Bridge prestressed reinforcement elastic compression loss prediction method and system and storage medium

Abstract

The invention relates to a bridge prestressed reinforcement elastic compression loss prediction method and system and a storage medium. The method comprises the following steps: simulating a bridge model through finite element software; the sum of all the prestressed reinforcement units is calculated, and temperature loads applied to all the reinforcement units are calculated; according to a prestressed reinforcement tensioning sequence and a life-death unit method, temperature loads are applied to reinforcement units one by one, and calculation is carried out; the method comprises the following steps: constructing an original data set, preprocessing, and dividing to obtain a training set, a verification set and a test set; constructing an MLP multilayer perceptron neural network model as a prediction model of the elastic compression loss of the bridge prestressed reinforcement to the concrete, and training the prediction model; and utilizing the trained prediction model to predict concrete elastic compression prestress loss caused by tensioning other prestressed reinforcements by the prestressed reinforcements. According to the invention, the elastic compression loss prediction of the bridge prestressed reinforcement can be quickly and accurately realized.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of bridge data processing, and in particular to a bridge prestressed steel bar elastic compression loss prediction method, system and storage medium. BACKGROUND

[0002] Prestressed concrete structures improve the crack resistance of bridges by reserving compressive stress in the bridge. Therefore, prestressed concrete bridges are basically used in actual bridge engineering. However, during the tensioning and anchoring process of the prestressed concrete bridge, the concrete will undergo elastic compression, and the prestressed steel bar will also be shortened, resulting in a decrease in the effective prestress of the prestressed steel bar and a decrease in the crack resistance of the bridge in actual operation. In addition, different tensioning sequences in post-tensioning construction will also result in different prestress losses of the steel bar. The prestressed steel bar that is tensioned first will be affected by the post-tensioned steel bar and will experience elastic compression prestress loss. In practical applications, the prediction of the elastic compression loss of the bridge prestressed steel bar is usually performed in the following three ways:

[0003] (1) Simplified calculation of the elastic compression loss caused by the tensioning of the bridge prestressed steel bar, but the accuracy of the simplified calculation is usually insufficient to meet the requirements;

[0004] (2) Finite element modeling according to a bridge with a prestressed steel bar elastic compression loss to be determined, but the finite element model needs to be constructed each time the prestressed steel bar elastic compression loss of the bridge is determined, which makes the prediction slow each time it is applied.

[0005] (3) Determining the elastic compression prestress loss of the concrete after the tensioning of the prestressed steel bar, which requires the steel bar to be tensioned, and the construction is complex.

[0006] Therefore, it is necessary to design a prestressed steel bar elastic compression loss prediction method and system that can accurately predict, has high prediction efficiency, and does not require tensioning of the prestressed steel bar. SUMMARY

[0007] Therefore, it is necessary to design a prestressed steel bar elastic compression loss prediction method and system that can accurately predict, has high prediction efficiency, and does not require tensioning of the prestressed steel bar.

[0008] To solve the above problems, the present disclosure adopts the following technical solutions:

[0009] In a first aspect, the present disclosure provides a bridge prestressed steel bar elastic compression loss prediction method, comprising:

[0010] obtaining bridge structure parameters of a plurality of bridges;

[0011] simulating a bridge model according to the bridge structure parameters through a finite element software; and calculating the elastic compression loss of each prestressed steel bar unit. and , calculate the temperature load applied to each prestressed steel unit , represent the prestress loss caused by the friction between the prestressed steel and the duct, represent the prestress loss caused by the deformation of the anchor, the retraction of the steel and the compression of the joint; according to the prestressed steel tensioning sequence and the live and dead unit method, the temperature load is applied to the corresponding prestressed steel unit in the bridge model one by one, and the concrete elastic compression prestress loss caused by the tensioning of other prestressed steels is calculated ;

[0012] construct an original data set including the bridge structure parameters, the prestressed steel tensioning sequence, the , preprocess the original data set, and divide the processed original data set into a training set, a validation set and a test set;

[0013] construct an MLP multi-layer perceptron neural network model as a prediction model of the bridge prestressed steel elastic compression loss to the concrete, and train the prediction model using the training set, the validation set and the test set;

[0014] use the trained prediction model to predict the concrete elastic compression prestress loss caused by the tensioning of other prestressed steels .

[0015] In a preferred embodiment, the temperature load is applied to the corresponding prestressed steel unit in the bridge model one by one according to the prestressed steel tensioning sequence and the live and dead unit method, and the concrete elastic compression prestress loss caused by the tensioning of other prestressed steels is calculated , including: activating the prestressed steel unit one by one according to the prestressed steel tensioning sequence, and applying the corresponding temperature load on each prestressed steel unit ; after applying the corresponding temperature load on a prestressed steel unit and before activating the next prestressed steel unit, the prestress of the prestressed steel is obtained; after applying the corresponding temperature load on the last prestressed steel unit, the prestress of all other prestressed steels is obtained; the concrete elastic compression prestress loss caused by the tensioning of other prestressed steels is calculated by prestress difference.

