A Comprehensive Solution Method for Flow Velocity in Complex Cross-Section Open Channels under a Multi-Point Launch Mechanism

Abstract

This invention relates to the field of numerical simulation modeling technology, specifically to a comprehensive solution method for flow velocity in complex open channels under a multi-point emission mechanism. In this invention, by jointly grouping, density-statistically analyzing, and sequence-filtering open channel cross-sectional parameters, disturbance source distribution characteristics, and water level gradient data, a dynamic partitioning process driven by structural parameters is established, constructing a lattice structure that better fits the characteristics of non-uniform boundaries. By extending and layering the disturbance projection direction and propagation path, and integrating a graph neural network to extract the correlation between disturbance nodes and construct the propagation path graph, the local propagation offset problem in particle propagation is effectively eliminated. By gradually constructing a multi-scale disturbance response mechanism and dynamic synchronous propagation logic, the continuity of multi-point disturbance response and the coordination of particle propagation paths in the open channel are ensured, significantly enhancing the ability to simulate the distribution of irregular boundaries and dynamic disturbance fields.

Description

Technical Field

[0001] This invention relates to the field of numerical simulation modeling technology, and in particular to a method for comprehensively solving the flow velocity of open channels with complex cross sections under a multi-point emission mechanism. Background Technology

[0002] The field of numerical simulation modeling technology mainly involves using computers to model and simulate natural systems. The goal is to accurately describe physical processes and discretize them into mathematical models, and then complete the simulation solution on a digital platform. The field of numerical simulation modeling usually combines partial differential equation modeling theory, finite element numerical methods, mesh generation technology and high-performance computing resources to solve complex boundary and multivariate coupling problems in natural systems.

[0003] A comprehensive solution method for flow velocity in open channels with complex cross-sections under a multi-point emission mechanism is proposed. By constructing a multi-point emission mechanism, multiple emission points are selected in the computational domain as initial disturbance sources and boundary drivers to achieve a comprehensive characterization of complex water flow behavior. The aim is to establish a flow field model applicable to irregular boundary conditions to improve the accuracy of open channel flow velocity distribution prediction. By introducing multi-point disturbances and joint solutions, the problem of insufficient response to local velocity anomalies under single-point boundary driving is overcome, and continuous, stable, and high-resolution simulation of the flow velocity state of the entire cross-section is achieved.

[0004] Existing technologies focus on perturbation modeling based on single-level boundary conditions and a unified launch point, which easily leads to problems such as response blind zones and failure to identify synchronous jumps. They lack differentiated response processing for the dynamic evolution of perturbation paths and particle propulsion processes, resulting in particle accumulation, misalignment of propulsion sequence, and failure to identify jump peaks in complex cross sections. Since the propulsion path does not establish parallel control for time difference and spatial expansion, it often results in abnormally high or low local flow velocity values, affecting the overall flow field calculation accuracy, causing the simulation structure to lose consistency, and reducing the robustness and applicability of the overall simulation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a comprehensive solution method for flow velocity in open channels with complex cross-sections under a multi-point launch mechanism.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for comprehensively solving the flow velocity of open channels with complex cross-sections under a multi-point emission mechanism, comprising the following steps:

[0007] Step 1: Based on the open channel cross-sectional structural parameters, the number of disturbance sources and the gradient fluctuation coefficient, calculate the ratio of cross-sectional coordinates to area, group the differences between the distance between disturbance sources and the flow inclination, statistically analyze the density of water level gradient and flow velocity intersection points, and generate a dynamic zoning structure delineation map of the watershed.

[0008] Step 2: Based on the aforementioned watershed dynamic zoning structure delineation map, cross-compare the velocity and water level sequences of the main stream section, filter the displacement and velocity angles of the boundary sections, and sort the boundary slope angle and depth jump points to generate a set of differential character-driven structures.

[0009] Step 3: Based on the differential character-driven structure set, extend the coordinates and projection direction of the disturbance source, use a graph neural network to analyze the correlation between the nodes of the disturbance path and construct a mutation propagation structure diagram, and sort the disturbance amplitude and propagation change rate hierarchically to generate a disturbance effect area grid synchronization instruction table.

[0010] Step 4: Based on the perturbation zone grid synchronization instruction table, obtain the particle propagation timing and propulsion difference, sort the propulsion time difference matrix of overlapping boundaries, match and number the propulsion direction density blocks, and generate a particle propulsion sequence parallel computing scheduling table.

[0011] Step 5: Based on the parallel computation scheduling table of the particle propulsion sequence, extract the difference between the state transition and propulsion cycle of the boundary segment, rearrange the propulsion order of the mutation zone, reconstruct the synchronous propulsion section of the overlapping boundary, and generate a multi-zone lattice state coupling propulsion structure diagram;

[0012] The specific steps for generating the dynamic watershed partitioning structure map are as follows:

[0013] Based on the open channel cross-sectional structural parameters, the number of disturbance sources, and the slope fluctuation coefficient, the segmental spacing between the cross-sectional coordinate point groups is determined and the area of ​​the contour closed curve is derived. The difference between the valley value of the length of the line connecting the disturbance sources and the angle between the adjacent water flow direction is extracted and the interval is classified to generate the cross-section and disturbance relationship extraction results.

