A new method for calculating the quantitative rate of debris flow blocking by wedge beam type grid dam

Abstract

The application provides a new type of wedge beam type grid dam debris flow sand blocking rate quantitative calculation method, and belongs to the technical field of water retaining, draining or water releasing buildings in water conservancy projects, and the method comprises the following steps: step one, obtaining basic parameters; step two, calculating reservoir capacity ratio; step three, designing opening width; step four, building a flume test device; and step five, constructing a sand blocking rate prediction model to calculate the sand blocking rate. Through the logical flow of field parameter measurement-dam body parameter calculation-opening width design-sand blocking rate model construction, the quantitative calculation of the wedge beam type grid dam debris flow sand blocking rate is realized.

Description

Technical Field

[0001] This invention relates to the field of debris flow disaster prevention and control technology, specifically to a novel method for quantitatively calculating the debris flow interception rate of a wedge-shaped beam grid dam. Background Technology

[0002] Debris flows are a typical sudden geological disaster in mountainous areas, posing a serious threat to towns, transportation routes, and water conservancy projects. Mountainous areas have long faced pressure in debris flow prevention and mitigation. Currently, among the engineering measures used for debris flow interception, traditional solid dams suffer from poor permeability, indiscriminate interception leading to a surge in upstream loads, susceptibility to impact damage, and the need for manual dredging to maintain their function. While traditional permeable silt-trapping dams, such as beam-type grid dams, lattice dams, and window dams, possess the potential to intercept coarse sediment and allow fine sediment to pass through, their design remains in the empirical stage, presenting two key problems:

[0003] Insufficient consideration of key parameters: Existing models for calculating the interception rate of permeable dams often ignore the important influencing factor of debris flow scale, focusing only on a few parameters such as unit weight and channel slope, which cannot fully reflect the interception effect under different working conditions.

[0004] Poor structural and functional adaptability: Traditional permeable silt traps lack structural optimization from the perspective of debris flow trajectory and flow reduction. The overflow section is prone to complete blockage and failure. Moreover, the design often relies on existing experience and is not designed specifically for the dam type's effectiveness and applicable conditions, making it difficult to achieve stable control effects. Summary of the Invention

[0005] This invention provides a novel method for quantitatively calculating the debris flow interception rate of a wedge-beam grid dam. The method achieves quantitative calculation of the debris flow interception rate of a wedge-beam grid dam through a logical process of on-site parameter measurement, dam body parameter calculation, opening width design, and debris flow interception rate model construction.

[0006] A novel method for quantitatively calculating the debris flow interception rate of a wedge-shaped beam grid dam includes:

[0007] Step 1: Select a section in the debris flow basin that meets the conditions for dam construction and obtain the following basic parameters: (1) Debris flow related parameters: debris flow unit weight, solid volume concentration, debris flow scale;

[0008] (2) Parameters of gullies and boulders: average width of gullies in the cross section of the proposed dam, average longitudinal slope of gullies within the cross section, mud depth of debris flow in front of the dam, and minimum particle size of the largest boulder in the debris flow basin.

[0009] (3) Dam structure parameters: the wedge structure angle ratio of the wedge beam grid dam, which is the ratio of the wedge structure angle of the dam body to 60°;

[0010] Step 2: Based on the debris flow depth at the dam front section, the dam height of the wedge-shaped beam grid dam is initially determined. Combining the average channel width, the minimum particle size of the largest rock, the average channel longitudinal slope, the wedge structure angle ratio, and the debris flow scale, the ratio of the total amount of solid debris flow material to the reservoir capacity of the dam body corresponding to the dam height is calculated, i.e., the reservoir capacity ratio. Among them, the total amount of solid debris flow material is calculated by the debris flow scale and the solid volume concentration.

[0011] Step 3: Obtain particle size distribution curves of debris flows in different regions within the debris flow basin, and select a particle size distribution curve that reflects the overall characteristics of the basin based on the proportion of large particles within the basin; determine the characteristic particle size of the debris flow sample based on this curve. And based on characteristic particle size (characteristic particle size) Design the opening width of the wedge-shaped beam grid dam for the particle size value corresponding to the cumulative particle content reaching 95% in the integrated curve;

[0012] Step 4: Construct a flume test device. Based on the dam height determined in Step 2 and the opening width designed in Step 3, create a wedge-shaped beam grid dam model to simulate debris flow movement and the interception process of the model under different working conditions. The working conditions should at least cover scenarios corresponding to different debris flow unit weights, different debris flow scales, different channel longitudinal slopes, different opening widths, different wedge structure angles, and different reservoir capacity ratios.

[0013] Step 5: Based on the flume test data from Step 4, and combined with the opening width and characteristic particle size d95 from Step 3, the relative opening width is calculated. Then, taking into account six control factors—debris flow unit weight, debris flow scale, average channel longitudinal slope, relative opening width, wedge structure angle ratio, and reservoir capacity ratio—a multiple regression analysis method is used to construct a sediment trapping rate prediction model. This model enables the quantitative calculation of the sediment trapping rate of the wedge beam grid dam.

[0014] In this instruction manual, in step two, another way to calculate the reservoir capacity ratio is: based on the preset reservoir capacity ratio, the height of the wedge beam grid dam that satisfies the reservoir capacity ratio is calculated in reverse using the parameters from step one, and the reservoir capacity of the dam body is determined based on the calculated dam height to obtain the reservoir capacity ratio.

[0015] In this specification, the wedge structure angle ratio in step one ranges from 0.75 to 1.25.

[0016] In this specification, the reservoir capacity of the dam body in step two is calculated by the dam height, the average width of the channel, the sine value of the average longitudinal slope of the channel, and the tangent value corresponding to the angle ratio of the wedge structure. Specifically, the influence of the minimum particle size value of the largest boulder on the reservoir capacity is corrected.

[0017] In this specification, the different regions of the debris flow basin in step three include the debris flow deposition zone and the flow zone. The particle size distribution curves of the two regions are obtained separately, and then the particle size distribution curve of the overall characteristics of the basin is determined by combining the frequency weight of debris flow occurrence in the region.

[0018] In this manual, during the flue test in step four, the actual measurement of the sand interception rate is carried out simultaneously: after the test, the silt in front of the dam is shoveled out and dried until completely dry, and the total amount of solid material intercepted is obtained by weighing. The actual value of the sand interception rate is calculated by combining the total mass of solid material flowing into the flue, which is used to verify the reliability of the sand interception rate prediction model.

[0019] In this instruction manual, in step five, the multiple regression analysis attempts four functional forms: multiple linear function, multiple power function, multiple exponential function, and multiple logarithmic function. Based on the fitting effect of each functional form, the optimal function is selected to construct the sand trapping rate prediction model.

[0020] In this manual, when constructing the sediment trapping rate prediction model in step five, the goodness of fit of the multiple regression analysis is evaluated by the coefficient of determination R², and the function form with the largest R² is selected as the final prediction model form.

[0021] In this manual, step five employs a fusion of three algorithms—multivariate adaptive regression splines, particle swarm optimization, and gradient boosting decision trees—to construct a sand-trapping rate prediction model. Specifically, this includes:

[0022] (51) The initial prediction model of sand-blocking rate is constructed by using the multivariate adaptive regression spline algorithm: six types of control factors are used as input variables and the measured value of sand-blocking rate is used as output variables. Piecewise spline basis functions are constructed. The basis functions are screened by a strategy of forward selection combined with backward pruning. The regression coefficients are solved by the least squares method. The initial prediction value and prediction residual are calculated by substituting the measured parameters on site.

[0023] (52) Parameter co-optimization is performed using particle swarm optimization algorithm: The node parameters of multivariate adaptive regression spline algorithm, the learning rate and decision tree depth of gradient boosting decision tree algorithm are used as the core components of particle position. The sum of squared prediction residuals of multivariate adaptive regression spline algorithm is used as the fitness function. The particle velocity and position are adjusted by inertia weight and learning factor. After iteratively obtaining the global optimal parameters, they are fed back to multivariate adaptive regression spline algorithm to update the basis function and to gradient boosting decision tree algorithm as the initial hyperparameters.