[0016] In a preferred embodiment, the temperature load is calculated by the following formula:

[0017] ;

[0018] wherein, This represents the temperature load applied to the prestressed steel reinforcement unit; This indicates the control stress for tensioning under the anchor; This represents the coefficient of linear expansion of prestressed steel bars.

[0019] In a preferred embodiment, the preprocessing includes at least one of missing value processing, outlier processing, derived feature calculation, and unified dimensionality.

[0020] In a preferred embodiment, the outlier processing is as follows: outliers are identified and removed using the interquartile range method.

[0021] In a preferred embodiment, the missing value processing is as follows: the median is used to fill numerical features, and the mode is used to fill categorical features.

[0022] In a preferred embodiment, the method includes the step of tuning the hyperparameters of the neural network model through hyperparameter random search and 3-fold cross-validation; the parameter configuration used in the training process of the training prediction model includes: an optimization algorithm using Dropout regularization and an Adam optimizer with adaptive learning rate.

[0023] In a preferred embodiment, the bridge structural parameters include beam height, beam length, bridge cross-sectional area, bridge cross-sectional moment of inertia, prestressed steel bar eccentricity, and steel bar cross-sectional area.

[0024] Thirdly, this disclosure provides a system for predicting the elastic compression loss of prestressed steel bars in bridges, including:

[0025] The first acquisition module is used to acquire the bridge structural parameters of several bridges;

[0026] The simulation calculation module is used to simulate the bridge model using finite element software based on the bridge structural parameters; it is used to calculate the prestressed steel reinforcement element. and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This represents the prestress loss caused by anchor deformation, rebar retraction, and joint compression; it is used to realize the sequential application of temperature loads according to the prestressing rebar tensioning sequence and the birth-death element method. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. ;

[0027] The dataset processing module is used to construct a dataset including the bridge structural parameters, the prestressed steel tensioning sequence, and the... The original dataset is used for preprocessing the original dataset, and the processed original dataset is used to divide it into training set, validation set and test set;

[0028] The model building and training module is used to build an MLP (Multilayer Perceptron) neural network model as a prediction model for the elastic compression loss of bridge prestressed steel bars on concrete, and to train the prediction model using the training set, validation set and test set.

[0029] The prediction module is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars, using a trained prediction model.

[0030] Thirdly, this disclosure provides a storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the method for predicting elastic compression loss of prestressed steel bars in bridges as described in the first aspect.

[0031] This disclosure discloses a method, system, and storage medium for predicting the elastic compression loss of prestressed steel bars in bridges. By constructing finite element models of various bridges, it designs a method to calculate the elastic compression prestress loss of concrete caused by the tensioning of other prestressed steel bars in each prestressed steel bar. The method constructs parameters including bridge structure parameters, prestressed steel tensioning sequence, and Using the original dataset, an MLP (Multilayer Perceptron) neural network model was constructed and trained as a prediction model for the elastic compression loss of prestressed steel reinforcement in bridge concrete. This model predicts the elastic compression prestress loss of prestressed reinforced concrete using bridge structural parameters and the prestressing steel tensioning sequence. The prediction method and system are described. Based on the obtained prediction model, the loss of prestress caused by the tensioning of other prestressed steel bars in the concrete elastic compression of bridge prestressed steel bars can be quickly and accurately predicted without the need for additional tensioning of the prestressed steel bars, i.e., no construction operation is required. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating a method for predicting the elastic compression loss of prestressed steel bars in bridges according to an embodiment of this disclosure.

[0033] Figure 2 This is a flowchart illustrating step 2 of the method for predicting elastic compression loss of prestressed steel bars in bridges according to an embodiment of this disclosure.

[0034] Figure 3 This is a loss curve diagram for training and validation according to an embodiment of the present disclosure;

[0035] Figure 4 This is a graph showing the learning rate variation according to an embodiment of the present disclosure;

[0036] Figure 5 This is a scatter regression plot of loos1 corresponding to the training set of an embodiment of this disclosure.

[0037] Figure 6 This is a scatter regression plot of loos1 corresponding to the test set of an embodiment of this disclosure;

[0038] Figure 7 This is a scatter regression plot corresponding to the training set loos2 in one embodiment of this disclosure;

[0039] Figure 8 This is a scatter regression plot of loos2 corresponding to the test set of an embodiment of this disclosure;

[0040] Figure 9 This is a scatter regression plot corresponding to the training set loos3 in one embodiment of this disclosure;

[0041] Figure 10 This is a scatter regression plot of loos3 corresponding to the test set of an embodiment of this disclosure;

[0042] Figure 11 This is a residual violin plot of loos1 corresponding to the training set of an embodiment of this disclosure;

[0043] Figure 12 This is a residual violin plot of loos1 corresponding to the test set of an embodiment of this disclosure;

[0044] Figure 13 This is a residual violin plot of loos2 corresponding to the training set of an embodiment of this disclosure;

[0045] Figure 14 This is a residual violin plot of loos2 corresponding to the test set of an embodiment of this disclosure;

[0046] Figure 15 This is a residual violin plot of loos3 corresponding to the training set of an embodiment of this disclosure;

[0047] Figure 16 This is a residual violin plot of loos3 corresponding to the test set of an embodiment of this disclosure;

[0048] Figure 17 This is a schematic diagram of the structure of a bridge prestressed steel bar elastic compression loss prediction system according to one embodiment of the present disclosure. Detailed Implementation

[0049] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the drawings, not the entire structure.

