Method and system for determining reliability of concrete filled steel tube arch bridge based on LSTM (Long Short Term Memory)
Through the alternative model method based on LSTM, the complexity and nonlinearity of structural parameters of large-span steel pipe concrete arch bridges are solved, and fast and accurate reliability calculation is achieved, which reduces the calculation cost and complexity and improves the calculation efficiency.
Patent Information
- Application Number
- CN202411022135.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-26
AI Technical Summary
It is difficult for the prior art to quickly and accurately calculate the overall reliability of large-span steel pipe concrete arch bridges, especially when the structural parameter coupling relationship is complex and nonlinear, the calculation cost is high and the complexity is high.
Using an alternative model method based on LSTM, we create learning samples through the finite element model, train the LSTM network model, predict the ultimate load coefficient under the combination of structural parameters, and then determine the stability and reliability of the arch bridge.
It effectively reduces the number of calls and calculations of the finite element model, improves the calculation efficiency, and can quickly and accurately predict the ultimate load coefficient and stability reliability of the arch bridge, and is suitable for dealing with nonlinear and high-dimensional problems.
Smart Images

Figure CN119026206A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of bridge design, and in particular to a reliability determination method and system of a steel tube concrete arch bridge based on LSTM. Background Art
[0002] With the development of science and technology, the span of bridges is increasing. Steel tube concrete arch bridges have now exceeded the 500-meter span and are further developing towards the 700-meter span. Extra-large span steel tube concrete arch bridges have large spans and relatively slender rods, which makes the stability of large-span arch bridges prominent. On the one hand, arch bridges need to withstand temporary loads from different construction operations during the construction phase, and structural stability directly affects construction safety. On the other hand, arch bridges in the operation phase need to withstand various loads, including personnel activities, equipment operation, changes in the natural environment, etc. Structural stability also controls the safety of arch bridges during the operation phase.
[0003] In order to better understand the probability of instability and failure of bridges, the concept of reliability was introduced and gradually became an important tool for evaluating their stability. The traditional reliability calculation methods are mainly the first-order second-moment method and the Monte Carlo method. For simple structures that can write explicit functional functions, the reliability calculation mainly adopts the first-order second-moment method. For large-span steel tube concrete arch bridges whose functional functions are very complex or even cannot be written as a relationship, the reliability analysis is mainly carried out by combining the Monte Carlo method with the finite element model, which usually requires a lot of computing resources and time. In order to overcome these limitations and solve the problems of high computing cost and complexity, the substitution model method was introduced and widely used in structural reliability analysis. How to quickly construct an accurate substitution model to accurately obtain the failure probability of large-span steel tube concrete arch bridges is an area that needs to be explored in the current field of arch bridge reliability analysis. Summary of the invention
[0004] The purpose of the present invention is to provide a method and system for determining the reliability of a steel tube concrete arch bridge based on LSTM, so as to overcome the problem that the related technology cannot quickly and accurately calculate the overall reliability of a large-span steel tube concrete arch bridge when the steel tube concrete arch bridge structure system is complex, the coupling relationship of the structural parameters is complex, and there is a strong nonlinearity in the relationship between the structural parameters and the structural response.
[0005] In order to achieve the above object, the present invention provides the following technical solutions:
[0006] In a first aspect, an embodiment of the present invention provides a method for determining the reliability of a steel tube concrete arch bridge based on LSTM, the method comprising the following steps:
[0007] S100, establishing a finite element model and setting a parameter set, wherein the parameter set includes a plurality of structural parameters;
[0008] S200, determining the variation range of each structural parameter in the parameter set, performing structural parameter orthogonal test design within the variation range of each structural parameter to obtain multiple structural parameter combinations, determining the number of structural parameters in each structural parameter combination, and selecting the structural parameter combination with the least number as a combination sample;
[0009] S300, calculating the load-displacement curve corresponding to the combined sample through a finite element model, finding the load coefficient K value corresponding to the instability state of the arch bridge on the load-displacement curve, assembling a learning sample with the combined sample as input data and the load coefficient K value as output data, and forming a data set with the assembled learning samples;
[0010] S400, establishing an LSTM network model, using the data set to train the LSTM network model, and using the trained LSTM network model as an LSTM replacement model;
[0011] S500, sampling is performed according to the distribution characteristics of the structural parameters, and the ultimate load coefficients corresponding to different structural parameter combinations are predicted by the LSTM substitution model, and the stability reliability of the arch bridge is determined according to the ultimate load coefficients.