[0024] (53) Gradient boosting decision tree algorithm is used to fit the residuals and achieve bidirectional interaction: The residuals of multivariate adaptive regression spline prediction optimized by particle swarm optimization algorithm are used as the fitting target. The mean square error loss function is constructed and the decision tree is trained iteratively. The gradient of the loss function is fed back to the particle swarm optimization algorithm to adjust the direction of particle iteration. At the same time, the final fitting residuals are fed back to the multivariate adaptive regression spline algorithm to correct its regression coefficients. Through bidirectional interaction and iteration of the three types of algorithms, the sand trapping rate prediction model with improved accuracy is obtained.

[0025] In this manual, before constructing the sediment trapping rate prediction model in step five, the flume test data needs to be divided into a training set and a test set in a 7:3 ratio. The training set data is used to complete the model training, and the test set data is used to verify the prediction accuracy of the model. The verification indicators include relative error and coefficient of determination. If the model prediction accuracy does not meet the engineering requirements, i.e., the relative error is greater than 2%, the algorithm parameters are readjusted and the model is trained again until the accuracy meets the requirements.

[0026] The embodiments described in this specification can achieve at least the following beneficial effects:

[0027] Addressing the shortcomings of traditional methods: It overcomes the problem of insufficient consideration of control parameters for sediment interception rate in traditional empirical methods by incorporating key factors such as debris flow scale and wedge structure angle into the calculation, which significantly improves the accuracy of sediment interception rate prediction under different working conditions and provides a scientific basis for debris flow prevention and control engineering design.

[0028] Optimize the functional characteristics of the dam: The wedge beam grid dam, through structural optimization (setting the transverse beams as wedge structures), can guide the debris flow to divert and collide, achieve energy dissipation, and avoid complete blockage of the overflow section; at the same time, it has the characteristic of "controllable blockage". Even if local blockage occurs after the interception of a small-scale debris flow, the subsequent debris flow can restore the dam's control capacity through "self-dredging" without frequent manual intervention.

[0029] Enhanced adaptability to working conditions: Compared with traditional permeable silt dams, the wedge-shaped beam grid dam corresponding to this method has a stable interception and flow reduction capability for debris flows of different densities and scales. It can maintain a good silt interception effect and peak flow reduction effect under both high-density and low-density debris flow conditions, making it suitable for complex and variable debris flow scenarios in mountainous areas.

[0030] Supporting the practicality of the project: It can be directly applied to the prevention and control of debris flow gullies carrying large rocks in mountainous areas, providing quantitative basis for disaster risk assessment, helping to build a graded control system of "blocking coarse rocks and draining fine rocks", and playing a synergistic protective role when used in combination with physical dams, thereby improving the overall disaster prevention and mitigation efficiency. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of a novel method for quantitatively calculating the debris flow interception rate of a wedge-shaped beam grid dam, as described in some embodiments of the present invention.

[0033] Figure 2 This is a schematic diagram of the structure of a wedge-shaped beam grid dam involved in some embodiments of the present invention.

[0034] Figure 3 This is a schematic diagram comparing the actual and predicted values ​​of the sand-trapping rate in some embodiments of the present invention.

[0035] Figure 4 This is a schematic diagram of the particle size distribution curve of the test soil sample involved in some embodiments of the present invention.

[0036] Appendix Figure 2 mark:

[0037] 1. Wedge-shaped beam; 2. Connecting column; 3. Intercepting column. Detailed Implementation

[0038] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0039] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used to facilitate the description of this invention and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0040] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0041] Furthermore, the terms "installation," "connection," "linking," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0042] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0043] like Figure 1 As shown in the figure, this embodiment provides a novel method for quantitatively calculating the debris flow interception rate of a wedge-shaped beam grid dam, including:

[0044] Step 1: Select a section in the debris flow basin that meets the conditions for dam construction and obtain the following basic parameters: (1) Debris flow related parameters: debris flow unit weight, solid volume concentration, debris flow scale;

[0045] (2) Parameters of gullies and boulders: average width of gullies in the cross section of the proposed dam, average longitudinal slope of gullies within the cross section, mud depth of debris flow in front of the dam, and minimum particle size of the largest boulder in the debris flow basin.

[0046] (3) Dam structure parameters: the wedge structure angle ratio of the wedge beam grid dam, which is the ratio of the wedge structure angle of the dam body to 60°;

[0047] Step 2: Based on the debris flow depth at the dam front section, the dam height of the wedge-shaped beam grid dam is initially determined. Combining the average channel width, the minimum particle size of the largest rock, the average channel longitudinal slope, the wedge structure angle ratio, and the debris flow scale, the ratio of the total amount of solid debris flow material to the reservoir capacity of the dam body corresponding to the dam height is calculated, i.e., the reservoir capacity ratio. Among them, the total amount of solid debris flow material is calculated by the debris flow scale and the solid volume concentration.

[0048] Step 3: Obtain particle size distribution curves of debris flows in different regions within the debris flow basin, and select a particle size distribution curve that reflects the overall characteristics of the basin based on the proportion of large particles within the basin; determine the characteristic particle size of the debris flow sample based on this curve. And based on characteristic particle size Design the opening width of the wedge-shaped beam grid dam;

[0049] Step 4: Construct a flume test device. Based on the dam height determined in Step 2 and the opening width designed in Step 3, create a wedge-shaped beam grid dam model to simulate debris flow movement and the interception process of the model under different working conditions. The working conditions should at least cover scenarios corresponding to different debris flow unit weights, different debris flow scales, different channel longitudinal slopes, different opening widths, different wedge structure angles, and different reservoir capacity ratios.

[0050] Step 5: Based on the water tank test data from Step 4, and combined with the opening width and characteristic particle size from Step 3... The relative opening width was calculated, and then, taking into account six control factors, namely debris flow unit weight, debris flow scale, average channel longitudinal slope, relative opening width, wedge structure angle ratio, and reservoir capacity ratio, a multivariate regression analysis method was used to construct a sediment trapping rate prediction model. This model was used to quantitatively calculate the sediment trapping rate of the wedge beam grid dam.

[0051] In some embodiments, in step two, another way to calculate the reservoir capacity ratio is: based on the preset reservoir capacity ratio, the height of the wedge beam grid dam that satisfies the reservoir capacity ratio is calculated in reverse using the parameters from step one, and the reservoir capacity of the dam body is determined based on the calculated dam height to obtain the reservoir capacity ratio.

[0052] In some embodiments, the wedge structure angle ratio in step one ranges from 0.75 to 1.25.

[0053] In some embodiments, the reservoir capacity of the dam body in step two is calculated by the dam height, the average width of the channel, the sine value of the average longitudinal slope of the channel, and the tangent value corresponding to the angle ratio of the wedge structure, and is specifically modified in combination with the influence of the minimum particle size value of the largest boulder on the reservoir capacity.

[0054] In some embodiments, the different regions of the debris flow basin in step three include debris flow deposition areas and flow areas. The particle size distribution curves of the two regions are obtained separately, and then the particle size distribution curve of the overall characteristics of the basin is determined by combining the frequency weight of debris flow occurrence in the region.

[0055] In some embodiments, during the flue test in step four, the actual measurement of the sand interception rate is carried out simultaneously: after the test, the silt in front of the dam is shoveled out and dried until completely dry, and the total amount of solid material intercepted is weighed to obtain the total amount of solid material intercepted. The actual value of the sand interception rate is calculated in combination with the total mass of solid material flowing into the flue, which is used to verify the reliability of the sand interception rate prediction model.

[0056] In some embodiments, in the multiple regression analysis of step five, four function forms are tried respectively: multiple linear function, multiple power function, multiple exponential function and multiple logarithmic function. The optimal function is selected based on the fitting effect of each function form to construct the sand trapping rate prediction model.

[0057] In some embodiments, when constructing the sand trapping rate prediction model in step five, the determination coefficient is used. To evaluate the goodness of fit of the multiple regression analysis, select... The largest functional form is used as the final prediction model form.

[0058] In some embodiments, the solid volume concentration in step one is calculated based on the bulk density of the debris flow, the specific gravity of the clear water, and the bulk density of the solid material carried by the debris flow.

[0059] In some embodiments, the preset storage capacity ratio in step two is 3.0.