[0050] See Figure 1 This embodiment provides a method for predicting the elastic compression loss of prestressed steel bars in bridges. The prediction method includes:

[0051] Obtain the bridge structural parameters of several bridges;

[0052] Using finite element analysis software and based on the bridge structural parameters, a bridge model was simulated; the parameters of each prestressed steel reinforcement unit were calculated. and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This indicates the prestress loss caused by anchor deformation, rebar retraction, and joint compression; the prestressing rebar tensioning sequence, combined with the birth and death element method (based on the prestressing rebar tensioning sequence and the birth and death element method), sequentially applies temperature loads. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. ;

[0053] The construction includes the bridge structural parameters, the prestressed steel tensioning sequence, and the... The original dataset is preprocessed, and the preprocessed dataset is divided into training set, validation set and test set.

[0054] A multilayer perceptron neural network model was constructed as a prediction model for the elastic compression loss of prestressed steel bars in bridge concrete. The prediction model was trained using the training set, validation set, and test set.

[0055] The trained prediction model is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars.

[0056] In this article, unless otherwise specified, all references to "steel bars" refer to prestressed steel bars.

[0057] The following section elaborates on a method for predicting the elastic compression loss of prestressed steel bars in bridges. The method includes:

[0058] Step 1, Data Acquisition: Collect the necessary bridge structural parameters for finite element simulation based on the construction drawings.

[0059] Obtain the structural parameters of the bridge from the bridge construction drawings. The structural parameters need to be simulated by a finite element model. For example, the structural parameters include beam height, beam length, bridge cross-sectional area, bridge cross-sectional moment of inertia, prestressed steel reinforcement eccentricity, and steel reinforcement cross-sectional area.

[0060] In this embodiment, data on two mainstream bridge types, T-beams and box girders, were collected to ensure the universality of the final model.

[0061] Step 2: Simulate the bridge model using finite element software based on the bridge structural parameters; calculate the prestressed steel reinforcement unit. and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This represents the prestress loss caused by anchor deformation, rebar retraction, and joint compression; based on the prestressing rebar tensioning sequence and the bridge model, the prestressing rebar tensioning sequence, combined with the birth and death element method, sequentially incorporates temperature loads. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. .

[0062] Based on all the bridge structural parameters from step 1, a corresponding number of bridge models were constructed. Numerical calculations were performed on each bridge model under different tensioning sequences. By tensioning the prestressed steel bars one by one, the effective prestress of each prestressed steel bar was obtained, and the loss of elastic compressive prestress in the concrete caused by tensioning the prestressed steel bars was calculated. Detailed results were obtained based on finite element simulation. The process is as follows Figure 2 As shown.

[0063] This embodiment uses solid reinforcement finite element modeling, using SOLID65 software to simulate concrete elements and LINK180 to simulate prestressed steel bars, and performs mesh generation separately. The concrete elements and prestressed steel bar elements (referred to as steel bar elements or elements) are connected by coupling.

[0064] This embodiment only considers the prestress loss generated during the tensioning of the steel bars, that is, only the short-term loss. Therefore, through the finite element model, the following results are obtained: , and The , Both refer to the prestress loss corresponding to a specific prestressed steel bar unit.

[0065] Because the prestressed steel bars are located in different positions, the prestress loss experienced by each prestressed steel bar is also different. Therefore, after constructing the bridge model, it is necessary to obtain the prestress after eliminating these losses and convert it into temperature load: first, calculate the prestress of each prestressed steel bar unit. and This will further control the anchorage stress corresponding to a single prestressed steel reinforcement unit. Subtract its and This allows us to obtain the results of removing each prestressed steel reinforcement unit. and The stress afterward is then removed using the equivalent temperature change method. and The subsequent stress is applied to each prestressed steel element as a temperature load.

[0066] Prestress loss caused by friction between the prestressed steel bars and the duct in each prestressed steel bar unit The calculation formula is as follows:

[0067]

[0068] in, This represents the coefficient that indicates the influence of local deviation per meter of pipeline on the friction coefficient. This indicates the length of the pipe from the tensioning end to the calculated cross-section; This represents the coefficient of friction between the prestressed steel bars and the pipe. This represents the sum of the angles between the tensioning end and the tangents of the curved pipe section at the calculated cross-section; This indicates the tension control stress under the anchor.