[0012] Optionally, in S400, the establishing of the LSTM network model, using the data set to train the LSTM network model, and using the trained LSTM network model as the LSTM replacement model includes:
[0013] S410, dividing the data set into a training set, a validation set and a test set;
[0014] S420, establish an LSTM network model, input the training set into the LSTM network model for optimization training, input the verification set into the LSTM network model for optimization verification, input the test set into the LSTM network model for testing, and obtain an LSTM replacement model.
[0015] Optionally, in S420, the establishing of the LSTM network model, inputting the training set into the LSTM network model for optimization training, inputting the verification set into the LSTM network model for optimization verification, and inputting the test set into the LSTM network model for testing to obtain the LSTM replacement model, includes:
[0016] S421, establishing an LSTM network model, wherein the LSTM network model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer includes an LSTM layer and a fully connected layer, the fully connected layer uses a LeakyReLU activation function, and the optimizer uses Adam;
[0017] S422, before each round of training and validation begins, randomly add noise to the input data for interference; during each round of training and validation, use a cross-validation method to form multiple hyperparameter combinations, and select the hyperparameter combination with the best performance; wherein the hyperparameters include one or more of a learning rate, a batch size, and a number of units;
[0018] S423, after each round of training is completed, determining a first evaluation index of the LSTM network model; after each round of verification is completed, determining a second evaluation index of the LSTM network model; wherein the first evaluation index and the second evaluation index both include mean square error MSE and mean absolute error MAE;
[0019] S424, inputting the combined samples in the test set into the LSTM network model, and determining a third evaluation index of the LSTM network model; wherein the third evaluation index includes a mean absolute error (MAE);
[0020] S425, determine whether the first evaluation index, the second evaluation index and the third evaluation index all meet the corresponding index thresholds. If so, obtain the LSTM replacement model; otherwise, execute step S422.
[0021] Optionally, in S423, after each round of training is completed, determining the first evaluation index of the LSTM network model includes:
[0022] After each round of training is completed, each combination sample in the training set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the training set and the load coefficient K value predicted by the LSTM network model for each combination sample in the training set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the first evaluation index.
[0023] Optionally, in S423, after each round of verification is completed, determining the second evaluation index of the LSTM network model includes:
[0024] After each round of verification is completed, each combination sample in the verification set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the verification set and the load coefficient K value predicted by the LSTM network model for each combination sample in the verification set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the second evaluation index.
[0025] Optionally, in S424, inputting the combined samples in the test set into the LSTM network model to determine the third evaluation index of the LSTM network model includes:
[0026] Each combination sample in the test set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value obtained by the LSTM network model and the load coefficient K value calculated by the finite element model for the corresponding combination sample in the test set are subjected to mean absolute error MAE calculation to obtain the third evaluation index.
[0027] Optionally, in S500, determining the stability reliability of the arch bridge according to the ultimate load coefficient includes:
[0028] Obtain the ultimate load coefficient Ku. If it is determined that the ultimate load coefficient Ku ≥ 1, it means that the arch bridge does not become unstable under this parameter combination.
[0029] If the ultimate load coefficient Ku is determined to be less than 1, it means that the arch bridge is unstable, and the failure probability of the arch bridge is determined by the following formula:
[0030] Pf = n / m;
[0031] Among them, m is the total number of random sampling, n is the total number of failures calculated by the LSTM substitution model, and Pf is the failure probability of the stability reliability of the steel tube concrete arch bridge.
[0032] In a second aspect, an embodiment of the present invention provides a reliability determination system for a steel tube concrete arch bridge based on LSTM, the system comprising:
[0033] at least one processor;
[0034] at least one memory for storing at least one program;
[0035] When the at least one program is executed by the at least one processor, the at least one processor implements the LSTM-based steel tube concrete arch bridge reliability determination method as described in any one of the above.
[0036] The beneficial effects of the present invention are:
[0037] The present invention uses the LSTM network model to replace the finite element model of the steel tube concrete arch bridge, quickly predicts the ultimate load coefficient of the arch bridge under different structural parameter combinations, and determines the stability reliability of the arch bridge according to the ultimate load coefficient. Give full play to the accuracy and speed of the LSTM replacement model prediction, effectively reduce the number of calls and calculations of the original finite element model, and improve the calculation efficiency.