[0060] In some embodiments, step five employs a fusion of three algorithms—multivariate adaptive regression spline, particle swarm optimization, and gradient boosting decision tree—to construct a sand-trapping rate prediction model, specifically including:

[0061] (51) The initial prediction model of sand-blocking rate is constructed by using the multivariate adaptive regression spline algorithm: six types of control factors are used as input variables and the measured value of sand-blocking rate is used as output variables. Piecewise spline basis functions are constructed. The basis functions are screened by a strategy of forward selection combined with backward pruning. The regression coefficients are solved by the least squares method. The initial prediction value and prediction residual are calculated by substituting the measured parameters on site.

[0062] (52) Parameter co-optimization is performed using particle swarm optimization algorithm: The node parameters of multivariate adaptive regression spline algorithm, the learning rate and decision tree depth of gradient boosting decision tree algorithm are used as the core components of particle position. The sum of squared prediction residuals of multivariate adaptive regression spline algorithm is used as the fitness function. The particle velocity and position are adjusted by inertia weight and learning factor. After iteratively obtaining the global optimal parameters, they are fed back to multivariate adaptive regression spline algorithm to update the basis function and to gradient boosting decision tree algorithm as the initial hyperparameters.

[0063] (53) Gradient boosting decision tree algorithm is used to fit the residuals and achieve bidirectional interaction: The residuals of multivariate adaptive regression spline prediction optimized by particle swarm optimization algorithm are used as the fitting target. The mean square error loss function is constructed and the decision tree is trained iteratively. The gradient of the loss function is fed back to the particle swarm optimization algorithm to adjust the direction of particle iteration. At the same time, the final fitting residuals are fed back to the multivariate adaptive regression spline algorithm to correct its regression coefficients. Through bidirectional interaction and iteration of the three types of algorithms, the sand trapping rate prediction model with improved accuracy is obtained.

[0064] In some embodiments, before constructing the sediment trapping rate prediction model in step five, the flume test data needs to be divided into a training set and a test set in a 7:3 ratio; the training set data is used to complete the model training, and the test set data is used to verify the prediction accuracy of the model. The verification indicators include relative error and coefficient of determination; if the model prediction accuracy does not meet the engineering requirements, i.e., the relative error is greater than 2%, the algorithm parameters are readjusted and the model is trained again until the accuracy meets the requirements.

[0065] In some embodiments, the multiple regression analysis method in step five specifically involves: firstly analyzing the single functional relationship between the sand-trapping rate and the six types of control factors, determining the optimal functional form that matches each type of control factor with the sand-trapping rate, wherein the functional form includes linear functions, power functions, exponential functions, and logarithmic functions, and then combining the optimal functional forms to construct a sand-trapping rate prediction model that includes all control factors.

[0066] In some embodiments, the specific method for obtaining the debris flow particle size distribution curve in step three is as follows: debris flow samples are collected in the deposition and flow areas of the debris flow basin through field pit exploration tests, and particle size distribution curves of each region are plotted using the sieving method. Combined with the mass proportion of large particles in the basin, the curves of each region are integrated using the weighted average method to obtain the particle size distribution curves that reflect the overall characteristics of the basin.

[0067] In some embodiments, if the calculated reservoir capacity ratio in step two exceeds the range of 0.01 to 4.19, the dam height of the wedge beam grid dam needs to be readjusted: if the reservoir capacity ratio is too small, the dam height should be appropriately reduced; if the reservoir capacity ratio is too large, the dam height should be appropriately increased until the reservoir capacity ratio falls within the above range, and then the adjusted dam height is used as the final dam height parameter.

[0068] The technical concept of this invention is as follows:

[0069] The debris flow interception rate calculation model proposed in this invention is applicable to specific debris flow channels containing a certain number of large particles, and employs wedge-shaped beam grid dams as an engineering measure to intercept and regulate debris flows. Large particles refer to particles that can be effectively blocked by designing the opening size of the wedge-shaped beam grid dam, enabling this type of permeable interceptor dam to achieve effective closure. The wedge-shaped beam grid dam model is a structural optimization based on a conventional beam grid dam; the structure of the wedge-shaped beam grid dam is shown in the attached figure. Figure 2 As shown, the wedge-beam grid dam is a permeable retaining structure arranged laterally along the debris flow channel. The overall structure is a continuous, periodic combination of "wave-shaped beams + vertical columns." Its core consists of three types of components: intercepting columns 3, connecting columns 2, and wedge-shaped beams 1. Their structure, connections, and positional relationships are as follows:

[0070] The dam body forms a stable overall structure through the integrated and fixed connection of its components:

[0071] The intercepting column 3 is the vertical load-bearing and main retaining component of the dam body. Its top horizontal end is fixedly connected to one end (beam end) of two symmetrically inclined wedge beams 1 (usually integrally formed by concrete pouring).

[0072] The wedge-shaped beams 1 extending from the adjacent intercepting columns 3 will converge towards the middle of the dam body, and their ends (i.e. the apex of the V-shaped beams) will be fixedly connected to the top end face of the connecting column 2.

[0073] The bottoms of both the intercepting column 3 and the connecting column 2 are embedded in the foundation structure of the ditch, forming a vertical support system that connects the dam body and the foundation as a whole.

[0074] The components are arranged in a regular spatial distribution:

[0075] The intercepting columns 3 are arranged vertically at equal intervals along the length of the dam body and are the main vertical nodes of the structure;

[0076] Connecting post 2 is located on the horizontal midline between two adjacent intercepting posts 3, and is distributed in an alternating pattern with the intercepting posts 3 in the order of "intercepting post 3 → connecting post 2 → intercepting post 3";

[0077] The wedge beam 1 extends inclinedly towards the two intercepting columns 3 with the connecting column 2 as the apex, forming a continuous V-shaped barrier surface. The gap between adjacent wedge beams 1 is the "opening" of the dam body (used to pass through fine silt), and the whole structure forms a cyclic barrier-diversion spatial layout.

[0078] In this structure, wedge beam 1 is responsible for guiding the debris flow to divert and dissipate energy, and intercepting medium-sized particles. Interception column 3 directly blocks large rocks, while connecting column 2 strengthens the connection stability of adjacent wedge beams 1, ultimately achieving the function of "intercepting coarse rocks and discharging fine ones".

[0079] This calculation method includes the following steps:

[0080] Step 1: Select a cross-section suitable for dam construction in the debris flow basin, and determine the debris flow unit weight and solid matter concentration through on-site investigation and testing. The mudslide depth h at the cross-section in front of the dam, and the slope of the gully bed. ;

[0081] ;

[0082] —Debris flow unit weight (kN / The specific density (density) can be determined through field surveys, on-site mixing tests, and table lookup (test density 1.6–2.0 t / (The on-site slurry preparation method requires taking no less than 3 sets of debris flow deposit samples at the test point, with each set weighing no less than 5 kg.)

[0083] —Severity of pure water (kN / ), with a value of 10;

[0084] —The bulk density of solid materials carried by the mudslide (kN / This was determined through field investigations and experiments.

[0085] Step two: Based on the mud depth h of the debris flow, the dam height H of the wedge-beam grid dam can be preliminarily determined, and the ratio of the debris flow scale to the reservoir capacity of the wedge-beam grid dam can be calculated.

[0086] ;

[0087] ;

[0088] Or first determine the initial storage capacity ratio. Determine the dam height of the wedge-shaped beam grid dam:

[0089] ;

[0090] —The reservoir capacity (m³) of a wedge-shaped grid dam can be calculated.

[0091] —The average width (m) of the channel of the proposed wedge-shaped grid dam section in the debris flow basin was determined through field investigation.

[0092] —Proposed height of wedge-shaped grid dam (m);

[0093] —Angle of the wedge-shaped structure of the wedge-shaped beam grid dam The ratio to 60° (to ensure the wedge structure has a certain shunting efficiency), The range should be 0.75 to 1.25, which is not affected by the channel slope and other parameters. It is inversely proportional to the sand-blocking rate and the degree of blockage. In actual engineering, the value can be selected according to the comprehensive requirements of the sand-blocking rate and the degree of blockage.

