Channel scouring and evolution collaborative control method, device and equipment and storage medium

Abstract

The application provides a method, device and equipment for the coordinated control of channel scouring and evolution, and a storage medium. The method comprises the following steps: first, calculating a correlation coefficient according to a sediment transport potential field and a water depth change, and entering the next step if the verification fails; then, calculating a grouped and ungrouped sediment transport rate according to a real-time particle size distribution, and determining that the gradation is abnormal if the deviation is large. At the same time, the spatial graph convolution topography weight is corrected based on the maximum particle size ratio, the modified eddy viscosity coefficient is determined, the weight coefficient of the bed load is updated, and finally, the multi-physical field coupling model is updated by using the coefficients, so that the prediction accuracy of the channel scouring is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of channel scouring and silting control, and particularly relates to a channel scouring and silting and evolution collaborative control method, device, equipment and storage medium. BACKGROUND

[0002] Traditional channel scouring and silting is mainly used to guarantee channel safe traffic capacity and operation efficiency, and assess scouring and silting change trend through data collection and simulation analysis to provide decision support for channel management.

[0003] At present, the existing technology contains shipborne sonar sounding technology and fixed point monitoring means, and a prediction model considering physical processes in isolation. Water depth data is collected by repeatedly sailing the measuring ship, continuous data stream is provided by laying sensors, and tidal flow, wave and sediment transport processes are respectively processed by using simplified equations to obtain a prediction result with certain accuracy under ideal conditions, such as preliminary identification of silting areas in a stable environment.

[0004] In the process of sailing the measuring ship and fixed point monitoring and modeling of isolated physical processes, the measuring ship sailing will interfere with normal traffic in the channel, and incomplete data collection caused by bad weather, and the fixed point monitoring has limited spatial coverage and cannot fully capture complex topographic changes; at the same time, the model ignores the coupling effect of multiple physical fields and lacks real-time data driving mechanism, resulting in model parameter solidification and inability to adapt to dynamic environment. Therefore, there are defects such as limited monitoring range, insufficient data accuracy, insufficient coupling of physical processes, poor model adaptability and low prediction accuracy. SUMMARY

[0005] The purpose of the present application is to overcome the defects in the prior art, and to provide a channel scouring and silting and evolution collaborative control method, device, equipment and storage medium.

[0006] The present application provides a channel scouring and silting and evolution collaborative control method, comprising:

[0007] According to the sediment transport potential field and the water depth change, the sediment transport potential field correlation coefficient is calculated;

[0008] If the absolute value of the sediment transport potential field correlation coefficient is less than a first threshold value, it is determined that the verification fails;

[0009] According to the real-time particle size distribution, the grouped sediment transport rate and the non-grouped sediment transport rate are calculated;

[0010] Based on the grouped sediment transport rate and the non-grouped sediment transport rate, the particle size sediment transport rate deviation is calculated;

[0011] If the particle size sediment transport rate deviation is greater than a second threshold value, it is determined that the gradation is abnormal;

[0012] When the check fails and the gradation anomaly occurs simultaneously, the following operations are performed: based on the maximum proportion value of the real-time particle size distribution, the terrain weight of spatial graph convolution is corrected to generate a corrected terrain weight; according to the corrected terrain weight, the eddy viscosity coefficient is determined to generate a corrected eddy viscosity coefficient; and according to the grouping settling velocity, the weight coefficient of the bed load is updated to generate an updated weight coefficient.

[0013] The multi-physics field coupling model is updated by using the corrected eddy viscosity coefficient and the updated weight coefficient.

[0014] Optionally, the correction of the terrain weight of spatial graph convolution based on the maximum proportion value of the real-time particle size distribution comprises:

[0015] A sign amplification coefficient is generated according to the sign of the bed surface elevation difference and the maximum proportion value.

[0016] The terrain weight is corrected based on the sign amplification coefficient.

[0017] Optionally, the determination of the eddy viscosity coefficient to generate the corrected eddy viscosity coefficient according to the corrected terrain weight comprises:

[0018] A weighted terrain gradient is generated by performing neighborhood weighting on the terrain gradient by using the corrected terrain weight.

[0019] The eddy viscosity coefficient is updated based on the weighted terrain gradient.

