ranges: a1 from 0.15 to 0.25, a2 from 0.20 to 0.30, a3 from 0.15 to 0.25, a4 from 0.15 to 0.25, and a5 from 0.10 to 0.20, ensuring that the sum of all coefficients is always equal to 1. This multi-dimensional allocation of weights avoids precipitous distortion in the final confidence score calculation due to local noise interference from a single evaluation index.
[0130] Example 8: Further explanation of the recursive Bayesian state estimation fusion process based on adaptive weights and the state physical projection truncation mechanism for the estimation results.
[0131] In one possible implementation, the following steps are included: Step 801, the state fusion adopts recursive Bayesian state estimation, with the bed surface elevation and the rate of change of bed surface elevation at preset measuring points as dynamic state variables; the process model of the recursive Bayesian state estimation includes a preset rate decay factor that reflects the scour approaching the equilibrium depth.
[0132] Since the candidate elevations extracted independently at each time point lack temporal correlation, this step introduces a state-space model to map discrete observation points into a physical state sequence that evolves smoothly over time.
[0133] For each preset measuring point, a two-dimensional dynamic state variable column vector is constructed. The first term of the column vector is the current bed elevation, and the second term is the current bed elevation change rate.
[0134] A process model is constructed to describe the evolution of this dynamic state variable over time. In the state transition matrix of this process model, a rate decay factor is configured for the bed elevation change rate term. The value of this rate decay factor is typically set to a real number greater than 0 and less than or equal to 1.
[0135] When the hydrodynamic force is strong in the early stage of the scouring process, the rate decay factor is set to a value close to 1 to maintain a high scouring rate for the next moment. When the scouring process approaches the equilibrium depth and the movable bed surface gradually stabilizes, the rate decay factor decreases, so that the rate of change in the prediction step gradually approaches zero.
[0136] The relevant process model prediction calculation logic is expressed by the following formula: x _i (t)=F*x _i (t _pre )+q _i (t); Where, x _i (t) represents the predicted dynamic state variables of the i-th preset measurement point at the current time, F is the state transition matrix including the rate decay factor, and x _i (t _pre q represents the estimated dynamic state variables at the previous sampling time. _i (t) represents the process noise vector representing an unknown external disturbance.
[0137] Step 802: Construct observation vectors using the elevations of each of the pre-screened candidate bed surfaces located at the same preset measurement point as the observations; convert the confidence levels corresponding to each of the pre-screened candidate bed surface elevations into the weighted reciprocals of the diagonal terms of the observation noise covariance matrix, which are then used as the observation noise weights.
[0138] After establishing the prediction step of the process model, discrete data extracted from multi-source sensors needs to be introduced into the update step to correct the prediction state. All pre-screened candidate bed elevations output in parallel from different channels such as acoustic, electrical, or pore pressure at the same preset measurement point are obtained, and the elevation values are spliced together in a fixed order to form a multi-dimensional observation vector.
[0139] Meanwhile, in order to measure the voice of each of the above observations during the fusion process, the confidence level corresponding to the elevation of each pre-screened candidate bed surface was extracted from the previous calculation.
[0140] Construct an observation noise covariance matrix to represent the observation uncertainty, and use each confidence level as a divisor to perform a division operation with a preset benchmark observation noise variance. Assign the result to the diagonal elements of the covariance matrix.
[0141] Specifically, S _i_j =(σ _0 2 ) / w _j ; In the formula, S _i_j σ is the value of the j-th diagonal term in the observation noise covariance matrix corresponding to the i-th preset measurement point. _0 2 w is the preset benchmark observation noise standard deviation. _j Let be the confidence level corresponding to the elevation of the j-th pre-screened candidate bed surface.
[0142] The specific value of the standard deviation of the benchmark observation noise can be determined based on the noise baseline data collected during the static water calibration phase of the sensor. Those skilled in the art can obtain it by calculating the standard deviation of the sensor elevation readings under static water conditions.