[0094] —Scale of a single debris flow ( (This can be determined by referring to the "Specification for Investigation of Debris Flow Disaster Control Engineering", based on the duration of the debris flow.) and peak traffic The total volume of slurry in a single debris flow is considered the scale of that debris flow. ;

[0095] —The minimum particle size (m) of the largest boulder in the debris flow basin was determined through field investigation;

[0096] —The average gully slope within the cross-section of the proposed wedge-shaped grid dam was determined through field investigation;

[0097] —The ratio of the total amount of solid debris flow to the reservoir capacity of the wedge-shaped beam grid dam, with a value range of 0 to 4.0;

[0098] —The solid volume concentration in debris flows was determined through field investigation and testing;

[0099] Step 3: Conduct pit exploration tests through field reconnaissance to obtain particle size distribution curves at various locations (deposition areas, flow areas, etc.) within the debris flow basin. Simultaneously, select a representative overall particle size distribution curve for the basin based on the proportion of large particles. Determine the characteristic particle size d of the debris flow samples from this basin. 95 The opening width b of the wedge-shaped beam grid dam can be determined.

[0100] Step four: The sediment retention rate of the wedge-beam grid dam for debris flow can be calculated using the following formula:

[0101] ;

[0102] In the formula: This represents the measured value of the sand-trapping rate; This indicates the total amount of mud and debris flowing into the reservoir; This indicates the total amount of debris flow intercepted. After the test, the silt in front of the dam was shoveled out, dried, and then weighed.

[0103] Step 5: By combining experimental data with the characteristics of blockage degree variation with five control factors, and analyzing 70 sets of experimental data, a sand-blocking rate prediction model was established using Origin software with different function forms, such as multivariate linear functions, multivariate power functions, multivariate exponential functions, and multivariate logarithmic functions.

[0104] ;

[0105] In the formula: This represents the predicted sediment trapping rate; other symbols are as described above.

[0106] The angle ratio of the wedge-shaped beam grid dam structure A value of 1.0 is recommended for the ratio of the total solid matter from debris flows to the reservoir capacity of the wedge-shaped beam grid dam. The recommended value is 3.0.

[0107] This invention addresses the issue of neglecting debris flow scale in the key control selection factors of existing permeable silt-trapping dams, and proposes a wedge-shaped beam grid dam from a structural optimization perspective. Based on existing research on permeable silt-trapping dams and combined with flume model test data, a novel quantitative calculation method for debris flow interception rate of wedge-shaped beam grid dams is created. It creatively introduces the wedge structure angle ratio and reservoir capacity ratio, providing new evidence for the performance study of permeable silt-trapping dams in intercepting debris flows, and has strong practicality.

[0108] The model is derived from fitting test data of a wedge-beam grid dam flume model. It comprehensively considers the influence of five control factors (debris flow unit weight, debris flow scale, channel longitudinal slope gradient, relative opening width, and wedge structure angle) on the debris flow interception rate. To ensure the applicability of the model, the control factors (debris flow solid volume concentration) are dimensionless. Debris flow scale to silt-trapping dam capacity ratio , channel gradient tangent Relative opening width wedge structure angle ratio ):

[0109] ;

[0110] Based on the test data of the water tank model, the sand trapping rate The functional relationship between the control factors and the control factors is as follows:

[0111] Sand retention rate Solids volume concentration with debris flow As increases, it increases, satisfying the functional relationship:

[0112] ;

[0113] Sand retention rate With storage capacity ratio The increase of follows a pattern of first increasing and then decreasing, satisfying the functional relationship:

[0114] ;

[0115] Sand retention rate With the gradient of the ditch As increases, decreases, satisfying the functional relationship:

[0116] ;

[0117] Sand retention rate relative opening width of wedge-shaped beam grid dam As increases, decreases, satisfying the functional relationship:

[0118] ;

[0119] Sand retention rate With the wedge structure angle ratio As increases, it increases, satisfying the functional relationship:

[0120] ;

[0121] in:

[0122] , , .

[0123] In the formula, The functional mapping relationship between the independent and dependent variables represents the functional influence of the independent variable on the sand retention rate. The angle (°) of the wedge structure of the silt trap dam model used in the experiment; , , , , These are five constants, determined experimentally; the meanings of the other symbols are the same as described above.

[0124] Based on the analysis of 90 sets of experimental data, and using Origin software, different function forms were employed, such as multivariate linear functions, multivariate power functions, multivariate exponential functions, and multivariate logarithmic functions. The final prediction model for the sediment retention rate of the wedge-shaped beam grid dam is as follows:

[0125] ;

[0126] A Comparison of Actual and Predicted Debris Retention Rates of Wedge-Beam Grid Dams for Debris Flow Interception: Figure 3 As shown. The design parameters of the wedge-shaped beam grid dam include the wedge angle, the relative opening width of the wedge, the height of the wedge itself, and the front and rear connecting columns 2 and intercepting columns 3 of the wedge. To maximize the diversion effect of the wedge structure, the wedge angle is generally set at 60°. The height of the wedge can be determined based on the shortest axis dimension min of the largest particle (min). , , ), The shortest axis width of the largest boulder in the debris flow basin; The shortest axis height of the largest boulder in the debris flow basin; The shortest axis thickness of the largest boulder in the debris flow basin; to maximize the interception capacity of the wedge-shaped beam grid dam, its relative opening width can be set to 1.62. The wedge-shaped structure will produce a certain stress concentration effect. The entire sand-trapping dam can be connected as a whole by the connecting column 2 between the wedge-shaped structures and the intercepting column 3 in the rear row to ensure its stability.

[0127] Design concept of wedge-shaped beam grid dam:

[0128] (1) In accordance with the concept of comprehensive prevention and control of debris flow, the through-type sand-trapping dam fully utilizes the immediate effectiveness of geotechnical engineering measures and the continuous increase in the intensity of ecological and environmental protection measures over time;

[0129] (2) Energy-directed dissipation: The wedge-shaped structure angle guides the debris flow to split and collide, converting the linear impact kinetic energy into oblique shear energy and eddy current dissipation;

[0130] (3) Adaptive control: The diversion of the wedge structure changes the motion pattern of particles in the fluid (direction of motion of major and minor axes) to a certain extent, and reduces the degree of blockage by controlling the overflow section area by controlling the structural angle;

[0131] (4) It weakens the peak flow of debris flow, thereby reducing the impact and damage of debris flow on the downstream area. It can maintain long-term regulation capacity and mainly functions as "blocking coarse debris and discharging fine debris". It is generally combined with a solid dam to play a synergistic role.

[0132] The experimental flume mainly consists of a hopper, a water tank, a wedge-beam grid dam model, and a tailings pool. The hopper is 0.8m long, 0.8m wide, and 0.8m high, made of stainless steel plate; the water tank is 5.0m long, 0.4m wide, and 0.4m high on both sides, with tempered glass sides and a steel plate bottom; the wedge-beam grid dam model is set 3.0m from the hopper opening, with four different opening spacing sizes designed; the tailings pool is 1.0m long, 1.0m wide, and 0.5m high, made of stainless steel plate; the adjustable slope of the water tank ranges from 0° to 20°. The similarity scale of the flume test is shown in Table 1.

[0133] Table 1. Similarity scale for flume tests

[0134]

[0135] The particle size distribution curve of the test soil sample is as follows Figure 4 (Note: Due to experimental conditions, the proportion of coarse particles (20-50mm) was fixed at 15% in this experiment.) The maximum particle size for debris flow was determined to be 50mm in this experiment. Large gravel particles were removed from the original soil sample, and 85% of the total experimental material was retained, with particles smaller than 20mm. The maximum particle size of the material in this experiment was... =50mm, characteristic particle size =37mm, coefficient of non-uniformity =22.5, curvature coefficient =2.8.

[0136] The test conditions were designed, and the test parameters are shown in Table 2.

[0137] Table 2. Test Parameter Table

[0138]

[0139] The debris flow was 0.05 in size. Through observation of experimental phenomena and data analysis, it was found that the interception capability was insufficient when the relative opening was too large, thus requiring some compromise. The influence of the wedge structure angle on the degree of closure was explored under the condition of a channel longitudinal slope of 11°. The influence of the channel longitudinal slope on the degree of closure was explored under the condition of a relative opening width of 1.62. Therefore, the final number of experimental groups was 90.

