A rapid prediction method for peak flow rate of dam burst
By employing a radial basis function neural network model to classify and dimensionlessly process landslide dam failure data, and combining the soil particle size and erosion degree of the landslide dam, the problem of not considering the influence of the erosion degree and soil particle size in existing technologies is solved, and high-precision prediction of the peak flow of landslide dam failure is achieved.
Patent Information
- Application Number
- CN202111042566.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-07
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2041-09-07
AI Technical Summary
Existing methods for predicting peak flow rates in landslide dam bursts do not consider the classification of erosion degree of the landslide dam and the influence of soil particle size, resulting in highly subjective prediction results, large data fluctuations, low accuracy, and neglect of nonlinear relationships between parameters.
A radial basis function neural network model was used to classify and dimensionlessly process the landslide dam failure data. Then, a classic three-layer neural network model was used to perform nonlinear fitting based on the soil particle size and erosion degree of the landslide dam to predict the peak flow rate of the landslide dam failure.
It has enabled more accurate and objective prediction of peak flow of landslide dam failure, improved the scientific nature and accuracy of the prediction results, and provided high-precision data support for landslide dam failure rescue and flood inversion analysis.
Smart Images

Figure QLYQS_1 
Figure QLYQS_2 
Figure QLYQS_3
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of water conservancy projects and relates to a method for quickly predicting a peak flow rate of a dam burst. Background Art
[0002] Rapid prediction of peak flow rate is very important in the rescue of landslide dam burst. However, most of the existing prediction methods are based on a small number of representative parameters, especially without considering the classification of landslide erosion degree and the important influence of landslide soil particle size on landslide burst. There is a lack of rapid prediction methods for peak flow rate of landslide burst based on comprehensive parameters. In the process of rapid prediction of peak flow rate of landslide burst, if only individual characteristic parameters are used to simulate the peak flow, there are defects such as strong subjectivity, large data fluctuations and low accuracy. At the same time, the existing methods do not classify the degree of erosion of the landslide body, do not consider the influence of the particle size of the landslide body soil on the peak flow rate of burst, and cannot accurately calculate the peak flow rate of landslide burst. In addition, most of the existing methods use linear regression method to fit the parameter model for calculation. This method only considers linear factors and ignores the nonlinear relationship between each parameter and the burst flow. At present, there are few rapid prediction models for peak flow rate of landslide burst that consider more comprehensive parameter information and nonlinear fitting relationships at the same time. Summary of the invention
[0003] The purpose of the present invention is to provide a method for quickly predicting the peak flow rate of a landslide body burst, so as to make up for the defects of the existing prediction methods that do not consider the classification of the erosion degree of the landslide body and the influence of the particle size of the landslide body soil on the burst flow, and provide data support for the rescue of the landslide lake burst, the demarcation of the evacuation range downstream of the landslide lake, and the inversion analysis of the landslide body burst flood, so as to obtain more accurate results.
[0004] The technical solution adopted by the present invention is a method for quickly predicting the peak flow rate of a dam burst, which is specifically implemented according to the following steps:
[0005] Step 1, collecting dam breach data samples;
[0006] Step 2, classifying the data samples collected in step 1 according to the degree of erosion of the dam body;
[0007] Step 3, performing dimensionless processing on the data samples collected in step 1;
[0008] Step 4, performing radial basis neural network model training on the data processed in step 3;
[0009] Step 5, the data sample of the landslide dam burst to be predicted is processed in step 3 and then input into the model obtained in step 4 to obtain the peak flow of the landslide dam burst.
[0010] The present invention is also characterized in that
[0011] In step 1, the dam body breach data samples include the dam body name, dam body location, breach time process, dam height, dam width, dam body volume, dam lake storage capacity, breach depth, dam body erosion degree, dam body soil particle size parameters, and breach peak flow.
[0012] The particle size parameters of the dam body soil include d 30 d 50 d 90 .