[0143] According to the above mapping rules, when the confidence value corresponding to a candidate elevation is high, the corresponding diagonal term in the observation noise covariance matrix is small, indicating that the observation data contains low noise interference. In the subsequent calculation of the Kalman gain matrix, the low-noise observation data will obtain a larger update gain, forcing the updated continuous bed elevation estimate to move closer to the high-confidence candidate elevation.
[0144] In this process, after obtaining the predicted state covariance matrix and the observation noise covariance matrix, the Kalman gain matrix is calculated according to the standard Kalman filter theory, and the predicted state is updated accordingly to obtain the posterior continuous bed surface elevation estimate and its uncertainty. The above calculation process is a well-known technique in this field and will not be elaborated here.
[0145] Step 803, after obtaining the estimated elevation value of the continuous bed surface at the preset measuring points, further includes performing a physical constraint projection on the estimated elevation value of the continuous bed surface: If the slope of the bed between adjacent preset measuring points exceeds the preset or dynamically updated dynamic stable slope angle, the estimated value of the continuous bed elevation will be reduced in the direction that satisfies the constraint. If the rate of change corresponding to the estimated elevation of the continuous bed exceeds the preset or dynamically updated maximum allowable scouring rate, then the rate of change is truncated.
[0146] When processing erosion data with strong non-Gaussian noise, conventional Bayesian filters may output divergent states that violate the laws of particle mechanics. For example, they may calculate an overhanging slope with a negative angle in a local area, or show instantaneous changes that exceed the fluid's erosion capacity.
[0147] Therefore, a post-event physical boundary forced correction mechanism is introduced after each recursive update.
[0148] Extract the estimated elevation values of the continuous bed surface between the current preset measuring point and the adjacent preset measuring point, calculate the elevation difference between the two points and divide it by the horizontal spatial distance to obtain the calculated slope.
[0149] The calculated slope is compared with the dynamically stable slope angle obtained from previous updates based on hydrodynamic intensity. If the calculated slope is greater than the dynamically stable slope angle, the slope reduction procedure is initiated.
[0150] The required elevation shrinkage amount dz for the current preset measuring point _i : dz _i =λ*(abs(z _i -z _k )-L _ik *tan(φ _max ))*sign(z _i -z _k ); In the formula, λ is a pre-set constraint contraction coefficient, abs is the absolute value operation, and z _i The estimated elevation of the continuous bed surface at the current preset measuring point, z _k L is the estimated elevation of the continuous bed surface at adjacent preset measuring points. _ikφ represents the horizontal spatial distance between the current preset measuring point and the adjacent preset measuring points, where tan is the tangent operation, and φ is the tangent. _max For the dynamically stable slope angle, sign is the function to extract the sign of the sign.
[0151] The value of the constraint contraction coefficient λ is between 0 and 1, and can be adaptively set by those skilled in the art according to the requirements of filter convergence stability.
[0152] After calculating the elevation shrinkage, it is added to or subtracted from the estimated elevation of the continuous bed surface at the current preset measuring point, forcing the local topography to revert to a safe state that satisfies the stability limit of the angle of repose.
[0153] Similarly, the absolute value of the rate of change of bed elevation in the state variables is extracted and compared with the dynamic maximum allowable scour rate obtained from the previous dynamic update. If the absolute value of the rate of change is greater than the dynamic maximum allowable scour rate, a rate truncation procedure is initiated. The system retains the original sign of the rate of change, but forcibly assigns its absolute value to the value corresponding to the dynamic maximum allowable scour rate, eliminating abnormal high-frequency oscillations. After the above forced physical projection truncation, the corrected continuous bed elevation estimate is output for subsequent 3D reconstruction.
[0154] Example 9: Further explanation of the specific implementation method of three-dimensional bed surface reconstruction based on the continuous bed surface elevation estimation values of each preset measuring point.
[0155] In one possible implementation, the following steps are included: Step 901: Construct a triangulation network based on the estimated elevation values of the continuous bed surface at each of the preset measuring points; The three-dimensional spatial physical coordinates of all preset measuring points at a specific time section are obtained, where the horizontal coordinates are determined by the physical mapping position of the sensor during initial installation, and the vertical coordinates are the estimated values of the continuous bed surface elevation output after state fusion and physical projection truncation.