[0069] Prestress loss caused by anchorage deformation, rebar retraction, and joint compression in each prestressed steel reinforcement unit. The calculation formula is as follows:

[0070]

[0071] in, This indicates the elastic modulus of prestressed steel bars; This represents the values ​​for anchorage deformation at the tensioning end, prestressed steel bar retraction, and joint compression. This indicates the distance between the tensioning end and the anchorage section.

[0072] The formula for the equivalent temperature change method is:

[0073]

[0074] in, This represents the temperature load applied to the prestressed steel reinforcement unit; This represents the coefficient of linear expansion of prestressed steel bars.

[0075] Given the material properties such as the elastic modulus of prestressed steel bars and concrete, This can be simulated and calculated in the finite element simulation used, so it is not necessary to subtract the control stress under the steel reinforcement anchorage during the process of converting the control stress into temperature. .

[0076] The prestressing steel tensioning sequence is simulated using the birth and death element method. Prestressing steel units are activated one by one according to the tensioning sequence, and a corresponding temperature load is applied to each activated unit. Excluding or including the last prestressing steel unit, for each prestressing steel unit, after applying the corresponding temperature load (before activating the next prestressing steel unit), the prestress F of that prestressing steel is obtained. i’,i’ After applying the corresponding temperature load to the last prestressed steel reinforcement unit, the F of the prestressed steel reinforcement is obtained. N,i’ Specifically, the prestress of all prestressed steel bars except the last one is obtained by subtracting the prestress, i.e., F. i’,i’ and the corresponding F N,i’ The absolute value of the difference is used to calculate the loss of elastic compressive prestress in the concrete caused by the tensioning of other prestressed steel bars. .

[0077] Finite element simulation of prestressed steel bar tensioning: Since later-tensioned steel bars can affect earlier-tensioned steel bars, the tensioning sequence is simulated using the birth and death element method. Before tensioning the prestressed steel bars, all steel bar elements are "killed." "Killing" doesn't mean deleting the steel bar elements, but rather multiplying their stiffness matrix by a very small attenuation factor, making their contribution to the overall stiffness matrix almost zero. When a steel bar element is activated, the stiffness matrix attenuation factor is restored to 1, and it re-participates in the assembly of the overall stiffness matrix.

[0078] According to the preset tensioning sequence of prestressed steel bars, the steel bars are tensioned one by one. Multiple tensioning sequences are preset (e.g., 200 tensioning sequences). After each tensioning, the prestress of the corresponding prestressed steel bar is obtained. Under each tensioning sequence, the tensioning, data acquisition, and data calculation are carried out one by one according to the following "one tensioning cycle".

[0079] Specifically, one tensioning cycle is as follows: when tensioning the first rebar, only the first rebar unit is activated, and the rebar is tensioned using the temperature load of that rebar unit to obtain the prestress F of the first prestressed rebar. 1,1 When tensioning the second reinforcing bar, the second reinforcing bar unit is activated again. At this time, both the first and second reinforcing bar units are activated, and the first reinforcing bar is in a tensioned state. The second reinforcing bar is tensioned using the temperature load of the second reinforcing bar unit to obtain the prestress F of the second prestressed reinforcing bar. 2,2When tensioning the third reinforcing bar, the third reinforcing bar unit is activated again. At this time, the first, second, and third reinforcing bar units are all activated, and the first and second reinforcing bars are in a tensioned state. The third reinforcing bar is tensioned using the temperature load of the third reinforcing bar unit to obtain the prestress F of the third prestressed reinforcing bar. 3,3 Similarly, assuming a total of N reinforcing bars, during the (N-1)th tensioning operation, when tensioning the (N-1)th reinforcing bar, the (N-1)th reinforcing bar unit is activated again. At this time, all reinforcing bars from the first to the (N-1)th are in a tensioned state. The (N-1)th reinforcing bar is tensioned using the temperature load of the (N-1)th reinforcing bar unit to obtain the prestress F of the (N-1)th prestressed reinforcing bar. N-1,N-1 During the Nth tensioning operation, when tensioning the Nth rebar, the Nth rebar unit is activated again. At this time, all rebar units are activated and rebars 1 to N-1 are under tension. The Nth rebar is tensioned using the temperature load of the Nth rebar unit. At this point, the prestress F of each prestressed rebar is obtained. N,i’ Calculate the prestressed steel bars. The prestress loss of the i'th prestressed steel bar is F. i’,i’ and F N,i’ The absolute value of the difference, where F i’,i’ F represents the prestress obtained when the i'-th prestressed steel bar is tensioned for the i'th time, i.e., when the i'-th prestressed steel bar is tensioned; N,i’ This refers to obtaining the prestress of the i'-th prestressed steel bar during the Nth tensioning of prestressed steel bars, i.e., when tensioning the Nth prestressed steel bar. i' is an integer in the range [1, N], where i' represents the prestressed steel bar unit / prestressed steel bar number, and also the tensioning number in a certain prestressed steel bar tensioning sequence; N represents the total number of prestressed steel bar units / prestressed steel bars. In other words, it obtains the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars in N-1 prestressed steel bars. .