[0014] Based on the extraction results of the relationship between the cross section and the disturbance, the elevation difference and gradient ratio between water level sampling points in the longitudinal slope area are statistically analyzed, the density distribution of the intersection points of the lateral velocity vectors is identified, and the overlapping areas of the turning section are screened for comparison, and a dynamic zoning structure delineation map of the watershed is generated.

[0015] The specific steps for generating the differential character-driven structure set are as follows:

[0016] Based on the aforementioned watershed dynamic zoning structure delineation map, the water level records at each cross-section in the main stream section are aligned with the velocity records at the corresponding time points. The differences between the water level and velocity at the same location in adjacent time steps are subtracted, the trend comparison is calculated, and the matching point sequences are labeled to generate a main stream slope change comparison set.

[0017] Based on the slope change comparison set of the main segment, the angle between the trajectory line of each control point in the boundary segment and the adjacent velocity vector is extracted to obtain control points with an angle smaller than the specified angle range. At the same time, the slope curvature change value of the control point is extracted and sorted with the depth difference sequence to generate a differential character-driven structure set.

[0018] The specific steps for generating the synchronization instruction table for the disturbance area grid are as follows:

[0019] Based on the differential grid cell-driven structure set, the coordinate points of the disturbance source are bound to the number of the starting line segment direction of the grid boundary. Line segments with fixed step sizes are added from the bound points along the diagonal direction. The coordinates of the displacement nodes are recorded, and the cumulative length and endpoint of the line segments are calculated and marked to generate a disturbance source extension path dataset.

[0020] Based on the disturbance source extension path dataset, the local velocity data of each extension node is subtracted from the disturbance frequency change value of the adjacent node to generate a difference sequence. A graph neural network is used to extract the disturbance correlation between nodes and construct a jump propagation path map to assist in matching the frequency transfer trajectory between nodes with a sudden change threshold. Nodes in each difference sequence that are greater than the jump threshold are paired in position, and the direction tracking mark is bound to the corresponding path identification code to generate a disturbance sudden change node matching set.

[0021] Based on the disturbance mutation node matching set, the peak disturbance amplitude of each node is extracted and the difference between the propagation speed of the nodes in the downstream propagation path is extracted. The amplitude values ​​in each segment are sorted, the speed difference is processed, the flag bit is updated and the target grid number is synchronized and recombined to generate the grid synchronization instruction table of the disturbance effect area.

[0022] As a further aspect of the present invention, the watershed dynamic zoning structure delineation map includes zoning boundary lines, disturbance response zone labels, and hydraulic slope change zone identification codes; the differential particle-driven structure set includes grid level identifiers, boundary difference zone numbers, and jump-related segment indexes; the disturbance effect zone grid synchronization instruction table includes node matching path numbers, synchronization propulsion level codes, and disturbance response grid indexes; the particle propulsion sequence parallel computing scheduling table includes propulsion task segment sequences, parallel sub-block numbers, and scheduling time step mapping tables; and the multi-zone grid state coupling propulsion structure diagram includes grid boundary connection maps, coupling propulsion channel mappings, and synchronization state marker fields.

[0023] As a further aspect of the present invention, the graph neural network extracts the perturbation correlation between nodes and constructs a jump propagation path graph. First, the extended nodes are regarded as independent nodes in the graph structure. A directed edge connection structure is established based on the spatial adjacency relationship and frequency change sequence between nodes. The perturbation frequency difference and spatial displacement between nodes are used as the basis for edge weight assignment. After generating the graph structure, the perturbation correlation index between each node is iteratively calculated, and the strongly correlated node pairs in the perturbation propagation chain are extracted to construct a frequency jump path index graph.

[0024] As a further aspect of the present invention, the addition of a fixed step length of line segments along the diagonal direction specifically refers to determining the diagonal extension direction with a relative peak angle offset in the grid structure, taking the direction of the initial line segment at the boundary of the grid where the disturbance source is located as the reference direction, and extending outward in the direction. In the two-dimensional coordinate system, the same value is added simultaneously in the horizontal and vertical directions as a fixed step length. A line segment consistent with the initial direction is drawn at each step length position. The starting point of the line segment is the ending point of the previous node, and the ending point is the coordinate point of the current step length. This process is repeated until the boundary of the region is reached and the preset distance condition is met. The starting and ending coordinates of the line segment must be recorded, and the total length and ending point information are updated based on the cumulative length of each step.

[0025] As a further aspect of the present invention, the specific steps for generating the parallel computation scheduling table for the particle propulsion sequence are as follows:

[0026] Based on the grid synchronization instruction table of the disturbance area, the grid number of each synchronization entry in the table is extracted and the time field is read. The number is extracted by reading the instruction line and a timestamp is generated and time-labeled. The particle propulsion sequence difference and node time difference are calculated to generate a particle propulsion timing difference list.

[0027] Based on the particle propulsion timing difference list, match the particle set and boundary unit number corresponding to the timing difference entry, read the particle ID and unit number mapping and identify the time difference aggregation and intersection areas within the group, sort the row and column indices of the timing difference matrix, and generate a boundary propulsion difference sorting matrix set.

[0028] Based on the boundary advancement difference sorting matrix set, the frequency of particle numbers at the matrix intersections is counted, the number list of each intersection is counted and the frequency continuous segment is identified, the block number allocation and advancement order are adjusted, and a particle advancement sequence parallel computing scheduling table is generated.