[0156] Based on spatially distributed discrete point cloud data, a spatial partitioning algorithm is applied to a horizontal projection plane to establish a grid topology connecting the various data points.
[0157] In one specific implementation, the constrained Delaunay subdivision algorithm is used to generate a spatial triangular mesh model. During the subdivision operation, the circumcircle of any triangle does not contain other measurement points, thus forming a spatially discretized basic network with local geometric properties.
[0158] Step 902: Identify grid edges in the triangular network whose slope gradient exceeds a preset threshold, and mark the grid edges exceeding the preset threshold as feature edges; When scour evolves around the foundation of a hydraulic structure, the scour pit often has a distinct boundary scarp. If global smoothing interpolation is performed directly over the entire computational domain, it will smooth out and weaken the true physical and geometric features of the scarp.
[0159] Therefore, local feature recognition of the mesh edges needs to be performed before interpolation. This involves traversing each mesh edge in the triangulated network and obtaining the elevation and horizontal coordinates of the two vertices at both ends of that edge.
[0160] Calculate the elevation difference between the two vertices and divide this elevation difference by the horizontal physical distance between the two vertices to obtain the slope gradient of the grid edge. Obtain the preset threshold set by the system, which is usually configured to be half of the tangent of the natural angle of repose of the test sediment in still water.
[0161] The slope gradients calculated for each grid edge are compared numerically with the preset threshold. When the slope gradient of a certain grid edge is determined to be greater than the preset threshold, the spatial topological location is confirmed to be on a steep slope of an erosion pit or in a locally eroded cutoff region. In the algorithm's underlying data structure, the attribute identifier of this grid edge is updated and recorded as a feature edge.
[0162] Step 903: Perform smooth interpolation on the region outside the feature edge, and perform independent interpolation on the regions on both sides of the feature edge to maintain the local bed surface boundary features, thereby completing the three-dimensional bed surface reconstruction.
[0163] Based on the mesh edge attribute identification classification completed in the previous steps, a differentiated and regional spatial interpolation operation strategy is adopted in the continuous bed surface generation stage.
[0164] Extract the set of grid spaces that are not marked as feature edges, and determine that the region corresponding to this set belongs to a non-severe erosion area or an uneroded area with a gentle change in the bed surface morphology.
[0165] For this flat region, a pre-built spatial interpolation algorithm library is used for surface smoothing. In alternative computational paths, radial basis function interpolation or inverse distance weighting algorithms are employed to generate grid node elevations with continuous spatial derivatives by calculating distance weighting factors.
[0166] Furthermore, the data space set marked as feature edges is extracted. When performing spatial weight allocation and interpolation operations, the feature edges within this set are set as physical isolation barriers for the interpolation algorithm, blocking the weight transfer operation path between discrete data points on both sides of the feature edge.
[0167] For the independent datasets on both sides of the feature edge, independent local surface fitting functions are initialized and constructed for calculation, so that the reconstructed system bed surface allows for discontinuous numerical jumps in spatial gradient parameters when crossing the physical location of the feature edge.
[0168] At the physical connection points of the feature edges, a numerical transition method is used to connect the boundary elevations generated by the independent interpolation functions on both sides. Following the aforementioned topological isolation and independent weight allocation reconstruction mechanism, the system calculates and outputs a three-dimensional continuous bed surface model with the ability to preserve the structural features of the steep sill, thus completing the three-dimensional bed surface reconstruction. This output three-dimensional reconstruction result will serve as the spatial mapping basis data for subsequent extraction of scour pit geometric evolution indices.
[0169] Example 10: The specific process of outputting scour pit evolution index is further explained, focusing on the method of separating and calculating complex multi-scour pit evolution index based on connected component segmentation.