[0080] Understandably, in all tensioning sequences, a particular rebar is not always the last one. After all the pre-set tensioning cycles, each rebar has several... .

[0081] Step 3: Construct the original dataset for the bridge. The original dataset includes bridge structural parameters and the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel bars in each prestressed steel bar tensioning sequence. The original dataset is preprocessed, and then divided into training, validation and test sets.

[0082] The preprocessing includes at least one of the following: missing value processing, outlier processing, derived feature calculation, and unified dimensionality.

[0083] In this embodiment, it is not required that missing value processing, outlier processing, and derived feature calculation be performed on the data set for each feature. Depending on the specific situation, at least one of these methods—missing value processing, outlier processing, derived feature calculation, and dimension unification—is selected for preprocessing. For example, for data sets targeting beam height and beam length, both missing value processing and dimension unification are performed; for data sets targeting steel reinforcement cross-sectional area, missing value processing, outlier processing, derived feature calculation, and dimension unification calculation are performed; for example, the elastic compressive prestress loss of concrete... Outlier handling is performed on the dataset. Specifically, for each prestressed steel bar, a dataset is created, and outlier handling is performed on the dataset; alternatively, for each tensioning sequence, a dataset with N-1 data sets is constructed. A set of values, and outlier handling for each such set.

[0084] Missing value handling: The median is used to impute numerical features, and the mode is used to impute categorical features. Numerical features are continuously distributed, while categorical features are discretely distributed. Therefore, different imputation strategies are used for different data types. Median imputation maintains the median of the data distribution and is unaffected by extreme values. Mode imputation ensures that the overall distribution of categorical features remains unchanged and does not alter the original category proportions. This approach ensures the consistency of the tensioning order, resists outliers in measurements, improves computational efficiency, and maintains the original feature distribution. The combination of these two methods satisfies both the robustness requirement for numerical features and the distribution preservation requirement for categorical features. It is understood that in this embodiment, the tensioning order of the prestressed steel reinforcement units is a categorical feature, while all others are numerical features.

[0085] Outlier Handling: This embodiment uses the interquartile range method to identify and process outliers in the original data. This is achieved by sorting the dataset in ascending order and then dividing the sorted dataset into quartiles. The dataset is divided into four equal parts, and the median, or second quartile, is determined first. Then find the median of the upper and lower halves, which is the first quartile. The third quartile That is, for example, sorting in ascending order. The value at 25% of the data set represents the lower limit of the data body; The values ​​at 75% of the data points represent the upper limit of the data set. The outlier boundaries are calculated using the following formula:

[0086]

[0087]

[0088]

[0089] in, Indicates the interquartile range; Indicates the upper boundary of the data outlier boundary; This indicates the lower boundary of the data outlier boundary.

[0090] Data outside the outlier boundary is considered an outlier. Outliers are removed, and data within the boundary is retained. This outlier handling method counteracts the interference of extreme values, avoids the model learning incorrect patterns, and enhances the robustness and accuracy of the model's predictions.

[0091] Derived Feature Calculation: While MLP models possess nonlinear learning capabilities, they often struggle to capture complex physical laws using only raw data features during the learning process. Complex neural networks may be required to capture the relationship between features and the target. By constructing derived features based on knowledge from the bridge engineering field, higher-order feature relationships are built, enriching the feature space and providing high-quality input features for the MLP neural network. This enables the model to efficiently capture the patterns between features and the target, improving prediction accuracy. Derived features include the ratio of prestressed steel reinforcement to concrete cross-sectional area and the ratio of prestressed steel reinforcement to concrete cross-sectional stiffness, calculated as follows:

[0092] The formula for the ratio of prestressed steel reinforcement to concrete cross-sectional area is as follows:

[0093]

[0094] in, Indicates the first The cross-sectional area of ​​the prestressed steel bars; This represents the cross-sectional area of ​​the concrete at mid-span. Indicates the first The ratio of the cross-sectional area of ​​the reinforcing steel bar to the area of ​​the concrete. This indicates the number of the prestressed steel reinforcement.

[0095] The ratio of prestressed steel reinforcement to concrete section stiffness is calculated using the following formula:

[0096]

[0097] in, This indicates the elastic modulus of concrete. Indicates the first The stiffness ratio of prestressed steel reinforcement to the mid-span concrete section; Indicates the first The eccentricity of the prestressed steel bars.

[0098] Standardizing Dimensions: Original features have significantly different dimensions. Features with larger dimensions dominate the model, while features with smaller dimensions are diluted during training. For MLP models, standardizing dimensions is well-suited to the gradient descent training mechanism. This embodiment uses the Robust Scaler standardization method to standardize the dimensions of features with different dimensions while maintaining the original distribution, stabilizing gradient updates, helping the optimizer efficiently find optimal parameters, accelerating model convergence, and improving model stability and robustness. The standardization formula is as follows:

[0099]

[0100] in This represents the standardized data; The raw data representing a certain feature; Represents the original data of a certain feature the median; Represents the original data of a certain feature The interquartile range, i.e. .