[0029] As a further aspect of the present invention, the particle propulsion order difference refers to the order difference between the time tags corresponding to different particles on their respective propulsion paths in particle propulsion scheduling. Specifically, it is based on the time field read by the particle in the synchronization instruction table to generate particle propulsion time tags, and then sorting the particles according to the grid number and propulsion path to calculate the difference in propulsion tag values ​​between adjacent particles.

[0030] As a further aspect of the present invention, the specific steps for generating the multi-zone lattice state coupling propulsion structure diagram are as follows:

[0031] Based on the parallel computing scheduling table of the particle propulsion sequence, the continuity of particle state labels and propulsion cycle marker fields in each group of boundary segments is read, the periodic sequence between the propulsion start node and the end node is extracted and the difference range is compared one by one, the peak and valley differences during the period are identified and the propulsion offset is sorted to generate a boundary segment propulsion difference dataset.

[0032] Based on the boundary segment advancement difference dataset, the particle number order and time period index value combination within the advancement offset segment are rearranged, the node number of the continuous mutation zone is located and the synchronous advancement number list is extracted. The advancement segment reorganization and the synchronous grid block path of the boundary coupling zone are combined to generate a multi-zone grid state coupled advancement structure diagram.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] 1. In this invention, by jointly grouping, density statistics and sequence screening the open channel cross-sectional parameters, disturbance source distribution characteristics and water level gradient data, a dynamic partitioning process driven by structural parameters is established, and a grid structure that better fits the characteristics of non-uniform boundaries is constructed, which achieves significant gains in the overall coordination accuracy of the velocity calculation field and the calculation efficiency of the grid system.

[0035] 2. In this invention, by extending and layering the perturbation projection direction and propagation path, and integrating graph neural networks to extract the correlation between perturbation nodes and construct the propagation path graph, the path recognition process becomes more accurate and dynamically adaptable, effectively eliminating the problem of local propulsion deviation in particle propulsion.

[0036] 3. In this invention, by gradually constructing a multi-scale disturbance response mechanism and a dynamic synchronous propagation logic, the continuity of the multi-point disturbance response and the coordination of particle propagation paths in the open channel are ensured, which significantly enhances the ability to simulate the distribution of irregular boundaries and dynamic disturbance fields. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the main steps of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0039] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0040] Example

[0041] Please see Figure 1 This invention provides a technical solution: a method for comprehensively solving the flow velocity of open channels with complex cross-sections under a multi-point emission mechanism, comprising the following steps:

[0042] Step 1: Based on the open channel cross-sectional structural parameters, the number of disturbance sources and the gradient fluctuation coefficient, calculate the ratio of cross-sectional coordinates to area, group the differences between the distance between disturbance sources and the flow inclination, statistically analyze the density of water level gradient and flow velocity intersection points, and generate a dynamic zoning structure delineation map of the watershed.

[0043] Step 2: Based on the dynamic zoning structure delineation map of the watershed, cross-compare the velocity and water level sequences of the main stream section, filter the displacement and velocity angles of the boundary sections, and sort the boundary slope angle and depth jump points to generate a set of differential character-driven structures.

[0044] Step 3: Based on the differential character-driven structure set, extend the coordinates and projection direction of the disturbance source, use a graph neural network to analyze the correlation between the nodes of the disturbance path and construct a mutation propagation structure graph, and sort the disturbance amplitude and propagation change rate hierarchically to generate a grid synchronization instruction table for the disturbance effect area.

[0045] Step 4: Based on the perturbation action area grid synchronization instruction table, obtain the particle propagation timing and propulsion difference, sort the propulsion time difference matrix of overlapping boundaries, match and number the propulsion direction density blocks, and generate a particle propulsion sequence parallel computing scheduling table;

[0046] Step 5: Based on the parallel computation scheduling table of particle propulsion sequence, extract the difference between the state transition and propulsion period of the boundary segment, rearrange the propulsion order of the mutation zone, reconstruct the synchronous propulsion section of the overlapping boundary, and generate a multi-zone lattice state coupling propulsion structure diagram;

[0047] The specific steps for generating a dynamic watershed zoning structure map are as follows:

[0048] Based on the open channel cross-sectional structural parameters, the number of disturbance sources, and the slope fluctuation coefficient, the segmental spacing between the cross-sectional coordinate point groups is determined and the area of ​​the contour closed curve is derived. The difference between the valley value of the length of the line connecting the disturbance sources and the angle between the adjacent water flow direction is extracted and the interval is classified to generate the cross-section and disturbance relationship extraction results.

[0049] Based on the results of the cross-section and disturbance relationship extraction, the elevation difference and gradient ratio between water level sampling points in the longitudinal slope area are statistically analyzed, the density distribution of the intersection points of the lateral velocity vectors is identified, and the overlapping areas of the turning section locations are screened for comparison, generating a dynamic zoning structure delineation map of the watershed.

[0050] The specific steps for generating the differential character-driven structure set are as follows:

[0051] Based on the dynamic zoning structure map of the watershed, the water level records at each cross section in the main stream are aligned with the velocity records at the corresponding time points. The differences between the water level and velocity at the same location in adjacent time steps are subtracted to calculate the trend comparison and label the matching point sequence to generate a comparison set of the slope changes in the main stream.