[0170] In one possible implementation, the following steps are included: Step 1001: Based on the scour depth after the three-dimensional bed surface reconstruction is completed, perform spatial connectivity analysis on the scour area to divide the scour area into several independent scour pit areas that are not connected to each other. Obtain the continuous three-dimensional bed surface reconstruction mesh model output by the aforementioned steps, and perform difference calculation between the spatial coordinates of the reconstruction surface and the initial bed surface elevation calibrated in the initial state of the experiment to obtain the scour depth of each spatial measuring point at the current moment.
[0171] When the flow field of a hydraulic structure is complex, such as the presence of pile foundations or asymmetric obstacles, the bed surface often forms multiple spatially unconnected scour zones, such as horseshoe vortex scour pits on the water-facing side of the structure, erosion grooves on the sides, and wake disturbance pits on the back side.
[0172] To independently quantify the aforementioned complex morphology, a scour identification threshold greater than zero is set. All grid nodes with a scour depth greater than or equal to this threshold are extracted to form a valid scour region point set. A spatial connectivity labeling algorithm is then invoked, using the physical shared edges between adjacent grid nodes as the connectivity criterion, to perform spatial topology analysis on the valid scour region point set.
[0173] This analysis process divides the overall effective scour area into multiple discrete sets of sub-regions, ensuring that there is no spatial connectivity between the different sub-regions. The resulting sets of sub-regions are then output as the independent scour pit regions.
[0174] Step 1002: For each of the independent scour pit areas, perform integral calculations with the scour depth at each location as the spatial weight to obtain the center position of each of the independent scour pit areas, which serves as the scour pit evolution index.
[0175] In the traditional single-mass centroid calculation mode, if the geometric center is directly calculated for a global flow field with multiple unconnected scour pits, the calculated global centroid coordinates will most likely fall into the unscoured solid region between the individual scour pits, causing the position parameter to lose its physical characterization meaning.
[0176] To address the distortion of metrics caused by multi-pit fusion calculations, this step initiates a multi-centroid integration program for independent regions, based on connected component segmentation. It iterates through each independent scour pit region output from the previous step, and for each specific independent scour pit region, it replaces the conventional mass parameters with local scour depth values, performing a spatially weighted double integration.
[0177] The relevant multi-centroid separation integral calculation logic is expressed by the following formula: x _c_m (t)=[iint _m (x*d _s (x,y,t)*dx*dy)] / [iint _m (d _s (x,y,t)*dx*dy)]; y _c_m (t)=[iint _m (y*d _s (x,y,t)*dx*dy)] / [iint _m (d _s (x,y,t)*dx*dy)]; Where, x _c_m (t) and y _c_m (t) represents the horizontal center coordinates of the m-th independent scour pit region obtained at time t, where iint _m For a dual spatial integral operator that is limited to the spatial boundary of the m-th independent scour pit region, x and y are the horizontal physical coordinates corresponding to the integral element, and d _s (x,y,t) represents the scour depth at the coordinate position at time t, and dx and dy represent the physical dimensions of the integral element in the horizontal dimension.
[0178] Through the above integral calculations, the geometrically weighted centroid coordinates of each independent scour pit region are extracted. Simultaneously, for each independent scour pit region, its local maximum scour depth, local scour area, and local scour erosion volume are calculated. The parameter sets obtained from the separate calculations for each independent object are packaged to obtain the scour pit evolution index, which is then output.
[0179] In some alternative implementations, for the discretized grid data format, the dual spatial integration operation is implemented by discrete summation.
[0180] Specifically, the weighted center of scour depth for each region is obtained by multiplying the scour depth value of each grid node by its horizontal spatial coordinate, summing the results, and then dividing by the sum of the scour depths within that independent region.
[0181] Furthermore, the center position of the same independent scour pit region at different time sections is extracted, the relative spatial displacement vector of the corresponding center position is calculated, and the displacement vector is divided by the time interval to obtain the spatial migration rate of the independent scour pit, which is output as a supplementary scour pit evolution index.
[0182] Example 11: Further explanation of the abnormal signal identification and dynamic weight adaptive adjustment mechanism under partial sensor failure. This mechanism serves as a safety fallback strategy before state fusion, used to handle special operating conditions where a single sensor channel suffers extreme environmental interference, ensuring the robustness of the system.