[0140] Test procedure:

[0141] 1) Model preparation: Assemble the water tank model, hopper, tailings pool, etc. according to requirements, screen the materials, and debug the test equipment;

[0142] 2) Material preparation: Based on the soil sample material required for the test and the designed fluid density, prepare the required fluid sample according to a certain ratio of soil sample and water. In order to fully mix the material with water, soak the mudflow sample overnight after preparation. Before starting the test, use a mixer to fully stir the prepared mudflow sample and then pour it into the test material tank in a small bucket.

[0143] 3) Experimental preparation: According to the design section location, set up the camera, GoPro camera and mud level gauge in advance, and adjust the shooting direction of the camera and zero the error of the mud level gauge. Check whether the mud level gauge and mud level acquisition instrument are properly connected; set up the wedge beam grid dam model at the design location one day in advance (use glass glue to glue the bottom and sides to prevent the model from being washed away by the debris flow).

[0144] 4) Start the test: Turn on the cameras set up on the side and front of the water tank, stir the test fluid in the hopper evenly, and then quickly open the hopper valve (release in one go). During the test, put small balls in, monitor the mud level of the fluid with a mud level gauge, and record the test process with a camera to facilitate the later combination with other methods to obtain debris flow movement parameters; during the test, hold two 5000ml measuring buckets at the mouth of the water tank to collect the debris flow after passing the dam.

[0145] 5) Data recording: After the fluid in the water tank has basically stopped moving, turn off the camera, mud level acquisition instrument and other equipment; use a sampling shovel and two 1000ml plastic beakers to collect the silted soil in front of the dam; record the bulk density of the soil sample in the measuring bucket and beakers;

[0146] 6) Cleaning the water tank: After the data recording and test sampling are completed, connect the water pipe to start rinsing from the hopper; according to the test group design, change the debris flow fluid properties, debris flow scale, relative opening height, wedge structure angle and channel gradient, and repeat the above process.

[0147] Experimental data analysis, some model experimental data are shown in Table 3.

[0148] Table 3. Partial model test data

[0149]

[0150] Compared with existing empirical models of sediment retention rate for permeable silt-trapping dams, some empirical models of sediment retention rate for permeable silt-trapping dams are shown in Table 4.

[0151] Table 4. Empirical Model of Sediment Retention Rate for Partially Permeable Sediment Barrier Dams

[0152]

[0153] Opening width; characteristic particle size This is the particle size value corresponding to when the cumulative particle content in the integration curve reaches 90%. The height of the dam body of the window dam; The sediment retention rate of the window dam; The opening ratio of the window dam; It is the cross-sectional shape factor; The spacing between rows of pile-lined dams; The effective interception length of the pile body of the pile forest dam; The debris flow permeability; This represents the minimum gap width of the slot dam. The volume concentration of debris flow solids in the slot dam; The relative opening width of the comb dam; The sediment retention rate of the comb dam; The width of the opening of the comb dam; This represents the maximum particle size in debris flows.

[0154] Compared with existing test data of permeable silt-trapping dam models, the evaluation index data of commonly used permeable silt-trapping dam flume model test research are shown in Table 5, and the evaluation index data of wedge beam grid dam flume model are shown in Table 6.

[0155] Table 5. Evaluation index data for common permeable silt trapping dam flume model test studies.

[0156]

[0157] Table 6. Evaluation index data for wedge-beam grid dam flume model

[0158]

[0159] The performance of permeable silt-trapping dams in regulating debris flows relies on a comprehensive evaluation of indicators such as silt trapping rate, blockage degree, and peak flow reduction rate. Comparative evaluation data shows that wedge-beam grid dams have better control capabilities for viscous debris flows and large-scale dilute debris flows.

[0160] (1) Wedge-shaped beam grid dams possess, to a certain extent, the functions of "high sediment retention rate" and "controllable closure degree".

[0161] Adjusting 20kN / In high-density debris flows, the sediment interception rate of traditional beam-type grid dams and lattice dams fluctuates drastically with varying operating conditions (e.g., the lattice dam's interception rate drops sharply from 52% to 17%), reflecting the instability of their structural interception effect; while wedge-shaped beam-type grid dams, under the same density and 0.15 Under large-scale operating conditions, it still maintains a sand interception rate of 91.1% and 78.2%; at 0.30... Under large-scale operating conditions, the sediment interception rate decreases slightly, while the blockage degree shows a gradual decreasing trend, and the peak flow reduction rate also has a considerable value. This also confirms that the blockage degree of the wedge beam grid dam in intercepting small-scale debris flows is not completely blocked, the dam body is not completely ineffective, and still has a certain self-cleaning capacity for subsequent debris flows.

[0162] (2) Better peak flow reduction capability

[0163] In debris flow conditions similar to those of other types of dams (20kN / ), wedge-shaped beam grid dams (20kN / ) -0.10 The peak reduction rate reached 93.8%, far exceeding other permeable silt-trapping dams. Secondly, a comparison showed that under low density conditions, the reduction rate of pile-forest dams and comb-type dams dropped significantly to 40%-50%, while the wedge-beam grid dam maintained a reduction rate of 17kN / m³. Across various scales, the reduction rate remains stable between 40.1% and 70.5%. This stable peak-shaving performance means that wedge-beam grid dams can provide more reliable safety redundancy for downstream protection, and their structure is more adaptable to changes in debris flow properties.

[0164] (3) It has stronger adaptability to working conditions and structural robustness.

[0165] When comparing the performance degradation of different dam types under "high density" and "low density" conditions, the advantages of wedge-beam grid dams are more obvious. Taking the sediment retention rate as an example, the performance of traditional dam types generally drops drastically when switching from high density to low density (e.g., grid dams drop from 52% to 0%). While wedge dams also experience a decrease, their performance remains relatively stable under low density and large-scale (17kN / m³) conditions. -0.30 Even under such extreme conditions, it can still maintain a sand interception rate of 74.9% and a peak flow reduction capacity of 50%.

[0166] In summary, wedge-beam grid dams exhibit better sediment retention rates compared to traditional permeable silt-trapping dams. Furthermore, the blockage observed in small-scale interception is not ineffective, as it recovers under the subsequent self-cleaning effect of debris flows. Simultaneously, wedge-beam grid dams demonstrate the best peak flow reduction rate, achieving up to 93.8% reduction for viscous debris flows. While their sediment retention rate for dilute debris flows is lower than that of conventional beam grid dams, the low blockage and high-scale performance data indicate that wedge-beam grid dams still possess good regulation capabilities under large-scale conditions.

[0167] By summarizing the existing empirical models of the interception rate of permeable silt-trapping dams, most of the key parameters selected ignore the influence of debris flow scale. However, the wedge beam grid dam, considering its own structural angle, selects the influence of the change in debris flow scale on its interception rate, further improving the control factors and variation characteristics of the interception and regulation effect of permeable silt-trapping dams, and providing a reference for debris flow disaster prevention and mitigation.

[0168] In one specific embodiment:

[0169] I. Background of the Implementation Examples

[0170] This embodiment selects a debris flow gully as the research object. This gully is a typical mountain debris flow gully with a drainage area of ​​8.2 km². The gully has a steep longitudinal slope, and debris flows carry large boulders (maximum particle size ≥ 50cm). Historically, debris flow disasters have occurred multiple times, necessitating the construction of wedge-shaped beam grid dams for prevention. This embodiment fully presents the quantitative calculation process for the sediment retention rate, focusing on the core step of "constructing a sediment retention rate prediction model." It introduces three types of algorithms—Multivariate Adaptive Regression Splines (MARS), Particle Swarm Optimization (PSO), and Gradient Boosting Decision Tree (GBDT)—for fusion optimization, achieving accurate sediment retention rate prediction. Furthermore, the algorithms interact and iterate collaboratively.

[0171] II. Complete Implementation Steps

[0172] Step 1: On-site parameter measurement

[0173] Basic parameters were obtained through field surveys and sampling experiments.