[0052] Based on the mainstream slope change comparison set, the angle between the trajectory line of each control point in the boundary segment and the adjacent velocity vector is extracted to obtain control points with an angle smaller than the specified angle range. At the same time, the slope curvature change value of the control point is extracted and sorted with the depth difference sequence to generate a differential character-driven structure set.

[0053] The specific steps for generating the synchronization command table for the disturbance area grid are as follows:

[0054] Based on the differential grid cell-driven structure set, the coordinate points of the disturbance source are bound to the number of the starting line segment direction of the grid boundary. Line segments with fixed step sizes are added from the bound points along the diagonal direction. The coordinates of the displacement nodes are recorded, and the cumulative length and endpoint of the line segments are calculated and marked to generate a disturbance source extension path dataset.

[0055] Based on the disturbance source extension path dataset, the local velocity data of each extension node is subtracted from the disturbance frequency change value of the adjacent node to generate a difference sequence. A graph neural network is used to extract the disturbance correlation between nodes and construct a jump propagation path map to help match the frequency transfer trajectory between nodes with a sudden change threshold. Nodes in each difference sequence that are greater than the jump threshold are paired in position, and the direction tracking mark is bound to the corresponding path identification code to generate a disturbance sudden change node matching set.

[0056] Based on the perturbation mutation node matching set, the peak value of the perturbation amplitude corresponding to each node and the difference between the propagation speed of the nodes in the downstream propagation path are extracted, and the amplitude values ​​in each segment are sorted. The speed difference is processed and the flag bit is synchronously updated and the target grid number is recombined to generate the grid synchronization instruction table of the perturbation effect area.

[0057] The watershed dynamic zoning structure delineation map includes zoning boundary lines, disturbance response zone labels, and hydraulic slope change zone identification codes. The differential particle-driven structure set includes grid level identifiers, boundary difference zone numbers, and jump-related segment indexes. The disturbance effect zone grid synchronization instruction table includes node matching path numbers, synchronization propulsion level codes, and disturbance response grid indexes. The particle propulsion sequence parallel computing scheduling table includes propulsion task segment sequences, parallel sub-block numbers, and scheduling time step mapping tables. The multi-zone grid state coupling propulsion structure diagram includes grid boundary connection maps, coupled propulsion channel mappings, and synchronization state marker fields.

[0058] The specific steps for generating a dynamic watershed zoning structure map are as follows:

[0059] Based on the open channel cross-sectional structural parameters, the number of disturbance sources, and the slope fluctuation coefficient, the segmental spacing between the cross-sectional coordinate point groups is determined and the area of ​​the contour closed curve is derived. The difference between the valley value of the length of the line connecting the disturbance sources and the angle between the adjacent water flow direction is extracted and the interval is classified to generate the cross-section and disturbance relationship extraction results.

[0060] Based on the results of the cross-section and disturbance relationship extraction, the elevation difference and gradient ratio between water level sampling points in the longitudinal slope area are statistically analyzed, the density distribution of the intersection points of the lateral velocity vectors is identified, and the overlapping areas of the turning section locations are screened for comparison, generating a dynamic zoning structure delineation map of the watershed.

[0061] Based on the structural parameters of the open channel cross section, the number of disturbance sources, and the slope fluctuation coefficient, the polygon node set of the cross section is obtained using complex coordinate representation. After the nodes are numbered sequentially, the contour closure curve is calculated according to Green's formula. The integration starting point is set as the lowest numbered point, and the calculation is performed in a clockwise direction. The calculated area is equal to half of the sum of the differences between the horizontal and vertical coordinate products of the node pairs. Then, the boundary envelope is defined by the elevation difference to generate the cross section and disturbance relationship extraction results.

[0062] Based on the results of the cross-section and disturbance relationship extraction, a two-dimensional elevation vector array is constructed for the sampling points in the longitudinal slope interval. The two-dimensional vector array composed of elevation difference and gradient ratio is clustered. The K value is set to 5, and the initial center point is constructed from the position of maximum gradient. After completing the classification of each type of sample, overlapping sections are screened. Then, a two-dimensional density analysis method is used to construct a planar distribution density map of the intersection points of the lateral velocity vectors using kernel density estimation. The kernel function is set to Gaussian kernel, and the bandwidth parameter is set to 0.8. The coordinate points exceeding the density threshold are screened for overlapping positions of the turning section, and a dynamic zoning structure delineation map of the watershed is generated.

[0063] The specific steps for generating the differential character-driven structure set are as follows:

[0064] Based on the dynamic zoning structure map of the watershed, the water level records at each cross section in the main stream are aligned with the velocity records at the corresponding time points. The differences between the water level and velocity at the same location in adjacent time steps are subtracted to calculate the trend comparison and label the matching point sequence to generate a comparison set of the slope changes in the main stream.

[0065] Based on the mainstream slope change comparison set, the angle between the trajectory line of each control point in the boundary segment and the adjacent velocity vector is extracted to obtain control points with an angle smaller than the specified angle range. At the same time, the slope curvature change value of the control point is extracted and sorted with the depth difference sequence to generate a differential character-driven structure set.