[0013] In step 2, the erosion degree of the dam body is divided according to the soil type of the dam body, as shown in the following table:
[0014]
[0015] The specific process of step 3 is:
[0016] Step 3.1, calculate the dam shape coefficient A1, the expression is:
[0017]
[0018] In formula (1), H d Indicates the dam height, m; V d represents the volume of the dam, 10 6 m 3 ;
[0019] Step 3.2, calculate the lake surface shape coefficient A2, the expression is:
[0020]
[0021] In formula (2), S represents the reservoir capacity of the barrier lake, 10 6 m 3 ;
[0022] Step 3.3, calculate the dam height breach depth ratio A3, the expression is:
[0023]
[0024] In formula (3), H w Indicates the breach depth, m;
[0025] Step 3.4, calculate the dam height to dam width ratio A4, the expression is:
[0026]
[0027] In formula (4), B d Indicates the dam width, m;
[0028] Step 3.5, calculate the peak flow parameter A5, the expression is:
[0029]
[0030] In formula (5), g represents the local gravity acceleration, Q p is the peak flow rate of the dam body, m 3 / s.
[0031] The specific process of step 4 is:
[0032] Step 4.1, divide the data samples classified in step 2 into training set, verification set and test set;
[0033] Step 4.2: Input the training set samples in each type of data into the classic three-layer neural network model for training, and obtain three types of trained models, among which the particle size d of the dam body soil is 30 , particle size of soil in the dam body d 50 , particle size of soil in the dam body d 90 Parameters, dam shape coefficient, lake shape coefficient, dam height breach depth ratio, dam height dam width ratio are put into the input layer of the classic three-layer neural network model as dimensionless variables, peak flow parameters are put into the output layer of the classic three-layer neural network model, and the radial basis function connecting the output layer and the hidden layer adopts the classic Gaussian function.
[0034] Step 4.3, input the test set samples in each type of data into the trained network model corresponding to step 4.2 to obtain the verification data, and use the root mean square error test method for verification. The root mean square error expression is:
[0035]
[0036] In formula (7), RMSE represents the root mean square error, n represents the total number of samples in the test set, and Q i represents the peak flow parameter corresponding to the i-th sample, represents the verification peak flow parameter corresponding to the i-th sample;
[0037] Setting reference value like Then the training of the network model in step 4.2 is ended, and the trained network model is obtained; if Then return to step 4.2, reselect the number of training times, expansion constant, and number of hidden layer units for retraining;
[0038] Step 4.4, input the test set samples in each type of data into the corresponding trained network model obtained in step 4.3 for testing, and use the calculation efficiency coefficient E to evaluate the accuracy of the trained network model. The expression of the efficiency coefficient E is:
[0039]
[0040] In formula (8), Represents the mean of the peak flow parameters of all samples.
[0041] In step 4.1, each type of data sample is divided into 7:1.5:1.5, where 70% is the training set, 15% is the verification set, and 15% is the test set.
[0042] In step 4.2, the expression of the classical Gaussian function is:
[0043]
[0044] In the formula, x represents the input layer vector, which should be a column vector composed of input layer parameters in this model. i represents the central parameter of each hidden layer, and γ represents the expansion constant of the basis function.
[0045] The specific process of step 5 is: collect the data samples of the dam body to be predicted, input the data samples into the tested model, obtain the peak flow parameter A5, and obtain the peak flow rate Q of the dam body according to formula (5): p .
[0046] The beneficial effect of the present invention is that the present invention is a method for quickly predicting the peak flow rate of a landslide body burst, which takes into account the influence of the degree of erosion of the landslide body and the particle size of the soil of the landslide body on the burst, adopts comprehensive parameters and takes into account the nonlinear relationship between the parameters and the peak flow rate of the burst, and performs a more scientific and accurate rapid prediction of the peak flow rate of the landslide body burst, so that the prediction result is more accurate, more objective and reasonable, and can provide high-precision data support and decision-making basis for work such as emergency rescue of landslide lake burst, demarcation of evacuation range downstream of the landslide lake, and inversion analysis and research on landslide lake flood. DETAILED DESCRIPTION
[0047] The present invention is described in detail below in conjunction with specific implementation modes.
[0048] The present invention provides a method for quickly predicting the peak flow rate of a dam burst, which is specifically implemented according to the following steps:
[0049] Step 1: Collect data samples of landslide body breach, including landslide body name, landslide body location, breach time, dam height, dam width, dam body volume, reservoir capacity of landslide lake, breach depth, landslide body erosion degree, landslide body soil particle size parameters (a parameter widely used in the field of geotechnical engineering, including d 30 d 50 d 90 , as shown in Table 1), peak flow rate;
[0050] Table 1. Granularity parameters of soil mass in dam body
[0051]
[0052] Step 2: Classify the data samples collected in step 1 according to the erosion degree of the dam body. As shown in Table 2, the erosion degree of the dam body is specifically divided into extremely high erosion degree I, high erosion degree II, medium erosion degree III, low erosion degree IV, extremely low erosion degree V, and no erosion degree according to the type of dam body soil.
