Construction method of crack water seepage prediction model, prediction model, prediction method, electronic device and storage medium

Abstract

This invention discloses a method for constructing a crack seepage prediction model, a prediction model, a prediction method, an electronic device, and a storage medium, comprising the following steps: S01, identifying the key factors affecting crack seepage; S02, listing corresponding orthogonal experimental tables of crack seepage characteristics based on the selected factors and levels; S03, using simulation software to simulate and analyze the flow field of crack seepage inside an aircraft under rain conditions; S04, setting corresponding simulation parameters according to the orthogonal experimental tables in step S02, and recording the simulation results as the crack outlet flow rate and velocity under each simulation condition; S05, building an adaptive learning algorithm model, using the simulation data obtained in step S04 as training samples to train the adaptive learning algorithm model, thereby obtaining the crack seepage prediction model. The prediction model constructed by this invention is highly practical and can be applied to the prediction and analysis of crack seepage under rain conditions for most equipment.

Description

Technical Field

[0001] This invention belongs to the field of computer modeling and simulation technology, specifically relating to a method for constructing a crack seepage prediction model, a prediction model, a prediction method, and electronic equipment and storage medium. Background Technology

[0002] Currently, whether there is water leakage in the gaps of aircraft structures is mainly verified by testing. However, testing has problems such as large scale, high cost of manpower and materials, long cycle, and low efficiency in solving problems.

[0003] With the rapid development of computing and simulation technologies, simulation has played a crucial role in process design and R&D. However, when it comes to the problem of water seepage through structural gaps in aircraft under rain conditions, predicting this using general simulation methods or small models typically takes several hours to a day, while large-scale models often require several days to several weeks. This results in low predictive analysis efficiency and makes it difficult to guarantee the accuracy of the predictions. Summary of the Invention

[0004] The purpose of this invention is to provide a method for constructing a crack seepage prediction model, a prediction model, a prediction method, an electronic device, and a storage medium, so as to solve the problems of low efficiency and low accuracy of existing methods for predicting and analyzing crack seepage in aircraft structures.

[0005] This invention is achieved through the following technical solution:

[0006] The method for constructing a crack seepage prediction model includes the following steps:

[0007] S01. Identify the key factors affecting water seepage through the cracks, including the length of the crack, the width of the crack, and the surface roughness of the wall.

[0008] S02. Select three specific factors as test factors, and select multiple levels corresponding to each factor according to the common working conditions of the internal gaps of the aircraft, including length level, width level, and roughness level. List the corresponding orthogonal test table of gap seepage characteristics according to the selected factors and levels.

[0009] S03. Simulation software was used to analyze the flow field of water seepage through the gaps inside the aircraft under rain conditions.

[0010] S04. Set the corresponding simulation parameters according to the orthogonal experimental table in step S02, and obtain the simulation results under various simulation conditions, including the flow rate and velocity at the gap outlet.

[0011] S05. Build an adaptive learning algorithm model. Use the simulation data obtained in step S04 as training samples to train the adaptive learning algorithm model and obtain the crack seepage prediction model.

[0012] In some embodiments of the present invention, in step S03, a velocity inlet for rainwater entry is set at the top of the gap, and a pressure outlet is set on the unrelated right side wall and the bottom gap outlet wall to establish a simulation model of the gap seepage flow field.

[0013] In some embodiments of the present invention, in step S04, the influence of each factor on the flow field of seepage in the crack is analyzed according to orthogonal experiments, and the range is calculated. The influence of each factor on the flow field of seepage in the crack is ranked according to the range, and the construction of the crack seepage prediction model is optimized according to the ranking result.

[0014] In some embodiments of the present invention, the constructed adaptive learning algorithm model includes an input layer L1, a hidden layer L2, and an output layer L3;

[0015] The input layer has n nodes, and the hidden layer has p nodes. Nodes are connected by a weight matrix W, and the weight vector is... Where i represents the number of nodes in the input layer and j represents the number of nodes in the hidden layer;

[0016] The outputs of the n processing units in the input layer are represented as X = (x1, …, x2) i , …,x n ) T ;

[0017] The threshold for the hidden layer processing unit is θ j This indicates that for any node j in the hidden layer processing unit, the obtained input weighted sum s j Represented as:

[0018] ;

[0019] Where n is the number of input layer nodes, w ij x is the weight from the i-th node in the input layer to the j-th node in the hidden layer. i It is the output value of the i-th node in the input layer;

[0020] The calculated weighted sum s j The input is a transition function, and the activation function of the hidden layer is a sigmoid function. This determines the output y of the j-th node in the hidden layer. j for:

[0021] ;

[0022] The weight vector between the hidden layer and the output layer is represented as follows: ,in, v jtThis represents the connection weight from the j-th node in the hidden layer to the t-th node in the output layer, where t represents the number of nodes in the output layer. The output layer has a total of q processing units, and the activation function of the output layer is a sigmoid function.