[0101] One-hot encoding is used to process the categorical features by stretching the order of the features, converting the discrete categorical values ​​into binary vectors. This prevents the model from interpreting the categorical features as numerical features and ensures that the model can learn them correctly.

[0102] Data partitioning: The feature data and target data of the data after the above operations are partitioned for the first time, into a 60% training set and a 40% test and validation set in a 6:4 ratio. The test and validation set is then partitioned for the second time, into a 1:1 ratio of 20% test set and 20% validation set. Both partitions are combined with a fixed random seed to ensure that the partitioning results are repeatable.

[0103] The training set, validation set, and test set are used for hierarchical evaluation. The training set is used for model training and parameter learning. The validation set is used to monitor overfitting, adjust hyperparameters, and dynamically optimize the learning rate. The test set is used to evaluate the model's generalization ability. Each set has its own function, ensuring data independence, avoiding data leakage that could lead to inflated model performance, and providing real-time feedback on loss through the validation set to prevent overfitting. This effectively improves the model's generalization ability and enables it to make accurate predictions on unfamiliar data.

[0104] Step 4: Construct an MLP (Multilayer Perceptron) neural network model as a prediction model for the elastic compression loss of bridge prestressed steel bars on concrete. Use the training set, validation set, and test set obtained in Step 3 to train the prediction model and obtain the trained prediction model.

[0105] A neural network model constructed using a multilayer perceptron (MLP) is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars in the prestressed steel bars. The model output is the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars in each prestressed steel bar in the bridge.

[0106] Different hyperparameter combinations were evaluated, the optimal hyperparameter combination was obtained, and an MLP multilayer perceptron neural network model was constructed. The model used Dropout regularization to improve generalization ability, and the Adam optimizer was used to adaptively adjust the learning rate and optimize the convergence process using learning rate scheduling. After training, the model that performed best on the validation set was obtained.

[0107] This embodiment optimizes the hyperparameters of the neural network model by combining randomized search with 3-fold cross-validation. The training set is divided into three mutually exclusive subsets: two training subsets for cross-validation and one validation subset for cross-validation. Training and validation are performed three times, meaning each subset is validated once. The negative mean squared error (NSF) of each validation is calculated, and the average performance of each set of parameters is evaluated using the NSF. The goal is to maximize the NSF and minimize the mean squared error, and the optimal hyperparameter combination is used for subsequent model training.

[0108] Next, the model is trained based on the preprocessed data. A ReLU activation function is introduced through forward propagation to calculate the loss, followed by backpropagation to update the parameters and construct the model. This embodiment uses mean squared error. As the loss function, its formula is as follows:

[0109]

[0110] in, Indicates the total number of samples; Indicates the first The true value of each sample; Indicates the first Predicted values ​​for each sample;

[0111] Dropout regularization is used to randomly discard neurons, preventing neuron co-adaptation, reducing the model's dependence on neurons, improving model robustness, and effectively preventing overfitting. Its forward propagation formula is as follows:

[0112] Training phase:

[0113] Prediction phase:

[0114] in, This represents the output value after Dropout; This represents the probability of dropping items in Dropout; This represents the neuron's raw output value; This represents the scaling factor, used to keep the expected data constant.

[0115] The loss gradient for each weight and bias is calculated using the chain rule of backpropagation, and the parameters are updated using the Adam optimizer with an adaptive learning rate, as shown in the following formula:

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] in, Indicates the number of iterations. Indicates the first The first moment of the next iteration; Indicates the first The first moment of the next iteration; Indicates the first The gradient of the next iteration; Indicates the first The second moment of the next iteration; Indicates the first The second moment of the next iteration; Indicates to Deviation correction; Indicates to Deviation correction; Indicates the learning rate; This represents the momentum decay rate, with a typical value of 0.9; This represents the gradient squared decay rate, with a typical value of 0.999; It is usually 1e-8 to prevent division by zero; The parameters representing the model update (used for the first) (parameters of the next iteration); Represents the current parameters of the model (the first parameter). (Parameters used in the next iteration).

[0122] Simultaneously, a learning rate scheduler is used. If the loss does not decrease for several consecutive rounds, the learning rate scheduler will automatically reduce the current learning rate to help the model escape local optima and avoid overfitting. This embodiment records the loss and learning rate of the training and validation sets for each iteration. Figure 3and Figure 4 As shown.

[0123] Step 5: Use the trained bridge prestressed steel reinforcement to predict the elastic compression loss of concrete.

