CN-120874462-B - Stage tensioning prestress long-term loss prediction method based on deep learning

CN120874462BCN 120874462 BCN120874462 BCN 120874462BCN-120874462-B

Abstract

The invention discloses a staged stretching prestress long-term loss prediction method based on deep learning, which belongs to the field of deep learning and solves the problem of poor robustness of a prediction result of a traditional prestress determination method, and comprises the steps of acquiring bridge original data based on a standard drawing set and bridge design specifications; the method comprises the steps of calculating concrete material properties based on construction environment conditions and concrete age, simulating the staged tensioning process of the prestressed reinforcement under different working conditions by utilizing finite element software based on bridge original data and the concrete material properties to generate a staged tensioning loss data set of the prestressed reinforcement of the bridge, constructing a prestress loss prediction model based on a multi-layer perceptron neural network, processing the staged tensioning loss data set of the prestressed reinforcement of the bridge based on the prestress loss prediction model, and outputting a long-term loss prediction result of the prestressed reinforcement caused by shrinkage creep. The method can effectively predict the stress loss of the prestressed reinforcement, and improves the robustness and accuracy of prediction.

Inventors

ZHANG YUNLONG
WANG HAOWEN
QIAN XUESONG
SUN YUN
QIAN YUCHENG

Assignees

吉林建筑大学

Dates

Publication Date: 20260508
Application Date: 20250901

Claims (9)

1. The method for predicting the long-term loss of the staged stretching prestress based on the deep learning is characterized by comprising the following steps of: acquiring bridge original data based on a standard drawing set and bridge design specifications; Calculating the properties of concrete materials based on construction environmental conditions and concrete age; Based on the bridge original data and the concrete material property, simulating a staged stretching and drawing process of the prestressed reinforcement under different working conditions by utilizing finite element software, and generating a staged stretching and drawing loss data set of the prestressed reinforcement of the bridge, wherein the process comprises the steps of establishing a finite element model of the bridge comprising a coupled concrete unit and a reinforcing steel unit by adopting a finite element modeling method based on the bridge original data, obtaining a concrete shrinkage strain value by adopting a shrinkage strain calculation method based on a concrete shrinkage coefficient in the concrete material property, converting the concrete shrinkage strain value into an equivalent temperature load, updating a material parameter by adopting a creep parameter calculation method based on the concrete creep coefficient in the concrete material property, solving the finite element model of the bridge based on the equivalent temperature load and the updated material parameter to obtain prestress loss data under each working condition, obtaining a standardized data set with unified dimension by adopting a data standardized processing method based on the prestress loss data, obtaining a training set, a verification set and a test set by adopting the data segmentation to obtain the standardized data set, and completing the staged stretching and drawing loss data set of the bridge; And constructing a prestress loss prediction model based on a multi-layer perceptron neural network, processing the bridge prestress steel bar staged tensioning loss data set based on the prestress loss prediction model, and outputting a long-term loss prediction result of the prestress steel bar caused by shrinkage creep.
2. The deep learning-based staged tension pre-stress long-term loss prediction method according to claim 1, wherein the concrete material properties include a concrete elastic modulus, a concrete shrinkage coefficient and a concrete creep coefficient; Wherein, the process of calculating the concrete material property comprises: calculating the elastic modulus of the concrete through a compressive strength development function based on the concrete age and the cement type; calculating a shrinkage coefficient of the concrete by adopting a shrinkage development function based on the concrete age and the theoretical thickness of the member; Based on the loading age and the average humidity of the environment, the creep coefficient of the concrete is calculated by adopting a viscoelastic theory and a statistical regression analysis.
3. The deep learning-based staged tension prestressing long-term loss prediction method according to claim 2, wherein the concrete shrinkage coefficient is calculated as: ; in the formula, Representation of To the point of The coefficient of shrinkage development of the concrete; the concrete age considering the calculation time is expressed in days; The concrete age at the beginning of shrinkage is expressed in days; The theoretical thickness of the component is expressed in mm; the reference theoretical thickness is expressed in mm.
4. The deep learning-based staged tension pre-stress long-term loss prediction method according to claim 3, wherein the concrete creep coefficient is calculated by the following expression: ; in the formula, Representation of To the point of Moment concrete creep coefficient; the concrete age considering the calculation time is expressed in days; Representing the loading age; Represents the average relative humidity of the environment over the years; representing the nominal creep coefficient of coagulation popular name obtained from the nominal creep coefficient table; representing an ambient reference humidity; The theoretical thickness of the component is expressed in mm; the reference theoretical thickness is expressed in mm.
5. The deep learning-based staged tension pre-stress long term loss prediction method of claim 1, wherein the expression for updating material parameters based on concrete creep coefficients in the concrete material properties is: ; in the formula, Representation of Modulus of elasticity of concrete at moment The concrete creep coefficient at the moment of calculation is represented; representing the concrete creep coefficient at a time immediately before the calculation time; Indicating the computing time age, wherein the unit is day; Considering the age of the moment before the calculation moment, wherein the unit is a day; Representing the updated material parameters.
6. The deep learning-based staged stretching pre-stress long-term loss prediction method is characterized in that a dual segmentation strategy is adopted to conduct data segmentation on the standardized data set to obtain a staged stretching loss data set of the bridge pre-stress steel bars; the bridge prestressed reinforcement staged tension loss data set comprises a training set, a verification set and a test set.
7. The deep learning-based staged stretching pre-stress long-term loss prediction method of claim 6, wherein the process of processing the staged stretching loss data set of the bridge pre-stress steel bar based on the pre-stress loss prediction model and outputting the long-term loss prediction result of the pre-stress steel bar caused by shrinkage creep comprises the following steps: based on the training set, model training is carried out by adopting a multi-layer perceptron neural network, a predicted value is calculated through forward propagation, and a mean square error loss function is adopted to evaluate the predicted deviation; based on the verification set and the prediction error, optimizing model super-parameters by adopting an early-stop method and a learning rate scheduling strategy; based on the optimization model hyper-parameters and the test set, evaluating the generalization performance of the prestress loss prediction model by adopting a mean square error, an average absolute error and a decision coefficient to obtain a trained prestress loss prediction model; and applying the trained prestress loss prediction model to new input data, and outputting a prestress loss predicted value.
8. A computer comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of claims 1-7 when executing the computer program.
9. A storage medium having stored thereon a computer program, which when executed by a processor performs the method of any of claims 1-7.