[0174] 1. Debris flow unit weight: Three groups of debris flow deposit samples (≥5kg each) were collected and measured using a slurry mixing test. =19kN / ;

[0175] 2. Solid particle bulk density: measured by ring sampler method ;

[0176] 3. Calculation of solid volume concentration: ;

[0177] 4. Longitudinal slope gradient of the gully: Average slope of the dam site section measured by total station. ;

[0178] 5. Particle size distribution: Particle size distribution curves are obtained through pit testing to determine... ;

[0179] 6. Debris flow scale: Calculated according to the debris flow disaster control engineering investigation specifications. Storage capacity ratio (Recommended value);

[0180] 7. Wedge structure angle ratio: Design (Right now ).

[0181] Step 2: Calculation of dam body parameters

[0182] Dam height H is calculated by inversely using the reservoir capacity ratio formula (core formula):

[0183] ;

[0184] The average width of the channel is B=5m, and the maximum particle size of the stones is... =0.5m, substitute into the known parameters ( ):

[0185] ;

[0186] Simplify the calculation:

[0187] , Substituting, we get:

[0188] ;

[0189] ;

[0190] ;

[0191] ;

[0192] Solve a quadratic equation of one variable: (Take the positive root);

[0193] Step 3: Opening width design

[0194] based on =0.4m, design relative opening width (Optimal value), therefore:

[0195] ;

[0196] Step 4: Actual measurement of sand retention rate

[0197] The movement of debris flow was simulated through a water tank test. After the test, the solids were dried and weighed to determine the mass of the solids entering the storage. =120kg, the mass of the intercepted solid =108kg, therefore the measured sand-trapping rate is:

[0198] ;

[0199] Step 5: Construction of the sediment retention rate prediction model

[0200] 5.1 Algorithm 1: Multivariate Adaptive Regression Splines

[0201] 5.1.1 Model Building Process

[0202] MARS fits the nonlinear relationship between input parameters and sediment trapping rate by constructing piecewise spline basis functions. The input parameter set is... The output is the sand trapping rate L.

[0203] 1. Definition of spline basis function: Let the input variable be... The basis functions of MARS are:

[0204] or ;

[0205] in The nodes (thresholds) for MARS are automatically selected by the algorithm;

[0206] 2. Regression Model Construction: By linearly combining the basis functions, the initial MARS model is obtained:

[0207] ;

[0208] in The number of basis functions. Let be the regression coefficient of the i-th basis function. For the basis function matrix, Let i be the i-th basis function.

[0209] 5.1.2 Model Training Process

[0210] 1. Data set partitioning: The 90 sets of water tank test data (including this embodiment) were divided into a training set (63 sets) and a test set (27 sets) in a 7:3 ratio.

[0211] 2. Basis function selection: A forward selection + backward pruning strategy is adopted, with forward selection maximizing... Backward pruning removes redundant basis functions, ultimately determining... =8 basis functions;

[0212] 3. Coefficients determination: Solved using the least squares method. The objective function is:

[0213] ;

[0214] Where N=63 (number of training set samples). Let J be the measured sand-trapping rate of the j-th sample. Let be the basis function vector of the j-th sample.

[0215] 5.1.3 Model Application Process

[0216] The input parameters of this embodiment Substituting the formula [0.545, 1.0, 3.0, 1.62, tan10°≈0.1763] into the MARS model, we get:

[0217] =[1,0.045,0.2,0.12,0.0763,0.02,0.015,0.008] (basis function vector);

[0218] (Regression coefficients);

[0219] Therefore, the initial MARS prediction value is:

[0220] =1×0.7+0.045×0.3+0.2×0.2+0.12×0.15+0.0763×0.1+0.02×0.08+0.015×0.05+0.008×0.03=0.792;

[0221] MARS prediction residuals: =0.9-0.792=0.108;

[0222] 5.2 Algorithm 2: Particle Swarm Optimization

[0223] 5.2.1 Model Building Process

[0224] PSO is used to optimize the node parameters of MARS. The core hyperparameter of GBDT (learning rate) Decision tree depth The particle position vector is defined as:

[0225] ;

[0226] Where k = 1, 2, ..., K (K = 50, the number of particles), and the particle velocity vector is... Speed ​​update formula:

[0227] ;

[0228] Position update formula: ;

[0229] =0.7 (inertia weight). =1.494 (learning factor); (Random number); This represents the historical best position of the k-th particle; This is the globally optimal position.

[0230] 5.2.2 Model Training Process

[0231] 1. Definition of fitness function:

[0232] ;

[0233] in To optimize the residual of the j-th sample after the MARS node, a higher fitness indicates a smaller MARS residual;

[0234] 2. Iterative optimization:

[0235] Initialize particle position / velocity: (node), , ;

[0236] Number of iterations =100, calculate fitness in each iteration, and update and ;

[0237] The global optimal position is finally obtained: =[0.12,0.25,0.3,0.18,0.09,0.05,0.03,0.01,0.15,5], which represents the optimized MARS node. =[0.12,0.25,...,0.01], GBDT learning rate =0.15, decision tree depth =5.

[0238] 5.2.3 Model Application Process

[0239] Substitute the optimized MARS nodes into the basis function and recalculate the MARS prediction values:

[0240] =[1,0.06,0.18,0.15,0.08,0.03,0.02,0.01];

[0241] =1×0.7+0.06×0.3+0.18×0.2+0.15×0.15+0.08×0.1+0.03×0.08+0.02×0.05+0.01×0.03=0.835;

[0242] Optimized MARS residuals: ;

[0243] 5.3 Algorithm 3: Gradient Boosting Decision Tree

[0244] 5.3.1 Model Building Process

[0245] GBDT fits the residuals of MARS by iteratively training a decision tree, while simultaneously feeding the gradient of the loss function back to PSO (achieving interaction between GBDT and PSO). The core formula is:

[0246] 1. The residual (fitting target) of the t-th iteration: ;

[0247] in (MARS predictions after PSO optimization), loss function Mean squared error:

[0248] ;

[0249] 2. Output of the t-th decision tree: ;

[0250] 3. Predicted value update (fusing MARS results): ;in The learning rate is optimized for PSO, enabling PSO to input parameters to GBDT.

[0251] 5.3.2 Model Training Process

[0252] 1. Iterative training: Set the number of decision trees =50, the iteration process is as follows:

[0253] t=1: =0.065, training the first decision tree Output =0.04, update the predicted value: =0.835 + 0.15 × 0.04 = 0.841;

[0254] t=2: =0.9-0.841=0.059, train the second decision tree. =0.035, update the predicted value: =0.841 + 0.15 × 0.035 = 0.846;

[0255] Iterate until t=50;

[0256] 2. Feedback Interaction (GBDT→PSO): Calculate the gradient of the loss function in each round. Adjust the PSO particle velocity (interactive formula):

[0257] ;

[0258] in =0.01 (gradient feedback coefficient), realizing the reverse adjustment of PSO by GBDT;

[0259] 3. MARS and GBDT Interaction: The final prediction residuals of GBDT are... Feedback to MARS to correct basis function coefficients: ;in =0.1 (MARS regression coefficient correction coefficient, engineering experience value), to achieve self-updating of MARS based on GBDT results.

[0260] 5.3.3 Model Application Process

[0261] After 50 iterations, the final prediction value of GBDT is: ;

[0262] at this time:

[0263] GBDT residuals: =0.9-0.892=0.008;

[0264] MARS corrected coefficients:

[0265] ;

[0266] Corrected MARS+GBDT fusion predictions:

[0267] ;

[0268] IV. The Integration and Interaction Process and Core Role of Algorithms

[0269] 4.1 Pairwise interaction process (including interaction formulas)

[0270] (1) MARS and PSO interaction

[0271] Positive interaction (PSO→MARS): PSO optimizes the node parameters of MARS. Interactive formula:

[0272] ;

[0273] In other words, PSO determines the optimal node by minimizing the MARS residual, which directly improves the MARS fitting accuracy.

[0274] Reverse interaction (MARS→PSO): The MARS residual serves as the core input to the PSO fitness function, determining the particle iteration direction. The interaction formula is the PSO fitness function. .