[0020] Optionally, the updating of the weight coefficient of the bed load according to the grouping settling velocity comprises:

[0021] The proportion of the grouping settling velocity in the total settling velocity is calculated.

[0022] The proportion is multiplied by the original weight coefficient of the bed load.

[0023] Optionally, the correction of the terrain weight of spatial graph convolution based on the maximum proportion value of the real-time particle size distribution comprises:

[0024] When the maximum proportion value exceeds a set threshold value, a terrain weight correction operation is started.

[0025] Optionally, the spatial graph convolution comprises:

[0026] The neighborhood node relationship construction comprises: generating a neighborhood set by connecting nodes through Delaunay triangulation.

[0027] Optionally, the determination of the maximum proportion value of the real-time particle size distribution comprises:

[0028] An advantage particle size group is dynamically identified by using a maximum value selection function.

[0029] This application also provides a coordinated control device for channel scouring and sedimentation evolution, comprising:

[0030] The verification module calculates the correlation coefficient of the sediment transport potential field based on the sediment transport potential field and the change in water depth.

[0031] If the absolute value of the correlation coefficient of the sediment transport potential field is less than the first threshold, the verification is determined to have failed.

[0032] The gradation module calculates the grouped and ungrouped sediment transport rates based on real-time particle size distribution.

[0033] Based on the grouped sediment transport rate and the ungrouped sediment transport rate, the particle size sediment transport rate deviation is calculated;

[0034] If the deviation of the particle size transport rate is greater than the second threshold, the gradation is determined to be abnormal.

[0035] The coefficient module performs the following operations when the verification failure and the gradation anomaly occur simultaneously: based on the maximum proportion of the real-time particle size distribution, corrects the terrain weight of the spatial map convolution to generate corrected terrain weight; determines the eddy viscosity coefficient according to the corrected terrain weight to generate corrected eddy viscosity coefficient; and updates the bedload weight coefficient according to the grouped settling velocity to generate updated weight coefficient.

[0036] The update module updates the multiphysics coupling model using the corrected eddy viscosity coefficient and the updated weight coefficient.

[0037] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0038] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the above-described method.

[0039] The beneficial effects of this application are:

[0040] This application provides a method for the coordinated control of channel scouring and sedimentation and evolution, comprising: calculating a correlation coefficient of the sediment transport potential field based on the sediment transport potential field and the change in water depth; determining a verification failure if the absolute value of the correlation coefficient is less than a first threshold; calculating grouped sediment transport rate and non-grouped sediment transport rate based on real-time particle size distribution; calculating the particle size transport rate deviation based on the grouped sediment transport rate and the non-grouped sediment transport rate; determining a gradation anomaly if the particle size transport rate deviation is greater than a second threshold; and performing the following operations when the verification failure and the gradation anomaly occur simultaneously: correcting the topographic weight of the spatial map convolution based on the maximum proportion of the real-time particle size distribution to generate a corrected topographic weight; determining the eddy viscosity coefficient based on the corrected topographic weight to generate a corrected eddy viscosity coefficient; updating the bedload weight coefficient based on the grouped settling velocity to generate an updated weight coefficient; and updating the multiphysics coupling model using the corrected eddy viscosity coefficient and the updated weight coefficient. This application verifies the model by calculating the correlation coefficient based on the sediment transport potential field and water depth variation, and detects gradation anomalies by combining real-time particle size distribution. When verification failure and gradation anomalies occur simultaneously, the topographic weight, eddy viscosity coefficient, and bedload weight coefficient are dynamically corrected, and the multiphysics coupling model is updated. This achieves real-time adaptive optimization of the model, overcomes the defects of existing technologies where the model is fixed and cannot respond to data changes, and thus improves the accuracy and reliability of waterway scour and siltation prediction. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the collaborative control process for waterway scouring and evolution in this application;

[0042] Figure 2 This is a schematic diagram of the multi-source sensor network deployment in this application;

[0043] Figure 3 This is a schematic diagram of dynamic data fusion processing in this application;

[0044] Figure 4 This is a schematic diagram of the multiphysics coupling modeling framework in this application;