Description

Stage tensioning prestress long-term loss prediction method based on deep learning Technical Field The invention belongs to the technical field of deep learning, and particularly relates to a stage tensioning prestress long-term loss prediction method based on deep learning. Background The prestressed concrete beam can shrink and deform in the hardening process of the concrete, and the concrete can creep under the action of the continuous pre-pressing stress of the prestressed steel bars. The shrinkage and creep of concrete can shorten the length direction of a girder, reduce the elongation of prestressed reinforcement, cause prestress loss, reduce the effective prestress provided by the prestressed reinforcement, reduce the crack resistance and rigidity of components, possibly lead the bending bearing capacity to not reach the design requirement, and for a large-span bridge, the long-term creep effect can lead to the far down deflection exceeding the design value in the later operation process. In order to solve the problems, a construction mode of staged and graded stretching can be adopted, a part of stress is stretched in advance in the early stage of strength establishment, shrinkage creep is accelerated, and then the stress is stretched to be controlled according to the standard requirement. The method can offset the partial shrinkage creep effect and is widely applied. The test research of the staged tensioning technology of the prestressed concrete rectangular beam is carried out as the paper ' staged tensioning technology test of the prestressed concrete rectangular beam ' published in journal of China highway school journal ', the prestressed steel bars are tensioned twice in different ages, and the prestress loss is detected, and the result shows that the staged tensioning technology of the prestressed concrete in the early age can effectively reduce the prestress loss of the concrete, improve the cracking load and the yield load of the beam body, and does not influence the flexural ductility failure form of the beam body. The Chinese patent with publication number of CN108396661B and name of construction method based on prestress staged tensioning for eliminating concrete creep provides a method for partially counteracting the creep effect and effectively reducing creep deformation by tensioning prestress three times respectively after the hollow slab is lifted, during secondary paving and laying and the full bridge deck construction is completed. However, the staged stretching technique cannot completely eliminate all shrinkage and creep, and the prestress loss still needs to be measured. The Chinese patent with publication number CN119756648B and name of a bridge prestress loss measurement system based on acoustic emission technology proposes a method for calculating prestress loss by collecting acoustic emission signals in the process of bridge prestress loss and correcting errors by combining environmental data. The technology of coupling multipoint quasi-distributed FBG to the central wire of the prestressed steel strand in the concrete beam is proposed to realize the monitoring of the prestress loss and the distribution of the prestress loss, as in the paper published in journal of railway science and engineering journal. Besides, the traditional prestress measuring method also comprises a counter force method, a strain gauge method and the like. The method can reflect the prestress loss condition of the bridge to a certain extent. In the prior art, the construction period of the staged stretching technology is long, the traditional prestress determination method depends on special equipment, consumes a large amount of manpower resources, cannot be continuously detected, depends on the experience of field personnel, lacks accuracy, has hysteresis in detection data, and can only correct prestress loss by a tensioning method. The method for obtaining the bridge original data through the field experiment is high in difficulty, high in cost and small in data size, and the robustness of the prediction result is poor. Therefore, the invention provides a stage tensioning prestress long-term loss prediction method based on deep learning. Disclosure of Invention In order to solve the technical problems, the invention provides a stage tensioning prestress long-term loss prediction method based on deep learning, which aims to solve the problems in the prior art. In order to achieve the above purpose, the invention provides a stage tensioning prestress long-term loss prediction method based on deep learning, which comprises the following steps: acquiring bridge original data based on a standard drawing set and bridge design specifications; Calculating the properties of concrete materials based on construction environmental conditions and concrete age; Based on the bridge original data and the concrete material properties, simulating the staged tensioning process of the prestressed reinforcement under