[0183] One possible implementation includes the following anomaly identification and weight configuration steps: Continuous multi-source non-optical signals were acquired within the scouring test section, and time-series anomaly detection was performed on the data characteristics of each sensor channel. In the turbid water environment of the moving bed scouring under hypergravity, the multi-source sensors are susceptible to interference from air bubbles, suspended sediment clumps, electrical signal saturation, or pore pressure spikes.
[0184] For the acoustic echo channel, candidate acoustic bed elevations are extracted over several consecutive sampling periods. The rate of change of these candidate bed elevations between adjacent sampling periods is calculated. This rate of change is then compared with the dynamically maximum permissible scour rate, which is pre-updated based on hydrodynamic intensity.
[0185] When the acoustic candidate bed elevation is identified to exhibit multi-peak drift characteristics, and the corresponding rate of change is greater than the dynamic maximum allowable scour rate, it is determined that the acoustic signal is abnormal due to bubble blockage or interference from suspended sediment masses. The system triggers a weight penalty instruction, reducing the confidence evaluation weight of the acoustic signal during that time period.
[0186] For the electrical sensing channel, acquire the real-time electrical responses of each vertically arranged electrical sensing sensor. Calculate the time derivative of this electrical response.
[0187] When instantaneous data saturation is detected in the electrical response of a certain electrical sensing channel, or when synchronous numerical changes in the electrical response occur at all vertically distributed measuring points at the same sampling time, it is determined that the electrical signal is affected by system electromagnetic noise or sudden changes in the overall water sediment concentration. The system triggers a weighted penalty instruction, reducing the voting weight of the electrical signal in the physical consistency score.
[0188] For the pore pressure sensing channel, the time-domain pulse characteristics of the pore water pressure signal are extracted and compared with the spectrum of the wave-generating equipment.
[0189] When a high-frequency spike unrelated to the wave period appears in the pore water pressure signal, and the spike does not exhibit a spatial propagation pattern of fluid pressure between adjacent spatial measurement points, it is determined to be a local abnormal disturbance pulse, and pore pressure candidate features containing the high-frequency spike period are directly eliminated.
[0190] When a certain type of sensor continuously outputs abnormal characteristics within a preset time window, the system initiates a dynamic weight adaptive allocation program. It calculates the confidence weight difference resulting from the forced removal of the failed sensor and redistributes this weight difference to the remaining evaluation factors that are operating normally, with a focus on increasing the weighting coefficients of the spatial continuity factor and the physical constraint factor of the bed surface.
[0191] The calculation logic for dynamic weight compensation is expressed by the following formula: a _new_k =a _old_k +δ _a *(a _old_k / Σ _a_valid ); Among them, a _new_k Let a be the weighting coefficient of the updated k-th effective factor. _old_k δ represents the initial weighting coefficient of the k-th effective factor before the update. _a Σ represents the total amount of compensation weights released by the failed sensor channels. _a_valid The weighted sum is obtained by accumulating the initial weighting coefficients of all factors currently in a valid state.
[0192] In a certain scenario, assuming that in the initial state, the weights of the signal quality factor, interface mutation intensity factor, spatial continuity factor, temporal continuity factor, and bed surface physical constraint factor in the confidence evaluation model are all configured to 0.2. During a certain period, the system detects dense free bubbles above the target measurement point, causing a sudden drop in the acoustic echo signal-to-noise ratio. The system determines that the acoustic channel has failed, forcibly resets its corresponding signal quality factor weighting coefficient to zero, and releases the total amount of weights to be compensated (0.2), i.e., δ in the aforementioned formula. _a The value is 0.2. The sum of the initial weighting coefficients of the remaining four factors in an effective state is 0.8, i.e., Σ. _a_validThe value is set to 0.8. Substituting the above parameters into the dynamic weight compensation formula to calculate the updated weight of the spatial continuity factor, the result is 0.2 + 0.2 * (0.2 / 0.8) = 0.25. Similarly, the weighting coefficients of the interface mutation intensity factor, the time continuity factor, and the physical constraint factor of the bed surface are all updated synchronously to 0.25. After the above parameter adaptive reconstruction calculation, even under the condition that some sensors experience local failure, the system can rely on spatial correlation and physical inertia to maintain the continuous calculation of the bed surface reconstruction results.