[0124] The model obtains the bridge structural parameters and prestressed steel tensioning sequence of the bridge to be predicted. The input to the model are the bridge structural parameters and the prestressed steel tensioning sequence. In one embodiment, the bridge structural parameters input to the model include beam height, beam length, bridge cross-sectional area, bridge cross-sectional moment of inertia, prestressed steel eccentricity, and steel cross-sectional area. The output of the model is the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel bars in the bridge under the given prestressed steel tensioning sequence.

[0125] Following step 4, the process further includes evaluating the neural network model using evaluation metrics. The model is evaluated on a test set, using metrics including root mean square error (RMSE). Mean absolute error ( ), coefficient of determination ( The calculation formula for the evaluation indicators is as follows:

[0126] Root mean square error :

[0127]

[0128] Mean Absolute Error :

[0129]

[0130] Coefficient of determination :

[0131]

[0132] in, This represents the sample mean.

[0133] This represents the error between the model's predicted value and the actual value. This represents the average magnitude of the error between the model's predicted values ​​and the actual values. When the predicted values ​​perfectly match the actual values... and The value is equal to 0, and the smaller the error, the smaller the value. This represents the proportion of the data variance captured by the model to the total variance. It indicates when the model's predicted values ​​match the actual values. The value of is 1. The larger the value, the better the model fit.

[0134] The evaluation indicators for predicting the prestress loss of each steel bar in this embodiment are shown in Table 1:

[0135] Table 1 Prediction and Evaluation Indicators

[0136]

[0137] Where loos1, loos2, and loos3 represent the prestress loss generated by the first, second, and third reinforcing bars under a certain tensioning sequence, respectively.

[0138] Additionally, the scatter regression plots and residual violin plots for the training and test sets are as follows: Figures 5 to 16 As shown.

[0139] See Figure 17 This disclosure also provides a system for predicting the elastic compression loss of prestressed steel bars in bridges, characterized in that it includes:

[0140] The first acquisition module is used to acquire the bridge structural parameters of several bridges;

[0141] The simulation calculation module is used to simulate the bridge model using finite element software based on the bridge structural parameters; it is used to calculate the prestressed steel reinforcement element. and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This represents the prestress loss caused by anchor deformation, rebar retraction, and joint compression; it is used to realize the sequential application of temperature loads according to the prestressing rebar tensioning sequence and the birth-death element method. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. ;

[0142] The dataset processing module is used to construct a dataset including the bridge structural parameters, the prestressed steel tensioning sequence, and the... The original dataset is used for preprocessing the original dataset, and the processed original dataset is used to divide it into training set, validation set and test set;

[0143] The model building and training module is used to build an MLP (Multilayer Perceptron) neural network model as a prediction model for the elastic compression loss of bridge prestressed steel bars on concrete, and to train the prediction model using the training set, validation set and test set.

[0144] The prediction module is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars, using a trained prediction model.

[0145] In this embodiment, the model building and training module is also used to fine-tune the hyperparameters of the neural network model through hyperparameter random search and 3-fold cross-validation.

[0146] This disclosure also provides a storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the method for predicting the elastic compression loss of prestressed steel bars in bridges as described in any embodiment.

[0147] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0148] This disclosure discloses a method, system, and storage medium for predicting the elastic compression loss of prestressed steel bars in bridges. By constructing finite element models of various bridges, it designs a method to calculate the elastic compression prestress loss of concrete caused by the tensioning of other prestressed steel bars in each prestressed steel bar. The method constructs parameters including bridge structure parameters, prestressed steel tensioning sequence, and Using the original dataset, an MLP (Multilayer Perceptron) neural network model was constructed and trained as a prediction model for the elastic compressive stress loss of prestressed steel reinforcement in bridge concrete. This yielded a method and system for predicting the elastic compressive prestress loss of prestressed steel reinforcement concrete using bridge structural parameters and the prestressing steel tensioning sequence. Based on the obtained prediction model, the elastic compressive prestress loss of prestressed steel reinforcement in bridge concrete caused by the tensioning of other prestressed steel reinforcements can be predicted quickly and accurately without requiring additional tensioning of the prestressed steel reinforcements, i.e., no construction operation is involved. This disclosure provides reliable decision support for bridge design and construction by enabling rapid and accurate prediction of the elastic compressive prestress loss of prestressed steel reinforcement in concrete caused by the tensioning of other prestressed steel reinforcements in practical engineering.

[0149] This disclosure employs finite element simulation technology, which can quickly and accurately obtain the original dataset, rapidly provide a large amount of data for model training, and ensure the accuracy of model predictions.

[0150] This disclosure presents a model with strong generalization ability: It constructs an MLP (Multilayer Perceptron) neural network model, employing hyperparameter optimization through random search and cross-validation to avoid overfitting caused by manual parameter tuning, ensuring greater robustness in the face of unknown data. Simultaneously, it utilizes the Adam optimizer for adaptive learning rate and learning rate scheduling to optimize the convergence process, while employing Dropout regularization and early stopping mechanisms to prevent overfitting, thereby improving the model's generalization performance.