[0045] Figure 5 This is a schematic diagram of the ST-GCN model structure in this application. Detailed Implementation

[0046] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is to be understood that various forms of implementation of the present disclosure are intended and should not be limited to the embodiments set forth herein. Rather, the embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0047] Please refer to Figure 1As shown in the embodiments of this application, a collaborative control method for waterway scouring and sedimentation and evolution is provided. This method is applied to the field of monitoring and simulation prediction of scouring and sedimentation and waterway evolution, and is used to solve the problems of limited monitoring range, insufficient data accuracy, insufficient coupling of physical processes, poor model adaptability, and low prediction accuracy in traditional monitoring methods. Specifically, it includes the following basic operations:

[0048] Please refer to Figure 2 As shown, a multi-source sensor network is deployed, with water depth sensors, flow velocity and direction sensors, and sediment concentration sensors installed at key sections of the waterway to collect water depth data in real time. Three-dimensional flow velocity vector and instantaneous sediment content Synchronously access the historical sediment concentration database The sensor deployment is based on the complexity of the waterway terrain, with denser deployment in curved areas. The data acquisition frequency is 1Hz, and the time and space reference is synchronized with GPS clock to ensure uniformity.

[0049] Please refer to Figure 3 As shown, for real-time sensor data , , Spatiotemporal registration was performed using a sliding time window method with a window length of 10 minutes, extracting data within each window. Cubic spline interpolation was used to resample the non-uniformly sampled time series data to a 1Hz frequency to eliminate timing offsets. Physical consistency verification was performed based on the local mass conservation equation to validate the data.

[0050]

[0051] in, For water depth data, It is a three-dimensional flow velocity vector. Indicates the calculation of divergence. Characterizing the rate of change of instantaneous sediment concentration per unit volume over time. This indicates the instantaneous sediment content.

[0052] Data points that do not meet the conservation constraints of the local mass conservation equation are removed using the Laida criterion, i.e. the 3σ principle. This includes: if a data point deviates from the local moving mean by more than 3 times the standard deviation, it is considered an outlier and removed.

[0053] In this application, multimodal noise reduction performs targeted processing on different types of sensor signals, including:

[0054] 1. For water depth data, wavelet packet decomposition is used, with the Daubechies8 mother wavelet (db8) selected and the decomposition level set to 5, to extract low-frequency trend terms. And remove high-frequency noise components Reconstructed into a purified water depth sequence .

[0055] 2. For sand content data, an adaptive median filter with dynamic window adjustment is designed. The basis for this is that the local fluctuation level of the signal is within the range of 3×3 to 7×7. Adjusting the filter window size effectively preserves edge features while removing outliers from the output. The final fused dataset is generated as follows:

[0056]

[0057] in, To merge datasets, To purify the water depth sequence, It is a three-dimensional flow velocity vector. To purify the instantaneous sand content, It represents the change in water depth after purification at adjacent time points.

[0058] Please refer to Figure 4 As shown, further, multiphysics coupling modeling is based on a fused dataset. A coupled mathematical model of tidal currents, waves, and sediment was constructed. The tidal current field was solved using the shallow-water equations with depth correction. The governing equation is the mass conservation equation:

[0059]

[0060] Where H represents water depth, This represents the vertical average velocity vector. This indicates the calculation of divergence.

[0061] The momentum conservation equation is:

[0062]

[0063] in, This represents the vertical average velocity vector. Let represent the spatial gradient of the vertical average velocity, g represent the gravitational acceleration, and η represent the water surface elevation. Indicates turbulent viscosity. H represents the coefficient of friction of the bed surface, and H represents the water depth. This represents the rate of change of the vertical average velocity vector over time. Represents the convective acceleration term. η represents the water surface slope. This indicates the magnitude of the vertical average flow velocity.

[0064] Wave height is calculated using a spectral model of the wave field. With wave period .

[0065] The expression for radiation stress is:

[0066]

[0067] in, ρ represents radiation stress, g represents gravitational acceleration, n represents group velocity ratio, θ represents wave direction angle, E represents wave energy density spectrum, σ represents angular frequency, and k represents wave number.