[0193] Example 12: This example provides a parameter migration and deployment implementation method for complex working conditions. For non-standard working conditions such as localized scour of offshore wind turbine monopile foundations and scour of cohesive soil beds, this example illustrates the spatial configuration mechanism of multi-source sensors and the dynamic reconstruction process of confidence weights.
[0194] In the first working condition, the target scenario is the localized scouring of an offshore wind turbine monopile foundation under wave-current coupling. In this condition, the flow field around the monopile includes horseshoe vortices on the upstream side, streamline compression on the side, and wake shedding on the downstream side, resulting in a highly spatially asymmetric evolution of the bed morphology. When spatially configuring the multi-source non-optical sensor array, acoustic echo sensor arrays are deployed above the upstream, side, and downstream regions of the monopile model.
[0195] Electrical sensing sensors are attached to the outer surface of the monopile model at preset vertical intervals, forming a vertical monitoring chain to detect abrupt changes in the medium from the water body to the bed surface. Micropore water pressure sensors are grouped and buried below the initial bed surface elevation on the water-facing and side surfaces. Multi-source non-optical signals are acquired through this spatial configuration and input into the state estimation and three-dimensional reconstruction model provided in the aforementioned embodiments.
[0196] Since the scour pit is usually divided into unconnected deep pits in front of the pile and trenches on the side of the pile under this working condition, the system calls the multi-centroid integral program of independent division to extract the geometric features of each independent region and output independent scour pit evolution indicators such as the depth of scour in front of the pile, the volume of the scour pit on the side, and the height of the wake back sedimentation.
[0197] In the second scenario, the target environment is the dynamic scouring of a cohesive or silty soil bed. Unlike the independent particle stripping mechanism of sandy beds, cohesive soils undergo surface softening, block erosion, and flocculation suspension under fluid shear. The presence of a softening transition layer reduces the acoustic impedance gradient between the water body and the dense bed, resulting in broadening of the echo signal peak and decreased sharpness acquired by the acoustic echo sensor. Under this physical context, maintaining the initial weighting configuration for sandy beds would lead to deviations in the acoustic feature-based recognition output. Therefore, the system activates a parameter migration configuration protocol for cohesive soil beds, redistributing the weighting coefficients in the confidence evaluation model.
[0198] Extract the confidence evaluation calculation logic established in the aforementioned embodiments: C _i_j (t)=a1*Q _i_j +a2*G _i_j +a3*S _i_j +a4*T _i_j +a5*Φ _i_j ; Among them, C _i_j (t) represents the confidence score, Q. _i_j G is the signal quality factor. _i_j S is the interface mutation intensity factor. _i_j T is the spatial continuity factor. _i_j Φ is the time continuity factor. _i_j A represents the physical constraint factor of the bed surface, and a1 to a5 represent the pre-configured weighting coefficients corresponding to each factor.
[0199] Under cohesive soil conditions, the system extracts the signal quality factor characterizing the acoustic echo features and lowers the corresponding weighting coefficient a1. Simultaneously, because the electrical sensing sensor exhibits high data response sensitivity to changes in porosity within the softened layer, and the pore water pressure sensor demonstrates clear physical boundary characteristics in its hysteretic response within the cohesive soil layer, the system extracts the interface abrupt change intensity factor involving electrical and pore pressure parameters and increases its corresponding weighting coefficient a2. Since the erosion evolution rate of cohesive soil is typically lower than that of non-cohesive sand, the system simultaneously increases the weighting coefficient a4 of the time continuity factor, enabling the recursively generated continuous bed elevation estimate to maintain stronger inertia over time. Through the above parameter reconstruction operations tailored to the physical properties of specific geological materials, the state fusion model, under the condition of blurred physical boundaries at the acoustic reflection interface, maintains continuous computational output for the stable bed interface by relying on electrical gradient abrupt changes, pore pressure phase evolution, and temporal state inertia.