[0023] In some embodiments of the present invention, the step of using the simulation data obtained in step S04 as training samples to train the adaptive learning algorithm model includes:

[0024] S06. Input one set of samples. Select three variables as input parameters: gap length, gap width, and wall roughness. The output parameter is the gap outlet flow rate c. t ;

[0025] S07. Repeat step S06, using the error backpropagation algorithm, combined with the desired output vector. Calculate the correction error values ​​for all processing units in the output layer, expressed as:

[0026] ;

[0027] In the formula, l t The weighted sum of the inputs to the output layer. The derivative of the transfer function;

[0028] Based on the correction error of the output layer, the correction error value of each unit in the hidden layer is calculated and expressed as:

[0029] ;

[0030] in, The derivative of the Sigmoid function;

[0031] Using learning rate Update the weights connecting the hidden and output layers. and the bias of the output layer ;

[0032] Use another learning rate Update the weights connecting the input layer and the hidden layer. and the bias of the hidden layer ;

[0033] Determine the degree to which the overall error E of the adaptive learning algorithm matches the required accuracy, and complete the training of the adaptive learning algorithm model.

[0034] In some embodiments of the present invention, in step S07, it is determined whether E is less than or equal to the accuracy setting value ε. If it is satisfied, the training ends; if it is not satisfied, the training continues to repeat to obtain the crack seepage prediction model.

[0035] On the other hand, the present invention also provides a crack seepage prediction model, which is constructed using the crack seepage prediction model construction method.

[0036] On the other hand, the present invention also provides a method for predicting crack seepage, which uses the aforementioned crack seepage prediction model to predict crack seepage.

[0037] On the other hand, the present invention also provides an electronic device, comprising:

[0038] Processor; and,

[0039] Memory for storing the executable instructions of the processor;

[0040] The processor is configured to execute the crevice seepage prediction method by executing the executable instructions.

[0041] On the other hand, the present invention also provides a computer storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned method for predicting water seepage through cracks.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] This invention qualitatively analyzes the order of influence of various factors on crack seepage by using orthogonal experiments and range analysis, which can improve simulation efficiency. A crack seepage prediction model is established based on an adaptive learning algorithm, and the model is trained using simulation data. The mean square error is used as the loss function, and the backpropagation algorithm, which adjusts its own parameters according to the error of the prediction results, improves the accuracy of the prediction results.

[0044] The predictive model constructed in this invention can quickly predict the water seepage of gaps based on the gap parameters. It uses simulation to analyze the impact of various factors on the internal gaps of aircraft under rain conditions, eliminating the need for rain tests, reducing the number of sample tests and test costs, and shortening the product development cycle.

[0045] The predictive model built using an adaptive learning algorithm can quickly predict the seepage flow rate of cracks under given parameters. Compared with manual modeling and simulation, it only requires inputting the parameters of the cracks during the design phase to quickly output the seepage flow rate of the cracks so as to determine whether the cracks will seep water under rain conditions, which greatly improves the efficiency of prediction.

[0046] The prediction model constructed in this invention is highly practical and can provide important reference for engineering design and water resource management in different fields. It has broad practical application prospects and can be applied to the prediction and analysis of water seepage in gaps of most equipment under rain conditions. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the gap geometry modeling in an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram of a simulation model of the seepage flow field in the gaps in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0051] In some embodiments of the present invention, a method for constructing a crack seepage prediction model includes the following steps:

[0052] S01. Based on known experience and laminar flow rate calculation formula, determine the key factors affecting water seepage in the gap, including the length of the gap, the width of the gap, and the surface roughness.

[0053] S02. Select the three factors determined in step S01 as test factors, and select three appropriate levels according to the common working conditions of the gaps inside the aircraft. List the corresponding orthogonal test table of gap seepage characteristics based on the three factors and three levels.

[0054] In orthogonal experimental design, the level refers to the specific parameter value selected for each factor. Taking the internal gap of an aircraft as an example, the gap length can be selected at three levels: 10mm, 20mm, and 30mm, which represent the typical assembly gap length range; the gap width can be selected at three levels: 0.1mm, 0.5mm, and 1.0mm, which represent the typical assembly gap width range; and the wall roughness can be selected at three levels: Ra0.8μm, Ra1.6μm, and Ra3.2μm, corresponding to different surface finishing grades.