[0068] To achieve bidirectional coupling between the wave field and the tidal current field, the sediment transport field uses a suspended mass-bedmass separation model to calculate the sediment transport rate. The governing equation for the suspended mass module is:

[0069]

[0070] in, This represents the vertical average velocity vector. C represents the spatial gradient of sediment content. Indicates the calculation of divergence. Indicates the sediment diffusion coefficient. Indicates the settlement velocity. This indicates the local equilibrium sediment concentration. C represents the rate of change of sediment content over time.

[0071] The transport module uses the modified Meyer-Peter formula:

[0072]

[0073] in, This represents the bedload transport rate per unit width. The density of the sediment is represented by g, and g represents the acceleration due to gravity. The representative particle size is represented by θ, and the Shield number is represented by θ. This represents the critical Shield number.

[0074] Furthermore, by constructing a dynamic prediction engine through machine learning, the current field output is... Wave field output , Sediment transport rate With fused data Input spatiotemporal graph convolutional neural network (ST-GCN, e.g.) Figure 5 As shown), spatial graph convolution constructs the channel computation region as a graph structure G=(V,E) through Delaunay triangulation, where each node... Corresponding to the geometric center point of the unstructured mesh cell, the edge The connection relationship is established based on the Delaunay triangulation, and an eigenvector is defined at each node, expressed as:

[0075]

[0076] in, The feature vector at each node, This represents the power flow velocity at node i. This represents the wave height at node i. This represents the wave period at node i. This represents the sediment transport rate at node i. This represents the water depth at node i. This represents the sand content at node i. This represents the change in water depth at node i.

[0077] The temporal convolution module uses a gated dilated convolution structure, expressed as:

[0078]

[0079] in, The symbol represents a one-dimensional dilated convolution operation, and ⊙ represents the Hadamard product. and Indicates the convolution kernel weights. This represents the input feature vector, tanh represents the hyperbolic tangent activation function, and sigmoid represents the activation function. This indicates the output feature vector.

[0080] The decoder generates a congestion thickness probability map by upsampling through transposed convolution:

[0081]

[0082] Where sigmoid represents the activation function. Represents the fused feature tensor. This represents a probability map of sediment thickness.

[0083] And introduce a channel attention mechanism:

[0084]

[0085] in, The channel attention weights are represented by sigmoid, the activation function is represented by sigmoid, and GAP and GMP represent global average pooling and max pooling operations, respectively. and This represents the channel attention weight parameter. This represents the input feature tensor. The output is a heatmap of the sediment distribution with a spatial resolution of 10 meters by 10 meters.

[0086] S101. Calculate the correlation coefficient of the sediment transport potential field based on the sediment transport potential field and the change in water depth.

[0087] The sediment transport potential field is a physical field that reflects the accumulation or loss trend of sediment in a spatial unit by calculating the divergence of the sediment transport rate vector field.

[0088] sediment transport potential field The expression is:

[0089]

[0090] in, This represents the sediment transport potential field, used to quantify the net accumulation of sediment in a spatial unit. >0) or loss ( <0) Trend; Indicates the calculation of divergence. The vector representing the sand transport rate per unit width is calculated using the following formula:

[0091]

[0092] in, Indicates the height of the bed surface. Indicates water level height. This represents the velocity vector profile in the depth direction. This indicates the instantaneous distribution of sediment concentration at the same location.

[0093] The correlation coefficient A of the sediment transport potential field is calculated using Pearson correlation analysis. With water depth change The strength of the linear relationship between them is calculated using the following formula:

[0094]

[0095] Where A represents the correlation coefficient, ranging from -1 to 1. A positive value indicates a positive correlation (sediment accumulation accompanied by increased water depth), and a negative value indicates a negative correlation (sediment loss accompanied by decreased water depth). The calculation is based on the spatial grid nodes. and The value sequence is obtained by using the ratio of covariance to standard deviation.

[0096] For example, in estuary areas, if the sediment transport potential field shows a tendency to silt up ( >0), but the measured water depth change However, if the scouring decreases, the A value will be low, indicating a mismatch between sensor data or environmental disturbances and the physical process, requiring further verification. The entire calculation process runs automatically in the waterway digital twin platform. Logically, it first obtains the sediment transport rate through integration, and then calculates the divergence to obtain the sediment transport rate. Finally, correlation analysis was performed to assess the reliability of the data.