[0038] The LSTM replacement model generated by the LSTM network model in the present invention has adaptability and generalization ability, and can well handle complex problems such as nonlinear, non-stationary, and non-Gaussian. Moreover, for high-dimensional problems, the LSTM replacement model only needs to achieve high approximation near the global optimal solution, and does not need to have high approximation in the entire design space, further reducing the computational burden of the original finite element model. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0040] Figure 1 It is a flow chart of a method for determining reliability of a steel tube concrete arch bridge based on LSTM in an embodiment of the present invention;
[0041] Figure 2 1 is a network architecture diagram of the LSTM substitution model in an embodiment of the present invention;
[0042] Figure 3 It is a structural schematic diagram of a reliability determination system for a steel tube concrete arch bridge based on LSTM in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention, so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.
[0044] In the related art, the methods of constructing alternative models can be roughly divided into response surface method, Kriging method and artificial neural network method. Among them, the concept of response surface method is clear. This method is based on mathematical models and experimental data, and predicts and optimizes the behavior of response variables by establishing a surface model. However, the response surface method may have difficulties in dealing with high-dimensional problems. This is because there may be complex interactions between the independent variables of high-dimensional problems. Therefore, it is mostly used in low-dimensional models in the early stage. With the increase of the span of the arch bridge, the superiority of the Kriging method and the traditional artificial neural network method is gradually reflected. However, in the convergence stage of the Kriging method, it can be observed that many sample points overlap and accumulate in space. The excessive concentration of training sample points means greater information redundancy, which wastes the calculation cost of the original function function. The traditional artificial neural network method has low fitting accuracy when the parameter nonlinearity is high, and there will be overfitting problems when the number of samples is small. These problems affect the efficiency and accuracy of these two reliability calculation methods. In view of these problems, the present invention studies the reliability of a super-large span steel tube concrete arch bridge based on LSTM, and uses LSTM to establish an alternative model of the finite element model to perform reliability analysis on the arch bridge.
[0045] See also Figure 1 , Figure 1 The present invention provides a method for determining the reliability of a steel tube concrete arch bridge based on LSTM, and the method comprises the following steps:
[0046] S100, establishing a finite element model and setting a parameter set, wherein the parameter set includes a plurality of structural parameters;
[0047] Specifically, a finite element model considering material nonlinearity and geometric nonlinearity is established, and structural parameters with complex coupling relationships such as the slenderness ratio of the main arch rib steel pipe, the steel content ratio of the main arch rib section, the diameter of the transverse connecting steel pipe, the wall thickness of the transverse connecting steel pipe, the elastic modulus of steel, and the elastic modulus of concrete are included in the combination of perturbed structural parameters;
[0048] S200, determining the variation range of each structural parameter in the parameter set, performing structural parameter orthogonal test design within the variation range of each structural parameter to obtain multiple structural parameter combinations, determining the number of structural parameters in each structural parameter combination, and selecting the structural parameter combination with the least number as a combination sample;
[0049] Specifically, the variation range of structural parameters is determined, and orthogonal experimental design of structural parameters is carried out within the range of structural parameters. The combination of structural parameters with the least number is used as the combination sample. The mechanical behavior characteristics of the finite element model are described with as few combination samples as possible to reduce the workload of finite element calculations.
[0050] S300, calculating the load-displacement curve corresponding to the combined sample through a finite element model, finding the load coefficient K value corresponding to the instability state of the arch bridge on the load-displacement curve, assembling a learning sample with the combined sample as input data and the load coefficient K value as output data, and forming a data set with the assembled learning samples;
[0051] S400, establishing an LSTM network model, using the data set to train the LSTM network model, and using the trained LSTM network model as an LSTM replacement model;
[0052] Specifically, there is a complex internal coupling relationship between the influence of structural parameters on the reliability of arch ribs, and the number of combined samples generated by orthogonal experimental design is small. The LSTM substitution model generated by the LSTM network model can well solve the problem of complex mapping relationship and small number of samples.
[0053] S500, sampling is performed according to the distribution characteristics of the structural parameters, and the ultimate load coefficients corresponding to different structural parameter combinations are predicted by the LSTM substitution model, and the stability reliability of the arch bridge is determined according to the ultimate load coefficients.
[0054] Specifically, after the LSTM substitution model is obtained, sampling is performed according to the distribution characteristics of the structural parameters, and the ultimate load coefficients corresponding to different structural parameter combinations are predicted by the LSTM substitution model. The stability reliability of the arch bridge is determined according to the ultimate load coefficients.