[0055] S03. Simulation software was used to analyze the flow field of water seepage through the gaps inside the aircraft under rain conditions.

[0056] To improve simulation efficiency and mesh quality, the flow field of water seepage in the crack is simplified. A multiphase flow model is used, and a laminar flow model is adopted. A velocity inlet for rainwater entry is set at the top of the crack, and a pressure outlet is set on the unrelated right side wall and the bottom crack outlet wall.

[0057] S04. Set the corresponding simulation parameters according to the orthogonal test table in step S02, and record the simulation results as the gap outlet flow rate and velocity under each simulation condition.

[0058] The influence of each factor on the flow field of seepage in the crack is analyzed by orthogonal experiment, and the range is calculated. The influence of each factor on the flow field of seepage in the crack is ranked in order of importance.

[0059] Based on the ranking results obtained from range analysis, when constructing the model, the weights of the hidden layer neurons can be assigned. The connection weights of the width parameter, which has a significant impact on the range, can be initialized to [-1, 1], while the connection weights of the roughness parameter, which has a smaller impact on the range, can be further narrowed during initialization. This improves the efficiency of the prediction model construction and the accuracy of the prediction. Simultaneously, range analysis can be applied to engineering optimization. For example, in engineering optimization, adjusting the width parameter to a suitable value can effectively reduce the penetration amount.

[0060] The flow velocity value output by the simulation can be used for orthogonal experimental range analysis. The flow rate can be calculated from the flow velocity and cross-sectional area, and this flow rate value is used as the final output index.

[0061] The flow velocity values ​​output from the simulation can be used to verify the model. For example, when the error between the predicted flow rate and the actual simulated flow rate is greater than 10%, the model can be verified by observing abnormal distribution of the flow velocity field.

[0062] S05. A three-layer adaptive learning algorithm model is used to predict the flow field of seepage water in the crack. The adaptive learning algorithm model includes an input layer L1, a hidden layer L2 and an output layer L3. The input layer is set with n processing units and is connected to the hidden layer by weights.

[0063] The input layer has n nodes, and the hidden layer has p nodes. Nodes are connected by a weight matrix W, and the weight vector is... , where i represents the number of nodes in the input layer and j represents the number of nodes in the hidden layer.

[0064] The outputs of the n processing units in the input layer are the column vector of inputs to all processing units in the hidden layer, represented as X=(x1, …, x i , …,x n ) T The threshold for the hidden layer processing unit is θ. j This indicates that for any node j in the hidden layer processing unit, the obtained input weighted sum s j It can be represented as:

[0065] ;

[0066] Where n is the number of input layer nodes, w ij x is the weight from the i-th node in the input layer to the j-th node in the L2 hidden layer. i It is the output value of the i-th node in the input layer.

[0067] The calculated weighted sum s j We input a transition function, here choosing a sigmoid activation function for the hidden layer, to determine the output y of the j-th node in the hidden layer. j for:

[0068] .

[0069] The weight vector between the hidden layer and the output layer is represented as follows: Where t represents the number of nodes in the output layer, and the output layer has a total of q processing units. v jt This represents the connection weight from the j-th node in the hidden layer to the t-th node in the output layer. j=1,2,3,4 corresponds to 4 hidden neurons, and t=1 corresponds to a single output node. The activation function of the output layer is also a sigmoid function.

[0070] S06. Based on the adaptive learning algorithm model built in step S05, input one set of samples. The input parameters are selected as three variables: gap length, gap width, and wall roughness. The output parameter is the gap outlet flow rate c. t .

[0071] S07. Select the data from step S04 as training sample data, repeat step S06, and use the backpropagation algorithm combined with the desired output vector. Calculate the correction error values ​​for all processing units in the output layer, expressed as:

[0072] ;

[0073] In the formula, l t The weighted sum of the inputs to the output layer. is the derivative of the transfer function.

[0074] The correction error value is the difference between the expected output and the actual output multiplied by the derivative of the transfer function.

[0075] Based on the correction error of the output layer, the correction error value of each unit in the hidden layer is calculated and expressed as:

[0076] ;

[0077] in, This is the derivative of the Sigmoid function, used for chain-like differentiation during error backpropagation.

[0078] Using learning rate To update the weights connecting the hidden layer and the output layer and the bias of the output layer Use another learning rate Update the weights connecting the input layer and the hidden layer. and the bias of the hidden layer .