[0053] Table 2 Classification of erosion degree of dam body
[0054]
[0055] Step 3, performing dimensionless processing on the data samples collected in step 1;
[0056] Step 3.1, calculate the dam shape coefficient A1, the expression is:
[0057]
[0058] In formula (1), H d Indicates the dam height, m; V d represents the volume of the dam, 10 6 m 3 ;
[0059] Step 3.2, calculate the lake surface shape coefficient A2, the expression is:
[0060]
[0061] In formula (2), S represents the reservoir capacity of the barrier lake, 10 6 m 3 ;
[0062] Step 3.3, calculate the dam height breach depth ratio A3, the expression is:
[0063]
[0064] In formula (3), H w Indicates the breach depth, m;
[0065] Step 3.4, calculate the dam height to dam width ratio A4, the expression is:
[0066]
[0067] In formula (4), B d Indicates the dam width, m;
[0068] Step 3.5, calculate the peak flow parameter A5, the expression is:
[0069]
[0070] In formula (5), g represents the local gravity acceleration, Q p is the peak flow rate of the dam body, m 3 / s;
[0071] Step 4, performing radial basis neural network model training on the data processed in step 3;
[0072] Step 4.1, divide the data samples classified in step 2 into training set, verification set and test set;
[0073] Each type of data sample is divided into 7:1.5:1.5, of which 70% is the training set, 15% is the verification set and 15% is the test set;
[0074] Step 4.2: Input the training set samples in each type of data into the classic three-layer neural network model for training, where the particle size d of the dam body soil is 30 , particle size of soil in the dam body d 50 , particle size of soil in the dam body d 90 The parameters, dam shape coefficient, lake surface shape coefficient, dam height breach depth ratio, and dam height width ratio are put into the input layer of the classic three-layer neural network model as dimensionless variables. The peak flow parameter is used as the output layer of the classic three-layer neural network model. The radial basis function connecting the output layer and the hidden layer adopts the classic Gaussian function, and the expression is:
[0075]
[0076] In the formula, x represents the input layer vector, which should be a column vector composed of input layer parameters in this model; c i represents the central parameter of each hidden layer, and γ represents the expansion constant of the basis function;
[0077] The present invention adopts Gaussian function as radial basis function, and takes into account the influence of erosion degree of dam body and particle size of dam body soil on the burst flow of dam body;
[0078] Step 4.3, input the test set samples in each type of data into the trained network model corresponding to step 4.2 to obtain the verification data, and use the root mean square error test method for verification. The root mean square error expression is:
[0079]
[0080] In formula (7), RMSE represents the root mean square error, n represents the total number of samples in the test set, and Q i represents the peak flow parameter corresponding to the i-th sample, represents the verification peak flow parameter corresponding to the i-th sample;
[0081] Setting reference value like Then the training of the network model in step 4.2 is ended, and the trained network model is obtained; if Then return to step 4.2, reselect the number of training times, expansion constant, and number of hidden layer units for retraining;
[0082] Step 4.4, input the test set samples in each type of data into the corresponding trained network model obtained in step 4.3 for testing, and use the calculation efficiency coefficient E to evaluate the model accuracy. The expression of the efficiency coefficient E is:
[0083]
[0084] In formula (8), It represents the mean of the peak flow parameters of all samples. The closer the efficiency coefficient E is to 1, the better the fitting effect is.
[0085] Step 5: Collect the data samples of the dam body to be predicted, input the data samples into the tested model, obtain the peak flow parameter A5, and obtain the peak flow rate Q of the dam body according to formula (5): p .