[0151] Enhancing data representativeness: This disclosure handles missing values ​​by imputation and outliers by IQR method, constructs derived features based on bridge engineering knowledge, improves the model's ability to capture key features, ensures the distribution trend of the original data by unifying the units of measurement, complements outlier handling, reduces the impact of extreme values ​​on model training, and enables more efficient learning of the relationship between features and target variables, thereby improving model training efficiency.

[0152] This disclosure enables rapid prediction of prestress loss in practical engineering projects, providing reliable decision support for bridge design and construction.

[0153] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods and systems according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a mechanism for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] Although preferred embodiments of the present disclosure have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this disclosure.

Claims

1. A method for predicting the elastic compression loss of prestressed steel bars in bridges, characterized in that, include: Obtain the bridge structural parameters of several bridges; The bridge model was simulated using finite element software based on the bridge's structural parameters. Calculate the prestressed steel reinforcement unit and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This represents the prestress loss caused by anchor deformation, rebar retraction, and joint compression; based on the prestressing rebar tensioning sequence and the birth-death element method, temperature loads are applied one by one. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. ; The construction includes the bridge structural parameters, the prestressed steel tensioning sequence, and the... The original dataset is preprocessed, and the preprocessed dataset is divided into training set, validation set and test set. A multilayer perceptron neural network model was constructed as a prediction model for the elastic compression loss of prestressed steel bars in bridge concrete. The prediction model was trained using the training set, validation set, and test set. The trained prediction model is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars.

2. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 1, characterized in that, According to the prestressed steel tensioning sequence and the birth-death unit method, the temperature load is applied one by one. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. This includes: activating prestressed steel reinforcement units one by one according to the prestressing steel tensioning sequence, and applying a corresponding temperature load to each activated prestressed steel reinforcement unit. After applying the corresponding temperature load to a prestressed steel bar unit and before activating the next prestressed steel bar unit, the prestress of that prestressed steel bar is obtained; after applying the corresponding temperature load to the last prestressed steel bar unit, the prestress of all other prestressed steel bars is obtained; the loss of concrete elastic compressive prestress caused by tensioning other prestressed steel bars is calculated by subtracting the prestress from the prestress.

3. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 1, characterized in that, The temperature load The calculation formula is: ； in, This represents the temperature load applied to the prestressed steel reinforcement unit; This indicates the control stress for tensioning under the anchor; This represents the coefficient of linear expansion of prestressed steel bars.

4. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 1, characterized in that, The preprocessing includes at least one of the following: missing value processing, outlier processing, derived feature calculation, and unified dimensionality.

5. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 4, characterized in that, The outlier handling method is as follows: outliers are identified and removed using the interquartile range method.

6. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 4, characterized in that, The missing value processing method is as follows: the median is used to fill in numerical features, and the mode is used to fill in categorical features.

7. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 1, characterized in that, The method includes the steps of tuning the hyperparameters of the neural network model through hyperparameter random search and 3-fold cross-validation; the parameter configuration used in the training process of the training prediction model includes: an optimization algorithm using Dropout regularization and an adaptive learning rate with Adam optimizer.

8. The method for predicting elastic compression loss of prestressed steel bars in bridges according to claim 1, characterized in that, The bridge structural parameters include beam height, beam length, bridge cross-sectional area, bridge cross-sectional moment of inertia, prestressed steel bar eccentricity, and steel bar cross-sectional area.

9. A system for predicting elastic compression loss of prestressed steel bars in bridges, characterized in that, include: The first acquisition module is used to acquire the bridge structural parameters of several bridges; The simulation calculation module is used to simulate the bridge model using finite element software based on the bridge structural parameters. Used to calculate each prestressed steel reinforcement unit and Calculate the temperature load applied to each prestressed steel element. , This indicates the loss of prestress caused by friction between the prestressed steel bars and the duct. This indicates the loss of prestress caused by anchor deformation, rebar retraction, and joint compression; This is used to implement the temperature load sequentially according to the prestressing steel tensioning sequence and the birth and death unit method. Apply the stress to the corresponding prestressed steel reinforcement element in the bridge model, and calculate the loss of elastic compressive prestress in the concrete caused by tensioning other prestressed steel reinforcements in each prestressed steel reinforcement. ; The dataset processing module is used to construct a dataset including the bridge structural parameters, the prestressed steel tensioning sequence, and the... The original dataset is used for preprocessing the original dataset, and the processed original dataset is used to divide it into training set, validation set and test set; The model building and training module is used to build an MLP (Multilayer Perceptron) neural network model as a prediction model for the elastic compression loss of bridge prestressed steel bars on concrete, and to train the prediction model using the training set, validation set and test set. The prediction module is used to predict the loss of elastic compressive prestress in concrete caused by the tensioning of other prestressed steel bars, using a trained prediction model.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for predicting elastic compression loss of prestressed steel bars in bridges as described in any one of claims 1-8.

Images