[0079] The overall error E of the adaptive learning algorithm is judged to be consistent with the required accuracy, i.e., E is less than or equal to ε. If this condition is met, the training ends; otherwise, the training continues to repeat until the crack seepage prediction model is obtained.

[0080] On the other hand, in some embodiments of the present invention, there is a crack seepage prediction model, which is constructed using the above-mentioned crack seepage prediction model construction method.

[0081] This crack seepage prediction model can be applied to the prediction and analysis of crack seepage under rain conditions for most equipment.

[0082] On the other hand, in some embodiments of the present invention, a method for predicting crack seepage is provided, which uses the above-mentioned crack seepage prediction model to predict crack seepage.

[0083] The crack seepage prediction model obtained from the above training is used to predict the crack seepage flow rate. By inputting the design parameters of the crack into the crack seepage prediction model, the crack seepage flow rate can be quickly output, thereby determining whether the crack will seep water under rain conditions, which greatly improves the efficiency of prediction analysis.

[0084] On the other hand, in some embodiments of the present invention, an electronic device is provided, comprising:

[0085] Processor; and,

[0086] Memory for storing the executable instructions of the processor;

[0087] The processor is configured to execute the crevice seepage prediction method in the above embodiments by executing executable instructions.

[0088] On the other hand, in some embodiments of the present invention, a computer storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the crack seepage prediction method in the above embodiments is implemented.

[0089] The following detailed description, in conjunction with specific embodiments, illustrates the construction of the crack seepage prediction model of the present invention and the method for predicting crack seepage based on the constructed prediction model.

[0090] Reference Figure 1 This is a geometric modeling diagram of the gap between two parts inside an aircraft, where h is the gap length and d is the gap width. The following section, based on this geometric modeling diagram, explains the construction method of the gap seepage prediction model.

[0091] S01. In one-dimensional longitudinal flow, the mass conservation equation for water seepage through cracks can be simplified to:

[0092] ;

[0093] in, y is the liquid density, A is the area of ​​the flow cross section, v is the flow velocity, y is the coordinate along the vertical direction, and t is time.

[0094] The momentum conservation equation can be written as:

[0095] ;

[0096] Where μ is the dynamic viscosity of the liquid, v is the flow velocity, and g is the acceleration due to gravity. It is the wall shear stress.

[0097] Based on the above formula, the influencing factors of water seepage through cracks under rain conditions are determined, including crack length, crack width, and wall roughness.

[0098] S02. Select the three factors from S01 (gap length, gap width, and wall roughness) as test factors, and refer to the common working conditions of aircraft internal gaps to select three appropriate levels respectively. List the factor level table for the orthogonal test of gap seepage characteristics, for a total of 3. 3 =27 groups of experiments.

[0099] S03. Simulation software was used to analyze the flow field of water seepage through the internal gaps of the aircraft under rain conditions. To improve simulation efficiency and mesh quality, the flow field was simplified using a multiphase flow model and a laminar flow model. A velocity inlet for rainwater entry was set at the top of the gap, and pressure outlets were set on the unrelated right side wall and the bottom gap outlet wall. A schematic diagram of the simulation model of the flow field of water seepage is shown below. Figure 2 As shown.

[0100] S04. Set the corresponding simulation parameters according to the orthogonal test table in step S02, and record the simulation results as the gap outlet flow rate and velocity under each simulation condition.

[0101] The influence of each factor on the flow field of seepage in the crack is analyzed by orthogonal experiment, and the range is calculated. The influence of each factor on the flow field of seepage in the crack is ranked in order of importance.

[0102] The steps for range analysis of orthogonal experiment results include:

[0103] S041. Calculate the total level value for each parameter, denoted by T. Column j displays the total level value of the i-th parameter. T ji express, T ji The value can be represented as:

[0104] ;

[0105] This indicates that the process is performed at the i-th level. Second result y i sum; This represents the total number of times the parameter in column j is at the same level.

[0106] S042. Calculate the average value of the experimental index corresponding to each level in column j. , represented as:

[0107] ;

[0108] S043, Calculate the range D j , which is equal to the maximum value minus the minimum value among the average values ​​of the test indicators corresponding to each level in column j. The calculation formula is as follows:

[0109] .