[0066] Based on the dynamic zoning structure map of the watershed, a dual-channel time series array containing water level and velocity values ​​is constructed for each cross-section location in the main stream. After aligning the water level and velocity sequences, the difference between the same cross-section location at adjacent time steps is calculated. The time interval is set to 5 seconds, and the data precision is retained to two decimal places. The difference values ​​of each group are calculated and a trend difference vector is constructed. Then, a dynamic time warping algorithm is used to construct a mapping for the matching points in the trend difference vector. A search path range with a constraint window of 6 is used, and the matching point locations are written into the matching array. Finally, a unique label value is assigned to each point in the matching array to construct a one-dimensional label sequence and generate a comparison set of the slope changes in the main stream.

[0067] Based on the mainstream slope change comparison set, the trajectory line coordinate sequence of the control points of the boundary segment is read and the tangential derivative is calculated. The angle between the derivative and the velocity vector of the adjacent node is calculated by dividing the dot product by the inverse cosine of the product of magnitude and length. The angle threshold is set to 30 degrees, and only the control point index less than the threshold is retained. At the same time, the curvature difference calculation method is used to perform second-order difference operation on the trajectory line of the retained points to construct a curvature sequence. After normalization with the maximum value, the curvature is sorted, and the sorted index and the depth difference are merged in coordinate order and joint sorting is performed. The priority rule is that the curvature order comes first and the depth order comes second, generating a differential character-driven structure set.

[0068] The specific steps for generating the synchronization command table for the disturbance area grid are as follows:

[0069] Based on the differential grid cell-driven structure set, the coordinate points of the disturbance source are bound to the number of the starting line segment direction of the grid boundary. Line segments with fixed step sizes are added from the bound points along the diagonal direction. The coordinates of the displacement nodes are recorded, and the cumulative length and endpoint of the line segments are calculated and marked to generate a disturbance source extension path dataset.

[0070] Based on the disturbance source extension path dataset, the local velocity data of each extension node is subtracted from the disturbance frequency change value of the adjacent node to generate a difference sequence. A graph neural network is used to extract the disturbance correlation between nodes and construct a jump propagation path map to help match the frequency transfer trajectory between nodes with a sudden change threshold. Nodes in each difference sequence that are greater than the jump threshold are paired in position, and the direction tracking mark is bound to the corresponding path identification code to generate a disturbance sudden change node matching set.

[0071] Based on the perturbation mutation node matching set, the peak value of the perturbation amplitude corresponding to each node and the difference between the propagation speed of the nodes in the downstream propagation path are extracted, and the amplitude values ​​in each segment are sorted. The speed difference is processed and the flag bit is updated synchronously and the target grid number is recombined to generate the grid synchronization instruction table of the perturbation effect area.

[0072] Based on the differential character-driven structure set, the starting line segment of the grid boundary where the coordinate point of the disturbance source is located is located. The line segment direction numbers 0 to 7 correspond to the main diagonal and secondary diagonal directions. After binding the numbers, the directional addition is performed according to the specified step size of 3 units. The endpoints of subsequent line segments are generated sequentially using an incremental iteration method. The coordinates of each displacement node are recorded and the cumulative length value is updated after generation. The termination condition is set as the cumulative length being greater than or equal to 15 units and encountering a boundary section. The endpoint coordinates and cumulative length information are written into the path array at the same time to generate the disturbance source extension path dataset.

[0073] Based on the dataset of extended paths of disturbance sources, a graph neural network method is used to construct a node graph structure. The local velocity data of each extended node and the disturbance frequency change value of adjacent nodes are used as input feature tensors to construct an adjacency matrix to connect upstream and downstream nodes in the path. A GCNConv layer is defined with 2 layers and ReLU activation function. The number of training rounds of the graph network is set to 300 rounds, the learning rate is 0.005, and the loss function is the mean squared error function. Forward propagation is performed on the input graph data to output the disturbance correlation matrix. Jump connection edges are filtered according to the maximum response value between nodes to construct a frequency jump propagation path graph. Nodes with a difference value greater than the specified jump threshold of 0.6 in each segment difference sequence are paired sequentially according to the propagation path graph. The paired nodes are labeled with direction vectors and path identification codes to generate a set of disturbance mutation node matching.

[0074] Based on the perturbation mutation node matching set, the amplitude peak value of each mutation node within the third time step range is extracted. A local maximum function with a sliding window length of 3 is used to identify the peak index of each segment. Then, the propagation speed is calculated node by node in the downstream path. The speed difference is the speed of the adjacent node minus the speed of the previous node. The amplitude peak values ​​in the same path are sorted by size. At the same time, the difference values ​​are classified by standard deviation. The interval boundaries are set to 0.2, 0.5 and 0.8. A hierarchical index is performed. The node array is labeled and updated according to the level. Then, the corresponding grid number is synchronously written to generate the perturbation effect area grid synchronization instruction table.

[0075] The graph neural network extracts the perturbation correlation between nodes and constructs a jump propagation path graph. First, the extended nodes are regarded as independent nodes in the graph structure. A directed edge connection structure is established based on the spatial adjacency relationship and frequency change sequence between nodes. The perturbation frequency difference and spatial displacement between nodes are used as the basis for edge weight assignment. After generating the graph structure, the perturbation correlation index between each node is iteratively calculated. Strongly correlated node pairs in the perturbation propagation chain are extracted, and a frequency jump path index graph is constructed.