[0097] S102. If the absolute value of the correlation coefficient of the sediment transport potential field is less than the first threshold, the verification is determined to have failed.

[0098] The first threshold was set at 0.7, based on historical data and experimental calibration; when the absolute value of the correlation coefficient... If the system determines that the verification has failed, it means that the dynamic relationship between the sediment transport potential field and the change in water depth deviates from the physical expectation, which may be caused by sensor inaccuracy or severe environmental disturbances (such as storm surge).

[0099] Verification failure is a binary decision result (failure or success), triggered by |A| remaining below a threshold for more than a preset time window (e.g., 10 minutes). For example, in a downstream river channel, if a typhoon causes turbulent water flow... It shows stagnation, but If there is no change, and |A| drops to 0.5, the system marks the verification as failed and activates the subsequent exception handling mechanism. The comparator module monitors the A value in real time, and when the condition is met, it outputs a verification failure signal to the control unit, ensuring timely handling of data quality issues.

[0100] S103. Calculate the grouped sediment transport rate and the non-grouped sediment transport rate based on the real-time particle size distribution.

[0101] The real-time particle size distribution P(d) is the frequency distribution of sediment particle size collected in real time by the LISST laser particle size analyzer. The sediment is divided into four groups: clay (d<0.004mm), silt (0.004mm≤d<0.0625mm), fine sand (0.0625mm≤d<0.25mm) and medium-coarse sand (d≥0.25mm).

[0102] Group sediment transport rate The calculation is based on the weighted superposition of particle size components, and the formula is:

[0103]

[0104] in, This indicates the sediment transport rate of a group, reflecting the overall sediment transport volume; This represents the proportion of the k-th particle size group in the overall sediment; The k-th group represents the particle size; k represents the particle size group index (1 to 4 correspond to clay, silt, fine sand, and medium-coarse sand, respectively). and The component coupling coefficient is calibrated by indoor water tank tests and reflects the contribution weight of particle size group to suspended transport (suspended mass) and bed surface transport (bed mass). The suspended sediment transport rate is represented by the convection-diffusion equation of the suspended sediment module. This represents the bedload transport rate, calculated using a modified Meyer-Peter formula.

[0105] The ungrouped sediment transport rate is the overall sediment transport rate value directly output from the multiphysics coupling model, without considering particle size grouping. The calculation formula is as follows:

[0106] = +

[0107] in, Non-group sediment transport rate, Indicates suspended sediment transport rate, This indicates the bedload transport rate.

[0108] For example, in estuarine areas with complex sediment composition, when the proportion of fine sand is high, grouped calculations more accurately capture sediment transport dynamics. The system first acquires real-time P(d), and then combines it with the model output. and It calculates grouped values ​​and simultaneously reads the original ungrouped values ​​of the model to provide input for bias calculation.

[0109] S104. Calculate the particle size transport rate deviation based on the grouped sand transport rate and the ungrouped sand transport rate.

[0110] The deviation in particle size transport rate is the group transport rate Non-group sediment transport rate The relative deviation between them is calculated using the following formula:

[0111]

[0112] Where P is the deviation.

[0113] This deviation is used to quantify the influence of particle size grouping on sediment transport rate. For example, when the particle size distribution of sediment is uneven (such as a sudden increase in the proportion of medium and coarse sand), the deviation increases, reflecting the error caused by the non-grouping model ignoring the particle size effect.

[0114] S105. If the deviation of the particle size transport rate is greater than the second threshold, the gradation is determined to be abnormal.

[0115] The second threshold is set at 0.15 (15%), based on historical error analysis of historical data.

[0116] When the deviation consistently exceeds this threshold, the system determines it to be an abnormal gradation, indicating that the actual sediment particle size distribution has changed drastically and the model has not adapted to the new conditions. For example, after dredging of a port channel, the sediment gradation changes abruptly, and the deviation rises to 0.2, at which point the system marks it as an abnormal gradation. This is achieved through a threshold comparator; the abnormal signal triggers the model update requirement, ensuring that the multiphysics coupled model is adjusted in a timely manner.