[0110] In this embodiment, range analysis is used to rank the influence of factors on the flow field of seepage in the cracks, resulting in: width range > length range > roughness range. The crack seepage prediction model is then optimized based on this ranking. For example, when assigning weights to neurons in the hidden layer during model construction, the connection weights for the width parameter, which has a significant impact on the range, are initialized to [-1, 1], while the connection weights for the roughness parameter, which has a smaller impact on the range, can be further narrowed during initialization. Furthermore, during engineering optimization, engineering parameters can be preferentially adjusted based on the ranking results; for example, prioritizing the adjustment of the width to a suitable value can effectively reduce leakage.

[0111] S05. A three-layer adaptive learning algorithm model is selected to predict the flow field of seepage water in the crack; wherein, the input layer L1 is set with 3 processing units, and the input mode vector is... The three features of the input data (i.e., parameters related to water seepage through cracks) correspond to the three nodes of the input layer: x1: length (normalized to the [0,1] interval), x2: width (normalized to the [0,1] interval), x3: roughness (normalized to the [0,1] interval); where m is the number of learning pattern pairs.

[0112] The expected output vector corresponding to the input pattern is The output layer has only one neuron, which is used to predict the flow rate of water seeping through the crack;

[0113] The hidden layer L2 has 4 processing units, and the net input vector of the hidden layer is... The output vector is ;

[0114] The net input vector in the output layer is The actual output vector is ;

[0115] The weight vector between the input layer and the hidden layer is The weight vector between the hidden layer and the output layer is The threshold of all units in the hidden layer is The threshold of the unit in the output layer is Where i=1,2,3, j=1,2,3,4, t=1; assign connection weights W, V and threshold values. , Let be any value between [-1, 1].

[0116] v jt This represents the connection weight from the j-th node in the hidden layer to the t-th node in the output layer. j=1,2,3,4 corresponds to 4 hidden neurons, and t=1 corresponds to a single output node.

[0117] The weighted sum of the inputs obtained by the hidden layer processing unit is as follows:

[0118] ;

[0119] Where j = 1, 2, 3, 4.

[0120] The activation function for the hidden layer is a sigmoid function, and the output of the hidden layer processing unit is expressed as follows:

[0121] ;

[0122] The activation function of the output layer is a sigmoid function.

[0123] S06, corresponding to the framework established in S05, selects the data from orthogonal experiments as training samples, and pairs of samples (X... k , Y k The adaptive learning algorithm is used to transmit the input training samples to the hidden layer units, and the net input vector and output vector of all processing units in the hidden layer are calculated according to the following formula:

[0124] ;

[0125] ;

[0126] Where j = 1, 2, 3, 4;

[0127] This is the output vector of the j-th unit in the hidden layer; This is the net input vector of the j-th unit in the hidden layer. The superscript k denotes the training sample number, and the input sample pair (X... k , Y k In ), X k Let Y be the gap parameters (length, width, roughness) for the k-th group, and Y be the normalized value of each input group in the orthogonal experiment. k Here, represents the flow rate value obtained from the simulation, and represents the normalized value of the flow rate value obtained from the simulation.

[0128] The net input vector and actual output vector of all processing units in the output layer are calculated using the following formula:

[0129] ;

[0130] ;

[0131] Where t=1;

[0132] This is the net input vector of the output layer. This is the actual output vector of the output layer.

[0133] Combined with the expected output vector Calculate the correction error values ​​for all processing units in the output layer:

[0134] ;

[0135] The correction error values ​​for all processing units in the hidden layer are calculated using the following formula:

[0136] ;

[0137] The weights V connecting the hidden layer and the output layer, as well as the thresholds of each neuron in the output layer, are changed according to the following formula. , represented as:

[0138] ;

[0139] ;

[0140] In the formula, This is the learning rate value, 0 < <1.

[0141] v jt The forward propagation weights are the connection weights from the j-th node in the hidden layer to the t-th node in the output layer. Update the weights; similarly, according to Output layer threshold Update.

[0142] The weights W connecting the input layer and the hidden layer, and the thresholds of the neurons in the hidden layer, are calculated using the following formula. , represented as:

[0143] ;

[0144] ;

[0145] In the formula, β is the learning rate value, 0 < β < 1.

[0146] The forward propagation is performed using the connection weights from the i-th node in the input layer to the j-th node in the hidden layer. Update the weights; similarly, according to Hidden layer threshold Update.

[0147] S07. Select another set of training sample data and apply it to the adaptive learning algorithm. Repeat step S06 until all training sample data has been trained.

[0148] Calculate the degree to which the overall error E of the adaptive learning algorithm matches the required accuracy, i.e., E is less than or equal to ε; if it meets the requirement, the process ends; otherwise, the training continues.