[0076] Graph neural networks follow the formula:

[0077]

[0078] in: The graph neural network represents the first... The first in the layer The result of updating the embedding vector of each node. The graph neural network represents the first... The first in the layer Embedding of adjacent nodes, Indicates the first The set of adjacent nodes of a node. The graph neural network represents the first... Layer weight matrix This represents the activation function. Indicates the first The node and the first The perturbation velocity difference factor between adjacent nodes Indicates the first The node and the first The significance coefficient of frequency jumps between adjacent nodes Indicates the first The node and the first Path direction consistency factor among adjacent nodes Indicates the first The node and the first Normalized constants between adjacent nodes.

[0079] Execution process: First, the open channel cross-section is discretized into multiple sets of disturbance source nodes, and the embedded representation of each node at the current time is extracted. It includes attribute values ​​such as speed, frequency, and direction, and iterates through the nodes. Adjacent nodes The velocity difference factor is calculated by normalizing the difference in propagation velocity between two nodes after normalizing it to the maximum velocity difference. The frequency jump factor is calculated by dividing the standard deviation of the perturbation frequency change by a specified jump threshold. The path direction consistency factor is calculated by using the cosine of the angle between the perturbation direction vector and the mainstream direction. The product of these three factors is used as the weight of the adjacent edge, and then divided by the normalization term. To balance the impact of adjacency density, each weight is multiplied by the characteristics of the adjacent nodes in the previous layer. Then summation is performed, and the activation function is input. Complete the first Layer node representation The update, after multiple layers of propagation, forms a jump propagation path diagram.

[0080] Specifically, adding a fixed step size line segment along the diagonal direction refers to determining the diagonal extension direction with a relative peak angle offset in the lattice structure, using the direction of the initial line segment at the boundary of the lattice where the disturbance source is located as the reference direction, and extending outward in the direction. In the two-dimensional coordinate system, the same value is added simultaneously in the horizontal and vertical directions as a fixed step size. At each step size position, a line segment consistent with the initial direction is drawn. The starting point of the line segment is the ending point of the previous node, and the ending point is the coordinate point of the current step size. This process is repeated until the boundary of the region is reached and the preset distance condition is met. The starting and ending coordinates of the line segment must be recorded, and the total length and ending point information are updated based on the cumulative length of each step.

[0081] The specific steps for generating the parallel computation scheduling table for particle propulsion sequences are as follows:

[0082] Based on the grid synchronization instruction table of the disturbance area, the grid number of each synchronization entry in the table is extracted and the time field is read. The number is extracted by reading the instruction line and a timestamp is generated and time-labeled. The particle propulsion sequence difference and node time difference are calculated to generate a particle propulsion timing difference list.

[0083] Based on the particle propulsion timing difference list, match the particle set and boundary unit number corresponding to the timing difference entry, read the particle ID and unit number mapping and identify the time difference aggregation and intersection areas within the group, sort the row and column indices of the timing difference matrix, and generate a set of boundary propulsion difference sorting matrices.

[0084] Based on the boundary-driven difference sorting matrix set, the frequency of particle numbers at the matrix intersections is counted, the number list of each intersection is counted and the frequency continuous segment is identified, the block number allocation and advancement order are adjusted, and a particle advancement sequence parallel computing scheduling table is generated.

[0085] Based on the grid synchronization instruction table of the disturbance area, the corresponding grid number and time record value of each synchronization instruction are extracted in sequence. The time records are uniformly converted into standard format time tags, and the advancement time between consecutive numbers is compared after sorting by number. The time difference of each pair of consecutive grids is calculated. Combined with the corresponding particle number identifier, the time difference of each node in the particle advancement path is sorted out and output as a particle advancement timing difference list.

[0086] Based on the particle propulsion timing difference list, the mapping relationship between particle number and propulsion difference is analyzed, the boundary unit number corresponding to the particle set is extracted, the change amplitude and aggregation degree of time difference of multiple particles in the same unit are statistically analyzed, the region with concentrated time difference is recorded, and the convergence and overlap of the propulsion rhythm of multiple particles in a certain region are analyzed. The row and column numbers of each particle path are sorted according to the propulsion difference, and the output is a boundary propulsion difference sorting matrix set.

[0087] Based on the boundary-advance difference sorting matrix set, the total number of times the particle number corresponding to the matrix intersection position appears is counted. The numbering is segmented according to the frequency of the intersection point. The numbering regions with temporal advancement continuity and repetition in the particle group are identified. The regions are divided into independent advancement block units, and the original particle advancement order is structurally adjusted. The particle advancement sequence parallel computing scheduling table is constructed and output according to the arrangement of particle positions and advancement rhythm.

[0088] The particle propulsion order difference refers to the difference in the order of time tags corresponding to different particles on their respective propulsion paths in particle propulsion scheduling. Specifically, it is generated based on the time field read by the particle in the synchronization instruction table, the particles are then sorted according to the grid number and propulsion path, and the difference in propulsion tag value between adjacent particles is calculated.

[0089] The specific steps for generating a multi-zone lattice state-coupled propulsion structure diagram are as follows:

[0090] Based on the parallel computing scheduling table of particle propulsion sequence, the continuity of particle state label and propulsion cycle label fields in each group of boundary segments is read. The periodic sequence between the propulsion start node and the end node is extracted and the difference range is compared one by one. The peak and valley differences during the period are identified and the propulsion offset is sorted to generate a boundary segment propulsion difference dataset.