[0055] In some embodiments, in S400, the establishing of the LSTM network model, training the LSTM network model using the data set, and using the trained LSTM network model as the LSTM replacement model includes:
[0056] S410, dividing the data set into a training set, a validation set and a test set;
[0057] In some embodiments, the ratio of the training set, validation set, and test set is 6:2:2;
[0058] S420, establish an LSTM network model, input the training set into the LSTM network model for optimization training, input the verification set into the LSTM network model for optimization verification, input the test set into the LSTM network model for testing, and obtain an LSTM replacement model.
[0059] Specifically, after the LSTM network model is established, the learning samples in the training set are used to train the LSTM network model, and the learning samples in the validation set are used to validate the LSTM network model. The training and validation of the LSTM network model are optimized through strategies such as noise interference and hyperparameter tuning.
[0060] refer to Figure 2 In some embodiments, in S420, the LSTM network model is established, the training set is input into the LSTM network model for optimization training, the verification set is input into the LSTM network model for optimization verification, and the test set is input into the LSTM network model for testing to obtain the LSTM replacement model, including:
[0061] S421, establishing an LSTM network model, wherein the LSTM network model includes an input layer, a hidden layer and an output layer, wherein the hidden layer includes an LSTM layer, a flattening layer and a fully connected layer, wherein the fully connected layer uses a LeakyReLU activation function, and an optimizer uses Adam;
[0062] S422, before each round of training and validation begins, randomly add noise to the input data for interference; during each round of training and validation, use a cross-validation method to form multiple hyperparameter combinations, and select the hyperparameter combination with the best performance; wherein the hyperparameters include one or more of a learning rate, a batch size, and a number of units;
[0063] In some specific embodiments, the specific method of optimizing the LSTM network model by using strategies such as noise interference and hyperparameter tuning is as follows:
[0064] (1) Noise interference: At the beginning of each round of training, noise with a signal-to-noise ratio of 30db to 40db is randomly added to the input data for interference. This can help the LSTM network model learn more robust feature representations, while reducing overfitting and improving the network's generalization ability for unknown data;
[0065] (2) Hyperparameter tuning: Hyperparameters include learning rate, batch size, number of units, etc. Use cross-validation to try different hyperparameter combinations and select the best performing combination;
[0066] The number of input layers is 1, and the number of units is consistent with the number of structural parameters. The number of output layers is 1, and the number of units is 1. The hidden layers include LSTM layers and fully connected layers. The number of LSTM layers is optimized between 2 and 4, and the number of fully connected layers is optimized between 1 and 2. A flattening layer is added between the LSTM layer and the fully connected layer. The flattening layer is used to convert the three-dimensional tensor output by the LSTM layer into a two-dimensional tensor, so that the fully connected layer can receive a flattened sequence as input. The number of units in the LSTM layer and the fully connected layer are both optimized between 4 and 32. The fully connected layer uses the LeakyReLU activation function with a slope α of 0.1. The optimizer uses Adam, and the learning rate is optimized in the range of 0.001 to 0.01.
[0067] S423, after each round of training is completed, determining a first evaluation index of the LSTM network model; after each round of verification is completed, determining a second evaluation index of the LSTM network model; wherein the first evaluation index and the second evaluation index both include mean square error MSE and mean absolute error MAE;
[0068] Specifically, the TensorFlow platform was used to establish an LSTM network model, the learning samples in the training set were used to train the LSTM network model, and the learning samples in the validation set were used to train and verify the LSTM network model. After optimizing the training and verification of the LSTM network model through strategies such as noise interference and hyperparameter tuning, multiple evaluation indicators were used to evaluate the stability and prediction accuracy of the LSTM network model during the training and verification stages.
[0069] S424, inputting the combined samples in the test set into the LSTM network model, and determining a third evaluation index of the LSTM network model; wherein the third evaluation index includes a mean absolute error (MAE);
[0070] Specifically, the mean absolute error (MAE) is used as the third evaluation indicator to verify the prediction accuracy of the trained LSTM network model.
[0071] S425, determine whether the first evaluation index, the second evaluation index and the third evaluation index all meet the corresponding index thresholds. If so, obtain the LSTM replacement model; otherwise, execute step S422.
[0072] Specifically, when the LSTM network model does not meet the accuracy requirements, steps S422 to S424 are repeated until the first evaluation indicator, the second evaluation indicator and the third evaluation indicator all meet the corresponding indicator thresholds. After the LSTM network model meets the corresponding indicator threshold requirements during training, verification and testing, the LSTM replacement model is determined.
[0073] In some embodiments, in S423, after each round of training is completed, determining the first evaluation index of the LSTM network model includes:
[0074] After each round of training is completed, each combination sample in the training set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the training set and the load coefficient K value predicted by the LSTM network model for each combination sample in the training set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the first evaluation index.