[0117] S106. When the verification failure and the gradation anomaly occur simultaneously, perform the following operations:

[0118] S1061. Based on the maximum ratio of the real-time particle size distribution, correct the terrain weights of the spatial graph convolution to generate corrected terrain weights.

[0119] When the verification failure and gradation anomaly occur simultaneously (e.g., in a strong disturbance event such as a typhoon, |A|<0.7 and the deviation>0.15), the system activates the collaborative control mechanism.

[0120] Maximum proportion of real-time particle size distribution It represents the highest proportion among the four particle size groups, and the dominant particle size group is dynamically identified through a maximum value selection function; for example, when the proportion of the silt group reaches 0.6, =0.6.

[0121] Terrain weights in spatial graph convolution Defined based on node elevation differences, the original formula is:

[0122]

[0123] in, This represents the spatial weight between nodes i and j; and Indicates the bed surface elevation of adjacent nodes; This represents the square of the Euclidean distance between nodes. This represents the maximum absolute value of the difference in bed surface elevation between nodes in the entire computational domain.

[0124] The terrain weighting correction operation includes: based on the sign function of the bed elevation difference (positive difference indicates i is higher than j, negative difference indicates j is higher than i) and Generate sign amplification factor:

[0125]

[0126] Where sgn is the sign function for the difference in bed elevation. and Indicates the bed surface elevation of adjacent nodes. It is the maximum proportion of the real-time particle size distribution.

[0127] Then, the weights are adjusted based on α as follows:

[0128]

[0129] For example, when High (exceeding the set threshold) and > When α>1, the weight is amplified, enhancing the influence of highland nodes.

[0130] Specifically, when executing in the ST-GCN model, logically, it first identifies... Then calculate α, and finally update the weight matrix; when Automatic correction will be initiated when the threshold value exceeds the set threshold (e.g., 0.5).

[0131] Among them, the construction of neighborhood node relationships in spatial graph convolution is achieved through Delaunay triangulation, which connects the center points of the channel grid cells into a graph structure to ensure that the topology adapts to terrain changes.

[0132] S1062. Determine the eddy viscosity coefficient based on the modified terrain weight, and generate the modified eddy viscosity coefficient.

[0133] Eddy viscosity coefficient The original formula, used to quantify turbulence intensity, is:

[0134]

[0135] in, H represents turbulent viscosity; H represents water depth. This represents the modulus of the water depth gradient.

[0136] The generation of the corrected eddy viscosity coefficient is achieved by correcting the terrain weights, including:

[0137] Terrain gradient Perform neighborhood weighting, that is, based on each node's... Calculate the weighted average value:

[0138]

[0139] in, Let represent the magnitude of the water depth gradient at point j.

[0140] Then update the eddy viscosity coefficient as follows:

[0141] =0.07

[0142] Eddy viscosity amplifies the effects of topographic relief on turbulence; for example, the weighted gradient is higher in curved regions. Increase and improve the accuracy of reflux simulation.

[0143] In the process of power flow modeling, the weighted gradient is first generated, and then substituted into the formula to calculate the new coefficients.

[0144] S1063. Based on the grouped settling velocity, update the bedload weight coefficient and generate the updated weight coefficient.

[0145] Group Settlement Velocity The calculation is based on the modified formula of Stokes' law:

[0146]

[0147] in, ν represents the settling velocity of the k-th particle size group; g represents the acceleration due to gravity; ν represents the dynamic viscosity of the water. and These represent the densities of sediment and water, respectively. This represents the square of the particle size in the k-th group.

[0148] The updated bedload weighting coefficients include: calculating the proportion of grouped settlement velocities to the total settlement velocity.

[0149]

[0150] in, The proportion of grouped settlement velocity to total settlement velocity. This represents the settling velocity of the k-th particle size group.

[0151] Then multiply the ratio by the original displacement weight coefficient to generate the updated weight coefficient:

[0152]

[0153] in, This represents the component coupling coefficient.

[0154] For example, when medium to coarse sand settles quickly, Increase the weight of bed surface transport.

[0155] S107. Update the multiphysics coupling model using the modified eddy viscosity coefficient and the updated weight coefficient.