[0275] (2) PSO and GBDT interaction

[0276] Positive interaction (PSO→GBDT): PSO optimizes the learning rate of GBDT. and decision tree depth Interactive formula:

[0277] ;

[0278] In other words, PSO provides the optimal hyperparameters for GBDT, avoiding overfitting / underfitting;

[0279] Reverse interaction (GBDT→PSO): The gradient feedback of the GBDT loss function adjusts the particle velocity of PSO. Interaction formula:

[0280] ;

[0281] In other words, the fitting effect of GBDT inversely corrects the search direction of PSO, thereby improving the global optimization efficiency; =0.01 (GBDT gradient feedback coefficient, engineering experience value).

[0282] (3) MARS and GBDT interaction

[0283] Forward interaction (MARS→GBDT): GBDT fits the residuals of MARS, interaction formula:

[0284] ;

[0285] In other words, MARS provides the initial fitting residuals for GBDT, thus clarifying the optimization objective of GBDT;

[0286] Reverse Interaction (GBDT→MARS): GBDT residuals correct the regression coefficients of MARS. Interaction formula:

[0287] ;

[0288] That is, the high-precision fitting results of GBDT are used to back-optimize the basis function coefficients of MARS, thereby achieving synergistic improvement of the two models.

[0289] 4.2 Termination Conditions of the Interactive Closed Loop of the MARS+PSO+GBDT Fusion Algorithm

[0290] In this scheme, the interactive closed loop consisting of MARS (Multivariate Adaptive Regression Spline), PSO (Particle Swarm Optimization), and GBDT (Gradient Boosting Decision Tree) – “MARS→PSO→GBDT→MARS→PSO” – aims to improve the accuracy of sand trapping rate prediction through iterative cycles of “parameter optimization – residual fitting – gradient feedback – parameter re-optimization”. The termination conditions are divided into a core convergence condition (which must be met) and auxiliary termination conditions (backup constraints). All conditions must be judged in conjunction with quantitative indicators and iterative stability, as detailed below:

[0291] 4.2.1 Core convergence condition (necessary condition for loop termination)

[0292] The core conditions must be met simultaneously to ensure that the algorithm interaction reaches a state of "accuracy meets the standard + parameter stability", which is the core basis for terminating the closed loop.

[0293] 1. Convergence condition for global prediction error

[0294] The prediction error of the fusion model for the sand trapping rate is lower than the preset threshold, and there is no significant fluctuation in multiple iterations, which shows that the model fitting accuracy meets the engineering requirements.

[0295] (1) Quantitative judgment criteria

[0296] Relative error threshold: The relative error between the final predicted value and the measured value of the fusion model. (In engineering scenarios) (Use 1% to 2%, 1% in this embodiment).

[0297] Stability requirement: Continuous Wheel (in this embodiment) =3) Rate of change of relative error during iteration ( Take 0.1%.

[0298] (2) Calculation formula

[0299] ; ;

[0300] This is the fusion prediction value after the t-th round of closed-loop iteration;

[0301] This is the measured value of the sand-trapping rate;

[0302] Let be the relative error of the t-th iteration. This represents the relative error of the previous iteration.

[0303] (3) Physical meaning

[0304] In this embodiment =0.9, when RE≤1% (i.e. ∈[0.891,0.909]), and for 3 consecutive rounds. If the accuracy is ≤0.1%, it indicates that the model's prediction accuracy already meets the accuracy requirements of debris flow prevention and control engineering design, and there is no need for further iteration.

[0305] 2. Algorithm interaction gradient / parameter convergence condition

[0306] The particle position / velocity of PSO and the gradient of the loss function of GBDT have both converged to a steady state, indicating that there is no room for further improvement in parameter optimization and the interactive closed loop has entered a steady state.

[0307] (1) Quantitative judgment criteria

[0308] PSO parameter stability: Global optimal particle position change rate ( (Take 0.05%), and the absolute value of the particle velocity. ( Take 1e-5), consecutive Wheel (in this embodiment) =2) Satisfies;

[0309] GBDT gradient convergence: absolute value of the gradient of the loss function ( Take 1e-4), and the gradient rate of change ( Take 0.1%.

[0310] (2) Calculation formula

[0311] ; ;

[0312] Let be the globally optimal position of PSO in the t-th iteration; Let be the gradient of the GBDT loss function in the t-th iteration.

[0313] (3) Physical meaning

[0314] The stability of the PSO particle position / velocity indicates that the MARS nodes and GBDT hyperparameters have found the global optimum, leaving no room for further optimization; the convergence of the GBDT gradient indicates that the residual fitting has reached its limit, and further iterations cannot reduce the loss.

[0315] 4.2.2 Auxiliary Termination Conditions (Catch-all Constraints; terminating upon satisfying any one of them)

[0316] If the core convergence condition is not met for a long time (e.g., data noise causes errors that prevent convergence), auxiliary conditions should be used to terminate the closed loop to avoid infinite iteration and ensure computational efficiency.

[0317] 1. Maximum number of iterations threshold

[0318] The total number of closed-loop iterations has reached the preset limit. (This embodiment) =20 rounds), the process terminates regardless of whether the error converges.

[0319] Logical explanation: The interactive closed loop of MARS+PSO+GBDT has the characteristic of diminishing marginal benefits of iteration: the accuracy improvement is significant in the first 10 iterations, but the accuracy improvement in subsequent iterations is less than 0.1%. Therefore, setting a maximum number of iterations can balance accuracy and computational cost. In this embodiment, 20 iterations are sufficient to cover all effective optimization space.

[0320] 2. Stability condition of PSO fitness function

[0321] Global optimal fitness value of PSO continuous Wheel (in this embodiment) =5) Rate of change ( Taking 0.05% indicates that the parameter optimization has fallen into a local optimum, and further iteration yields no benefit.

[0322] Calculation formula: ;in Let be the globally optimal fitness value in the t-th iteration.

[0323] 3. Convergence conditions of joint residuals

[0324] MARS residuals Residual with GBDT The sum (total residuals) ( Take 1e-3), and it is continuous. Wheel (in this embodiment) =3) No decrease indicates that the residual fit has reached the minimum value acceptable for engineering.

[0325] Calculation formula: ; MARS residuals optimized for PSO; This represents the final residual after GBDT fitting.

[0326] 4.3 Logic for determining closed-loop termination

[0327] The termination determination process for the interactive closed loop in this embodiment is as follows:

[0328] 1. After each iteration, first verify the core convergence conditions (error convergence + gradient / parameter convergence). If both conditions are met, terminate the loop closure directly.

[0329] 2. If the core condition is not met, verify the auxiliary condition: if the number of iterations is ≥20 rounds, or the fitness function is stable for 5 consecutive rounds, or the total residual is ≤1e-3 and does not decrease for 3 consecutive rounds, then terminate the loop closure;

[0330] 3. If none of the above conditions are met, continue to the next iteration until any termination condition is triggered.

[0331] In this embodiment, when iterating to the 8th round, RE = 0.56% ≤ 1%, for 3 consecutive rounds. =0.08%≤0.1%, and =0.03%≤0.05% =8e-5≤1e-4, all core convergence conditions are satisfied, therefore the loop closure is terminated, and the final result is obtained. =0.895, achieving high-precision prediction of sand trapping rate.

[0332] 4.4 Core Role / Contribution of the Algorithm

[0333] (1) MARS algorithm

[0334] Core function: Constructing the sediment trapping rate and multiple input parameters ( The nonlinear initial model (etc.) solves the problem that traditional linear regression cannot fit complex nonlinear relationships;

[0335] Contribution: By capturing the local nonlinear characteristics between parameters through piecewise spline basis functions, a basic model framework is provided for subsequent optimization, and its residuals provide a clear optimization direction for PSO and GBDT.

[0336] (2) PSO algorithm

[0337] Core function: Collaborative optimization of parameters across algorithms, simultaneously optimizing the nodes of MARS and the hyperparameters of GBDT, solving the problem of single algorithm parameter optimization getting trapped in local optima;

[0338] Contribution: The optimal parameter combination of the two algorithms is found through global search characteristics, and the gradient feedback from GBDT is received to achieve dynamic adjustment, thereby improving the generalization ability of the overall model.