[0075] In some embodiments, in S423, after each round of verification is completed, determining the second evaluation index of the LSTM network model includes:
[0076] After each round of verification is completed, each combination sample in the verification set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the verification set and the load coefficient K value predicted by the LSTM network model for each combination sample in the verification set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the second evaluation index.
[0077] In some embodiments, in S424, inputting the combined sample in the test set into the LSTM network model to determine the third evaluation index of the LSTM network model includes:
[0078] Each combination sample in the test set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value obtained by the LSTM network model and the load coefficient K value calculated by the finite element model for the corresponding combination sample in the test set are subjected to mean absolute error MAE calculation to obtain the third evaluation index.
[0079] It should be noted that the first evaluation index of the LSTM network model in the training phase and the second evaluation index in the verification phase both include the loss function Loss and the evaluation index Metrics. The loss function Loss uses the mean square error MSE, which is calculated by comparing the square of the difference between the predicted output and the actual output of the model:
[0080]
[0081] Where n is the number of combined samples, y i is the load factor K value calculated by the finite element model for the i-th combination sample, It is the load coefficient K value predicted by the LSTM network model for the i-th combination sample.
[0082] The evaluation metric is the mean absolute error (MAE). The smaller the MAE value, the smaller the difference between the model's prediction result and the true value, indicating that the model's prediction accuracy is higher. Compared with other evaluation indicators, MAE is easy to interpret and has strong robustness to outliers. Its calculation formula is as follows:
[0083]
[0084] The third evaluation indicator of the LSTM network model in the test phase uses the mean absolute error MAE.
[0085] In some embodiments, determining the stability reliability of the arch bridge according to the ultimate load coefficient includes:
[0086] Obtain the ultimate load coefficient Ku. If it is determined that the ultimate load coefficient Ku ≥ 1, it means that the arch bridge does not become unstable under this parameter combination.
[0087] If the ultimate load coefficient Ku is determined to be less than 1, it means that the arch bridge is unstable, and the failure probability of the arch bridge is determined by the following formula:
[0088] Pf = n / m;
[0089] Among them, m is the total number of random sampling, n is the total number of failures calculated by the LSTM substitution model, and Pf is the failure probability of the stability reliability of the steel tube concrete arch bridge.
[0090] Specifically, when calculating the failure probability, the total number of random samplings is defined as m, and the total number of failures calculated by the LSTM substitution model is n, and the failure probability of the stability reliability of the steel tube concrete arch bridge is obtained.
[0091] and Figure 1 Corresponding to the method, refer to Figure 3 The embodiment of the present invention provides a reliability determination system of a steel tube concrete arch bridge based on LSTM, including:
[0092] at least one processor;
[0093] at least one memory for storing at least one program;
[0094] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.
[0095] It can be seen that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0096] In addition, an embodiment of the present invention further discloses a computer program product or a computer program, which is stored in a computer-readable storage medium. A processor of a computer device can read the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the above method. Similarly, the contents of the above method embodiment are all applicable to the storage medium embodiment, and the functions specifically implemented by the storage medium embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method embodiment.
[0097] It will be appreciated by those skilled in the art that all or some of the methods disclosed above, the system can be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, and the computer-readable medium can include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0098] The above is a specific description of the preferred implementation of the present disclosure, but the present disclosure is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions without violating the spirit of the present disclosure. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present disclosure.
Claims
1. Reliability determination method of steel tube concrete arch bridge based on LSTM, It is characterized in that The method comprises the following steps: S100, establishing a finite element model and setting a parameter set, wherein the parameter set includes a plurality of structural parameters; S200, determining the variation range of each structural parameter in the parameter set, performing structural parameter orthogonal test design within the variation range of each structural parameter to obtain multiple structural parameter combinations, determining the number of structural parameters in each structural parameter combination, and selecting the structural parameter combination with the least number as a combination sample; S300, calculating the load-displacement curve corresponding to the combined sample through a finite element model, finding the load coefficient K value corresponding to the instability state of the arch bridge on the load-displacement curve, assembling a learning sample with the combined sample as input data and the load coefficient K value as output data, and forming a data set with the assembled learning samples; S400, establishing an LSTM network model, using the data set to train the LSTM network model, and using the trained LSTM network model as an LSTM replacement model; S500, sampling is performed according to the distribution characteristics of the structural parameters, and the ultimate load coefficients corresponding to different structural parameter combinations are predicted by the LSTM substitution model, and the stability reliability of the arch bridge is determined according to the ultimate load coefficients.