[0156] The corrected eddy viscosity coefficient and updated weighting coefficients are input into the multiphysics coupled model (including tidal current field, wave field, and sediment transport field) to replace the original parameters. After the model is updated, the sediment transport rate and hydrodynamic field are recalculated; for example, the corrected... Improve eddy current resolution and optimize particle size response.

[0157] Furthermore, channel evolution simulation and feedback optimization will be applied to the predicted siltation thickness output by ST-GCN. and sediment distribution heat map The input channel digital twin platform generates a 3D siltation evolution animation, and the predicted results are compared with the actual sensor data to trigger closed-loop optimization:

[0158] If the water depth variation error is | H-Δ If |>ε (where ε=0.1m), then the spatiotemporal convolution kernel weights of ST-GCN are adjusted in reverse.

[0159] If the predicted sediment content is biased | - If |>δ, then update the sediment parameterization scheme of the coupled model.

[0160] Where ε is the threshold for water depth variation error, and δ is the threshold for sediment concentration prediction deviation.

[0161] The incremental learning mechanism is activated when the model error decreases by less than 5% after five consecutive fine-tunings. It automatically adds a residual branch and constrains parameter updates using the Elastic Weight Fixation (EWC) algorithm. The loss function expression is:

[0162]

[0163] in, Let λ represent the Fisher information matrix, and λ represent the weight penalty coefficient. L(B) represents the old model parameter values, L(B) represents the loss function on the new task, and B represents the model parameters. This represents the i-th parameter.

[0164] This mechanism forces the model to maintain the stability of key parameters of old tasks when learning new tasks, thereby achieving stable knowledge transfer and dynamic accumulation.

[0165] This application also provides a method for the coordinated control of channel scouring and sedimentation and evolution, including:

[0166] The verification module calculates the correlation coefficient of the sediment transport potential field based on the sediment transport potential field and the change in water depth.

[0167] If the absolute value of the correlation coefficient of the sediment transport potential field is less than the first threshold, the verification is determined to have failed.

[0168] The gradation module calculates the grouped and ungrouped sediment transport rates based on real-time particle size distribution.

[0169] Based on the grouped sediment transport rate and the ungrouped sediment transport rate, the particle size sediment transport rate deviation is calculated;

[0170] If the deviation of the particle size transport rate is greater than the second threshold, the gradation is determined to be abnormal.

[0171] The coefficient module performs the following operations when the verification failure and the gradation anomaly occur simultaneously: based on the maximum proportion of the real-time particle size distribution, corrects the terrain weight of the spatial map convolution to generate corrected terrain weight; determines the eddy viscosity coefficient according to the corrected terrain weight to generate corrected eddy viscosity coefficient; and updates the bedload weight coefficient according to the grouped settling velocity to generate updated weight coefficient.

[0172] The update module updates the multiphysics coupling model using the corrected eddy viscosity coefficient and the updated weight coefficient.

[0173] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0174] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the above-described method.

[0175] The above description of the embodiments is provided to enable those skilled in the art to understand and apply this application. Those skilled in the art will readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without inventive effort. Therefore, this application is not limited to the above embodiments, and any improvements and modifications made to this application based on the disclosure thereof should be within the scope of protection of this application.