[0149] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for constructing a crack seepage prediction model, characterized in that, Includes the following steps: S01. Identify the key factors affecting water seepage through the cracks, including the length of the crack, the width of the crack, and the surface roughness of the wall. S02. Select three specific factors as test factors, and select multiple levels corresponding to each factor according to the common working conditions of the internal gaps of the aircraft, including length level, width level, and roughness level. List the corresponding orthogonal test table of gap seepage characteristics according to the selected factors and levels. S03. Simulation software was used to simulate and analyze the flow field of water seepage through the gaps inside the aircraft under rain conditions. S04. Set the corresponding simulation parameters according to the orthogonal experimental table in step S02, and obtain the simulation results under various simulation conditions, including the flow rate and velocity at the gap outlet. S05. Build an adaptive learning algorithm model. Use the simulation data obtained in step S04 as training samples to train the adaptive learning algorithm model and obtain a crack seepage prediction model. The adaptive learning algorithm model consists of an input layer L1, a hidden layer L2, and an output layer L3. The input layer has n nodes, and the hidden layer has p nodes. Nodes are connected by a weight matrix W, and the weight vector is... Where i represents the number of nodes in the input layer and j represents the number of nodes in the hidden layer; The outputs of the n processing units in the input layer are represented as X = (x1, …, x2) i , …,x n ) T ; Threshold of hidden layer processing unit This indicates that for any node j in the hidden layer processing unit, the obtained input weighted sum s j Represented as: ； Where n is the number of input layer nodes, w ij x is the weight from the i-th node in the input layer to the j-th node in the hidden layer. i It is the output value of the i-th node in the input layer; The calculated weighted sum s j The input is a transition function, and the activation function of the hidden layer is a sigmoid function. This determines the output y of the j-th node in the hidden layer. j for: ； The weight vector between the hidden layer and the output layer is represented as follows: ,in, v jt This represents the connection weight from the j-th node in the hidden layer to the t-th node in the output layer, where t represents the number of nodes in the output layer. The output layer has a total of q processing units, and the activation function of the output layer is a sigmoid function.

2. The method for constructing a crack seepage prediction model according to claim 1, characterized in that, In step S03, a velocity inlet for rainwater entry is set at the top of the gap, and a pressure outlet is set on the unrelated right side wall and the bottom gap outlet wall to establish a simulation model of the gap seepage flow field.

3. The method for constructing a crack seepage prediction model according to claim 1, characterized in that, In step S04, the influence of each factor on the flow field of seepage in the crack is analyzed based on orthogonal experiments, and the range is calculated. The influence of each factor on the flow field of seepage in the crack is ranked according to the range. The construction of the crack seepage prediction model is optimized based on the ranking results.

4. The method for constructing a crack seepage prediction model according to claim 1, characterized in that, The steps for training the adaptive learning algorithm model using the simulation data obtained in step S04 as training samples include: S06. Input one set of samples. Select three variables as input parameters: gap length, gap width, and wall roughness. The output parameter is the gap outlet flow rate c. t ; S07. Repeat step S06, using the error backpropagation algorithm, combined with the desired output vector. Calculate the correction error values ​​for all processing units in the output layer, expressed as: ； In the formula, The weighted sum of the inputs to the output layer. The derivative of the transfer function; Based on the correction error of the output layer, the correction error value of each unit in the hidden layer is calculated and expressed as: ； in, The derivative of the Sigmoid function; Using learning rate Update the weights connecting the hidden and output layers. and the bias of the output layer ; Use another learning rate Update the weights connecting the input layer and the hidden layer. and the bias of the hidden layer ; Determine the degree to which the overall error E of the adaptive learning algorithm matches the required accuracy, and complete the training of the adaptive learning algorithm model.

5. The method for constructing a crack seepage prediction model according to claim 4, characterized in that, In step S07, it is determined whether E satisfies the condition of being less than or equal to the precision setting value. Training ends when the condition is met; otherwise, training continues until the crack seepage prediction model is obtained.

6. A crack seepage prediction model, characterized in that, The model was constructed using the method described in any one of claims 1-5 for predicting crack seepage.

7. A method for predicting water seepage through cracks, characterized in that, The crack seepage prediction model described in claim 6 is used to predict crack seepage.

8. An electronic device, characterized in that, include: processor; as well as, Memory for storing the executable instructions of the processor; The processor is configured to execute the crevice seepage prediction method of claim 7 by executing the executable instructions.

9. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the crack seepage prediction method of claim 7.

Images