[0091] Based on the boundary segment advancement difference dataset, the particle numbering order and time period index value combination within the advancement offset segment are rearranged, the node numbering of the continuous mutation zone is located and the synchronous advancement numbering list is extracted. The advancement segment reorganization and the synchronous grid block path of the boundary coupling zone are combined to generate a multi-zone grid state coupled advancement structure diagram.

[0092] Based on the parallel computation scheduling table of particle propulsion sequence, a linear cursor is used to read the particle number sequence and propulsion cycle marker field according to the propulsion time index. The starting position of the field is set as the propulsion start index, and a sliding window interval with an interval step size of 1 is set to perform continuity checks on the cycle value, screen out the propulsion start and end nodes and obtain intermediate sequence data, compare the mean and range of each propulsion cycle, and set the peak and valley recognition threshold value difference between 0.25 and 0.45 as the jump judgment criterion. The nodes matching the condition in the propulsion difference sequence within the cycle segment are classified into secondary intervals and indexed and rearranged. The sorting marker field and the difference table correspond to the numbering and construct a table structure for output, generating a boundary segment propulsion difference dataset.

[0093] Based on the boundary segment advancement difference dataset, the particle number field and the advancement time field are set as joint primary keys. A particle index cache is constructed within the advancement offset segment, and the position numbers of the corresponding row and column matrix are read. The particles are arranged in reverse order from high to low advancement difference according to the advancement trend direction and a sequence mapping is performed. The continuous particle number intervals in the advancement mutation segment are extracted and high-frequency continuous segments are marked. The synchronous advancement number is written into a temporary marker column. The coupling path is divided using the overlap of the inter-cell boundary cross coordinates of the grid cells as the coupling criterion. The advancement path set is constructed according to the coupling relationship and the grid block position is merged. The multi-zone grid state coupled advancement structure diagram is output.

[0094] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for comprehensively solving the flow velocity in a complex cross-section open channel under a multi-point launch mechanism, characterized in that, Includes the following steps: Step 1: Based on the open channel cross-sectional structural parameters, the number of disturbance sources and the gradient fluctuation coefficient, calculate the ratio of cross-sectional coordinates to area, group the differences between the distance between disturbance sources and the flow inclination, statistically analyze the density of water level gradient and flow velocity intersection points, and generate a dynamic zoning structure delineation map of the watershed. Step 2: Based on the aforementioned watershed dynamic zoning structure delineation map, cross-compare the velocity and water level sequences of the main stream section, filter the displacement and velocity angles of the boundary sections, and sort the boundary slope angle and depth jump points to generate a set of differential character-driven structures. Step 3: Based on the differential character-driven structure set, extend the coordinates and projection direction of the disturbance source, use a graph neural network to analyze the correlation between the nodes of the disturbance path and construct a mutation propagation structure diagram, and sort the disturbance amplitude and propagation change rate hierarchically to generate a disturbance effect area grid synchronization instruction table. Step 4: Based on the perturbation zone grid synchronization instruction table, obtain the particle propagation timing and propulsion difference, sort the propulsion time difference matrix of overlapping boundaries, match and number the propulsion direction density blocks, and generate a particle propulsion sequence parallel computing scheduling table. Step 5: Based on the parallel computation scheduling table of the particle propulsion sequence, extract the difference between the state transition and propulsion cycle of the boundary segment, rearrange the propulsion order of the mutation zone, reconstruct the synchronous propulsion section of the overlapping boundary, and generate a multi-zone lattice state coupling propulsion structure diagram; The specific steps for generating the dynamic watershed partitioning structure map are as follows: Based on the open channel cross-sectional structural parameters, the number of disturbance sources, and the slope fluctuation coefficient, the segmental spacing between the cross-sectional coordinate point groups is determined and the area of ​​the contour closed curve is derived. The difference between the valley value of the length of the line connecting the disturbance sources and the angle between the adjacent water flow direction is extracted and the interval is classified to generate the cross-section and disturbance relationship extraction results. Based on the extraction results of the relationship between the cross section and the disturbance, the elevation difference and gradient ratio between water level sampling points in the longitudinal slope area are statistically analyzed, the density distribution of the intersection points of the lateral velocity vectors is identified, and the overlapping areas of the turning section are screened for comparison, and a dynamic zoning structure delineation map of the watershed is generated. The specific steps for generating the differential character-driven structure set are as follows: Based on the aforementioned watershed dynamic zoning structure delineation map, the water level records at each cross-section in the main stream section are aligned with the velocity records at the corresponding time points. The differences between the water level and velocity at the same location in adjacent time steps are subtracted, the trend comparison is calculated, and the matching point sequences are labeled to generate a main stream slope change comparison set. Based on the slope change comparison set of the main segment, the angle between the trajectory line of each control point in the boundary segment and the adjacent velocity vector is extracted to obtain control points with an angle smaller than the specified angle range. At the same time, the slope curvature change value of the control point is extracted and sorted with the depth difference sequence to generate a differential character-driven structure set. The specific steps for generating the synchronization instruction table for the disturbance area grid are as follows: Based on the differential grid cell-driven structure set, the coordinate points of the disturbance source are bound to the number of the starting line segment direction of the grid boundary. Line segments with fixed step sizes are added from the bound points along the diagonal direction. The coordinates of the displacement nodes are recorded, and the cumulative length and endpoint of the line segments are calculated and marked to generate a disturbance source extension path dataset. Based on the disturbance source extension path dataset, the local velocity data of each extension node is subtracted from the disturbance frequency change value of the adjacent node to generate a difference sequence. A graph neural network is used to extract the disturbance correlation between nodes and construct a jump propagation path map to assist in matching the frequency transfer trajectory between nodes with a sudden change threshold. Nodes in each difference sequence that are greater than the jump threshold are paired in position, and the direction tracking mark is bound to the corresponding path identification code to generate a disturbance sudden change node matching set. Based on the disturbance mutation node matching set, extract the peak disturbance amplitude of each node and the difference between the propagation speed of the nodes in the downstream propagation path, sort the amplitude values ​​in each segment, process the speed difference, synchronously update the flag bit recombination and target grid number, and generate a grid synchronization instruction table for the disturbance effect area. The graph neural network extracts the perturbation correlation between nodes and constructs a jump propagation path graph. First, the extended nodes are regarded as independent nodes in the graph structure. A directed edge connection structure is established based on the spatial adjacency relationship and frequency change sequence between nodes. The perturbation frequency difference and spatial displacement between nodes are used as the basis for edge weight assignment. After generating the graph structure, the perturbation correlation index between each node is iteratively calculated, and the strongly correlated node pairs in the perturbation propagation chain are extracted to construct a frequency jump path index graph.