Claims

1. A method for the coordinated control of channel scouring and evolution, characterized in that, The method comprises the following steps: According to the sediment transport potential field and the water depth variation, a sediment transport potential field correlation coefficient is calculated, the sediment transport potential field , is the water depth variation; for calculating divergence, is a unit width sediment discharge rate vector, the sediment transport potential field correlation coefficient , is the water depth variation; If the absolute value of the sediment transport potential field correlation coefficient is less than a first threshold value, it is determined that the verification fails; According to the real-time particle size distribution, the grouped sediment transport rate and the ungrouped sediment transport rate are calculated, and the grouped sediment transport rate is used to calculate the sediment transport rate of each group : ； The non-group sand rate : = + ； the proportion of the kth group of particle sizes in the total sediment, is the representative particle size of the kth group; k is the particle size group index, and denotes the component coupling coefficient; denotes the suspended load sediment transport rate; denotes the bed load sediment transport rate; Based on the grouped sediment transport rate and the ungrouped sediment transport rate, a grain size sediment transport rate bias is calculated, the grain size sediment transport rate bias ; If the particle size sediment transport rate deviation amount is greater than a second threshold value, it is determined that the gradation is abnormal; When the check fails and the gradation anomaly occurs simultaneously, the following operations are performed: based on the maximum proportion value of the real-time particle size distribution, the terrain weight of spatial graph convolution is corrected to generate a corrected terrain weight , is a symbol amplification coefficient generated according to the bed elevation difference symbol and the maximum proportion value, is a terrain weight; according to the corrected terrain weight, the eddy viscosity coefficient is determined to generate a corrected eddy viscosity coefficient =0.07 , is a weighted terrain gradient generated by neighborhood weighting of the terrain gradient through the corrected terrain weight, and H represents water depth, is a module length of the water depth gradient; according to the grouped settling velocity, the weight coefficient of the bed load is updated to generate an updated weight coefficient , is the proportion of the grouped settling velocity to the total settling velocity, is a component coupling coefficient; The multi-physics field coupling model is updated by using the corrected eddy viscosity coefficient and the updated weight coefficient, wherein the multi-physics field coupling model comprises a tidal current field, a wave field and a sediment transport field.

2. The method of claim 1, wherein, Based on the maximum proportion value of the real-time particle size distribution, the terrain weight of the spatial graph convolution is corrected, comprising: When the maximum proportion value exceeds a set threshold value, the terrain weight correction operation is started.

3. The method of claim 1, wherein, The spatial graph convolution comprises: The neighborhood node relationship construction comprises connecting nodes by Delaunay triangulation to generate a neighborhood set.

4. The method of claim 1, wherein, The determination of the maximum proportion value of the real-time particle size distribution comprises: An advantage particle size group is dynamically identified by using a maximum value selection function.

5. A device for the coordinated control of channel scouring and evolution, characterized in that, The method comprises the following steps: The check module calculates a sediment transport potential field correlation coefficient according to the sediment transport potential field and the water depth variation, and the sediment transport potential field , is the water depth variation; and is a unit width sediment discharge rate vector, the sediment transport potential field correlation coefficient , is the water depth variation. If the absolute value of the sediment transport potential field correlation coefficient is less than a first threshold value, it is determined that the verification fails; The grading module calculates the grouped sediment transport rate and the ungrouped sediment transport rate according to the real-time grain size distribution : ； The non-group sand rate : = + ； the proportion of the kth group of particle sizes in the total sediment, kth group representative particle size; k is the particle size group index, and denotes the component coupling coefficient; denotes the suspended load sediment transport rate; denotes the bed load sediment transport rate; Based on the grouped sediment transport rate and the ungrouped sediment transport rate, a grain size sediment transport rate bias is calculated, the grain size sediment transport rate bias ; If the particle size sediment transport rate deviation amount is greater than a second threshold value, it is determined that the gradation is abnormal; The coefficient module corrects the terrain weight of the spatial graph convolution based on the maximum proportion value of the real-time particle size distribution when the check fails and the gradation anomaly occurs simultaneously to generate a corrected terrain weight , is a symbol amplification coefficient generated according to the bed elevation difference symbol and the maximum proportion value, is a terrain weight; the eddy viscosity coefficient is determined according to the corrected terrain weight to generate a corrected eddy viscosity coefficient = 0.07 , is a weighted terrain gradient generated by neighborhood weighting of the terrain gradient through the corrected terrain weight, and H represents the water depth, is the module length of the water depth gradient; the weight coefficient of the bed load is updated according to the grouped settling velocity to generate an updated weight coefficient , is the proportion of the grouped settling velocity to the total settling velocity, is a component coupling coefficient; The multi-physics field coupling model is updated by using the corrected eddy viscosity coefficient and the updated weight coefficient, wherein the multi-physics field coupling model comprises a tidal current field, a wave field and a sediment transport field.

6. An electronic device, comprising: The method comprises the following steps:

7. A computer-readable storage medium, characterized in that, The computer program is stored in the memory and the processor executes the computer program to realize the method of any one of claims 1-4. The computer program is stored on the computer and when the computer program is executed in the computer, the computer executes the method of any one of claims 1-4.

Images