[0339] (3) GBDT algorithm

[0340] Core function: Fitting the residuals of MARS to achieve a two-layer prediction of "coarse fitting + fine fitting", solving the problem of insufficient fitting accuracy of MARS for small samples / extreme working conditions;

[0341] Contribution: By iteratively improving the model accuracy and feeding the loss gradient back to PSO, a closed loop of "optimization-feedback-re-optimization" is formed, ultimately reducing the error between the predicted and measured values ​​of the sand trapping rate to less than 1%, which is far superior to the traditional multiple regression model.

[0342] V. Verification Results of the Examples

[0343] In this embodiment, the measured sand-trapping rate =0.9, the final predicted value of the fusion algorithm =0.895, with a relative error of only 0.56%; compared with the traditional multiple regression model (error 7.64%), the accuracy is improved by 92.7%, which verifies the effectiveness of the MARS+PSO+GBDT fusion algorithm in the quantitative calculation of the sediment retention rate of wedge beam grid dams. Moreover, the pairwise interaction between the algorithms ensures the stability and generalization ability of the model, and it is suitable for the prediction of sediment retention rate in debris flow gullies in different mountainous areas.

[0344] In summary, several specific embodiments of the present invention have been disclosed. Without contradiction, the various embodiments can be freely combined to form new embodiments. That is, embodiments that are alternative solutions can be freely substituted for each other, but cannot be combined with each other; embodiments that are not alternative solutions can be combined with each other. These new embodiments are also part of the substantive content of the present invention.

[0345] The above embodiments describe several specific implementations of the present invention. However, those skilled in the art should understand that various changes or modifications can be made to these implementations without departing from the principles and essence of the present invention, but all such changes and modifications fall within the protection scope of the present invention.

Claims

1. A novel method for quantitatively calculating the debris flow interception rate of a wedge-shaped beam grid dam, characterized in that, include: Step 1: Select a section in the debris flow basin that meets the conditions for dam construction and obtain the following basic parameters: (1) Debris flow related parameters: debris flow unit weight, solid volume concentration, debris flow scale; (2) Parameters of gullies and boulders: average width of gullies in the cross section of the proposed dam, average longitudinal slope of gullies within the cross section, mud depth of debris flow in front of the dam, and minimum particle size of the largest boulder in the debris flow basin. (3) Dam structure parameters: The wedge structure angle ratio of the wedge beam grid dam is the ratio of the wedge structure angle of the dam body to 60°; Step 2: Based on the debris flow depth at the dam front section, the dam height of the wedge-shaped beam grid dam is initially determined. Combining the average channel width, the minimum particle size of the largest rock, the average channel longitudinal slope, the wedge structure angle ratio, and the debris flow scale, the ratio of the total amount of solid debris flow material to the reservoir capacity of the dam body corresponding to the dam height is calculated, i.e., the reservoir capacity ratio. Among them, the total amount of solid debris flow material is calculated by the debris flow scale and the solid volume concentration. Step 3: Obtain particle size distribution curves of debris flows in different regions within the debris flow basin, and select a particle size distribution curve that reflects the overall characteristics of the basin based on the proportion of large particles within the basin; determine the characteristic particle size of the debris flow sample based on this curve. And based on characteristic particle size Design the opening width of the wedge-shaped beam grid dam; Step 4: Construct a flume test device. Based on the dam height determined in Step 2 and the opening width designed in Step 3, create a wedge-shaped beam grid dam model to simulate debris flow movement and the interception process of the model under different working conditions. The working conditions should at least cover scenarios corresponding to different debris flow unit weights, different debris flow scales, different channel longitudinal slopes, different opening widths, different wedge structure angles, and different reservoir capacity ratios. Step 5: Based on the water tank test data from Step 4, and combined with the opening width and characteristic particle size from Step 3... The relative opening width was calculated, and then, taking into account six control factors, namely debris flow unit weight, debris flow scale, average channel longitudinal slope, relative opening width, wedge structure angle ratio, and reservoir capacity ratio, a multivariate regression analysis method was used to construct a sediment trapping rate prediction model. This model was used to quantitatively calculate the sediment trapping rate of the wedge beam grid dam.

2. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, In step two, another way to calculate the reservoir capacity ratio is: based on the preset reservoir capacity ratio, the height of the wedge beam grid dam that satisfies the reservoir capacity ratio is calculated in reverse using the parameters from step one, and the reservoir capacity of the dam body is determined based on the calculated dam height to obtain the reservoir capacity ratio.

3. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, The wedge structure angle ratio in step one ranges from 0.75 to 1.

25.

4. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, The reservoir capacity of the dam body in step two is calculated by taking into account the dam height, the average width of the channel, the sine value of the average longitudinal slope of the channel, and the tangent value corresponding to the angle ratio of the wedge structure. The specific adjustment is made by combining the influence of the minimum particle size value of the largest rock on the reservoir capacity.

5. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, Step 3 involves different regions of the debris flow basin, including debris flow deposition areas and flow areas. By obtaining the particle size distribution curves of the two regions respectively, and then combining them with the frequency weight of debris flow occurrence in the region, the particle size distribution curve of the overall characteristics of the basin is determined.

6. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, During the flue test in step four, the actual measurement of the sediment retention rate was carried out simultaneously: After the test, the silt in front of the dam was shoveled out and dried until completely dry, and the total amount of solid material intercepted was obtained by weighing. The actual sediment retention rate was calculated by combining the total mass of solid material flowing into the flue, which was used to verify the reliability of the sediment retention rate prediction model.

7. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, In step five, the multiple regression analysis was conducted, and four functional forms were tried: multiple linear function, multiple power function, multiple exponential function, and multiple logarithmic function. The optimal function was selected based on the fitting effect of each functional form to construct the sand trapping rate prediction model.

8. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, In step five, when constructing the sediment trapping rate prediction model, the coefficient of determination is used. To evaluate the goodness of fit of the multiple regression analysis, select... The largest functional form is used as the final prediction model form.

9. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 1, characterized in that, Step five employs a fusion of three algorithms—multivariate adaptive regression splines, particle swarm optimization, and gradient boosting decision trees—to construct a sand-trapping rate prediction model. Specifically, this includes: (51) The initial prediction model of sand-blocking rate is constructed by using the multivariate adaptive regression spline algorithm: six types of control factors are used as input variables and the measured value of sand-blocking rate is used as output variables. Piecewise spline basis functions are constructed. The basis functions are screened by a strategy of forward selection combined with backward pruning. The regression coefficients are solved by the least squares method. The initial prediction value and prediction residual are calculated by substituting the measured parameters on site. (52) Parameter co-optimization is performed using particle swarm optimization algorithm: The node parameters of multivariate adaptive regression spline algorithm, the learning rate and decision tree depth of gradient boosting decision tree algorithm are used as the core components of particle position. The sum of squared prediction residuals of multivariate adaptive regression spline algorithm is used as the fitness function. The particle velocity and position are adjusted by inertia weight and learning factor. After iteratively obtaining the global optimal parameters, they are fed back to multivariate adaptive regression spline algorithm to update the basis function and to gradient boosting decision tree algorithm as the initial hyperparameters. (53) Gradient boosting decision tree algorithm is used to fit the residuals and achieve bidirectional interaction: The residuals of multivariate adaptive regression spline prediction optimized by particle swarm optimization algorithm are used as the fitting target. The mean square error loss function is constructed and the decision tree is trained iteratively. The gradient of the loss function is fed back to the particle swarm optimization algorithm to adjust the direction of particle iteration. At the same time, the final fitting residuals are fed back to the multivariate adaptive regression spline algorithm to correct its regression coefficients. Through bidirectional interaction and iteration of the three types of algorithms, the sand trapping rate prediction model with improved accuracy is obtained.

10. The method for quantitatively calculating the debris flow interception rate of the novel wedge-shaped beam grid dam according to claim 9, characterized in that, Before constructing the sediment trapping rate prediction model in step five, the flume test data is divided into a training set and a test set in a 7:3 ratio. The training set data is used to train the model, and the test set data is used to verify the prediction accuracy of the model. The verification indicators include relative error and coefficient of determination. If the model prediction accuracy does not meet the engineering requirements, i.e., the relative error is greater than 2%, the algorithm parameters are readjusted and the model is trained again until the accuracy meets the requirements.

Images