[0200] To address the common challenge of single sensors being susceptible to data jumps and time discontinuities caused by strong local disturbances, a framework for heterogeneous physical response consistency pre-screening and adaptive state estimation was constructed. By mapping the comprehensive evaluation confidence score, which includes signal quality, abrupt change intensity, and spatiotemporal continuity, to observation noise weights, it is beneficial to improve the smooth evolution of discrete observations into continuous physical state sequences, suppress interference from non-Gaussian isolated noise, and achieve high stability in the time-series tracking process.
[0201] Furthermore, to overcome the technical bottlenecks of traditional fixed filtering thresholds causing distortion in terrain evolution during alternating strong and weak hydrodynamics, and the difficulty in quantifying and analyzing multiple scour pits under complex flow fields, this scheme adopts dynamic updates of near-bottom hydrodynamic intensity to allow for scour rate and stable slope angle, enabling the mathematical filtering boundary to adapt to the evolution of the hydrodynamic boundary. At the same time, the combination of connected domain spatial segmentation and local depth weighted integration is beneficial to improving the independent extraction of the evolution trajectory of complex multi-morphological scour pits, and reducing the physical coordinate drift deviation that may occur in traditional global centroid calculation.
Claims
1. A non-optical identification and three-dimensional reconstruction method of a moving bed scour pit in a high gravity field, characterized in that, include: Multiple non-optical signals were acquired during the dynamic bed scouring model test in a hypergravity field, and multiple candidate bed surface elevations were extracted based on the multiple non-optical signals. Each of the candidate bed surface elevations is pre-screened and evaluated to obtain the confidence level of each candidate bed surface elevation; Based on the candidate bed surface elevations and their corresponding confidence levels, state fusion is performed to obtain the estimated continuous bed surface elevations at preset measurement points. Based on the estimated elevation of the continuous bed surface at each of the preset measuring points, a three-dimensional bed surface is reconstructed, and the scour pit evolution index is output. Before extracting multiple candidate bed surface elevations based on the multi-source non-optical signals, the process also includes: The sound wave propagation path is divided into several thin layers; The sound velocity within each layer is determined based on the sand concentration and temperature of each layer, and the propagation direction and propagation time of the acoustic signal in the multi-source non-optical signal are corrected by combining the law of refraction, thus obtaining the corrected multi-source non-optical signal. Accordingly, the step of extracting multiple candidate bed surface elevations based on the multi-source non-optical signals specifically involves: extracting multiple candidate bed surface elevations based on the modified multi-source non-optical signals; the modified multi-source non-optical signals include vertically distributed electrical response signals; The extraction of multiple candidate bed surface elevations based on the corrected multi-source non-optical signal includes: Based on the pre-calibrated mapping relationship between electrical characteristics and sediment concentration, the electrical response signal is converted into a vertical sediment concentration profile; Identify the locations of abrupt changes in the vertical gradient in the vertical sediment concentration profile, and extract the electrical candidate bed elevation as one of the candidate bed elevations; Correction of multi-source non-optical signals also includes acoustic echo signals; Extracting multiple candidate bed surface elevations based on the corrected multi-source non-optical signals also includes: The vertical sand concentration profile is input into the acoustic attenuation model to calculate the expected echo intensity at the reflection interface; Calculate the ratio of the measured echo intensity to the desired echo intensity of the acoustic echo signal, and classify the physical properties of the echo peaks in the acoustic echo signal based on the ratio; The distances corresponding to the echo peaks classified as high-probability bed surface peaks are converted into acoustic candidate bed surface elevations, which are then used as one of the candidate bed surface elevations. The modified multi-source non-optical signal also includes at least one of the following: pore water pressure signal, near-bottom pressure pulsation signal, or flexible contact signal; Extracting multiple candidate bed surface elevations based on the corrected multi-source non-optical signals also includes at least one of the following operations: Identify the boundary location where the pore water pressure signal changes from a hysteretic response within the soil to a synchronous response within the water body, and extract the pore pressure candidate bed elevation as one of the candidate bed elevations; Identify the locations where pressure pulsation energy abruptly changes in the near-bottom pressure pulsation signal, and extract the pressure pulsation candidate bed surface elevation as one of the candidate bed surface elevations; Identify the location where the contact state changes abruptly in the flexible contact signal, and extract the strain candidate bed surface elevation as one of the candidate bed surface elevations; Based on the candidate bed surface elevations and their corresponding confidence levels, state fusion is performed to obtain the estimated continuous bed surface elevation values for preset measuring points, specifically: The state fusion is achieved using recursive Bayesian state estimation. A recursive state space model is established using the bed elevation and its rate of change at the preset measurement points as dynamic state variables. Within each data sampling period, the candidate bed elevations located at the same preset measurement point are used as observation inputs, and the confidence levels corresponding to each candidate bed elevation are converted into observation noise weights in the state update. Candidate bed elevations with higher confidence levels are assigned higher confidence weights in the fusion update. Through recursive iteration of prediction and update, continuous bed elevation estimates for the preset measurement points are output.