2. According to the LSTM-based steel tube concrete arch bridge reliability determination method of claim 1, It is characterized in that In S400, the LSTM network model is established, the LSTM network model is trained using the data set, and the trained LSTM network model is used as the LSTM replacement model, including: S410, dividing the data set into a training set, a validation set and a test set; S420, establish an LSTM network model, input the training set into the LSTM network model for optimization training, input the verification set into the LSTM network model for optimization verification, input the test set into the LSTM network model for testing, and obtain an LSTM replacement model.
3. According to the LSTM-based steel tube concrete arch bridge reliability determination method of claim 2, It is characterized in that In S420, the LSTM network model is established, a training set is input into the LSTM network model for optimization training, a verification set is input into the LSTM network model for optimization verification, and a test set is input into the LSTM network model for testing to obtain an LSTM replacement model, including: S421, establishing an LSTM network model, wherein the LSTM network model includes an input layer, a hidden layer, and an output layer, wherein the hidden layer includes an LSTM layer and a fully connected layer, the fully connected layer uses a LeakyReLU activation function, and the optimizer uses Adam; S422, before each round of training and validation begins, randomly add noise to the input data for interference; during each round of training and validation, use a cross-validation method to form multiple hyperparameter combinations, and select the hyperparameter combination with the best performance; wherein the hyperparameters include one or more of a learning rate, a batch size, and a number of units; S423, after each round of training is completed, determining a first evaluation index of the LSTM network model; after each round of verification is completed, determining a second evaluation index of the LSTM network model; wherein the first evaluation index and the second evaluation index both include mean square error MSE and mean absolute error MAE; S424, inputting the combined samples in the test set into the LSTM network model, and determining a third evaluation index of the LSTM network model; wherein the third evaluation index includes a mean absolute error (MAE); S425, determine whether the first evaluation index, the second evaluation index and the third evaluation index all meet the corresponding index thresholds. If so, obtain the LSTM replacement model; otherwise, execute step S422.
4. According to the LSTM-based steel tube concrete arch bridge reliability determination method of claim 3, It is characterized in that In S423, after each round of training is completed, determining the first evaluation index of the LSTM network model includes: After each round of training is completed, each combination sample in the training set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the training set and the load coefficient K value predicted by the LSTM network model for each combination sample in the training set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the first evaluation index.
5. According to the LSTM-based steel tube concrete arch bridge reliability determination method of claim 3, It is characterized in that In S423, after each round of verification is completed, determining the second evaluation index of the LSTM network model includes: After each round of verification is completed, each combination sample in the verification set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value corresponding to the combination sample in the verification set and the load coefficient K value predicted by the LSTM network model for each combination sample in the verification set are obtained, and then the mean square error MSE and the mean absolute error MAE are calculated to obtain the second evaluation index.
6. The method for determining the reliability of a steel tube concrete arch bridge based on LSTM according to claim 3, It is characterized in that In S424, the step of inputting the combined samples in the test set into the LSTM network model and determining the third evaluation index of the LSTM network model includes: Each combination sample in the test set is input into the LSTM network model to obtain the load coefficient K value predicted by the LSTM network model for the combination sample; the load coefficient K value obtained by the LSTM network model and the load coefficient K value calculated by the finite element model for the corresponding combination sample in the test set are subjected to mean absolute error MAE calculation to obtain the third evaluation index.
7. The method for determining the reliability of a steel tube concrete arch bridge based on LSTM according to claim 1, It is characterized in that In S500, determining the stability reliability of the arch bridge according to the ultimate load coefficient includes: Obtain the ultimate load coefficient Ku. If it is determined that the ultimate load coefficient Ku ≥ 1, it means that the arch bridge does not become unstable under this parameter combination. If the ultimate load coefficient Ku is determined to be less than 1, it means that the arch bridge is unstable, and the failure probability of the arch bridge is determined by the following formula: Pf = n / m; Among them, m is the total number of random sampling, n is the total number of failures calculated by the LSTM substitution model, and Pf is the failure probability of the stability reliability of the steel tube concrete arch bridge.
8. Reliability determination system of steel tube concrete arch bridge based on LSTM, It is characterized in that The system comprises: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the LSTM-based steel tube concrete arch bridge reliability determination method as described in any one of claims 1 to 7.
Citation Information
Cited By
Cloud computing roadbed construction slope displacement monitoring system
CN120337684A