2. The method for comprehensively solving the flow velocity of complex cross-section open channels under a multi-point launch mechanism as described in claim 1, characterized in that, The watershed dynamic zoning structure delineation map includes zoning boundary lines, disturbance response zone labels, and hydraulic slope change zone identification codes. The differential particle-driven structure set includes grid level identifiers, boundary difference zone numbers, and jump-related segment indexes. The disturbance effect zone grid synchronization instruction table includes node matching path numbers, synchronization propulsion level codes, and disturbance response grid indexes. The particle propulsion sequence parallel computing scheduling table includes propulsion task segment sequences, parallel sub-block numbers, and scheduling time step mapping tables. The multi-zone grid state coupling propulsion structure diagram includes grid boundary connection maps, coupling propulsion channel mappings, and synchronization state marker fields.

3. The method for comprehensively solving the flow velocity of complex cross-section open channels under a multi-point launch mechanism as described in claim 1, characterized in that, Specifically, the process of adding a fixed step size to the line segment along the diagonal direction refers to determining the diagonal extension direction with a relative peak angle offset in the lattice structure, using the direction of the initial line segment at the boundary of the lattice where the disturbance source is located as the reference direction, and extending outward in the direction. In the two-dimensional coordinate system, the same value is added simultaneously in the horizontal and vertical directions as a fixed step size. At each step size position, a line segment consistent with the initial direction is drawn. The starting point of the line segment is the ending point of the previous node, and the ending point is the coordinate point of the current step size. This process is repeated until the boundary of the region is reached and the preset distance condition is met. The starting and ending coordinates of the line segment must be recorded, and the total length and ending point information are updated based on the cumulative length of each step.

4. The method for comprehensively solving the flow velocity of complex cross-section open channels under the multi-point launch mechanism according to claim 1, characterized in that, The specific steps for generating the parallel computation scheduling table for the particle propulsion sequence are as follows: Based on the grid synchronization instruction table of the disturbance area, the grid number of each synchronization entry in the table is extracted and the time field is read. The number is extracted by reading the instruction line and a timestamp is generated and time-labeled. The particle propulsion sequence difference and node time difference are calculated to generate a particle propulsion timing difference list. Based on the particle propulsion timing difference list, match the particle set and boundary unit number corresponding to the timing difference entry, read the particle ID and unit number mapping and identify the time difference aggregation and intersection areas within the group, sort the row and column indices of the timing difference matrix, and generate a boundary propulsion difference sorting matrix set. Based on the boundary advancement difference sorting matrix set, the frequency of particle numbers at the matrix intersections is counted, the number list of each intersection is counted and the frequency continuous segment is identified, the block number allocation and advancement order are adjusted, and a particle advancement sequence parallel computing scheduling table is generated.

5. The method for comprehensively solving the flow velocity of complex cross-section open channels under a multi-point launch mechanism as described in claim 4, characterized in that, The particle propulsion order difference refers to the difference in the order of time tags corresponding to different particles on their respective propulsion paths during particle propulsion scheduling. Specifically, it is generated based on the time field read by the particle in the synchronization instruction table, and then the particles are sorted according to the grid number and propulsion path to calculate the difference in propulsion tag values ​​between adjacent particles.

6. The method for comprehensively solving the flow velocity of complex cross-section open channels under a multi-point launch mechanism according to claim 1, characterized in that, The specific steps for generating the multi-zone lattice state coupling propulsion structure diagram are as follows: Based on the parallel computing scheduling table of the particle propulsion sequence, the continuity of particle state labels and propulsion cycle marker fields in each group of boundary segments is read, the periodic sequence between the propulsion start node and the end node is extracted and the difference range is compared one by one, the peak and valley differences during the period are identified and the propulsion offset is sorted to generate a boundary segment propulsion difference dataset. Based on the boundary segment advancement difference dataset, the particle number order and time period index value combination within the advancement offset segment are rearranged, the node number of the continuous mutation zone is located and the synchronous advancement number list is extracted. The advancement segment reorganization and the synchronous grid block path of the boundary coupling zone are combined to generate a multi-zone grid state coupled advancement structure diagram.

Images