2. The method of claim 1, wherein, Pre-screening of each of the candidate bed elevations includes: The number of heterogeneous sensor physical responses supported within a preset range of the candidate bed surface elevation is counted to determine the cross-sensor physical consistency score. If the cross-sensor physical consistency score is lower than a preset threshold, the weight of the candidate bed elevation is reduced or the candidate bed elevation is removed to obtain a pre-screened candidate bed elevation.
3. The method of claim 1, wherein, The candidate bed elevations are pre-screened and evaluated, and the evaluation process specifically includes: Obtain the near-bottom hydrodynamic intensity within the scour test section, and dynamically update the physical constraint factor for bed evolution based on the near-bottom hydrodynamic intensity; The confidence level is determined by weighting the following factors, including: Signal quality factor, used to characterize the stability of sensor response; Interface mutation intensity factor is used to characterize the degree of mutation of physical quantities at the candidate bed surface elevation after pre-screening; The spatial continuity factor is used to characterize the coordination between the current measuring point and adjacent high-confidence measuring points; The time continuity factor is used to characterize the continuity of the elevation change magnitude between the current moment and the previous moment; The physical constraint factor for bed evolution is determined based on preset or dynamically updated physical constraints for bed evolution.
4. The method of claim 3, wherein, The near-bottom hydrodynamic intensity within the scour test section was obtained, specifically: Real-time acquisition of near-bottom pressure pulsation signals within the scour test section; The root mean square value of the near-bottom pressure pulsation signal within a preset short time window is calculated, and the near-bottom hydrodynamic intensity is determined based on the ratio of the root mean square value to a comprehensive empirical coefficient.
5. The method of claim 3, wherein, The near-bottom hydrodynamic intensity within the scour test section was obtained, specifically: Real-time acquisition of near-bottom pressure pulsation signals within the scour test section; By combining a preset velocity scale and processing the near-bottom pressure pulsation signal using the turbulent kinetic energy method, the near-bottom hydrodynamic intensity is obtained.
6. The method according to claim 4 or 5, characterized in that, The physical constraint factors for bed evolution are dynamically updated based on the near-bottom hydrodynamic intensity, including: The near-bottom hydrodynamic intensity is compared with the pre-calibrated critical starting hydrodynamic intensity for sediment; Based on the comparison results, the dynamic maximum allowable scour rate and dynamic stable slope angle in the physical constraint factors of the bed surface evolution are dynamically updated. Specifically, when the near-bottom hydrodynamic intensity exceeds the critical starting hydrodynamic intensity of sediment, the maximum allowable dynamic scour rate is increased, and the natural angle of repose of sediment under still water conditions is reduced to decrease the dynamic stable slope angle.
Citation Information
Patent Citations
Three-dimensional numerical simulation method for downstream scouring process of flood discharge gate of hydropower station
CN110046398A
Channel earthwork three-dimensional dynamic detection system based on multi-modal data driving
CN122592410A