Claims
1. A method for rapidly predicting peak flow of a dam burst, characterized in that: Please follow the steps below to implement it: Step 1: Collect dam breach data samples; In step 1, the landslide body breach data sample includes the name of the landslide body, the location of the landslide body, the breach time, the dam height, the dam width, the dam body volume, the reservoir capacity of the landslide lake, the breach depth, the degree of erosion of the landslide body, the particle size parameters of the landslide body soil, and the breach peak flow rate; The particle size parameters of the dam body soil include d 30 d 50 d 90 ; Step 2: classify the data samples collected in step 1 according to the degree of erosion of the dam body; Step 3: Perform dimensionless processing on the data samples collected in step 1; The specific process of step 3 is: Step 3.1, calculate the dam shape coefficient A1, the expression is: In formula (1), H d Indicates the dam height, m; V d Indicates the volume of the dam, m 3 ; Step 3.2, calculate the lake surface shape coefficient A2, the expression is: In formula (2), S represents the reservoir capacity of the barrier lake, m 3 ; Step 3.3, calculate the dam height breach depth ratio A3, the expression is: In formula (3), H w Indicates the depth of the breach, m; Step 3.4, calculate the dam height to width ratio A4, the expression is: In formula (4), B d Indicates the dam width, m; Step 3.5, calculate the peak flow parameter A5, the expression is: In formula (5), g represents the local gravity acceleration, Q p is the peak discharge of the dam body, m 3 / s; Step 4, performing radial basis neural network model training on the data processed in step 3; The specific process of step 4 is: Step 4.1: Divide the data samples classified in step 2 into training set, verification set and test set; Step 4.2: Input the training set samples in each type of data into the classic three-layer neural network model for training, and obtain three types of trained models, in which the particle size d of the landslide body soil is 30 , particle size of dam soil d 50 , particle size of dam soil d 90 The parameters, dam shape coefficient, lake surface shape coefficient, dam height breach depth ratio, and dam height width ratio are put into the input layer of the classic three-layer neural network model as dimensionless variables. The peak flow parameter is used as the output layer of the classic three-layer neural network model. The radial basis function connecting the output layer and the hidden layer adopts the classic Gaussian function. In step 4.3, input the test set samples in each category of data into the trained network model corresponding to step 4.2 to obtain the verification data, and use the root mean square error test method for verification. The root mean square error expression is: In formula (7), RMSE represents the root mean square error, n represents the total number of test set samples, Q i represents the peak flow parameter corresponding to the i-th sample, represents the verification peak flow parameter corresponding to the i-th sample; Setting reference value like Then the training of the network model in step 4.2 is completed and the trained network model is obtained; if Then return to step 4.2 and reselect the number of training times, expansion constant, and number of hidden layer units for retraining; In step 4.4, the test set samples in each category of data are input into the corresponding trained network model obtained in step 4.3 for testing. The accuracy of the trained network model is evaluated using the calculation efficiency coefficient E. The expression of the efficiency coefficient E is: In formula (8), represents the mean of the peak flow parameters of all samples; Step 5: After the data samples of the landslide dam failure to be predicted are processed in step 3, they are input into the model obtained in step 4 to obtain the peak flow rate of the landslide dam failure.
2. A rapid prediction method for peak flow of a landslide dam according to claim 1, characterized in that: In step 2, the erosion degree of the landslide dam is divided according to the landslide dam soil type, as shown in the following table:
3. A method for rapid prediction of peak flow of landslide dam burst according to claim 1, characterized in that: In step 4.1, each type of data sample is divided into 7:1.5:1.5, of which 70% is the training set, 15% is the verification set, and 15% is the test set.
4. A method for rapid prediction of peak flow of a landslide dam according to claim 1, characterized in that: In step 4.2, the expression of the classical Gaussian function is: Where x represents the input layer vector, which should be a column vector composed of input layer parameters in this model; c i represents the central parameter of each hidden layer, and γ represents the expansion constant of the basis function.
5. The method for rapid prediction of peak flow rate of landslide dam burst according to claim 1, characterized in that: The specific process of step 5 is: collect the data samples of the landslide body to be predicted, input the data samples into the tested model, obtain the peak flow parameter A5, and obtain the peak flow rate Q of the landslide body according to formula (5) p .
Citation Information
Patent Citations
Weir dam burst mechanism analysis method for drainage channel excavation measures
CN109657281A
Barrier lake dangerous classification and landslide-type debris flow and flash flood warning method
CN110533884A
Landslide barrier dam breach depth measuring and calculating method and landslide barrier dam burst flood peak flow measuring and calculating method
CN112182993A
